Omgxs

Principles of Measurement Systems
We work with leading authors to develop the strongest educational materials in engineering, bringing cutting-edge thinking and best learning practice to a global market. Under a range of well-known imprints, including Prentice Hall, we craft high quality print and electronic publications which help readers to understand and apply their content, whether studying or at work. To find out more about the complete range of our publishing, please visit us on the World Wide Web at: www.pearsoned.co.uk
Principles of Measurement Systems Fourth Edition
John P. Bentley Emeritus Professor of Measurement Systems University of Teesside
Pearson Education Limited Edinburgh Gate Harlow Essex CM20 2JE England and Associated Companies throughout the world Visit us on the World Wide Web at: www.pearsoned.co.uk First published 1983 Second Edition 1988 Third Edition 1995 Fourth Edition published 2005 © Pearson Education Limited 1983, 2005 The right of John P. Bentley to be identified as author of this work has been asserted by him in accordance w th the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without either the prior written permission of the publisher or a licence permitting restricted copying in the United Kingdom issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP. ISBN 0 130 43028 5 British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data Bentley, John P., 1943– Principles of measurement systems / John P. Bentley. – 4th ed. p. cm. Includes bibliographical references and index. ISBN 0-13-043028-5 1. Physical instruments. 2. Physical measurements. 3. Engineering instruments. 4. Automatic control. I. Title. QC53.B44 2005 530.8–dc22 2004044467 10 10
9 8 7 6 09 08 07
5 4 3 06 05
2
1
Typeset in 10/12pt Times by 35 Printed in Malaysia The publisher’s policy is to use paper manufactured from sustainable forests.
To Pauline, Sarah and Victoria
Contents Preface to the fourth edition Acknowledgements
Part A
xi xiii
General Principles
1
1
The 1.1 1.2 1.3 1.4
General Measurement System Purpose and performance of measurement systems Structure of measurement systems Examples of measurement systems Block diagram symbols
3 3 4 5 7
2
Static Characteristics of Measurement System Elements 2.1 Systematic characteristics 2.2 Generalised model of a system element 2.3 Statistical characteristics 2.4 Identification of static characteristics – calibration
9 9 15 17 21
3
The Accuracy of Measurement Systems in the Steady State 3.1 Measurement error of a system of ideal elements 3.2 The error probability density function of a system of non-ideal elements 3.3 Error reduction techniques
35 35
4
Dynamic Characteristics of Measurement Systems 4.1 Transfer function G(s) for typical system elements 4.2 Identification of the dynamics of an element 4.3 Dynamic errors in measurement systems 4.4 Techniques for dynamic compensation
51 51 58 65 70
5
Loading Effects and Two-port Networks 5.1 Electrical loading 5.2 Two-port networks
77 77 84
6
Signals and Noise in Measurement Systems 6.1 Introduction 6.2 Statistical representation of random signals 6.3 Effects of noise and interference on measurement circuits 6.4 Noise sources and coupling mechanisms 6.5 Methods of reducing effects of noise and interference
36 41
97 97 98 107 110 113
viii CONTENTS 7
Part B
Reliability, Choice and Economics of Measurement Systems 7.1 Reliability of measurement systems 7.2 Choice of measurement systems 7.3 Total lifetime operating cost
Typical Measurement System Elements
125 125 140 141
147
8
Sensing Elements 8.1 Resistive sensing elements 8.2 Capacitive sensing elements 8.3 Inductive sensing elements 8.4 Electromagnetic sensing elements 8.5 Thermoelectric sensing elements 8.6 Elastic sensing elements 8.7 Piezoelectric sensing elements 8.8 Piezoresistive sensing elements 8.9 Electrochemical sensing elements 8.10 Hall effect sensors
149 149 160 165 170 172 177 182 188 190 196
9
Signal Conditioning Elements 9.1 Deflection bridges 9.2 Amplifiers 9.3 A.C. carrier systems 9.4 Current transmitters 9.5 Oscillators and resonators
205 205 214 224 228 235
10
Signal Processing Elements and Software 10.1 Analogue-to-digital (A/D) conversion 10.2 Computer and microcontroller systems 10.3 Microcontroller and computer software 10.4 Signal processing calculations
247 247 260 264 270
11
Data Presentation Elements 11.1 Review and choice of data presentation elements 11.2 Pointer–scale indicators 11.3 Digital display principles 11.4 Light-emitting diode (LED) displays 11.5 Cathode ray tube (CRT) displays 11.6 Liquid crystal displays (LCDs) 11.7 Electroluminescence (EL) displays 11.8 Chart recorders 11.9 Paperless recorders 11.10 Laser printers
285 285 287 289 292 295 299 302 304 306 307
CONTENTS
Part C
Specialised Measurement Systems Measurement Systems Essential principles of fluid mechanics Measurement of velocity at a point in a fluid Measurement of volume flow rate Measurement of mass flow rate Measurement of flow rate in difficult situations
ix
311
12
Flow 12.1 12.2 12.3 12.4 12.5
13
Intrinsically Safe Measurement Systems 13.1 Pneumatic measurement systems 13.2 Intrinsically safe electronic systems
351 353 362
14
Heat 14.1 14.2 14.3
Transfer Effects in Measurement Systems Introduction Dynamic characteristics of thermal sensors Constant-temperature anemometer system for fluid velocity measurements 14.4 Katharometer systems for gas thermal conductivity and composition measurement
367 367 369
15
Optical Measurement Systems 15.1 Introduction: types of system 15.2 Sources 15.3 Transmission medium 15.4 Geometry of coupling of detector to source 15.5 Detectors and signal conditioning elements 15.6 Measurement systems
385 385 387 393 398 403 409
16
Ultrasonic Measurement Systems 16.1 Basic ultrasonic transmission link 16.2 Piezoelectric ultrasonic transmitters and receivers 16.3 Principles of ultrasonic transmission 16.4 Examples of ultrasonic measurement systems
427 427 428 436 447
17
Gas Chromatography 17.1 Principles and basic theory 17.2 Typical gas chromatograph 17.3 Signal processing and operations sequencing
461 461 465 468
18
Data 18.1 18.2 18.3 18.4 18.5 18.6 18.7
475 476 477 478 479 487 490 493
Acquisition and Communication Systems Time division multiplexing Typical data acquisition system Parallel digital signals Serial digital signals Error detection and correction Frequency shift keying Communication systems for measurement
313 313 319 321 339 342
374 378
x CONTENTS 19
The Intelligent Multivariable Measurement System 19.1 The structure of an intelligent multivariable system 19.2 Modelling methods for multivariable systems
503 503 507
Answers to Numerical Problems Index
515 521
Preface to the fourth edition Measurement is an essential activity in every branch of technology and science. We need to know the speed of a car, the temperature of our working environment, the flow rate of liquid in a pipe, the amount of oxygen dissolved in river water. It is important, therefore, that the study of measurement forms part of engineering and science courses in further and higher education. The aim of this book is to provide the fundamental principles of measurement which underlie these studies. The book treats measurement as a coherent and integrated subject by presenting it as the study of measurement systems. A measurement system is an information system which presents an observer with a numerical value corresponding to the variable being measured. A given system may contain four types of element: sensing, signal conditioning, signal processing and data presentation elements. The book is divided into three parts. Part A (Chapters 1 to 7) examines general systems principles. This part begins by discussing the static and dynamic characteristics that individual elements may possess and how they are used to calculate the overall system measurement error, under both steady and unsteady conditions. In later chapters, the principles of loading and two-port networks, the effects of interference and noise on system performance, reliability, maintainability and choice using economic criteria are explained. Part B (Chapters 8 to 11) examines the principles, characteristics and applications of typical sensing, signal conditioning, signal processing and data presentation elements in wide current use. Part C (Chapters 12 to 19) examines a number of specialised measurement systems which have important industrial applications. These are flow measurement systems, intrinsically safe systems, heat transfer, optical, ultrasonic, gas chromatography, data acquisition, communication and intelligent multivariable systems. The fourth edition has been substantially extended and updated to reflect new developments in, and applications of, technology since the third edition was published in 1995. Chapter 1 has been extended to include a wider range of examples of basic measurement systems. New material on solid state sensors has been included in Chapter 8; this includes resistive gas, electrochemical and Hall effect sensors. In Chapter 9 there is now a full analysis of operational amplifier circuits which are used in measurement systems. The section on frequency to digital conversion in Chapter 10 has been expanded; there is also new material on microcontroller structure, software and applications. Chapter 11 has been extensively updated with new material on digital displays, chart and paperless recorders and laser printers. The section on vortex flowmeters in Chapter 12 has been extended and updated. Chapter 19 is a new chapter on intelligent multivariable measurement systems which concentrates on structure and modelling methods. There are around 35 additional problems in this new edition; many of these are at a basic, introductory level.
xii P REFACE TO THE FOURTH EDITION Each chapter in the book is clearly divided into sections. The topics to be covered are introduced at the beginning and reviewed in a conclusion at the end. Basic and important equations are highlighted, and a number of references are given at the end of each chapter; these should provide useful supplementary reading. The book contains about 300 line diagrams and tables and about 140 problems. At the end of the book there are answers to all the numerical problems and a comprehensive index. This book is primarily aimed at students taking modules in measurement and instrumentation as part of degree courses in instrumentation/control, mechanical, manufacturing, electrical, electronic, chemical engineering and applied physics. Much of the material will also be helpful to lecturers and students involved in HNC/HND and foundation degree courses in technology. The book should also be useful to professional engineers and technicians engaged in solving practical measurement problems. I would like to thank academic colleagues, industrial contacts and countless students for their helpful comments and criticism over many years. Thanks are again especially due to my wife Pauline for her constant support and help with the preparation of the manuscript. John P. Bentley Guisborough, December 2003
Acknowledgements We are grateful to the following for permission to reproduce copyright material: Figure 2.1(b) from Repeatability and Accuracy, Council of the Institution of Mechanical Engineers (Hayward, A.T.J., 1977); Figure 2.17(a) from Measurement of length in Journal Institute Measurement & Control, Vol. 12, July (Scarr, A., 1979), Table 5.1 from Systems analysis of instruments in Journal Institute Measurement & Control, Vol. 4, September (Finkelstein, L. and Watts, R.D., 1971), Table 7.3 from The application of reliability engineering to high integrity plant control systems in Measurement and Control, Vol. 18, June (Hellyer, F.G., 1985), and Figures 8.4(a) and (b) from Institute of Measurement and Control; Tables 2.3 and 2.4 from Units of Measurement poster, 8th edition, 1996, and Figures 15.22(a) and (b) from Wavelength encoded optical fibre sensors in N.P.L. News, No. 363 (Hutley, M.C., 1985), National Physical Laboratory; Figure 7.1 from The Institution of Chemical Engineers; Table 7.1 from Instrument reliability in Instrument Science and Technology: Volume 1 (Wright, R.I., 1984), and Figure 16.14 from Medical and industrial applications of high resolution ultrasound in Journal of Physics E: Scientific Instruments, Vol. 18 (Payne, P.A., 1985), Institute of Physics Publishing Ltd.; Table 7.2 from The reliability of instrumentation in Chemistry and Industry, 6 March 1976, Professor F. Lees, Loughborough University; Table 8.2 from BS 4937: 1974 International Thermocouple Reference Tables, and Table 12.1 and Figure 12.7 from BS 1042: 1981 Methods of measurement of fluid flow in closed conduits, British Standards Institution; Figure 8.2(a) from Instrument Transducers: An Introduction to their Performance and Design, 2nd edition, Oxford University Press (Neubert, H.K.P., 1975); Figure 8.3(a) from Technical Information on Two-point NTC Thermistors, 1974, Mullard Ltd.; Table 8.4 from Technical Data on Ion Selective Electrodes, 1984, E.D.T. Research; Figures 8.4(b) and (c) from Thick film polymer sensors for physical variables in Measurement and Control, Vol. 33, No. 4, May, Institute of Measurement and Control and Professor N. White, University of Southampton (Papakostas, T.V. and White, N., 2000); Figures 8.8(a), (b) and (c) from Thick film chemical sensor array allows flexibility in specificity in MTEC 1999, Sensor and Transducer Conference, NEC Birmingham, Trident Exhibitions and Dr. A Cranny, University of Southampton (Jeffrey, P.D. et al., 1999); Figure 8.10 from Ceramics put pressure on conventional transducers in Process Industry Journal, June, Endress and Hauser Ltd. (Stokes, D., 1991); Figure 8.23(b) from Piezoelectric devices: a step nearer problem-free vibration measurement in Transducer Technology, Vol. 4, No. 1 (Purdy, D., 1981), and Figure 8.24 from IC sensors boost potential of measurement systems in Transducer Technology, Vol. 8, No. 4 (Noble, M., 1985), Transducer Technology; Figure 8.25(b) from Analysis with Ion Selective Electrodes, John Wiley
xiv ACKNOWLEDGEMENTS and Sons Ltd. (Bailey, P.L., 1976); Figure 8.25(c) from pH facts – the glass electrode in Kent Technical Review, Kent Industrial Measurements Ltd., E.I.L Analytical Instruments (Thompson, W.); Figure 8.26(a) from Electrical Engineering: Principles and Applications, 2nd edition, reprinted by permission of Pearson Education Inc., Upper Saddle River, NJ, USA (Hambley, A.R.); Table 10.5 from Appendix A, MCS BASIC-52 User’s Manual, Intel Corporation; Figure 11.10(c) from Instrumentation T 292 Block 6, part 2 Displays, 1986, The Open University Press; Figures 11.12(a) and (b) from Trident Displays technical literature on EL displays, Trident Microsystems Ltd. and M.S. Caddy and D. Weber; Figure 12.11(a) from Kent Process Control Ltd., Flow Products; Figure 15.10 from Optical Fibre Communications (Keiser, G., 1983), and Figure 15.12(b) from Measurement Systems: Application and Design (Doebelin, E.O., 1976), McGraw-Hill Book Co. (USA); Table 16.1 from Piezoelectric transducers in Methods of Experimental Physics, Vol. 19, Academic Press (O’Donnell, M., Busse, L.J. and Miller, J.G., 1981); Table 16.2 from Ultrasonics: Methods and Applications, Butterworth and Co. (Blitz, J., 1971); Figure 16.15 from ultrasonic image of Benjamin Stefan Morton, Nottingham City Hospital NHS Trust and Sarah Morton; Figure 17.5 from Process gas chromatography in Talanta 1967, Vol. 14, Pergamon Press Ltd. (Pine, C.S.F., 1967); Error Detection System in Section 18.5.2 from Technical Information on Kent P4000 Telemetry Systems, 1985, Kent Automation Systems Ltd. In some instances we have been unable to trace the owners of copyright material, and we would appreciate any information that would enable us to do so.
Part A General Principles
1 The General Measurement System 1.1
Purpose and performance of measurement systems We begin by defining a process as a system which generates information. Examples are a chemical reactor, a jet fighter, a gas platform, a submarine, a car, a human heart, and a weather system. Table 1.1 lists information variables which are commonly generated by processes: thus a car generates displacement, velocity and acceleration variables, and a chemical reactor generates temperature, pressure and composition variables.
Table 1.1 Common information/measured variables.
Acceleration Velocity Displacement Force–Weight Pressure Torque Volume Mass Flow rate Level
Density Viscosity Composition pH Humidity Temperature Heat/Light flux Current Voltage Power
We then define the observer as a person who needs this information from the process. This could be the car driver, the plant operator or the nurse. The purpose of the measurement system is to link the observer to the process, as shown in Figure 1.1. Here the observer is presented with a number which is the current value of the information variable. We can now refer to the information variable as a measured variable. The input to the measurement system is the true value of the variable; the system output is the measured value of the variable. In an ideal measurement system, the measured Figure 1.1 Purpose of measurement system.
4 TH E G ENERAL MEASUREMENT SY STEM value would be equal to the true value. The accuracy of the system can be defined as the closeness of the measured value to the true value. A perfectly accurate system is a theoretical ideal and the accuracy of a real system is quantified using measurement system error E, where E = measured value − true value E = system output − system input Thus if the measured value of the flow rate of gas in a pipe is 11.0 m3/h and the true value is 11.2 m3/h, then the error E = −0.2 m3/h. If the measured value of the rotational speed of an engine is 3140 rpm and the true value is 3133 rpm, then E = +7 rpm. Error is the main performance indicator for a measurement system. The procedures and equipment used to establish the true value of the measured variable will be explained in Chapter 2.
1.2
Structure of measurement systems The measurement system consists of several elements or blocks. It is possible to identify four types of element, although in a given system one type of element may be missing or may occur more than once. The four types are shown in Figure 1.2 and can be defined as follows.
Figure 1.2 General structure of measurement system.
Sensing element This is in contact with the process and gives an output which depends in some way on the variable to be measured. Examples are:
• • •
Thermocouple where millivolt e.m.f. depends on temperature Strain gauge where resistance depends on mechanical strain Orifice plate where pressure drop depends on flow rate.
If there is more than one sensing element in a system, the element in contact with the process is termed the primary sensing element, the others secondary sensing elements.
Signal conditioning element This takes the output of the sensing element and converts it into a form more suitable for further processing, usually a d.c. voltage, d.c. current or frequency signal. Examples are:
• • •
Deflection bridge which converts an impedance change into a voltage change Amplifier which amplifies millivolts to volts Oscillator which converts an impedance change into a variable frequency voltage.
1.3 EXAMPLES OF MEASUREMENT SYSTEMS
5
Signal processing element This takes the output of the conditioning element and converts it into a form more suitable for presentation. Examples are:
• •
Analogue-to-digital converter (ADC) which converts a voltage into a digital form for input to a computer Computer which calculates the measured value of the variable from the incoming digital data.
Typical calculations are:
• • •
Computation of total mass of product gas from flow rate and density data Integration of chromatograph peaks to give the composition of a gas stream Correction for sensing element non-linearity.
Data presentation element This presents the measured value in a form which can be easily recognised by the observer. Examples are:
• • • • 1.3
Simple pointer–scale indicator Chart recorder Alphanumeric display Visual display unit (VDU).
Examples of measurement systems Figure 1.3 shows some typical examples of measurement systems. Figure 1.3(a) shows a temperature system with a thermocouple sensing element; this gives a millivolt output. Signal conditioning consists of a circuit to compensate for changes in reference junction temperature, and an amplifier. The voltage signal is converted into digital form using an analogue-to-digital converter, the computer corrects for sensor non-linearity, and the measured value is displayed on a VDU. In Figure 1.3(b) the speed of rotation of an engine is sensed by an electromagnetic tachogenerator which gives an a.c. output signal with frequency proportional to speed. The Schmitt trigger converts the sine wave into sharp-edged pulses which are then counted over a fixed time interval. The digital count is transferred to a computer which calculates frequency and speed, and the speed is presented on a digital display. The flow system of Figure 1.3(c) has an orifice plate sensing element; this gives a differential pressure output. The differential pressure transmitter converts this into a current signal and therefore combines both sensing and signal conditioning stages. The ADC converts the current into digital form and the computer calculates the flow rate, which is obtained as a permanent record on a chart recorder. The weight system of Figure 1.3(d) has two sensing elements: the primary element is a cantilever which converts weight into strain; the strain gauge converts this into a change in electrical resistance and acts as a secondary sensor. There are two signal conditioning elements: the deflection bridge converts the resistance change into
Figure 1.3 Examples of measurement systems.
6 TH E G ENERAL MEASUREMENT SY STEM
CONCLUSION
7
millivolts and the amplifier converts millivolts into volts. The computer corrects for non-linearity in the cantilever and the weight is presented on a digital display. The word ‘transducer’ is commonly used in connection with measurement and instrumentation. This is a manufactured package which gives an output voltage (usually) corresponding to an input variable such as pressure or acceleration. We see therefore that such a transducer may incorporate both sensing and signal conditioning elements; for example a weight transducer would incorporate the first four elements shown in Figure 1.3(d). It is also important to note that each element in the measurement system may itself be a system made up of simpler components. Chapters 8 to 11 discuss typical examples of each type of element in common use.
1.4
Block diagram symbols A block diagram approach is very useful in discussing the properties of elements and systems. Figure 1.4 shows the main block diagram symbols used in this book.
Figure 1.4 Block diagram symbols.
Conclusion This chapter has defined the purpose of a measurement system and explained the importance of system error. It has shown that, in general, a system consists of four types of element: sensing, signal conditioning, signal processing and data presentation elements. Typical examples have been given.
2 Static Characteristics of Measurement System Elements
Figure 2.1 Meaning of element characteristics.
2.1
In the previous chapter we saw that a measurement system consists of different types of element. The following chapters discuss the characteristics that typical elements may possess and their effect on the overall performance of the system. This chapter is concerned with static or steady-state characteristics; these are the relationships which may occur between the output O and input I of an element when I is either at a constant value or changing slowly (Figure 2.1).
Systematic characteristics Systematic characteristics are those that can be exactly quantified by mathematical or graphical means. These are distinct from statistical characteristics which cannot be exactly quantified and are discussed in Section 2.3.
Range The input range of an element is specified by the minimum and maximum values of I, i.e. IMIN to IMAX. The output range is specified by the minimum and maximum values of O, i.e. OMIN to OMAX. Thus a pressure transducer may have an input range of 0 to 104 Pa and an output range of 4 to 20 mA; a thermocouple may have an input range of 100 to 250 °C and an output range of 4 to 10 mV.
Span Span is the maximum variation in input or output, i.e. input span is IMAX – IMIN, and output span is OMAX – OMIN . Thus in the above examples the pressure transducer has an input span of 104 Pa and an output span of 16 mA; the thermocouple has an input span of 150 °C and an output span of 6 mV.
Ideal straight line An element is said to be linear if corresponding values of I and O lie on a straight line. The ideal straight line connects the minimum point A(IMIN, OMIN ) to maximum point B(IMAX, OMAX ) (Figure 2.2) and therefore has the equation:
10 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS
O – OMIN =
Ideal straight line equation
G OMAX – OMIN J (I – IMIN ) I IMAX – IMIN L
OIDEAL = KI + a
[2.1]
[2.2]
where: K = ideal straight-line slope =
OMAX – OMIN IMAX – IMIN
and a = ideal straight-line intercept = OMIN − KIMIN Thus the ideal straight line for the above pressure transducer is: O = 1.6 × 10−3I + 4.0 The ideal straight line defines the ideal characteristics of an element. Non-ideal characteristics can then be quantified in terms of deviations from the ideal straight line.
Non-linearity In many cases the straight-line relationship defined by eqn [2.2] is not obeyed and the element is said to be non-linear. Non-linearity can be defined (Figure 2.2) in terms of a function N(I) which is the difference between actual and ideal straight-line behaviour, i.e. N(I ) = O(I ) − (KI + a)
[2.3]
O(I ) = KI + a + N(I )
[2.4]
or
Non-linearity is often quantified in terms of the maximum non-linearity ; expressed as a percentage of full-scale deflection (f.s.d.), i.e. as a percentage of span. Thus: ; Max. non-linearity as = × 100% a percentage of f.s.d. OMAX – OMIN
Figure 2.2 Definition of non-linearity.
[2.5]
2.1 SYSTEMATIC CHARACTERISTICS
11
As an example, consider a pressure sensor where the maximum difference between actual and ideal straight-line output values is 2 mV. If the output span is 100 mV, then the maximum percentage non-linearity is 2% of f.s.d. In many cases O(I ) and therefore N(I ) can be expressed as a polynomial in I: q m
O(I ) = a0 + a1I + a2 I 2 + . . . + aq I q + . . . + am I m =
∑a I q
q
[2.6]
q 0
An example is the temperature variation of the thermoelectric e.m.f. at the junction of two dissimilar metals. For a copper–constantan (Type T) thermocouple junction, the first four terms in the polynomial relating e.m.f. E(T ), expressed in µV, and junction temperature T °C are: E(T ) = 38.74T + 3.319 × 10−2T 2 + 2.071 × 10−4T 3 − 2.195 × 10−6T 4 + higher-order terms up to T 8
[2.7a]
for the range 0 to 400 °C.[1] Since E = 0 µV at T = 0 °C and E = 20 869 µV at T = 400 °C, the equation to the ideal straight line is: EIDEAL = 52.17T
[2.7b]
and the non-linear correction function is: N(T ) = E(T ) − EIDEAL = −13.43T + 3.319 × 10−2T 2 + 2.071 × 10− 4T 3 − 2.195 × 10−6T 4 + higher-order terms
[2.7c]
In some cases expressions other than polynomials are more appropriate: for example the resistance R(T ) ohms of a thermistor at T °C is given by: R(T ) = 0.04 exp
A 3300 D C T + 273F
[2.8]
Sensitivity This is the change ∆O in output O for unit change ∆I in input I, i.e. it is the ratio ∆O/∆I. In the limit that ∆I tends to zero, the ratio ∆O/∆I tends to the derivative dO/dI, which is the rate of change of O with respect to I. For a linear element dO/dI is equal to the slope or gradient K of the straight line; for the above pressure transducer the sensitivity is 1.6 × 10−3 mA/Pa. For a non-linear element dO/dI = K + dN/dI, i.e. sensitivity is the slope or gradient of the output versus input characteristics O(I). Figure 2.3 shows the e.m.f. versus temperature characteristics E(T ) for a Type T thermocouple (eqn [2.7a] ). We see that the gradient and therefore the sensitivity vary with temperature: at 100 °C it is approximately 35 µV/°C and at 200 °C approximately 42 µV/°C.
Environmental effects In general, the output O depends not only on the signal input I but on environmental inputs such as ambient temperature, atmospheric pressure, relative humidity, supply voltage, etc. Thus if eqn [2.4] adequately represents the behaviour of the element under ‘standard’ environmental conditions, e.g. 20 °C ambient temperature,
12 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Figure 2.3 Thermocouple sensitivity.
1000 millibars atmospheric pressure, 50% RH and 10 V supply voltage, then the equation must be modified to take account of deviations in environmental conditions from ‘standard’. There are two main types of environmental input. A modifying input IM causes the linear sensitivity of an element to change. K is the sensitivity at standard conditions when IM = 0. If the input is changed from the standard value, then IM is the deviation from standard conditions, i.e. (new value – standard value). The sensitivity changes from K to K + KM IM, where KM is the change in sensitivity for unit change in IM . Figure 2.4(a) shows the modifying effect of ambient temperature on a linear element. An interfering input II causes the straight line intercept or zero bias to change. a is the zero bias at standard conditions when II = 0. If the input is changed from the standard value, then II is the deviation from standard conditions, i.e. (new value – standard value). The zero bias changes from a to a + KI II, where KI is the change in zero bias for unit change in II . Figure 2.4(b) shows the interfering effect of ambient temperature on a linear element. KM and KI are referred to as environmental coupling constants or sensitivities. Thus we must now correct eqn [2.4], replacing KI with (K + KM IM)I and replacing a with a + KI II to give: O = KI + a + N(I) + KM IM I + KI II
Figure 2.4 Modifying and interfering inputs.
[2.9]
2.1 SYSTEMATIC CHARACTERISTICS
13
An example of a modifying input is the variation ∆VS in the supply voltage VS of the potentiometric displacement sensor shown in Figure 2.5. An example of an interfering input is provided by variations in the reference junction temperature T2 of the thermocouple (see following section and Section 8.5). If x is the fractional displacement, then VOUT = (VS + ∆VS )x = VS x + ∆VS x Figure 2.5
Hysteresis For a given value of I, the output O may be different depending on whether I is increasing or decreasing. Hysteresis is the difference between these two values of O (Figure 2.6), i.e. Hysteresis H(I ) = O(I )I↓ − O(I)I ↑
[2.10]
Again hysteresis is usually quantified in terms of the maximum hysteresis à expressed as a percentage of f.s.d., i.e. span. Thus: Maximum hysteresis as a percentage of f.s.d. =
à × 100% [2.11] OMAX – OMIN
A simple gear system (Figure 2.7) for converting linear movement into angular rotation provides a good example of hysteresis. Due to the ‘backlash’ or ‘play’ in the gears the angular rotation θ, for a given value of x, is different depending on the direction of the linear movement.
Resolution Some elements are characterised by the output increasing in a series of discrete steps or jumps in response to a continuous increase in input (Figure 2.8). Resolution is defined as the largest change in I that can occur without any corresponding change in O.
Figure 2.6 Hysteresis.
Figure 2.7 Backlash in gears.
14 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Figure 2.8 Resolution and potentiometer example.
Thus in Figure 2.8 resolution is defined in terms of the width ∆IR of the widest step; resolution expressed as a percentage of f.s.d. is thus: ∆IR × 100% IMAX – IMIN A common example is a wire-wound potentiometer (Figure 2.8); in response to a continuous increase in x the resistance R increases in a series of steps, the size of each step being equal to the resistance of a single turn. Thus the resolution of a 100 turn potentiometer is 1%. Another example is an analogue-to-digital converter (Chapter 10); here the output digital signal responds in discrete steps to a continuous increase in input voltage; the resolution is the change in voltage required to cause the output code to change by the least significant bit.
Wear and ageing These effects can cause the characteristics of an element, e.g. K and a, to change slowly but systematically throughout its life. One example is the stiffness of a spring k(t) decreasing slowly with time due to wear, i.e. k(t) = k0 − bt
[2.12]
where k0 is the initial stiffness and b is a constant. Another example is the constants a1, a2, etc. of a thermocouple, measuring the temperature of gas leaving a cracking furnace, changing systematically with time due to chemical changes in the thermocouple metals.
Error bands Non-linearity, hysteresis and resolution effects in many modern sensors and transducers are so small that it is difficult and not worthwhile to exactly quantify each individual effect. In these cases the manufacturer defines the performance of the element in terms of error bands (Figure 2.9). Here the manufacturer states that for any value of I, the output O will be within ±h of the ideal straight-line value OIDEAL. Here an exact or systematic statement of performance is replaced by a statistical statement in terms of a probability density function p(O). In general a probability density function p(x) is x defined so that the integral ∫ x12 p(x) dx (equal to the area under the curve in Figure 2.10 between x1 and x2) is the probability Px1, x2 of x lying between x1 and x2 (Section 6.2). In this case the probability density function is rectangular (Figure 2.9), i.e.
2.2 GE N E R ALI S E D MO D E L O F A S Y S TE M E LE ME NT
15
Figure 2.9 Error bands and rectangular probability density function.
Figure 2.10 Probability density function.
1 1 4 = 2h p(O) 2 4= 0 3= 0
Oideal − h ≤ O ≤ Oideal + h O > Oideal + h Oideal − h > O
[2.13]
We note that the area of the rectangle is equal to unity: this is the probability of O lying between OIDEAL − h and OIDEAL + h.
2.2
Generalised model of a system element If hysteresis and resolution effects are not present in an element but environmental and non-linear effects are, then the steady-state output O of the element is in general given by eqn [2.9], i.e.: O = KI + a + N(I) + KM IM I + KI II
[2.9]
Figure 2.11 shows this equation in block diagram form to represent the static characteristics of an element. For completeness the diagram also shows the transfer function G(s), which represents the dynamic characteristics of the element. The meaning of transfer function will be explained in Chapter 4 where the form of G(s) for different elements will be derived. Examples of this general model are shown in Figure 2.12(a), (b) and (c), which summarise the static and dynamic characteristics of a strain gauge, thermocouple and accelerometer respectively.
16 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Figure 2.11 General model of element.
Figure 2.12 Examples of element characteristics: (a) Strain gauge (b) Copper–constantan thermocouple (c) Accelerometer.
2.3 STATISTICAL CHARACTERISTICS
17
The strain gauge has an unstrained resistance of 100 Ω and gauge factor (Section 8.1) of 2.0. Non-linearity and dynamic effects can be neglected, but the resistance of the gauge is affected by ambient temperature as well as strain. Here temperature acts as both a modifying and an interfering input, i.e. it affects both gauge sensitivity and resistance at zero strain. Figure 2.12(b) represents a copper–constantan thermocouple between 0 and 400 °C. The figure is drawn using eqns [2.7b] and [2.7c] for ideal straight-line and non-linear correction functions; these apply to a single junction. A thermocouple installation consists of two junctions (Section 8.5) – a measurement junction at T1 °C and a reference junction at T 2 °C. The resultant e.m.f. is the difference of the two junction potentials and thus depends on both T1 and T2, i.e. E(T1, T2) = E(T1) − E(T2); T2 is thus an interfering input. The model applies to the situation where T2 is small compared with T1, so that E(T2) can be approximated by 38.74 T2, the largest term in eqn [2.7a]. The dynamics are represented by a first-order transfer function of time constant 10 seconds (Chapters 4 and 14). Figure 2.12(c) represents an accelerometer with a linear sensitivity of 0.35 mV m−1 s2 and negligible non-linearity. Any transverse acceleration aT , i.e. any acceleration perpendicular to that being measured, acts as an interfering input. The dynamics are represented by a second-order transfer function with a natural frequency of 250 Hz and damping coefficient of 0.7 (Chapters 4 and 8).
2.3 2.3.1
Statistical characteristics Statistical variations in the output of a single element with time – repeatability Suppose that the input I of a single element, e.g. a pressure transducer, is held constant, say at 0.5 bar, for several days. If a large number of readings of the output O are taken, then the expected value of 1.0 volt is not obtained on every occasion; a range of values such as 0.99, 1.01, 1.00, 1.02, 0.98, etc., scattered about the expected value, is obtained. This effect is termed a lack of repeatability in the element. Repeatability is the ability of an element to give the same output for the same input, when repeatedly applied to it. Lack of repeatability is due to random effects in the element and its environment. An example is the vortex flowmeter (Section 12.2.4): for a fixed flow rate Q = 1.4 × 10−2 m3 s−1, we would expect a constant frequency output f = 209 Hz. Because the output signal is not a perfect sine wave, but is subject to random fluctuations, the measured frequency varies between 207 and 211 Hz. The most common cause of lack of repeatability in the output O is random fluctuations with time in the environmental inputs IM, II : if the coupling constants KM , KI are non-zero, then there will be corresponding time variations in O. Thus random fluctuations in ambient temperature cause corresponding time variations in the resistance of a strain gauge or the output voltage of an amplifier; random fluctuations in the supply voltage of a deflection bridge affect the bridge output voltage. By making reasonable assumptions for the probability density functions of the inputs I, IM and II (in a measurement system random variations in the input I to
18 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Figure 2.13 Normal probability density function with P = 0.
a given element can be caused by random effects in the previous element), the probability density function of the element output O can be found. The most likely probability density function for I, IM and II is the normal or Gaussian distribution function (Figure 2.13): Normal probability density function
p( x ) =
 ( x − P)2  exp −  2σ 2   σ 2π 1
[2.14]
where: P = mean or expected value (specifies centre of distribution) σ = standard deviation (specifies spread of distribution). Equation [2.9] expresses the independent variable O in terms of the independent variables I, IM and II. Thus if ∆O is a small deviation in O from the mean value /, caused by deviations ∆I, ∆IM and ∆II from respective mean values -, -M and -I , then: ∆O =
A ∂O D A ∂O D A ∂O D ∆I + ∆I + ∆I C ∂I F C ∂IM F M C ∂II F I
[2.15]
Thus ∆O is a linear combination of the variables ∆I, ∆IM and ∆II ; the partial derivatives can be evaluated using eqn [2.9]. It can be shown[2] that if a dependent variable y is a linear combination of independent variables x1, x 2 and x3, i.e. y = a1x1 + a2 x2 + a3 x3
[2.16]
and if x1, x2 and x3 have normal distributions with standard deviations σ1, σ2 and σ3 respectively, then the probability distribution of y is also normal with standard deviation σ given by:
σ =
a12σ 12 + a 22 σ 22 + a32 σ 32
[2.17]
From eqns [2.15] and [2.17] we see that the standard deviation of ∆O, i.e. of O about mean O, is given by:
2.3 STATISTICAL CHARACTERISTICS
Standard deviation of output for a single element
2
σ0 =
2
19
2
 ∂O  2  ∂O  2  ∂O  2  σI  σI +   σI +    ∂II   ∂IM   ∂I  M
I
[2.18]
where σI , σIM and σII are the standard deviations of the inputs. Thus σ0 can be calculated using eqn [2.18] if σI , σ IM and σ II are known; alternatively if a calibration test (see following section) is being performed on the element then σ0 can be estimated directly from the experimental results. The corresponding mean or expected value / of the element output is given by: Mean value of output for a single element
/ = K- + a + N(- ) + KM -M - + KI - I
[2.19]
and the corresponding probability density function is: p (O ) =
2.3.2
1
σ0
 − (O − / ) 2  exp    2σ 02  2π
[2.20]
Statistical variations amongst a batch of similar elements – tolerance Suppose that a user buys a batch of similar elements, e.g. a batch of 100 resistance temperature sensors, from a manufacturer. If he then measures the resistance R0 of each sensor at 0 °C he finds that the resistance values are not all equal to the manufacturer’s quoted value of 100.0 Ω. A range of values such as 99.8, 100.1, 99.9, 100.0 and 100.2 Ω, distributed statistically about the quoted value, is obtained. This effect is due to small random variations in manufacture and is often well represented by the normal probability density function given earlier. In this case we have: p (R0 ) =
1
σR
0
 − (R0 − =0 )2  exp   2σ R2 2π  
[2.21]
0
where = 0 = mean value of distribution = 100 Ω and σR0 = standard deviation, typically 0.1 Ω. However, a manufacturer may state in his specification that R0 lies within ±0.15 Ω of 100 Ω for all sensors, i.e. he is quoting tolerance limits of ±0.15 Ω. Thus in order to satisfy these limits he must reject for sale all sensors with R0 < 99.85 Ω and R0 > 100.15 Ω, so that the probability density function of the sensors bought by the user now has the form shown in Figure 2.14. The user has two choices: (a)
(b)
He can design his measurement system using the manufacturer’s value of R 0 = 100.0 Ω and accept that any individual system, with R 0 = 100.1 Ω say, will have a small measurement error. This is the usual practice. He can perform a calibration test to measure R0 as accurately as possible for each element in the batch. This theoretically removes the error due to uncertainty in R 0 but is time-consuming and expensive. There is also a small remaining uncertainty in the value of R 0 due to the limited accuracy of the calibration equipment.
20 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Figure 2.14 Tolerance limits.
This effect is found in any batch of ‘identical’ elements; significant variations are found in batches of thermocouples and thermistors, for example. In the general case we can say that the values of parameters, such as linear sensitivity K and zero bias a, for a batch of elements are distributed statistically about mean values . and ã.
2.3.3
Summary In the general case of a batch of several ‘identical’ elements, where each element is subject to random variations in environmental conditions with time, both inputs I, IM and II and parameters K, a, etc., are subject to statistical variations. If we assume that each statistical variation can be represented by a normal probability density function, then the probability density function of the element output O is also normal, i.e.: p (O ) =
1
σ0
 − (O − / ) 2  exp    2σ 02  2π
[2.20]
where the mean value / is given by: Mean value of output for a batch of elements
/ = .- + å(- ) + ã + .M -M - + .I -I
[2.22]
and the standard deviation σ0 is given by: Standard deviation of output for a batch of elements
2
σ0 =
2
2
2
2
 ∂O  2  ∂O  2  ∂O  2  ∂O  2  ∂O  2  σa + . . .  σK +   σI +   σI +   σI +    ∂a   ∂K   ∂II   ∂IM   ∂I  M
I
[2.23] Tables 2.1 and 2.2 summarise the static characteristics of a chromel–alumel thermocouple and a millivolt to current temperature transmitter. The thermocouple is
2.4 I D E N T I F I CATI O N O F S T AT I C CH AR ACTE RI S T I CS – CALI BRATI ON
Table 2.1 Model for chromel–alumel thermocouple.
Table 2.2 Model for millivolt to current temperature transmitter.
21
Model equation
ET,Ta = a0 + a1(T − Ta ) + a 2(T 2 − T 2a ) (50 to 150 °C)
Mean values
ã0 = 0.00, ã1 = 4.017 × 10−2, ã2 = 4.66 × 10 −6, >a = 10
Standard deviations
σa0 = 6.93 × 10 −2, σa1 = 0.0, σa2 = 0.0, σTa = 6.7
Partial derivatives
∂E = 1.0, ∂a0
Mean value of output
+T,Ta = ã0 + ã1(> − >a ) + ã2(> 2 − > 2a )
Standard deviation of output
σ E2 =
∂E = −4.026 × 10 −2 ∂Ta
A ∂E D 2 2 A ∂E D 2 2 σ a0 + σ Ta C ∂a0 F C ∂TaF
Model equation
i = KE + KM E∆Ta + KI ∆Ta + a 4 to 20 mA output for 2.02 to 6.13 mV input ∆Ta = deviation in ambient temperature from 20 °C
Mean values
. = 3.893, ã = 3.864, ∆>a = −10 .M = 1.95 × 10 − 4, .I = 2.00 × 10−3
Standard deviations
σa = 0.14, σ∆Ta = 6.7 σK = 0.0, σKM = 0.0, σKI = 0.0
Partial derivatives
∂i = 3.891, ∂E
∂i = 2.936 × 10−3, ∂∆Ta
Mean value of output
j = .+ + .M +∆>a + .I ∆>a + ã
Standard deviation of output
σ 2i =
∂i = 1.0 ∂a
A ∂i D 2 2 A ∂i D 2 2 A ∂i D 2 2 σE + σ ∆T + σa C ∂E F C ∂a F C ∂∆TaF
characterised by non-linearity, changes in reference junction (ambient temperature) Ta acting as an interfering input, and a spread of zero bias values a0. The transmitter is linear but is affected by ambient temperature acting as both a modifying and an interfering input. The zero and sensitivity of this element are adjustable; we cannot be certain that the transmitter is set up exactly as required, and this is reflected in a non-zero value of the standard deviation of the zero bias a.
2.4 2.4.1
Identification of static characteristics – calibration Standards The static characteristics of an element can be found experimentally by measuring corresponding values of the input I, the output O and the environmental inputs IM and II, when I is either at a constant value or changing slowly. This type of experiment is referred to as calibration, and the measurement of the variables I, O, IM and II must be accurate if meaningful results are to be obtained. The instruments and techniques used to quantify these variables are referred to as standards (Figure 2.15).
22 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Figure 2.15 Calibration of an element.
The accuracy of a measurement of a variable is the closeness of the measurement to the true value of the variable. It is quantified in terms of measurement error, i.e. the difference between the measured value and the true value (Chapter 3). Thus the accuracy of a laboratory standard pressure gauge is the closeness of the reading to the true value of pressure. This brings us back to the problem, mentioned in the previous chapter, of how to establish the true value of a variable. We define the true value of a variable as the measured value obtained with a standard of ultimate accuracy. Thus the accuracy of the above pressure gauge is quantified by the difference between the gauge reading, for a given pressure, and the reading given by the ultimate pressure standard. However, the manufacturer of the pressure gauge may not have access to the ultimate standard to measure the accuracy of his products. In the United Kingdom the manufacturer is supported by the National Measurement System. Ultimate or primary measurement standards for key physical variables such as time, length, mass, current and temperature are maintained at the National Physical Laboratory (NPL). Primary measurement standards for other important industrial variables such as the density and flow rate of gases and liquids are maintained at the National Engineering Laboratory (NEL). In addition there is a network of laboratories and centres throughout the country which maintain transfer or intermediate standards. These centres are accredited by UKAS (United Kingdom Accreditation Service). Transfer standards held at accredited centres are calibrated against national primary and secondary standards, and a manufacturer can calibrate his products against the transfer standard at a local centre. Thus the manufacturer of pressure gauges can calibrate his products against a transfer standard, for example a deadweight tester. The transfer standard is in turn calibrated against a primary or secondary standard, for example a pressure balance at NPL. This introduces the concept of a traceability ladder, which is shown in simplified form in Figure 2.16. The element is calibrated using the laboratory standard, which should itself be calibrated using the transfer standard, and this in turn should be calibrated using the primary standard. Each element in the ladder should be significantly more accurate than the one below it. NPL are currently developing an Internet calibration service.[3] This will allow an element at a remote location (for example at a user’s factory) to be calibrated directly against a national primary or secondary standard without having to be transported to NPL. The traceability ladder is thereby collapsed to a single link between element and national standard. The same input must be applied to element and standard. The measured value given by the standard instrument is then the true value of the input to the element, and this is communicated to the user via the Internet. If the user measures the output of the element for a number of true values of input, then the characteristics of the element can be determined to a known, high accuracy.
2.4 I D E N T I F I CATI O N O F S T AT I C CH AR ACTE RI S T I CS – CALI BRATI ON
23
Figure 2.16 Simplified traceability ladder.
2.4.2
SI units Having introduced the concepts of standards and traceability we can now discuss different types of standards in more detail. The International System of Units (SI) comprises seven base units, which are listed and defined in Table 2.3. The units of all physical quantities can be derived from these base units. Table 2.4 lists common physical quantities and shows the derivation of their units from the base units. In the United Kingdom the National Physical Laboratory (NPL) is responsible for the
Table 2.3 SI base units (after National Physical Laboratory ‘Units of Measurement’ poster, 1996[4]). Time: second (s)
The second is the duration of 9 192 631 770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the caesium-133 atom.
Length: metre (m)
The metre is the length of the path travelled by light in vacuum during a time interval of 1/299 792 458 of a second.
Mass: kilogram (kg)
The kilogram is the unit of mass; it is equal to the mass of the international prototype of the kilogram.
Electric current: ampere (A)
The ampere is that constant current which, if maintained in two straight parallel conductors of infinite length, of negligible circular cross-section, and placed 1 metre apart in vacuum, would produce between these conductors a force equal to 2 × 10−7 newton per metre of length.
Thermodynamic temperature: kelvin (K)
The kelvin, unit of thermodynamic temperature, is the fraction 1/273.16 of the thermodynamic temperature of the triple point of water.
Amount of substance: mole (mol)
The mole is the amount of substance of a system which contains as many elementary entities as there are atoms in 0.012 kilogram of carbon-12.
Luminous intensity: candela (cd)
The candela is the luminous intensity, in a given direction, of a source that emits monochromatic radiation of frequency 540 × 1012 hertz and that has a radiant intensity in that direction of (1/683) watt per steradian.
24 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Table 2.4 SI derived units (after National Physical Laboratory Units of Measurement poster, 1996[4] ).
Examples of SI derived units expressed in terms of base units Quantity
SI unit
area volume speed, velocity acceleration wave number density, mass density specific volume current density magnetic field strength concentration (of amount of substance) luminance
Name
Symbol
square metre cubic metre metre per second metre per second squared 1 per metre kilogram per cubic metre cubic metre per kilogram ampere per square metre ampere per metre mole per cubic metre candela per square metre
m2 m3 m/s m/s2 m−1 kg/m3 m3/kg A/m2 A/m mol/m3 cd/m2
SI derived units with special names Quantity
plane angleb solid angleb frequency force pressure, stress energy, work quantity of heat power, radiant flux electric charge, quantity of electricity electric potential, potential difference, electromotive force capacitance electric resistance electric conductance magnetic flux magnetic flux density inductance Celsius temperature luminous flux illuminance activity (of a radionuclide) absorbed dose, specific energy imparted, kerma dose equivalent
SI unit Expression in terms of other units
Expressiona in terms of SI base units
rad sr Hz N Pa
N/m2
m ⋅ m−1 = 1 m2 ⋅ m−2 = 1 s−1 m kg s−2 m−1 kg s−2
joule watt
J W
Nm J/s
m2 kg s−2 m2 kg s−3
coulomb
C
volt farad ohm siemens weber tesla henry degree Celsius lumen lux becquerel
V F Ω S Wb T H °C lm lx Bq
W/A C/V V/A A/V Vs Wb/m2 Wb/A
gray sievert
Gy Sv
J/kg J/kg
Name
Symbol
radian steradian hertz newton pascal
sA
cd sr lm/m2
m2 kg s−3 A−1 m−2 kg−1 s4 A2 m2 kg s−3 A−2 m−2 kg−1 s3 A2 m2 kg s−2 A−1 kg s−2 A−1 m2 kg s−2 A−2 K cd ⋅ m2 ⋅ m−2 = cd cd ⋅ m2 ⋅ m− 4 = cd ⋅ m−2 s−1 m2 s−2 m2 s−2
2.4 I D E N T I F I CATI O N O F S T AT I C CH AR ACTE RI S T I CS – CALI BRATI ON
Table 2.4 (cont’d)
25
Examples of SI derived units expressed by means of special names Quantity
dynamic viscosity moment of force surface tension heat flux density, irradiance heat capacity, entropy specific heat capacity, specific entropy specific energy thermal conductivity energy density electric field strength electric charge density electric flux density permittivity permeability molar energy molar entropy, molar heat capacity exposure (X and γ rays) absorbed dose rate
SI unit Name
Symbol
Expression in terms of SI base units
pascal second newton metre newton per metre
Pa s Nm N/m
m−1 kg s−1 m2 kg s−2 kg s−2
watt per square metre joule per kelvin
W/m2 J/K
kg s−3 m2 kg s−2 K−1
joule per kilogram kelvin joule per kilogram watt per metre kelvin joule per cubic metre volt per metre coulomb per cubic metre coulomb per square metre farad per metre henry per metre joule per mole
J/(kg K) J/kg W/(m K) J/m3 V/m C/m3 C/m2 F/m H/m J/mol
m2 s−2 K−1 m2 s−2 m kg s−3 K−1 m−1 kg s−2 m kg s−3 A−1 m−3 s A m−2 s A m−3 kg−1 s4 A2 m kg s−2 A−2 m2 kg s−2 mol−1
joule per mole kelvin coulomb per kilogram gray per second
J/(mol K) C/kg Gy/s
m2 kg s−2 K−1 mol−1 kg−1 s A m2 s−3
Examples of SI derived units formed by using the radian and steradian Quantity
angular velocity angular acceleration radiant intensity radiance a b
SI unit Name
Symbol
radian per second radian per second squared watt per steradian watt per square metre steradian
rad/s rad/s2 W/sr W m−2 sr −1
Acceptable forms are, for example, m ⋅ kg ⋅ s−2, m kg s −2 ; m/s, –ms or m ⋅ s−1. The CIPM (1995) decided that the radian and steradian should henceforth be designated as dimensionless derived units.
physical realisation of all of the base units and many of the derived units mentioned above. The NPL is therefore the custodian of ultimate or primary standards in the UK. There are secondary standards held at United Kingdom Accreditation Service (UKAS) centres. These have been calibrated against NPL standards and are available to calibrate transfer standards. At NPL, the metre is realised using the wavelength of the 633 nm radiation from an iodine-stabilised helium–neon laser. The reproducibility of this primary standard
26 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Figure 2.17 Traceability ladders: (a) length (adapted from Scarr [5] ) (b) mass (reprinted by permission of the Council of the Institution of Mechanical Engineers from Hayward[6] ).
is about 3 parts in 1011, and the wavelength of the radiation has been accurately related to the definition of the metre in terms of the velocity of light. The primary standard is used to calibrate secondary laser interferometers which are in turn used to calibrate precision length bars, gauges and tapes. A simplified traceability ladder for length[5] is shown in Figure 2.17(a). The international prototype of the kilogram is made of platinum–iridium and is kept at the International Bureau of Weights and Measures (BIPM) in Paris. The British national copy is kept at NPL and is used in conjunction with a precision balance to calibrate secondary and transfer kilogram standards. Figure 2.17(b) shows a simplified traceability ladder for mass and weight.[6] The weight of a mass m is the force mg it experiences under the acceleration of gravity g; thus if the local value of g is known accurately, then a force standard can be derived from mass standards. At NPL deadweight machines covering a range of forces from 450 N to 30 MN are used to calibrate strain-gauge load cells and other weight transducers. The second is realised by caesium beam standards to about 1 part in 1013; this is equivalent to one second in 300 000 years! A uniform timescale, synchronised to 0.1 microsecond, is available worldwide by radio transmissions; this includes satellite broadcasts. The ampere has traditionally been the electrical base unit and has been realised at NPL using the Ayrton-Jones current balance; here the force between two currentcarrying coils is balanced by a known weight. The accuracy of this method is limited by the large deadweight of the coils and formers and the many length measurements necessary. For this reason the two electrical base units are now chosen to be the farad and the volt (or watt); the other units such as the ampere, ohm, henry and joule are derived from these two base units with time or frequency units, using
2.4 I D E N T I F I CATI O N O F S T AT I C CH AR ACTE RI S T I CS – CALI BRATI ON
27
Ohm’s law where necessary. The farad is realised using a calculable capacitor based on the Thompson–Lampard theorem. Using a.c. bridges, capacitance and frequency standards can then be used to calibrate standard resistors. The primary standard for the volt is based on the Josephson effect in superconductivity; this is used to calibrate secondary voltage standards, usually saturated Weston cadmium cells. The ampere can then be realised using a modified current balance. As before, the force due to a current I is balanced by a known weight mg, but also a separate measurement is made of the voltage e induced in the coil when moving with velocity u. Equating electrical and mechanical powers gives the simple equation: eI = mgu Accurate measurements of m, u and e can be made using secondary standards traceable back to the primary standards of the kilogram, metre, second and volt. Ideally temperature should be defined using the thermodynamic scale, i.e. the relationship PV = Rθ between the pressure P and temperature θ of a fixed volume V of an ideal gas. Because of the limited reproducibility of real gas thermometers the International Practical Temperature Scale (IPTS) was devised. This consists of: (a) (b)
several highly reproducible fixed points corresponding to the freezing, boiling or triple points of pure substances under specified conditions; standard instruments with a known output versus temperature relationship obtained by calibration at fixed points.
The instruments interpolate between the fixed points. The numbers assigned to the fixed points are such that there is exactly 100 K between the freezing point (273.15 K) and boiling point (373.15 K) of water. This means that a change of 1 K is equal to a change of 1 °C on the older Celsius scale. The exact relationship between the two scales is:
θ K = T °C + 273.15 Table 2.5 shows the primary fixed points, and the temperatures assigned to them, for the 1990 version of the International Temperature Scale – ITS90.[7] In addition there are primary standard instruments which interpolate between these fixed points. Platinum resistance detectors (PRTD – Section 8.1) are used up to the freezing point of silver (961.78 °C) and radiation pyrometers (Section 15.6) are used at higher temperatures. The resistance RT Ω of a PRTD at temperature T °C (above 0 °C) can be specified by the quadratic equation RT = R0(1 + αT + βT 2) where R0 is the resistance at 0 °C and α, β are constants. By measuring the resistance at three adjacent fixed points (e.g. water, gallium and indium), R0, α and β can be calculated. An interpolating equation can then be found for the standard, which relates RT to T and is valid between the water and indium fixed points. This primary standard PRTD, together with its equation, can then be used to calibrate secondary and transfer standards, usually PRTDs or thermocouples depending on temperature range. The standards available for the base quantities, i.e. length, mass, time, current and temperature, enable standards for derived quantities to be realised. This is illustrated in the methods for calibrating liquid flowmeters.[8] The actual flow rate through the meter is found by weighing the amount of water collected in a given time, so that the accuracy of the flow rate standard depends on the accuracies of weight and time standards.
28 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS Table 2.5 The primary fixed points defining the International Temperature Scale 1990 – ITS90.
Equilibrium state
θK
T °C
Triple point of hydrogen Boiling point of hydrogen at a pressure of 33 321.3 Pa Boiling point of hydrogen at a pressure of 101 292 Pa Triple point of neon Triple point of oxygen Triple point of argon Triple point of mercury Triple point of water Melting point of gallium Freezing point of indium Freezing point of tin Freezing point of zinc Freezing point of aluminium Freezing point of silver Freezing point of gold Freezing point of copper
13.8033 17.035 20.27 24.5561 54.3584 83.8058 234.3156 273.16 302.9146 429.7485 505.078 692.677 933.473 1234.93 1337.33 1357.77
−259.3467 −256.115 −252.88 −248.5939 −218.7916 −189.3442 −38.8344 0.01 29.7646 156.5985 231.928 419.527 660.323 961.78 1064.18 1084.62
National primary and secondary standards for pressure above 110 kPa use pressure balances.[3] Here the force pA due to pressure p acting over an area A is balanced by the gravitational force mg acting on a mass m, i.e. pA = mg or p = mg/A Thus standards for pressure can be derived from those for mass and length, though the local value of gravitational acceleration g must also be accurately known.
2.4.3
Experimental measurements and evaluation of results The calibration experiment is divided into three main parts.
O versus I with IM = II = 0 Ideally this test should be held under ‘standard’ environmental conditions so that IM = II = 0; if this is not possible all environmental inputs should be measured. I should be increased slowly from IMIN to IMAX and corresponding values of I and O recorded at intervals of 10% span (i.e. 11 readings), allowing sufficient time for the output to settle out before taking each reading. A further 11 pairs of readings should be taken with I decreasing slowly from IMAX to IMIN. The whole process should be repeated for two further ‘ups’ and ‘downs’ to yield two sets of data: an ‘up’ set (Ii, Oi)I ↑ and a ‘down’ set (Ij, Oj)I↓ i, j = 1, 2, . . . , n (n = 33). Computer software regression packages are readily available which fit a polyq nomial, i.e. O(I ) = ∑ q=m q=0 aq I , to a set of n data points. These packages use a ‘least squares’ criterion. If di is the deviation of the polynomial value O(Ii) from the data value Oi, then di = O(Ii ) – Oi. The program finds a set of coefficients a0, a1, a2, etc.,
2.4 I D E N T I F I CATI O N O F S T AT I C CH AR ACTE RI S T I CS – CALI BRATI ON
29
Figure 2.18 (a) Significant hysteresis (b) Insignificant hysteresis.
i=n 2 such that the sum of the squares of the deviations, i.e. ∑ i=1 d i , is a minimum. This [9] involves solving a set of linear equations. In order to detect any hysteresis, separate regressions should be performed on the two sets of data (Ii, Oi)I ↑, (Ij, Oj)I↓, to yield two polynomials: q m
q m
O(I)I↑ =
∑ a↑q I q
and
O(I)I↓ =
q 0
∑a I
↓ q q
q 0
If the hysteresis is significant, then the separation of the two curves will be greater than the scatter of data points about each individual curve (Figure 2.18(a)). Hysteresis H(I ) is then given by eqn [2.10], i.e. H(I ) = O(I )I↓ − O(I )I ↑. If, however, the scatter of points about each curve is greater than the separation of the curves (Figure 2.18(b)), then H is not significant and the two sets of data can then be combined to give a single polynomial O(I). The slope K and zero bias a of the ideal straight line joining the minimum and maximum points (IMIN, OMIN) and (IMAX, OMAX ) can be found from eqn [2.3]. The non-linear function N(I ) can then be found using eqn [2.4]: N(I ) = O(I ) − (KI + a)
[2.24]
Temperature sensors are often calibrated using appropriate fixed points rather than a standard instrument. For example, a thermocouple may be calibrated between 0 and 500 °C by measuring e.m.f.’s at ice, steam and zinc points. If the e.m.f.–temperature relationship is represented by the cubic E = a1T + a2T 2 + a3T 3, then the coefficients a1, a2, a3 can be found by solving three simultaneous equations (see Problem 2.1).
O versus IM , II at constant I We first need to find which environmental inputs are interfering, i.e. which affect the zero bias a. The input I is held constant at I = IMIN , and one environmental input is changed by a known amount, the rest being kept at standard values. If there is a resulting change ∆O in O, then the input II is interfering and the value of the corresponding coefficient KI is given by KI = ∆O/∆II . If there is no change in O, then the input is not interfering; the process is repeated until all interfering inputs are identified and the corresponding KI values found. We now need to identify modifying inputs, i.e. those which affect the sensitivity of the element. The input I is held constant at the mid-range value –12 (IMIN + IMAX ) and each environmental input is varied in turn by a known amount. If a change in input
30 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS produces a change ∆O in O and is not an interfering input, then it must be a modifying input IM and the value of the corresponding coefficient KM is given by KM =
1 ∆O 2 ∆O = I ∆IM IMIN + IMAX ∆IM
[2.25]
Suppose a change in input produces a change ∆O in O and it has already been identified as an interfering input with a known value of KI . Then we must calculate a non-zero value of KM before we can be sure that the input is also modifying. Since ∆O = KI ∆II,M + KM ∆II,M
IMIN + IMAX 2
then KM =
2 J G ∆O − KI L IMIN + IMAX I ∆II,M
[2.26]
Repeatability test This test should be carried out in the normal working environment of the element, e.g. out on the plant, or in a control room, where the environmental inputs IM and II are subject to the random variations usually experienced. The signal input I should be held constant at mid-range value and the output O measured over an extended period, ideally many days, yielding a set of values Ok , k = 1, 2, . . . , N. The mean value of the set can be found using /=
1 N
k N
∑O
[2.27]
k
k 1
and the standard deviation (root mean square of deviations from the mean) is found using
σ0 =
1 N
N
∑ (O
k
− / )2
[2.28]
k =1
A histogram of the values Ok should then be plotted in order to estimate the probability density function p(O) and to compare it with the normal form (eqn [2.20]). A repeatability test on a pressure transducer can be used as an example of the construction of a histogram. Suppose that N = 50 and readings between 0.975 and 1.030 V are obtained corresponding to an expected value of 1.000 V. The readings are grouped into equal intervals of width 0.005 V and the number in each interval is found, e.g. 12, 10, 8, etc. This number is divided by the total number of readings, i.e. 50, to give the probabilities 0.24, 0.20, 0.16, etc. of a reading occurring in a given interval. The probabilities are in turn divided by the interval width 0.005 V to give the probability densities 48, 40 and 32 V−1 plotted in the histogram (Figure 2.19). We note that the area of each rectangle represents the probability that a reading lies within the interval, and that the total area of the histogram is equal to unity. The mean and standard deviation are found from eqns [2.27] and [2.28] to be / = 0.999 V and σ0 = 0.010 V respectively. Figure 2.19 also shows a normal probability density function with these values.
REFERENCES
31
Figure 2.19 Comparison of histogram with normal probability density function.
Conclusion The chapter began by discussing the static or steady-state characteristics of measurement system elements. Systematic characteristics such as non-linearity and environmental effects were first explained. This led to the generalised model of an element. Statistical characteristics, i.e. repeatability and tolerance, were then discussed. The last section explained how these characteristics can be measured experimentally, i.e. calibration and the use and types of standards.
References [1] [2] [3] [4] [5] [6] [7] [8]
[9]
IEC 584.1:1995 International Thermocouple Reference Tables, International Electrotechnical Committee. KENNEDY J B and NEVILLE A 1986 Basic Statistical Methods for Engineers and Scientists, 3rd edn, pp. 345–60. Harper and Row, New York. STUART P R 1987 ‘Standards for the measurement of pressure’, Measurement and Control, vol. 20, no. 8, pp. 7–11. National Physical Laboratory 1996 Units of Measurement, poster, 8th edn. SCARR A 1979 ‘Measurement of length’, Journal of the Institute of Measurement and Control, vol. 12, July, pp. 265–9. HAYWARD A T J 1977 Repeatability and Accuracy, p. 34. Mechanical Engineering Publications, London. TC Ltd 2001 Guide to Thermocouple and Resistance Thermometry, issue 6.0. HAYWARD A T J 1977 ‘Methods of calibrating flowmeters with liquids – a comparative survey’, Transactions of the Institute of Measurement and Control, vol. 10, pp. 106–16. FREUND J E 1984 Modern Elementary Statistics, 6th edn, pp. 401–34. Prentice Hall International, Englewood Cliffs, NJ.
32 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS
Problems 2.1
The e.m.f. at a thermocouple junction is 645 µV at the steam point, 3375 µV at the zinc point and 9149 µV at the silver point. Given that the e.m.f.–temperature relationship is of the form E(T ) = a1T + a2T 2 + a3T 3 (T in °C), find a1, a2 and a3.
2.2
The resistance R(θ ) of a thermistor at temperature θ K is given by R(θ ) = α exp( β/θ ). Given that the resistance at the ice point (θ = 273.15 K) is 9.00 kΩ and the resistance at the steam point is 0.50 kΩ, find the resistance at 25 °C.
2.3
A displacement sensor has an input range of 0.0 to 3.0 cm and a standard supply voltage VS = 0.5 volts. Using the calibration results given in the table, estimate: (a) (b) (c)
The maximum non-linearity as a percentage of f.s.d. The constants KI, KM associated with supply voltage variations. The slope K of the ideal straight line.
Displacement x cm Output voltage millivolts (VS = 0.5) Output voltage millivolts (VS = 0.6) 2.4
0.0 0.0 0.0
0.5 16.5 21.0
1.0 32.0 41.5
1.5 44.0 56.0
2.0 51.5 65.0
2.5 55.5 70.5
3.0 58.0 74.0
A liquid level sensor has an input range of 0 to 15 cm. Use the calibration results given in the table to estimate the maximum hysteresis as a percentage of f.s.d. Level h cm 0.0 1.5 3.0 4.5 6.0 7.5 9.0 10.5 12.0 13.5 15.0 Output volts h increasing 0.00 0.35 1.42 2.40 3.43 4.35 5.61 6.50 7.77 8.85 10.2 Output volts h decreasing 0.14 1.25 2.32 3.55 4.43 5.70 6.78 7.80 8.87 9.65 10.2
2.5
A repeatability test on a vortex flowmeter yielded the following 35 values of frequency corresponding to a constant flow rate of 1.4 × 10−2 m3 s−1: 208.6; 208.3; 208.7; 208.5; 208.8; 207.6; 208.9; 209.1; 208.2; 208.4; 208.1; 209.2; 209.6; 208.6; 208.5; 207.4; 210.2; 209.2; 208.7; 208.4; 207.7; 208.9; 208.7; 208.0; 209.0; 208.1; 209.3; 208.2; 208.6; 209.4; 207.6; 208.1; 208.8; 209.2; 209.7 Hz. (a) (b) (c)
2.6
Using equal intervals of width 0.5 Hz, plot a histogram of probability density values. Calculate the mean and standard deviation of the data. Sketch a normal probability density function with the mean and standard deviation calculated in (b) on the histogram drawn in (a).
A platinum resistance sensor is used to interpolate between the triple point of water (0 °C), the boiling point of water (100 °C) and the freezing point of zinc (419.6 °C). The corresponding resistance values are 100.0 Ω, 138.5 Ω and 253.7 Ω. The algebraic form of the interpolation equation is: RT = R0(1 + α T + βT 2 ) where
RT Ω = resistance at T °C R0 Ω = resistance at 0 °C α, β = constants.
Find the numerical form of the interpolation equation.
PROBLEMS
2.7
33
The following results were obtained when a pressure transducer was tested in a laboratory under the following conditions: I II III
Ambient temperature 20 °C, supply voltage 10 V (standard) Ambient temperature 20 °C, supply voltage 12 V Ambient temperature 25 °C, supply voltage 10 V
Input (barg) Output (mA) I II III (a) (b)
0
2
4
6
8
10
4 4 6
7.2 8.4 9.2
10.4 12.8 12.4
13.6 17.2 15.6
16.8 21.6 18.8
20 28 22
Determine the values of KM, KI, a and K associated with the generalised model equation O = (K + KM IM)I + a + KI II . Predict an output value when the input is 5 barg, VS = 12 V and ambient temperature is 25 °C.
Basic problems 2.8
A force sensor has an output range of 1 to 5 V corresponding to an input range of 0 to 2 × 105 N. Find the equation of the ideal straight line.
2.9
A differential pressure transmitter has an input range of 0 to 2 × 104 Pa and an output range of 4 to 20 mA. Find the equation to the ideal straight line.
2.10
A non-linear pressure sensor has an input range of 0 to 10 bar and an output range of 0 to 5 V. The output voltage at 4 bar is 2.20 V. Calculate the non-linearity in volts and as a percentage of span.
2.11
A non-linear temperature sensor has an input range of 0 to 400 °C and an output range of 0 to 20 mV. The output signal at 100 °C is 4.5 mV. Find the non-linearity at 100 °C in millivolts and as a percentage of span.
2.12
A thermocouple used between 0 and 500 °C has the following input–output characteristics: Input T °C Output E µV (a) (b)
0 0
100 5268
200 10 777
300 16 325
500 27 388
Find the equation of the ideal straight line. Find the non-linearity at 100 °C and 300 °C in µV and as a percentage of f.s.d.
2.13
A force sensor has an input range of 0 to 10 kN and an output range of 0 to 5 V at a standard temperature of 20 °C. At 30 °C the output range is 0 to 5.5 V. Quantify this environmental effect.
2.14
A pressure transducer has an output range of 1.0 to 5.0 V at a standard temperature of 20 °C, and an output range of 1.2 to 5.2 V at 30 °C. Quantify this environmental effect.
2.15
A pressure transducer has an input range of 0 to 104 Pa and an output range of 4 to 20 mA at a standard ambient temperature of 20 °C. If the ambient temperature is increased to 30 °C, the range changes to 4.2 to 20.8 mA. Find the values of the environmental sensitivities KI and KM .
34 STATIC CHARACTERISTICS OF MEASUREMENT SYSTEM ELEMENTS
2.16
An analogue-to-digital converter has an input range of 0 to 5 V. Calculate the resolution error both as a voltage and as a percentage of f.s.d. if the output digital signal is: (a) (b)
2.17
8-bit binary 16-bit binary.
A level transducer has an output range of 0 to 10 V. For a 3 metre level, the output voltage for a falling level is 3.05 V and for a rising level 2.95 V. Find the hysteresis as a percentage of span.
3 The Accuracy of Measurement Systems in the Steady State In Chapter 1 we saw that the input to a measurement system is the true value of the variable being measured. Also, if the measurement system is complete, the system output is the measured value of the variable. In Chapter 2 we defined accuracy in terms of measurement error, i.e. the difference between the measured and true values of a variable. It follows therefore that accuracy is a property of a complete measurement system rather than a single element. Accuracy is quantified using measurement error E where: E = measured value – true value = system output – system input
[3.1]
In this chapter we use the static model of a single element, developed previously, to calculate the output and thus the measurement error for a complete system of several elements. The chapter concludes by examining methods of reducing system error.
3.1
Measurement error of a system of ideal elements Consider the system shown in Figure 3.1 consisting of n elements in series. Suppose each element is ideal, i.e. perfectly linear and not subject to environmental inputs. If we also assume the intercept or bias is zero, i.e. a = 0, then:
Input/output equation for ideal element with zero intercept
Oi = Ki Ii
[3.2]
for i = 1, . . . , n, where Ki is the linear sensitivity or slope (eqn [2.3]). It follows that O2 = K2 I2 = K2 K1 I, O3 = K3 I3 = K3K2K1 I, and for the whole system: O = On = K1 K2 K3 . . . Ki . . . Kn I Figure 3.1.
[3.3]
36 THE ACCURACY OF MEASUREMENT SYS T E MS I N TH E S T E AD Y S TATE Figure 3.2 Simple temperature measurement system.
If the measurement system is complete, then E = O − I, giving: E = (K1K2K3 . . . Kn − 1)I
[3.4]
Thus if K1K2K3 . . . Kn = 1
[3.5]
we have E = 0 and the system is perfectly accurate. The temperature measurement system shown in Figure 3.2 appears to satisfy the above condition. The indicator is simply a moving coil voltmeter (see Chapter 11) with a scale marked in degrees Celsius so that an input change of 1 V causes a change in deflection of 25 °C. This system has K1K2K3 = 40 × 10−6 × 103 × 25 = 1 and thus appears to be perfectly accurate. The system is not accurate, however, because none of the three elements present is ideal. The thermocouple is non-linear (Chapter 2), so that as the input temperature changes the sensitivity is no longer 40 µV °C−1. Also changes in reference junction temperature (Figure 2.12(b)) cause the thermocouple e.m.f. to change. The output voltage of the amplifier is also affected by changes in ambient temperature (Chapter 9). The sensitivity K3 of the indicator depends on the stiffness of the restoring spring in the moving coil assembly. This is affected by changes in environmental temperature and wear, causing K3 to deviate from 25 °C V −1. Thus the condition K1K2K3 = 1 cannot be always satisfied and the system is in error. In general the error of any measurement system depends on the non-ideal characteristics – e.g. non-linearity, environmental and statistical effects – of every element in the system. Thus in order to quantify this error as precisely as possible we need to use the general model for a single element developed in Sections 2.2 and 2.3.
3.2
The error probability density function of a system of non-ideal elements In Chapter 2 we saw that the probability density function of the output p(O) of a single element can be represented by a normal distribution (eqn [2.20]). The mean value / of the distribution is given by eqn [2.22], which allows for non-linear and environmental effects. The standard deviation σ0 is given by [2.23], which allows for statistical variations in inputs I, IM and II with time, and statistical variations in parameters K, a, etc., amongst a batch of similar elements. These equations apply to each element in a measurement system of n elements and can be used to calculate the system error probability density function p(E) as shown in Table 3.1. Equations [3.6] (based on [2.22]) show how to calculate the mean value of the output of each element in turn, starting with /1 for the first and finishing with /n = / for the nth. The mean value + of the system error is simply the difference between the mean value of system output and mean value of system input (eqn [3.7]). Since the probability densities of the outputs of the individual elements are normal, then, using the result outlined in Section 2.3, the probability density function of the
3.2 THE ERROR PROBABILIT Y D E N S I TY F UN CT I O N O F A S Y S T E M O F N O N - I D E AL E LE ME N T S
37
Table 3.1 General calculation of system p(E). Mean values of element outputs -l = -2 = /1 = .1 -1 + å1(-1) + ã1 + .M1 -M1 -1 + .I1 -I1 -3 = /2 = .2 -2 + å2(-2 ) + ã2 + .M2 -M2 -2 + .I2 -I2 7
[3.6]
7
-i+1 = /i = .i -i + åi (-i ) + ãi + .Mi -Mi -i + .Ii -Ii 7
7
/ = /n = .n -n + ån(-n ) + ãn + .Mn -Mn -n + .In -In Mean value of system error +=/−-
[3.7]
Standard deviations of element outputs
σ I21 = 0 σ I22 = σO2 1 =
A ∂O1D 2 2 A ∂O1 D 2 2 A ∂O1D 2 2 A ∂O1D 2 2 σ I1 + σ IM + σI + σ K1 + . . . C ∂I1 F C ∂IM1F C ∂II1 F I C ∂K1 F
σ I23 = σO2 2 =
A ∂O2D 2 2 A ∂O2 D 2 2 A ∂O2D 2 2 A ∂O2D 2 2 σ I2 + σ IM + σI + σ K2 + . . . C ∂I2 F C ∂IM2 F C ∂II2 F I C ∂K2 F
7
1
1
2
2
[3.8]
7
σ I2i +1 = σO2 i = σ O2 = σ O2 n =
A ∂Oi D 2 2 A ∂Oi D 2 2 A ∂Oi D 2 2 A ∂Oi D 2 2 σI + σI + σI + σ Ki + . . . C ∂Ii F i C ∂IMi F M C ∂IIi F I C ∂Ki F i
i
A ∂OnD 2 2 A ∂On D 2 2 A ∂OnD 2 2 A ∂OnD 2 2 σ In + σ IM + σI + σ Kn + . . . C ∂In F C ∂IMn F C ∂IIn F I C ∂Kn F n
n
Standard deviation of system error
σE = σO
[3.9]
Error probability density function p (E ) =
1
σE
 1  exp − 2 ( E − + )2  2π  2σ E 
[3.10]
system output O and system error E is also normal (eqn [3.10]). Equations [3.8] (based on [2.23]) show how to calculate the standard deviation of the output of each element in turn, starting with σO1 for the first, and finishing with σOn for the nth. We note that the standard deviation of the system input is zero and that the standard deviation of the error is equal to that of the system output (eqn [3.9]). We now use the temperature measurement system shown in Figure 3.3 as an example of the calculation of + and σE. This consists of a platinum resistance temperature detector, current transmitter and recorder. Table 3.2 gives the models for each of the elements in the temperature measurement system. In (a), the platinum resistance temperature detector is characterised by
38 THE ACCURACY OF MEASUREMENT SYS T E MS I N TH E S T E AD Y S TATE Figure 3.3 Temperature measurement system.
Table 3.2 Models for temperature measurement system elements.
(a) Platinum resistance temperature detector Model equation
RT = R0 (1 + αT + βT 2 )
Individual mean values
=0 = 100.0 Ω, [ = 3.909 × 10−3, ] = −5.897 × 10−7 (between 100 and 130 °C)
Individual standard deviations
σR 0 = 4.33 × 10−2, σα = 0.0, σβ = 0.0
Partial derivatives
∂RT = 1.449 at T = 117 °C ∂R0
Overall mean value
=T = =0 (1 + [> + ]> 2 )
Overall standard deviation
σ R2T =
A ∂RT D 2 2 σR0 C ∂R0 F
(b) Current transmitter Model equation
4 to 20 mA output for 138.5 to 149.8 Ω input (100 to 130 °C) ∆Ta = deviation of ambient temperature from 20 °C i = KRT + KM RT ∆Ta + KI ∆Ta + a
Individual mean values
. = 1.4134, .M = 1.4134 × 10 −4, .I = −1.637 × 10−2 ã = −191.76, ∆>a = −10 σK = 0.0, σKM = 0.0, σKI = 0.0 σa = 0.24, σ∆Ta = 6.7
Individual standard deviations
Partial derivatives
∂i ∂i ∂i = 1.413, = 4.11 × 10 −3, = 1.00 ∂RT ∂∆Ta ∂a
Overall mean value
j = .=T + .M =T ∆>a + .I ∆>a + ã
Overall standard deviation
σ 2i =
A ∂i D 2 2 A ∂i D 2 2 A ∂i D 2 2 σ RT + σ ∆Ta + σa C ∂RT F C ∂∆TaF C ∂aF
(c) Recorder Model equation
TM = Ki + a
Individual mean values
. = 1.875, ã = 92.50 (100 to 130 °C record for 4 to 20 mA input)
Individual standard deviations
σk = 0.0, σa = 0.10
Partial derivatives
∂TM ∂T = 1.875, M = 1.00 ∂i ∂a
Overall mean value
>M = .j + ã
Overall standard deviation
σ T2M =
A ∂TMD 2 2 A ∂TMD 2 2 σi + σa C ∂i F C ∂a F
3.2 THE ERROR PROBABILIT Y D E N S I TY F UN CT I O N O F A S Y S T E M O F N O N - I D E AL E LE ME N T S
Table 3.3 Summary of calculation of + and σE for temperature system.
39
Mean + > = 117 °C =T = 144.93 Ω j = 13.04 mA >M = 116.95 °C + = >M − > = −0.005 °C Standard deviation σE
σ T2 = 0 σ R2T = σ 2i =
A ∂RT D 2 2 σ R0 = 39.4 × 10− 4 C ∂R0 F A ∂i D 2 2 A ∂i D 2 2 A ∂i D 2 2 σ RT + σ ∆Ta + σa C ∂RT F C ∂∆TaF C ∂aF
= 78.7 × 10−4 + 8.18 × 10−4 + 5.76 × 10−2 = 6.62 × 10−2
σ T2M =
A ∂TM D 2 2 A ∂TM D 2 2 σi + σ a = 24.3 × 10−2 C ∂i F C ∂a F
σE = σTM = 0.49 °C
a small amount of non-linearity and a spread of values of R0 (resistance at 0 °C). The current transmitter (b) is linear but temperature acts as both a modifying and an interfering input. The zero bias and sensitivity are adjustable: we cannot be certain that the transmitter will be set up exactly as stated in the table, and this is reflected in the non-zero value of σa . In (c) the recorder is linear but again calibration uncertainties are modelled by a non-zero value of σa . Table 3.3 summarises the calculation of + and σE for the system when the mean value > of the true temperature is 117 °C. The corresponding mean values for the system are resistance =T = 144.93 Ω (Table 3.2(a)), current j = 13.04 mA (Table 3.2(b)) and measured temperature >M = 116.95 °C (Table 3.2(c)). The mean error is therefore + = −0.05 °C. The standard deviations σRT , σi and σTM are calculated using Table 3.2 to give σE = 0.49 °C.
Modelling using error bands In Section 2.1 we saw that in situations where element non-linearity, hysteresis and environmental effects are small, their overall effect is quantified using error bands. Here a systematic statement of the exact element input/output relationship (e.g. eqn [2.9]) is replaced by a statistical statement. The element output is described by a rectangular probability density function, of width 2h, centred on the ideal straight line value OIDEAL = KI + a. If every element in the system is described in this way, then the mean output value /i will have the ideal value /i = Ki -i + ai for each element in the system. In the special case ai = 0 for all i, /i = Ki -i, the mean value of system output is given by / = K1K2 . . . Ki . . . Kn and the mean value of error + = / − -. In the special case K1K2 . . . Ki . . . Kn = 1, + = 0 (Table 3.4). The error probability density function p(E) is the result of
40 THE ACCURACY OF MEASUREMENT SYS T E MS I N TH E S T E AD Y S TATE Table 3.4 + and σE for system of elements described by error bands. /i = Ki -i / = K1K2 . . . Ki . . . Kn + = / − - [= 0, if K1K2 . . . Kn = 1]
σ I21 = 0 σ I22 = σO2 1 =
h12 3
σ I23 = σO2 2 = K 22σ I22 + 7
7
σ I2i+1 = σO2 i = K 2i σ I2i + 7
h 22 3 h2i 3
7
σ O2 = σO2 n = K 2n σ I2n +
h2n 3
σE = σ0
combining n rectangular distributions, each of width 2hi, i = 1, 2, . . . , n. If n > 3, then the resultant distribution p(E) approximates to a normal distribution; the larger the value of n the closer the distribution is to normal. In order to calculate the standard deviation σE of p(E) we return to eqns [2.16] and [2.17]. If a dependent variable y is a linear combination of independent random variables x1, x2, x3, i.e. y = a1x1 + a2 x2 + a3 x3
[2.16]
then the standard deviation σ of the probability distribution of y is given by
σ 2 = a 21σ 21 + a22σ 22 + a23σ 23
[2.17]
Equation [2.17] applies whatever the probability distributions of the individual x1, x2, x3 are, provided that the individual standard deviations σ1, σ2, σ3 are small.[1] Consider the ith element in Table 3.4 with sensitivity Ki and error band width 2hi . The standard deviation σOi of the output Oi is equal to
σO2 i = σ 2 due to input + σ 2 due to element If σIi is the standard deviation of the input Ii , then this contributes Ki σIi to the output standard deviation. The standard deviation of the rectangular distribution of width 2hi , due to the element, is hi /ê3. We therefore have:
σO2 i = K 2i σ I2i + hi2/3 Table 3.4 gives the calculation procedure for a complete system of n elements described by error bands.
3.3 E RRO R RE D UCT I O N T E CH N I Q UE S
3.3
41
Error reduction techniques In the two previous sections we saw that the error of a measurement system depends on the non-ideal characteristics of every element in the system. Using the calibration techniques of Section 2.4, we can identify which elements in the system have the most dominant non-ideal behaviour. We can then devise compensation strategies for these elements which should produce significant reductions in the overall system error. This section outlines compensation methods for non-linear and environmental effects. One of the most common methods of correcting a non-linear element is to introduce a compensating non-linear element into the system. This method is illustrated in Figure 3.4. Given a non-linear element, described by U(I), we need a compensating element C(U), such that the overall characteristics C[U(I)] of the elements together are as close to the ideal straight line as possible. The method is illustrated in Figure 3.4 by the use of a deflection bridge to compensate for the non-linear characteristics of a thermistor. A detailed procedure for the design of the bridge is given in Section 9.1. The most obvious method of reducing the effects of environmental inputs is that of isolation, i.e. to isolate the transducer from environmental changes so that effectively IM = II = 0. Examples of this are the placing of the reference junction of a thermocouple in a temperature-controlled enclosure, and the use of spring mountings to isolate a transducer from the vibrations of the structure to which it is attached. Another obvious method is that of zero environmental sensitivity, where the element is completely insensitive to environmental inputs, i.e. KM = KI = 0. An example of this is the use of a metal alloy with zero temperature coefficients of expansion and resistance as a strain gauge element. Such an ideal material is difficult to find, and in practice the resistance of a metal strain gauge is affected slightly by changes in ambient temperature. A more successful method of coping with environmental inputs is that of opposing environmental inputs. Suppose that an element is affected by an environmental
Figure 3.4 Compensating non-linear element.
42 THE ACCURACY OF MEASUREMENT SYS T E MS I N TH E S T E AD Y S TATE Figure 3.5 Compensation for interfering inputs: (a) Using opposing environmental inputs (b) Using a differential system.
input; then a second element, subject to the same environmental input, is deliberately introduced into the system so that the two effects tend to cancel out. This method is illustrated for interfering inputs in Figure 3.5 and can be easily extended to modifying inputs. An example is compensation for variations in the temperature T2 of the reference junction of a thermocouple. For a copper–constantan thermocouple (Figure 2.12(b)), we have KI II equal to −38.74T2 µV so that a compensating element giving an output equal to +38.74T2 µV is required. The design of the reference junction compensation element is discussed in Sections 8.5 and 9.1 and Problem 9.2. An example of a differential system (Figure 3.5(b)) is the use of two matched strain gauges in adjacent arms of a bridge to provide compensation for ambient temperature changes. One gauge is measuring a tensile strain +e and the other an equal compressive strain −e. The bridge effectively subtracts the two resistances so that the strain effect is doubled and the environmental effects cancel out. The use of high-gain negative feedback is an important method of compensating for modifying inputs and non-linearity. Figure 3.6 illustrates the technique for a force transducer. The voltage output of a force-sensing element, subject to a modifying input, is amplified by a high-gain amplifier. The amplifier output is fed back to an element (e.g. a coil and permanent magnet) which provides a balancing force to oppose the input force.
3.3 E RRO R RE D UCT I O N T E CH N I Q UE S
43
Figure 3.6 Closed loop force transducer.
Ignoring the effects of the modifying input for the moment we have: ∆F = FIN − FB VOUT = KKA ∆F FB = KF VOUT
[3.11] [3.12] [3.13]
i.e. VOUT = FIN − KF VOUT KKA giving Equation for force transducer with negative feedback
VOUT =
KK A FIN 1 + KK A K F
[3.14]
If the amplifier gain KA is made large such that the condition KKA KF 0002 1
[3.15]
1 F . This means that the system output depends only on KF IN the gain KF of the feedback element and is independent of the gains K and KA in the forward path. This means that, provided the above condition is obeyed, changes in K and KA due to modifying inputs and/or non-linear effects have negligible effect on VOUT. This can be confirmed by repeating the above analysis with K replaced by K + KM IM , giving is satisfied, then VOUT ≈
VOUT =
( K + KM IM ) KA FIN 1 + ( K + KM IM ) KA KF
[3.16]
FIN if (K + KM IM)KA KF 0002 1. We now, of course, have KF to ensure that the gain KF of the feedback element does not change due to nonlinear or environmental effects. Since the amplifier provides most of the required power, the feedback element can be designed for low power-handling capacity, which again reduces to VOUT ≈
44 THE ACCURACY OF MEASUREMENT SYS T E MS I N TH E S T E AD Y S TATE Figure 3.7 Models for system element: (a) Direct (b) Inverse.
giving greater linearity and less susceptibility to environmental inputs. A commonly used device which employs this principle is discussed in Section 9.4. The rapid fall in the cost of digital integrated circuits in recent years has meant that microcontrollers are now widely used as signal-processing elements in measurement systems (Chapter 10). This means that the powerful techniques of computer estimation of measured value can now be used. For this method, a good model of the elements in the system is required. In Sections 2.1 and 2.2 we saw that the steadystate output O of an element is in general given by an equation of the form: Direct equation
O = KI + a + N(I) + KM IM I + KI II
[2.9]
This is the direct equation (Figure 3.7(a)); here O is the dependent variable, which is expressed in terms of the independent variables I, IM and II . In Section 2.4.2 we saw how the direct equation could be derived from sets of data obtained in a calibration experiment. In Section 3.2 eqn [2.9] was used to derive the error probability density function for a complete measurement system. The steady-state characteristics of an element can also be represented by an alternative equation. This is the inverse equation (Figure 3.7(b)); here the signal input I is the dependent variable and the output O and environmental inputs II and IM are the independent variables. The general form of this equation is: Inverse equation
I = K′O + N′(O) + a′ + K M′ IM O + KI′I
[3.17]
3.3 E RRO R RE D UCT I O N T E CH N I Q UE S
45
where the values of K′, N′( ), a′, etc., are quite different from those for the direct equation. For example, the direct and inverse equations for a copper–constantan (type T) thermocouple with reference junction at 0 °C are: DIRECT E = 3.845 × 10−2T + 4.682 × 10−5T 2 − 3.789 × 10 −8T 3 + 1.652 × 10−11T 4 mV INVERSE T = 25.55E − 0.5973E 2 + 2.064 × 10−2E 3 − 3.205 × 10− 4E 4 °C
[3.18]
where E is the thermocouple e.m.f. and T the measured junction temperature between 0 and 400 °C. Both equations were derived using a least squares polynomial fit to IEC 584.1 data; for the direct equation E is the dependent variable and T the independent variable; for the inverse equation T is the dependent variable and E the independent variable. While the direct equation is more useful for error estimation, the inverse equation is more useful for error reduction. The use of the inverse equation in computer estimation of measured value is best implemented in a number of stages; with reference to Figure 3.8(a), these are: 1.
Treat the uncompensated system as a single element. Using the calibration procedure of Section 2.4.2 (or any other method of generating data) the parameters K′, a′, etc. in the inverse model equation: I = K′U + N′(U) + a′ + K M′ IM U + KI′II
2.
3.
representing the overall behaviour of the uncompensated system can be found. This procedure will enable major environmental inputs IM and II to be identified (there may be more than one of each type). The uncompensated system should be connected to the estimator. This consists firstly of a computer which stores the model parameters K′, a′, N′( ), etc. If errors due to environmental inputs are considered significant, then environmental sensors to provide the computer with estimates IM′ , II′, of these inputs are also necessary. The output U of the uncompensated system is also fed to the computer. The computer then calculates an initial estimate I′ of I using the inverse equation: I ′ = K′U + N′(U ) + a′ + KM′ IM′ U + KI′I I′
4.
5.
6.
[3.19]
[3.20]
The data presentation element then displays the measured value O, which should be close to I ′. In applications not requiring the highest accuracy the procedure can be terminated at this stage. If high accuracy is required, then it may be possible to further improve the estimator by calibrating the complete system. Values of system output O are measured for a range of known standard inputs I and the corresponding values of system error E = O − I are calculated. These error values will be mainly due to random effects but may also contain a small systematic component which can be corrected for. An attempt should now be made to fit the data set (Oi , Ei ), i = 1, 2, . . . n, by a least squares straight line of the form: E = kO + b
[3.21]
where b is any residual zero error and k specifies any residual scale error. There is little point in attempting a polynomial fit at this stage.
46 THE ACCURACY OF MEASUREMENT SYS T E MS I N TH E S T E AD Y S TATE Figure 3.8 Computer estimation of measured value using inverse model equation: (a) Principles (b) Example of displacement measurement system.
7.
The correlation coefficient: i= n
r=
8.
∑ i=1 Oi Ei i= n
i= n
∑ i=1 O i2 × ∑ i=1 E i2
[3.22]
between E and O data should now be evaluated. If the magnitude of r is greater than 0.5, then there is reasonable correlation between the E and O data; this means the systematic error of eqn [3.7] is present and we can proceed to stage 8 to correct for it. If the magnitude of r is less than 0.5, then there is no correlation between the E and O data; this means that the errors E are purely random and no correction can be made. If appropriate, eqn [3.21] can be used to calculate an improved measured value: O′ = O − E = O − (kO + b)
[3.23]
The displacement measurement system of Figure 3.8(b) shows this method. The uncompensated system consists of an inductive displacement sensor, an oscillator (Section 9.5) and a Schmitt trigger (Section 10.1.4). The sensor has a non-linear
PROBLEMS
47
relation between inductance L and displacement x, the oscillator a non-linear relation between frequency f and inductance L. This means that the inverse model equation, relating displacement x and frequency f of the Schmitt trigger output signal, has the non-linear form shown. The estimator consists of a 16-bit pulse counter and a computer. The computer reads the state of the counter at the beginning and end of a fixed time interval and thus measures the frequency f of the pulse signal. The computer then calculates x from the inverse model equation using model coefficients stored in memory.
Conclusion This chapter has shown how to find the error of a complete measurement system under steady-state conditions. Measurement error was first defined and then the error probability density function was derived, firstly for a general system of non-ideal elements and then for the typical example of a temperature measurement system. The last section discussed a range of methods for error reduction.
Reference [1]
kennedy j b and neville a 1986 Basic Statistical Methods for Engineers and Scientists, 3rd edn, pp. 345–60. Harper and Row, New York.
Problems 3.1
A measurement system consists of a chromel–alumel thermocouple (with cold junction compensation), a millivolt-to-current converter and a recorder. Table Prob. 1 gives the model equations, and parameters for each element. Assuming that all probability distributions are normal, calculate the mean and standard deviation of the error probability distribution, when the input temperature is 117 °C.
Table Prob. 1.
Chromel–alumel thermocouple
e.m.f-to-current converter
Recorder
Model equation
E = C 0 + C 1T + C 2T 2 i = K1E + KM E∆Ta + KI ∆Ta + a1 TM = K2i + a2
Mean values
™0 = 0.00 ™1 = 4.017 × 10−2 ™2 = 4.66 × 10 − 6
Standard deviations σC 0 = 6.93 × 10−2 σC1 = σC2 = 0
.1 = 3.893 ∆Ta = −10 ã1 = −3.864 .M = 1.95 × 10 − 4 .I = 2.00 ×10−3
.2 = 6.25 ã2 = 25.0
σa1 = 0.14, σ∆Ta = 10 σK1 = σKM = σKI = 0
σa 2 = 0.30 σK 2 = 0.0
48 THE ACCURACY OF MEASUREMENT SYS T E MS I N TH E S T E AD Y S TATE Table Prob. 2.
3.2
Element
Linear sensitivity K
Error bandwidth ± h
Pressure sensor Deflection bridge Amplifier Recorder
10 − 4 Ω Pa−1 4 × 10−2 mV Ω−1 103 mV mV−1 250 Pa mV −1
±0.005 Ω ±5 × 10 − 4 mV ±0.5 mV ±100 Pa
A pressure measurement system consists of a pressure sensor, deflection bridge, amplifier and recorder. Table Prob. 2 gives the linear sensitivities and error bandwidths for each element in the system. (a) (b)
3.3
Calculate the standard deviation σE of the error distribution function. Given that the recorder is incorrectly adjusted so that its sensitivity is 225 Pa mV−1, calculate the mean error + for an input pressure of 5 × 103 Pa.
Figure Prob. 3 shows a block diagram of a force transducer using negative feedback. The elastic sensor gives a displacement output for a force input; the displacement sensor gives a voltage output for a displacement input. VS is the supply voltage for the displacement sensor. (a)
Calculate the output voltage V0 when (i) (ii)
(b)
VS = 1.0 V, F = 50 N VS = 1.5 V, F = 50 N.
Comment on the practical significance of the variation of the supply voltage VS .
Figure Prob. 3.
3.4
Figure Prob. 4 is a block diagram of a voltmeter. The motor produces a torque T proportional to voltage V, and the output angular displacement θ of the spring is proportional to T. The stiffness Ks of the spring can vary by ±10% about the nominal value of 5 × 10−2 rad N−1m−1. Given that the following are available: (i) (ii) (iii) (a) (b)
Figure Prob. 4.
a d.c. voltage amplifier of gain 1000 a voltage subtraction unit a stable angular displacement transducer of sensitivity 100 V rad−1,
draw a block diagram of a modified system using these components, which reduces the effect of changes in Ks ; calculate the effect of a 10% increase in Ks on the sensitivity of the modified system.
PROBLEMS
3.5
A temperature measurement system consists of a thermocouple, an amplifier, an 8-bit analogueto-digital converter and a microcontroller with display facilities. Table Prob. 5 gives the model equations and parameters for each element in the system. The temperature of the thermocouple measurement junction is T1 °C and the temperature of the reference junction is T2 °C. The microcontroller corrects for T2 having a non-zero mean value. (a)
(b)
Estimate the mean and standard deviation of the error probability density function when the input temperature T1 is 100 °C. Treat rectangular distribution as normal with σ = h/ê3. Explain briefly what modifications should be made to the system to reduce the quantities calculated in (a).
Table Prob. 5.
Model equations
3.6
49
Thermocouple
Amplifier
Analogue-todigital converter
ET1,T2 = a0 + a1(T1 − T2 ) + a2 (T 21 − T 22 )
V = K1ET1,T2 + b1 n = K2V + b2
Microcontroller with display Tm = K3n + b3
Mean values ã0 = 0 .1 = 255 ã1 = 4.3796 × 10−2 ç1 = 0.0 ã2 = −1.7963 × 10−5 >2 = 20
.2 = 0.1 ç2 = 0.0 n rounded to nearest integer
Statistical Normal with distributions σa0 = 0.05 σT2 = 2.0 σa1 = σa2 = 0.0
b2 has a rectangular σK3 = 0.0 distribution of σb3 = 0.0 width ±0.5 σK 2 = 0.0
Normal with σb1 = 5.0 σK1 = 0.0
.3 = 1.0 ç3 = 20
A fluid velocity measurement system consists of a pitot tube, a differential pressure transmitter, an 8-bit analogue-to-digital converter and a microcontroller with display facilities. Table Prob. 6 gives the model equations and parameters for each element in the system. The microcontroller calculates the measured value of velocity assuming a constant density. (a)
(b)
Estimate the mean and standard deviation of the error probability density function assuming the true value of velocity vT is 14.0 m s−1. Use the procedure of Table 3.1, treating the rectangular distributions as normal with σ = h/ê3. Explain briefly what modifications could be made to the system to reduce the quantities calculated in (a).
Table Prob. 6.
Pitot tube
Differential pressure transmitter
Analogue-todigital converter
Microcontroller with display
Model equations
∆P = –12 ρ v T2 i = K1∆P + a1
n = K2i + a2
vM = K3 ên ú−ú ú5ú1ú
Mean values
ë = 1.2
.1 = 0.064 ã1 = 4.0
.2 = 12.80 ã2 = 0.0 n rounded off to nearest integer
.3 = 1.430
Half-widths of rectangular distribution
hρ = 0.1
ha1 = 0.04
ha2 = 0.5
hK3 = 0.0
50 THE ACCURACY OF MEASUREMENT SYS T E MS I N TH E S T E AD Y S TATE
3.7
A temperature measurement system consists of a thermistor, bridge and recorder. Table Prob. 7 gives the model equations, mean values and standard deviations for each element in the system. Assuming all probability distributions are normal, calculate the mean and standard deviation of the error probability density function for a true input temperature of 320 K.
Table Prob. 7.
Thermistor
Bridge A βD C θF
 1  − a1  3.3 V0 = VS   1 +  R θ  
Recorder
θM = K2V0 + a2
Model equations
Rθ = K1 exp
Mean values
.1 = 5 × 10 − 4 kΩ ] = 3 × 103 K
S = −3.00 V ã1 = 0.77
.2 = 50.0 K/V ã2 = 300 K
Standard deviations
σK1 = 0.5 × 10 − 4 σβ = 0
σVS = 0.03 σa1 = 0.01
σK2 = 0.0 σa2 = 3.0
Basic problems 3.8
A force measurement system consists of four elements with sensitivities 10−2, 5 × 10−2, 103 and 1.9. Find the system error for a true value input of 10 kN.
3.9
A level measurement system consists of three linear elements with sensitivities of 0.050, 21.5 and 0.99. Find the system error for a true value input of 5.0 metres.
4 Dynamic Characteristics of Measurement Systems If the input signal I to an element is changed suddenly, from one value to another, then the output signal O will not instantaneously change to its new value. For example, if the temperature input to a thermocouple is suddenly changed from 25 °C to 100 °C, some time will elapse before the e.m.f. output completes the change from 1 mV to 4 mV. The ways in which an element responds to sudden input changes are termed its dynamic characteristics, and these are most conveniently summarised using a transfer function G(s). The first section of this chapter examines the dynamics of typical elements and derives the corresponding transfer function. The next section examines how standard test signals can be used to identify G(s) for an element. If the input signal to a multi-element measurement system is changing rapidly, then the waveform of the system output signal is in general different from that of the input signal. Section 4.3 explains how this dynamic error can be found, and the final section outlines dynamic compensation methods that can be used to minimise errors.
4.1 4.1.1
Transfer function G(s) for typical system elements First-order elements A good example of a first-order element is provided by a temperature sensor with an electrical output signal, e.g. a thermocouple or thermistor. The bare element (not enclosed in a sheath) is placed inside a fluid (Figure 4.1). Initially at time t = 0− ( just before t = 0), the sensor temperature is equal to the fluid temperature, i.e. T(0−) = TF (0−). If the fluid temperature is suddenly raised at t = 0, the sensor is no longer in a steady state, and its dynamic behaviour is described by the heat balance equation: rate of heat inflow − rate of heat outflow =
Figure 4.1 Temperature sensor in fluid.
rate of change of sensor heat content
[4.1]
Assuming that TF > T, then the rate of heat outflow will be zero, and the rate of heat inflow W will be proportional to the temperature difference (TF − T). From Chapter 14 we have
52 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS W = UA(TF − T) watts −2
[4.2]
−1
where U W m °C is the overall heat transfer coefficient between fluid and sensor and A m2 is the effective heat transfer area. The increase of heat content of the sensor is MC [T − T(0−)] joules, where M kg is the sensor mass and C J kg−1 °C −1 is the specific heat of the sensor material. Thus, assuming M and C are constants: rate of increase of sensor heat content = MC
d [T − T(0−)] dt
[4.3]
Defining ∆T = T − T(0−) and ∆TF = TF − TF (0−) to be the deviations in temperatures from initial steady-state conditions, the differential equation describing the sensor temperature changes is UA(∆TF − ∆T ) = MC
d∆T dt
i.e. MC d∆T + ∆T = ∆TF UA dt
[4.4]
This is a linear differential equation in which d∆T/dt and ∆T are multiplied by constant coefficients; the equation is first order because d∆T/dt is the highest derivative present. The quantity MC/UA has the dimensions of time: kg × J × kg−1 × °C −1 J = = seconds −2 −1 2 W × m × °C × m W and is referred to as the time constant τ for the system. The differential equation is now: Linear first-order differential equation
τ
d ∆T + ∆T = ∆TF dt
[4.5]
While the above differential equation is a perfectly adequate description of the dynamics of the sensor, it is not the most useful representation. The transfer function based on the Laplace transform of the differential equation provides a convenient framework for studying the dynamics of multi-element systems. The Laplace transform g(s) of a time-varying function is defined by: Definition of Laplace transform
∞
g(s) =
0002e
−st
f(t) dt
[4.6]
0
where s is a complex variable of the form σ + jω where j = ê−ú1ú . Table 4.1 gives Laplace transforms for some common standard functions f(t). In order to find the transfer function for the sensor we must find the Laplace transform of eqn [4.5]. Using Table 4.1 we have:
τ [s∆>(s) − ∆T(0−)] + ∆>(s) = ∆>F (s)
[4.7]
4.1 TRAN S F E R F UN CTI O N G( S ) F O R T Y PI CAL S Y S T E M E LE ME N T S
Table 4.1 Laplace transforms of common time functions f(t).a
53
∞
0002[ f(t)] = g (s) =
0002e
−st
f(t)dt
0
Function
Symbol
1st derivative
d f (t) dt
sg(s) − f(0− )
2nd derivative
d2 f(t) dt 2
s2 g(s) − sf(0− ) − Ã(0−)
Unit impulse
δ (t)
1
Unit step
µ (t)
1 s
Exponential decay
exp(−α t)
1 s+α
Exponential growth
1 − exp(−α t)
α s(s + α )
Sine wave
sin ω t
Phase-shifted sine wave
sin(ω t + φ)
Exponentially damped sine wave
exp(−α t)sin ω t
Ramp with exponential decay
t exp(−α t)
a
Graph
Transform
ω s2 + ω 2 ω cos φ + s sin φ s2 + ω 2 ω (s + α )2 + ω 2
1 (s + α )2
Initial conditions are at t = 0−, just prior to t = 0.
where ∆T(0−) is the temperature deviation at initial conditions prior to t = 0. By definition, ∆T(0−) = 0, giving:
τ s∆>(s) + ∆>(s) = ∆>F (s) i.e. (τ s + 1)∆>(s) = ∆>F (s)
[4.8]
The transfer function G(s) of an element is defined as the ratio of the Laplace transform of the output to the Laplace transform of the input, provided the initial conditions are zero. Thus:
54 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS Definition of element transfer function
G( s ) =
g 0 (s) g i (s)
[4.9]
and g0(s) = G(s)gi (s); this means the transfer function of the output signal is simply the product of the element transfer function and the transfer function of the input signal. Because of this simple relationship the transfer function technique lends itself to the study of the dynamics of multi-element systems and block diagram representation (Figure 4.2). Figure 4.2 Transfer function representation.
From eqns [4.8] and [4.9] the transfer function for a first-order element is: Transfer function for a first-order element
G( s ) =
∆> (s) 1 = ∆>F (s) 1 + τs
[4.10]
The above transfer function only relates changes in sensor temperature to changes in fluid temperature. The overall relationship between changes in sensor output signal O and fluid temperature is: ∆/(s) ∆O ∆>(s) = ∆>F (s) ∆T ∆>F (s)
[4.11]
where ∆O/∆T is the steady-state sensitivity of the temperature sensor. For an ideal element ∆O/∆T will be equal to the slope K of the ideal straight line. For non-linear elements, subject to small temperature fluctuations, we can take ∆O/∆T = dO/dT, the derivative being evaluated at the steady-state temperature T(0−) around which the fluctuations are taking place. Thus for a copper–constantan thermocouple measuring small fluctuations in temperature around 100 °C, ∆E/∆T is found by evaluating dE/dT at 100 °C (see Section 2.1) to give ∆E/∆T = 35 µV °C −1. Thus if the time constant of the thermocouple is 10 s the overall dynamic relationship between changes in e.m.f. and fluid temperature is: ∆+(s) 1 = 35 × ∆>F (s) 1 + 10s
[4.12]
In the general case of an element with static characteristics given by eqn [2.9] and dynamic characteristics defined by G(s), the effect of small, rapid changes in ∆I is evaluated using Figure 4.3, in which steady-state sensitivity (∂O/∂I )I0 = K + KM IM + (dN/dI )I0, and I0 is the steady-state value of I around which the fluctuations are Figure 4.3 Element model for dynamic calculations.
4.1 TRAN S F E R F UN CTI O N G( S ) F O R T Y PI CAL S Y S T E M E LE ME N T S
55
Table 4.2 Analogous first-order elements. 1 (P − P) RF IN Pressures PIN = hIN ρg, P = hρg ρg dh =Q= (h − h) AF RF IN dt AF RF dh + h = hIN ρg dt i.e. A R dh τF + h = hIN, τF = F F ρg dt Volume flow rate Q =
Fluidic
VIN − V = iR
Electrical
Charge q = CV, current i = RC
dq CdV = dt dt
dV + V = VIN dt
i.e. dV τE + V = VIN, τE = RC dt
Mechanical
FIN − F = λ
dx , dt
Displacement x =
λ N s m−1 = damping constant F , k
k N m−1 = spring stiffness
λ dF + F = FIN k dt dF λ τM + F = FIN, τM = dt k
Thermal
τTh =
MC 1 = RThCTh; RTh = , C = MC UA UA Th
Fluidic
τF =
AF RF A = RF CF ; RF = RF , CF = F ρg ρg
Electrical
τE = RC = RE CE;
RE = R,
CE = C
λ = RM CM ; k
RM = λ,
CM =
Mechanical τM =
1 k
taking place. Table 4.2 shows analogous fluidic, electrical and mechanical elements, which are also described by a first-order transfer function G(s) = 1/(1 + τs). All four elements are characterised by ‘resistance’ and ‘capacitance’ as illustrated in the table. Temperature, pressure, voltage and force are analogous ‘driving’ or effort variables; heat flow rate, volume flow rate, current and velocity are analogous ‘driven’ or flow variables. These analogies are discussed further in Section 5.2.
56 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS Figure 4.4 Mass–spring–damper model of elastic force sensor.
4.1.2
Second-order elements The elastic sensor shown in Figure 4.4, which converts a force input F into a displacement output x, is a good example of a second-order element. The diagram is a conceptual model of the element, which incorporates a mass m kg, a spring of stiffness k N m−1, and a damper of constant λ N s m−1. The system is initially at rest at time t = 0− so that the initial velocity x(0−) = 0 and the initial acceleration y(0−) = 0. The initial input force F(0−) is balanced by the spring force at the initial displacement x(0−), i.e. F(0−) = kx(0−)
[4.13]
If the input force is suddenly increased at t = 0, then the element is no longer in a steady state and its dynamic behaviour is described by Newton’s second law, i.e. resultant force = mass × acceleration i.e. F − kx − λ x = my
[4.14]
and my + λ x + kx = F Defining ∆F and ∆x to be the deviations in F and x from initial steady-state conditions: ∆F = F − F(0−), ∆x = x,
∆x = x − x(0−) ∆y = y
[4.15]
the differential equation now becomes: m∆y + λ ∆x + kx(0−) + k∆x = F(0−) + ∆F which, using [4.13], reduces to: m∆y + λ ∆x + k∆x = ∆F i.e. m d2∆x λ d∆x 1 + + ∆x = ∆F k dt 2 k dt k
[4.16]
This is a second-order linear differential equation in which ∆x and its derivatives are multiplied by constant coefficients and the highest derivative present is d2∆x/dt 2. If we define undamped natural frequency ωn =
k rad/s m
4.1 TRAN S F E R F UN CTI O N G( S ) F O R T Y PI CAL S Y S T E M E LE ME N T S
57
and damping ratio ξ =
λ 2 km
[4.17]
then m/k = 1/ω n2, λ /k = 2ξ /ωn, and [4.16] can be expressed in the standard form: Linear second-order differential equation
1 d 2∆ x 2ξ d ∆ x 1 + + ∆ x = ∆F ω n2 dt 2 ω n dt k
[4.18]
In order to find the transfer function for the element we require the Laplace transform of eqn [4.18]. Using Table 4.1 we have: 1 2 2ξ [s ∆P(s) − s∆x(0−) − ∆ x(0−)] + [s∆P(s) − ∆ x(0−)] + ∆ P(s) ω n2 ωn 1 = ∆,(s) k
[4.19]
Since ∆x(0−) = x(0−) = 0 and ∆x(0−) = 0 by definition, [4.19] reduces to: 1 J G 1 2 2ξ s + s + 1 ∆P(s) = ∆,(s) L I ω n2 ωn k
[4.20]
Thus ∆P(s) 1 = G(s) ∆,(s) k where l/k = steady-state sensitivity K, and Transfer function for a second-order element
G (s) =
1 1 2 2ξ s + s +1 ω n2 ωn
[4.21]
Figure 4.5 shows an analogous electrical element, a series L–C–R circuit. Comparing eqns [4.14] and [4.22] we see that q is analogous to x, V is analogous to F, and L, R and 1/C are analogous to m, λ and k respectively (see Table 5.1). The electrical circuit is also described by the above second-order transfer function with ωn = 1/êLúCú and ξ = (R/2)êC/úLú . Figure 4.5 Series L–C–R circuit.
[4.22]
58 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS
4.2
Identification of the dynamics of an element In order to identify the transfer function G(s) of an element, standard input signals should be used. The two most commonly used standard signals are step and sine wave. This section examines the response of first- and second-order elements to step and sine wave inputs.
4.2.1
Step response of first- and second-order elements From Table 4.1 we see that the Laplace transform of a step of unit height u(t) is g(s) = 1/s. Thus if a first-order element with G(s) = 1/(1 + τ s) is subject to a unit step input signal, the Laplace transform of the element output signal is: go (s) = G(s) g i (s) =
1 (1 + τ s)s
[4.23]
Expressing [4.23] in partial fractions, we have: go (s) =
1 A B = + (1 + τ s)s (1 + τ s) s
Equating coefficients of constants gives B = 1, and equating coefficients of s gives 0 = A + Bτ, i.e. A = −τ. Thus: go (s) =
1 τ 1 1 − = − s (1 + τ s) s (s + 1/τ )
[4.24]
Using Table 4.1 in reverse, i.e. finding a time signal f(t) corresponding to a transform g(s), we have: fo (t) = u(t) − exp
A −t D C τF
and since u(t) = 1 for t > 0: Response of first-order element to unit step
 −t  fo (t ) = 1 − exp   τ
[4.25]
The form of the response is shown in Figure 4.6. As an example of the use of eqn [4.25], consider the temperature sensor of Section 4.1.1. Initially the temperature of the sensor is equal to that of the fluid, i.e. T(0−) = TF (0−) = 25 °C, say. If TF is suddenly raised to 100 °C, then this represents a step change ∆TF of height 75 °C. The corresponding change in sensor temperature is given by ∆T = 75(1 − e−t/τ ) and the actual temperature T of the sensor at time t is given by: T(t) = 25 + 75(1 − e−t/τ )
[4.26]
Thus at time t = τ, T = 25 + (75 × 0.63) = 72.3 °C. By measuring the time taken for T to rise to 72.3 °C we can find the time constant τ of the element.
4.2 I D E N T I F I CATI O N O F TH E D Y N AMI CS O F AN E LE ME N T
59
Figure 4.6 Response of a first-order element to a unit step.
If the second-order element with transfer function 1 1 2 2ξ s + s +1 ω n2 ωn
G (s) =
is subject to a unit step input signal, then the Laplace transform of the element output signal is: g o ( s) =
1
[4.27]
 1 2  2ξ  ω 2 s + ω s + 1 s  n  n
Expressing [4.27] in partial fractions we have: g o ( s) =
As + B  1 2 2ξ   ω 2 s + ω s + 1  n  n
+
C s
[4.27]
where A = −1/ω 2n , B = −2ξ /ωn, C = 1. This gives: go (s) =
1 (s + 2ξωn ) − s s 2 + 2ξωn s + ω 2n
=
1 (s + 2ξωn ) − s (s + ξωn )2 + ω 2n (1 − ξ 2 )
=
1 (s + ξωn ) ξωn − − 2 2 2 2 s (s + ξωn ) + ω n (1 − ξ ) (s + ξωn ) + ω 2n (1 − ξ 2 )
[4.28]
There are three cases to consider, depending on whether ξ is greater than 1, equal to 1, or less than 1. For example, if ξ = 1 (critical damping) then: go (s) =
1 1 ωn − − s s + ωn (s + ωn )2
[4.29]
60 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS Using Table 4.1 we have: Response of secondorder element to a unit step, critical damping ξ = 1 Second-order step response, underdamping ξ < 1
fo (t) = 1 − e− ω n t (1 + ωnt)
[4.30]
Using standard tables it can be shown that if ξ < 1 (underdamping) then:  fo (t ) = 1 − e−ξω t cos ω n (1 − ξ 2 ) t +  n
 sin ω n (1 − ξ 2 ) t   1 − ξ2
ξ
[4.31]
and if ξ > 1 (overdamping) then: Second-order step response, overdamping ξ > 1
 fo (t ) = 1 − e−ξω t cos h ω n (ξ 2 − 1) t +  n
 sinh ω n (ξ 2 − 1) t   ξ2 − 1
ξ
[4.32]
We consider the case of the underdamped response with ξ < 1, given by eqn [4.31]. Here the damped angular frequency of the oscillations is given by:
ωd = ωn 1 − ξ 2 where ωn is the natural or undamped angular frequency. The corresponding period Td of the damped oscillations is given by: Td =
2π 2π = ωd ωn 1 − ξ 2
The time Tp at which the first oscillation peak occurs is correspondingly given by: Tp = Td /2 =
π
ωn 1 − ξ 2
The settling time Ts taken for the response to settle out at the steady-state value is governed by the exponential decay term e−ξωn t. When t = 5/ξωn, i.e. ξωnt = 5, then e−5 = 0.0067. The time for the response to settle out approximately within 1% is therefore: Ts = 5/ξωn From Figure 4.7, we see that Ts is minimum when ξ = 0.7. The response fo(t) has a maximum value f op at the peak of the first oscillation. The difference ( f op − 1) between maximum and steady-state values of fo(t), for ξ < 1, is termed the maximum overshoot and is given by:  − πξ  exp   2  1 − ξ  and depends only on ξ. Thus for ξ < 1, an estimate of ξ can be found from measurement of maximum overshoot and then, knowing ξ, ωn can be estimated from measurements of Td, Tp and/or Ts.
4.2 I D E N T I F I CATI O N O F TH E D Y N AMI CS O F AN E LE ME N T
61
Figure 4.7 Response of a second-order element to a unit step.
The form of the responses is shown in Figure 4.7. As an example consider the step response of a force sensor with stiffness k = 103 N m−1, mass m = 10−1 kg and damping constant λ = 10 N s m−1. The steady-state sensitivity K = 1/k = 10−3 m N −1, natural frequency ωn = êkú /m ú ú = 102 rad s−1 and damping coefficient ξ = λ /2 êkú mú = 0.5. Initially at time t = 0−, a steady force F(0−) = 10 N causes a steady displacement of (1/103) × 10 metre, i.e. 10 mm. Suppose that at t = 0 the force is suddenly increased from 10 to 12 N, i.e. there is a step change ∆F of 2 N. The resulting change ∆x(t) in displacement is found using ∆x(t) = steady-state sensitivity × step height × unit step response fo (t) [4.33] i.e. ∆x(t) =
1 × 2 × [1 − e−50t (cos 86.6t + 0.58 sin 86.6t)] metre 103
= 2[1 − e−50t(cos 86.6t + 0.58 sin 86.6t)] mm
[4.34]
Here the damped angular frequency ωd = 86.6 rad/s and the period of the damped oscillations Td = 72.6 ms. The maximum overshoot is 0.16, giving f op = 1.16 so that the maximum value of ∆x is 2.32 mm. This occurs at time Tp = 36.3 ms. Eventually as t becomes large ∆x tends to 2 mm, i.e. x settles out at a new steady-state value of 12 mm with a settling time Ts of 100 ms.
4.2.2
Sinusoidal response of first- and second-order elements From Table 4.1 we see that the Laplace transform of sine wave f (t) = sin ω t, with unit amplitude and angular frequency ω, is g(s) = ω /(s 2 + ω 2). Thus if a sine wave of amplitude Î is input to a first-order element, then the Laplace transform of the output signal is
62 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS
go (s) =
1 Î⋅ ω 2 (1 + τ s) (s + ω 2 )
[4.35]
Expressing [4.35] in partial fractions we have go (s) =
A Bs + C + 2 (1 + τ s) s + ω 2
[4.36]
where: A=
ωτ 2Î −ωτ Î ωÎ , B= , C= 2 2 2 2 (1 + τ ω ) (1 + τ ω ) (1 + τ 2ω 2 )
so that: go (s) =
− ωτ s + ω  ωτ 2 ï 1 ï +   (1 + τ 2ω 2 ) (1 + τ s) (1 + τ 2ω 2 )  s 2 + ω 2 
ωτ 2 ï 1 = + 2 2 (1 + τ ω ) (1 + τ s) ωτ 2 ï 1 + (1 + τ 2ω 2 ) (1 + τ s)
=
1 − ωτ  +s ω 2 2  1+τ ω 1 + τ 2ω 2 1 + τ 2ω 2  s2 + ω 2 
    
 ω cos φ + s sin φ    s2 + ω 2  1 + τ 2ω 2 
[4.37]
ï
ï
where cos φ =
1 1+τ ω 2
, 2
sin φ =
−ωτ 1 + τ 2ω 2
Using Table 4.1 we have: fo (t ) =
ωτï ï e−t /τ + sin(ω t + φ ) 11+ τ 22 ω 2 3 11 + τ 2ω 22 3 Transient term Sinusoidal term
[24.38]
In a sine wave test experiment, we wait until the transient term has decayed to zero and measure the sinusoidal signal: fo (t ) =
ï 1 + τ 2ω 2
sin(ω t + φ )
[4.39]
which remains. We see therefore that the output signal is also a sine wave of frequency ω but with amplitude Î/ê1ú ú+ú úτ ú2úωú 2ú , and shifted in phase by φ = −tan−1(ωτ) relative to the input sine wave. These amplitude and phase results can be found directly from the transfer function G(s) = 1/(1 + τs) without having to use the table of transforms. If we replace s by jω ( j = ê −ú1ú ) in G(s) we form the complex number G( jω) = 1/(1 + jτω). The magnitude |G( jω )| = 1/ê1ú ú+ú úτ ú2úωú 2ú of this complex number is equal to the ratio of output amplitude to input amplitude, and the angle or argument arg G( jω) = −tan−1(ωτ) is equal to the phase difference φ between output and input sine waves. Figure 4.8 shows amplitude ratio versus frequency and phase versus frequency graphs for a first-order element; these are known as the frequency response characteristics of the element. From the above equations we see that when ωτ = 1,
4.2 I D E N T I F I CATI O N O F TH E D Y N AMI CS O F AN E LE ME N T
63
Figure 4.8 Frequency response characteristics of first-order element with 1 G(s) = . 1 + τs
Figure 4.9 Frequency response of an element with linear dynamics.
i.e. ω = 1/τ, the amplitude ratio = 1/ê2 and phase difference φ = −45°. These results enable the value of τ to be found from experimental frequency response data. The above results can be generalised to an element with steady-state sensitivity K (or ∂O/∂I ) and transfer function G(s), subject to a sinusoidal input signal I = Î sin ω t as in Figure 4.9. In the steady state we can make four statements about the output signal: (a) (b) (c) (d)
O is also a sine wave. The frequency of O is also ω. The amplitude of O is Ô = K |G( jω)|Î. The phase difference between O and I is φ = arg G( jω).
Using the above rules we can quickly find the amplitude ratio and phase relations for a second-order element with: G (s) =
1 1 2 2ξ s + s +1 ω n2 ωn
Here we have: G ( jω ) =
1 1 2ξ ( jω )2 + ( jω ) + 1 2 ωn ωn
64 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS so that Frequency response characteristics of second-order element
Amplitude ratio = |G( jω)| =
1 2
 ω2 ω2 2 1 − 2  + 4ξ 2 ωn  ωn 
 2ξω /ω n  Phase difference = arg G( jω) = − tan−1   2 2 1 − ω /ω n 
[4.40]
These characteristics are shown graphically in Figure 4.10; both amplitude ratio and phase characteristics are critically dependent on the value of ξ. We note that for ξ < 0.7, |G( jω) | has a maximum value which is greater than unity. This maximum value is given by: |G( jω)| MAX =
1 2ξ 1 − ξ 2
and occurs at the resonant frequency ωR where:
ωR = ω n 1 − 2ξ 2 , ξ < 1/ 2 Figure 4.10 Frequency response characteristics of second-order element with G(s) =
1 1 2 2ξ s + s +1 ω n2 ωn
4.3 D Y N AMI C ER R O R S I N MEAS UR EMEN T S Y S T EM S
65
Thus by measuring | G( jω)| MAX and ωR, ξ and ωn can be found. An alternative to plotting |G( jω)| versus ω is a graph of the number of decibels N dB versus ω, where N = 20 log10 |G( jω)|. Thus if |G( jω)| = 1, N = 0 dB; if |G( jω)| = 10, N = +20 dB; and if |G( jω)| = 0.1, N = −20 dB.
4.3
Dynamic errors in measurement systems Figure 4.11 shows a complete measurement system consisting of n elements. Each element i has ideal steady-state and linear dynamic characteristics and can therefore be represented by a constant steady-state sensitivity Ki and a transfer function Gi (s).
Figure 4.11 Complete measurement system with dynamics.
We begin by assuming that the steady-state sensitivity K1 K2 . . . Ki . . . Kn for the overall system is equal to 1, i.e. the system has no steady-state error (Section 3.1). The system transfer function G(s) is the product of the individual element transfer functions, i.e. Transfer function for complete measurement system
∆/(s) = G(s) = G1(s)G2(s) . . . Gi(s) . . . Gn(s) ∆-(s)
[4.41]
In principle we can use eqn [4.41] to find the system output signal ∆O(t) corresponding to a time-varying input signal ∆I(t). We first find the Laplace transform ∆-(s) of ∆I(t), then using [4.9] the Laplace transform of the output signal is ∆/(s) = G(s)∆-(s). By expressing ∆/(s) in partial fractions, and using standard tables of Laplace transforms, we can find the corresponding time signal ∆O(t). Expressing this mathematically: ∆O(t) = 0002 −1[G(s)∆-(s)]
[4.42]
−1
where 0002 denotes the inverse Laplace transform. The dynamic error E(t) of the measurement system is the difference between the measured signal and the true signal, i.e. the difference between ∆O(t) and ∆I(t): Dynamic error of a measurement system
E(t) = ∆O(t) − ∆I(t)
[4.43]
Using [4.42] we have: E(t) = 0002 −1[G(s)∆-(s)] − ∆I(t)
[4.44]
The simple temperature measurement system (Figure 4.12), first introduced in Section 3.1, provides a good example of dynamic errors. The thermocouple has a time constant of 10 s, the amplifier has a time constant of 10−4 s (Chapter 9), and the recorder (Chapter 11) is a second-order element with ωn = 200 rad/s and ξ = 1.0. The overall steady-state sensitivity of the system is unity.
66 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS Figure 4.12 Simple temperature measurement system with dynamics.
We can now calculate the dynamic error of the system for a step input of +20 °C, i.e. ∆TT (t) = 20u(t) and ∆>T (s) = 20(1/s). Thus the Laplace transform of the output signal is: ∆>M (s) = 20
1 1 1 1 −4 s (1 + 10s) (1 + 10 s) (1 + 1/200s)2
 1 A B Cs + D  − − = 20  −  4  s ( s + 0.1) ( s + 10 ) (s + 200)2  Using Table 4.1 and eqn [4.30],
[4.45]
∆TM (t) = 20{u(t) − A e−0.1t − B e−10 t − E e−200t(l + 200t)} 4
and the dynamic error: E(t) = ∆TM (t) − ∆TT (t) 4 = −20{A e−0.1t + B e−10 t − E e−200t(1 + 200t)}
[4.46] −10 4t
where the negative sign indicates too low a reading. The B e term decays to zero after about 5 × 10− 4 s, and the E e−200t(1 + 200t) term decays to zero after about 25 ms. The A e−0.1t term, which corresponds to the 10 s time constant of the thermocouple, takes about 50 s to decay to zero and so has the greatest effect on the dynamic error. We can use the rules developed in Section 4.2.2 to find the dynamic error of a system, with transfer function G(s) subject to a sinusoidal input ∆I(t) = Î sin ω t. From Figure 4.9 we have: ∆O(t) = |G( jω)| Î sin(ω t + φ) giving E(t) = Î{|G( jω)| sin(ω t + φ) − sin ω t}
[4.47]
where φ = arg G( jω). Suppose that the above temperature measurement system is measuring a sinusoidal temperature variation of amplitude ñT = 20 °C and period T = 6 s, i.e. angular frequency ω = 2π/T ≈ 1.0 rad s−1. The frequency response function G( jω) is: G( jω) =
1 (1 + 10 jω )(1 + 10− 4 jω )[1 + 10 −2 jω + 2.5 × 10−5 ( jω )2 ]
[4.48]
so that at ω = 1 |G( jω)|ω =1 =
1 −8
(1 + 100)(1 + 10 )[(1 − 2.5 × 10−5 )2 + 10− 4 ]
≈ 0.10
and
[4.49] arg G( jω)ω =1 ≈ 0 − tan−1(10) − tan−1(10−4 ) − tan−1(10−2) ≈ −85°
4.3 D Y N AMI C ER R O R S I N MEAS UR EMEN T S Y S T EM S
67
We note from the above equations that the values of |G( jω)| and arg G( jω) at ω = 1 are determined mainly by the 10 s time constant; the dynamic characteristics of the other elements will only begin to affect the system performance at much higher frequencies. Since TT (t) = 20 sin t and TM (t) = 0.1 × 20 sin(t − 85°), the error is: E(t) = 20{0.1 sin(t − 85°) − sin t}
[4.50]
We note that in the case of a sine wave input, the output recording is also a sine wave, i.e. the waveform of the signal is unaltered even though there is a reduction in amplitude and a phase shift. In practice the input signal to a measurement system is more likely to be periodic rather than a simple sine wave. A periodic signal is one that repeats itself at equal intervals of time T, i.e. f (t) = f(t + T ) = f (t + 2T ) etc. where T is the period. One example of a periodic measurement signal is the time variation of the temperature inside a diesel engine; another is the vibration of the casing of a centrifugal compressor. In order to calculate dynamic errors for periodic signals, we need to use Fourier analysis. Any periodic signal f (t) with period T can be expressed as a series of sine and cosine waves; these have frequencies which are harmonics of the fundamental frequency ω1 = 2π/T rad s−1, i.e. Fourier series for periodic signal
f(t) = a0 +
n ∞
∑a
n
cos nω1t +
n 1
n ∞
∑b
n
sin nω 1t
[4.51]
n 1
where
0002 2 b = T0002 an =
2 T
+T/2
f(t) cos nω 1t dt −T/2
[4.52]
+T/2
f(t) sin nω 1 t dt
n
−T/2
and a0 =
1 T
0002
+T/2
f(t) dt = average value of f (t) over T −T/2
If f(t) = ∆I(t), where ∆I(t) is the deviation of measurement input signal I(t) from steadystate or d.c. value I0, then a0 = 0. If we also assume that f(t) is odd, i.e. f(t) = −f(−t), then an = 0 for all n, i.e. there are only sine terms present in the series. This simplifying assumption does not affect the general conclusions drawn in the following section. The system input signal is thus given by ∆I(t) =
n ∞
∑Î
n
sin nω 1t
[4.53]
n 1
where În = bn is the amplitude of the nth harmonic at frequency nω 1. In order to find ∆O(t), let us first suppose that only the nth harmonic În sin nω 1t is input to the system. From Figure 4.9 the corresponding output signal is În |G( jnω1)| sin(nω 1t + φn ) where φn = arg G( jnω1).
68 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS We now require to use the principle of superposition, which is a basic property of linear systems (i.e. systems described by linear differential equations). This can be stated as follows: If an input I1(t) causes an output O1(t) and an input I2(t) causes an output O2(t), then an input I1(t) + I2(t) causes an output O1(t) + O2(t), provided the system is linear. This means that if the total input signal is the sum of many sine waves (equation [4.53]), then the total output signal is the sum of the responses to each sine wave, i.e. n ∞
∑ Î |G( jnω )| sin(nω t + φ )
∆O(t) =
n
1
1
[4.54]
n
n 1
The dynamic error is thus Dynamic error of system with periodic input signal
n ∞
∆E(t) = ∑ În {|G( jnω1)| sin(nω 1t + φn ) − sin nω 1t}
[4.55]
n 1
As an example, suppose that the input to the temperature measurement system is a square wave of amplitude 20 °C and period T = 6 s (i.e. ω 1 = 2π/T ≈ 1 rad s−1). The Fourier series for the input signal is: ∆TT (t) =
80 [sin t + –13 sin 3t + –15 sin 5t + –17 sin 7t + . . . ] π
[4.56]
Figure 4.13 shows the waveforms of the input square wave and the first four Fourier components with frequencies 1, 3, 5 and 7 rad s−1. Figure 4.14 shows the amplitude–frequency and phase–frequency relationships for the input temperature; these define the frequency spectrum of the signal. The spectrum consists of a number of lines at frequencies ω 1, 3ω 1, 5ω 1, etc., of decreasing length to represent the smaller amplitudes of the higher harmonics. In practical cases we can terminate or truncate the series at a harmonic where the amplitude is negligible; in this case we choose n = 7. In order to find the output signal, i.e. the recorded waveform, we need to evaluate the magnitude and argument of G( jω) at ω = 1, 3, 5 and 7 rad s−1. Thus |G( j)| ≈ 0.100,
| G(3j)| ≈ 0.033,
|G(5j)| ≈ 0.020,
|G(7j )| ≈ 0.014
and arg G( j) ≈ −85°,
arg G(3j) ≈ −90°,
arg G(5j) ≈ −92°, arg G(7j) ≈ −93°
[4.57]
Again the above values are determined mainly by the 10 s thermocouple time constant; the highest signal frequency ω = 7 is still well below the natural frequency of the recorder ωn = 200. The system output signal is: ∆TM (t) =
80 [0.100 sin(t − 85°) + 0.011 sin(3t − 90°) π + 0.004 sin(5t − 92°) + 0.002 sin(7t − 93°)]
[4.58]
4.3 D Y N AMI C ER R O R S I N MEAS UR EMEN T S Y S T EM S
69
Figure 4.13 Waveforms for input square wave and Fourier components.
Figure 4.14 shows the system frequency response characteristics, the output signal frequency spectrum and the output waveform. We note that, in the output signal, the amplitudes of the 3rd, 5th and 7th harmonics have been reduced relative to the amplitude of the fundamental. The recorded waveform has therefore a different shape from the input signal as well as being reduced in amplitude and changed in phase.
70 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS
Figure 4.14 Calculation of dynamic errors with periodic input signal.
4.4
The above ideas can be extended to calculating the dynamic error for random input signals. Random signals can be represented by continuous frequency spectra (Chapter 6).
Techniques for dynamic compensation From eqn [4.55] we see that in order to have E(t) = 0 for a periodic signal, the following conditions must be obeyed: |G( jω 1)| = |G( j2ω 1)| = . . . = |G( jnω 1)| = . . . = |G( jmω 1)| = 1 arg G( jω 1) = arg G( j2ω 1) = . . . = arg G( jnω 1) = . . . = arg G( jmω 1 ) = 0 [4.59]
4.4 TECHNIQUES FOR DYNAMIC COMPENSATION
71
Figure 4.15 Percentage limits and bandwidth.
where m is the order of the highest significant harmonic. For a random signal (Chapter 6) with a continuous frequency spectrum containing frequencies between 0 and ω MAX, we require: |G( jω)| = 1 and arg G( jω) = 0
for 0 < ω ≤ ω MAX
[4.60]
The above conditions represent a theoretical ideal which will be difficult to realise in practice. A more practical criterion is one which limits the variation in |G( jω)| to a few per cent for the frequencies present in the signal. For example the condition 0.98 < |G( jω)| < 1.02 for 0 < ω ≤ ω MAX
[4.61]
will ensure that the dynamic error is limited to ≈ ±2 per cent for a signal containing frequencies up to ω MAX /2π Hz (Figure 4.15). Another commonly used criterion is that of bandwidth. The bandwidth of an element or a system is the range of frequencies for which | G( jω)| is greater than 1/ê2. Thus the bandwidth of the system, with frequency response shown in Figure 4.15, is 0 to ωB rad s−1. The highest signal frequency ω MAX must be considerably less than ωB. Since, however, there is a 30% reduction in |G( jω)| at ωB, bandwidth is not a particularly useful criterion for complete measurement systems. Bandwidth is commonly used in specifying the frequency response of amplifiers (Chapter 9); a reduction in | G( jω)| from 1 to 1/ê2 is equivalent to a decibel change of N = 20 log(1/ê2) = −3.0 dB. A first-order element has a bandwidth between 0 and l/τ rad s−1. If a system fails to meet the specified limits on dynamic error E(t), i.e. the system transfer function G(s) does not satisfy a condition such as [4.61], then the first step is to identify which elements in the system dominate the dynamic behaviour. In the temperature measurement system of the previous section, the dynamic error is almost entirely due to the 10 s time constant of the thermocouple. Having identified the dominant elements in the system, the most obvious method of improving dynamic response is that of inherent design. In the case of a first-order temperature sensor with τ = MC/UA, τ can be minimised by minimising the mass/ area ratio M/A – for example by using a thermistor in the form of a thin flake. In the case of a second-order force sensor with ωn = êk ú/úmú , ωn can be maximised by maximising k/m, i.e. by using high stiffness k and low mass m. Increasing k, however, reduces the steady-state sensitivity K = 1/k.
72 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS Figure 4.16 Open-loop dynamic compensation.
From second-order step and frequency response graphs we see that the optimum value of damping ratio ξ is around 0.7. This value ensures minimum settling time for the step response and | G( jω)| closest to unity for the frequency response. Another possible method is that of open-loop dynamic compensation (Figure 4.16). Given an uncompensated element or system GU (s), a compensating element GC (s) is introduced into the system, such that the overall transfer function G(s) = GU (s)GC (s) satisfies the required condition (for example eqn [4.61]). Thus if a lead/lag circuit (Figure 9.12) is used with a thermocouple (Figure 4.16), the overall time constant is reduced to τ2 so that | G( jω)| is close to unity over a wider range of frequencies. The main problem with this method is that τ can change with heat transfer coefficient U, thus reducing the effectiveness of the compensation (Chapter 14). Another method is to incorporate the element to be compensated into a closedloop system with high-gain negative feedback. An example of this is the constant temperature anemometer system for measuring fluid velocity fluctuations (Section 14.3). Another example is the closed-loop accelerometer shown in schematic and block diagram form in Figure 4.17. Figure 4.17 Schematic and block diagram of closed-loop accelerometer.
CONCLUSION
73
The applied acceleration a produces an inertia force ma on the seismic mass m (Chapter 8). This is balanced by the force of the permanent magnet on the current feedback coil. Any imbalance of forces is detected by the elastic force element to produce a displacement which is detected by a potentiometric displacement sensor (Chapter 5). The potentiometer output voltage is amplified, giving a current output which is fed to the feedback coil through a standard resistor to give the output voltage. Analysis of the block diagram shows that the overall system transfer function is:  1 ∆ ( s) mR    = k ks  1 2 2ξ 1 s + + 1 + ∆ ã( s ) KF   2 K A KD KF  ω n K A KD KF   K A KD KF ω n
    
[4.62]
If KA is made large so that KA KD KF /k 0002 1 (Chapter 3), then the system transfer function can be expressed in the form: ∆ ( s) ∆ ã ( s)
=
Ks 1
ω
2 ns
s2 +
2ξ s
ω ns
s +1
where system steady-state sensitivity
system natural frequency
system damping ratio
Ks =
mR KF
ω ns = ω n ξs = ξ
KA KD KF k k KA KD KF
[4.63]
We see that the system natural frequency ω ns is now much greater than that of the elastic force element itself. The system damping ratio ξ s is much less than ξ, but by making ξ large a value of ξ s ≈ 0.7 can be obtained. Furthermore the system steady-state sensitivity depends only on m, KF and R, which can be made constant to a high degree.
Conclusion The dynamic characteristics of typical measurement system elements were initially discussed; in particular the transfer functions of first- and second-order elements were derived. The response of both first- and second-order elements to step and sine wave inputs was then studied. A general description of the dynamic error of a complete measurement system was then developed and applied to a temperature measurement system subject to step, sine wave and periodic input signals. Finally methods of dynamic compensation, which reduce dynamic error, were explained.
74 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS
Problems 4.1
A temperature measurement system consists of linear elements and has an overall steady-state sensitivity of unity. The dynamics of the system are determined by the first-order transfer function of the sensing element. At time t = 0, the sensing element is suddenly transferred from air at 20 °C to boiling water. One minute later the element is suddenly transferred back to air. Using the data given below, calculate the system dynamic error at the following times: t = 10, 20, 50, 120 and 300 s. Sensor data Mass = 5 × 10−2 kg Surface area = 10−3 m2 Specific heat = 0.2 J kg−1 °C −1 Heat transfer coefficient for air = 0.2 W m−2 °C −1 Heat transfer coefficient for water = 1.0 W m−2 °C −1
4.2
A force sensor has a mass of 0.5 kg, stiffness of 2 × 102 N m−1 and a damping constant of 6.0 N s m−1. (a) (b) (c)
4.3
Calculate the steady-state sensitivity, natural frequency and damping ratio for the sensor. Calculate the displacement of the sensor for a steady input force of 2 N. If the input force is suddenly increased from 2 to 3 N, derive an expression for the resulting displacement of the sensor.
A force measurement system consists of linear elements and has an overall steady-state sensitivity of unity. The dynamics of the system are determined by the second-order transfer function of the sensing element, which has a natural frequency ωn = 40 rad s−1 and a damping ratio ξ = 0.1. Calculate the system dynamic error corresponding to the periodic input force signal: F(t) = 50(sin 10t + –13 sin 30t + –15 sin 50t)
4.4
An uncompensated thermocouple has a time constant of 10 s in a fast-moving liquid. (a) (b) (c)
(d)
4.5
Calculate the bandwidth of the thermocouple frequency response. Find the range of frequencies for which the amplitude ratio of the uncompensated thermocouple is flat within ±5%. A lead /lag circuit with transfer function G(s) = (1 + 10s)/(1 + s) is used to compensate for thermocouple dynamics. Calculate the range of frequencies for which the amplitude ratio of the compensated system is flat within ±5%. The velocity of the liquid is reduced, causing the thermocouple time constant to increase to 20 s. By sketching |G( jω )| explain why the effectiveness of the above compensation is reduced.
An elastic force sensor has an effective seismic mass of 0.1 kg, a spring stiffness of 10 N m−1 and a damping constant of 14 N s m−1. (a)
Calculate the following quantities: (i) (ii) (iii)
(b)
sensor natural frequency sensor damping ratio transfer function relating displacement and force.
The above sensor is incorporated into a closed-loop, force balance accelerometer. The following components are also present: Potentiometer displacement sensor: sensitivity 1.0 V m−1 Amplifier: voltage input, current output, sensitivity 40 A V−1 Coil and magnet: current input, force output, sensitivity 25 N A−1 Resistor: 250 Ω.
PROBLEMS
(i) (ii) (iii) 4.6
Draw a block diagram of the accelerometer. Calculate the overall accelerometer transfer function. Explain why the dynamic performance of the accelerometer is superior to that of the elastic sensor.
A load cell consists of an elastic cantilever and a displacement transducer. The cantilever has a stiffness of 102 N m−1, a mass of 0.5 kg and a damping constant of 2 N s m−1. The displacement transducer has a steady-state sensitivity of 10 V m−1. (a)
(b)
4.7
75
A package of mass 0.5 kg is suddenly dropped onto the load cell. Use eqn [4.31] to derive a numerical equation describing the corresponding time variation of the output voltage (g = 9.81 m s−2). The load cell is used to weigh packages moving along a conveyor belt at the rate of 60 per minute. Use the equation derived in (a) to explain why the load cell is unsuitable for this application. Explain what modifications to the load cell are necessary.
A force measurement system consisting of a piezoelectric crystal, charge amplifier and recorder is shown in Figure Prob. 7. (a)
Calculate the system dynamic error corresponding to the force input signal: F(t) = 50(sin 10t + –13 sin 30t + –15 sin 50t)
(b)
Explain briefly the system modifications necessary to reduce the error in (a) (hint: see Figure 8.22).
Figure Prob. 7.
4.8
A thermocouple is used to measure the temperature inside a vessel, which is part of a high-speed batch process. At time t = 0, with the vessel at an initial temperature of 50 °C, the vessel is instantaneously filled with gas at 150 °C. One minute later, instantaneously the gas is removed and the vessel is filled with liquid at 50 °C. The thermocouple can be regarded as having linear steady-state characteristics and first-order dynamics. (a)
(b) (c)
Use the data given below to sketch a graph of how the thermocouple e.m.f. changes with time. The axes of the graph should have suitable scales and the answer should include supporting numerical calculations. Comment on whether the thermocouple is suitable for this application. What modifications should be made?
Data Thermocouple sensitivity = 40 µV °C −1 Thermocouple mass = 5 × 10−2 kg Thermocouple specific heat = 0.2 J kg−1 °C −1 Thermocouple surface area = 10−3 m2 Heat transfer coefficient for gas = 0.2 W m−2 °C −1 Heat transfer coefficient for liquid = 1.0 W m−2 °C −1 Response of first-order element, with unit sensitivity, to unit step: F 0(t) = 1 − exp
A −t D CτF
76 DYNAMIC CHARACTERISTICS OF MEASUREMENT SYSTEMS
4.9
A temperature measurement system for a gas reactor consists of linear elements and has an overall steady-state sensitivity of unity. The temperature sensor has a time constant of 5.0 s; an ideal low-pass filter with a cut-off frequency of 0.05 Hz is also present. The input temperature signal is periodic with period 63 s and can be approximated by the Fourier series: T (t) = 10(sin ω 0 t + –12 sin 2ω 0 t + –13 sin 3ω 0 t + –14 sin 4ω 0 t) where ω 0 is the angular frequency of the fundamental component. (a)
Calculate expressions for the time response of: (i) (ii)
(b)
the system output signal the system dynamic error.
Explain what modifications are necessary to the system to minimise the dynamic error in (a).
Note An ideal low-pass filter has a gain of one and zero phase shift up to the cut-off frequency. The gain is zero above the cut-off frequency.
5 Loading Effects and Two-port Networks In our discussion of measurement systems no consideration has yet been given to the effects of loading. One important effect is that of inter-element loading where a given element in the system may modify the characteristics of the previous element (for example by drawing current). In turn, the characteristics of this element may be modified by the following element in the system. Inter-element loading is normally an electrical loading effect which is described in the first section of this chapter using Thévenin and Norton equivalent circuits. The second section begins by discussing the analogies between electrical and non-electrical variables. This means that mechanical and thermal systems can be described by equivalent circuits and sensing elements by two-port networks. Two-port networks are then used to describe process loading; here the introduction of the sensing element into the process or system being measured causes the value of the measured variable to change. Finally two-port networks are used to describe bilateral transducers which use reversible physical effects.
5.1
Electrical loading We have so far represented measurement systems as blocks connected by single lines where the transfer of information and energy is in terms of one variable only. Thus in the temperature measurement system of Figure 3.2 the information transfer between elements is in terms of voltage only. No allowance can therefore be made for the amplifier drawing current from the thermocouple and the indicator drawing current from the amplifier. In order to describe both voltage and current behaviour at the connection of two elements, we need to represent each element by equivalent circuits characterised by two terminals. The connection is then shown by two lines.
5.1.1
Thévenin equivalent circuit Thévenin’s theorem states that any network consisting of linear impedances and voltage sources can be replaced by an equivalent circuit consisting of a voltage source ETh and a series impedance ZTh (Figure 5.1). The source ETh is equal to the open circuit voltage of the network across the output terminals, and ZTh is the impedance looking back into these terminals with all voltage sources reduced to zero and replaced by their internal impedances. Thus connecting a load ZL across the output
78 LOA DING EFFECTS AND TWO-PORT NE TWO RK S Figure 5.1 Thévenin equivalent circuit.
terminals of the network is equivalent to connecting ZL across the Thévenin circuit. The current i in ZL is then simply given by: i=
ETh ZTh + Z L
[5.1]
and the voltage VL across the load by: Loading of Thévenin equivalent circuit
VL = iZ L = ETh
ZL ZTh + Z L
[5.2]
From eqn [5.2] we see that if ZL 0002 Z Th, then VL → ETh; i.e. in order to get maximum voltage transfer from the network to the load, the load impedance should be far greater than the Thévenin impedance for the network. In order to get maximum power transfer from network to load, the load impedance should be equal to the network impedance; i.e. ZL = ZTh.[1] (An example of the calculation of ETh and ZTh for a potentiometer displacement transducer is given in the following section and for a deflection bridge in Section 9.l.) We can now discuss the Thévenin equivalent circuit for the temperature measurement system of Figure 3.2. The thermocouple may be represented by ZTh = 20 Ω (resistive) and ETh = 40T µV, where T is the measurement junction temperature, if non-linear and reference junction temperature effects are ignored. The amplifier acts both as a load for the thermocouple and as a voltage source for the indicator. Figure 5.2 shows a general equivalent circuit for an amplifier with two pairs of terminals. Using typical amplifier data (Chapter 9), we have input impedance Z IN = R IN = 2 × 106 Ω, closed-loop voltage gain A = 10 3, and output impedance Z OUT = R OUT = 75 Ω. The indicator is a resistive load of 104 Ω. The complete equivalent circuit for the system is shown in Figure 5.3, and using eqn [5.2] we have: V IN = 40 × 10−6T VL = 1000 VIN
A 2 × 106 D C 2 × 106 + 20 F
A 104 D C 104 + 75 F
[5.3]
If the indicator scale is drawn so that a change of 1 V in VL causes a change in deflection of 25 °C, then the measured temperature TM = 25VL. This gives:
5.1 E LE CT R I CAL LO AD I N G
79
Figure 5.2 Equivalent circuit for amplifier.
Figure 5.3 Thévenin equivalent to temperature measurement system.
TM =
A 2 × 106 D A 104 D T = 0.9925T C 2 × 106 + 20 F C 104 + 75 F
[5.4]
i.e. we have to introduce the factor ZL /(ZTh + ZL) at every interconnection of two elements to allow for loading. The loading error = −0.0075T; this is in addition to the steady-state error due to element inperfections calculated in Chapter 3. The loading error in the above example is small, but if care is not taken it can be very large. Suppose a pH glass electrode (Chapter 8), with sensitivity 59 mV per pH, i.e. ETh = 59pH mV and ZTh = RTh = 109 Ω, is directly connected to an indicator with ZL = RL = 104 Ω and a scale of sensitivity 59 –1 pH/mV. The measured pH is: pHM = 59pH
A 104 D 1 ≈ 10−5 pH C 104 + 109 F 59
[5.5]
i.e. there will be effectively a zero indication for any non-zero value. This problem is solved by connecting the electrode to the indicator via a buffer amplifier. This is characterised by large Z IN, small Z OUT and unity gain A = 1. For example, an operational amplifier with a field effect transistor (FET) input stage connected as a voltage follower (Figure 9.9) would have Z IN = 1012 Ω, Z OUT = 10 Ω. The indicated pH value for the modified system (Figure 5.4) is: pHM =
1012 104 × 4 pH 9 10 + 10 10 + 10 12
and the loading error is now −0.002pH, which is negligible.
Figure 5.4 Equivalent circuit for pH measurement system.
[5.6]
80 LOA DING EFFECTS AND TWO-PORT NE TWO RK S Figure 5.5 A.C. loading of tachogenerator.
An example of a.c. loading effects is given in Figure 5.5, which shows the equivalent circuit for a variable reluctance tachogenerator connected to a recorder. The Thévenin voltage ETh for the tachogenerator is a.c. with an amplitude : and angular frequency ω, both proportional to the mechanical angular velocity ω r (Section 8.4). In this example, : = (5.0 × 10−3) ω r volts and ω = 6ω r rad s −1. The Thévenin impedance ZTh for the tachogenerator is an inductance and resistance in series (coil surrounding magnet), i.e. ZTh = RTh + jω LTh. Thus if ω r = 103 rad s−1: : = 5 V, ω = 6 × 103 rad s−1 and ZTh = 1.5 + 6.0j kΩ so that the amplitude of the recorded voltage is ÄL = :
RL | ZTh + RL |
=5
10 (11.5)2 + (6.0)2
= 3.85 V
[5.7]
If the recorder scale sensitivity is set at 1/(5 × 10−3) rad s−1 V −1, then the recorded angular velocity is 770 rad s−1. This error can be removed either by increasing the recorder impedance, or by changing the recorder sensitivity to allow for loading effects. A better alternative is to replace the recorder by a counter which measures the frequency rather than the amplitude of the tachogenerator signal (Section 10.3).
5.1.2
Example of Thévenin equivalent circuit calculation: potentiometric displacement sensor Figure 5.6 shows a schematic diagram of a potentiometric sensor for measuring displacements d. The resistance of the potentiometer varies linearly with displacement.
Figure 5.6 Potentiometer displacement sensor and Thévenin equivalent circuit.
5.1 E LE CT R I CAL LO AD I N G
81
Figure 5.7 Calculation of RTh for potentiometer.
Thus if x = d/dT is the fractional displacement, the corresponding resistance is RP x, where RP Ω is the total resistance of the potentiometer. The Thévinin voltage ETh is the open circuit voltage across the output terminals AB. The ratio between ETh and supply voltage VS is equal to the ratio of fractional resistance RP x to total resistance RP; that is ETh R x = P , giving VS RP
ETh = VS x
[5.8]
The Thévenin impedance ZTh is found by setting supply voltage VS = 0, replacing the supply by its internal impedance (assumed to be zero), and calculating the impedance looking back into the terminals AB as shown in Figure 5.7. Thus: 1 1 1 = + RTh RP x RP (1 − x ) giving RTh = RP x(1 − x)
[5.9]
Thus the effect of connecting a resistive load RL (recorder or indicator) across the terminals AB is equivalent to connecting RL across the Thévenin circuit. The load voltage is thus: VL = ETh
RL RTh + RL
= VS x
RL RP x(1 − x ) + RL
i.e. Voltage–displacement relationship for a loaded potentiometer
VL = VS x
1 (RP / RL ) x(1 − x ) + 1
[5.10]
The relationship between VL and x is non-linear, the amount of non-linearity depending on the ratio RP /RL (Figure 5.8). Thus the effect of loading a linear potentiometric sensor is to introduce a non-linear error into the system given by: 1   N ( x ) = ETh − VL = VS x 1 −  (RP / RL ) x(1 − x ) + 1  i.e.  x 2 (1 − x )(RP / RL )  N ( x ) = VS   1 + ( RP / RL ) x(1 − x ) 
[5.11]
82 LOA DING EFFECTS AND TWO-PORT NE TWO RK S Figure 5.8 Non-linear characteristics of loaded potentiometer.
which reduces to: N(x) ≈ VS (RP /RL)(x 2 − x 3) if RP /RL 0003 1 (the usual situation). N(x) has a maximum value of ; = –27–4 VS (RP /RL ) when x = –23 , corresponding to dN/dx = 0 and negative d 2N/dx 2. Expressing ; as a percentage of full-scale deflection or span VS volts gives: ;=
400 RP RP % ≈ 15 % 27 RL RL
[5.12]
Non-linearity, sensitivity and maximum power requirements are used to specify the values of RP and VS for a given application. Suppose that a 10 cm range potentiometer is to be connected to a 10 kΩ recorder. If the maximum non-linearity must 20 – × 103 Ω; thus a 1 kΩ not exceed 2%, then we require 15RP /RL ≤ 2, i.e. RP ≤ –15 potentiometer would be suitable. Since sensitivity dVL /dx ≈ VS , the greater VS the higher the sensitivity, but we must satisfy the requirement that power dissipation VS2 /RP should not exceed maximum value Å watts. If Å = 0.1 W we require VS ≤ ê0.ú1ú ú×ú ú1ú0ú3ú , i.e. VS ≤ 10 V; if VS = 10 V, then corresponding sensitivity = 1.0 V cm−1.
5.1.3
Norton equivalent circuit Norton’s theorem states that any network consisting of linear impedances and voltage sources can be replaced by an equivalent circuit consisting of a current source iN in parallel with an impedance Z N (Figure 5.9). ZN is the impedance looking back into the output terminals with all voltage sources reduced to zero and replaced by their internal impedances, and iN is the current which flows when the terminals are short circuited. Connecting a load ZL across the output terminals of the network is equivalent to connecting ZL across the Norton circuit. The voltage VL across the load is given by VL = iN Z, where 1/Z = 1/ZN + 1/ZL, giving:
5.1 E LE CT R I CAL LO AD I N G
83
Figure 5.9 Norton equivalent circuit.
Loading of Norton equivalent circuit
VL = iN
ZN ⋅ ZL ZN + ZL
[5.13]
From eqn [5.13] we note that if ZL 0003 ZN, then VL → iN ZL; i.e. in order to develop the maximum current through the load, the load impedance should be far smaller than the Norton impedance for the network. A common example of a current source is an electronic differential pressure transmitter giving an output current signal, range 4 to 20 mA, proportional to an input differential pressure, typical range 0 to 2 × 104 Pa (Section 9.4). Figure 5.10 shows a typical equivalent circuit for the transmitter connected to a recorder via a cable. Using eqn [5.13], the voltage across the total load RC + RR of recorder and cable is VL = iN
RN (RC + RR) RN + RC + RR
[5.14]
and the ratio VR /VL = RR /(RC + RR), giving the recorder voltage: VR = iN RR
RN RN + RC + RR
[5.15]
Using the data given, we have VR = 0.9995iN RR, so that the recorded voltage deviates from the desired range of 1 to 5 V by only 0.05%. A second example of a current generator is provided by a piezoelectric crystal acting as a force sensor. If a force F is applied to any crystal, then the atoms of the crystal undergo a small displacement x proportional to F. For a piezoelectric material the crystal acquires a charge q proportional to x, i.e. q = Kx. The crystal can therefore be regarded as a Norton current source of magnitude iN = dq/dt = K(d x/dt), Figure 5.10 Typical current source and load.
84 LOA DING EFFECTS AND TWO-PORT NE TWO RK S Figure 5.11 Piezoelectric force measurement system.
where dx/dt is the velocity of the atomic deformations. This effect is discussed more fully in Section 8.7, where we see that the crystal acts as a capacitance CN in parallel with the current source iN . Figure 5.11 shows the equivalent circuit and typical component values for a crystal connected via a capacitive cable CC to a recorder acting as a resistive load RL. The voltage VL across the load is given by iN Z, where Z is the impedance of CN, CC and RL in parallel. Since 1 1 = CN s + CC s + Z RL Z=
RL 1 + RL(CN + CC)s
where s denotes the Laplace operator. The transfer function relating dynamic changes in source current and recorder voltage is thus: ∆L(s) RL = ∆jN (s) 1 + RL(CN + CC)s
[5.16]
Thus the effect of electrical loading in this example is to introduce a first-order transfer function into the force measurement system; this will affect dynamic accuracy.
5.2 5.2.1
Two-port networks Generalised effort and flow variables We have seen in the previous section how electrical loading effects can be described using a pair of variables, voltage and current. Voltage is an example of an across or effort variable y; current is an example of a through or flow variable x. An effort variable drives a flow variable through an impedance. Other examples of effort/flow pairs are force/velocity, torque/angular velocity, pressure difference/volume flow rate and temperature difference/heat flow rate.[2] Each y–x pair has the following properties: (a) (b)
The product yx represents power in watts. The ratio y/x represents impedance.
The only exception is the thermal variables where the product has the dimensions of watts and temperature. Table 5.1 lists the effort/flow pairs for different forms of energy and for each pair defines the related quantities of impedance, stiffness, compliance
heat
Thermal
Source: adapted from Finkelstein and Watts, 1971[2].
volume
charge
d (current) dt
Electrical
Fluidic
angular displacement
angular acceleration
Mechanicalrotation
displacement
∫ x dt
acceleration
d x dt
Mechanicaltranslation
Variables
heat flow rate
volume flow rate
current
angular velocity
temperature
pressure
voltage
torque
force
angular disp. = torque 1 mech. stiffness
torque = angular disp. mechancial stiffness voltage = charge 1 elect. capacitance pressure = volume 1 fluid capacitance temperature = heat 1 therm. capacitance
torque = ang. velocity damping constant voltage = current electrical resistance pressure = vol. flow rate fluidic resistance temperature = heat flow rate thermal resistance
thermal capacitance
heat = temperature
fluidic capacitance
volume = pressure
electrical capacitance
charge = voltage
1 mech. stiffness
mechanical stiffness
damping constant
=
fluidic inertance
d (flow rate) dt
pressure
voltage = d (current) dt inductance
moment of inertia
torque = angular accn
mass
force = acceleration
y dx/dt
∫ x dt y displacement = force
Inertance
Compliance
force = displacement
∫ x dt
y
Stiffness
force = velocity
y x
y
x
velocity
Impedance
Effort
Flow
Table 5.1 Flow/effort variables and related quantities.
5.2 TWO-PORT NETWORKS
85
86 LOA DING EFFECTS AND TWO-PORT NE TWO RK S
Figure 5.12 Equivalent circuit for a mechanical system: (a) Parallel mechanical system (b) Series electrical circuit (c) Equivalent mechanical circuit.
and inertance. Thus we see that the concept of impedance is applicable to mechanical, fluidic and thermal systems as well as electrical. For a mechanical system, mass is analogous to electrical inductance, damping constant is analogous to electrical resistance, and 1/stiffness is analogous to electrical capacitance. For a thermal system, thermal resistance is analogous to electrical resistance, and thermal capacitance is analogous to electrical capacitance. This means we can generalise the electrical equivalent circuits of Thévenin and Norton to non-electrical systems. Figure 5.12(a) shows a parallel mechanical system consisting of a mass m, spring stiffness k and damper constant λ. Figure 5.12(b) shows a series electrical circuit consisting of an inductance L, a capacitance C and a resistance R. Since mechanical impedance is the ratio of force/velocity, the mechanical impedance transfer function is ZM (s) =
∆, k = ms + λ + ∆á s
[5.17]
The impedance transfer function for the electrical circuit is ZE (s) =
∆ 1 + Ls + R + ∆j Cs
[5.18]
We see that these have a similar form, with m corresponding to L, λ corresponding to R, and k corresponding to 1/C. Thus the parallel mechanical system can be represented by an equivalent circuit consisting of an inductive element m, a resistive element λ and a capacitive element 1/k in series (Figure 5.12(c)). Figure 5.13(a) shows a thermal system consisting of a body at temperature T immersed in a fluid at temperature TF. The body has a mass M, specific heat CH and surface area A; U is the heat transfer coefficient between the body and the fluid. Figure 5.13(b) shows a series electrical circuit with resistance R, capacitance C, input voltage VIN and output voltage VOUT. The differential equation for the thermal system is:
5.2 TWO-PORT NETWORKS
87
Figure 5.13 Equivalent circuit for a thermal system: (a) Thermal system (b) Electrical circuit (c) Equivalent thermal circuit.
MCH
dT = UA(TF − T ) dt
[5.19]
and the differential equation for the electrical circuit is: C
dVOUT 1 = (VIN − VOUT) dt R
[5.20]
We see that the equations have a similar form and the temperature effort variables TF and T are analogous to the voltage effort variables VIN and VOUT. Heat flow W is analogous to current flow i, UA is analogous to 1/R, i.e. the reciprocal of electrical resistance, and MCH is analogous to electrical capacitance C. Thus the thermal system can be represented by an equivalent circuit consisting of a resistive element 1/UA in series with a capacitive element MCH as in Figure 5.13(c).
5.2.2
Two-port networks We saw in Section 5.1 that the electrical output of a sensing element such as a thermocouple or piezoelectric crystal can be represented by a Thévenin or Norton equivalent circuit. The sensor has therefore two output terminals which allow both voltage and current flow to be specified; this is referred to as an electrical output port. The sensing element will have a mechnical, thermal or fluidic input; we saw in the previous section that mechanical and thermal systems can be represented by equivalent circuits which show the relation between the corresponding effort and flow variables. Thus the input to a mechanical or thermal sensor can be represented by two input terminals which allow both the effort and flow variables to be specified; this is either a mechanical or a thermal input port. Thus a sensing element can be represented by a two-port or four-terminal network. Figure 5.14(a) shows a two-port representation of a mechanical sensor with input mechanical port and output electrical port; Figure 5.15(a) shows the two-port representation of a thermal sensor. Figures 5.14(b) and (c) show the detailed two-port networks for a range of mechanical sensing elements. Figure 5.14(b) shows the equivalent circuit for sensing elements with a Thévenin equivalent circuit at the electrical output port. ZM is the input mechanical impedance; ETh and ZTh are the Thévenin voltage and impedance. For a displacement sensor ETh is proportional to displacement x, i.e.
88 LOA DING EFFECTS AND TWO-PORT NE TWO RK S Figure 5.14 Mechanical sensing elements as two-port networks: (a) Overall two-port representation (b) Equivalent circuit with Thévenin output (c) Equivalent circuit with Norton output.
Figure 5.15 Thermal sensing elements as two-port networks: (a) Overall two-port representation (b) Equivalent circuit with Thévenin output (c) Equivalent circuit with Norton output.
ETh = Kx x
[5.21]
where Kx is the sensitivity and x = ∫ x dt. For a potentiometer displacement sensor (Section 5.1.2) K x = VS (supply voltage), and for a linear variable differential transformer (LVDT, Section 8.3.2) Kx is the slope of the linear portion of the a.c. voltage versus displacement characteristics. For a velocity sensor, ETh is proportional to velocity x, i.e.
5.2 TWO-PORT NETWORKS
ETh = KV x
89
[5.22]
For an electromagnetic linear velocity sensor (Section 12.5.1) KV = Bl where B is the applied magnetic field and l the length of the conductor. In the case of an electromagnetic angular velocity sensor or tachogenerator (Section 8.4) we have ETh = KV ωr, where ωr is the angular velocity and KV = dN/dθ the rate of change of flux N with angle θ. Figure 5.14(c) shows the equivalent circuit for sensing elements with a Norton equivalent circuit iN, ZN at the electrical output port. For a piezoelectric sensor iN is proportional to velocity x (Section 8.7), i.e. iN = Kx
[5.23]
where K = dk, d is the charge sensitivity to force and k is the stiffness of the crystal. Figures 5.15(b) and (c) show the detailed two-port networks for two examples of thermal sensing elements. Figure 5.15(b) shows the equivalent circuit for a sensing element with a Thévenin equivalent circuit ETh, ZTh at the electrical output port; ZH is the thermal input impedance. For a thermocouple temperature sensor (Section 8.5) with the reference junction at 0 °C we have ETh = ET,0 where ET,0 is the e.m.f. at the measured junction at T °C and is given by the power series ET,0 = a 1T + a 2 T 2 + a 3 T 3 + . . .
[5.24]
Figure 5.15(c) shows the equivalent circuit for a sensing element with a Norton equivalent circuit iN , ZN at the electrical output port. For a pyroelectric detector (Section 15.5.1) iN is proportional to rate of change of temperature dT/dt, i.e. iN = K
dT dt
[5.25]
where K = A(dP/dT ), A is the area of the electrodes and dP/dT is the slope of the polarisation temperature characteristics.
5.2.3
Process loading Having introduced the concepts of equivalent circuits and two-port networks for mechanical and thermal systems, we can now use these concepts to study examples of how a primary sensing element can ‘load’ the process or element being measured. Figure 5.16 shows a mechanical system or ‘process’ represented by a mass, spring and damper. The force F applied to the ‘process’ is being measured by a force sensor, consisting of an elastic element in conjunction with a potentiometric displacement sensor. The elastic force sensor can also be represented by a mass spring and damper (Section 4.1.2). Under steady-state conditions when both velocity x = 0 and acceleration y = 0, we have the following force balance equations: process F = kP x + FS sensor FS = k S x
[5.26]
showing that the relationship between the measured force FS and the true force F is: Steady-state loading of mechanical system
FS =
kS 1 F= F kS + kP 1 + k P /k S
[5.27]
90 LOA DING EFFECTS AND TWO-PORT NE TWO RK S Figure 5.16 Loading of mechanical system by force sensor.
We see that in order to minimise the loading error in the steady state the sensor stiffness kS should be very much greater than the process stiffness kP. Under unsteady conditions when x and y are non-zero, Newton’s second law gives the following differential equations: process F − kP x − λP x − FS = mP y sensor FS − kS x − λS x = mS y
[5.28]
i.e. mP
dx + λP x + kP ∫ x dt = F − FS dt
[5.29]
dx + λ S x + kS ∫ x dt = FS mS dt Using the analogues given earlier, the sensor can be represented by FS driving x through the mechanical L, C, R circuit mS, 1/kS, λ S ; and the process can be represented by F − FS driving x through the mechanical L, C, R circuit mP, 1/kP, λP. If ∆x, ∆F and ∆FS are deviations from initial steady conditions, then the Laplace transforms of eqns [5.29] are: A kPD mP s + λP + ö = ∆, − ∆,S C sF
[5.30]
A kS D m s + λS + ö = ∆,S C S sF
Using Table 5.1 we can define mechanical impedance transfer functions by ZM (s) = ∆,(s)/ö(s), so that: process impedance ZMP(s) = mP s + λP + sensor impedance
kP s
kS ZMS (s) = mS s + λS + s
[5.31]
5.2 TWO-PORT NETWORKS
91
Figure 5.17 Equivalent circuit for complete system showing a force sensor as a two-port network.
From [5.30] and [5.31] the relationship between measured and actual dynamic changes in force is: Dynamic loading of mechanical system
∆ , S ( s) =
Z MS ∆ , ( s) Z MS + Z MP
[5.32]
Thus in order to minimise dynamic loading effects, sensor impedance ZMS should be very much greater than process impedance ZMP. Figure 5.17 shows the equivalent circuit for the system: process, force sensor and recorder. Figure 5.18 shows a hot body, i.e. a thermal ‘process’ whose temperature TP is being measured by a thermocouple sensor. Under unsteady conditions, heat flow rate considerations give the following differential equations (Section 4.1): process MP CP sensor
Figure 5.18 Loading of thermal ‘process’ by thermocouple temperature sensor.
dTP = WP − WS , dt
dTS MS CS = WS , dt
WP = UP AP (TF − TP) [5.33] WS = US AS (TP − TS)
92 LOA DING EFFECTS AND TWO-PORT NE TWO RK S Figure 5.19 Equivalent circuit for thermal system showing thermocouple as a two-port network.
where M denotes masses C denotes specific heats U denotes heat transfer coefficients A denotes heat transfer areas. The quantities MPCP and MS CS have the dimensions of heat/temperature and are analogous to electrical capacitance. The quantities UP AP and US AS have the dimensions of heat flow rate/temperature and are analogous to 1/(electrical resistance). The equivalent circuit for the process and thermocouple is shown in Figure 5.19. We see that the relationship between TF and TP depends on the potential divider 1/UP AP, MP CP, and the relationship between TP and TS depends on the potential divider 1/(US AS), MS CS . Again the thermocouple can be represented as a two-port network with a thermal input port and an electrical output port. In conclusion we see that the representation of measurement system elements by two-port networks enables both process and inter-element loading effects to be quantified.
5.2.4
Bilateral transducers Bilateral transducers are associated with reversible physical effects. In a reversible effect the same device can, for example, convert mechanical energy into electrical energy and also convert electrical energy into mechanical energy. When the device converts electrical energy into mechanical energy it acts as a transmitter or sender. This can be represented by a two-port network with an input electrical port and an output mechanical port as in Figure 5.20(a), When the device converts mechanical energy into electrical energy it acts as a receiver or sensor, and can be represented by an input mechanical port and an output electrical port as in Figure 5.20(b). The piezoelectric effect is a common example of a reversible effect (Section 8.7). In the direct effect a force F applied to the crystal produces a charge q, proportional to F, according to: q = dF
[5.34]
5.2 TWO-PORT NETWORKS
93
Figure 5.20 Bilateral transducers: (a) Transmitter/sender (b) Receiver/sensor.
i.e. this is a conversion of mechanical energy to electrical energy and the device acts as a receiver or sensor. In the inverse effect a voltage V applied to the crystal produces a mechanical deformation x, proportional to V, according to: x = dV
[5.35]
This is a conversion of electrical energy to mechanical energy and the device acts as a transmitter or sender. The detailed equivalent circuits for a piezoelectric transmitter and receiver are given in Section 16.2.1. Another reversible physical effect is the electromagnetic effect. In the direct effect, a conductor of length l moving with velocity x perpendicular to a magnetic field B has a voltage E = Bl x
[5.36]
induced across the ends of the conductor. This is a conversion of mechanical to electrical energy and the device acts as a receiver or sensor. In the inverse effect a conductor of length l carrying a current i in a transverse magnetic field B experiences a force F = Bli
[5.37]
This is a conversion of electrical to mechanical energy and the device acts as a transmitter or sender. Figure 5.21 gives the detailed equivalent circuits for an electromagnetic transmitter and receiver. At the electrical ports the applied voltage drives a current through the electrical impedance L, R; at the mechanical ports the applied force drives a velocity through the mechanical impedance m, 1/k, λ. The transmitter can be used as a voltage indicator (Section 11.2) and the receiver as a velocity sensor (Sections 5.1.1 and 8.4).
Figure 5.21 Equivalent circuits for bilateral electromagnetic transducers: (a) Transmitter/sender (b) Receiver/sensor.
94 LOA DING EFFECTS AND TWO-PORT NE TWO RK S
Conclusion We have seen how the use of equivalent circuits and two-port networks has enabled both inter-element and process loading effects to be described.
References [1] [2]
bell e c and whitehead r w 1979 Basic Electrical Engineering and Instrumentation for Engineers, Granada, London, pp. 46–7. finkelstein l and watts r d 1971 ‘Systems analysis of instruments’, Journal of the Institute of Measurement and Control, vol. 4, Sept., pp. 236–7.
Problems 5.1
A glass pH electrode with a sensitivity of 59 mV pH −1 and a resistance of 109 Ω is used to measure pH in the range 0 to 15. The electrode is to be connected to a recorder of input range 0 to 100 mV and resistance 100 Ω using a buffer amplifier of unity gain and output resistance 100 Ω. (a) (b)
5.2
Calculate the input impedance of the amplifier, and the sensitivity of the recorder scale necessary to obtain an accurate recording of pH. The resistance of the electrode increases to 2 × 109 Ω due to chemical action. Calculate the resulting measurement error in the above system, as a percentage of full scale, for a true pH of 7.
The motion of a hydraulic ram is to be recorded using a potentiometer displacement sensor connected to a recorder. The potentiometer is 25 cm long and has linear resistance displacement characteristics. A set of potentiometers with maximum power rating of 5 W and resistance values ranging from 250 to 2500 Ω in 250 Ω steps is available. The recorder has a resistance of 5000 Ω and the non-linear error of the system must not exceed 2% of full scale. Find: (a) (b)
the maximum potentiometer sensitivity that can be obtained; the required potentiometer resistance and supply voltage in order to achieve maximum sensitivity.
5.3
An electronic differential transmitter gives a current output of 4 to 20 mA linearly related to a differential pressure input of 0 to 104 Pa. The Norton impedance of the transmitter is 105 Ω. The transmitter is connected to an indicator of impedance 250 Ω via a cable of total resistance 500 Ω. The indicator gives a reading between 0 and 104 Pa for an input voltage between 1 and 5 V. Calculate the system measurement error, due to loading, for an input pressure of 5 × 103 Pa.
5.4
A sensor with mass 0.1 kg, stiffness 103 N m−1 and damping constant 10 N s m−1 is used to measure the force on a mechanical structure of mass 5 kg, stiffness 102 N m−1 and damping constant 20 N s m−1. Find the transfer function relating measured and actual changes in force.
PROBLEMS
95
Basic problems 5.5
A linear thermocouple with a sensitivity of 0.04 mV/°C and resistance of 100 Ω is connected to a load with a resistance of 1 kΩ. Find the voltage across the load for a temperature of 250 °C.
5.6
A potentiometer displacement sensor has a supply voltage of 15 V and a resistance of 50 kΩ. The fractional displacement of the wiper is 0.3. Find the Thévenin voltage and resistance for the circuit.
5.7
A potentiometer has a supply voltage of 10 V, a resistance of 10 kΩ and a length of 10 cm. A recorder of resistance 10 kΩ is connected across the potentiometer. Calculate the Thévenin equivalent circuit for the sensor and the recorder voltage for each of the following displacements: (a) (b) (c)
5.8
A potentiometer has a total length of 10 cm and a resistance of 100 Ω. (a) (b) (c)
5.9
2 cm 5 cm 8 cm.
Calculate the supply voltage so that power dissipation = 1 W. Draw the Thévenin equivalent circuit for 7 cm displacement. The potentiometer is connected to a recorder with a resistance RL. Find RL such that the recorder voltage is 5% less than the open circuit voltage at 7 cm displacement.
A pressure transducer consists of a Bourdon tube elastic element connected to a potentiometer displacement sensor. The input range of the Bourdon tube is 0 to 104 Pa and the output range is 0 to 1 cm. The potentiometer has a length of 1 cm, a resistance of 10 kΩ and a supply voltage of 10 V. If the input pressure is 5 × 103 Pa, calculate: (a) (b) (c)
the displacement of the potentiometer wiper (assume a linear Bourdon tube) the open circuit transducer output voltage the voltage indicated by a voltmeter of resistance 10 kΩ connected across the potentiometer.
6 Signals and Noise in Measurement Systems 6.1
Introduction In Chapter 4 we studied the dynamic response of measurement systems to step, sine wave and square wave input signals. These signals are examples of deterministic signals: a deterministic signal is one whose value at any future time can be exactly predicted. Thus if we record these signals for an observation period TO (Figure 6.1), the future behaviour of the signal, once the observation period is over, is known exactly. The future behaviour of real processes, such as chemical reactors, blast furnaces and aircraft, will depend on unknown factors such as the type of feedstock, reliability of equipment, changes in throughput and atmospheric conditions, and cannot be known in advance. This means that the future value of measured variables, such as reactor temperature, flow in a pipe and aircraft speed, cannot be exactly predicted. Thus in real measurement applications the input signal to the measurement system is not deterministic but random. If a random signal is recorded for an observation period TO (Figure 6.1) the behaviour of the signal, once the observation period is over, is not known exactly. However, five statistical quantities – mean, standard deviation, probability density function, power spectral density and autocorrelation function – are used to estimate the behaviour of random signals. These are defined and explained in Section 6.2. Random variations in the measured input variable produce corresponding random variations in the electrical output signals of elements in the system. More precisely the Thévenin voltage and Norton current signals defined in Chapter 5 vary randomly with time and will be referred to as the measurement signal. Examples are random fluctuations in the millivolt output of a thermocouple, random fluctuations in the current output of a differential pressure transmitter, and random fluctuations in the amplitude and frequency of the a.c. output voltage of a variable reluctance tachogenerator. However, unwanted electrical signals may also be present in the measurement circuit. These may be due to sources inside the measurement circuit or caused by coupling to sources outside the circuit. The magnitude of the unwanted signal may be comparable to or larger than that of the measurement signal itself, resulting in a measurement error for the overall system. This is an example of the interfering inputs discussed in Chapter 2. The unwanted signal then may be either random, e.g. signals caused by the random motion of electrons, or deterministic, e.g. sinusoidal signals at 50 Hz caused by power cables. Unwanted random signals are usually referred to as noise signals and unwanted deterministic signals as interference signals.
98 S IG NA LS AND NOISE IN MEASUREMEN T S Y S T E MS Figure 6.1 Deterministic and random signals.
Sections 6.3 and 6.4 discuss the sources of noise and interference signals and how they affect the measurement circuit. Section 6.5 examines ways of reducing the effects of noise and interference.
6.2
Statistical representation of random signals Figure 6.2 shows a recording of a section of a random signal obtained during an observation period TO. Since the signal is random we cannot write down a continuous algebraic equation y(t) for the signal voltage y at time t. We can, however, write down the values y1 to yN of N samples taken at equal intervals ∆T during TO . The first sample y1 is taken at t = ∆T, the second y2 is taken at t = 2∆T, and the ith yi is taken at t = i∆T, where i = 1, . . . , N. The sampling interval ∆T = TO /N must satisfy the Nyquist sampling theorem, which is explained in Section 10.1. We can now use these samples to calculate statistical quantities for the observed section of the signal. These observed statistical quantities will provide a good estimate of the future behaviour of the signal, once the observation period is over, provided:
Figure 6.2 Sampling of a random signal.
6.2 S T AT I S TI CAL R E PRE S E N T AT I O N O F R AN D O M S I GN AL S
(a) (b)
6.2.1
99
TO is sufficiently long, i.e. N is sufficiently large; the signal is stationary, i.e. long-term statistical quantities do not change with time.
Mean æ For a signal defined in terms of a continuous function y(t) in the interval 0 to TO , the mean is given by:
Mean for continuous signal
R=
0002
TO
1 TO
y(t ) dt
[6.1]
0
If, however, the signal is represented by the set of sampled values yi we have Mean for sampled signal
6.2.2
R=
1 N
i=N
∑y
[6.2]
i
i =1
Standard deviation σ This is a measure of the average spread or deviation of the signal from the mean value R. In the continuous case:
Standard deviation for continuous signal
σ2 =
1 TO
0002
TO
[ y(t ) − R]2 dt
[6.3]
0
and in the sampled case: Standard deviation for sampled signal
σ2 =
1 N
i=N
∑(y
i
− R )2
[6.4]
i =1
In the special case that R = 0, the standard deviation σ is equal to the root mean square (r.m.s.) value yrms, where: yrms =
1 TO
0002
TO
y 2 dt
[6.5]
0
or yrms =
1 N
i=N
∑y i =1
2 i
[6.6]
100 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS Figure 6.3 Probability and probability density.
6.2.3
Probability density function p( y) This is a function of signal value y and is a measure of the probability that the signal will have a certain range of values. Figure 6.3 shows the set of sample values yi and the y axis divided into m sections each of width ∆y. We can then count the number of samples occurring within each section, i.e. n1 in section 1, n2 in section 2, nj in section j, etc., where j = 1, . . . , m. The probability Pj of the signal occurring in the jth section is thus: Pj = =
number of times sample occurs in the jth section total number of samples nj , N
[6.7]
j = 1, . . . , m
The cumulative probability Cj is the total probability that the signal will occur in the first j sections and is given by:
6.2 S T AT I S TI CAL R E PRE S E N T AT I O N O F R AN D O M S I GN ALS
Cumulative probability
101
C j = P1 + P2 + . . . + Pj =
1 N
(n1 + n2 + . . . + n j )
[6.8]
Figure 6.3 shows the corresponding forms of Pj and Cj. The final value of Cj when j = m is: Cm = =
1 (n1 + n2 + . . . + nm) N 1 × N = 1 (since n1 + n2 + . . . + nm = N ) N
[6.9]
i.e. the total probability of finding the signal in all m sections is unity. In the limit as the interval ∆y tends to zero the discrete cumulative probability Cj tends to a continuous function (Figure 6.3). This is the cumulative probability distribution function (c.d.f.) P( y), which is defined by: Cumulative probability distribution function
P( y ) = lim C j ∆ y →0
[6.10]
The probability density function (p.d.f.) p( y) is more commonly used and is the derivative of P( y), i.e. Probability density function
p( y ) =
dP dy
[6.11]
Thus the probability Py,y+∆y that the signal will lie between y and y + ∆y is given by: Py, y +∆ y = ∆P = p( y)∆y
[6.12]
i.e. by the area of a strip of height p( y) and width ∆y. Similarly the probability Py1, y2 that the signal will lie between y1 and y2 is given by: Py1, y2 =
0002
y2
p( y) dy
[6.13]
y1
i.e. the area under the probability density curve between y1 and y2. The total area under the probability density curve is equal to unity, corresponding to the total probability of the signal having any value of y. The normal probability density function (Section 2.3), i.e. p ( y) =
 ( y − R )2  exp −  2σ 2  σ 2π  1
[6.14]
usually provides an adequate description of the amplitude distribution of random noise signals.
102 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS
6.2.4
Power spectral density φ (ω ) If we record a random signal for several observation periods, each of length TO (Figure 6.4), the waveform will be different for each period. However, the average signal power will be approximately the same for each observation period. This means that signal power is a stationary quantity which can be used to quantify random signals. In Section 4.3 we saw that a periodic signal can be expressed as a Fourier series, i.e. a sum of sine and cosine waves with frequencies which are harmonics of the fundamental frequency. The power in a periodic signal is therefore distributed amongst these harmonic frequencies. A random signal is not periodic and cannot be represented by Fourier series but does contain a large number of closely spaced frequencies. Power spectral density is a stationary quantity which is a measure of how the power in a random signal is distributed amongst these different frequencies. In order to explain the meaning of φ (ω) we approximate the random signal by a periodic signal (Figure 6.4) in which the waveform recorded during the first observation period TO is exactly repeated during each subsequent observation period. This periodic approximation is valid: (a) (b)
Figure 6.4 Power spectrum and power spectral density.
in the limit that the observation period, i.e. the signal period TO, becomes large, provided we use it to calculate the power distribution in the signal.
6.2 S T AT I S TI CAL R E PRE S E N T AT I O N O F R AN D O M S I GN ALS
103
The Fourier series for a voltage signal with period TO is y(t) = a0 +
n ∞
n ∞
n 1
n 1
∑ an cos nω 1 t +
∑b
n
sin nω 1 t
[4.51a]
where the fundamental frequency ω1 = 2π/TO , a0 = mean R, and Fourier series coefficients for sampled signal
an =
i=N
2
∑ yi cos (nω1i∆T ) =
N
i =1
2 N
i= N
∑y
i
i =1
 i cos  2πn  N 
 i bn = yi sin (nω1i∆T ) = yi sin  2πn  ∑ ∑ N i =1 N i =1 N  i= N
2
2
i= N
[6.15]
Equations [6.15] use the set of sample values y1 and are equivalent to eqns [4.52] for the Fourier coefficients of a continuous signal. If the nth harmonic an cos nω 1 t is applied across a 1 Ω resistor, the instantaneous power in the resistor at time t is a 2n cos 2nω 1 t watts, and the average power over period TO is: a 2n
1 TO
0002
TO
cos2nω 1 t dt =
0
a 2n 2
[6.16]
Similarly the average power in a 1 Ω resistor due to bn sin nω 1 t is b 2n /2, so that the power due to the nth harmonic at frequency nω 1 is: Power due to nth harmonic
wn = –12 (a 2n + b 2n )
[6.17]
Figure 6.4 shows that the relationship between wn and ω is a series of lines at the harmonic frequencies nω 1. This is referred to as the power spectrum of the signal and is terminated at ω MAX, the harmonic frequency beyond which wn becomes negligible. The cumulative power Wn is the total power in a 1 Ω resistor due to the first n harmonics and the d.c. component a0, i.e. Cumulative power
Wn = w0 + w1 + w2 + . . . + wn
[6.18]
The diagram shows the relation between Wn and ω ; it is in the form of a staircase, each step having width ω1. At ω = ω MAX, Wn = WTOT, the total power in the signal. However, in the limit that signal period TO → ∞, ω 1 → 0 and Wn becomes a continuous function of ω. This is the cumulative power function (c.p.f.) W(ω), which is defined by: Cumulative power function
W (ω ) = lim Wn ω 1→ 0
[6.19]
The power spectral density (PSD) φ (ω) is more commonly used and is the derivative of W(ω) (Figure 6.4), i.e.
104 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS Power spectral density
φ (ω ) =
dW dω
[6.20]
The PSD φ has units watt s rad−1 or W/Hz. Thus the power produced in a 1 Ω resistor due to frequencies between ω and ω + ∆ω is: Wω,ω +∆ω = ∆W = φ (ω)∆ω watts
[6.21]
i.e. the area of a strip of height φ (ω) and width ∆ω. Similarly the power due to frequencies between ω 1 and ω 2 is: Wω 1, ω 2 =
0002
ω2
φ (ω) dω watts
[6.22]
ω1
i.e. the area under the power spectral density curve between ω 1 and ω 2. The total area under the PSD curve is the total power in the signal, i.e. ω MAX
WTOT =
0002
φ (ω) dω watts
[6.23]
0
Internal noise sources in electrical circuits can often be regarded as white noise, which has a uniform PSD over an infinite range of frequencies, i.e.
φ (ω) = A, 0 ≤ ω ≤ ∞
[6.24]
Another useful representation for both noise and measurement signals is a power spectral density which is constant up to a cut-off frequency ωC and zero for higher frequencies (band limited white noise):
φ (ω) =
!A, 0 ≤ ω ≤ ωC @0, ω > ωC
[6.25]
i.e. WTOT = AωC . The maximum frequency ω MAX for a measurement signal depends on the nature of the measured variable. Thus the vibration displacement of part of a machine may contain frequencies up to many kHz, whereas the temperature variations in a chemical reactor may only contain frequencies up to 0.01 Hz. In order to accurately measure random fluctuations in a measured variable, the transfer function G(s) for the measurement system must satisfy the conditions of Section 4.4, i.e. |G( jω)| = 1 and arg G( jω) = 0 for all ω up to ω MAX.
6.2.5
Autocorrelation function Figure 6.5 shows a block diagram of a simple correlator. The input signal y(t) is passed through a variable time delay unit to give a delayed signal y(t − β ). The signals y(t) and y(t − β ) are multiplied together to give the product waveform y(t)y(t − β ). This product waveform is passed through an averager and the average value y(t)y(t − β ) is displayed on a meter: this is the autocorrelation coefficient R yy. If the time delay β is altered, the shape of the product waveform changes, causing the meter reading R yy to change. The relationship between R yy and time delay β is the autocorrelation function R yy ( β ) of the signal. We note that R yy ( β ) has a maximum value
6.2 S T AT I S TI CAL R E PRE S E N T AT I O N O F R AN D O M S I GN ALS
105
Figure 6.5 Autocorrelation and evaluation of autocorrelation coefficient Ryy(3∆T).
R yy (0) when β = 0; this is because the corresponding product waveform is y 2(t), which is always positive and has a maximum average value. If the signal is defined by a continuous function y(t) in the interval 0 to TO, then R yy( β ) can be evaluated using: Autocorrelation function of a continuous signal
Ryy (β ) = lim
TO →∞
1 TO
0002
TO
y(t ) y(t − β ) dt
[6.26]
0
Thus, if y(t) = b sin(ω 1 t + φ), R yy( β ) = b2 lim
TO →∞
= b lim 2
TO →∞
1 TO
0002
sin(ω 1 t + φ)sin[ω 1(t − β ) + φ ] dt
1 TO
0002
–12 {cos ω 1 β − cos[ω 1(2t − β ) + 2φ]} dt
TO
0
TO
0
2
=
b 1 cos ω 1 β lim T →∞ 2 TO O
=
b2 cos ω 1 β 2
0002
TO
0
dt −
b2 1 lim T →∞ 2 TO O
0002
TO
cos[ω 1(2t − β ) + 2φ] dt
0
[6.27]
since the average value of cos[ω1(2t − β ) + 2φ ] is zero. Thus the autocorrelation function of a sinusoidal signal is a cosine function of the same frequency, but the phase information φ in the sine wave is lost. The autocorrelation function of any periodic signal has the same period as the signal itself (see Problem 6.2).
106 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS A random signal is usually characterised by a set of N sample values yi. Since information is only available at discrete time intervals, the time delay β is normally an integer multiple of the sampling interval ∆T, i.e.
β = m∆T, m = 0, 1, 2 . . .
[6.28]
In this case the autocorrelation function of the signal is found by evaluating the set of autocorrelation coefficients R yy (m∆T ), which are given by: Autocorrelation coefficient for sampled signal
Ryy (m∆T ) = lim
N →∞
1 N
i=N
∑yy
i i −m
[6.29]
i =1
where yi is the sample value at time i∆T and yi− m the value at time (i − m)∆T (m sampling intervals earlier). Figure 6.5 shows the evaluation of Ryy(3∆T ) for a sampled waveform; this involves calculating the products y1 y−2, y2 y−1, y3 y0, . . . yN yN−3, summing, and dividing by N. The autocorrelation function of a random signal can also be found from the power spectral density φ (ω). To illustrate this we first consider a periodic signal which is the sum of three harmonics: y(t) = b1 sin ω 1 t + b2 sin 2ω 1 t + b3 sin 3ω 1 t Using eqn [6.27] the autocorrelation function is: R yy ( β ) =
b 21 b 22 b 23 cos ω 1 β + cos 2ω 1 β + cos 3ω 1 β 2 2 2
which has period 2π/ω 1 (Figure 6.6). Using eqn [6.17], the power spectrum of the signal consists of three lines at frequencies ω 1, 2ω 1 and 3ω 1 with heights b 21 /2, b 22 /2 and b 23 /2 respectively. Thus the power spectrum can be obtained from the autocorrelation function by Fourier analysis. Similarly the autocorrelation function can be obtained from the power spectrum by adding the harmonics together as in Fourier synthesis. For random signals, Ryy( β ) and φ (ω) are related by the Fourier transform or Wiener–Khinchin relations: Figure 6.6 Relationships between power spectrum and autocorrelation function for periodic and random signals.
6.3 EFFECTS O F N O I S E AN D I N TE RF E RE N CE O N ME AS URE ME N T CI R CUI TS
R yy ( β ) =
0002
∞
φ (ω) cos ωβ dω 0
0002
2 φ ( ω) = π
107
[6.30]
∞
R yy ( β ) cos ωβ dβ 0
Thus for a signal with φ (ω) constant up to ωC and zero for higher frequencies we have: ωC
Ryy( β ) =
0002
ω
A cos ωβ dω = A
0
sin ω C β G sin ωβ J C =A I β L0 β
[6.31]
The form of both functions is shown in Figure 6.6; we see that R yy ( β ) has its first zero crossings at β = ±π/ωC , i.e. the width of the central ‘spike’ is 2π/ωC. Thus a rapidly varying random signal has a high value of ωC, i.e. a broad power spectrum but a narrow autocorrelation function, that falls off sharply as β is increased. A slowly varying random signal, however, has a low value of ωC, i.e. a narrow power spectrum but a broad autocorrelation function that falls slowly as β is increased.
6.2.6
Summary In order to specify a random signal we need to know: !either probability density function # to specify amplitude @or mean and standard deviation $ behaviour and !either power spectral density # to specify frequency/time @or autocorrelation function $ behaviour Important relations between these different quantities can be derived by considering Ryy(0), the autocorrelation coefficient at zero time delay. From eqns [6.26] and [6.30] we have R yy (0) = lim
TO →∞
1 TO
0002
TO
y dt =
0
2
0002
∞
φ (ω) dω
[6.32]
0
2 From eqn [6.5], the first expression is equal to y RMS in the limit of infinitely long observation time TO. From equation [6.23] the second expression is equal to WTOT, the total power produced by the signal in a 1 Ω resistor. Thus:
R yy (0) = y 2RMS = WTOT
6.3
[6.33]
Effects of noise and interference on measurement circuits In Section 5.1 we saw that the interconnection of two measurement system elements, e.g. a thermocouple and an amplifier, or a differential pressure transmitter and a recorder, could be represented by an equivalent circuit in which either a Thévenin voltage
108 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS Figure 6.7 Effects of interference on measurement circuit: (a) Voltage transmission – series mode interference (b) Current transmission – series mode interference (c) Voltage transmission – common mode interference.
source or a Norton current source is connected to a load. In industrial installations, source and load may be typically 100 metres apart and noise and/or interference voltages may also be present. Figure 6.7(a) shows a voltage transmission system subject to series mode interference; here a noise or interference voltage VSM is in series with the measurement signal voltage ETh. The current i through the load is: i=
ETh + VSM ZTh + R C + Z L
and the corresponding voltage across the load is: VL =
ZL (ETh + VSM ) ZTh + R C + Z L
[6.34]
Normally we make ZL 0002 RC + ZTh to obtain maximum voltage transfer to the load (Section 5.1.1); under these conditions eqn [6.34] becomes: VL ≈ ETh + VSM
[6.35]
This means that with a voltage transmission system all of VSM is across the load; this affects the next element in the system and possibly results in a system measurement error. We define signal-to-noise or signal to interference ratio S/N in decibels by:
6.3 EFFECTS O F N O I S E AN D I N TE RF E RE N CE O N ME AS URE ME N T CI R CUI TS
Signal-to-noise ratio
W  E  S = 20 log10  Th  = 10 log10  S  dB N  WN   VSM 
109
[6.36]
where ETh and VSM are the r.m.s. values of the voltages, and WS and WN are the corresponding total signal and noise powers. Thus if ETh = 1 V and VSM = 0.1 V, S/N = +20 dB. Figure 6.7(b) shows a current transmission system subject to the same series mode interference voltage VSM. The Norton source current iN divides into two parts, one part through the source impedance ZN, the other part through ZL. Using the current divider rule, the current through the load due to the source is: i = iN
ZN ZN + R C + Z L
In addition there is an interference current iSM =
VSM ZN + R C + Z L
through the load due to the interference voltage. The total voltage across the load is therefore: VL = iZL + iSM ZL = iN ZL ⋅
ZN ZL + VSM ⋅ ZN + R C + Z L ZN + R C + Z L
[6.37]
Normally we make RC + ZL 0003 ZN to obtain maximum current transfer to the load (Section 5.1.3); under these conditions eqn [6.37] becomes: VL ≈ iN ZL +
ZL VSM ZN
[6.38]
Since ZL /ZN 0003 1, this means that with a current transmission system only a small fraction of VSM is across the load. Thus a current transmission system has far greater inherent immunity to series mode interference than a voltage transmission system. In a thermocouple temperature measurement system, therefore, it may be better to convert the thermocouple millivolt e.m.f. into a current signal (Section 9.4.1) prior to transmission, rather than transmit the e.m.f. directly. Figure 6.7(c) shows a voltage transmission system subject to common mode interference in which the potentials of both sides of the signal circuit are raised by VCM relative to a common earth plane. If, as above, ZL 0002 RC + ZTh , then current i → 0 so that the potential drops iRC /2, etc., can be neglected. Under these conditions: Potential at B = VCM Potential at A = VCM + ETh and VL = VB − VA = ETh This means that the voltage across the load is unaffected by VCM; there is, however, the possibility of conversion of a common mode voltage to series mode.
110 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS
6.4 6.4.1
Noise sources and coupling mechanisms Internal noise sources The random, temperature-induced motion of electrons and other charge carriers in resistors and semiconductors gives rise to a corresponding random voltage which is called thermal or Johnson noise. This has a power spectral density which is uniform over an infinite range of frequencies (white noise) but proportional to the absolute temperature θ K of the conductor, i.e.
φ = 4Rkθ W/Hz
[6.39]
where RΩ is the resistance of the conductor and k is the Boltzmann constant = 1.4 × 10−23 J K−1. From eqn [6.22] the total thermal noise power between frequencies f 1 and f 2 is: W=
0002
f2
4Rkθ df = 4Rkθ ( f 2 − f 1) W
[6.40]
f1
and from [6.33] the corresponding r.m.s. voltage is: VRMS = W = 4Rkθ ( f 2 − f1 ) V
[6.41]
Thus if R = 106 Ω, f 2 − f 1 = 106 Hz and θ = 300 K, VRMS = 130 µV and is therefore comparable with low-level measurement signals such as the output from a strain gauge bridge. A similar type of noise is called shot noise; this occurs in transistors and is due to random fluctuations in the rate at which carriers diffuse across a junction. This is again characterised by a uniform power spectral density over a wide range of frequencies.
6.4.2
External noise and interference sources[1] The most common sources of external interference are nearby a.c. power circuits which usually operate at 240 V, 50 Hz. These can produce corresponding sinusoidal interference signals in the measurement circuit, referred to as mains pick-up or hum. Power distribution lines and heavy rotating machines such as turbines and generators can cause serious interference. D.C. power circuits are less likely to cause interference because d.c. voltages are not coupled capacitively and inductively to the measurement circuit. However, switching often occurs in both a.c. and d.c. power circuits when equipment such as motors and turbines is being taken off line or brought back on line. This causes sudden large changes in power, i.e. steps and pulses, which can produce corresponding transients in the measurement circuit. The air in the vicinity of high voltage power circuits can become ionised and a corona discharge results. Corona discharge from d.c. circuits can result in random noise in the measurement circuit and that from a.c. circuits results in sinusoidal interference at the power frequency or its second harmonic. Fluorescent lighting is another common interference source; arcing occurs twice per cycle so that most of the interference is at twice the power frequency.
6.4 NOISE SOURCES AND COUPLING MECHANISMS
111
Radio-frequency transmitters, welding equipment and electric arc furnaces can produce r.f. interference at frequencies of several MHz.
6.4.3
Coupling mechanisms to external sources Inductive coupling[2] Figure 6.8(a) shows inductive or electromagnetic coupling between the measurement circuit and a nearby power circuit. If the circuits are sufficiently close together, then there may be a significant mutual inductance M between them. This means that an alternating current i in the power circuit induces a series mode interference voltage in the measurement circuit of magnitude VSM = M
di dt
[6.42]
Thus if M ≈ 1 µH and di/dt ≈ 103 A s−1 (possible in a 1 horsepower motor) then VSM ≈ 1 mV, which can be comparable with the measurement signal. The mutual inductance M depends on the geometry of the two circuits, namely on the overlapping length and separation, but is distributed over the entire length of the circuits rather than the ‘lumped’ equivalent value shown in Figure 6.8. Inductive coupling will occur even if the measurement circuit is completely isolated from earth. Figure 6.8 Coupling mechanisms to external sources: (a) Electromagnetic coupling (b) Electrostatic coupling (c) Multiple earths.
112 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS
Capacitive coupling[1,2] Another important coupling mechanism is capacitive or electrostatic coupling, which is illustrated in Figure 6.8(b). The diagram shows the measurement circuit close to an a.c. power cable which is at a potential of 240 V (r.m.s.) relative to the earth plane. The power cable, earth plane and signal leads are all conductors, so that there may be some capacitance between the power cable and the signal leads and between the signal leads and the earth plane. These capacitances will be distributed over the entire length of the measurement circuit, but are represented by ‘lumped’ equivalents. C1 and C2 are the capacitances between the power cable and signal leads, and C1E and C2E the capacitances between the signal lead and the earth plane; all four capacitances will be proportional to the length of the measurement circuit, which could be tens of metres in an industrial installation. Ignoring the measurement signal voltage ETh for the moment, the potentials at B and E are determined by the potential dividers ABC and DEF:   1/( jω C1E ) C1 VB = 240   = 240 ω ω 1 /( ) 1 /( ) j C + j C C + C1E  1E 1  1    1/( jω C2 E ) C2 VE = 240   = 240 ω ω 1 /( ) 1 /( ) j C + j C C  2E 2  2 + C2 E 
[6.43]
Thus we have a common mode interference voltage VCM = VE and a series mode interference voltage:   C1 C2 VSM = VB − VE = 240  −   C1 + C1 E C2 + C2 E 
[6.44]
Thus series mode interference is zero only if there is perfect balance between the coupling capacitances, i.e. C1 = C2 and C1E = C2E; in practice small imbalances are present due to slightly different distances between each signal lead and the power cable/earth plane.
Multiple earths[1] The above explanation assumes an earth plane having a potential of 0 volts at every point on its surface. Heavy electrical equipment can, however, cause currents to flow through the earth, causing different potentials at different points. If the measurement circuit is completely isolated from the earth plane there is no problem. In practice, however, there may be a leakage path connecting the signal source to one earth point and another leakage path connecting the recorder or indicator to a different earth point, some distance away. If the two earth points are at different potentials, then common and series mode interference voltages are produced in the measurement circuit. Figure 6.8(c) illustrates the general problem of multiple earths. The measurement signal source ETh is connected via a resistive cable to a receiver represented by a resistive load RL. Provided RL 0002 RC + RTh , the current flow in PQRS is negligible and VL ≈ ETh, provided the circuit is completely isolated from earth. However, leakage paths ZSE and ZRE exist between source/source earth and receiver/receiver earth. If
6.5 ME TH O D S O F R E D UCI N G E F F E CT S O F N O I S E AN D I N T E R F E R E N CE
113
VE is the difference in potential between source earth and receiver earth, then a current iE flows in the circuit UPST, given by: iE =
VE ZE + ZSE + (RC /2) + ZRE
[6.45]
Thus Potential at P = VE − iE (ZE + ZSE) Potential at Q, R = VE − iE (ZE + ZSE) + ETh Potential at S = iE ZRE Thus there is a common mode interference voltage: VCM = VS = VE
ZRE ZE + ZSE + (RC /2) + ZRE
[6.46]
To find the series mode interference voltage we need to calculate the voltage across RL: VL = VR − VS = VE − iE (ZE + ZSE + ZRE) + ETh = ETh + iE RC /2 i.e. there is a series mode interference voltage: VSM = VE
RC /2 ZE + ZSE + (RC /2) + ZRE
[6.47]
Ideally we require both ZSE and ZRE to be as large as possible in order to minimise iE and VSM ; this, however, is not always possible in an industrial application. A common example is a thermocouple installation where in order to achieve as good a speed of response as possible the tip of the thermocouple touches the thermowell or sheath (Section 14.2). The thermowell is itself bolted to a metal vessel or pipe which is in turn connected to one point in the earth plane. Thus ZSE will be very small, say ZSE = 10 Ω (resistive), so that the receiver must be isolated from earth to minimise VSM . Taking ZE = 1 Ω, RC /2 = 10 Ω, VE = 1 V and ZRE = 106 Ω, we have: VSM = 1 ×
10 V ≈ 10 µV 1 + 10 + 10 + 106
The worst case is when the receiver is also directly connected to earth, i.e. ZRE ≈ 0, giving VSM = 0.48 V. Thus if the measurement circuit must be connected to earth, the connection must be made at one point only.
6.5 6.5.1
Methods of reducing effects of noise and interference Physical separation Since mutual inductances and coupling capacitances between measurement and power circuits are inversely proportional to the distance between them, this distance should be as large as possible.
114 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS Figure 6.9 Reduction of electromagnetic coupling by twisted pairs.
6.5.2
Electromagnetic shielding The simplest way of reducing the effects of inductive coupling to an external interference source is shown in Figure 6.9. The two conductors A and B of the measurement circuit are twisted into loops of approximately equal area. This arrangement is commonly known as twisted pairs and is explained in ref. [2]. The magnitude of the interference voltage induced in a given loop is proportional to the area of the loop and the rate of change of the external magnetic field. The sign of the induced voltage depends on the orientation of conductors A and B. Thus if a voltage VX Y is induced in the j th loop between points X and Y, then an opposing voltage VY Z is induced in the ( j + 1)th loop between Y and Z. In the ideal case of both loops having the same area and experiencing the same magnetic fields, |VXY | = |VYZ |, i.e. there is a zero resultant induced voltage between X and Z. This process is repeated for the whole length of the twisted pair, giving a reduced overall interference voltage.
6.5.3
Electrostatic screening and shielding The best method of avoiding the problem of capacitive coupling to a power circuit (Section 6.4.3) is to enclose the entire measurement circuit in an earthed metal screen or shield. Figure 6.10(a) shows the ideal arrangement; the screen is connected directly to earth at a single point, at either the source or the receiver. There is no direct connection between the screen and the measurement circuit, only high impedance leakage paths via the small (screen/measurement circuit) capacitances CSM. The screen provides a low impedance path to earth for the interfering currents i; the currents through CSM and CE are small, thus reducing series and common mode interference. The above ideal of the measurement circuit completely insulated from the screen and the screen earthed at one point only may be difficult to achieve in practice for the following reasons: (a)
(b)
(c)
The signal source may be directly connected to a local earth point via the structure on which it is mounted; an example is the thermocouple installation mentioned in Section 6.4.3. The receiver may be directly connected to a local earth; an example is in a computer-based system where the receiver must be directly connected to the computer earth. There may be indirect connections via leakage impedances.
Figure 6.10(b) illustrates the general problem. The measurement circuit PQRS is connected to the screen (impedance ZS) via source/screen impedance ZSS and receiver/ screen impedance ZRS . The screen is connected to earth point U at the source end via ZSU and to earth point T at the receiver end via ZST. The measurement circuit can be affected by interference voltages from both VE (potential difference between U and T) and nearby power circuits.
6.5 ME TH O D S O F R E D UCI N G E F F E CT S O F N O I S E AN D I N T E R F E R E N CE
115
Figure 6.10 Reduction of electrostatic coupling using screening: (a) Ideal arrangement (b) Practical equivalent circuit.
Analysis of circuits UXYT and XPSY using Kirchhoff’s laws gives: UXYT XPSY
VE = i1ZE + i1ZSU + (i1 − i2)ZS + i1ZST 0 = −(i1 − i2)ZS + i2ZSS + i2RC /2 + i2ZRS
[6.48] [6.49]
Solution of these equations gives: i1 = i2 =
(ZE + ZSU
(ZS + ZSS + RC /2 + ZRS )VE + ZS + ZST )(ZS + ZSS + RC /2 + ZRS ) − Z S2
[6.50]
(ZE + ZSU
ZS VE + ZS + ZST )(ZS + ZSS + RC /2 + ZRS ) − Z S2
[6.51]
The series mode interference voltage in the measurement circuit PQRS is the voltage drop across PS, i.e.
116 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS
VSM = i2RC /2 =
(ZE + ZSU
ZS (RC /2)VE + ZS + ZST )(ZS + ZSS + RC /2 + ZRS ) − Z S2
[6.52]
The common mode interference voltage is the voltage drop across ST, i.e. VCM = VST = VSY + VYT = i2ZRS + i1ZST =
[6.53]
[ZRS ZS + ZST (ZS + ZSS + RC /2 + ZRS )]VE (ZE + ZSU + ZS + ZST )(ZS + ZSS + RC /2 + ZRS ) − Z S2
To minimise VSM and VCM, we want the product term: (ZE + ZSU + ZS + ZST)(ZS + ZSS + RC /2 + ZRS ) to be large. Since in practice the earth impedance ZE, screen impedance ZS and cable resistance RC /2 are all small, the above condition reduces to: (ZSU + ZST) (ZSS + ZRS) to be large
[6.54]
However, we cannot have both ZSU and ZST large; either ZST or ZSU must be small, otherwise the screen will not be earthed and there is therefore no low impedance path to earth for the capacitively coupled interference currents. Condition (6.54) therefore reduces to the two conditions: ZSU (ZSS + ZRS ) = HIGH; ZST = LOW OR
[6.55] ZST (ZSS + ZRS ) = HIGH; ZSU = LOW
which are satisfied if: ZSU = HIGH AND (ZSS OR ZRS OR both = HIGH), ZST = LOW OR
[6.56] ZST = HIGH AND (ZSS OR ZRS OR both = HIGH), ZSU = LOW
As mentioned above, in many practical situations it may be impossible to have the measurement circuit completely isolated from the screen, i.e. both ZSS and ZRS high. In this situation, possible confusion is avoided if ZSU and ZSS are both high, or ZST and ZRS are both high. In this situation the conditions become: ZSU = HIGH AND ZSS = HIGH, ZRS = LOW AND ZST = LOW (Isolated source) OR
[6.57] ZST = HIGH AND ZRS = HIGH, ZSS = LOW AND ZSU = LOW (Isolated receiver)
6.5.4
Use of differential amplifiers Common mode interference voltages can be successfully rejected by the use of a differential amplifier (Figure 6.11 and Section 9.2). An ideal differential amplifier has an output: VOUT =
RF RF (V2 − V1) = − ETh R1 R1
[6.58]
6.5 ME TH O D S O F R E D UCI N G E F F E CT S O F N O I S E AN D I N T E R F E R E N CE
117
Figure 6.11 Use of differential amplifier.
i.e. only the sensor voltage ETh is amplified. The output of a practical amplifier (Section 9.2.2) contains a contribution proportional to VCM; from eqn [9.39] we have: VOUT = −
 R  V RF ETh + 1 + F  CM R1  CMRR R1 
[6.59]
The common mode rejection ratio (CMRR) of the amplifier is the ratio of differential voltage gain to common mode voltage gain and should be as large as possible to minimise this effect. Thus if we have ETh = 1 mV, R1 = 1 kΩ, RF = 1 MΩ, VCM = 1 V and CMRR = 105 (100 dB) then: VOUT ≈ −1.0 + 0.01 V i.e. the resultant series mode interference is only 1%.
6.5.5
Filtering A filter is an element which transmits a certain range (or ranges) of frequencies and rejects all other frequencies. An analogue filter is an electrical network, consisting usually of resistors, capacitors and operational amplifiers, which conditions continuous signals. A digital filter is usually a digital computer programmed to process sampled values of a signal (Chapter 10). Provided that the power spectrum of the measurement signal occupies a different frequency range from that of the noise or interference signal, then filtering improves the signal-to-noise ratio. Figures 6.12(a)–(d) show the use of low pass, high pass, band pass and band stop filters in rejecting noise. In all cases the filter transmits the measurement signal but rejects the noise signal, which occupies a different frequency range. The diagrams show the amplitude ratio |G( jω)| for each filter and the power spectral densities φ (ω) for signal and noise. In order to transmit the measurement signal without distortion the transfer function G(s) of the filter must, ideally, satisfy the conditions of Section 4.4, i.e. that | G( jω)| = 1 and argG( jω) = 0 for all the frequencies present in the measurement signal spectrum. Analogue filtering can be implemented at the signal conditioning stage; the a.c. amplifier of Figure 9.12 is an example of a band pass filter. Digital filtering can be implemented at the signal processing stage (Chapter 10). If, however, measurement signal and noise spectra overlap, filtering has limited value. Figure 6.13(a) shows a measurement signal affected by wide band noise; a low-pass filter with bandwidth matched to the signal spectrum removes as much of the noise as possible, but the noise inside the filter bandwidth still remains. Figure 6.13(b)
118 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS Figure 6.12 Use of filtering to reject noise.
shows the measurement signal affected by 50 Hz interference; a narrow band-stop filter centred on 50 Hz rejects the interference but also rejects measurement signal frequencies around 50 Hz and causes amplitude and phase distortion over a wider range of frequencies.
6.5.6
Modulation The problem of Figure 6.13(b) can be solved by modulating the measurement signal onto a higher frequency carrier signal, e.g. a 5 kHz sine wave as shown in Figure 6.13(c). The simplest form of modulation is amplitude modulation; this involves the multiplication of measurement and carrier signals and is discussed in detail in Section 9.3. Modulation causes the spectrum of the measurement signal to be shifted to around 5 kHz (Figure 6.13(d)). If the 50 Hz interference is added after modulation, i.e. during transmission from sensor/modulator to a remote amplifier/demodulator, the interference spectrum is not shifted. The interference can then be easily rejected by an a.c. amplifier, i.e. a band pass filter with bandwidth matched to the spectrum of the amplitude modulated signal. Modulation, however, does not help the problem
6.5 ME TH O D S O F R E D UCI N G E F F E CT S O F N O I S E AN D I N T E R F E R E N CE
119
Figure 6.13 Limitations of filtering and use of modulation.
of Figure 6.13(a); the noise has a uniform power spectral density over a wide band of frequencies, so that moving the measurement signal to a different frequency range does not improve the signal-to-noise ratio.
6.5.7
Averaging Signal averaging can be used to recover a repetitive measurement signal affected by random noise, even if the signal r.m.s. value is much less than that of the noise.[3] The process is shown in Figure 6.14. Suppose that T is the time for each complete cycle of the repetitive signal; p sections of the noise-affected signal, each of duration T, are fed into the averager. N samples from each section are taken and stored, giving pN samples in total; typically
120 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS Figure 6.14 Signal averaging.
we may have p = 50, N = 100. The sampling is exactly synchronised: i.e. if the ith sample of the 1st section is taken at time t0, the ith sample of the 2nd section is taken at t0 + T, the ith sample of the 3rd at t0 + 2T, and so on. Corresponding samples from each section are then averaged: e.g. the first sampled values from each of the p sections are added together and divided by p. Thus the average value of the ith sample is: y iAV =
1 ( yi1 + yi2 + . . . + yip ), p
i = 1, . . . , N
[6.60]
Each of these N average sample values are then displayed at the appropriate time to give the averaged signal. Corresponding sample values A, B, C, . . . of the signal component are approximately equal, so that the average value Z has a similar magnitude. Corresponding sample values A′, B′, C′, . . . of the noise component are very different, some positive and some negative, so that the average value Z′ is reduced in magnitude. Averaging therefore maintains the r.m.s. value of the measurement signal while reducing the r.m.s. value of the random noise. This improvement in signal-to-noise ratio can be readily calculated for random noise with a normal probability density function. Suppose we have p normal signals y1(t) to yp(t), with standard deviations σ1 to σp respectively; then from Section 2.3 the average signal yAV(t) =
1 [ y1(t) + y2(t) + . . . + yp(t)] p
[6.61]
is also normal with standard deviation
σAV =
σ 2p σ 12 σ 22 + + . . . + p2 p2 p2
If σ1 = σ2 = σp = σ, then
[6.62]
6.5 ME TH O D S O F R E D UCI N G E F F E CT S O F N O I S E AN D I N T E R F E R E N CE
Reduction in noise standard deviation due to averaging
σ AV =
pσ 2 σ = p2 p
121
[6.63]
Thus if we average 50 sections σAV = σ /ê50ú ≈ σ /7, i.e. the noise r.m.s. value is reduced by a factor of 7, giving an increase in signal-to-noise ratio of 17dB.
6.5.8
Autocorrelation Autocorrelation can be used to detect the presence of a sinusoidal or any periodic signal buried in random noise. The actual waveform of the measurement signal cannot be recovered because phase information is lost in correlation, but we can measure the amplitude and period of the signal from the autocorrelation function (ACF) of the noise-affected signal. The ACF for the (signal + noise) is the sum of the signal ACF and noise ACF, i.e. S N R S+N yy ( β ) = R yy( β ) + R yy( β )
[6.64]
Thus using eqns [6.27] and [6.31] the ACF for a sinusoidal signal affected by bandlimited white noise is: R S+N yy ( β ) =
b2 sin ωC β cos ω1 β + A 2 β
[6.65]
The form of R S+N yy ( β ) is shown in Figure 6.15; at large values of β the A(sin ωc β )/β term due to the noise decays to zero, leaving the (b2/2) cos ω 1 β term due to the signal. Thus the amplitude b and period 2π/ω 1 of the original signal can be found from the amplitude and period of the autocorrelation function at large values of time delay. This method can be used in the vortex flowmeter (Chapter 12) to measure the vortex frequency in the presence of random flow turbulence.
Figure 6.15 Autocorrelation detection of periodic signal buried in noise.
122 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS
Conclusion The chapter began by defining random signals and deterministic signals and explained that in many practical situations the wanted signal may be random. Unwanted signals may also be present in the measurement circuit; these can be classified as either interference (deterministic) or noise (random). The chapter then explained how random signals can be quantified using the following statistical functions: mean, standard deviation, probability density function, power spectral density and autocorrelation function. The effects of noise and interference voltage on measurement circuits using both voltage and current transmission were then discussed. In the following section internal noise and external interference sources were discussed and the mechanisms whereby external sources are coupled to the measurement circuit were explained. The chapter concluded by explaining methods of reducing the effects of noise and interference, including electromagnetic shielding, electrostatic screening, filtering, modulation, averaging and autocorrelation.
References [1] [2] [3]
oliver f j 1972 Practical Instrumentation Transducers. Pitman, London, pp. 290–333. cook b e 1979 ‘Electronic noise and instrumentation’, Journal of Instrumentation Measurement and Control, vol. 12, no. 8. Hewlett Packard, Technical Literature on Model 3721A Correlator.
Problems 6.1
Table Prob. 1 gives 50 sample values of a random signal. (a) (b)
Table Prob. 1.
−0.59 −1.00 0.65 2.15 0.42 −0.63 6.2
Estimate the mean and standard deviation of the signal. Using an interval ∆y = 0.5 V, draw the Pj and Cj discrete probability distributions for the signal. 1.02 1.36 0.11 0.91 0.05 −0.06
−0.25 0.03 0.99 0.89 0.26 −0.61
−0.34 0.04 0.17 1.43 0.33 0.77
0.95 −0.13 0.39 −1.69 −0.42 1.90
1.24 −0.71 2.61 −0.25 0.79
−0.30 −1.23 −0.08 2.47 −0.07
0.21 0.03 −0.33 −1.97 −0.32
−0.89 −1.00 0.99 −2.26 −0.66
Two complete periods of a square wave can be represented by the following 20 sample values: +1+1+1+1+1
−1−1−1−1−1
+1+1+1+1+1
−1−1−1−1−1
Find the autocorrelation function of the signal over one complete period by evaluating the coefficients Ryy (m∆T ) for m = 0, 1, 2, . . . ,10.
PROBLEMS
6.3
A sinusoidal signal of amplitude 1.4 mV and frequency 5 kHz is ‘buried’ in Gaussian noise with zero mean value. The noise has a uniform power spectral density of 100 pW Hz−1 up to a cut-off frequency of 1 MHz. (a) (b) (c) (d) (e)
6.4
123
Find the total power, r.m.s. value and standard deviation for the noise signal. What is the signal-to-noise ratio in dB? Sketch the autocorrelation function for the combined signal and noise. The combined signal is passed through a band-pass filter with centre frequency 5 kHz and bandwidth 1 kHz. What improvement in signal-to-noise ratio is obtained? The filtered signal is then passed through a signal averager which averages corresponding samples of 100 sections of signal. What further improvement in signal-to-noise ratio is obtained?
A thermocouple giving a 10 mV d.c. output voltage is connected to a high impedance digital voltmeter some distance away. A difference in potential exists between earth at the thermocouple and earth at the voltmeter. Using the equivalent circuit given in Figure Prob. 4. (a) (b)
Calculate the r.m.s. values of the series mode and common mode interference voltages at the voltmeter input. If the digital voltmeter has a common mode rejection ratio of 100 dB, find the minimum and maximum possible measured voltages.
Figure Prob. 4.
6.5
A sinusoidal signal is transmitted over a noisy transmission link to a remote correlator acting as a receiver. Figure Prob. 5 shows a typical autocorrelation function. Use the figure to estimate the following quantities: (a) (b) (c) (d) (e) (f)
Signal power Noise power Signal-to-noise ratio in decibels Signal amplitude Signal frequency Noise standard deviation (assume zero mean).
Hint: use eqns [6.27] and [6.33].
124 S IGNALS AND NOISE IN MEASUREME N T S Y S T E MS Figure Prob. 5.
7 Reliability, Choice and Economics of Measurement Systems In Chapters 3 and 4 we defined the accuracy of a measurement system and explained how measurement error can be calculated, under both steady-state and dynamic conditions. Reliability is another important characteristic of a measurement system; it is no good having an accurate measurement system which is constantly failing and requiring repair. The first section of this chapter deals with the reliability of measurement systems, first explaining the fundamental principles of reliability and the reliability of practical systems, then failure rate data, and finally examining ways of improving reliability. The following section examines the problems of how to choose the most appropriate measurement system, for a given application, from several competing possibilities. Initially a specification for the required application can be drawn up: this will be a list of important parameters such as accuracy, reliability, etc., each with a desired value. This can then be compared with the manufacturer’s specification for each of the competing measurement systems and the system with the closest specification is chosen. Even if all the required information is available this procedure is far from satisfactory because it takes no account of the relative importance of each parameter. A better method, explained in the final section, is to choose the system with minimum total lifetime operating cost.
7.1 7.1.1
Reliability of measurement systems Fundamental principles of reliability Probability P If a large number of random, independent trials are made, then the probability of a particular event occurring is given by the ratio: P=
number of occurrences of the event total number of trials
[7.1]
126 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS in the limit that the total number of trials tends to infinity. Thus the probability of a tossed coin showing heads tends to the theoretical value of –12 over a large number of trials.
Reliability R(t) The reliability of a measurement element or system can be defined as: ‘the probability that the element or system will operate to an agreed level of performance, for a specified period, subject to specified environmental conditions’. In the case of a measurement system ‘agreed level of performance’ could mean an accuracy of ±1.5%. If the system is giving a measurement error outside these limits, then it is considered to have failed, even though it is otherwise working normally. The importance of environmental conditions on the reliability of measurement systems will be discussed more fully later. Reliability decreases with time; a measurement system that has just been checked and calibrated should have a reliability of 1 when first placed in service. Six months later, the reliability may be only 0.5 as the probability of survival decreases.
Unreliability F(t) This is ‘the probability that the element or system will fail to operate to an agreed level of performance, for a specified period, subject to specified environmental conditions’. Since the equipment has either failed or not failed the sum of reliability and unreliability must be unity, i.e. R(t) + F(t) = 1
[7.2]
Unreliability also depends on time; a system that has just been checked and calibrated should have an unreliability of zero when first placed in service, increasing to, say, 0.5 after six months.
7.1.2
Practical reliability definitions Since R(t) and F(t) are dependent on time, it is useful to have measures of reliability which are independent of time. We will consider two cases: in the first the items are non-repairable and in the second the items are repairable.
Non-repairable items Suppose that N individual items of a given non-repairable component are placed in service and the times at which failures occur are recorded during a test interval T. We further assume that all the N items fail during T and that the ith failure occurs at time Ti, i.e. Ti is the survival time or up time for the ith failure. The total up time for i=N N failures is therefore ∑ i=1 Ti and the mean time to failure is given by Mean time to fail =
total up time number of failures
7.1 RE LI ABI LI TY O F ME AS UR E ME N T S Y S T E MS
127
i.e. 1 N
MTTF =
i=N
∑T
[7.3]
i
i =1
The mean failure rate õ is correspondingly given by: Mean failure rate =
number of failures total up time
i.e.
õ=
N ∑
i= N i =1
[7.4]
Ti
i.e. mean failure rate is the reciprocal of MTTF. There are N survivors at time t = 0, N − i at time t = Ti, decreasing to zero at time t = T; Figure 7.1(a) shows how the probability of survival, i.e. reliability, Ri = (N − i)/N decreases from Ri = 1 at t = 0, to Ri = 0 at t = T. The ith rectangle has height 1/N, length Ti and area Ti /N. Therefore from eqn [7.3] we have: MTTF = total area under the graph In the limit that N → ∞, the discrete reliability function Ri becomes the continuous function R(t). The area under R(t) is ∫ 0T R(t) dt so that we have in general: MTTF =
0002
∞
R(t) dt
[7.5]
0
The upper limit of t = ∞ corresponds to N being infinite.
Repairable items Figure 7.1(b) shows the failure pattern for N items of a repairable element observed over a test interval T. The down time TDj associated with the jth failure is the total time that elapses between the occurrence of the failure and the repaired item being put back into normal operation. The total down time for NF failures is therej=N fore ∑ j=1 TDj and the mean down time is given by F
Mean down time =
total down time number of failures
i.e.
MDT =
1 NF
j = NF
∑T
Dj
j =1
[7.6]
128 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS Figure 7.1 Failure patterns: (a) Non-repairable items (b) Repairable items.
7.1 RE LI ABI LI TY O F ME AS UR E ME N T S Y S T E MS
129
The total up time can be found by subtracting the total down time from NT, i.e.: Total up time = NT −
j NF
∑T
Dj
j 1
= NT − NF MDT The mean up time or the mean time between failures (MTBF) is therefore given by: Mean time between failures =
total up time number of failures
i.e. MTBF =
NT − N F MDT NF
[7.7]
The mean failure rate õ is correspondingly given by: Mean failure rate =
number of failures total up time
i.e. õ=
NF NT − N F MDT
[7.8]
Again mean failure rate is the reciprocal of MTBF. Thus if 150 faults are recorded for 200 transducers over 1.5 years with a mean down time of 0.002 years, then the observed MTBF is 1.998 years and the mean failure rate is 0.5005 yr −1. The availability A of the element is the fraction of the total test interval over which it is performing within specification, i.e. up; thus we have: Availability =
total up time test interval
=
total up time total up time + total down time
=
NF × MTBF NF × MTBF + NF × MDT
i.e. A=
MTBF MTBF + MDT
[7.9]
Using the above data of MTBF = 1.998 years and MDT = 0.002 years gives A = 0.999.
130 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS Unavailability U is similarly defined as the fraction of the total test interval over which it is not performing to specification, i.e. failed or down; thus we have: Unavailability =
total down time test interval
giving: U=
MDT MTBF + MDT
[7.10]
It follows from eqns [7.9] and [7.10] that: A+U=1
7.1.3
[7.11]
Instantaneous failure rate and its relation to reliability We assume to begin with that n items of an element survive up to time t = ξ and that ∆n items fail during the small time interval ∆ξ between ξ and ξ + ∆ξ. The probability of failure during interval ∆ξ (given survival to time ξ ) is therefore equal to ∆n/n. Assuming no repair during ∆ξ the corresponding instantaneous failure rate or hazard rate at time ξ is, from eqn [7.8], given by:
λ (ξ ) =
∆n failure probability = n∆ξ ∆ξ
[7.12]
The unconditional probability ∆ F that an item fails during the interval ∆ξ is: ∆ F = probability that item survives up to time ξ AND probability that item fails between ξ and ξ + ∆ξ (given survival to ξ ). The first probability is given by R(ξ ) and from eqn [7.12] the second probability is λ (ξ )∆ξ. The combined probability ∆F is the product of these probabilities: ∆F = R(ξ )λ (ξ )∆ξ i.e. ∆F = R(ξ )λ (ξ ) ∆ξ Thus in the limit that ∆ξ → 0, we have: dF = R(ξ )λ (ξ ) dξ Also since F(ξ ) = 1 − R(ξ ), dF/dξ = −(dR/dξ ), giving: −
dR = R(ξ )λ (ξ ) dξ
[7.13]
7.1 RE LI ABI LI TY O F ME AS UR E ME N T S Y S T E MS
131
i.e.
0002
R(t)
t
0002 λ(ξ ) dξ
dR =− R R(0)
[7.14]
0
In eqn [7.14], the left-hand integral is with respect to R and the right-hand integral with respect to ξ. Since at t = 0, R(0) = 1, we have: t
0002 λ(ξ ) dξ
[loge R] 1R(t) = −
0
t
loge R(t) = −
0002 λ(ξ ) dξ 0
i.e. Relation between reliability and instantaneous failure rate
7.1.4
 R(t ) = exp −  
 λ (ξ ) dξ   0 
0002
t
[7.15]
Typical forms of failure rate function In the previous section instantaneous failure rate or hazard rate λ (t) was defined. Figure 7.2 shows the most general form of λ (t) throughout the lifetime of an element. This is the so-called bathtub curve and consists of three distinct phases: early failure, useful life and wear-out failure. The early failure region is characterised by λ (t) decreasing with time. When items are new, especially if the element is of a new design, early failures can occur due to design faults, poor quality components, manufacturing faults, installation errors, operator and maintenance errors; the latter may be due to unfamiliarity with the product. The hazard rate falls as design faults are rectified, weak components are removed, and the user becomes familiar with installing, operating and maintaining the element. The useful life region is characterised by a low,
Figure 7.2 Typical variation in instantaneous failure rate (hazard rate) during the lifetime of element – ‘bathtub curve’.
132 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS Figure 7.3 Reliability and unreliability with constant failure rate model.
constant failure rate. Here all weak components have been removed: design, manufacture, installation, operating and maintenance errors have been rectified so that failure is due to a variety of unpredictable lower-level causes. The wear-out region is characterised by λ (t) increasing with time as individual items approach the end of the design life for the product; long-life components which make up the element are now wearing out. Many measurement elements have a useful life region lasting many years, so that a constant failure rate model is often a good approximation. Here we have:
λ (t) = λ (ξ ) = λ = constant
[7.16]
so that:  R(t ) = exp −λ  
 ξ  = exp(−λ t )  0  
0002
t
[7.17]
F (t ) = 1 − exp(−λ t ) Thus a constant failure or hazard rate gives rise to an exponential reliability time variation or distribution as shown in Figure 7.3.
7.1.5
Reliability of systems Series systems We saw in Chapter 1 that a complete measurement system consists of several elements usually in series or cascade. Figure 7.4 shows a series system of m elements with individual reliabilities R1, R2, . . . , Ri, . . . , R m respectively. The system will only survive if every element survives; if one element fails then the system fails.
Figure 7.4 Reliability of series system.
7.1 RE LI ABI LI TY O F ME AS UR E ME N T S Y S T E MS
133
Assuming that the reliability of each element is independent of the reliability of the other elements, then the probability that the system survives is the probability that element 1 survives and the probability that 2 survives and the probability that 3 survives, etc. The system reliability RSYST is therefore the product of the individual element reliabilities, i.e. Reliability of series system
RSYST = R1 R2 . . . R i . . . R m
[7.18]
If we further assume that each of the elements can be described by a constant failure rate λ (Section 7.1.4), and if λ i is the failure rate of the ith element, then R i is given by the exponential relation (eqn [7.17]): R i = e−λ i t
[7.19]
R SYST = e−λ1 t e−λ 2 t . . . e−λ i t . . . e−λ m t
[7.20]
Thus so that if λ SYST is the overall system failure rate: RSYST = e−λ SYST t = e−(λ1 +λ 2 +...+λ i +...+λ m )t
[7.21]
and Failure rate of system of m elements in series
λ SYST = λ 1 + λ 2 + . . . + λ i + . . . + λ m
[7.22]
This means that the overall failure rate for a series system is the sum of the individual element or component failure rates. Equations [7.18] and [7.22] show the importance of keeping the number of elements in a series system to a minimum; if this is done the system failure rate will be minimum and the reliability maximum. A measurement system consisting of a thermocouple (λ 1 = 1.1), a millivolt to current converter (λ 2 = 0.1) and a recorder (λ 3 = 0.1) in series will therefore have a failure rate λ SYST = 1.3.
Parallel systems Figure 7.5 shows an overall system consisting of n individual elements or systems in parallel with individual unreliabilities F1, F2, . . . , Fj , . . . , Fn respectively. Only one individual element or system is necessary to meet the functional requirements placed on the overall system. The remaining elements or systems increase the reliability of the overall system; this is termed redundancy. The overall system will only fail if every element/system fails; if one element/system survives the overall system survives. Assuming that the reliability of each element/system is independent of the reliability of the other elements, then the probability that the overall system fails is the probability that element/system 1 fails and the probability that 2 fails and the probability that 3 fails, etc. The overall system unreliability FSYST is therefore the product of the individual element system unreliabilities, i.e. Unreliability of parallel system
FSYST = F1 F2 . . . Fj . . . Fn
[7.23]
134 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS Figure 7.5 Reliability of parallel system.
Comparing eqns [7.18] and [7.23] we see that, for series systems, system reliability is the product of element reliabilities, whereas for parallel systems system unreliability is the product of element unreliabilities. Often the individual elements/systems are identical, so that F1 = F2 = . . . = Fi = . . . = Fn = F, giving: FSYST = F n
[7.24]
The temperature measurement system of the previous section has a failure rate λ SYST = 1.3 yr−1; the corresponding unreliability F is given by F = 1 − e−λ SYST t Thus if t = 0.5 year then F = 0.478. Figure 7.6 shows a redundant system consisting of three single temperature measurement systems in parallel. The overall system unreliability is therefore: FOVERALL = F 3 = 0.109 so that the probability of a failure with the overall system is less than a quarter of that of a single system. The above parallel system, while reliable, is expensive. Since the thermocouple failure rate is 11 times greater than the converter and recorder failure rates, a more
Figure 7.6 Reliability of three thermocouple temperature measurement systems in parallel.
7.1 RE LI ABI LI TY O F ME AS UR E ME N T S Y S T E MS
135
Figure 7.7 Reliability of system with three thermocouples and middle value selector.
cost-effective redundant system would have three thermocouples in parallel and only one converter and recorder. One possible system is shown in Figure 7.7. The three thermocouple e.m.f.’s E1, E2 and E3 are input into a middle value selector element. The selector output signal is that input e.m.f. which is neither the lowest nor the highest; thus if E1 = 5.0 mV, E2 = 5.2 mV and E3 = 5.1 mV, the output signal is E3. If, however, thermocouple 3 fails so that E3 = 0 mV, the selector output signal is E1. The reliability of this system can be analysed by replacing the three thermocouples by a single element of unreliability F1 = (1 − e−λ 1 t )3 = (1 − e−0.55)3 = 0.076 or reliability R1 = 1 − 0.076 = 0.924. The reliability of the other elements with λ = 0.1 is: R2 = R3 = R4 = e−λ t = e−0.05 = 0.951 Using [7.18], the overall system reliability is: R OVERALL = R1 R2 R3 R 4 = 0.924(0.951)3 = 0.795 i.e. FOVERALL = 0.205. This is almost twice the unreliability of the parallel system but less than half that of a single system.
7.1.6
Failure rate data and models A distinction must now be made between components and elements. A component is defined as a ‘non-repairable’ device, i.e. when it fails it is removed and thrown away. Examples are a resistor or an integrated circuit. An element, however, is a repairable part of a system, which is usually made up of several components. Examples are a pressure sensor, a temperature transmitter and a recorder. Failure rate data for both components and elements can be found experimentally by direct measurements of the frequency of failure of a number of items of a given type. Equation [7.4] can be used to find the observed failure rate of non-repairable components and eqn [7.8] to find õ for repairable elements. Table 7.1 gives observed average failure rates for typical instruments. These data have been taken from the UK data bank operated by the Systems Reliability Service (SRS).[1] The table specifies the environment in which each type of instrument is located. For an element located in the process fluid, the environment is determined by the nature of the fluid, e.g. temperature, corrosion properties, presence of dirt or solid particles.
136 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS Table 7.1 Observed average failure rates for instruments.
Instrument
Environment
Chemical analyser, Oxygen pH meter Conductivity indicator Fire detector head Flow transmitter, pneumatic Level indicator, pneumatic Pressure controller Pressure indicator, dial, mechanical Pressure sensor, differential, electronic Pressure transmitter Recorder, pen Temperature indicator and alarm Temperature indicator, resistance thermometer Temperature indicator, bimetal Temperature trip unit Thermocouple Valve, gate non-return solenoid
Poor, chemical/ship Poor, chemical/ship Average, industrial Average, industrial Average, industrial Average, industrial Average, industrial Average, industrial
Experience (item-years)
No. of failures
Failure rate (failures/yr)
4.34 28.08 7.53 1 470 125 898 40 575
30 302 18 128 126 201 63 178
6.92 10.75 2.39 0.09 1.00 0.22 1.58 0.31
Poor, chemical/ship Average, industrial Average, industrial Fair, laboratory
225 85 045 26.02 47.2
419 806 7 101
1.86 0.01 0.27 2.14
Fair, laboratory Average, industrial Average, industrial Poor, chemical/ship Poor, chemical/ship Poor, chemical/ship Poor, chemical/ship
212.3 165 120 317 11 564 1 530 1 804
68 215 70 127 841 101 66
0.32 1.30 0.58 0.40 0.07 0.07 0.04
Source: after Wright[1].
For an element located in the atmosphere, the environment is determined by atmospheric conditions, e.g. temperature, humidity, salinity, presence of dust. The failure rate of a given type of element will depend on the environment in which it is located: an iron–copper thermocouple will have a higher failure rate in a corrosive acid than in water. Table 7.2 shows the observed failure rates for given types of elements at three works A, B, C which process different materials and fluids and have different background environments.[2] The observed failure rates can be regarded as the product of a base failure rate λ B and an environmental correction factor πE: Element failure rate model
λ OBS = π E × λ B
[7.25]
Here λ B corresponds to the best environmental conditions and π E has values 1, 2, 3 or 4, the highest figure corresponding to the worst environment. The failure rate of elements can alternatively be calculated from the failure rate data/models for the basic components which make up the element. Table 7.3 shows the calculation of the overall failure rate, from basic component data, for an electronic square root extractor module.[3] The module gives an output voltage signal in the range 1–5 V, proportional to the square root of the input signal with range 4–20 mA; this type of module is commonly used in fluid flow rate control systems. The module is made up from basic electronic components of various types, all connected in series. Several components of each type are present. From
137
7.1 RE LI ABI LI TY O F ME AS UR E ME N T S Y S T E MS
Table 7.2 Observed failure rates for instruments in different chemical plant environments.
Instrument (p = pneumatic)
Control valve (p) – Works A Works B Works C Differential pressure transmitter (p) – Works A (flow) Works A (level) Works B (flow) Works C (flow) Variable area flowmeter transmitter (p) – Works A Thermocouple Works A Works B Works C Controller – Works A Works B Works C Pressure switch – Works A Works B
Observed failure rate, faults/year
Environmental correction factor
Base failure rate, faults/year
0.25 0.57 2.27 0.127
1 2 4 2
0.25 0.29 0.57 0.064
0.76 1.86 1.71 2.66 1.22
1 3 4 4 2
0.76 0.62 0.43 0.67 0.61
0.68 1.01
1 3
0.68 0.34
0.40 1.34 1.00
3 4 4
0.13 0.34 0.25
0.38 0.26 1.80 0.32
1 1 1 1
0.38 0.26 1.80 0.32
0.14 0.30 1.00
1 2 4
0.14 0.15 0.25
Source: after Lees[2].
eqn [7.22] the failure rate λ of a module containing m different component types in series with failure rates λ 1, λ 2, . . . , λ i, . . . , λ m, and one of each type present is:
λ = λ1 + λ 2 + . . . + λ i + . . . + λ m
[7.26]
If there are multiple components of each type, all connected in series, then the module failure rate is given by: Module failure rate – multiple components of each type
λ = N1λ 1 + N2 λ 2 + . . . + Ni λ i + . . . + Nm λ m
[7.27]
where N1, N2, . . . , Ni, . . . , Nm are the quantities of each component type. The failure rate of each component type is calculated using the model equation:
λ i = (F1i + F2i + F3i ) × Kli × K2i × K3i × K4i
[7.28]
Table 7.3 gives values of F1, F2, F3, K1, K2, K3, K4 and failure rate λ i for each component. Each failure rate is then multiplied by the appropriate quantity Ni and the Ni λ i values are added together to give a total module failure rate of 3.01 × 10−6 per hour.
138 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS Table 7.3 Calculation of overall failure rate for electronic square root module.a
Failure rates per 1010 hours
Component F1
F2
F3
K1
K2
RESISTORS Carbon film 0 < R ≤ 100 K 100 0 0 1 1 100 K < R ≤ 1 M 100 0 0 1 1.1 1 M < R ≤ 10 M 100 0 0 1 1.6 Metal film 0 < R ≤ 100 K 150 0 0 1 1 100 K < R ≤ 1 M 150 0 0 1 1.1 POTENTIOMETERS 0 < R ≤ 50 K 700 0 0 1 1 50 K < R ≤ 100 K 700 0 0 1 1 CAPACITORS Metal film 0 < C ≤ 33 nF 200 0 0 1 1 33 nF < C ≤ 1 µF 200 0 0 1 1 1 µF < C ≤ 10 µF 200 0 0 1 1 Ceramic 0 < C ≤ 3.3 nF 150 0 0 1 1 Electrolytic 3.2 < C ≤ 62 µF 500 0 0 0.29 1 DIODES Silicon LP 200 0 0 0.55 1 Zener 1000 0 0 1.3 1 TRANSISTORS NPN LP 400 0 0 1.4 1 INTEGRATED CIRCUITS OP AMP 160 50 600 1 1 Quad switch 38 320 560 1 1 OTHERS Edge connectors 300 0 0 1 1 Soldered joints 20 0 0 1 1 PCB 60 0 0 1 1 Total rate: 3.01 × 10−6 per hour MTBF: 3.32 × 105 hours a
b
K3
K4
Rateb
Qty
Value
1 1 1
1 1 1
100 110 160
17 2 3
1700 220 480
1 1
1 1
150 165
17 3
2550 495
1 1.1
1 1
700 770
8 1
5600 770
1 1.3 1.5
1 1 1
200 260 300
3 2 1
600 520 300
1
1
150
1
150
0.7
1
102
1
102
1 1
1 1
110 1300
2 1
220 1300
1
1
560
2
1120
1 1
1 1
810 918
9 1
7290 918
1 1 1
1 1 1
300 20 60
8 167 1
2400 3340 60
Data sources: 1: Electronic Reliability Data, National Centre of Systems Reliability, Application Code 2, 25C 2: Reliability Prediction Manual for Guided Weapon Systems, MOD 3: Component supplier’s information. Rate = (F1 + F2 + F3) × K1 × K2 × K3 × K4.
Source: after Hellyer[3].
7.1 RE LI ABI LI TY O F ME AS UR E ME N T S Y S T E MS
7.1.7
139
Design and maintenance for reliability Design for reliability The following general principles should be observed. Element selection. Only elements with well-established failure rate data/models should be used. Furthermore some technologies are inherently more reliable than others. Thus an inductive LVDT displacement sensor (Chapter 8) is inherently more reliable than a resistive potentiometer; the latter involves a contact sliding over a wire track, which will eventually become worn. A vortex flowmeter (Chapter 12) involves no moving parts and is therefore likely to be more reliable than a turbine flowmeter which incorporates a rotor assembly. Environment. The environment in which the element is to be located should first be defined and the element should consist of components and elements which are capable of withstanding that environment. Thus the diaphragm of a differential pressure transmitter on a sulphuric acid duty should be made from a special alloy, e.g. Hastelloy C, which is resistant to corrosion. Minimum complexity. We saw above that, for a series system, the system failure rate is the sum of the individual component/element failure rates. Thus the number of components/elements in the system should be the minimum required for the system to perform its function. Redundancy. We also saw that the use of several identical elements/systems connected in parallel increases the reliability of the overall system. Redundancy should be considered in situations where either the complete system or certain elements of the system have too high a failure rate. Diversity. In practice faults can occur which cause either more than one element in a given system, or a given element in each of several identical systems, to fail simultaneously. These are referred to as common mode failures and can be caused by incorrect design, defective materials and components, faults in the manufacturing process, or incorrect installation. One common example is an electronic system where several of the constituent circuits share a common electrical power supply; failure of the power supply causes all of the circuits to fail. This problem can be solved using diversity; here a given function is carried out by two systems in parallel, but each system is made up of different elements with different operating principles. One example is a temperature measurement system made up of two subsystems in parallel, one electronic and one pneumatic.
Maintenance The mean down time, MDT, for a number of items of a repairable element has been defined as the mean time between the occurrence of the failure and the repaired element being put back into normal operation. It is important that MDT is as small as possible in order to minimise the financial loss caused by the element being out of action.
140 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS There are two main types of maintenance strategy used with measurement system elements. Breakdown maintenance simply involves repairing or replacing the element when it fails. Here MDT or mean repair time, TR, is the sum of the times taken for a number of different activities. These include realisation that a fault has occurred, access to the equipment, fault diagnosis, assembly of repair equipment, components and personnel, active repair/replacement and finally checkout. Preventive maintenance is the servicing of equipment and/or replacement of components at regular fixed intervals; the corresponding maintenance frequency is m times per year. Here MDT or mean maintenance time, TM , is the sum of times for access, service/replacement and checkout activities and therefore should be significantly less than mean repair time with breakdown maintenance.
7.2
Choice of measurement systems The methods to be used and problems involved in choosing the most appropriate measurement system for a given application can be illustrated by a specific example. The example used will be the choice of the best system to measure the volume flow rate of a clean liquid hydrocarbon, range 0 to 100 m3 h−1, in a 0.15 m (6 inch) diameter pipe. The measured value of flow rate must be presented to the observer in the form of a continuous trend on a chart recorder. The first step is to draw up a specification for the required flow measurement system. This will be a list of all important parameters for the complete system such as measurement error, reliability and cost, each with a desired value or range of values. The first two columns of Table 7.4 are an example of such a ‘job specification’. As explained in Chapter 3, system measurement error in the steady state can be quantified in terms of the mean + and standard deviation σE of the error probability distribution p(E). These quantities depend on the imperfections, e.g. non-linearity and repeatability, of every element in the system. System failure rate λ and repair time TR were defined in Section 7.1. Initial cost CI is the cost of purchase, delivery, installation and commissioning of the complete system. CR is the average cost of materials for each repair.
Table 7.4 Comparison table for selection of flow measurement system.
Parameter
Job specification
System 1 System 2 System 3 System 4 Orifice plate Vortex Turbine Electromagnetic
Measurement error (at 50 m3 h−1)
+ ≤ 0.25 σE ≤ 0.8 m3 h−1
0.2 0.7
0.1 0.3
0.03 0.1
Initial cost
CI ≤ £4000
3500
3000
4200
Annual failure rate
λ ≤ 2.0 1.8 failures yr−1
1.0
2.0
6
5
7
100
100
300
Average repair time TR ≤ 8 h Material repair cost
CR ≤ £200
5 4 4 4 4 6 Not technically 4 feasible 4 4 4 7
7.3 T O T AL LI F E T I ME O PE RATI N G CO S T
141
In this example the choice could be between four competing systems based on the orifice plate, vortex, turbine and electromagnetic primary sensing elements. The principles and characteristics of all four elements are discussed in Chapter 12. The configuration of the four systems could be: 1. 2. 3. 4.
Orifice plate – electronic D/P transmitter – square rooter – recorder. Vortex element – frequency to voltage converter – recorder. Turbine element – frequency to voltage converter – recorder. Electromagnetic element – recorder.
The next step is to decide whether all the competing systems are technically feasible. The electromagnetic device will not work with electrically non-conducting fluids such as hydrocarbons, so that system 4 is technically unsuitable. Systems 1, 2 and 3 are feasible and the specification for each competing system must then be written down to see whether it satisfies the job specification. Table 7.4 gives possible specifications for the orifice plate, vortex and turbine systems. This data is entirely fictitious and is given only to illustrate the problems of choice. In practice a prospective user may not have all the information in Table 7.4 at his disposal. The manufacturer will be able to give estimates of initial cost CI and the limits of measurement error, e.g. ±2% of full scale, for the orifice plate system at 50 m3 h−1. He will not, however, be able to give values of mean error +, failure rates and repair times. The last two quantities will depend on the environment of the user’s plants and the maintenance strategy used. This information may be available within the user’s company if adequate maintenance records have been kept. From Table 7.4 we see that the turbine flowmeter system 3 does not satisfy the job specification; both initial cost and material repair cost are outside the limits set. This would appear to rule out system 3, leaving 1 and 2. Both these systems satisfy the job specification but the vortex system 2 is cheaper, more accurate and more reliable than the orifice plate system 1. Thus, based on a straightforward comparison of job and system specification, the vortex measurement system 2 would appear to be the best choice for this application. Under certain circumstances, however, the above conclusion could be entirely wrong. The turbine system is more expensive and less reliable than the vortex system, but is three times more accurate. We must now ask how much this increased accuracy is worth. Suppose the market value of the hydrocarbon is £100 m−3. A measurement error of one standard deviation in the turbine system, where σE = 0.1 m3 h−1, represents a potential cash loss of £10 per hour or approximately £80 000 per annum. A corresponding error in the vortex system, where σE = 0.3 m3 h−1, represents an approximate potential loss of £240 000 per annum. The difference between these two figures far outweighs the extra initial and maintenance costs, so that the turbine system 3 is the best choice in this case. We can conclude, therefore, that in order to choose the correct system for a given application, the financial value of each parameter in the job specification must be taken into account. In a costing application of this type, a digital printout of flow rate is more suitable than an analogue trend record.
7.3
Total lifetime operating cost The total lifetime operating costing (TLOC) of a measurement system is the total cost penalty, incurred by the user, during the lifetime of the system. The TLOC is given by
142 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS TLOC = initial cost of system (purchase, delivery, installation and commissioning) + cost of failures and maintenance over lifetime of the system + cost of measurement error over lifetime of the system [7.29] and therefore takes account of the financial value of each parameter in the job specification. The best system for a given application is then the one with minimum TLOC. This method also enables the user to decide whether a measurement system is necessary at all. If no system is installed, TLOC may still be very large because no measurement implies a large measurement error. A measurement system should be purchased if it produces a significant reduction in TLOC. Using eqn [7.29], we can derive an algebraic expression for TLOC using the parameters listed in Table 7.4. The initial cost of the system is £CI . If the system lifetime is T years and average failure rate λ faults yr −1, then the total number of faults is λT. Since the average repair time is TR hours then the total ‘downtime’ due to repair is λTTR hours. The total lifetime cost of failures is the sum of the repair cost (materials and labour) and the process cost, i.e. the cost of lost production and efficiency while the measurement system is withdrawn for repair. If we define £CR = average materials cost per repair £CL = repair labour cost per hour £CP = process cost per hour then the total repair cost is (CRλT + CLTRλT ) and the total process cost is CPTR λT, giving: Total lifetime cost of failures = [CR + (CL + CP)TR ]λT
[7.30]
The above costs only apply to breakdown maintenance; many users also practise preventive maintenance in order to reduce failure rates. Suppose preventive maintenance is carried out on a measurement system m times yr −1, the average maintenance time is TM hours and the materials cost per service is £CM. The total number of services is mT and the total time taken for preventive maintenance is mTTM hours. Usually preventive maintenance of measurement systems is carried out at a time when the process or plant itself is shut down for repair and maintenance. This means that no process costs are incurred during preventive maintenance, giving: Total lifetime maintenance cost = (CM + CLTM)mT
[7.31]
The last term in eqn [7.29] involves the total lifetime cost of measurement error. In order to evaluate this we first need to evaluate the cost penalty function C(E) (£ yr−1) associated with a given steady-state measurement system error E. The form of C(E) depends on the economics of the process on which the measurement is being made. A good example is temperature measurement in a chemical reactor where a degradation reaction is taking place.[4] Here two reactions occur simultaneously: the feedstock A is converted into a desired product B but B is also degraded to an undesired product C. The rates of both reactions increase sharply with temperature, the rate of B to C being more temperature sensitive than the rate of A to B. There is an optimum temperature at which the yield of B is maximum. If the reactor is operated at either above or below this optimum temperature, then the yield of B is sharply reduced and a cost penalty is incurred. This situation will occur if there is a measurement system error E between measured and true values of reactor temperature: the system tells the operator that the reactor is at optimum temperature
7.3 T O T AL LI F E T I ME O PE RATI N G CO S T
143
Figure 7.8 Error cost penalty function C(E) for different processes.
when it is not. Figure 7.8(a) shows the form of the cost penalty function C(E) in this case. We see that when E = 0, C(E) = 0 corresponding to optimum temperature; but C(E) increases rapidly with positive and negative values of E as the yield of B decreases. In Figure 7.8(b) C(E) represents the cost penalty, due to imperfect flow measurement, incurred by a customer receiving a fluid by pipeline from a manufacturer. If E is positive, the customer is charged for more fluid than he actually receives and so is penalised, i.e. C(E) is positive. If E is negative, the customer is charged for less fluid than was actually received and C(E) is negative. Figure 7.8(c) refers to a noncritical liquid level measurement in a vessel. The vessel should be about half full, but plant problems will occur if the vessel is emptied or completely filled. Thus a cost penalty is incurred only if there is gross measurement error, i.e. the measurement system shows the vessel to be half full when almost empty. We saw in Chapter 3 that the exact value of measurement error E, for a given measurement system at a given time, cannot be found. We can, however, find the probability that the system will have a certain error. This is quantified using a normal probability density function p(E), with mean + and standard deviation σE (Table 3.1). The probability of getting a measurement error between E and E + dE is p(E) dE; the corresponding cost penalty is C(E)p(E) dE per year, or TC(E)p(E) dE throughout the system lifetime. The total lifetime cost of measurement error is then found by integrating the above expression over all possible values of E, i.e. Total lifetime cost of measurement error = T
0002
∞
C(E)p(E) d E
[7.32]
−∞
The integral has a finite value, because the value of p(E) becomes negligible for | E| greater than 3 or 4 standard deviations; it can be evaluated numerically[4] using values of the normalised Gaussian distribution.[5] From eqns [7.29]–[7.32] we have TLOC = CI + [CR + (CL + CP)TR ]λT + (CM + CM TM )mT +T
0002
∞
C(E)p(E) d E −∞
[7.33]
144 REL IABILITY, CHOICE AND ECONOMICS O F ME AS URE ME N T S Y S TE MS The relative importance of the terms in the above equation will depend on the application. In the chemical reactor C(E) and CP will be the major factors so that accuracy and reliability will be far more important than initial cost. In the tank level application, measurement error is unimportant and minimum TLOC will be obtained by the best trade-off between initial cost, reliability, and maintainability.
Conclusion The first section of this chapter discussed the reliability of measurement systems. The fundamental principles and practical definitions of reliability were first explained and the relationship between reliability and instantaneous failure rate was derived. The typical variation in instantaneous failure rate throughout the lifetime of an element was then discussed and the reliability of series and parallel systems examined. The section concluded by looking at failure rate data and models and general strategies in design and maintenance for reliability. The second section dealt with the problem of choice of measurement systems approached by comparing the job specification with those of the competing systems. This method does not take account of the financial value of each item in the specification. A better method, discussed in the final section, is to choose the systems with minimum lifetime operating cost.
References [1] [2] [3]
[4]
[5]
wright r i (SRD Warrington) 1984 ‘Instrument reliability’, Instrument Science and Technology, Institute of Physics, Bristol, vol. 1, pp. 82–92. lees f p 1976 ‘The reliability of instrumentation’, Chemistry and Industry, March, pp. 195–205. hellyer f g (Protech Instruments and Systems) 1985 ‘The application of reliability engineering to high integrity plant control systems’, Measurement and Control, 18 June, pp. 172–6. bentley j p 1979 ‘Errors in industrial temperature measurement systems and their effect on the yield of a chemical degradation reaction’, 8th IMEKO Congress of the International Measurement Confederation, Moscow, May. white j, yeats a and skipworth g 1974 Tables for Statisticians, Stanley Thornes, London, pp. 18–19.
Problems 7.1
A batch of 100 identical thermocouples were tested over a 12-week period. Twenty failures were recorded and the corresponding down times in hours were as follows: 5, 6, 7, 8, 4, 7, 8, 10, 5, 4, 8, 5, 4, 5, 6, 5, 4, 9, 8, 6
PROBLEMS
145
Calculate: (a) (b) (e) (d) 7.2
Mean down time Mean time between failures Mean failure rate Availability.
A flow measurement system consists of an orifice plate (λ = 0.75), differential pressure transmitter (λ = 1.0), square root extractor (λ = 0.1) and recorder (λ = 0.1). Calculate the probability of losing the flow measurement after 0.5 year for the following: (a) (b) (c)
A single flow measurement system Three identical flow measurement systems in parallel A system with three orifice plates, three differential pressure transmitters and a middle value selector relay (λ = 0.1). The selected transmitter output is passed to a single square root extractor and recorder.
Annual failure rate data are given in brackets; assume that all systems were initially checked and found to be working correctly. 7.3
Use the data given in Table Prob. 3 to decide which level measurement system should be purchased. Assume a breakdown maintenance only strategy is practised, each system has the same measurement error, and there is a 10-year total lifetime.
Table Prob. 3. Initial cost Materials cost per repair Labour cost per hour Process cost per hour Repair time Annual failure rate
Parameter
System 1
System 2
£ £ £ £ h yr −1
1000 20 10 100 8 2.0
2000 15 10 100 12 1.0
Part B Typical Measurement System Elements
8 Sensing Elements
In Chapter 1 we saw that, in general, a measurement system consists of four types of element: sensing, signal conditioning, signal processing and data presentation elements. The sensing element is the first element in the measurement system; it is in contact with, and draws energy from, the process or system being measured. The input to this element is the true value of the measured variable; the output of the element depends on this value. The purpose of this chapter is to discuss the principles and characteristics of sensing elements in wide current use; more specialised elements are discussed in later chapters. Table 8.1 lists the sensing elements described in this book according to the physical principle involved, e.g. inductive or thermoelectric. The elements are classified according to whether the output signal is electrical, mechanical, thermal or optical. Elements with an electrical output are further divided into passive and active. Passive devices such as resistive, capacitive and inductive elements require an external power supply in order to give a voltage or current output signal; active devices, e.g. electromagnetic and thermoelectric elements, need no external power supply. The table also lists the input measured variables which a given type of element is used to sense, by giving the relevant section number in the book. Thus capacitive elements can sense pressure, liquid level, displacement and humidity. Elastic elements can sense force, pressure torque, acceleration and density. Sensors with a mechanical output are commonly used as the primary sensing element in measurement systems for mechanical variables such as force or flow rate. In order to obtain an electrical signal, this primary element is followed by a secondary sensing element with an electrical output signal. Examples are a resistive strain gauge sensing the strain in an elastic cantilever in a force measurement system, and an electromagnetic tachogenerator sensing the angular velocity of a turbine in a flow measurement system.
8.1 8.1.1
Resistive sensing elements Potentiometers for linear and angular displacement measurement Figure 8.1 shows potentiometers for the measurement of (a) linear (rectilinear) and (b) angular (rotary) displacement. They consist of a former with a cylindrical
Heat transfer Various
Optical output
Differential pressure Turbine Vortex Pneumatic Coriolis
Elastic
15.6
15.5
15.5
15.6
13.1
8.6 9.4 9.5
16.2
15.6
13.1
4.1 8.6 9.5
8.7
8.6
15.6
13.1
9.4 9.5
Density
Thermal output
Mechanical output
15.5
15.5
12.4
12.3 12.3 12.3
14.3
12.2
15.6
13.1
5.1 8.1 8.2 8.3
15.6
8.4
Velocity
Piezoelectric Electrochemical Pyroelectric
8.5 15.5
Flow rate 12.4
14.3
8.6
8.7
8.6 4.4
15.6
14.4
8.9
8.9
8.1 14.4
8.9
8.9
Ionic concentration
Electromagnetic Thermoelectric
8.2
Level
8.8
Torque 8.6
Flow velocity
8.2
Force 8.6
Displacement/strain
15.5 15.5
Pressure 8.6
8.2
8.1
8.10
Magnetic field
Electrical output active
Heat/light flux 15.5
Acceleration
Capacitive Inductive Piezoresistive Photovoltaic Photoconductive FET Hall effect
Temperature 8.1 15.5
Gas composition
Resistive
Humidity
Electrical output passive
Physical principle
Input measured variable
Table 8.1 Sensing elements and measured variables.
150 S ENSING ELEMENTS
8.1 RESISTIVE SENSING ELEMENTS
151
Figure 8.1 Potentiometer displacement sensors.
cross-section which is either a straight cylinder or an arc of a circle. Resistive material is then placed on the former so that the resistance per unit length is constant (the usual case). This means that resistance is proportional to the distance d travelled by the wiper between A and B. From Figure 8.1(a), the ratio of open circuit voltage ETh to supply voltage VS is given by: ETh voltage across AB resistance across AB = = VS voltage across CB resistance across CB where: resistance of CB = total resistance of potentiometer = RP resistance of AB = fractional resistance = RP d/dT = RP x x = fractional displacement = d/dT . Therefore the open circuit voltage for a linear displacement potentiometer is: ETh = VS x = VS d/dT
[8.1]
i.e. voltage is proportional to displacement. From Figure 8.1(b), the open circuit voltage for an angular displacement potentiometer is: ETh = VS θ/θT = VS x
[8.2]
where x = θ /θT is the fractional angular displacement. In Section 5.1.2 the Thévenin resistance of the potentiometer was shown to be RTh = RP x(1 − x). It was also shown that in general the relationship between the voltage across an external load RL and fractional displacement x is non-linear (eqn [5.10]). The choice of a potentiometer for a given application involves four main parameters:
•
Maximum travel dT , θT Depends on range of displacement to be measured, e.g. 0 to 5 cm, 0 to 300°.
152 S ENSING ELEMENTS
• • •
Supply voltage VS Set by required output range, e.g. for a range of 0 to 5 V d.c., we need VS = 5 V d.c. Resistance RP For a given load RL , choose RP to be sufficiently small compared with RL so that maximum non-linearity is acceptable (eqn [5.12] ). Power rating Wmax Wmax should be greater than actual power V S2 /RP produced in RP.
In wire wound potentiometers the resistive track, of total length dT or θT , consists of n discrete turns of wire. The resistance between A and B therefore increases in a series of steps for a smooth continuous increase in displacement d or θ. The corresponding resolution error is therefore dT /n, θ T /n or (100/n)% (Section 2.1). A typical family of rectilinear precision potentiometers covers displacement spans from 12 to 2500 mm, non-linearity from ±0.2% and resolution from 0.008% with resistance values of 40 Ω/mm. In conductive plastic film potentiometers the track is continuous so that there is zero resolution error and less chance of contact wear than with wire wound; the temperature coefficient of resistance is, however, higher. A family of rectilinear precision potentiometers covers displacement spans from 25 to 250 mm, non-linearities to ±0.04% and resistance values from 500 Ω to 80 kΩ. A typical angular (rotary) potentiometer has a diameter of 42 mm, an input range of 0 to 350°, a corresponding resistance range of 1 to 220 kΩ and non-linearity of ±0.5%. The most modern development is the hybrid track potentiometer, which is manufactured by depositing a conductive film on a precision wire wound track and incorporates the best features of both types.
8.1.2
Resistive metal and semiconductor sensors for temperature measurement The resistance of most metals increases reasonably linearly with temperature in the range −100 to +800 °C. The general relationship between the resistance RT Ω of a metal element and temperature T °C is a power series of the form: RT = R0(1 + α T + βT 2 + γ T 3 + . . . )
[8.3]
where R 0 Ω is the resistance at 0 °C and α, β, γ are temperature coefficients of resistance. The magnitude of the non-linear terms is usually small. Figure 8.2(a) shows the variation in the ratio RT /R0 with temperature for the metals platinum, copper and nickel. Although relatively expensive, platinum is usually chosen for industrial resistance thermometers; cheaper metals, notably nickel and copper, are used for less demanding applications. Platinum is preferred because it is chemically inert, has linear and repeatable resistance/temperature characteristics, and can be used over a wide temperature range (−200 to +800 °C) and in many types of environments. It can be refined to a high degree of purity, which ensures that statistical variations in resistance, between similar elements at the same temperature (Section 2.3.2), are small. A typical platinum element has R 0 = 100.0 Ω, R100 = 138.50 Ω, R200 = 175.83 Ω, α = 3.91 × 10−3 °C−1 and β = −5.85 × 10−7 °C−2. The change in resistance between
8.1 RESISTIVE SENSING ELEMENTS
Figure 8.2 Metal resistive temperature sensors: (a) Resistance/ temperature characteristics of commonly used metals (b) Typical construction of platinum element probe.[1]
153
the ice point and the steam point, i.e. R100 − R0, is called the fundamental interval; in the above element this is 38.5 Ω. The maximum non-linearity as a percentage of f.s.d. (eqn [2.5]) between 0 and 200 °C is +0.76%. The standard IEC 751: 1983 (BS EN 60751: 1996)[2] lays down tolerance limits on the maximum variation in resistance between platinum elements at a given temperature. For class A elements the tolerance limits are ±0.06 Ω at 0 °C and ±0.20 Ω at 200 °C; for class B elements the tolerance limits are ±0.12 Ω at 0 °C and ±0.48 Ω at 200 °C. The amount of electrical power produced in the element should be limited in order to avoid self-heating effects (Chapter 14); in a typical element 10 mW of power causes a temperature rise of 0.3 °C. One type of element is constructed using the partially supported arrangement shown in Figure 8.2(b).[1] Here fine platinum wire is wound into a very small spiral and is inserted into axial holes in a high purity alumina insulator. A small quantity of glass adhesive is introduced into the holes and the unit is fired, thus securely fixing a part of each turn onto the alumina; the remainder of the wire is free to move. The diagram also shows the element housed in a stainless steel protective sheath. Resistive temperature elements made from semiconductor materials are often referred to as thermistors. The most commonly used type is prepared from oxides of the iron group of transition metal elements such as chromium, manganese, iron, cobalt and nickel. The resistance of these elements decreases with temperature – in other words there is a negative temperature coefficient (NTC) – in a highly non-linear way. Figure 8.3 shows typical thermistor resistance–temperature characteristics which can be described by the relationship:  β Rθ = K exp   θ
[8.4]
154 S ENSING ELEMENTS Figure 8.3 Thermistor resistance–temperature characteristics and types (graph (a) after Mullard Ltd[3] ).
where Rθ is the resistance at temperature θ kelvin; K and β are constants for the thermistor. A commonly used alternative equation is: 1 1 Rθ = Rθ exp β  −  θ θ1  1
[8.5]
where Rθ 1 Ω is the resistance at reference temperature θ 1 K, usually θ 1 = 25 °C = 298 K. Thermistors are usually in the form of either beads, rods or discs (Figure 8.3); bead thermistors are enclosed in glass envelopes. A typical NTC thermistor has a resistance of 12 kΩ at 25 °C (298 K), falling to 0.95 kΩ at 100 °C (373 K), and β = 3750 K.[3] The manufacturer’s tolerance limits on the above figures are ±7%, i.e. ±840 Ω, at 25 °C, and ±5%, i.e. ±47.5 Ω, at 100 °C, which is far wider than for metal elements. The element time constant is 19 seconds in air and 3 seconds in oil, and the self-heating effect is 1 °C rise for every 7 mW of electrical power. Thermistors with positive temperature coefficients (P.T.C.) are also available; the resistance of a typical element increases from 100 Ω at −55 °C to 10 kΩ at 120 °C. Thick-film polymers can be used as resistive sensors for temperature and humidity measurement.[4] These are pastes consisting of a polymer matrix (usually epoxy, silicone or phenolic resin) that binds together the filler particles. If the filler material is metallic, e.g. silver flakes or copper particles, then the paste has similar resistive properties to a metal. If carbon particles are used as the filler material, then the paste has the resistive properties of a semiconductor. Solvents are then added to give the paste the thixotropic properties of an ink so it can be screen printed on to a substrate (often alumina). Following the printing process, the solvents are dried off and the pastes are cured at temperatures that do not affect the physical and chemical stability of the substrates. The printed patterns used for carbon and silver paste are shown in Figure 8.4(a). Figure 8.4(b) shows the variation in resistances with temperature for silver and carbon polymers. Silver polymers exhibit a linear increase in resistance with temperature. This is described by a linear approximation to eqn [8.3], i.e.
8.1 RESISTIVE SENSING ELEMENTS
Figure 8.4 Thick-film polymer resistive sensors: (a) Printed patterns for carbon paste (left) and silver paste (right) (b) Changes in resistance of silver and carbon filled pastes with temperature (c) Changes in resistance of carbon polymer pastes with relative humidity (Figures 8.4(b) and (c) are reprinted by permissions of the Institute of Measurement and Control and the University of Southampton and N. M. White[4] ).
RT = R0(1 + α T )
155
[8.6]
An alternative linear approximation for T between 25 and 200 °C is: RT = R25 [1 + α (T − 25)]
[8.7]
where α = 3.732 × 10−3 °C −1.[4] In contrast the carbon paste cured at 220 °C exhibits semiconductor or thermistorlike characteristics, described by eqns [8.4] and [8.5]: here the constant β = 136 K. Carbon polymers also show a variation in resistance with humidity. The polymer matrix absorbs water, causing it to swell. This causes the carbon particles embedded in the matrix to move further away from each other, thus reducing the number of conductive paths and consequently increasing the resistivity of the material. Figure 8.4(c) shows the variation in resistance with humidity in the range 30% to 90% for carbon polymers cured at 160 °C and 200 °C and operated at 25 °C and 60 °C. The resistance–humidity characteristics are non-linear; the higher the curing temperature the greater the non-linearity.[4]
156 S ENSING ELEMENTS Figure 8.5 Stress and strain: (a) Effect of tensile stress (b) Effect of compressive stress.
8.1.3
Metal and semiconductor resistive strain gauges Before discussing strain gauges we must first briefly explain the concepts of stress, strain, elastic modulus and Poisson’s ratio. Stress is defined by force/area, so that in Figure 8.5(a) the stress experienced by the body is +F/A, the positive sign indicating a tensile stress which tends to increase the length of the body. In Figure 8.5(b) the stress is −F/A, the negative sign indicating a compressive stress which tends to reduce the length of the body. The effect of the applied stress is to produce a strain in the body which is defined by (change in length)/(original unstressed length). Thus in Figure 8.5(a) the strain is e = +∆l/l (tensile), and in 8.5(b) the strain is e = −∆l/l (compressive); in both cases the strain is longitudinal, i.e. along the direction of the applied stress. The relationship between strain and stress is linear for a given body over a certain range of values; the slope of the straight line is termed the elastic modulus of the body:
Elastic modulus Elastic modulus =
stress strain
[8.8]
For linear tensile or compressive stress the elastic modulus is called Young’s modulus E; for shear stress the relevant elastic modulus is shear modulus S. Returning to Figure 8.5(a) we note that the increase in length of the body is accompanied by a decrease in cross-sectional area, i.e. a reduction in width and thickness. Thus in Figure 8.5(a) the longitudinal tensile strain is accompanied by a transverse compressive strain, and in Figure 8.5(b) the longitudinal compressive strain is accompanied by a transverse tensile strain. The relation between longitudinal strain eL and accompanying transverse strain eT is: eT = −νeL
[8.9]
where ν is Poisson’s ratio, which has a value between 0.25 and 0.4 for most materials. A strain gauge is a metal or semiconductor element whose resistance changes when under strain. We can derive the relationship between changes in resistance and strain by considering the factors which influence the resistance of the element. The resistance of an element of length l, cross-sectional area A and resistivity ρ (Figure 8.6) is given by: R=
ρl A
[8.10]
8.1 RESISTIVE SENSING ELEMENTS
157
Figure 8.6 Strain gauges.
In general with strain gauges ρ, l and A can change if the element is strained, so that the change in resistance ∆R is given by: ∆R =
A ∂R D A ∂R D A ∂R D ∆l + ∆A + ∆ρ C ∂l F C ∂A F C ∂ρ F
∆R =
ρ ρl l ∆l − 2 ∆A + ∆ρ A A A
[8.11]
i.e.
Dividing throughout by R = ρl/A yields ∆R ∆l ∆A ∆ ρ = − + R l A ρ
[8.12]
The ratio ∆l/l is the longitudinal strain eL in the element. Since cross-sectional area A = wt (Figure 8.6), ∆A ∆w ∆t = + = 2eT A w t where eT is the transverse strain in the element. From [8.9] and [8.12] we have: ∆R ∆ρ = eL − 2(−νeL ) + R ρ = (1 + 2ν)eL +
∆ρ ρ
[8.13]
We now define the gauge factor G of a strain gauge by the ratio (fractional change in resistance)/(strain), i.e. G=
∆R/R 0 e
158 S ENSING ELEMENTS Hence Resistance/strain relationship for a strain gauge
∆R = Ge R0
[8.14]
where R0 is the unstrained resistance of the gauge. From [8.13] the gauge factor is given by: Gauge factor of a strain gauge
G = 1 + 2ν +
1 ∆ρ e ρ
[8.15]
For most metals ν ≈ 0.3, and the term (1/e) (∆ ρ /ρ) representing strain-induced changes in resistivity (piezoresistive effect) is small (around 0.4), so that the overall gauge factor G is around 2.0. A popular metal for strain gauges is the alloy ‘Advance’; this is 54% copper, 44% nickel and 1% manganese. This alloy has a low temperature coefficient of resistance (2 × 10−5 °C −1) and a low temperature coefficient of linear expansion. Temperature is both an interfering and a modifying input (Section 2.2), and the above properties ensure that temperature effects on zero values and sensitivity are small. The most common strain gauges are of the bonded type, where the gauge consists of metal foil, cut into a grid structure by a photoetching process, and mounted on a resin film base. The film backing is then attached to the structure to be measured with a suitable adhesive. The gauge should be positioned so that its active axis is along the direction of the measured strain: the change in resistance, due to a given strain, along the passive axis is very small compared with that produced by the same strain along the active axis. A typical gauge has:
• • • • • •
Gauge factor 2.0 to 2.2 Unstrained resistance 120 ± 1 Ω Linearity within ±0.3% Maximum tensile strain +2 × 10−2 Maximum compressive strain −1 × 10−2 Maximum operating temperature 150 °C.
The change in resistance at maximum tensile strain is ∆R = +4.8 Ω, and ∆R = −2.4 Ω at maximum compressive strain. A maximum gauge current between 15 mA and 100 mA, depending on area, is specified in order to avoid self-heating effects. Unbonded strain gauges consisting of fine metal wire stretched over pillars are used in some applications. In semiconductor gauges the piezoresistive term (1/e) (∆ ρ /ρ) can be large, giving large gauge factors. The most common material is silicon doped with small amounts of p-type or n-type material. Gauge factors of between +100 and +175 are common for p-type silicon, and between −100 and −140 for n-type silicon. A negative gauge factor means a decrease in resistance for a tensile strain. Thus semiconductor gauges have the advantage of greater sensitivity to strain than metal ones, but have the disadvantage of greater sensitivity to temperature changes. Typically a rise in ambient temperature from 0 to 40 °C causes a fall in gauge factor from 135 to 120. Also the temperature coefficient of resistance is larger, so that the resistance of a typical unstrained gauge will increase from 120 Ω at 20 °C to 125 Ω at 60 °C. Strain gauge elements are incorporated in deflection bridge circuits (Section 9.1).
8.1 RESISTIVE SENSING ELEMENTS
8.1.4
159
Semiconductor resistive gas sensors Metal oxide sensors have semiconducting properties which are affected by the presence of gases. The resistance of chromium titanium oxide is affected by reducing gases such as carbon monoxide (CO) and hydrocarbons. Here oxygen atoms near the surface react with reducing gas molecules; this reaction takes up conduction electrons so that fewer are available for conduction. This causes a decrease in electrical conductivity and a corresponding increase in resistance. The resistance of tungsten oxide is affected by oxidising gases such as oxides of nitrogen (NOx ) and ozone. Here atoms near the surface react with oxidising gas molecules; this reaction takes up conduction electrons, again causing a decrease in electrical conductivity and an increase in resistance with gas concentration. To aid the oxidation/reduction process, these sensors are operated at elevated temperatures well above ambient temperature. Figure 8.7 shows a typical construction of a metal oxide sensor using thick film technology. This consists of an alumina substrate with a film of oxide printed on one side and a platinum heater grid on the other. A typical NOX sensor has an ambient temperature range of −20 °C to +60 °C and operating power of 650 mW. The resistance is typically 6 kΩ in air, 39 kΩ in 1.5 ppm NO2 and 68 kΩ in 5.0 ppm NO2.[5] A typical CO sensor has an ambient temperature range of −20 °C to +60 °C and an operating power of 650 mW. The resistance is typically 53 kΩ in air, 85 kΩ in 100 ppm CO and 120 kΩ in 400 ppm CO.[5] Gas sensing reactions occur at or near the surface of organic polymer semiconductors so that only a thin film of the material is required in a sensor. The metal phthalocyanines (MPc) have high sensitivities to certain gases. Thin films of MPc act as p-type semiconductors where electrical conductivity is due to positive holes. Electron acceptor gases such as chlorine and nitrogen dioxide are absorbed by the film and capture MPc electrons. This causes an increase in the population density of free holes and a corresponding increase in electrical conductivity, i.e. a decrease in resistance, with gas concentration. The metal M is chosen so that it has the greatest sensitivity to a particular gas. Thus lead has the highest sensitivity to nitrogen dioxide (NO2 ) and copper the highest sensitivity to hydrogen chloride (HCl). For a lead phthalocyanine film detecting NO2, the relationship between resistance R and concentration [NO2 ] is approximately logarithmic, i.e. log10 R = −K log10[NO2 ] + a
[8.16]
Here K and a are constants and concentrations are in the parts per billion (ppb) range (Figure 8.8(a)).[6] A given metal phthalocyanine will also respond to other gases in addition to the one it has been selected for: thus copper has maximum response to
Figure 8.7 Typical construction of metal oxide gas sensor.
160 S ENSING ELEMENTS
Figure 8.8 Organic polymer resistive gas sensors: (a) Characteristics of NO2 sensor (b) Typical construction of single sensor (c) Array of six sensors (after Trident Exhibitions and Dr A. Cranny[6]).
HCl, but also responds to NO2, Cl2 and H2S.[6] Thus the other gases act as environmental modifying and interfering inputs (Sections 2.1 and 19.2.4). Figure 8.8(b) shows the construction of a typical individual sensor. Here a thin film of the metal phthalocyanine is deposited onto a planar configuration of gold electrodes deposited on an alumina substrate.[6] Figure 8.8(c) shows an array of six sensors; each sensor is formed by depositing a thin film of a given MPc on a configuration of thick film platinum electrodes. By using a different metal M for each sensor, the concentration of each of six different gases in a mixture can in principle be found. Methods of compensation for modifying cross-sensitivity effects between the gases will need to be developed (Chapter 19).
8.2
Capacitive sensing elements The simplest capacitor or condenser consists of two parallel metal plates separated by a dielectric or insulating material (Figure 8.9). The capacitance of this parallel plate capacitor is given by: C=
ε 0ε A d
[8.17]
where ε0 is the permittivity of free space (vacuum) of magnitude 8.85 pF m−1, ε is the relative permittivity or dielectric constant of the insulating material, A m2 is the area of overlap of the plates, and d m is their separation. From [8.17] we see that C
8.2 CAPACITIVE SENSING ELEMENTS
Figure 8.9 Capacitive sensing elements.
Variable separation displacement sensor
161
can be changed by changing either d, A or ε ; Figure 8.9 shows capacitive displacement sensors using each of these methods. If the displacement x causes the plate separation to increase to d + x the capacitance of the sensor is: C=
ε 0ε A d+x
[8.18]
i.e. there is a non-linear relation between C and x. In the variable area type, the displacement x causes the overlap area to decrease by ∆A = wx, where w is the width of the plates, giving: Variable area displacement sensor
C=
ε 0ε ( A − wx ) d
[8.19]
In the variable dielectric type, the displacement x changes the amount of dielectric material ε 2 (ε 2 > ε 1 ) inserted between the plates. The total capacitance of the sensor
162 S ENSING ELEMENTS is the sum of two capacitances, one with area A1 and dielectric constant ε1, and one with area A2 and dielectric constant ε 2, i.e.
ε 0 ε 1 A1 ε 0 ε 2 A2 + d d
C=
Since A1 = wx, A2 = w(l − x), when w is the width of the plates, Variable dielectric displacement sensor
C=
ε 0w [ε 2l − (ε 2 − ε 1) x] d
[8.20]
A commonly used capacitive pressure sensor is shown in Figure 8.9. Here one plate is a fixed metal disc, the other is a flexible flat circular diaphragm, clamped around its circumference; the dielectric material is air (ε ≈ 1). The diaphragm is an elastic sensing element (Section 8.6) which is bent into a curve by the applied pressure P. The deflection y at any radius r is given by: y=
3 (1 − ν 2 ) 2 (a − r 2 )2P 16 Et 3
[8.21]
where a = radius of diaphragm t = thickness of diaphragm E = Young’s modulus ν = Poisson’s ratio. The deformation of the diaphragm means that the average separation of the plates is reduced. The resulting increase in capacitance ∆C is given by[7] Capacitive pressure sensor
∆C (1 − ν 2 )a 4 = P C 16 Edt 3
[8.22]
where d is the initial separation of the plates and C = ε 0 πa 2 /d the capacitance at zero pressure. The variable separation displacement sensor has the disadvantage of being nonlinear (eqn [8.18]). This problem is overcome by using the three-plate differential or push-pull displacement sensor shown in Figure 8.9. This consists of a plate M moving between two fixed plates F1 and F2; if x is the displacement of M from the centre line AB, then the capacitances C1 and C2 formed by MF1 and MF2 respectively are: Differential capacitive displacement sensor
C1 =
εε 0 A εε 0 A , C2 = d+x d−x
[8.23]
The relations between C 1, C 2 and x are still non-linear, but when C 1 and C 2 are incorporated into the a.c. deflection bridge described in Section 9.1.3, the overall relationship between bridge output voltage and x is linear (eqn [9.23]).
8.2 CAPACITIVE SENSING ELEMENTS
163
The next sensing element shown in Figure 8.9 is a level sensor consisting of two concentric metal cylinders. The space between the cylinders contains liquid to the height h of the liquid in the vessel. If the liquid is non-conducting (electrical conductivity less than 0.1 µmho cm−3), it forms a suitable dielectric and the total capacitance of the sensor is the sum of liquid and air capacitances. The capacitance/unit length of two coaxial cylinders, radii b and a (b > a), separated by a dielectric ε is 2π ε 0 ε /loge(b/a). Assuming the dielectric constant of air is unity, the capacitance of the level sensor is given by: Ch =
Capacitive level sensor Ch =
2π ε 0ε h  b loge    a 2π ε 0  b loge    a
+
2π ε 0 (l − h)  b loge    a
[l + (ε − 1)h]
[8.24]
The sensor can be incorporated into the a.c. deflection bridge of Figure 9.5(a). The final sensor shown in Figure 8.9 is a thin-film capacitive humidity sensor.[8] The dielectric is a polymer which has the ability to absorb water molecules; the resulting change in dielectric constant and therefore capacitance is proportional to the percentage relative humidity of the surrounding atmosphere. One capacitor plate consists of a layer of tantalum deposited on a glass substrate; the layer of polymer dielectric is then added, followed by the second plate, which is a thin layer of chromium. The chromium layer is under high tensile stress so that it cracks into a fine mosaic which allows water molecules to pass into the dieletric. The stress in the chromium also causes the polymer to crack into a mosaic structure. A sensor of this type has a input range of 0 to 100% RH, a capacitance of 375 pF at 0% RH and a linear sensitivity of 1.7 pF/% RH. The capacitance–humidity relation is therefore the linear equation: C = 375 + 1.7 RH pF The maximum departure from this line is 2% due to non-linearity and 1% due to hysteresis. The basic capacitive pressure sensor shown in Figure 8.9 has some limitations. It consists of two circular metal plates or diaphragms, one flexible, one rigid, with air as dielectric. Because the dielectric constant ε for air is only 1, changes in capacitance ∆C caused by the applied pressure P will be very small. To increase ∆C, the thickness t of the flexible plate/diaphragm can be reduced; this gives greater deformation y for a given pressure P but reduces mechanical strength. Also many metals have limited resistance to corrosion. These problems can be largely overcome by using liquid-filled ceramic membrane sensors. Dielectric liquids such as silicon oil have a much higher ε than air, so that significant capacitance changes ∆C can be obtained with small membrane displacements of the order of 10 µm. These small displacements, together with lower Young’s modulus for ceramics than metals, mean that ceramic membranes can be made much thicker than metal diaphragms – between 0.25 and 1.5 mm is typical. This gives greater mechanical strength and ease
164 S ENSING ELEMENTS Figure 8.10 Ceramic liquid-filled differential pressure sensor (after Endress and Hauser Ltd.).
of maintenance; furthermore, ceramics such as aluminium oxide have excellent resistance to corrosion. Figure 8.10 shows a differential pressure sensor using this technology. Process fluid at pressure p1 in the left-hand chamber and at pressure p2 in the right-hand chamber is in contact with the ceramic membrane. The membrane is supported on a ceramic substrate. Two capacitances C1 and C2 are formed using gold plates and silicon oil as dielectric. An increase in differential pressure p1 − p2 causes the lefthand membrane to move to the right, causing C1 to increase by ∆C. The pressure change is transmitted through the fluid to the right, causing the right-hand membrane to move to the right and C2 to decrease by ∆C. C1 and C2 are then incorporated in an a.c. bridge circuit (Sections 9.1.3 and 9.4.2). Capacitive sensing elements are incorporated in either a.c. deflection bridge circuits (Section 9.1) or oscillator circuits (Section 9.5.1). Capacitive sensors are not pure capacitances but have an associated resistance R in parallel, to represent losses in the dielectric; this has an important influence on the design of circuits, particularly oscillator circuits. For example, a typical capacitive humidity sensor would have an approximate parallel dielectric loss resistance R of 100 kΩ at 100 kHz. The quality of a dielectric is often expressed in terms of its ‘loss tangent’ tan δ where: tan δ =
1 ω CR
In this example, if C = 500 pF then tan δ ≈ 0.03. Figure 9.25(b) shows the capacitive sensor incorporated with a pure inductance into an oscillator circuit. The ‘quality factor’ Q of the circuit is given by Q=R
C = ω nCR L
and thus depends crucially on R. If the natural frequency fn of the circuit with the above humidity sensor is fn = 105 Hz, then the circuit Q factor is approximately 30.
8.3 INDUCTIVE SENSING ELEMENTS
165
Precautions should also be taken to minimise the effect of cable and stray capacitances on these circuits.
8.3 8.3.1
Inductive sensing elements Variable inductance (variable reluctance) displacement sensors In order to discuss the principles of these elements we must first introduce the concept of a magnetic circuit. In an electrical circuit an electromotive force (e.m.f.) drives a current through an electrical resistance and the magnitude of the current is given by e.m.f. = current × resistance
[8.25]
A simple magnetic circuit is shown in Figure 8.11(a): it consists of a loop or core of ferromagnetic material on which is wound a coil of n turns carrying a current i. By analogy we can regard the coil as a source of magnetomotive force (m.m.f.) which drives a flux φ through the magnetic circuit. The equation corresponding to [8.25] for a magnetic circuit is: m.m.f. = flux × reluctance = φ × 0002
[8.26]
so that reluctance 0002 limits the flux in a magnetic circuit just as resistance limits the current in an electrical circuit. In this example m.m.f. = ni, so that the flux in the magnetic circuit is:
φ=
ni weber 0002
[8.27]
This is the flux linked by a single turn of the coil; the total flux N linked by the entire coil of n turns is: N = nφ =
n2i 0002
[8.28]
By definition the self-inductance L of the coil is the total flux per unit current, i.e. Self-inductance of a coil
L=
N n2 = 0002 i
[8.29]
The above equation enables us to calculate the inductance of a sensing element given the reluctance of the magnetic circuit. The reluctance 0002 of a magnetic circuit is given by: 0002=
l µµ 0 A
[8.30]
where l is the total length of the flux path, µ is the relative permeability of the circuit material, µ 0 is the permeability of free space = 4π × 10−7 H m−1 and A is the
166 S ENSING ELEMENTS Figure 8.11 Variable reluctance elements: (a)(b) Basic principle of reluctance sensing elements (c) Reluctance calculation for typical element (d) Differential or push/pull reluctance displacement sensor.
cross-sectional area of the flux path. Figure 8.11(b) shows the core separated into two parts by an air gap of variable width. The total reluctance of the circuit is now the reluctance of both parts of the core together with the reluctance of the air gap. Since the relative permeability of air is close to unity and that of the core material many thousands, the presence of the air gap causes a large increase in circuit reluctance and a corresponding decrease in flux and inductance. Thus a small variation in air gap causes a measurable change in inductance so that we have the basis of an inductive displacement sensor.
8.3 INDUCTIVE SENSING ELEMENTS
167
Figure 8.11(c) shows a typical variable reluctance displacement sensor, consisting of three elements: a ferromagnetic core in the shape of a semitoroid (semicircular ring), a variable air gap and a ferromagnetic plate or armature. The total reluctance of the magnetic circuit is the sum of the individual reluctances, i.e. 0002 TOTAL = 0002CORE + 0002GAP + 0002ARMATURE
[8.31]
The length of an average, i.e. central, path through the core is πR and the crosssectional area is πr 2, giving: 0002CORE =
πR R = 2 µ 0 µC πr µ 0 µC r 2
[8.32]
The total length of the flux path in air is twice the air gap, i.e. 2d; also if there is little bending or fringing of the lines of flux in the air gap, then the cross-sectional area of the flux path in air will be close to that of the core. Assuming the relative permeability of air is unity, 0002GAP =
2d µ 0 πr 2
[8.33]
The length of an average central flux path in the armature is 2R; the calculation of the appropriate cross-sectional area is more difficult. A typical flux distribution is shown in Figure 8.11(c) and for simplicity we assume that most of the flux is concentrated within an area 2rt, giving 0002 ARMATURE =
2R R = µ 0 µA 2rt µ 0 µA rt
[8.34]
Thus 0002 TOTAL =
R 2d R + + 2 2 µ 0 µC r µ 0 πr µ 0 µA rt
i.e. 0002 TOTAL = 0002 0 + kd
[8.35]
where 00020 =
k=
R  1 1  +   = reluctance at zero air gap µ 0 r  µ C r µ A t  2 µ 0 πr 2
A typical element with n = 500 turns, R = 2 cm, r = 0.5 cm, t = 0.5 cm, µ C = µ A = 100, has 0002 0 = 1.3 × 107 H −1, k = 2 × 1010 H −1 m−1. This gives L = 19 mH at d = 0 (zero air gap) and L = 7.6 mH at d = 1 mm. From [8.29] and [8.35] we have Inductance of reluctance displacement sensor
L=
n2 L0 = 0002 0 + kd 1 + α d
[8.36]
168 S ENSING ELEMENTS where L 0 = n 2 /0002 0 = inductance at zero gap and α = k/0002 0. Equation [8.36] is applicable to any variable reluctance displacement sensor; the values of L0 and α depend on core geometry and permeability. We see that the relationship between L and d is non-linear, but this problem is often overcome by using the push-pull or differential displacement sensor shown in Figure 8.11(d). This consists of an armature moving between two identical cores, separated by a fixed distance 2a. From eqn [8.36] and Figure 8.11(d) we have: Differential reluctance displacement sensor
L1 =
L0 , 1 + α (a − x )
L2 =
L0 1 + α (a + x )
[8.37]
The relationship between L1, L2 and displacement x is still non-linear, but if the sensor is incorporated into the a.c. deflection bridge of Figure 9.5(b), then the overall relationship between bridge out of balance voltage and x is linear (eqn [9.25]). A typical sensor of this type would have an input span of 12 mm, a coil inductance (L0) of 25 mH, a coil resistance of 70 Ω and a maximum non-linearity of 0.5%. Thus inductive sensors are not pure inductances but have an associated resistance R in series; this has an important influence on the design of oscillator circuits (Figure 9.25(a)).
8.3.2
Figure 8.12 LVDT and connections to phase-sensitive detector.
Linear Variable Differential Transformer (LVDT) displacement sensor This sensor is a transformer with a single primary winding and two identical secondary windings wound on a tubular ferromagnetic former (Figure 8.12). The primary winding is energised by an a.c. voltage of amplitude ÄP and frequency f Hz; the two secondaries are connected in series opposition so that the output voltage ÄOUT sin(2π ft + φ) is the difference (V1 − V2 ) of the voltages induced in the secondaries. A ferromagnetic core or plunger moves inside the former; this alters the mutual inductance between the primary and secondaries. With the core removed the secondary voltages are ideally equal so that ÄOUT = 0. With the core in the former, V1 and V2 change with core position x, causing amplitude ÄOUT and phase φ to change.
8.3 INDUCTIVE SENSING ELEMENTS
Figure 8.13 LVDT secondary waveforms.
169
Figure 8.13 shows a.c. waveforms for secondary voltages V1, V2 and the difference V1 − V2 at three positions A, C and B. C is the halfway point at x = –12 l and A and B are equidistant from C. Figure 8.13(a) shows the waveforms at A; here the lower secondary is strongly coupled to the primary and the upper secondary is weakly coupled to the primary; V2 has therefore greater amplitude than V1 and V1 − V2 is 180° out of phase with the primary voltage VP . Figure 8.13(b) shows the waveforms at C; here the secondaries are equally coupled to the primary and V1 and V2 have approximately equal amplitudes; V1 − V2 has therefore minimum amplitude (ideally zero) and C is termed the null point. Figure 8.13(c) shows the waveforms at B; here the upper secondary is strongly coupled to the primary and the lower secondary is weakly coupled to the primary; V1 has greater amplitude than V2 and V1 − V2 is in phase with the primary voltage. Therefore V1 − V2 has the same amplitude at positions A and B but there is a phase difference of 180°. The a.c. voltage V1 − V2 is converted to d.c. in a way which distinguishes between positions such as A and B either side of the null point C. This is done using a phase sensitive detector or demodulator (Section 9.3). This senses the 180° phase difference and gives a negative d.c. voltage at A and a positive d.c. voltage of equal magnitude at B. Figure 8.14 shows how the amplitude ÄOUT and phase φ of the a.c. output and the d.c. voltage VDC vary with displacement x. The relationship between VDC and x is linear over the central portion of the range 0 to l, but non-linear effects
170 S ENSING ELEMENTS Figure 8.14 A.C. and D.C. characteristics of LVDT.
occur at either end (D and E) as the core moves to the edge of the former. This nonlinearity can be reduced by using only a central portion of the available range. LVDT displacement sensors are available to cover ranges from ±0.25 mm to ±25 cm. For a typical sensor of range ±2.5 cm, the recommended ÄP is 4 to 6 V, the recommended f is 5 kHz (400 Hz minimum, 50 kHz maximum), and maximum nonlinearity is 1% f.s.d. over the above range.
8.4
Electromagnetic sensing elements These elements are used for the measurement of linear and angular velocity and are based on Faraday’s law of electromagnetic induction. This states that if the flux N linked by a conductor is changing with time, then a back e.m.f. is induced in the conductor with magnitude equal to the rate of change of flux, i.e. E=−
dN dt
[8.38]
In an electromagnetic element the change in flux is produced by the motion being investigated; this means that the induced e.m.f. depends on the linear or angular velocity of the motion. A common example of an electromagnetic sensor is the variable reluctance tachogenerator for measuring angular velocity (Figure 8.15). It consists of a toothed wheel of ferromagnetic material (attached to the rotating shaft) and a coil wound onto a permanent magnet, extended by a soft iron pole piece. The wheel moves in close proximity to the pole piece, causing the flux linked by the coil to change with time, thereby inducing an e.m.f. in the coil. The magnitude of the e.m.f. can be calculated by considering the magnetic circuit formed by the permanent magnet, air gap and wheel. The m.m.f. is constant with time and depends on the field strength of the permanent magnet. The reluctance of the circuit will depend on the width of the air gap between the wheel and pole piece. When a tooth is close to the pole piece the reluctance is minimum but will increase as the tooth moves away. The reluctance is maximum when a ‘gap’ is adjacent to the pole
8.4 E LE CT R O MAGN E T I C S E N S I N G E LE ME N TS
171
Figure 8.15 Variable reluctance tachogenerator, angular variations in reluctance and flux.
piece but falls again as the next tooth approaches the pole piece. Figure 8.15 shows the resulting cyclic variation in reluctance 0002 with angular rotation θ. From eqn [8.26] the flux in the circuit is given by φ = m.m.f./0002, and the total flux N linked by a coil of n turns is: N = nφ =
n × m.m.f. 0002
[8.39]
so that N ∝ 1/0002. The corresponding variation of flux N with θ is also shown in Figure 8.15. We see that a reluctance minimum corresponds to a flux maximum and vice versa. This relation may be approximated by N(θ ) ≈ a + b cos mθ
[8.40]
where a is the mean flux, b is the amplitude of the flux variation and m is the number of teeth. Using eqn [8.38] the induced e.m.f. is given by: E=−
dN dN dθ =− × dt dθ dt
where dN = −bm sin mθ dθ dθ = ω r (angular velocity of wheel) dt and
θ = ω r t (assuming θ = 0 at t = 0)
[8.41]
172 S ENSING ELEMENTS Thus Output signal for variable reluctance tachogenerator
E = bmω r sin mω r t
[8.42]
This is a sinusoidal signal of amplitude : = bmω r and frequency f = mω r /(2π); i.e. both amplitude and frequency are proportional to the angular velocity of the wheel. In principle ω r can be found from either the amplitude or the frequency of the signal. In practice the amplitude measured by a distant recorder or indicator may be affected by loading effects (Section 5.1) and electrical interference (Section 6.3). A frequency system is therefore preferred (Chapter 10), where the number of cycles in a given time is counted to give a digital signal corresponding to the angular velocity. A variable reluctance tachogenerator is incorporated in the turbine flowmeter (Section 12.2) to give an accurate measurement of volume flow rate or total volume of fluid.
8.5
Thermoelectric sensing elements Thermoelectric or thermocouple sensing elements are commonly used for measuring temperature. If two different metals A and B are joined together, there is a difference in electrical potential across the junction called the junction potential. This junction potential depends on the metals A and B and the temperature T °C of the junction, and is given by a power series of the form: E TAB = a1T + a 2T 2 + a 3T 3 + a4T 4 + . . .
[8.43]
The values of constants a1, a2, etc., depend on the metals A and B. For example, the first four terms in the power series for the e.m.f. of an iron v. constantan (Type J ) junction are as follows,[9] expressed in µV: E TJ = 5.037 × 10+1T + 3.043 × 10−2T 2 − 8.567 × 10−5T 3 + 1.335 × 10−7T 4 + higher-order terms up to T 7
[8.44]
A thermocouple is a closed circuit consisting of two junctions (Figure 8.16), at different temperatures T1 and T2 °C. If a high-impedance voltmeter is introduced into the circuit, so that current flow is negligible, then the measured e.m.f. is, to a close approximation, the difference of the junction potentials, i.e. ETAB1,T 2 = ETAB1 − ETAB2 = a1(T1 − T2 ) + a 2(T 21 − T 22 ) + a3(T 31 − T 23 ) + . . .
[8.45]
Thus the measured e.m.f. depends on the temperatures T1, T2 of both junctions. In the following discussion T1 will be the temperature to be measured, i.e. the temperature of the measurement junction, and T2 will be the temperature of the reference
8.5 T H E R MO E LE CTRI C S E N S I N G E LE ME N T S
173
Figure 8.16 Thermocouple principles.
junction. In order to accurately infer T1 from the measured e.m.f., the reference junction temperature T2 must be known. Figure 8.16 summarises five ‘laws’ of thermocouple behaviour which are vital in temperature measurement.[10] Law 1 states that the e.m.f. of a given thermocouple depends only on the temperatures of the junctions and is independent of the temperatures of the wires connecting the junctions. This is important in industrial installations, where the leads connecting measurement and reference junctions may be exposed to large changes in ambient temperature. Law 2 states that if a third metal C is introduced into A (or B) then, provided the two new junctions are at the same temperature (T3), the e.m.f. is unchanged. This means that a voltmeter can be introduced into the circuit without affecting the voltage produced. If a third metal C is inserted between A and B at either junction, then Law 3 states that, provided the two new junctions AC and CB are both at the same temperature (T1 or T2 ), then the e.m.f. is unchanged. This means that at the measurement junction, wires A and B can be soldered or brazed together with a third metal without affecting the e.m.f. A voltage measuring device can be introduced at the reference junction again without affecting the measurement. Law 4 (law of intermediate metals) can be used, for example, to deduce the e.m.f. of a copper–iron (AB) thermocouple, given the e.m.f. values for copper–constantan (AC) and constantan–iron (CB) thermocouples. The fifth law (law of intermediate temperatures) is used in interpreting e.m.f. measurements. For a given pair of metals we have:
174 S ENSING ELEMENTS Law of intermediate temperatures
ET 1,T 2 = ET 1,T 3 + ET 3,T 2
[8.46]
where T3 is the intermediate temperature. If T2 = 0 °C, then ET 1,0 = ET 1,T 3 + ET 3,0
[8.47]
Suppose that we wish to measure the temperature T1 °C of a liquid inside a vessel with a chromel v. alumel thermocouple. The measurement junction is inserted in the vessel and the reference junction is outside the vessel, where the temperature is measured to be 20 °C, i.e. T3 = 20 °C. The measured e.m.f. is 5.3 mV using a voltmeter inserted at the reference junction, i.e. ET1,T 3 = ET 1,20 = 5.3 mV. The value of ET 3,0 = E20,0 is found to be 0.8 mV using thermocouple tables.[9] These tables give the e.m.f. ET,0 for a particular thermocouple, with measured junction at T °C and reference junction at 0 °C. From eqn [8.47], we have ET 1,0 = 5.3 + 0.8 = 6.1 mV; T1 is then found to be 149 °C by looking up the temperature corresponding to 6.1 mV. The importance of correct installation of thermocouples can be illustrated by the problem shown in Figure 8.17. Here we wish to measure the temperature of highpressure steam in a pipe, at around 200 °C, with a chromel v. alumel thermocouple, for which E200,0 = 8.1 mV. Figure 8.17 Thermocouple installations.
8.5 T H E R MO E LE CTRI C S E N S I N G E LE ME N T S
175
Installation (a) with the meter located just outside the pipe is completely useless. The reference junction temperature T2 can vary widely from sub-zero temperatures in cold weather to possibly +50 °C if a steam leak occurs; the measured e.m.f. is therefore meaningless. Installation (b) with the meter located in the control room and connected to the thermocouple with copper leads is equally useless – the reference junction is still located outside the pipe. In installation (c), the thermocouple is extended to the control room using extension or compensation leads made of chromel and alumel. This is an improvement because the reference junction is now in the control room where the variation in ambient temperature is smaller, possibly 10 °C at most. However, this is still not satisfactory for most applications, and one obvious solution is to place the reference in a temperature-controlled environment, e.g. a refrigerator at 0 °C. An alternative solution which utilises the law of intermediate temperatures is shown in installation (d). The thermocouple e.m.f. is ET 1, T 2 for a measured junction temperature of T1 °C and a reference junction temperature of T2 °C (T2 around 20 °C). If we introduce a second source of e.m.f. of magnitude ET 2 ,0 into the circuit in series with ET 1,T 2, then the voltmeter measures ET 1,T 2 + ET 2 ,0 which is equal to ET 1,0. Thus the voltmeter measures an e.m.f. relative to an apparent fixed reference temperature of 0 °C, even though the actual reference temperature is varying about a mean of 20 °C. The e.m.f. source producing ET 2 ,0 is known as an automatic reference junction compensation circuit (ARJCC). From eqn [8.45] we have: ET 2 ,0 = a1T2 + a2T 22 + a3T 32 + . . .
[8.48]
but since T2 is small, we can approximate by ET 2 ,0 ≈ a1T2. Thus we require a circuit giving a millivolt output signal proportional to reference temperature T2. This can be obtained with a metal resistance temperature sensor incorporated into a deflection bridge circuit, with a large value of R3 /R2 (Section 9.1). The output voltage of the bridge must be equal to ET 2 ,0, so that using eqn [9.15] we require: ET 2 ,0 = a1T2 = VS
R2 α T2 R3
i.e. Automatic reference junction compensation bridge
VS
R2 R3
=
a1
α
[8.49]
Thus any change in T2 which causes the thermocouple e.m.f. to alter is sensed by the metal resistive sensor, producing a compensating change in bridge output voltage. The thermocouple signal E T 1,0 is at a low level, typically a few millivolts, and often requires amplification prior to processing and presentation. The open-loop temperature transmitter described in Section 9.4.2 is often used to convert a thermocouple e.m.f. to a current signal in a standard range, e.g. 4 to 20 mA. In order to find an accurate estimate of T1 from E T 1,0 an inverse equation of the form T1 = a′1 ET 1,0 + a′2 E 2T 1,0 + a′3 ET3 1,0 + a 4′ ET4 1,0 + . . . should be solved using a microcontroller (Section 3.3 and Chapter 10).
[8.50]
176 S ENSING ELEMENTS Table 8.2 Thermocouple data and characteristics. e.m.f. values (µV)a
Tolerances
Extension leads
Characteristics
= 5 269 = 10 779 = 16 327 = 27 393
Class 1 −40 °C to +375 °C ±1.5 °C
As for thermocouple
Reducing atmospheres have little effect. Should be protected from moisture, oxygen and sulphur-bearing gases
E−100,0 E+100,0 E200,0 E400,0
= −3 379 = 4 279 = 9 288 = 20 872
Class 1 −40 °C to +125 °C ±0.5 °C
As for thermocouple
Recommended for low and sub-zero temperatures. Resists oxidising and reducing atmospheres up to approximately 350 °C. Requires protection from acid fumes
0 to +1100
E100,0 E250,0 E500,0 E1000,0
= 4 096 = 10 153 = 20 644 = 41 276
Class 1 −40 °C to +375 °C ±1.5 °C
As for thermocouple
Recommended for oxidising and neutral conditions. Rapidly contaminated in sulphurous atmospheres. Not suitable for reducing atmospheres
0 to +1600
E300,0 E600,0 E900,0 E1200,0
= 2 401 = 5 583 = 9 205 = 13 228
Class 1 0 to 1100 °C ±1.0 °C
Copper Copper–nickel
High resistance to oxidation and corrosion. Particularly susceptible to many metal vapours, therefore important that non-metal sheaths are used
Type
Temperature ranges (°C)
Iron v. constantan Type J
−20 to +700
E100,0 E200,0 E300,0 E500,0
Copper v. constantan Type T
−185 to +400
Nickel–chromium v. nickel–aluminium Chromel v. alumel Type K
Platinum v. platinum– 13% rhodium Type R
a
e.m.f. values are after International Standard IEC 584.1: 1995[9].
Figure 8.18 Mineral insulated thermocouple.
Table 8.2 summarises the measurement range, e.m.f. values, tolerances and characteristics of four thermocouples in common industrial use. The table can be used to quantify non-linearity: for example, a copper v. constantan thermocouple, used between 0 and +400 °C, has an e.m.f. of 9288 µV at 200 °C compared with an ideal straight line value of 10 436 µV. Thus the non-linearity at 200 °C is −1148 µV or −5.5% of f.s.d. Typical tolerances are of the order of ±1%, i.e. about 10 times greater than for platinum resistance thermometers. For base metal thermocouples (types J, T and K), the extension or compensation leads are made of the same metals as the thermocouple itself. For rare metal thermocouples (e.g. type R), the metals in the extension leads are copper and copper–nickel, which have similar thermoelectric properties to platinum and platinum–rhodium but are far cheaper. In order to give mechanical and chemical protection, the thermocouple is often enclosed in a thermowell or sheath (Figure 14.1). An alternative is the mineral insulated thermocouple (Figure 8.18). This is a complete package, where the space between the thermocouple wires and metal case is filled with material which is both a good heat conductor and an electrical insulator. The dynamic characteristics of both types of installation are discussed in Section 14.3.
8.6 ELASTIC SENSING ELEMENTS
8.6
177
Elastic sensing elements If a force is applied to a spring, then the amount of extension or compression of the spring is approximately proportional to the applied force. This is the principle of elastic sensing elements which convert an input force into an output displacement. Elastic elements are also commonly used for measuring torque, pressure and acceleration, which are related to force by the equations: torque = force × distance pressure =
force area
acceleration =
force mass
[8.51]
In a measurement system an elastic element will be followed by a suitable secondary displacement sensor, e.g. potentiometer, strain gauge or LVDT, which converts displacement into an electrical signal. The displacement may be translational or rotational. Elastic sensing elements have associated mass (inertance) and damping (resistance) as well as spring characteristics. In Section 4.1.2 the dynamics of a mass–spring– damper force sensor were analysed, and shown to be represented by the second-order transfer function: G( s ) =
1 1 2 2ξ s + s +1 ω n2 ωn
Figure 8.19 shows dynamic models of elastic elements for measuring linear acceleration, torque, pressure and angular acceleration. The dynamics of all four elements are represented by second-order differential equations and transfer functions. The figure gives the values of steady-state gain K, natural frequency ω n and damping ratio ξ for each type of sensor. The differential equation for a torque sensor is an analogue of eqn [4.14] for a force sensor; the torque equation involves rotational motion whereas the force equation involves translational motion. Here an input torque T is opposed by the spring torque cθ and damping torque b2; the resultant unbalanced torque is equal to the product of moment of inertia I and angular acceleration 3. In the pressure sensor, the input pressure P produces a force AP over the area A of the bellows which is opposed by the bellows spring force kx and damping force λ x. The conceptual model of an accelerometer, shown in Figure 8.19, is a casing containing a ‘seismic’ mass which moves on frictionless guide rails against a spring. If the casing is given an acceleration a, then the mass m experiences an inertia force ma in an opposite direction. This is the same force a car driver experiences when he is thrown back against his seat as the car accelerates, or thrown forwards during sudden braking. This force is opposed by the spring force kx, where x is the displacement of the mass relative to the casing. Under steady conditions at t = 0– these two forces are in equilibrium, thus: ma 0− = kx0−
[8.52]
178 S ENSING ELEMENTS
Figure 8.19 Dynamic models of elastic elements: (a) Linear accelerometer (b) Pressure sensor (c) Angular accelerometer (d) Torque sensor.
i.e. steady-state sensitivity K=
x0− m 1 = = 2 a 0− k ω n
If the input acceleration is suddenly increased to a at time t = 0−, then the forces acting on the mass are no longer in equilibrium and the mass moves relative to the casing, i.e. velocity x and acceleration y are non-zero. The resultant unbalanced force is ma − kx − λ x, giving: ma − kx − λ x = my i.e. my + λ x + kx = ma
[8.53]
where we note that the acceleration y of the mass relative to the casing is completely different from the acceleration a of the casing itself. By defining deviation variables ∆x and ∆a, we can use eqns [8.52] and [8.53] to derive the element transfer function (see Section 4.1). The differential equation for an angular acceleration sensor is the rotational analogue of [8.53]; here the inertia torque Iφ, due to the input acceleration φ, is opposed by spring torque cθ and damping torque b2. In Chapter 4 we saw that the optimum value of damping ratio ξ is 0.7; this ensures minimum settling time for step response and |G( jω)| closest to unity for the frequency response. Many practical force sensors and accelerometers incorporate liquid or electromagnetic damping in order to achieve this value. However, elements such as diaphragm or bellows pressure sensors with only air damping have a very low damping ratio, typically ξ = 0.1 or less. In this case, the natural frequency ω n of the sensor must be several times greater than the highest signal frequency ω MAX, in order to limit the variation in |G( jω)| to a few per cent.
8.6 ELASTIC SENSING ELEMENTS
179
Suppose we wish to measure pressure fluctuations containing frequencies up to 10 Hz, keeping the dynamic error within ±2%, using an elastic pressure element with ξ = 0.1. We thus require: 0.98 ≤ |G( jω)| ≤ 1.02 for ω ≤ 62.8
[8.54]
Since ξ is small and ω < ω n, eqn [4.40] for |G( jω)| can be approximated to: |G( jω)| ≈
1 1 − ω 2 /ω 2n
(ξ small, ω /ω n 0002 1)
[8.55]
In this situation |G( jω)| cannot be less than unity, so that the above condition simplifies to: 1 ≤ 1.02 for 1 − ω 2 /ω 2n
ω ≤ 62.8
[8.56]
Thus
ω ≤ 0.14 for ω ≤ 62.8, i.e. ωn ≥ 450 ωn Since ω n = êkú/úmú , ω n can be increased by increasing the ratio k/m, giving a high stiffness, low mass sensor. However, increasing k reduces the steady-state sensitivity K = A/k of the pressure element, so that a more sensitive secondary displacement sensor is required. The design of any elastic sensing element is therefore a compromise between the conflicting requirements of high steady-state sensitivity and high natural frequency. Figure 8.20 shows four practical elastic sensing elements. All four elements are fairly stiff, i.e. k and ω n are high, but steady-state sensitivity K and displacement x are small, so that strain gauges are used as secondary displacement sensors. In the cantilever force element or load cell, the applied force F causes the cantilever to bend so that the top surface experiences a tensile strain +e and the bottom surface an equal compressive strain −e. The magnitude of strain e is given by: e =
6(l − x) F wt 2E
[8.57]
where E is Young’s modulus for the cantilever material and the other quantities are defined in Figure 8.20. Strain gauges 1 and 3 sense a tensile strain +e so that their resistance increases by ∆R. Gauges 2 and 4 sense a compressive strain −e so that their resistance decreases by an equal amount. From eqn [8.14], ∆R = GR 0 e
[8.58]
where G is the gauge factor and R 0 the unstrained resistance of the gauges. This gives: R1 = R3 = R 0 + ∆R = R 0(1 + Ge) R2 = R4 = R 0 − ∆R = R 0(1 − Ge)
[8.59]
The four gauges are connected into a deflection bridge (see Figure 9.4(b) and eqn [9.18]).
180 S ENSING ELEMENTS Figure 8.20 Practical elastic sensing elements using strain gauges: (a) Cantilever load cell (b) Pillar load cell (c) Torque sensor (d) Unbonded strain gauge accelerometer.
In the pillar load cell the applied force F causes a compressive stress −F/A, where A is the cross-sectional area of the pillar. This produces a longitudinal, compressive strain: eL = −
F AE
[8.60]
which is accompanied by a transverse, tensile strain (eqn [8.9]): eT = −νeL =
+ νF AE
[8.61]
where E and ν are respectively Young’s modulus and Poisson’s ratio for the pillar material. Strain gauges are bonded onto the pillar so that gauges 1 and 3 sense eT and gauges 2 and 4 sense eL; thus A G νF D R1 = R3 = R 0 + R 0GeT = R0 1 + C AE F A GF D R2 = R4 = R 0 + R 0GeL = R0 1 − C AE F
[8.62]
The output voltage of the deflection bridge is given by eqn [9.19]. Figure 8.20(c) shows a cylindrical shaft acting as a torque sensing element. The applied torque T produces a shear strain φ in the shaft and corresponding linear tensile and compressive strains on the shaft surface. Gauge 1 is mounted with its active axis at +45° to the shaft axis, where the tensile strain has a maximum value +e, and
8.6 ELASTIC SENSING ELEMENTS
181
gauge 2 at −45° to the shaft axis where the compressive strain has a maximum value −e. Gauges 3 and 4 are mounted at similar angles on the other side of the shaft and experience strains +e and −e respectively. This maximum strain is given by: e=
T πSa 3
[8.63]
where S is the shear modulus of the shaft material and a is the radius of the shaft. The resistances of the gauges are R1 = R3 = R 0(1 + Ge) and R2 = R4 = R 0(1 − Ge) and the corresponding bridge output voltage is given by eqn [9.18]. A simplified diagram of a practical accelerometer using four unbonded strain gauges is shown in Figure 8.20(d). The space between the seismic mass and casing is filled with liquid to provide damping. The unbonded strain gauges are stretched fine metal wires, which provide the spring restoring force as well as acting as secondary displacement sensors. The gauges are prestressed, so that at zero acceleration each gauge experiences a tensile strain e0 and has a resistance R 0(1 + Ge0). If the casing is given an acceleration a, then the resultant displacement of the seismic mass m relative to the casing is (eqn [8.52]): x=
1 m a= 2a ωn k
[8.64]
where k is the effective stiffness of the strain gauges. Gauges 1 and 3 increase in length from L to L + x, and gauges 2 and 4 decrease in length from L to L − x. The tensile strain in gauges 1 and 3 increases to e0 + e, and that in gauges 2 and 4 decreases to e0 − e, where: e=
a x = L ω 2n L
[8.65]
The four gauges are connected into a deflection bridge circuit with out of balance voltage given by eqn [9.20]. In order to ensure that all four gauges are kept in tension over the whole range of movement of the mass, the maximum acceleration induced strain is only one-half of the initial strain, i.e. e MAX =
a MAX e0 = ω 2n L 2
or
a MAX =
e0 Lω 2n 2
[8.66]
Thus the acceleration input span is proportional to the square of the natural frequency. A family of accelerometers of this type, using 350 Ω gauges, cover the ranges ±5 g to ±500 g with natural frequencies between 300 and 3000 Hz and a damping ratio of 0.7 ± 0.1.[11] Figure 8.21(a) shows elastic pressure elements in current use. The bourdon tubes (7)–(9) are characterised by low stiffness and low natural frequency but large displacement sensitivity. A typical C-shaped tube of 50 mm bending diameter has a typical displacement travel of up to 4 mm, and is used with a potentiometer displacement sensor. C-shaped tubes have an approximately rectangular cross-section, are usually made from brass or phosphor-bronze, and cover pressure ranges from 35 kPa up to 100 MPa. The membranes, diaphragms and capsules (1)–(5) are far stiffer devices with higher natural frequency but lower displacement sensitivity. These elements are used with capacitive (Section 8.2) or strain gauge elements which are capable of sensing small displacements of 0.1 mm or less.
182 S ENSING ELEMENTS
Figure 8.21 Elastic pressure sensing elements: (a) Different types (after Neubert[12]) (b) Flat circular diaphragm.
The performance equations for a flat circular diaphragm, clamped around its circumference, are summarised in Figure 8.21(b). We see that the relation between applied pressure P and centre deflection yc is in general non-linear, but this nonlinearity can be minimised by keeping the ratio yc /t small. Thus if the non-linearity is to be limited to 1% maximum we require: 2
0.5
A yc D 1 ≤ C tF 100
or
yc ≤ 0.14t
[8.67]
The equations demonstrate that high steady-state sensitivity K and high natural frequency fn cannot both be achieved. A higher sensitivity can be achieved by reducing the effective stiffness γ = Et 2/[3a4(1 − ν 2 )], but since f n is proportional to êγ this has the effect of reducing the natural frequency.
8.7
Piezoelectric sensing elements If a force is applied to any crystal, then the crystal atoms are displaced slightly from their normal positions in the lattice. This displacement x is proportional to the applied force F: i.e., in the steady state, x=
1 F k
[8.68]
The stiffness k of the crystal is large, typically 2 × 109 N m−1. The dynamic relation between x and F can be represented by the second-order transfer function: ∆P 1/k ( s) = ∆, 1 2 2ξ s + s +1 ω n2 ωn
[8.69]
8.7 PIEZOELECTRIC SENSING ELEMENTS
183
where ωn = 2π fn is large, typically fn = 10 to 100 kHz, and ξ is small, typically ξ ≈ 0.01. In a piezoelectric crystal, this deformation of the crystal lattice results in the crystal acquiring a net charge q, proportional to x, i.e. q = Kx
[8.70]
Thus from [8.68] and [8.70] we have Direct piezoelectric effect
K F = dF k
q=
[8.71]
where d = K/k C N −1 is the charge sensitivity to force. Thus a piezoelectric crystal gives a direct electrical output, proportional to applied force, so that a secondary displacement sensor is not required. The piezoelectric effect is reversible; eqn [8.71] represents the direct effect where an applied force produces an electric charge. There is also an inverse effect where a voltage V applied to the crystal causes a mechanical displacement x, i.e. Inverse piezoelectric effect
x = dV
[8.72]
The inverse effect is important in ultrasonic transmitters (Chapter 16). The dimensions of d in eqn [8.71], i.e. C N −1, are identical with the dimensions m V −1 of d in eqn [8.72]. In order to measure the charge q, metal electrodes are deposited on opposite faces of the crystal to give a capacitor. The capacitance of a parallel plate capacitor formed from a rectangular block of crystal of thickness t (Figure 8.9 and eqn [8.17]) is given by: CN =
ε 0ε A t
[8.73]
The crystal can therefore be represented as a charge generator q in parallel with a capacitance CN, or alternatively by a Norton equivalent circuit (Section 5.1.3) consisting of a current source iN in parallel with CN. The magnitude of iN is: iN =
dq dx =K dt dt
[8.74]
or, in transfer function form: ∆jN (s) = Ks ∆P
[8.75]
where d /dt is replaced by the Laplace operator s. We note that for a steady force F, F and x are constant with time, so that d x/dt and iN are zero. If the piezoelectric sensor is connected directly to a recorder (assumed to be a pure resistive load RL ) by a cable (assumed to be a pure capacitance CC ) then the complete equivalent circuit is shown in Figure 5.11. The transfer function relating recorder voltage VL to current iN is shown to be:
184 S ENSING ELEMENTS ∆L (s) RL = ∆jN (s) 1 + RL(CN + CC )s
[5.16]
The overall system transfer function relating recorder voltage VL to input force F is: ∆L ∆L ∆jN ∆P (s) = ∆, ∆jN ∆P ∆,
[8.76]
From [8.69], [8.75] and [5.16] we have: ∆ L RL 1/k ( s) = Ks ∆, 1 + RL (C N + CC )s 1 2 2ξ s + s +1 ω n2 ωn =
Transfer function for basic piezoelectric force measurement system
1 1 RL (C N + CC )s k (C N + CC ) 1 + RL (C N + CC )s 1 2 2ξ s + s +1 ω n2 ωn
K
∆VL d τs 1 (s) =  ∆, (C N + CC ) (1 + τ s)  1 2 2ξ s + 1  2s + ωn  ωn
[8.77]
where τ = RL(CN + CC). The above transfer function emphasises the two disadvantages of this basic system: (a) (b)
The steady-state sensitivity is equal to d/(CN + CC ). Thus the system sensitivity depends on the cable capacitance CC , i.e. on the length and type of cable. The dynamic part of the system transfer function (ignoring recorder dynamics) is: G( s ) =
1 τs  (τ s + 1)  1 2 2ξ s + 1  2s + ωn  ωn
[8.78]
The second term is characteristic of all elastic elements and cannot be avoided; however, it causes no problems if the highest signal frequency ω MAX is well below ωn. The first term τ s/(τ s + 1) means that the system cannot be used for measuring d.c. and slowly varying forces. To illustrate this we can plot the frequency response characteristics |G( jω)| and arg G( jω) for a typical system (Figure 8.22). Here Amplitude ratio | G( jω )| =
τω 1 + τ 2ω 2
1
[8.79] 2
2  ω2 2ω 1 − 2  + 4ξ 2 ωn ωn 
Phase difference arg G( jω) = 90° − tan−1(ωτ) − tan−1
A 2ξ(ω /ω n ) D C 1 − (ω 2/ω 2n ) F
8.7 PIEZOELECTRIC SENSING ELEMENTS
185
Figure 8.22 Approximate frequency response characteristics and equivalent circuit for a piezoelectric measurement system.
The τω / 1 + τ 2ω 2 term causes a low frequency roll-off so that |G( jω)| = 0 at ω = 0 and the system cannot be used for frequencies much below 1/τ. Using the values of Figure 5.11, i.e. CN = 1600 pF, CC = 600 pF, RL = 1 MΩ, τ = 2.2 × 10−3 s and 1/τ = 455 rad s−1 or 72 Hz, we find there is a useful operating range between 3/τ and 0.2ω n, i.e. 216 Hz and 5.4 kHz, where 0.95 ≤ |G( jω)| ≤ 1.05 and φ is close to zero. This is, however, insufficient for many applications. These disadvantages are largely overcome by introducing a charge amplifier into the system, as shown in Figure 8.22. This is an integrator which ideally gives an output proportional to ∫ iN dt, i.e. an output proportional to charge q, since iN = dq/dt. Thus the system gives a non-zero output for a steady-force input. From Figure 8.22 we have: i1 = iF + i−
[8.80]
and the charge on feedback capacitor CF is qF = CF (V− − VOUT)
[8.81]
For an ideal operational amplifier (Section 9.2) we have i+ = i− = 0 and V− = V+. In this case we have V− = V+ = 0 and iF = dqF /dt, so that:
186 S ENSING ELEMENTS
i1 = iF =
dqF dVOUT = −CF dt dt
[8.82]
Since there is zero potential drop across CN and CC , there is no current flow through either capacitor, so that: i1 = iN =
dq dt
[8.83]
Thus from [8.82] and [8.83] we have: Transfer characteristics of ideal charge amplifier
dVOUT 1 dq =− dt CF dt
i.e. VOUT = −
q CF
[8.84]
assuming that VOUT = 0 when q = 0. From eqns [8.84], [8.70] and [8.69], the overall transfer function for the force measurement system with a charge amplifier is: Transfer function for piezoelectric system with ideal charge amplifier
1 ∆ OUT d (s) =  ∆, CF  1 2 2 ξ s + 1  2s + ωn  ωn
[8.85]
The steady-state sensitivity is now d/CF, i.e. it depends only on the capacitance CF of the charge amplifier and is independent of transducer and cable capacitance. In the ideal case the τ s/(1 + τ s) term is not present, so that |G( jω)| = 1 at ω = 0. In practice a resistor RF must be connected across feedback capacitor CF to provide a path for d.c. current. This reintroduces a τF s/(1 + τF s) term into the system transfer function, where τF = RF CF. By making RF and CF large, the frequency response of the charge amplifier will extend down to well below 1 Hz. For example if RF = 108 Ω, CF = 104 pF and τ F = 1.0 s, then |G( jω)| = 0.95 at ω = 3 rad s−1, i.e. at f ≈ 0.5 Hz. Table 8.3 (based on Table 16.1) shows properties of four commonly used piezoelectric materials. All figures are approximate and may depend on the direction of the applied force. Quartz occurs naturally in rocks and is piezoelectric when cut along certain directions. PZT and barium titanate are man-made ceramics possessing ferroelectric properties. If a strong electric field (typically 500–1000 kV/cm) is applied Table 8.3 Properties of piezoelectric materials. Material
Charge sensitivity d pC N −1
Dielectric constant ε
Voltage sensitivity g × 10 −3 V m N −1
Young’s modulus E × 109 N m−2
Damping ratio ξ
Quartz PZT BaTi 2 O3 PVDF
2.3 110 78 23
4.5 1200 1700 12
50 10 5.2 230
80 80 80 2
2 × 10 −5 7 × 10 −3 1 × 10 −3 5 × 10 −2
8.7 PIEZOELECTRIC SENSING ELEMENTS
187
to the crystal, at elevated temperatures (typically 80–110 °C), then the molecular dipoles remain aligned after the removal of the electric field. The material is then said to be poled and possesses piezoelectric and pyroelectric (Section 15.5.1) properties. Polyvinylidene fluoride (PVDF) is a long chain semi-crystalline polymer of the repeat unit CH2–CF2. After poling, this again possesses piezoelectric properties and is available in thin sheets with thickness between 10 and 1000 µm.[13] From the table we see that the charge sensitivity d for PZT and BaTi2O3 is high; this is, however, offset for some applications by a high dielectric constant and therefore capacitance. This means that both ceramics have only moderate electric field sensitivity to force g. In contrast PVDF has a lower value of d but a much lower value of ε, giving the highest value of g of the four materials. Also PVDF has a much lower value of Young’s modulus E and therefore elastic stiffness k than the other materials, giving greater displacement for the same force. The damping ratio ξ is also much greater for PVDF, giving a much flatter frequency response. The natural frequency fn of PVDF film can vary from low MHz to GHz frequencies depending on film thickness. Piezoelectric elements are commonly used for the measurement of acceleration and vibration. Here the piezoelectric crystal is used in conjunction with a seismic mass m. If the accelerometer casing is given an acceleration a, then there is a resulting inertia force F = ma acting on the seismic mass and the piezoelectric crystal. This results in the crystal acquiring a charge q = dF, where d C N−1 is the crystal charge sensitivity to force. The resulting steady-state charge sensitivity of the piezoelectric accelerometer is: ∆q = dm C m−1 s2 ∆a
[8.86]
Usually charge sensitivity is expressed in C g−1, where 1 g = 9.81 m s−2, i.e. ∆q = 9.81dm C g−1 ∆a
[8.87]
If the accelerometer is used with a charge amplifier with feedback capacitance CF, the corresponding steady-state system voltage sensitivity is: ∆V 9.81dm = ∆a CF
V g−1
[8.88]
The dynamic and frequency response characteristics of piezoelectric accelerometers, with and without charge amplifiers, are as described earlier in this section. Figure 8.23(a) shows a compression mode accelerometer, in which a crystal disc is sandwiched between the seismic mass and the accelerometer casing. The whole is bolted together to produce a large static pre-load compression force on the crystal. Depending on direction, the acceleration to be measured causes the compression force on the crystal to be greater or less than the static pre-load value. Figure 8.23(b) shows a shear-type accelerometer; here a piezoelectric element in the shape of a truncated conical tube is sandwiched between a tapered central pillar and the outer seismic mass.[14] The acceleration to be measured causes corresponding shear forces in the piezoelectric element. A general purpose shear-mode accelerometer of this type has a charge sensitivity of 25 pC g−1, a capacitance CN of 1600 pF, a natural frequency fn of 27 kHz and a weight of 18 g.
188 S ENSING ELEMENTS Figure 8.23 Piezoelectric accelerometers: (a) Compression mode (b) Shear mode (after Purdy[14] ).
Consider a body, e.g. a part of a machine such as the casing of a pump or a compressor, which is executing sinusoidal vibrations with displacement amplitude n at frequency f Hz, i.e. displacement x = n sin 2π ft
m
[8.89] −1
velocity x = 2πfn cos 2π ft m s = υˆ cos 2π ft, where υˆ = 2πf n acceleration y = −4π2 f 2 n sin 2π ft m s−2 = −â sin 2π ft, where â = 4π 2 f 2 n
[8.90] [8.91]
In the monitoring of machine vibration it is important to measure at least one of the amplitudes n, υˆ, â of displacement, velocity and acceleration respectively. This is especially so where the machine has certain resonant frequencies; here the above amplitudes may be large. It is possible to use a displacement transducer, differentiating the output once to obtain velocity information and twice to obtain acceleration information. However, displacement transducers often have insufficient sensitivity, inadequate frequency response or too large a mass for these applications; moreover the operation of differentiation amplifies unwanted high frequency noise. A better solution is to use a piezoelectric accelerometer, which has sufficient sensitivity, adequate frequency response and low mass; the accelerometer output can be integrated once for velocity information and twice for displacement information. The operation of integration tends to attenuate high frequency noise.
8.8
Piezoresistive sensing elements In Section 8.1 the piezoresistive effect was defined as the change in resistivity ρ of a material with applied mechanical strain e, and is represented by the term (1/e) (∆ ρ /ρ) in the equation for gauge factor of a strain gauge. We saw that silicon doped with
8.8 PIEZORESISTIVE SENSING ELEMENTS
189
Figure 8.24 Piezoresistive pressure sensors (after Noble[16]).
small amounts of n- or p-type material exhibits a large piezoresistive effect and is used to manufacture strain gauges with high gauge factors. The traditional way of making diaphragm pressure sensors is to cement metal foil strain gauges onto the flat surface of a metal diaphragm. In piezoresistive pressure sensors the elastic element is a flat silicon diaphragm. The distortion of the diaphragm is sensed by four piezoresistive strain elements made by introducing doping material into areas of the silicon, where the strain is greatest. One method of introducing the doping material is diffusion at high temperatures; the resulting four strain gauges are connected into a deflection bridge circuit in the normal way. A typical sensor of this type has an input range of 0 to 100 kPa, a sensitivity of around 3 mV/kPa (for a 10 V bridge supply voltage), a natural frequency of 100 kHz and combined non-linearity and hysteresis of ±0.5%.[15] Figure 8.24(a) shows a piezoresistive pressure sensor where p-type doping material is introduced into an n-type silicon diaphragm using ion implantation technology.[16] Four piezoresistive strain elements are thus produced (two in tension, two in compression), which are connected into a deflection bridge circuit. Figure 8.24(b) shows how the element may be used to measure absolute, differential and gauge pressures. A typical sensor of this type has an input range of 0 to 100 kPa, a sensitivity of around 1 mV/kPa (for a 12 V bridge supply voltage), a 10% to 90% rise time of 100 µs and typical combined non-linearity and hysteresis of ±0.1%.
190 S ENSING ELEMENTS
8.9 8.9.1
Electrochemical sensing elements Ion selective electrodes Ion selective electrodes (ISEs) are sensors which directly measure the activity or concentration of ions in solution. They could, for example, be used to measure the concentration of lead, sodium or nitrate ions in drinking water. When an ISE is immersed in a solution, a reaction takes place between the charged species in the solution and those on the sensor surface. An equilibrium is then established between these species: there is a corresponding equilibrium potential difference between the sensor and solution, which depends mainly, but not entirely, on the activity of a single ion. This output signal depends also, to some extent, on the activity of other ions present in the solution; the electrodes are therefore selective rather than specific. For example, the output signal of a potassium electrode depends not only on the activity of potassium (K+) ions in solution, but on some fraction of the sodium (Na+) ions also present. The Thévenin or open circuit e.m.f. of ISEs for monovalent ions is given by the modified Nernst equation:
Thévenin e.m.f. for monovalent ion ISE
ETh = E0 +
where
Rθ loge (a X + K X/ Y a Y ) F
[8.92]
E0 = constant depending on electrode composition R = universal gas constant (8.314 J K−1) θ = absolute temperature (K) F = Faraday number (96 493 C) a X = activity of ion X in solution, e.g. K+ a Y = activity of ion Y in solution, e.g. Na+ KX/ Y = selectivity coefficient of X electrode to Y (0 < K < 1)
The smaller the value of KX/ Y the more selective the electrode is to X; typically KK + / Na+ ≈ 2.6 × 10−3. The above equation is applicable to any monovalent ion, i.e. where a single electron is involved in the reaction. If n electrons are involved in the reaction then eqn [8.92] should be modified by replacing Rθ /F with Rθ /nF. The activity aA of an ion A in solution depends on, but is not proportional to, the concentration CA of that ion. The activity of an ion of particular concentration in a given solution is a function not only of its own concentration, but also of the concentration of all other ions and species. The relation between aA and CA is of the form: aA = fACA
[8.93]
where fA is the activity coefficient for the ion A. In general fA < 1 and is a nonlinear function of CA. At low concentrations (<10 −4 mole litre−1), fA ≈ 1 and aA ≈ CA. Figure 8.25(a) shows the basic system for measuring the activity or concentration of an ion in solution. Because an ion-selective electrode is only a single electrode and therefore only one half of a complete electrochemical cell, it must be used with a suitable standard reference electrode, immersed in the same test solution.
8.9 ELECTROCHEMICAL SENSING ELEMENTS
191
Figure 8.25 Ion-selective electrodes: (a) Basic system for ion concentration measure and equivalent circuit (b) Calomel reference electrode (after Bailey[17], © P. L. Bailey 1980. Reprinted by permission of John Wiley & Sons Ltd) (c) Practical pH sensor (after Thompson[18]).
Reference electrodes are electrochemical half-cells whose potential is held at a constant value by the chemical equilibrium maintained inside them. Figure 8.25(a) also shows the Thévenin equivalent circuit for the system; the Thévenin resistance RTh is made up of the resistance of the ISE in series with the resistance of the reference electrode and can be as high as 109 Ω. This high resistance means that the potential difference ETh between ion selective and reference electrodes must be measured using either a high input impedance millivoltmeter or a buffer amplifier with Z IN ≈ 1012 Ω (Section 5.1). One commonly used reference electrode is the calomel electrode shown in Figure 8.25(b).[17] This consists of a mercury–mercurous chloride (Hg/Hg2Cl2)
192 S ENSING ELEMENTS electrode system which is in contact with a saturated solution of potassium chloride (KCl). The glass container has a porous ceramic plug; this provides a liquid junction, which allows ions to diffuse between the KCl and test solutions. An important industrial application of ISEs is the measurement of the activity of hydrogen ions in solution. Acidity and alkalinity are quantified using pH where: pH = −log10 a H + = −
1 loge a H + 2.303
[8.94]
where a H + is the activity of the H + ion in solution. Hydrogen ion activity can vary from 100 for a strong acid, through 10−7 for a neutral solution (e.g. pure water), to 10 −14 for a strong alkali. The corresponding pH values are 0, 7 and 14. A typical pH sensor is shown in Figure 8.25(c).[18] Here the ISE is a membrane of a specially formulated glass. A thin layer (10 −7 m thick) of hydrated silica forms on the wetted surface of the glass. An equilibrium is set up between the activities of H+ ions in this layer and those in the test solution; an increase in acidity causes more H+ ions to be deposited in the layer which raise its potential relative to the bulk solution. Because the H + ion is the only positive ion small enough to fit into the silica lattice, the glass electrode is highly selective to H+ and the selectivity coefficient K to other positive ions can usually be assumed to be zero. The corresponding Thévenin e.m.f. for the sensor is: ETh = E0 +
Rθ loge a H + F
[8.95]
Using eqn [8.94], at θ = 298 K (25 °C) we have: ETh = E0 −
Thévenin e.m.f. for glass ISE
8.314 × 298 × 2.303 pH 96 493
ETh = E0 − 0.0592 pH V
[8.96]
Thus the sensor output voltage is proportional to pH with a sensitivity of −59.2 mV/pH at 25 °C. The constant E0 depends on the construction of the glass electrode and the choice of the reference electrode; the sensor shown has an internal reference electrode consisting of a silver wire coated with silver chloride immersed in a potassium chloride/silver gel. The Thévenin resistance RTh of the sensor is made up of the glass electrode resistance (100 to 300 MΩ) in series with the reference electrode resistance (10 kΩ).[18] Table 8.4 gives the type and concentration range of 19 ISEs in common use.[19] Many of these are solid-state electrodes, where the membrane is either a single crystal or a compacted disc of a highly insoluble inorganic salt. Other ISEs use neutral carriers as the active material; these are non-ionic organic molecules which are incorporated into an inert matrix of PVC or silicone rubber. ISEs for sensing soluble gases such as ammonia and sulphur dioxide are also available. The gas diffuses through a gas-permeable membrane and then dissolves in an internal filling solution; the resulting change in the activity of a given ion in the solution (e.g. H+) is measured using an ion selective and a reference electrode in the usual way.
8.9 ELECTROCHEMICAL SENSING ELEMENTS
Table 8.4 Ion-selective electrodes.
193
Model no.
Electrode
Type
Concentration range
ISE 301 ISE 302 ISE 303 ISE 304 ISE 305 ISE 306 ISE 307 ISE 308 ISE 309 ISE 310 ISE 311 ISE 312 ISE 313 ISE 314 ISE 315 ISE 321 ISE 322 ISE 323 ISE 324
Chloride Bromide Iodide Cyanide Sulphide Copper Lead Silver Cadmium Calcium Nitrate Barium Water hardness Potassium Sodium Ammonia Sulphur dioxide Carbon dioxide Nitrogen oxides
Solid-state Solid-state Solid-state Solid-state Solid-state Solid-state Solid-state Solid-state Solid-state PVC membrane PVC membrane PVC membrane PVC membrane PVC membrane PVC membrane Gas-permeable membrane Gas-permeable membrane Gas-permeable membrane Gas-permeable membrane
100 − 5 × 10−5 M 100 − 5 × 10−6 M 100 − 10−6 M 10−2 − 10−6 M 100 − 10−7 M 100 − 10−6 M 10−1 − 10−4 M 100 − 10−6 M 10−1 − 10−5 M 100 − 10−5 M 100 − 10−5 M 100 − 10−4 M 100 − 10−4 M (as Ca 2+ + Mg 2+ ) 100 − 10−5 M 10−1 − 10−5 M 10−1 − 10−5 M 5 × 10−2 − 5 × 10−5 M 10−2 − 10−4 M 10−2 − 5 × 10−6 M
Source: adapted from Ref. [19].
8.9.2
Electrochemical gas sensors Some solid-state materials give an electrochemical response to certain gases. An example is zirconia, which is sensitive to oxygen. Zirconia is based on zirconium oxide (ZrO2) with small amounts of other metal oxides present. These atoms replace Zr atoms at lattice sites and enable the material to conduct both electrons and oxygen O2− ions. Opposite surfaces of a slab of zirconia are coated with a thin layer of platinum, which is porous to oxygen molecules, to give two electrodes. If a surface is exposed to a gas containing oxygen, then oxygen molecules diffuse into the zirconia. If the sensor is heated to around 600 to 700 °C, the following reversible electrochemical reaction takes place between oxygen molecules and electrons: O2 + 4e− ì 2O2− The electrons and oxygen ions migrate between the electrodes until an equilibrium is established; there will be a corresponding electrochemical voltage across the electrodes. A practical sensor consists of a small hollow cone of zirconia, coated on both the inside and outside with a layer of porous platinum and held at a constant temperature of 640 °C.[20] A sample of the gas to be analysed is present on the inside of the cone and a reference gas (often air) is present on the outside. Under these conditions, the open circuit (Thévenin) e.m.f. of the sensor is given by the Nernst equation:
Thévenin e.m.f. for zirconia oxygen sensor
ETh = E0 +
 P REF  Rθ O loge  SAM   nF  PO  2
2
[8.97]
194 S ENSING ELEMENTS where
E0 = constant n = number of electrons (4) involved in reaction P REF O2 = partial pressure of oxygen in reference gas (cone outside) SAM P O2 = partial pressure of oxygen in sample gas (cone inside).
Provided both the sensor temperature θ and the partial pressure P REF O2 of the reference gas outside the sensor are held constant, then ETh depends only on the partial pressure of oxygen in the sample gas inside the sensor. Typically, the sensitivity of the sensor is 60 mV/%O2 at 640 °C.
8.9.3
Chemically sensitive field effect transistors (CHEMFET) Figure 8.26(a) shows the construction of an n-channel enhancement-mode MOSFET (metal oxide semiconductor field effect transistor).[21] It is a chip of silicon crystal with impurities added to create areas of n-type and p-type material. The device has four terminals. The source S and drain D are regions of enriched n-type material, the body or substrate (B) is p-type material and the gate G is metal or polysilicon material. The body is often connected to the source to give a three-terminal device. The gate is insulated from the substrate by a thin layer of silicon dioxide so that negligible current is drawn through the gate terminal. Figures 8.26(c) and (d) show typical characteristics for the device. If the voltage VGS between gate and source is greater than a threshold value VT, then electrons are attracted to the region under the gate, and a channel of n-channel material is induced between the drain and source. The thickness of this channel increases as VGS is increased. Suppose that a given voltage VGS2 (>VT) is applied between gate and source and the voltage VDS, between drain and source, is increased from zero. There is now a flow of electrons from source to drain, i.e. a current flow iD from drain to source. Initially iD increases linearly with VDS, but increased iD causes the voltage in the drain area to rise and a tapering of the channel towards the drain. This means that as VDS increases further, the increase in iD gets smaller until it reaches a saturation value which is independent of VDS (Figure 8.26(c)). Figure 8.26(d) shows how the saturation current iD increases with VGS for a fixed value of VDS = 5 V, for VGS > VT; the relationship is non-linear. In many applications VGS is the input signal and iD the output signal, so that Figure 8.26(d) is the input–output characteristic for the device. Because the gate current is negligible, FETs have extremely high-input impedance and are used as the input stage of buffer amplifiers, which are used with high impedance sensors such as ion-selective electrodes (Section 8.9.1). The metal gate G is normally inert, but by using a layer which is sensitive to a given gas, a gas-sensitive FET (gasFET) sensor can be produced. One example uses a platinum or palladium layer on the gate to detect the presence of hydrogen. Here the gate–source voltage VGS will be the equilibrium potential between the gas and this layer and will depend on the gas concentration a via a Nernst equation similar to eqn [8.97]. The relationship between drain current iD and a will be broadly similar to that shown in Figure 8.26(d). Figure 8.26(b) shows the general construction of an ion-sensitive FET (ISFET). ISFETs have been developed from ion-selective electrodes; instead of connecting an ion-selective electrode to the FET input stage of an amplifier, the FET is used directly as a sensor for ion concentration. There is now no metal gate but a reference
8.9 ELECTROCHEMICAL SENSING ELEMENTS
Figure 8.26 Chemically sensitive field effect transistors: (a) Construction of n-channel enhancement MOSFET (Figure 8.26(a) is reprinted from Electrical Engineering Principles and Applications 2/e by allan r. hambley by permission of Pearson Education Inc., Upper Saddle River, NJ, USA[21]) (b) Constructi on of ion-sensitive FET (ISFET) (Figure 8.26(b) adapted from Hambley[21]) (c), (d) Typical characteristics.
195
196 S ENSING ELEMENTS electrode separate from the transistor. The solution to be analysed is introduced into the space between the reference electrode and the gate area. In a pH sensor the gate is simply a silicon nitride layer which is both an electrical insulator and sensitive to H+ ions. By depositing an ion-sensitive layer on top of the silicon nitride layer, sensors for a range of ions can be made. For example, a layer of sodium aluminosilicate glass will be sensitive to Na+ ion concentration. The gate–source voltage will be the sum of the reference electrode bias potential and the equilibrium potential between the solution and the membrane. This equilibrium potential will depend on ion activity a via a Nernst equation similar to eqn [8.92]; iD will correspondingly depend on a.
8.10
Hall effect sensors Figure 8.27(a) shows a rectangular slab of metal or semiconductor material and a corresponding xyz coordinate system. A current I flows through the slab in the x-direction, producing a corresponding current density J. A magnetic field of flux density B is applied to the slab in the z-direction. In a metal there are electron (n) charge carriers only; in a semiconductor there may be both electron and hole ( p) charge carriers. Figure 8.27(b) shows the motion of electrons in the xy-plane; the magnetic field is perpendicular to the plane of the paper. The electrons experience a force in the positive y-direction, due to the magnetic field, causing them to be deflected as shown. The top of the slab becomes negatively charged and the bottom positively
Figure 8.27 Hall effect sensors: (a) Geometry (b) Motion of electrons (c) Motion of holes (d) Typical sensor.
CONCLUSION
197
charged, resulting in an electric field E in the negative y-direction. The force on the electrons due to this electric field is in the opposite direction to the magnetic force and an equilibrium is established, with the two forces equal. In this equilibrium situation, E is given by: E = −(1/ne)JB = −RH JB
[8.98]
where: E = Hall electric field = V/w V/m V = Hall voltage V w = width of slab m J = current density = I/wt Am−2 t = slab thickness m B = magnetic flux density T (tesla) R H = Hall coefficient = 1/ne n = density of electrons m−3 e = charge on the electron C. The Hall voltage V is therefore given by: V = +(R H /t)IB
[8.99]
Figure 8.27(c) shows the motion of holes through the slab in the xy-plane; the magnetic field is again in the plane of the paper. The top of the slab becomes positively charged, the bottom negatively charged, with a resulting electric field E in the positive y-direction. The corresponding Hall voltage is given by: V = +(R H /t)IB
[8.100]
where: R H = 1/pe p = density of holes m−3. An important application of Hall devices is as magnetic field sensors; from eqns [8.99] and [8.100] we see that, for a constant current I, V is proportional to B. Silicon, doped with n- and p-type material, has a high Hall coefficient and is a suitable material. A typical sensor, shown in Figure 8.27(d), consists of an n-type conducting layer on a p-type substrate. There are two pairs of electrodes, one pair for current, one pair for Hall voltage, formed by diffusing highly doped and therefore highly conducting regions of n-type material into the n layer. A typical device has thickness t of 1 mm and Hall coefficient RH of 3 × 10−2; if current I = 10 mA then V = 30 mV for a magnetic field B of 0.1 tesla.
Conclusion This chapter has discussed the principles, characteristics and applications of a wide range of sensing elements.
198 S ENSING ELEMENTS
References [1] bentley j p 1984 ‘Temperature sensor characteristics and measurement system design’, J. Phys. E: Scientific Instruments, vol. 17, pp. 430–9. [2] IEC 751: 1983 (BS EN 60751: 1996) Industrial Platinum Resistance Detector Elements, International Electroterminal Commission. [3] Mullard Ltd 1974 Technical Information on Two-point NTC Thermistors. [4] papakostas t v and white n m 2000 ‘Thick-film polymer sensors for physical variables’, Measurement and Control, vol. 33, no. 4, pp. 105–8. [5] van ewyk r 1996 ‘Flammable gas sensors and sensing’, Measurement and Control, vol. 29, no. 1, pp. 13–16. [6] jefferey p d et al. 1998 ‘Thick film chemical sensor array allows flexibility in specification’, Sensor and Transducer Conf., NEC Birmingham. [7] neubert h k p 1975 Instrument Transducers: An Introduction to their Performance and Design (2nd edn), Oxford University Press, London, pp. 237–8. [8] lee-integer 1985 ‘Advanced relative humidity sensor’, Electronic Product Review, June. [9] IEC 584.1: 1995 International Thermocouple Reference Tables, International Electroterminal Commission. [10] doebelin e o 1975 Measurement Systems: Application and Design (2nd edn), McGraw-Hill, New York, pp. 520–1. [11] Consolidated Electrodynamics, Bulletin 4202B/1167 on Type 4-202 Strain Gauge Accelerometer. [12] neubert h k p Instrument Transducers, p. 56 (see Ref. 4). [13] chatigny j v and robb l e 1986 ‘Sensors: making the most of piezo film’, Sensor Review, vol. 7, no. 1, pp. 15–20. [14] purdy d 1981 ‘Piezoelectric devices. A step nearer problem-free vibration measurement’, Transducer Technology, vol. 4, no. 1. [15] Endevco 1980 Product Development News, vol. 16, issue 3. [16] noble m 1985 ‘IC sensors boost potential of measurement systems’, Transducer Technology, vol. 8, no. 4. [17] bailey p l 1976 Analysis with Ion Selective Electrodes, Heyden. [18] thompson w ‘pH facts – the glass electrode’, Kent Technical Review, pp. 16–22. [19] E.D.T. Research 1984 Technical Data on Ion Selective Electrodes. [20] Sirius Instruments Ltd 1983 in Technical Bulletin on Zirconia Oxygen Analysers. [21] hambley a r 2002 Electrical Engineering: Principles and Applications (2nd edn), Prentice-Hall, Englewood Cliffs, NJ, pp. 521–8.
Problems 8.1
A platinum resistance sensor is to be used to measure temperatures between 0 and 200 °C. Given that the resistance R T Ω at T °C is given by RT = R0(1 + αT + β T 2 ) and R0 = 100.0, R100 = 138.50, R200 = 175.83 Ω, calculate:
PROBLEMS
(a) (b) 8.2
199
the values of α and β; the non-linearity at 100 °C as a percentage of full-scale deflection.
Four strain gauges are bonded onto a cantilever as shown in Figure 8.20(a). Given that the gauges are placed halfway along the cantilever and the cantilever is subject to a downward force of 0.5 N, use the data given below to calculate the resistance of each strain gauge: Cantilever data Length l = 25 cm Width w = 6 cm Thickness t = 3 mm Young’s modulus E = 70 × 109 Pa
Strain gauge data Gauge factor G = 2.1 Unstrained resistance R0 = 120 Ω
8.3
A variable dielectric capacitive displacement sensor consists of two square metal plates of side 5 cm, separated by a gap of 1 mm. A sheet of dielectric material 1 mm thick and of the same area as the plates can be slid between them as shown in Figure 8.9. Given that the dielectric constant of air is 1 and that of the dielectric material 4, calculate the capacitance of the sensor when the input displacement x = 0.0, 2.5 and 5.0 cm.
8.4
A variable reluctance sensor consists of a core, a variable air gap and an armature. The core is a steel rod of diameter 1 cm and relative permeability 100, bent to form a semi-circle of diameter 4 cm. A coil of 500 turns is wound onto the core. The armature is a steel plate of thickness 0.5 cm and relative permeability 100. Assuming the relative permeability of air = 1.0 and the permeability of free space = 4π × 10−7 H m−1, calculate the inductance of the sensor for air gaps of 1 mm and 3 mm.
8.5
By taking a central flux path, estimate the inductance of the sensor shown in Figure Prob. 5 (a) (b)
for zero air gap; for a 2 mm air gap.
Assume the relative permeability of core and armature is 104 and that of air is unity. Figure Prob. 5.
8.6
A variable reluctance tachogenerator consists of a ferromagnetic gear wheel with 22 teeth rotating close to a magnet and coil assembly. The total flux N linked by the coil is given by: N(θ ) = 4.0 + 1.5 cos 22θ milliwebers where θ is the angular position of the wheel relative to the axis of the magnet. Calculate the amplitude and frequency of the output signal when the angular velocity of the wheel is 1000 and 10 000 r.p.m.
200 S ENSING ELEMENTS
8.7
An iron v. constantan thermocouple is to be used to measure temperatures between 0 and 300 °C. The e.m.f. values are as given in Table 8.2. (a) (b) (c)
8.8
Find the non-linearity at 100 °C and 200 °C as a percentage of full scale. Between 100 °C and 300 °C the thermocouple e.m.f. is given by ET,0 = a1T + a2T 2. Calculate a1 and a2. The e.m.f. is 12 500 µV relative to a reference junction of 20 °C and the corresponding reference junction circuit voltage is 1000 µV. Use the result of (b) to estimate the measured junction temperature.
An accelerometer consisting of an elastic element and a potentiometric displacement sensor has to meet the following specification: Input range = 0 to 5g Output range = 0 to 10 V Natural frequency = 10 Hz (a) (b) (c)
Damping ratio = 0.8 Maximum non-linearity = 2% of f.s.d. Seismic mass = 0.005 kg
Calculate the required spring stiffness and damping constant. What should the input displacement range of the potentiometer be? ( g = 9.81 m s−2 ) If the accelerometer is to be used with a recorder of 10 kΩ resistance, what is the maximum allowable potentiometer resistance?
8.9
An accelerometer is to measure the angular acceleration of a rotating mixing vessel. The angular position of the vessel varies sinusoidally with time with amplitude 2.5 rad and period 2 s. The rotating seismic mass is equivalent to a mass of 0.1 kg on a weightless arm of length 5 cm. The stiffness of the spring is 2.5 × 10 −2 N m rad −1 and the damping ratio is 1/ê2. The angular position of the seismic mass is measured with a secondary potentiometric sensor. What should the input range of this sensor be?
8.10
A flat circular diaphragm of density 6 × 103 kg m−3 is to be used as a pressure sensor. The element should fulfil the following specification: Input range = 0 to 104 Pa Maximum non-linearity = 1% of f.s.d Amplitude ratio to be flat with ±3% up to 100 Hz Using the equations given in Figure 8.21, and assuming a damping ratio of 0.01, calculate: (a) (b)
8.11
the thickness t of the diaphragm; the output displacement range of the sensor.
A piezoelectric crystal, acting as a force sensor, is connected by a short cable of negligible capacitance and resistance to a voltage detector of infinite bandwidth and purely resistive impedance of 10 MΩ. (a) (b)
(c)
Use the crystal data below to calculate the system transfer function and to sketch the approximate frequency response characteristics of the system. The time variation in the thrust of an engine is a square wave of period 10 ms. Explain carefully, but without performing detailed calculations, why the above system is unsuitable for this application. A charge amplifier with feedback capacitance CF = 1000 pF and feedback resistance RF = 100 MΩ is incorporated into the system. By sketching the frequency response characteristics of the modified system, explain why it is suitable for the application of part (b).
PROBLEMS
Crystal data
8.12
(a)
(b)
Charge sensitivity to force = 2 pC N−1 Capacitance = 100 pF Natural frequency = 37 kHz Damping ratio = 0.01
The casing of a compressor is executing sinusoidal vibrations with a displacement amplitude of 10 −4 m and a frequency of 500 Hz. Calculate the amplitude of the acceleration of the casing in units of g (g = 9.81 m s−2 ). A piezoelectric crystal accelerometer has a steady-state sensitivity of 2.0 pC/g, a natural frequency of 20 kHz, a damping ratio of 0.1 and a capacitance of 100 pF. It is connected to an oscilloscope (a resistive load of 10 MΩ) by a cable of capacitance 100 pF. (i) (ii) (iii)
8.13
Without performing detailed calculations, sketch magnitude and phase frequency response characteristics for the accelerometer. Using these characteristics: Estimate the amplitude of the voltage measured by the oscilloscope when the accelerometer is mounted on the casing described in (a). Explain why this system is unsuitable for measuring vibrations at 20 Hz. How should the system be modified to make these measurements possible?
A seismic mass accelerometer with a variable reluctance displacement sensor is shown in Figure Prob. 13. The undamped natural frequency of the accelerometer is to be 25 Hz and the range of the acceleration input is to be 0 to 5g. Using the data given below, calculate: (a) (b) (c)
the required spring stiffness; the air gap at maximum acceleration; the range of variation of the coil inductance.
Data Relative permeability of armature and coil material = 100 Relative permeability of air =1 Permeability of free space = 4π × 10−7 H/m Effective mass of armature and spring = 0.025 kg Acceleration of gravity g = 9.81 m/s2 Number of coil turns = 500 Air gap at zero acceleration = 3 mm Figure Prob. 8.13.
201
202 S ENSING ELEMENTS
Basic problems 8.14
A platinum resistance sensor has a resistance of 100.0 Ω at 0 °C and a temperature coefficient of resistance α of 4 × 10 −3 °C −1. If the resistance of the sensor is 125 Ω, find the corresponding temperature of the sensor.
8.15
A platinum resistance sensor has a resistance of 100 Ω at 0 °C and a temperature coefficient of resistance of 4 × 10 −3 °C −1. Find the resistance of the sensor at −100 °C.
8.16
A thermistor has constants K = 0.1 Ω, β = 3200 K. Find the resistance at the ice point.
8.17
A steel bar has a cross-sectional area of 10−3 m2, Young’s modulus of 2 × 1011 Pa and Poisson’s ratio of 0.4. If the bar is subject to a compressive force of 10 5 N, find the corresponding longitudinal and transverse strains.
8.18
Two strain gauges, each with an unstrained resistance of 120 Ω and a gauge factor of 2.0, are bonded onto the steel bar of Problem 8.17 so that one measures the longitudinal strain and one the transverse strain. Find the corresponding gauge resistances for the longitudinal and transverse strains.
8.19
A parallel plate capacitive displacement sensor consists of two square metal plates 8 cm × 8 cm, separated by a gap of 2 mm. The space between the plates is completely filled by a slab with a dielectric constant of 6.0. If the permittivity of free space is 8.85 pF/m, find the capacitance of the sensor.
8.20
A parallel plate capacitance pressure sensor consists of two circular plates of diameter 2 cm separated by an air gap of 1 mm. If the dielectric constant of air is 1.0 and the permittivity of free space is 8.85 pF/m, find the capacitance of the sensor.
8.21
A strain gauge having an unstrained resistance of 120 Ω and a gauge factor of 2.1 is bonded onto a steel girder so that it experiences a tensile stress of 108 Pa. If Young’s modulus for steel is 2 × 1011 Pa, calculate the strained resistance of the gauge.
8.22
Table Prob. 22 gives values for the e.m.f. of a Type T thermocouple relative to a reference junction temperature of 0 °C. Use the table to calculate: (a) (b)
the percentage non-linearity at 150 °C if the temperature range is 0 to 300 °C; the temperature inside a vessel if an e.m.f. of 11.5 mV is measured relative to a reference junction of 20 °C.
8.23
The e.m.f. of a type T thermocouple is measured to be 8.561 mV relative to a reference junction temperature of 20 °C. Use the table provided in Problem 8.22 to find the temperature of the measured junction.
8.24
A variable reluctance tachogenerator consists of a ferromagnetic wheel with 20 teeth rotating close to a bar magnet and coil. If the wheel is rotating at 6000 rpm, what is the frequency of the a.c. voltage induced in the coil?
8.25
A force sensor includes an elastic spring with a stiffness of 106 N m−1. If the force input has a range of 0 to 5 kN, find the corresponding displacement output range.
8.26
A doped silicon wafer has a thickness of 0.5 mm and a Hall coefficient of 2 × 10−2. If the current through the wafer is 7.5 mA, find the Hall voltage for a magnetic flux density of 0.2 T.
PROBLEMS
Table Prob. 22 Temp. °C
203
e.m.f. in mV for type T thermocouple (measured junction at T °C, reference junction at 0 °C).
0
1
2
3
4
5
6
7
8
9
10
0 10 20 30 40
0.000 0.391 0.789 1.196 1.611
0.039 0.430 0.830 1.237 1.653
0.078 0.470 0.870 1.279 1.695
0.117 0.510 0.911 1.320 1.738
0.156 0.549 0.951 1.361 1.780
0.195 0.589 0.992 1.403 1.822
0.234 0.629 1.032 1.444 1.865
0.273 0.669 1.073 1.486 1.907
0.312 0.709 1.114 1.528 1.950
0.351 0.749 1.155 1.569 1.992
0.391 0.789 1.196 1.611 2.035
50 60 70 80 90
2.035 2.467 2.908 3.357 3.813
2.078 2.511 2.953 3.402 3.859
2.121 2.555 2.997 3.447 3.906
2.164 2.599 3.042 3.493 3.952
2.207 2.643 3.087 3.538 3.998
2.250 2.687 3.131 3.584 4.044
2.294 2.731 3.176 3.630 4.091
2.337 2.775 3.221 3.676 4.137
2.380 2.819 3.266 3.721 4.184
2.424 2.864 3.312 3.767 4.231
2.467 2.908 3.357 3.813 4.277
100 110 120 130 140
4.277 4.749 5.227 5.712 6.204
4.324 4.796 5.275 5.761 6.254
4.371 4.844 5.324 5.810 6.303
4.418 4.891 5.372 5.859 6.353
4.465 4.939 5.420 5.908 6.403
4.512 4.987 5.469 5.957 6.452
4.559 5.035 5.517 6.007 6.502
4.607 5.083 5.566 6.056 6.552
4.654 5.131 5.615 6.105 6.602
4.701 5.179 5.663 6.155 6.652
4.749 5.227 5.712 6.204 6.702
150 160 170 180 190
6.702 7.207 7.718 8.235 8.757
6.753 7.258 7.769 8.287 8.810
6.803 7.309 7.821 8.339 8.863
6.853 7.360 7.872 8.391 8.915
6.903 7.411 7.924 8.443 8.968
6.954 7.462 7.975 8.495 9.021
7.004 7.513 8.027 8.548 9.074
7.055 7.564 8.079 8.600 9.127
7.106 7.615 8.131 8.652 9.180
7.156 7.666 8.183 8.705 9.233
7.207 7.718 8.235 8.757 9.286
200 210 220 230 240
9.286 9.820 10.360 10.905 11.456
9.339 9.874 10.414 10.960 11.511
9.392 9.928 10.469 11.015 11.566
9.446 9.982 10.523 11.070 11.622
9.499 10.036 10.578 11.125 11.677
9.553 10.090 10.632 11.180 11.733
9.606 10.144 10.687 11.235 11.788
9.659 10.198 10.741 11.290 11.844
9.713 10.252 10.796 11.345 11.900
9.767 10.306 10.851 11.401 11.956
9.820 10.360 10.905 11.456 12.011
250 260 270 280 290
12.011 12.572 13.137 13.707 14.281
12.067 12.628 13.194 13.764 14.339
12.123 12.684 13.251 13.821 14.396
12.179 12.741 13.307 13.879 14.454
12.235 12.797 13.364 13.936 14.512
12.291 12.854 13.421 13.993 14.570
12.347 12.910 13.478 14.051 14.628
12.403 12.967 13.535 14.108 14.686
12.459 13.024 13.592 14.166 14.744
12.515 13.080 13.650 14.223 14.802
12.572 13.137 13.707 14.281 14.860
300 310 320 330 340
14.860 15.443 16.030 16.621 17.217
14.918 15.501 16.089 16.681 17.277
14.976 15.560 16.148 16.740 17.336
15.034 15.619 16.207 16.800 17.396
15.092 15.677 16.266 16.859 17.456
15.151 15.736 16.325 16.919 17.516
15.209 15.795 16.384 16.978 17.576
15.267 15.853 16.444 17.038 17.636
15.326 15.912 16.503 17.097 17.696
15.384 15.971 16.562 17.157 17.756
15.443 16.030 16.621 17.217 17.816
9 Signal Conditioning Elements As stated in Chapter 1, signal conditioning elements convert the output of sensing elements into a form suitable for further processing. This form is usually a d.c. voltage, a d.c. current or a variable frequency a.c. voltage.
9.1
Deflection bridges Deflection bridges are used to convert the output of resistive, capacitive and inductive sensors into a voltage signal.
9.1.1
Thévenin equivalent circuit for a deflection bridge In Chapter 5 we saw that any linear network can be represented by a Thévenin equivalent circuit consisting of a voltage source ETh together with a series impedance ZTh. Figure 9.1 shows a general deflection bridge network. ETh is the open-circuit output voltage of the bridge, i.e. when current i in BD = 0. Using Kirchhoff’s laws:
Figure 9.1 Calculation of Thévenin equivalent circuit for a deflection bridge.
Loop PABCQ VS = i1 Z2 + i1 Z3
i.e.
i1 =
VS Z2 + Z3
[9.1]
Loop PADCQ VS = i2 Z1 + i2 Z4
i.e. i2 =
VS Z1 + Z4
[9.2]
206 S IGNAL CONDITIONING ELEMENTS Assuming Q is at earth potential, then: Potential at P and A = VS Potential at B = VS − i1 Z2 Potential at D = VS − i2 Z1 ETh is equal to the potential difference between B and D, i.e. ETh = (VS − i1 Z 2 ) − (VS − i2 Z 1 ) = i2 Z1 − i1 Z2 Using [9.1] and [9.2] we have Thévenin voltage for general deflection bridge
 Z1 Z2  ETh = VS  −   Z1 + Z 4 Z 2 + Z3 
[9.3]
ZTh is the impedance, looking back into the circuit, between the output terminals BD, when the supply voltage VS is replaced by its internal impedance. Assuming the internal impedance of the supply is zero, then this is equivalent to a short circuit across AC (see Figure 9.1). We see that ZTh is equal to the parallel combination of Z2 and Z3 in series with the parallel combination of Z1 and Z4, i.e. Thévenin impedance for general deflection bridge
Z Th =
Z 2 Z3 Z1 Z 4 + Z 2 + Z3 Z1 + Z 4
[9.4]
If a load, e.g. a voltmeter or amplifier, of impedance ZL is connected across the output terminals BD, then the current through the load is i = ETh /(ZTh + ZL). The corresponding voltage across the load is VL = ETh ZL /(ZTh + ZL). Thus in the limit that |ZL | 0002 |ZTh |, VL → ETh.
9.1.2
Design of resistive deflection bridges In a resistive or Wheatstone bridge all four impedances Z1 to Z4 are pure resistances R1 to R4. From [9.3] we have
Output voltage for resistive deflection bridge
 R1 R2  ETh = VS  −   R1 + R4 R2 + R3 
[9.5]
We first consider the case when only one of the resistances is a sensing element. Here R1 depends on the input measured variable I, i.e. R1 = RI, and R2, R3 and R4 are fixed resistors. This gives ETh = VS
A D 1 1 − C 1 + R4 /RI 1 + R3 /R2 F
[9.6]
9.1 D E F LE CT I O N BR I D GE S
207
from which we see that to design a single element bridge we need to specify the three parameters VS, R4 and R3 /R2. The individual values of R2 and R3 are not critical; it is their ratio which is crucial to the design. The three parameters can be specified by considering the range and linearity of the output voltage and electrical power limitations for the sensor. Thus if IMIN and IMAX are minimum and maximum values of the measured variable, and RI MIN and RI MAX are the corresponding sensor resistances, then in order for the bridge output voltage to have a range from VMIN to VMAX the following conditions must be obeyed: VMIN = VS
A D 1 1 − C 1 + R4 /RI MIN 1 + R3 /R2 F
[9.7]
VMAX = VS
A D 1 1 − C 1 + R4 /RI MAX 1 + R3 /R2 F
[9.8]
Often we require VMIN = 0, i.e. the bridge to be balanced when I = IMIN; in this case [9.7] reduces to: Relationship between resistances in a balanced Wheatstone bridge
R4 R = 3 RI R2
[9.9]
MIN
A third condition is required to complete the design. One important consideration is the need to limit the electrical power i 22 RI in the sensor to a level which enables it to be dissipated as heat flow to the surrounding fluid; otherwise the temperature of the sensor rises above that of the surrounding fluid, thereby affecting the sensor resistance (Chapter 14). Thus if w ˆ watts is the maximum power dissipation, using [9.2] we require: V S2
RI ≤w ˆ (RI + R4)2
[9.10]
for I MIN ≤ I ≤ I MAX. Another requirement which is often important is the need to keep the non-linearity of the overall relationship between ETh and I within specified limits. Assuming VMIN = 0, then the ideal relationship between V and I is the ideal straight line: VIDEAL =
A VMAX D A VMAX D I− I C I MAX − I MIN F C I MAX − I MIN F MIN
[9.11]
From Chapter 2, the non-linear function N(I ) = ETh − VIDEAL. Suppose we require the non-linearity as a percentage of full scale to be less than ; per cent over the measurement range; we then have the condition: ETh − VIDEAL × 100 ≤ ; VMAX
[9.12]
for IMIN ≤ I ≤ IMAX, where ETh and VIDEAL are given by [9.6] and [9.11] respectively. The value of ratio R3 /R2 varies according to the type of resistive sensor used; it is useful to have a graph which gives some insight into this. From [9.9] we have R4 = (R3 /R2)RIMIN. Substituting for R4 in [9.6] gives:
208 S IGNAL CONDITIONING ELEMENTS 1 1 ETh = − 1 + (R /R )(R /R ) 1 + R VS 3 2 I MIN I 3 /R2 or
ν=
1 1 x 1 − = − 1 + r/x 1 + r x + r 1 + r
[9.13]
where
ν=
ETh , VS
r=
R3 , R2
x=
RI RI MIN
Figure 9.2 shows graphs of ν versus x, for x in the range 0.1 to 2.0, and for r = 0.1, 1.0, 10.0 and 100. We note that ν is always zero at x = 1, corresponding to the bridge being balanced at I = IMIN; also that ν (x) is in general non-linear, the degree of nonlinearity depending on r. For a strain gauge (Section 8.1) the change in resistance ∆R = R0Ge is very small; this means that x is very close to 1. We require the sensitivity of the bridge to be as high as possible, i.e. (∂ν /∂x)x≈1 to be maximum. From Figure 9.2 we see that this occurs when r = R3 /R2 = 1, i.e. R3 = R2. Using eqn [9.9], this means that R4 = RI MIN = R0. Thus for a bridge with a single strain gauge, we require R2, R3 and R4 to all equal the unstrained
Figure 9.2 The deflection bridge function ν (x) = x/(x + r) − 1/(1 + r).
9.1 D E F LE CT I O N BR I D GE S
209
gauge resistance R0. The value of VS is determined by the power condition [9.10]. Putting r = 1 in [9.13] gives:
ν(x) =
x−1 ≈ 2(x + 1)
1 4
(x − 1)
so that: D 1 A RI − RIMIN D 1 ∆R 1 ETh 1 A RI = −1 = = = –4 Ge C F 4 C RI MIN F 4 R0 VS 4 RI MIN Thus: Output voltage for single-element strain gauge bridge
ETh =
VS Ge 4
[9.14]
i.e. the relationship between ETh and e is linear. The resistance RT Ω of a metal resistance sensor, e.g. platinum, at T °C is given approximately by (Section 8.1) RT = R0(1 + α T ). Typically x varies between 1 and 2 (R0 = 100 Ω, R250 ≈ 200 Ω). Since this device has only a small non-linearity (less than 1%) a linear bridge is required. From Figure 9.2 we see this means a large value of r, e.g. r ≈ 100, linearity being obtained at the expense of low sensitivity. For large r (i.e. r 0002 1) eqn [9.13] approximates to:
ν≈
1 (x − 1) i.e. r
D ETh R2 A RT = −1 F VS R3 C RTMIN
If TMIN = 0 °C, then since RT /R0 = 1 + α T: Output voltage for metal resistance sensor bridge (large r)
ETh = VS
R2 αT R3
[9.15]
i.e. a linear relationship between ETh and T. For given VS (satisfying [9.10]), R3 /R2 can be found from [9.15] and R4 from [9.9]. Figure Prob. 16 in this chapter shows an accurate arrangement for temperature measurement involving a platinum resistance sensor with four-lead connection to a deflection bridge. With a copper or nickel resistance sensor, this circuit provides the voltage source ET2,0 necessary for automatic reference junction compensation of a thermocouple e.m.f. (see Section 8.5 and Problem 9.2). The resistance Rθ of a thermistor varies non-linearly with absolute temperature θ K according to the relationship Rθ = K exp( β /θ ) (Section 8.1). A typical thermistor has a resistance of 12 kΩ at 298 K (25 °C) falling to 2 kΩ at 348 K (75 °C); i.e. x varies from 1.0 to 0.17 over this measurement range. By choosing a suitable value of r, usually between 0.25 and 0.30, it is possible to use the bridge non-linearity to partially compensate for the thermistor non-linearity; this means the overall relationship between ETh and θ is reasonably linear over this range (see Figures 9.2 and 9.3). Supposing we require an output range of 0 to 1.0 V, corresponding to a temperature range of 298 to 348 K; then almost minimum non-linearity is obtained if we design
210 S IGNAL CONDITIONING ELEMENTS Figure 9.3 Design conditions and characteristics for thermistor bridge.
the bridge so that ETh = 0.5 V at θ = 323 K (Figure 9.3). The bridge is then designed by solving the following three equations for VS, R4 and R3 /R 2: 0.0 = VS
A D 1 1 − C 1 + R4 /R298 1 + R3 /R2 F
(point O)
0.5 = VS
A D 1 1 − C 1 + R4 /R323 1 + R3 /R2 F
(point Q)
1.0 = VS
A D 1 1 − C 1 + R4 /R348 1 + R3 /R2 F
(point P)
[9.16]
A bridge with two metal resistance sensor elements can be designed to give an output voltage approximately proportional to temperature differences T1 − T2. The bridge incorporates one element at T1 °C and another at T2 °C so that R1 = R0(1 + αT1) and R2 = R0(1 + αT2), as shown in Figure 9.4(a). In order to balance the bridge when T1 − T2 = 0, we require R4 /R1 = R3 /R2 when T1 = T2; this implies R4 /R0 = R3 /R0, i.e. R4 = R3. Using [9.6], we have: Figure 9.4 (a) Two-element resistance sensor bridge (b) Four-element strain gauge bridge.
ETh
 1 1 −  = VS  R3  1  R3  1     1+  1+ R0  1 + α T 2  R0  1 + α T1  
    
9.1 D E F LE CT I O N BR I D GE S
211
and if we choose R3 so that R3 /R0 0002 1, this approximates to: ETh ≈ VS
A 1 + α T1 1 + α T2 D − C R3 /R0 R3 /R0 F
ETh = VS
R0 α (T1 − T2 ) R3
i.e. Output voltage for two-element resistance sensor bridge
[9.17]
which has a similar form to [9.15]. This bridge has several applications, including the measurement of humidity with wet and dry bulb sensors and as a detector in a thermal radiation measurement system (Section 15.2). Bridges with four active strain gauges, mounted on elastic elements, are commonly used for the measurement of force, torque, acceleration and pressure. Provided the gauges are correctly connected into the bridge, so that one opposite pair (e.g. R1, R3) are in tension and the other opposite pair (e.g. R2, R4) are in compression, then the sensitivity is greater than for a single-element bridge. This bridge also compensates for changes in gauge resistance due to temperature. For metal strain gauges the effect of temperature is to multiply each gauge resistance by the factor (1 + α T ); this cancels out in the output voltage eqn [9.5]. The only design decision to be made is the value of supply voltage; VS is determined either by the maximum power condition [9.10] or by the maximum specified gauge current (Section 8.1). For the cantilever load cell and torque element of Section 8.6 (Figures 8.20(a) and (c)), we have R1 = R3 = R0(1 + Ge) and R2 = R4 = R0(1 − Ge), where R0 is the unstrained gauge resistance. Here e=
6(l − x)F wt 2E
for the cantilever and e=
T πSa3
for the torque element. Using [9.5] the output voltage of the bridge (Figure 9.4) is thus given by: R0(1 + Ge) R0(1 − Ge) # − @ R0(1 + Ge) + R0(1 − Ge) R0(1 − Ge) + R0(1 + Ge) $
ETh = VS ! = VS
A 1 + Ge 1 − Ge D − C 2 2 F
i.e. Output voltage for cantilever and torque elements
ETh = VS Ge
[9.18]
Here the output voltage is four times that of the single gauge bridge (eqn [9.14]). For the pillar load cell (Figure 8.20(b)):
212 S IGNAL CONDITIONING ELEMENTS A GνF D R1 = R3 = R0 1 + C AE F
and
A GF D R2 = R4 = R0 1 − C AE F
Using eqn [9.5] and assuming (GF/AE) 0003 1, the approximate bridge output voltage is: Output voltage for pillar load cell
ETh ≈
VS GF (1 + ν ) 2 AE
[9.19]
Finally for the unbonded strain gauge accelerometer (Figure 8.20(d)), we have R1 = R3 = R0[1 + G(e0 + e)] and
R2 = R4 = R0 [l + G(e0 − e)]
where e0 = strain due to prestressing which is present at zero acceleration and e is the acceleration-induced strain = a/ω 2n L. The output voltage of the bridge is given by: Output voltage for strain gauge accelerometer
ETh =
VSGe 1 + Ge0
[9.20]
The output voltage of these bridges is usually small and requires high amplification; typically if VS = 15 V, G = 2 and e = 10−5, then ETh ≈ VS Ge = 300 µV. Because of the problems in amplifying low-level d.c. signals (see following section), an a.c. supply voltage VS = ÄS sin ωS t is often used to give an a.c. output voltage; a.c. carrier systems are discussed fully in Section 9.3.
9.1.3
Design of reactive deflection bridges A reactive bridge has an a.c. supply voltage; two arms are usually reactive impedances and two arms resistive impedances. Figure 9.5(a) shows the bridge to be used with the capacitance level transducer of Section 8.2; here we have Ch =
2πε 0 [l + (ε − 1)h] loge(b/a)
Also Z1 = 1/( jω C0), Z2 = R2, Z3 = R3 and Z4 = 1/( jω Ch). Figure 9.5 (a) Bridge for capacitive level sensor (b) Bridge for inductive push-pull displacement sensor.
9.1 D E F LE CT I O N BR I D GE S
213
This gives ETh = VS
A D 1 1 − C 1 + C0 /Ch 1 + R3 /R2 F
Thus in order to get :Th = 0 at minimum level h MIN, we require C0 = Ch MIN (R3 /R2), giving: Output voltage for capacitance level bridge
ETh
 1 1  −   = VS  Ch R3 R3  1 + 1 +   R2  Ch R 2  MIN
[9.21]
Again if R3 /R2 is made large compared with 1, this approximates to the linear form: ETh ≈ VS
D R2 A Ch −1 C F R3 Ch MIN
The most common two-element reactive bridges incorporate either capacitive or inductive push-pull displacement sensors (Sections 8.2 and 8.3). The capacitive sensor (Figure 8.9) has C1 = εε 0 A/(d + x) and C2 = εε 0 A/(d − x); if this is connected into an a.c. bridge so that Z1 = l/( jω C1), Z2 = Z3 = R, Z4 = 1/( jω C2) we have: ETh = VS
A C2 1D − C C1 + C2 2 F
[9.22]
This gives: Output voltage for capacitance push-pull bridge
ETh =
VS x 2d
[9.23]
i.e. the relationship between ETh and x is linear and independent of frequency ω . We note that the alternative way of connecting the sensor into the bridge so that Z1 = l/( jω C1) and Z2 = 1/( jω C2) gives an output voltage which is non-linearly related to x and dependent on the supply frequency ω. A similar result is obtained with the variable reluctance push-pull displacement sensor (Figure 8.11). This has: L1 =
L0 , 1 + α (a − x)
L2 =
L0 1 + α (a + x)
and from Figure 9.5(b) we have Z1 = jω L1, Z2 = Z3 = R, Z4 = jω L2, giving: ETh = VS
A L1 1D − C L1 + L2 2 F
[9.24]
from which Output voltage for inductive push-pull bridge
ETh =
VS α x 2(1 + α a)
[9.25]
214 S IGNAL CONDITIONING ELEMENTS Again the relationship between ETh and x is linear and frequency independent. In practice, however, these sensors will have associated resistance so that eqns [9.23] and [9.25] are only approximate.
9.2
Amplifiers Amplifiers are necessary in order to amplify low-level signals, e.g. thermocouple or strain gauge bridge output voltages, to a level which enables them to be further processed.
9.2.1
The ideal operational amplifier and its applications Ideal operational amplifier characteristics The operational amplifier can be regarded as the basic building block for modern amplifiers. It is a high gain, integrated circuit amplifier designed to amplify signals from d.c. up to many kHz. It is not normally used by itself but with external feedback networks to produce precise transfer characteristics which depend almost entirely on the feedback network. Usually there are two input terminals and one output terminal, the voltage at the output terminal being proportional to the difference between the voltages at the input terminals. Figure 9.6 shows the circuit symbol and a simplified equivalent circuit for an operational amplifier. Table 9.1 summarises the main characteristics of an ideal operational amplifier together with those of a typical practical amplifier OPA27.[1]
Figure 9.6 Circuit symbol and simplified equivalent circuit for operational amplifier.
Table 9.1 Ideal and typical operational amplifier characteristics.
Parameter
Ideal op-amp
OPA27[1]
D.C. open-loop gain A OL Input impedance Z IN Output impedance Z OUT Input offset voltage VOS Temp. coeff. of input offset voltage γ Input bias current iB 3 dB bandwidth 0 to fB Common mode rejection ratio, CMMR
∞ ∞ 0 0 0 0 0 to ∞ ∞
120 dB (106) 6 MΩ 70 Ω 25 µV 0.6 µV °C −1 40 nA 0 to 10 Hz 114 dB
9.2 AMPLIFIERS
215
If we assume ideal behaviour, then the calculations of transfer characteristics of operational amplifier feedback networks are considerably simplified. These calculations can then be modified if it is necessary to take into account the non-ideal behaviour of practical amplifiers. The transfer characteristics are derived using the following two ideal conditions: Summing point conditions for ideal operational amplifier with negative feedback
i− = i+ = 0 V− = V+
[9.26]
The first condition results from assuming input impedance Z IN = ∞. The second condition results from assuming that AOL → ∞: since VOUT = AOL(V+ − V−) and VOUT is finite, then the differential voltage (V+ − V− ) → 0 i.e. V+ = V− . We now derive the transfer characteristics for six voltage amplifier circuits.
Inverting amplifier Figure 9.7 shows a signal source VIN , RIN connected to an inverting amplifier. Since V+ = V− = 0: iIN =
VIN , RIN
iF =
VOUT RF
Also iIN + iF = i− = 0 giving VIN VOUT + =0 RIN RF VOUT =
− RF VIN RIN
[9.27]
Thus the output voltage VOUT is proportional to the input voltage VIN with gain −RF /RIN. The resistance R is set equal to the resistance of R IN in parallel with RF , i.e. RIN RF R= . This is to balance the d.c. bias currents to inverting and non-inverting RIN + RF inputs. Figure 9.7 Inverting amplifier.
216 S IGNAL CONDITIONING ELEMENTS Figure 9.8 Non-inverting amplifier.
Non-inverting amplifier Figure 9.8 shows a non-inverting amplifier circuit. Since i+ = 0, V+ = VIN; also RF and R1 form a potential divider so that: V− = VOUT
A R1 D C R1 + RF F
Also since V+ = V− VOUT
A R1 D = VIN C R1 + RF F
 RF  VOUT = 1 +  VIN R1  
[9.28]
A RF D . Thus output voltage VOUT is proportional to input voltage VIN with gain 1 + C R1 F
Voltage follower Figure 9.9 shows a signal source VIN, RIN connected to a voltage follower circuit. Here V− = VOUT, and since i+ = 0, V+ = VIN. Since V− = V+ then: VOUT = VIN
[9.29]
This circuit has unity gain, high input impedance and low output impedance and is used as a buffer amplifier to connect a voltage signal source with high output impedance to a low impedance load.
Differential amplifier Figure 9.10 shows two voltage sources V1, RIN and V2, RIN connected to a differential amplifier. V2, RIN and RF form a potential divider, so that:
9.2 AMPLIFIERS
Figure 9.9 Voltage follower.
Figure 9.10 Differential amplifier.
V+ = V2
A RF D C RF + RINF
Also: iIN + iF = i− = 0 where iIN =
(V1 − V−) , RIN
iF =
(VOUT − V−) RF
giving: (V1 − V−) (VOUT − V−) + =0 RIN RF and V− =
RIN ⋅ VOUT RFV1 + (RF + RIN) (RF + RIN)
Since V− = V+ RINVOUT RFV1 RFV2 + = (RF + RIN) (RF + RIN) (RF + RIN)
217
218 S IGNAL CONDITIONING ELEMENTS Figure 9.11 Strain gauge bridge connected to differential amplifier.
giving:
VOUT =
RF (V2 − V1 ) RIN
[9.30]
The output voltage VOUT is therefore proportional to the voltage difference (V2 − V1), and the closed-loop gain is RF /RIN. Figure 9.11 shows a strain gauge bridge connected to a differential amplifier. Here R1 and R3 experience tensile strain +e so that: R1 = R3 = R0(1 + Ge) while R2 and R4 experience compressive strain −e so that: R2 = R4 = R0(1 − Ge) RIN is the Thévenin resistance of the potential divider R1, R4 (also R2, R3): RIN is therefore equal to the resistance of R1 and R4 in parallel (Figure 5.7). Since the strain e is much less than 1, we have: RIN ≈ R0 /2 V2 is the potential at B and is set by the potential divider R3, R2 giving: V2 = VS
A R3 D VS = (1 + Ge) C R3 + R2 F 2
V1 is the potential at D and is set by the potential divider R4, R1 giving: V1 = VS
A R4 D VS = (1 − Ge) C R4 + R1 F 2
VOUT =
RF (V2 − V1) RIN
Since
we obtain:
9.2 AMPLIFIERS
219
Figure 9.12 A.C. amplifier.
VOUT = 2
RF VS Ge R0
[9.31]
Typically VS = 15 V, e = 10−5, R = 120 Ω and G = 2; therefore to give VOUT = 5 V, RF must be 1 MΩ.
A.C. amplifier Figure 9.12 shows an a.c. signal source VIN connected to an inverting amplifier with input resistance RIN, capacitance CIN, feedback resistance RF and capacitance CF. The circuit can be analysed using eqn [9.27] with RIN and RF replaced by complex impedances ZIN and ZF to give: VOUT = −
ZF VIN ZIN
Here ZIN is the series combination of RIN and CIN, i.e.: ZIN = RIN + (1/CIN s) where s is the Laplace operator. ZF is the parallel combination of RF and CF , i.e.: ZF =
RF RF CF s + 1
From these equations, the transfer function relating output voltage VOUT to input voltage VIN, for the a.c. amplifier, is:   RIN C IN s  R  OUT 1 =− F    RIN  RF CF s + 1  RIN C IN s + 1 IN
[9.32]
Figure 9.13 shows the amplitude frequency response characteristics of the amplifier. It has a mid-band gain of RF /R1, and a bandwidth between f1 and f2 and can be used in a.c. carrier systems (Section 9.3).
220 S IGNAL CONDITIONING ELEMENTS Figure 9.13 Ideal frequency response characteristics of a.c. amplifier.
Figure 9.14 Voltage adder.
Voltage adder Figure 9.14 shows a set of voltages V1, V2, . . . , Vn connected into a summing amplifier. Here: V− = V+ = 0 and i1 =
V1 V2 Vn , i2 = , . . . , in = R1 R2 Rn
giving: I = i1 + i2 + . . . + in =
V1 V2 Vn + +...+ R1 R2 Rn
Also iF =
VOUT , I + iF = i− = 0, iF = −I RF
giving: V V V V  VOUT = − RF  1 + 2 + 3 + . . . + n  Rn   R1 R2 R3
[9.33]
9.2 AMPLIFIERS
221
The output voltage V is the sum of the voltages V1, V2, . . . , Vn; each voltage is weighted by a factor RF /R. This voltage adder forms the basis of an analogue-todigital converter (Section 10.1). R is set equal to the resistance of R1, R2, . . . , Rn, RF in parallel in order to balance the d.c. bias currents to inverting and non-inverting inputs.
9.2.2
Limitations of practical operational amplifiers Table 9.1 summarises the characteristics of the practical OPA27 operational amplifier. The parameters VOS, γ, iB and CMRR influence the d.c. performance of the amplifier and are discussed first. The existence of input offset voltage VOS means that VOUT is unequal to zero when both V− and V+ = 0 volts, i.e. VOUT = AOL(V+ − V−) + AOLVOS
[9.34]
Some operational amplifiers have facilities for adjusting VOS to zero, i.e. for obtaining VOUT = 0 when V+ = V− = 0. However, VOS is dependent on the temperature TE °C of the amplifier environment, where γ is the appropriate temperature coefficient. Thus supposing VOS is set to zero at TE = 15 °C; then if TE subsequently increases to 25 °C, the resulting input offset voltage is γ (25 − 15), i.e. ≈ 6 µV, which causes a change of approximately 0.6 × 106 µV, i.e. 0.6 V in the output of the open-loop operational amplifier. The effect on a closed-loop amplifier is considerably reduced but may be still important. For an inverting amplifier we have: VOUT = −
A RF RF D VIN + 1 + VOS C RIN RIN F
[9.35]
With a thermocouple temperature sensor, VIN ≈ 40T µV, i.e. 40 µV change for every 1 °C change in measured temperature T. The above variation in VOS with environmental temperature therefore causes the output of the inverting amplifier to drift significantly. Ideally the output voltage should depend only on the differential voltage (V+ − V−) and should be independent of the common mode voltage VCM = (V+ + V−)/2. For a practical operational amplifier, however, we have: VOUT = AOL(V+ − V−) + ACMVCM
[9.36]
where ACM is the common mode gain. A more commonly used term is Common Mode Rejection Ratio (CMRR), where: CMMR =
AOL ACM
or in decibels (CMMR)dB = 20 log10
A AOL D C ACM F
[9.37]
222 S IGNAL CONDITIONING ELEMENTS This gives: VCM J G VOUT = AOL (V+ − V−) + I CMRRL
[9.38]
VCM CMRR
i.e. the equivalent circuit for an open-loop amplifier. Typically (CMRR)dB = 110 dB, i.e. CMRR ≈ 3 × 105. For a closed-loop differential amplifier (Figure 9.10), we have: VOUT ≈
A RF RF D VCM (V2 − V1) + 1 + C RIN RIN F CMRR
[9.39]
If the differential amplifier is used with the strain gauge bridge described in Section 9.2.1 (Figure 9.11), then we have: V2 − V1 = VS Ge = 300 µV Also VCM = 7.5 V, so that VCM /CMRR = 25 µV. Thus the common mode voltage can have a significant effect. However, much greater errors can occur in a differential amplifier if there is any mismatch between the values of the two input resistances RIN and the two RF resistances. The a.c. performance of a practical operational amplifier is determined by its dynamic characteristics. These are adequately represented by a first-order lag, i.e. ∆VOUT AOL (s) = ∆(V+ − V−) 1 + τs
[9.40]
where AOL is the d.c. open-loop gain, τ is the time constant, and ∆(V+ − V−) and ∆VOUT are corresponding small changes in input and output. The variation in open-loop amplifier gain with frequency f is thus given by: Gain–frequency relation for open-loop amplifier
AOL( f ) =
Ä OUT AOL (f) = (Ê ) 1 + ( f / f B )2
[9.41]
where f B = 1/2πτ is the −3 dB cut-off frequency. The open-loop gain–bandwidth product AOL × fB is typically ≈ 106 × 10 = 107 Hz. Figure 9.15 shows the variation in gain with frequency for both open- and closedloop amplifiers. We note that the lower the gain of the closed-loop amplifier the greater the bandwidth. This is because the gain–bandwidth product of a closed-loop amplifier is also equal to the open-loop value of AOL fB. The frequency response of a practical a.c. amplifier (Figures 9.12 and 9.13) depends, to some extent, on the frequency characteristics of the operational amplifier as well as on those of the external components.
9.2 AMPLIFIERS
223
Figure 9.15 Typical gain–frequency characteristics for operational amplifier.
9.2.3
Instrumentation amplifiers An instrumentation amplifier is a high-performance differential amplifier system consisting of several closed-loop operational amplifiers. An ideal instrumentation amplifier gives an output voltage which depends only on the difference of two input voltages V1 and V2, i.e. VOUT = K(V2 − V1)
[9.42]
where the gain K is precisely known and can be varied over a wide range. A practical instrumentation amplifier should have a gain which can be set by a single external resistor and should combine the following:
• • • •
High input impedance High common mode rejection ratio Low input offset voltage Low temperature coefficient of offset voltage.
The differential amplifier already discussed (Figure 7.4(d)), which uses a single operational amplifier, is inadequate: in order to obtain high gain, RIN must be low. This means low input impedance; also the differential amplifier is highly susceptible to any mismatch in resistance values. Figure 9.16 shows a typical instrumentation amplifier system consisting of three operational amplifiers, A1, A2 and A3. The two input non-inverting amplifiers A1 and A2 provide an overall differential gain of (1 + 2R1/RG) and a common mode gain of unity. The output amplifier A3 is a unity gain differential amplifier. A typical precision instrumentation amplifier, INA 115, has:[2]
• • • •
ZIN ≈ 1010 Ω in parallel with 6 pF. (CMRR)dB ≈ 115 dB min. Offset voltage VOL ≈ 50 µV max. Temperature coefficient of offset voltage γ ≈ 0.25 µV °C −1.
224 S IGNAL CONDITIONING ELEMENTS Figure 9.16 Typical instrumentation amplifier.
9.3
A.C. carrier systems Two of the most difficult problems in conditioning low-level d.c. signals from sensors are external interference in the signal circuit (Section 6.4) and amplifier drift (Section 9.2.2). These problems can be avoided if the signal is converted into a.c. form, amplified and then reconverted into d.c. The initial or primary elements in an a.c. carrier system are usually of one of the following types: (a) (b)
R, L, C sensors in a suitable deflection bridge (Section 9.1) An LVDT sensor (Section 8.3).
The output voltage of a four-element strain gauge bridge with a cantilever load cell is given by ETh = GVS e (eqn [9.18]), i.e. the output signal is proportional to the product of the supply voltage VS and the strain signal e. Similar results are obtained with other bridge circuits (eqns [9.15], [9.17], [9.19], [9.20], [9.23] and [9.25]); in each case the bridge output voltage is proportional to the product of the supply voltage and the measured variable. The output voltage of the LVDT sensor (Section 8.3) is also approximately proportional to the product of supply voltage and displacement x of the core from the null point. In an a.c. carrier system the supply voltage is a.c., i.e. VS = ÄS sin 2πfS t, with a frequency fS of typically a few kHz. The measured variable is also a function of time; in the case of a step change in force on the cantilever, the strain variation e(t) is a damped sine wave. In the general case e(t) is a complicated function of time which can be expressed as a sum of many sine waves (Sections 4.3 and 6.2): i m
e(t) =
∑ ê sin 2πf t i
i
i 1
Figure 9.17 shows a typical measurement signal frequency spectrum containing frequencies from f1 = 0 Hz up to fm (often only a few Hz). The bridge output voltage is thus given by: i m
ETh =
∑ Ä sin 2πf t sin 2πf t i
i 1
i
S
[9.43]
9.3 A.C. CARRIER SYSTEMS
225
Figure 9.17 Corresponding waveforms and frequency spectra for amplitude modulation: (a) Measurement signal (b) Supply voltage (c) AM with sinusoidal modulating signal (d) Complete AM waveform.
where Äi = GÄS êi. If we, for the moment, consider only the ith component: Vi = Äi sin 2πfi t sin 2πfS t then we have a sine wave of frequency fS and amplitude Äi sin 2πfi t. This is amplitude modulation (AM); here the amplitude of the carrier signal frequency fS is being altered according to the instantaneous value of the modulating signal frequency fi (Figure 9.17). Using the identity: sin A sin B = –12 cos(A − B) − –12 cos(A + B) the amplitude modulated signal Vi = Äi sin 2πfS t sin 2πfi t
226 S IGNAL CONDITIONING ELEMENTS can be expressed in the form: Vi =
Äi Äi cos 2π( fS − fi )t − cos 2π( fS + fi )t 2 2
[9.44]
Thus the AM signal is the sum of two cosinusoidal signals with frequencies fS − fi and fS + fi; the corresponding frequency spectrum (Figure 9.17) contains two lines. This form of amplitude modulation is called balanced amplitude modulation, because when êi = 0, Äi = 0, i.e. the bridge is balanced. This means also that the frequency spectrum does not contain the carrier frequency fS; there are only two lines at the sideband frequencies fS ± fi. Hence the alternative name is double sideband carrier suppressed amplitude modulation. We see that each frequency fi in the modulating signal spectrum contributes two lines to the AM spectrum, e.g. if fS = 1 kHz, then fi = 1 Hz gives lines at 999 and 1001 Hz. The AM spectrum corresponding to the total modulating signal (0 to fm Hz) consists therefore of frequencies between fS − fm and fS + fm, e.g. between 995 and 1005 Hz if fS = 1 kHz and fm = 5 Hz. We see therefore that the operation of multiplication and thence modulation has the effect of shifting the spectrum of the measurement signal from low frequencies to higher frequencies. The amplitudes Äi are small but an a.c. amplifier (Figures 9.12 and 9.13) can now be used. The bandwidth frequencies f1 and f2 are chosen so that they include the measurement signal between fS − fm and fS + fm (Figure 9.17), but exclude low frequency drift voltages and 50 Hz mains interference. In the above example, an amplifier with f1 = 900 Hz and f2 = 1100 Hz will successfully amplify the measurement signal and reject drift and interference. It is important to realise that this system only rejects drift and interference signals which are introduced after the modulator; for example, drift in bridge output voltage due to drift in bridge supply voltage will be regarded as part of the measurement signal. The AM signal must now be demodulated in order to obtain an output voltage which is proportional to the strain signal e(t). Figure 9.18 shows corresponding waveforms for e(t) and supply voltage VS (t); we see that at time interval t1, e(t1) = +e1 and at time interval t2, e(t2) = −e1. Thus at time t1 the cantilever is below the neutral position and gauges 1 and 3 are in tension. At time t2 the cantilever is above the neutral position and gauges 1 and 3 are in compression. The corresponding bridge output voltages are: ETh = +Ge1ÄS sin 2πfS t
at t1
and ETh = −Ge1ÄS sin 2πfS t = Ge1ÄS sin(2πfS t + 180) at t2 i.e. the signals have equal amplitude but are 180° out of phase (Figure 9.18). Thus in order to distinguish between the situations at t1 and t2, we need a demodulator which detects this phase difference and gives a positive output voltage when e is positive and a negative voltage when e is negative. Such a demodulator is called a phase-sensitive demodulator (PSD). The simplest form of PSD is an amplifier with a gain of ±1 according to the sign of the supply voltage; that is, VPSD = ETh when VS > 0 # VPSD = −ETh when VS < 0 $
[9.45]
9.3 A.C. CARRIER SYSTEMS
227
Figure 9.18 Phasesensitive demodulation.
The action of this element is shown in Figure 9.18. We see that a low-pass filter (Section 6.5) is required to give the mean value of the output signal. The bandwidth of this filter should be less than fS in order to filter out carrier frequencies, but greater than fm in order not to filter out measurement signal frequencies. The complete a.c. carrier system is shown in Figure 9.19. We have seen above that an a.c. bridge can be used to generate an a.c. voltage from a resistive, capacitive or inductive sensor. This raises the question of how to convert a low-level d.c. voltage signal, e.g. a thermocouple e.m.f., into a.c. form. The usual solution here is to sample the signal at high speed, around 1000 times per second, using a transistor switch or ‘chopper’ driven by an oscillator. The resulting ‘chopped’ signal is a series of pulses whose amplitudes correspond to the instantaneous value of the d.c. voltage; this is pulse amplitude modulation (PAM). The PAM signal can be processed by an a.c. amplifier and phase-sensitive demodulator so that the chopper simply replaces the multiplier in Figure 9.19. The resulting chopperstabilised amplifier has excellent drift characteristics.
Figure 9.19 Complete a.c. carrier system.
228 S IGNAL CONDITIONING ELEMENTS
9.4
Current transmitters In Section 5.1 we saw that sensing elements could be represented by either a Thévenin voltage source or a Norton current source. In Section 6.3 we saw that a current transmission system rejects series interference more efficiently than a voltage transmission system. In this section we discuss transmitters used in the process industries to give an output d.c. current signal in the standard range 4 to 20 mA proportional to the input measured variable. The output has a live zero of 4 mA: this enables a distinction to be made between a zero input situation (output = 4 mA) and a fault situation (output = 0 mA). These transmitters were initially closed-loop systems with high-gain negative feedback (Section 3.3). Here the overall transmitter sensitivity depends mainly on the sensitivity of the elements in the feedback path and is largely unaffected by changes in the characteristics of forward path elements, for example amplifier drift. The zero and span of these transmitters can be adjusted but, in the case of force balance systems, the adjustments have to be made mechanically, for example by adjusting the position of a pivot. Closed-loop transmitters have now been largely replaced by open-loop transmitters. Here the overall transmitter sensitivity depends on the sensitivity of every element in the system; these element sensitivities can change due to modifying inputs and/or non-linear effects. However, by using precision mechanical components and high-performance electronic integrated circuits, the above effects are minimised and open-loop transmitters are now widely used. They have the advantage that zero and span adjustments can be made electronically, which is more convenient than mechanical adjustment. The most recent development is that of intelligent or ‘smart’ transmitters: here a microcontroller is incorporated into an open-loop transmitter.[3] This results in an improvement in overall accuracy and enhanced flexibility and versatility.
9.4.1
Closed-loop transmitters Figure 9.20(a) shows a simplified block diagram of a closed-loop differential pressure transmitter operating on the force balance or, more accurately, torque balance principle. The diaphragm force (P1 − P2)AD produces a clockwise torque AD(P1 − P2)a on the lever arm, which is supported by a clockwise torque F0 b due to the zero spring. The output current i is fed back to a coil, inside a permanent magnet; this produces a force KM i on the lever arm and a corresponding anticlockwise torque KM ib. Any imbalance of these torques causes the lever to rotate; the resulting displacement x is measured by a sensor–bridge–amplifier system. The resulting change in output current is fed back to the coil and magnet to adjust the anticlockwise torque until an approximate torque balance is again obtained. Inspection of Figure 9.20(a) gives KD KBKA /k [aAD(P1 − P2 ) + bF0] 1 + KD KB KA bKM /k
[9.46]
If the gain KA of the amplifier system is high so that
KD KB KA bKM is very much greater k
i=
than 1, then eqn [9.46] reduces to:
9.4 CURRE N T TRAN S MI T T E R S
229
Figure 9.20 Closed-loop differential pressure transmitters: (a) Simplified block diagram (b) Simplified schematic diagram.
i=
a AD F0 (P1 − P2) + b KM KM
[9.47]
From eqn [9.47] we see that transmitter sensitivity = (a/b)(AD /KM) and zero bias = F0 /KM . The closed-loop transmitter has the advantage, therefore, that the overall sensitivity is independent of diaphragm stiffness, displacement sensor sensitivity and amplifier gain, so that once set it should remain constant and not drift. The main disadvantage of this type of transmitter is that sensitivity (input span) and zero adjustments must be made mechanically; the input span is adjusted by altering the
230 S IGNAL CONDITIONING ELEMENTS ratio a/b and the zero by altering F0. For example, suppose AD = 10−2 m2 (circular diaphragm approximately 11 cm in diameter) and KM = 103 N A−1 Then in order to obtain an output range of 4 to 20 mA, corresponding to 0 to 104 Pa (0 to 1 metre water), we require a/b = 0.16, f0 = 4 N. Figure 9.20(b) shows a simplified schematic diagram of a two-lever torque balance differential pressure transmitter using an inductive displacement sensor.
9.4.2
Open-loop transmitters Open-loop differential pressure transmitters Figure 9.21(a) shows a general simplified block diagram for an open-loop differential pressure transmitter. The input differential pressure (P1 − P2) acting over the area AD of the sensing diaphragm produces a deflecting force AD(P1 − P2) which is opposed by the elastic spring force kx. The resulting diaphragm deformation x is small (≈ 10−5 m) and approximately proportional to (P1 − P2). This displacement can be measured using variable capacitance, variable reluctance or resistive strain gauge sensors. Another possibility is to use a piezoresistive sensor consisting of a silicon diaphragm with diffused p- or n-type material to give four strain gauges. With strain gauge elements, a d.c. four-element strain gauge bridge is used; the amplifier system consists of an instrumentation amplifier and output current regulator. With capacitive or inductive sensors, an a.c. bridge is used with a carrier amplifier system consisting of an a.c. amplifier, phase sensitive demodulator, d.c. amplifier and output current regulator. Span and zero adjustments are made electronically via the amplifier system. Figure 9.21(b) shows a differential capacitance displacement sensor. The input differential pressure causes the sensing diaphragm to be displaced a small distance x to the left. This in turn causes C1 to fall from C0 to C0 − ∆C and C2 to increase from C0 to C0 + ∆C, where ∆C is the change in capacitance proportional to (P1 − P2). The capacitances are incorporated in an a.c. bridge circuit, similar to Figure 9.5(a), which has: Z1 =
1 1 , Z2 = Z3 = R, Z4 = jω C1 jω C2
The bridge output voltage is therefore given by: ∆V = VS = VS
A C2 1D − C C1 + C2 2 F A C0 + ∆C 1D − C C0 − ∆C + C0 + ∆C 2 F
[9.48]
which simplifies to:
∆V =
VS ∆C 2C0
[9.49]
9.4 CURRE N T TRAN S MI T T E R S
Figure 9.21 Open-loop differential pressure transmitters: (a) General block diagram; (b) Differential capacitance – principle; (c) Differential capacitance – block diagram; (d) Application – flow measurement; (e) Application – level measurement open vessel; (f) Application – level measurement closed vessel.
231
232 S IGNAL CONDITIONING ELEMENTS Thus ∆V is proportional to ∆C and the bridge sensitivity KB is equal to VS /2C0. Figure 9.21(c) shows a block diagram of the complete transmitter; typically the overall accuracy is ±0.25%. This type of transmitter is widely used in the process industries for measuring pressure, flow rate and liquid level. The transmitter can be used for measuring gauge pressure (pressure relative to atmosphere) by opening the low pressure chamber to atmosphere, i.e. setting P2 equal to atmospheric pressure Pa . By evacuating and sealing the low pressure chamber, i.e. setting P2 equal to zero, the transmitter can be used for measuring absolute pressure (pressure relative to vacuum). Figure 9.21(d) shows the use of the transmitter with a differential pressure flowmeter to measure mass flow rate Ç of a fluid in a pipe; here differential pressure (P1 − P2) is proportional to Ç 2 (Section 12.3.1). Figures 9.21(e) and (f ) show the application of the transmitter to the measurement of liquid level in open and closed vessels; in both cases differential pressure is proportional to the height h of liquid.
Open-loop temperature transmitters
Figure 9.22 Open-loop temperature transmitters: (a) Millivolt input (b) Resistance input.
Figure 9.22 shows block diagrams for two types of open-loop transmitter. Both use an a.c. carrier amplifier system (Section 9.3) and give a 4–20 mA d.c. current output. In the millivolt input type, Figure 9.22(a), the input is a thermocouple e.m.f. ET1,T2. The automatic reference junction circuit provides an e.m.f. equal to ET2,0; this is added to ET1,T2 to give a compensated e.m.f. ET1,0 which is relative to a fixed reference junction temperature of 0 °C (Section 8.5). This e.m.f. is then ‘chopped’ to give the a.c. voltage necessary for the carrier system. This type of transmitter can be located close to the process so that the signal is transmitted to a remote control room as a 4–20 mA current using normal leads. This is simpler and less prone to
9.4 CURRE N T TRAN S MI T T E R S
233
external interference than the alternative of transmitting the basic millivolt e.m.f. ET1,T2 using extension leads as in Figure 8.17(d). In the resistance input type, Figure 9.22(b), the input is the resistance RT of a resistance sensor. This is incorporated into a deflection bridge circuit which gives a voltage output ∆V proportional to temperature T (Section 9.1.1 and eqn [9.15]). Problem 9.16 shows a four-lead resistance sensor with two sensor leads and two dummy leads, incorporated into a deflection bridge circuit. Here the bridge output voltage is independent of the lead resistance RC and therefore the temperature of the plant environment. The a.c. bridge output voltage necessary for the carrier system is obtained either by using an a.c. bridge supply or by using a d.c. bridge supply and ‘chopping’ the resulting d.c. bridge output voltage.
9.4.3
Intelligent or smart transmitters By incorporating a microcontroller into the basic open-loop transmitter, not only is the accuracy of the transmitter increased but flexibility and versatility are enhanced. An intelligent or smart transmitter can perform the following functions. Computer calculation of measured value (Section 3.3). Here the microcontroller solves the inverse equation to calculate an accurate value of the measured variable from the sensor output signal. This means that non-linear, hysteresis and environmental effects in the sensor can be compensated for. The overall accuracy of a smart transmitter is therefore typically ±0.1% of span compared with ±0.25% for a conventional transmitter. This accurate measured value in engineering units is then converted into a standard 4–20 mA current output. Remote diagnosis of faults. Signals at different points in the transmitter are continuously monitored and the actual values compared with expected values. Both the existence and the nature of a fault can be detected in this way. Remote re-ranging. Span and zero adjustments can be made remotely from the control room rather than at the transmitter itself. Remote identification. The transmitter identification number and maintenance records are stored in the computer memory; this information can again be accessed remotely. Digital communications. A digital communications signal can be superimposed on the 4–20 mA analogue current signal. Local indication of measured value. An analogue indication or digital display of measured value is available at the transmitter. Figure 9.23 is a block diagram of a smart differential pressure transmitter.[3] The differential capacitance pressure sensor of Figure 9.21(b) gives a change in capacitance ∆C which depends on input differential pressure (P1 − P2). This is converted into a d.c. voltage V1 by the signal conditioning system of Figure 9.21(c). The relationship between (P1 − P2) and V1 is non-linear and is also affected by changes in the temperature of the sensor. ∆T is the deviation in temperature from standard conditions.
234 S IGNAL CONDITIONING ELEMENTS
Figure 9.23 ‘Smart’ differential pressure transmitter.
Calibration of a given sensor/signal conditioning element gives corresponding values of (P1 − P2), V1 and ∆T and an inverse equation of the form (eqn [3.17]): P1 − P2 = K′1 V1 + N′1 (V1) + a′1 + K′M ∆TV1 + K′I ∆T
[9.50]
The coefficients K 1′, a 1′, etc., are specific to an individual element. Similarly ∆T is measured with a temperature sensor and appropriate signal conditioning element to give a d.c. output voltage V2. Here the inverse equation is: ∆T = K′2 V2 + N′2 (V2) + a′2
[9.51]
where again K′2, N 2( ) and a′2 are specific to an individual element. The coefficients K′1, N′1( ), K 2′, N′2( ), etc., are stored in a sensor memory which is identified with the individual differential pressure and temperature sensors. The voltages V1 and V2 are then input to a time division multiplexer (Section 18.1) which transfers them one at a time to an analogue-to-digital converter (ADC) (see Section 10.1). The ADC gives a parallel digital output signal proportional to V1 or V2. This is fed to a microcontroller along with the inverse equation coefficients from the sensor memory. The microcontroller solves eqns [9.51] and [9.50] to give an accurate measured value of the differential pressure (P1 − P2) in engineering units. If the transmitter is being used to measure flow rate or liquid level, the measured value of flow rate or level is calculated in engineering units from (P1 − P2). This measured value is then output from the microcontroller to a digital-to-analogue converter (DAC) (see Section 10.1). The DAC forms a 4–20 mA current loop with a d.c. power
9.5 OSCILLATORS AND RESONATORS
235
supply and a series resistor; an analogue indicator and/or recorder is connected into the loop to display the measured value of (P1 − P2). The measured value of (P1 − P2 ) is also transferred from the computer to a digital communications module. This gives a serial digital signal corresponding to (P1 − P2) using frequency shift keying (FSK) (see Section 18.6). This is a sinusoidal signal at two distinct frequencies, one corresponding to a binary ‘1’ and the other to binary ‘0’. The FSK signal (typical amplitude of 0.5 mA) is then superimposed on the 4–20 mA analogue signal; this means that no d.c. component is added to the 4–20 mA signal (Section 18.7). Communication between the transmitter and the outside world is via a hand-held communicator which is connected into the current loop. There is two-way communication between the communicator and the microcontroller via the digital communications module. Thus a technician can transmit information requesting a range change and receive information back confirming that the change has been implemented.
9.5
Oscillators and resonators In Section 9.3 we studied a.c. carrier systems, where the amplitude of a sinusoidal voltage depends on the magnitude of the measured variable – amplitude modulation (AM); these systems reject external interference added after the modulator. An alternative is frequency modulation (FM), where the frequency of a sinusoidal voltage depends on the magnitude of the measured variable. FM systems also reject interference and have the following advantages over AM systems: (a) (b)
The external interference affects signal amplitude more than it affects signal frequency; FM is therefore inherently more resistant to interference than AM. By counting pulses over a fixed time interval, a frequency signal can easily be converted to digital form (Section 10.1.4).
Examples of sensors which directly generate frequency signals are the variable reluctance tachogenerator (Section 8.4), the turbine flowmeter (Section 12.3.2) and the vortex flowmeter (Section 12.3.3). Because of the advantages of frequency signals, however, we need general methods for producing them. These are oscillators, which are purely electrical, and resonators, which are electromechanical. Oscillators and resonators are feedback systems which rely on the dynamic property of a closed-loop system, consisting of elements with dynamic characteristics, to sustain continuous oscillations, if certain conditions are satisfied. Figure 9.24 shows a closed-loop system consisting of a forward path element with transfer function G(s) and a feedback path element with transfer function H(s). Suppose that an input signal I = Î sin ω t is applied to the system with the feedback link broken at A. The corresponding feedback signal F is also sinusoidal and is given by (Section 4.2): F = |G( jω)| |H( jω)|Î sin(ω t + φ)
[9.52]
where:
φ = arg G( jω) + arg H( jω)
[9.53]
If we make the product of magnitudes | G | × |H | equal to 1 and the sum of arguments (phase angles) equal to −180°, then F = −Î sin ω t, i.e. F has the same amplitude as
236 S IGNAL CONDITIONING ELEMENTS Figure 9.24 Principle of oscillator/resonator.
the input signal but is 180° out of phase. If the feedback link is now closed, then e = 2Î sin ω t, but if I(t) is reduced to zero, then e = Î sin ω t and the system remains in continuous oscillation at frequency ω. The conditions for continuous oscillation at ω are therefore: Conditions for continuous oscillation at ω
|G( jω)| × | H( jω)| = 1 arg G( jω) + arg H( jω) = −180°
[9.54]
Under these conditions there is a sinusoidal output signal at frequency ω for no input signal.
9.5.1
Oscillators Here the feedback element H(s) is an L–C–R circuit and G(s) is an amplifier (maintaining amplifier). In Figure 9.25(a) the L–C–R circuit consists of an inductive sensor (pure inductance with series loss resistance) in series with a fixed pure capacitance; in Figure 9.25(b) the L–C–R circuit consists of a capacitive sensor (pure capacitance with parallel loss resistance) in series with a fixed pure inductance. From Figure 9.25 we see that each circuit is described by a second-order differential equation and a corresponding second-order transfer function of the form: H (s) =
1 1 2 2ξ s + s +1 ω n2 ωn
[9.55]
The figure also gives expressions for the natural frequency ωn, the damping ratio ξ and the Q factor for the circuits. Since ωn = 1/êLúCú ú for both circuits, a change in L for the inductive sensor or a change in C for the capacitive sensor both result in a change in circuit natural frequency. We therefore design the system to oscillate at ωn. From eqn [9.55] we have: H ( jω ) =
1  ω  ω 1 − ω 2  + j 2ξ ω  n n 2
and
H ( jω n ) =
1 j 2ξ
[9.56]
9.5 OSCILLATORS AND RESONATORS
237
Figure 9.25 Oscillators: (a) Inductive sensor (b) Capacitive sensor.
LC
d 2VF dt
2
H (s) =
ωn =
+ RC
dVF + VF = VO dt
LC
∆F 1 (s) = ∆O LCs2 + RCs + 1 1 = 1 2 2ξ s + s +1 ωn ω n2 1 LC
,ξ =
R 2
C 1 ,Q = L R
for small changes: ∆ω n ≈ −
L C
1 ω n0 ∆L 2 LO
(a)
d 2VF dt 2
H (s) =
ωn =
+
L dVF + VF = VO R dt
∆F (s) = ∆O
1
L s +1 R 1 = 1 2 2ξ s + s +1 ωn ω n2
1 LC
,ξ =
LCs +
1 2R
2
L C ,Q = R C L
for small changes: ∆ω n ≈ −
1 ω n0 ∆C 2 CO
(b)
From eqns [9.54], in order to sustain oscillations at ω = ωn, the gain and phase of the maintaining amplifier must satisfy the conditions: Gain and phase conditions for maintaining amplifier in oscillator
| G( jω n) | =
1 1 = 2ξ = | H( jωn) | Q
[89.57]
arg G( jω n) = −180° − arg H( jωn ) = −90° over the range of ωn corresponding to the range of variation of either L or C. Figure 9.26 shows a frequency system for measuring sinusoidal variations in pressure. The capacitive pressure sensor of Figure 8.9 is used; for small changes, the
Figure 9.26 Frequency pressure measurement system.
238 S IGNAL CONDITIONING ELEMENTS Figure 9.27 Block diagram of resonator.
increase in capacitance ∆C is proportional to the applied pressure P (eqn [8.22]). The sensor is incorporated in the oscillator circuit of Figure 9.25(b); the change ∆ωn in circuit natural frequency is approximately proportional to ∆C. The frequency of the oscillator output signal varies sinusoidally with time, giving rise to the frequency modulated signal shown.
9.5.2
Resonators Figure 9.27 shows the block diagram for a resonator.[4] The forward path element G(s) is again an amplifier; the feedback element H(s) consists of:
• • •
A drive element to convert voltage to force An elastic force sensor to convert force to displacement A displacement sensor to convert displacement into voltage.
From Figure 9.27 we have: H (s) =
KS KD 1 k 1 2 2ξ s + s +1 ω n2 ωn
[9.58]
The figure also gives expressions for natural frequency ωn, damping ratio ξ and Q factor. Since ωn = êk/úmú , changes in either effective stiffness k or effective mass
9.5 OSCILLATORS AND RESONATORS
239
m cause corresponding changes in ωn. Thus provided the system is designed to oscillate at ωn, it could be used, for example, to sense force variations (affecting k) or density variations (affecting m). From eqns [9.54] and [9.58], in order to sustain oscillations at ω = ωn, the gain and phase of the maintaining amplifier must satisfy the conditions: Gain and phase conditions for maintaining oscillations in a resonator
| G( jω n ) | =
1 2ξ k λ = = | H ( jω n ) | K S KD K S KD
k m
[9.59]
arg G( jωn) = −180° − arg H( jωn) = −90° over the range of ωn, corresponding to the range of variation in either m or k. Figure 9.28(a) shows a vertical section through a resonant sensor for measuring absolute pressure. The sensor uses a vibrating flat circular plate. If the plate is clamped around the circumference and subject to a radial tensile force F, then the fundamental frequency ωn of transverse vibration is given by:
ωn = 2πfn = K
F σ
[9.60]
where σ is the mass per unit area of the disc and K is a constant. Therefore if F is changed, fn changes. The flat plate is a circular silicon wafer which is clamped around its circumference by pier supports; these form part of a silicon structure. The Figure 9.28 Examples of resonators: (a) Vibrating plate element (b) Vibrating tube element.
240 S IGNAL CONDITIONING ELEMENTS plate is enclosed in a sealed vacuum chamber. An increase in the downward pressure P applied to the pressure sensitive diaphragm at the top produces an increase in the tensile force F on the plate and a corresponding increase in natural frequency fn. Two metal electrodes, an inner circular electrode and an outer annular electrode, are deposited on the bottom surface of the plate. Corresponding and concentric electrodes are deposited on the upper surface of the substrate to form two capacitances. The outer capacitance CS is the sensor capacitance, which detects the transverse displacement x of the plate. Changes in CS are converted into a d.c. voltage signal using an a.c. bridge, an a.c. amplifier and a phase sensitive detector. The inner capacitance CD provides the drive element; if a voltage V is applied across the plates then the plates acquire equal and opposite charges, causing an attractive force between the plates. The drive force per unit voltage KD is given by: KD =
ε 0 AV d2
[9.61]
where ε 0 is the permittivity of free space, A the area of the plates and d their separation. The relationship between pressure P and natural frequency fn is non-linear and is also affected by the temperature T of the plate. Temperature is measured using a forward-biased diode within the substrate; the relationship between output voltage V and T is again non-linear. The overall inverse equation for pressure is:[5] 3
P=
3
∑∑ K
ij
f niV j
[9.62]
i 0 j 0
This equation involves powers of fn, powers of V and cross-terms such as fn V 2; there are 16 calibration constants Kij. Figure 9.28(b) shows a vibrating thin-walled tube element, inside which a fluid is flowing.[6] The tube is clamped at each end and a transverse force is applied to the tube using a drive coil connected to the maintaining amplifier. This causes transverse deflection of the tube centre, which is sensed by a pick-up coil and fed back to the amplifier. By correct choice of amplifier gain and phase characteristics, the tube is continuously maintained in transverse oscillation at a fundamental frequency fn = (1/2π)(êk/úmú ). The total vibrating mass m and therefore fn depend on the fluid density ρ, as well as on the mass of the tube itself. The inverse equation for density is:
ρ=
A B + +C f n2 f n
[9.63]
where A, B and C are constants. This tube arrangement is particularly useful for liquids and liquid–solid mixtures.
Conclusion This chapter has studied methods of converting the output of sensing elements into a form suitable for further processing; this is usually a d.c. voltage, d.c. current or variable frequency a.c. voltage. The chapter began by examining the principles of bridge circuits, which convert a change in resistance, capacitance or inductance into a d.c. or a.c. voltage. It then went on to study the principles and characteristics of ideal and practical operational amplifiers. The third section looked at a.c. carrier
PROBLEMS
241
systems where amplitude modulation is used to reject external interference. In the fourth section three types of current transmitter were studied: closed-loop, openloop and intelligent; these give a standard 4 to 20 mA output signal. The final section examined oscillators and resonators, which give a variable frequency signal, i.e. frequency modulation.
References [1] [2] [3] [4] [5] [6]
Burr-Brown 1999 I C Data Book – Linear Products, OPA27: ultra low noise precision operational amplifier. Burr-Brown 1999 I C Data Book – Linear Products, INA115: precision instrumentation amplifier. Rosemount Ltd 1989 Technical Information on Model 3051C Smart Differential Pressure Transmitter. langdon r m 1985 ‘Resonator sensors – a review’, J. Phys. E: Scientific Instruments, vol. 18, pp. 103–15. frost d (Druck Ltd) 1999 ‘Resonant silicon pressure transducers’, Sensors and Transducers Conf., NEC Birmingham. Sarosota Automation Instrumentation Division 1985 Technical Information on FD 800 Series Density Meters.
Problems 9.1
A platinum resistance sensor has a resistance of 100 Ω at 0 °C and a temperature coefficient of resistance of 4 × 10−3 °C −1. Given that a 15 V supply is available, design a deflection bridge giving an output range of 0 to 100 mV for an input range of 0 to 100 °C: (a) (b)
using the procedure summarised by eqns [9.7] and [9.8]; using the linear approximation of eqn [9.15].
Give values for all circuit components and assume a high impedance load. (c)
How should the circuit be altered if the input range is changed to 50 to 150 °C?
9.2
A thermocouple has an e.m.f. of 4.1 mV at 100 °C and 16.4 mV at 400 °C relative to a cold junction of 0 °C. A deflection bridge incorporating a nickel metal resistance sensor is to be used as the voltage source ET2,0 necessary for automatic reference compensation of the thermocouple e.m.f. The nickel sensor has a resistance of 10 Ω at 0 °C and a temperature coefficient of resistance of 6.8 × 10−3 °C −1. Design the deflection bridge assuming a 1 V supply is available. (Hint: eqns [8.49] and [9.15].)
9.3
The resistance Rθ kΩ of a thermistor at θ K is given by: G A1 1 DJ − Rθ = 1.68 exp 3050 I C θ 298F L The thermistor is incorporated into the deflection bridge circuit shown in Fig. Prob. 3. (a)
Assuming that VOUT is measured with a detector of infinite impedance, calculate: (i) (ii)
(b)
the range of VOUT corresponding to an input temperature range of 0 to 50 °C; the non-linearity at 12 °C as a percentage of full-scale deflection.
Calculate the effect on the range of VOUT of reducing the detector impedance to 1 kΩ.
242 S IGNAL CONDITIONING ELEMENTS Figure Prob. 3.
9.4
The resistance Rθ Ω of a thermistor varies with temperature θ K according to the following equation: Rθ = 0.0585 exp
A 3260 D C θ F
Design a deflection bridge, incorporating the thermistor, to the following specification:
• • • 9.5
Input range 0 to 50 °C Output range 0 to 1.0 V Relationship between output and input to be approximately linear.
Four strain gauges, with specification given below, are available to measure the torque on a cylindrical shaft 4 cm in diameter connecting a motor and load. (a)
Draw clearly labelled diagrams showing: (i) (ii)
(b)
the arrangement of the gauges on the shaft; the arrangement of the gauges in the bridge circuit, for optimum accuracy and sensitivity.
Calculate the maximum achievable bridge out-of-balance voltage for an applied torque T of 103 N m given the following: Tensile and compressive strains = ±T/π Sa3 where S = 1.1 × 1011 N m−2 is the shear modulus of the shaft material and a is the radius of the shaft in metres. Strain gauge data: Resistance = 120 Ω Gauge factor = 2.1 Maximum current = 50 mA
9.6
A load cell consists of a domed vertical steel cylinder 20 cm high and 15 cm in diameter. Four flat surfaces, at right angles to each other, are cut on the vertical surface so as to form 10 cm squares. Resistance strain gauges are attached to these flat surfaces so that two gauges (on opposite faces) suffer longitudinal compression and two gauges (on the other pair of opposite faces) suffer transverse tension. The strain gauges have the following specification: Resistance = 100 Ω Gauge factor = 2.1 Maximum gauge current = 30 mA
PROBLEMS
243
The gauges are connected in a temperature-compensated bridge and the out-of-balance signal is input to a differential amplifier. Calculate the minimum amplifier gain if the amplifier output voltage is to be 1 V for a compressive force of 105 N. Young’s modulus for steel = 2.1 × 1011 N m−2 Poisson’s ratio for steel = 0.29 9.7
The unbonded strain gauge accelerometer of Figure 8.20 and Section 9.1.2 has a natural frequency of 175 Hz and an input range of 0 to 5g. Given that the strain due to prestressing is twice the maximum acceleration-induced strain, use the strain gauge data given below to calculate the range of the bridge output voltage. Resistance = 120 Ω Gauge factor = 2.1 Maximum current = 50 mA Length at zero acceleration = 2.3 cm
9.8
The capacitance level transducer of Section 8.2 and Figure 8.9 is to be used to measure the depth h of liquid in a tank between 0 and 7 m. The total length l of the transducer is 8 m and the ratio b/a of the diameters of the concentric cylinders is 2.0. The dielectric constant ε of the liquid is 2.4 and the permittivity of free space ε0 is 8.85 pF m−1. The transducer is incorporated into the deflection bridge of Figure 9.5(a) with R2 = 100 Ω, R3 = 10 kΩ and Äs = 15 V. (a) (b) (c)
9.9
Figure Prob. 9 shows a variable reluctance force sensor which is incorporated into the bridge circuit of Figure 9.5(b). When the applied force is zero the armature is positioned along the centre line AB. (a)
Figure Prob. 9.
Calculate the value of C0 so that the amplitude ÊTh is zero when the tank is empty. Using this value of C0 calculate ÊTh at maximum level. Explain why the relationship between ÊTh and h is non-linear and calculate the nonlinearity at h = 3.5 m as a percentage of full-scale deflection.
Explain why the sensor would be suitable for measuring force signals containing frequencies between 0 and 10 Hz.
244 S IGNAL CONDITIONING ELEMENTS (b) (c) (d)
Sketch the frequency spectrum of the bridge output voltage. Use the data given below to calculate the form of the bridge output voltage when F = +1.0 N and F = −1.0 N. Using the results of (c), explain how to demodulate the bridge output voltage. Overall spring stiffness = 103 N m−1 Effective mass of spring and armature = 25 × 10−3 kg Damping ratio = 0.7 20 Inductance of each coil = mH (d = air gap in mm) 1 + 2d Amplitude of bridge supply = 1 V Frequency of bridge supply = 1000 Hz
9.10
A strain in the range 0 to 10−3 is to be measured using a four-strain-gauge bridge connected to a differential amplifier (Figure 9.11). The strain gauges have resistance 120 Ω and gauge factor 2.1, and the supply voltage is 15 V. Find the value of the amplifier feedback resistance if the output signal range is 0 to 1.0 V.
9.11
Design an a.c. amplifier system which incorporates an ideal operational amplifier, to meet the following specification:
• • • 9.12
Midband input impedance = 10 kΩ Midband gain = 100 Bandwidth = 100 to 1000 Hz.
The electronic torque balance D/P transmitter of Figure 9.20(a) is to have an output range of 4 to 20 mA for input ranges between 0 to 0.5 m of water and 0 to 5.0 m of water. The diaphragm has a diameter of 10 cm and the maximum value of a/b is 2.0. Take g = 9.81 m s−2, and the density of water = 103 kg m−3. (a) (b)
Complete the design by calculating the electromagnetic force constant and the zero spring force. Find the values of a/b corresponding to input ranges of 0 to 1 m and 0 to 5 m.
9.13
The variable reluctance displacement transducer of Problem 8.4 is incorporated into an electrical oscillator circuit with a fixed capacitance of 500 pF. Find the variation in frequency of the oscillator output signal corresponding to a variation in air gap of between 1 and 3 mm.
9.14
The natural frequency fn Hz of a thin-walled tube executing circumferential vibrations in a fluid of density ρ kg m−3 is given by: f n2 =
1.25 × 1010 ρ + 350
The stiffness of the tube is 109 N m−1 and the damping ratio ξ = 0.1. The tube is incorporated in a closed-loop system which also includes a drive coil of sensitivity 104 N V −1 and a displacement sensing coil of sensitivity 103 V m−1. The system is required to give a sinusoidal voltage output signal whose frequency changes with fluid density in the range ρ = 250 to 1500 kg m−3. Find the gain and phase characteristics of the maintaining amplifier. What is the effect of changes in fluid viscosity on the system? 9.15
A solid-state capacitive humidity sensor has a capacitance given by: C = 1.7 RH + 365 pF where RH is the percentage relative humidity. The sensor has an associated parallel resistance of 100 kΩ. The sensor is incorporated into a feedback oscillator system with a pure inductance
PROBLEMS
245
and a maintaining amplifier. The oscillator is to give a sinusoidal output voltage at the natural frequency of the L–C circuit for a relative humidity between 5 and 100%. (a) (b) (c)
Draw a diagram of a suitable oscillator system. If the frequency of the output signal is to be 100 kHz at RH = 100%, calculate the required inductance. Find the gain and phase characteristics of the maintaining amplifier.
9.16
Figure Prob. 16 shows a four-lead bridge circuit; Rc is the resistance of the leads connecting the sensor to the bridge circuit. Show that ETh ≈ VS (R0 /R3)α T, i.e. the bridge output voltage is unaffected by changes in Rc.
9.17
A resonant pressure sensor consists of a vibrating circular silicon plate. The natural frequency fn Hz of the plate is given by:
Figure Prob. 16.
P = 104 − 30fn + 0.12f n2 where P Pa is the input pressure. The stiffness of the plate is 107 N/m and the damping ratio is 0.01. The sensitivity of the capacitance displacement sensor is 2 × 102 V/m and the sensitivity of the drive element is 5 N/V. (a) (b)
If the sensor is to have an output frequency range of 1.0 to 3.0 kHz, what is the input pressure range? What are the gain and phase characteristics of the maintaining amplifier?
Basic problems 9.18
A deflection bridge has a supply voltage of 10 V. Find the bridge output voltage when: (a) (b)
9.19
R1 = 101 Ω, R2 = 99 Ω, R3 = 101 Ω, R4 = 99 Ω R1 = 100.1 Ω, R2 = 99.9 Ω, R3 = 100.1 Ω, R4 = 99.9 Ω.
A platinum resistance sensor has a resistance of 100 Ω at 0 °C and a temperature coefficient of resistance of 4 × 10−3 °C −1. The above sensor is incorporated into a bridge circuit which has R3 /R2 = 100. Find the value of R4 such that VOUT = 0 at 0 °C. (b) Complete the bridge design by calculating the supply voltage required to give VOUT = 100 mV at 100 °C. A simple weight sensor consists of a steel cantilever clamped at one end with the free end subject to a downward force F.
(a)
9.20
(a)
If F = 102 N, use the data given below to calculate the strain halfway along the cantilever.
246 S IGNAL CONDITIONING ELEMENTS (b)
Four identical strain gauges (specified below) are to be bonded onto the cantilever halfway along its length as shown in Figure 8.20(a). Use the data given below to calculate the bridge output voltage corresponding to the above conditions.
Strain e =
6(l − x)F wt 2E
Cantilever: length width thickness Young’s modulus Gauge factor Bridge supply voltage
l = 150 mm w = 50 mm t = 3 mm E = 2 × 1011 N m−2 G = 2.0 VS = 15 V
10 Signal Processing Elements and Software The output signal from the conditioning elements is usually in the form of a d.c. voltage, d.c. current or variable frequency a.c. voltage. In many cases calculations must be performed on the conditioning element output signal in order to establish the value of the variable being measured. Examples are the calculation of temperature from a thermocouple e.m.f. signal, and the calculation of total mass of product gas from flow rate and density signals. These calculations are referred to as signal processing and are usually performed digitally using a computer. The first section of this chapter discusses the principles of analogue-to-digital conversion and the operation of typical analogue-to-digital converters. The following section explains the structure and operation of typical computer systems. The various forms of computer software, both low- and high-level languages, are discussed in the next section. The final section looks at the two main types of signal processing calculation in measurement systems: steady-state calculation of measured value, and dynamic digital compensation and filtering.
10.1
Analogue-to-digital (A /D) conversion This section commences by discussing the three operations involved in A/D conversion: these are sampling, quantisation and encoding. The first operation is performed by a sample-and-hold device; the second and third are combined in an analogue-todigital converter.
10.1.1
Sampling In Chapter 6 we saw that a continuous signal y(t) could be represented by a set of samples yi, i = 1, . . . , N, taken at discrete intervals of time ∆T (sampling interval). The operation is shown in Figure 10.1; the switch is closed fS times per second, where sampling frequency fS = 1/(∆T ). In order for the sampled signal yS (t) to be an adequate representation of y(t), fS should satisfy the conditions of the Nyquist sampling theorem, which can be stated as follows: A continuous signal can be represented by, and reconstituted from, a set of sample values provided that the number of samples per second is at least twice the highest frequency present in the signal.
248 S IGNAL PROCESSING ELEMENTS AND S OF TWARE Figure 10.1 Time waveform and frequency spectrum of sampled signal.
Mathematically we require: Nyquist sampling theorem
fS ≥ 2f MAX
[10.1]
where f MAX is the frequency beyond which the continuous signal power spectral density φ ( f ) becomes negligible (Chapter 6 and Figure 10.1). The above result can be explained by examining the power spectral density φS ( f ) of the sampled signal. Because the sampled signal is a series of sharp pulses, φS ( f ) contains additional frequency components which are centred on multiples of the sampling frequency. Figure 10.1 shows φS ( f ) in the three situations fS > 2f MAX, fS = 2f MAX and fS < 2f MAX. If fS > 2f MAX, then the additional frequency components
10.1 AN ALO GUE - TO - D I GI TAL ( A/D ) CO N VE R S I O N
249
Figure 10.2 Aliasing.
can easily be filtered out with an ideal low-pass filter of bandwidth 0 to f MAX and the original signal reconstituted. If f = 2f MAX, it is just possible to filter out the sampling components and reconstitute the signal. If f < 2f MAX, the sampling components occupy the same frequency range as the original signal and it is impossible to filter them out and reconstitute the signal. The effect of sampling at too low a frequency is shown in Figure 10.2. Here a sine wave of period 1 s, i.e. frequency 1 Hz, is being sampled approximately once every second, i.e. the sampling frequency is below the Nyquist minimum of 2 samples/ second. The diagram shows that it is possible to reconstruct an entirely different sine wave of far lower frequency from the sample values. This is referred to as the ‘alias’ of the original signal; it is impossible to decide whether the sample values are derived from the original signal or its alias. The phenomenon of two different signals being constructed from a given set of sample values is referred to as aliasing. The operation of analogue-to-digital conversion can take up to a few milliseconds; it is necessary therefore to hold the output of the sampler constant at the sampled value while the conversion takes place. This is done using a sample-and-hold device as shown in Figure 10.3. In the sample state the output signal follows the input signal; in the hold state the output signal is held constant at the value of the input signal at the instant of time the hold command is sent. The sample-and-hold waveform shown is ideal; in practice errors can occur due to the finite time for the transition between sample-and-hold states (aperture time) and reduction in the hold signal (droop).
10.1.2
Quantisation Although the above sample values are taken at discrete intervals of time, the values yi can take any value in the signal range yMIN to yMAX (Figure 10.4). In quantisation the sample voltages are rounded either up or down to one of Q quantisation values or levels Vq, where q = 0, 1, 2, . . . , Q − 1. These quantum levels correspond to the Q decimal numbers 0, 1, 2, . . . , Q − 1. If V0 = yMIN and VQ−1 = yMAX, then there are (Q − 1) spacings occupying a span of yMAX − yMIN. The spacing width or quantisation interval ∆V is therefore: ∆V =
yMAX − yMIN Q−1
[10.2]
250 S IGNAL PROCESSING ELEMENTS AND S OF TWARE Figure 10.3 Sample and hold.
Figure 10.4 Quantisation.
The operation of quantisation produces an error eq = Vq − yi termed the quantisation error. Normally if yi is above the halfway point between two levels q, q + 1 it is rounded up to Vq+1; if yi is below halfway it is rounded down to Vq (Figure 10.4). The maximum quantisation error e MAX is therefore ±∆V/2, or expressed as a percentage q of span yMAX − yMIN:
10.1 AN ALO GUE - TO - D I GI TAL ( A/D ) CO N VE R S I O N
Maximum percentage quantisation error
eqMAX = ±
∆V 100 × 100% = ± % 2( yMAX − yMIN ) 2(Q − 1)
251
[10.3]
We see that the relationship between Vq and yi is characterised by a series of discrete steps or jumps; this is an example of resolution errors discussed in Chapter 2. Obviously the greater the number of levels Q, the lower the quantisation errors.
10.1.3
Encoding The encoder converts the quantisation values Vq into a parallel digital signal corresponding to a binary coded version of the decimal numbers 0, 1, 2, . . . , Q − 1. The commonly used decimal or denary number system uses a base or radix of 10, so that any positive integer is expressed as a series of powers of 10 (decades): d n × 10n + dn−1 × 10n−1 + . . . + d i × 10i + . . . + d1 × 101 + d 0 × 100
[10.4]
where the di are the respective weights or digits, which take the values 0 to 9. In digital computers the binary number system is used. This has a base of 2 so that any positive integer can be expressed as a series of powers of 2: bn × 2n + bn−1 × 2n−1 + . . . + bi × 2 i + . . . + b1 × 2 1 + b0 × 20 [10.5] ↑ ↑ MSB LSB (Most significant bit) (Least significant bit) where the weightings bi are referred to as bits or digits. A bit can take only the values 0 or 1 so that calculations on binary numbers are easily performed by logic circuits which distinguish between two states – on or off, true or false. In order to convert the decimal number 183 to binary, we express it in the form: 1 × (128) + 0 × (64) + 1 × (32) + 1 × (16) + 0 × (8) + 1 × (4) + 1 × (2) + 1 × (1) i.e. 1 × 27 + 0 × 26 + 1 × 25 + 1 × 24 + 0 × 23 + 1 × 22 + 1 × 21 + 1 × 20 giving the 8-bit binary number 10110111. To produce an electrical signal corresponding to this number we require eight wires in parallel, the voltage on each wire being typically 5 V for a ‘1’ and 0 V for a ‘0’; this signal is an 8-bit parallel digital signal. The number of binary digits n required to encode Q decimal numbers is given by: Q = 2n i.e. Number of digits in binary code
n = log 2 Q =
log10 Q log10 2
[10.6]
Thus if Q = 200, n = log10 200/log10 2 = 2.301/0.301 = 7.64. Since, however, n must be an integer, we require eight bits, which corresponds to Q = 28 = 256. From eqn [10.3], the corresponding maximum quantisation error is ±100/2(255)% = +0.196%. If the
252 S IGNAL PROCESSING ELEMENTS AND S OF TWARE Table 10.1.
Analogue input (V )
Decimal number
Digital output
0
0
00 000 000
1.2
1.2 × 255 = 61.2 ≈ 61 5.0
00 111 101
3.7
3.7 × 255 = 188.7 ≈ 189 5.0
10 111 101
5
255
11 111 111
input range of the converter is 0 to 5 V, then the corresponding analogue input, decimal numbers and digital output signals are as shown in Table 10.1. Other codes are as follows.
Binary coded decimal (b.c.d.) Here each decade of the decimal number is separately coded into binary. Since 23 = 8 and 24 = 16, four binary digits DCBA are required to encode the 10 numbers 0 to 9 in each decade. In 8:4:2:1 b.c.d. A ≡ 20 = 1, B ≡ 21 = 2, C ≡ 22 = 4, D ≡ 23 = 8; thus the decimal number 369 becomes DCBA 0011
DCBA 0110
DCBA 1001
3
6
9
The number of decades p of b.c.d. required to encode Q decimal numbers is given by Q = 10 p, i.e. p = log10Q, and the corresponding total number of binary digits is: Number of digits in b.c.d.
4p = 4 log10Q
[10.7]
The input signal to character displays (Chapter 11) is normally in b.c.d. form; since the signal is already separated into decades the conversion into seven segment or 7 × 5 dot matrix code is easier than with pure binary.
Octal code Here a base of 8 is used so that the weights take the values 0 to 7. In order to convert 183 decimal, i.e. (183)10, into octal we express it in the form: 2(64) + 6(8) + 7(1) i.e. 2 × 82 + 6 × 81 + 7 × 80 i.e. 267 octal or (267)8
10.1 AN ALO GUE - TO - D I GI TAL ( A/D ) CO N VE R S I O N
253
A binary number is easily coded into octal by arranging the digits in groups of 3, where for each group A ≡ 20 = 1, B ≡ 21 = 2, C ≡ 22 = 4. Thus 010110111 (binary code for (183)10) becomes: CBA 010
CBA 110
CBA 111
2
6
7
i.e.
267 octal
Hexadecimal code (hex) Here a base of 16 is used and the digits are the ten numbers 0 to 9 and the six letters ABCDEF. Since 24 = 16, each hexadecimal digit corresponds to four binary digits: Binary Hex Binary Hex
0000 0 1000 8
0001 1 1001 9
0010 2 1010 A
0011 3 1011 B
0100 4 1100 C
0101 5 1101 D
0110 6 1110 E
0111 7 1111 F
Some corresponding decimal, binary and hexadecimal numbers are Decimal 94 167 238
Binary 01011110 10100111 11101110
Hexadecimal 5E A7 EE
Floating point representation Here a number is expressed in the form aN b where a is the mantissa, b the exponent and N the base of the number system. Thus for a denary or decimal system with N = 10, we have: −43 700 = −.437 × 10+5 +0.000 259 = +.259 × 10−3 Both the mantissa and the exponent, each with a corresponding sign bit, can then be separately represented inside a computer using either binary, b.c.d. or hexadecimal code. Thus if three-decade b.c.d. is used for the mantissa and four-digit binary for the exponent, with 0 corresponding to − and 1 to +, the above numbers can be repesented by:
This representation is used when the range of numbers in a given computation is very wide. For example, the above format covers the range from 0.001 × 10−15 to 0.999 × 10+15. Arithmetic operations are performed on exponent and mantissa separately.
254 S IGNAL PROCESSING ELEMENTS AND S OF TWARE According to the IEEE standard 754-1985, the standard floating-point representation is: 23-bit mantissa, 8-bit exponent, 1-bit sign (single precision) 52-bit mantissa, 11-bit exponent, 1-bit sign (double precision)
10.1.4
Frequency to digital conversion In several cases the output signal from primary sensing or signal conditioning elements is an a.c. voltage with a frequency which depends on the measured variable. Examples are:
• • •
The variable reluctance tachogenerator (Section 8.4) Electrical and electromechanical oscillators for displacement, density and differential pressure measurement (Section 9.4) Turbine, vortex and Doppler flow meters (Sections 12.2 and 16.4).
There are two main methods of converting a variable frequency sinusoidal signal into a parallel digital output signal. The sine wave must first be converted into a square wave signal with sharp edges using a Schmitt trigger circuit. In the first method the frequency fS of the signal is measured by counting the number of pulses during a fixed time interval T. The principle is shown in Figure 10.5(a). The number NS of positive-going edges during T is counted, giving: fS =
Figure 10.5 Frequency to digital conversion: (a) Frequency measurement – principle (b) Frequency measurement – system
NS , T
i.e.
N S = fS T
[10.8]
10.1 AN ALO GUE - TO - D I GI TAL ( A/D ) CO N VE R S I O N
255
Figure 10.5 (cont’d) (c) Period measurement – principle (d) Period measurement – system.
Figure 10.5(b) shows one possible system for implementing this method. Both counters are initially reset to zero. The number NC is then loaded into the clock counter; this sets the counting interval T to be: T=
NC fC
[10.9]
where fC is the clock frequency. Clock pulses are input to the clock counter, which then counts down to zero. Signal pulses are input to the signal counter; this counts up until the clock counter reaches zero, when the count is stopped. The signal count is then: NS =
NC ⋅ fS fC
[10.10]
The parallel digital output signal corresponds to signal count NS , which is proportional to input signal frequency fS. The resolution of this method is limited to ±1 signal count; this can mean poor percentage resolution at low frequencies. If we consider the example when fS = 20 Hz
256 S IGNAL PROCESSING ELEMENTS AND S OF TWARE and fC = 10 kHz, then to get T = 1 s, NC = 104, giving NS = 20 ± 1, i.e. a percentage resolution of ±5%. Percentage resolution can be increased by increasing the counting interval, but this is only possible if T is small compared with the time scale of the dynamic variations in the measurement signal. This problem can be solved using the second method. Here the period TS of the signal is measured by counting the number of clock pulses within TS. The principle is shown in Figure 10.5(c). The number NC of positive-going edges is counted, giving: TS =
NC , fC
i.e.
NC = fC TS
[10.11]
where 1/fC is the clock period. Figure 10.5(d) shows one possible system for implementing this method. Both counters are initially set to zero; clock pulses are input to the clock counter and signal pulses to the signal counter. The clock counter counts up until a count of 1 is registered in the signal counter; the count then stops. The clock count NC is then given by eqn [10.11] and is proportional to TS. The parallel digital output signal is proportional to this clock count. The resolution of this method is ±1 clock count. Using the above example with fS = 20 Hz, TS = 0.05 s and fC = 10 kHz, NC = 500 ± 1, i.e. a percentage resolution of ±0.002%. However, some frequency signals are subject to random fluctuations which cause the signal period to vary, even though the input true value of the measured variable is constant. For example, the period of a nominal 20 Hz signal from a vortex flow meter may vary randomly between 0.04 s and 0.06 s about a mean value of 0.05 s. It is therefore essential to measure the mean frequency or mean period of the signal. The first method gives the mean frequency of 20 cycles measured over 1 second; the second method gives the period of one cycle only. This problem is solved using the second method but now the clock counter counts up until a count of 20 is registered in the signal counter. The mean clock count NC for one signal cycle is then the total clock count divided by 20; the mean signal period TS is then given by eqn [10.11].
10.1.5
Digital-to-analogue converters (DACs) A DAC gives an analogue output voltage which is proportional to an input parallel digital signal, e.g. an 8-bit binary signal b7b6 . . . b1b0. In Figure 10.6 an operational amplifier is used to sum a number of currents which are either zero or non-zero depending on whether the corresponding bit is 0 or 1. The current corresponding to the most significant bit is twice that corresponding to the next significant bit and so on. This is achieved in Figure 10.6(a) by using a network of binary-weighted resistors 20R, 21R, 22R, . . . , 27R. The problem with this arrangement is that a very large range of resistance values is required. A better alternative is the ladder network shown in Figure 10.6(b); the advantage of this circuit is that the required current distribution can be obtained with only two values of resistance, R and 2R.
10.1.6
Analogue-to-digital converters (ADCs) One commonly used type of analogue to digital converter is the dual-slope converter. Here the input is a continuous voltage VIN rather than sample values; dual slope
10.1 AN ALO GUE - TO - D I GI TAL ( A/D ) CO N VE R S I O N
Figure 10.6 Summing amplifier DACs: (a) Binary weighted resistor network (b) R–2R ladder network.
257
converters are used in digital voltmeters and indicators. Figures 10.7(a) and (b) show a schematic diagram of the system and the principle of operation. Firstly the control logic switches VIN onto the integrator input for a fixed time interval. During this time interval the integrator output VI is a positive ramp with a slope proportional to VIN (Figure 10.7(b)) so that the output VC of the comparator is 0. At the end of the fixed time interval VI is proportional to VIN (voltages V1, V2, V3). At this point the control logic switches the reference voltage −VREF onto the integrator and resets the counter to zero. The integrator output is then a negative ramp with a fixed rate of decrease. This means that the time taken for VI to decrease to zero is proportional to VIN. During this time the control logic routes clock pulses to the counter and the counter increments. When VI falls to zero the comparator output VC changes to 1 and the count is stopped. The count is proportional to the fall time and therefore to VIN: the parallel digital output signal is proportional to the count. This method has the advantage that integration tends to average out the effects of mains interference voltages. Figure 10.7(c) shows a schematic diagram of a successive approximation analogueto-digital converter. This method involves making successive guesses at the binary code corresponding to the input voltage yi. The trial code is converted into an analogue voltage using a DAC, and a comparator is used to decide whether the guess
258 S IGNAL PROCESSING ELEMENTS AND S OF TWARE Figure 10.7 Analogueto-digital converters: (a) Dual slope – system (b) Dual slope – principle (c) Successive approximation.
10.1 AN ALO GUE - TO - D I GI TAL ( A/D ) CO N VE R S I O N
Table 10.2 Typical sequence of guesses in successive approximation ADC.
259
Input voltage yi = 0.515 V Clock pulse
Initiate conversion →
Data valid →
1 Clear register 2 First guess
DAC input
00000000 01111111 (127)10 3 Next guess 00111111 (63)10 4 00011111 (31)10 5 00101111 (47)10 6 00110111 (55)10 7 00110011 (51)10 8 00110101 (53)10 9 Final guess 00110100 (52)10 Output digital signal = 00110100
DAC Vq output (volts)
Comparator output VC
Result
0 1.27
0 1 HIGH
b7 = 0
0.63
1 HIGH
b6 = 0
0.31
0 LOW
b5 = 1
0.47
0 LOW
b4 = 1
0.55
1 HIGH
b3 = 0
0.51
0 LOW
b2 = 1
0.53
1 HIGH
b1 = 0
0.52
1 HIGH
b0 = 0
is too high or too low. On the basis of this result another guess is made, and the process is repeated until Vq is within half a quantisation interval of yi. Table 10.2 shows a series of guesses for an 8-bit binary converter with an input range of 0 to 2.55 V. The first guess is always 01111111 corresponding to (127)10, i.e. approximately half full scale: this guess is high so that b7 is set to 0; if the guess had been low b7 would be set to 1. The next guess is 00111111 corresponding to (63)10, i.e. approximately one-quarter full scale; this guess is also high so that b6 is confirmed as 0. The process continues until all the remaining bits have been confirmed; the DATA VALID signal then changes state. Successive approximation converters (SAC) can be used for sample rates up to over 106 samples/s; even 16-bit types can be used up to over 105 samples/s. For the fastest applications up to 109 samples/s, such as video digitisation, flash converters are used. SAC converters can be linked to microcontrollers using two-way serial communication over a pair of wires. Here the successive approximation logic is provided by the microcontroller; the SAC consists only of a DAC and a comparator. The microcontroller sends out clock pulses to operate the DAC switches and receives the digital code in serial form. Figure 10.8 shows a general schematic diagram of a flash or parallel analogueto-digital converter.[1] In any n-digit binary ADC there are Q quantisation voltage levels V0 to VQ−1, where Q = 2n. In a flash ADC there are Q − 1 comparators in parallel and Q − 1 corresponding voltage levels V1 to VQ−1. There is no need to provide the V0 voltage level. In each comparator q, the input sample value yi is compared with the corresponding voltage level Vq . If yi is less than or equal to Vq , the output is zero corresponding to 0. If yi is greater than Vq , the output is non-zero corresponding to a 1, i.e.
260 S IGNAL PROCESSING ELEMENTS AND S OF TWARE Figure 10.8 Flash analogue-to-digital converter.
V qc = 0,
yi ≤ Vq
V = 1,
yi > Vq
c q
[10.12]
Thus if yi lies between Vq and Vq+1, i.e. Vq < yi ≤ Vq+1, the output of the lowest q comparators 1 to q will all be 1 and the output of the remaining comparators q + 1 to Q − 1 will all be 0. Thus the comparators provide a Q − 1 digit parallel input code to a priority encoder which generates an n-digit binary parallel output code corresponding to the value of q. The main advantage of the flash converter is the short conversion time; the main disadvantage is that the large number of comparators required to give acceptable resolution mean that it is relatively expensive.
10.2 10.2.1
Computer and microcontroller systems The general computer system Figure 10.9 is a block diagram of a general computer system; the system is made up of the following elements.
10.2 COMPUTER AND MICROCONTROLLER SYSTEMS
Figure 10.9 General computer system.
261
The microprocessor consists of three main parts. The control unit controls the processing of instructions by providing the necessary control and timing signals. The arithmetic logic unit carries out arithmetic operations such as addition and subtraction and logic operations such as NOT, AND, OR. The register array is a collection of registers required to temporarily store instructions, addresses and data during the execution of an instruction. The read only memory (ROM) consists of a large number of permanent storage elements, each specified by a unique address. Information can only be read from ROM; no information can be written into ROM. Reading is an operation whereby a copy of a data word or instruction is transferred from a given storage location to another device without changing the contents of the store. Writing is an operation whereby a data word or instruction is placed in a given storage location. Instructions are normally stored in ROM. The random access memory (RAM) consists of a large number of nonpermanent storage elements, each specified by a unique address. Information can both be read from and written into RAM. Both instructions and data can be stored in RAM. The input/output interface consists of a number of ports to which external devices can be physically connected. The ports can be configured to accept input data in parallel and/or serial format and output data in parallel and/or serial format. For example, data from an ADC can be received at the parallel input port and data transmitted to a display or printer at the parallel output port. The address bus has a one-way flow of information; the microprocessor sends out address codes to memory locations and input/output ports. The data bus has a two-way flow of information. Here, for example, data is transferred from the input port or RAM to the microprocessor, and instructions are
262 S IGNAL PROCESSING ELEMENTS AND S OF TWARE
Figure 10.10 Layout of typical microprocessor.[1]
transferred from the ROM to the microprocessor. The results of calculations or operations are transferred from the microprocessor to RAM or the output port. The control bus also has a two-way flow of information. The microprocessor sends out enable and clock signals to all elements in the system to achieve the coordination and synchronisation necessary for the above information transfers to take place. It receives information back on the status of each element. A small-scale computer system based on an 8-bit microprocessor will correspondingly have an 8-bit data bus. A 16-bit address bus will enable 216 = 65 536 different locations to be addressed. A typical system would have two 4 K ROMs, each capable of storing 4096 8-bit bytes, and two RAMs, each capable of storing 128 8-bit bytes. The computer operates sequentially by executing a series of steps; the timing of these steps is determined by the clock. Each step is performed as a result of an instruction; the sequence of instructions is called a program. The first step in the execution of an instruction is for the CPU to address the memory location in which the instruction is stored. The instruction is then fetched and decoded; any data required by the instruction is also fetched. The instruction is then executed and the results of the instruction are transferred to memory or an output port. After completing the instruction the CPU moves on to the next one. Figure 10.10 shows the layout of a typical microprocessor. The processor is divided into two separate units. The executive unit (EU) executes the instructions;
10.2 COMPUTER AND MICROCONTROLLER SYSTEMS
263
the bus interface unit (BIU) interfaces with the computer buses, fetches instructions, reads operands and writes results.[1] The EU control system fetches an instruction from the instruction queue, decodes it and sends out control signals to execute the instruction. The arithmetic logic unit (ALU) performs all arithmetic and logic operations; the data registers are used to store data before and after these operations. The flag register contains status flags associated with arithmetic operations such as zero, sign, carry. The pointer and index registers are used to create an address in the BIU; the result of the instruction is then transferred to this temporary address. While a given instruction is being executed in the EU, the bus control logic fetches the next instructions from memory via the computer bus and places them in the instruction queue. The address segment registers and instruction pointer are used to create parts of the instruction address. These are then combined in the address summing block to create the full address of the next instruction to be fetched.
10.2.2
Microcontrollers In a microcontroller the processor, memory and input/output are all combined on a single chip. Microcontrollers are widely used for signal processing in measurement systems. Figure 10.11 shows the layout of a typical microcontroller.[2] This is an 8-bit system with CPU, memory, four parallel ports, two serial ports, 12 MHz clock and a single bus. There are two 4 K EPROMs (Electrically Programmable Read Only Memory). This memory is non-volatile and programmed electrically; once programmed it cannot be changed, but memory contents are unchanged when the electrical power is switched off. Program instructions are stored in EPROM. There are two 128-byte RAMs, which are used to store data; this is volatile memory whose contents are lost when the power is switched off. An alternative to RAM is EEPROM (Electrically Erasable PROM); this will hold data without a continuous power supply, and program instructions can be changed or updated in the field.
Figure 10.11 Layout of a typical microcontroller.
264 S IGNAL PROCESSING ELEMENTS AND S OF TWARE
10.3 10.3.1
Microcontroller and computer software Assembly language for microcontrollers Microcontrollers are programmed in assembly language. Here each instruction is of the form: operator operand The operator specifies the instruction to be carried out and is defined by a threeto six-letter mnemonic called an opcode. A typical microcontroller has a set of about 35 instructions, which are stored in EEPROM. Table 10.3 lists some typical instructions. The working register W is at the heart of the microcontroller CPU. It is used to perform arithmetic and logic operations. This means that data must first be transferred from a file register to W; the operation is then performed in W and the result transferred from W to another register. Consequently several instructions involve the W register. The operand is either data or an address location. In an 8-bit microcontroller the basic data or address word is eight bits or one byte long; this is referred to as a file. A single file can therefore cover the range of numbers 0 to 255 decimal, 0000 0000 to 1111 1111 binary or 00 to FF hexadecimal. Typically there are 16 special-purpose registers, specified by the hex addresses 00 to 0F. These include, for example, timer/counter, program counter, status, file select, input and output port registers. In addition there are typically 70 general-purpose registers, specified by the hex addresses 10 to 55. These are within the RAM and are used to store user data files; they are referred to as file registers or simply files. To accommodate data numbers greater than 255, two 8-bit data files are combined; this allows numbers up to 216, i.e. 65 536. These two files are then stored in two adjacent file registers.
Table 10.3 Typical microcontroller instructions.
Opcode
Description
ADDLW ADDWF CALL CLRF CLRW DECF GOTO INCF MOVF MOVLW MOVWF RETURN SUBLW SUBWF
Adds a number to contents of W Adds contents of W to contents of file Calls a subroutine Clears file F Clears the W register Subtracts 1 from file Unconditional jump to specified location in program Adds 1 to file Moves contents of file into W Moves data file directly into W Moves contents of W to file Returns from a subroutine to main program Subtracts contents of W from a number Subtracts contents of W from contents of file
10.3 MI CR O CO N TRO LLE R AN D CO MPUTE R S O F T WAR E
Table 10.4 Typical microcontroller program.
265
Opcode
Operand
Description
CLRW MOVF ADDLW SUBWF MOVWF
2C $1A 2D result
Clears W register Moves contents of file 2C to W Adds the number 1A (hex) to W Subtracts contents of W from contents of file 2D Moves contents of W to reserved register with name ‘result’
Table 10.4 shows a typical assembler program for a microcontroller. The contents of file register 2C are transferred to the W register, the hex number 1A is added to W, and the contents of W are then subtracted from the contents of file register 2D. The final result is then transferred from W to a reserved register with the name ‘result’.
10.3.2
Higher-level languages Assembly language programs generally take less time to execute and require less memory than those written in other languages; they also support any type of input, output and peripheral device. However, assembly programming is far more difficult; we can see from the example in Table 10.4 that several lines of code are necessary to perform a simple arithmetic operation. Several types of error are possible in assembly programming which are much less likely in a high-level language. If we assume that it takes about the same time, per line of code, to write, debug, test, execute and comment irrespective of the language, then assembly language implementation will be far more expensive. Another disadvantage of assembly language is that it is specific to one type of processor and is therefore not portable to other microprocessors or microcontrollers. Assembly code written for one type of processor cannot be moved to another processor with a different instruction set. High-level languages have the following characteristics: 1. 2.
3.
They include procedures: a procedure is a sequence of operations which defines exactly how a task is to be performed. The control of program flow is determined by the way the program is structured, i.e. by the sequential ordering of operations or by explicit linkages between them. They include assignment statements; an assignment statement assigns a new value to a variable, usually by computing the numerical value of an algebraic expression specified in the statement. Examples are: A=B+C Y = 5.0 * sin T Y = exp(−7.2 * X)
4.
Several data types are possible; examples are integer, real and Boolean. Highlevel languages are processor-independent and use a compiler to translate the complete high-level language program into the assembly code for a given processor. They therefore require more memory (the compiler is itself a large program) and take more time to execute. However, high-level programs are easier to write and correct.
266 S IGNAL PROCESSING ELEMENTS AND S OF TWARE Table 10.5 Typical BASIC statements and operators.
Statements
Operators
BAUD CALL CLEAR CLEAR(S&I) CLOCK(1&0) DATA READ RESTORE DIM DO-WHILE DO-UNTIL END FOR-TO-STEP NEXT GOSUB RETURN GOTO ON-GOTO ON-GOSUB IF-THEN-ELSE INPUT LET ONERR ONEX1 ONTIME PRINT PRINT#
ADD (+) DIVIDE (/) EXPONENTIATION (**) MULTIPLY (*) SUBTRACT(—) LOGICAL AND (.AND.) LOGICAL OR (.OR.) LOGICAL X-OR (. XOR.) LOGICAL NOT (.OR.) ABS() INT() SGN() SQR() RND LOG() EXP() SIN() COS() TAN() ATN() =, >, >=, <, <=, <> ASC() CHR() CBY() DBY() XBY() GET
BASIC is a very popular, easy-to-learn language. It began as an interpreted rather than a compiled language. This means that it is translated and executed statementby-statement, using a program called an interpreter, instead of being fully translated into machine code before execution. There are several ‘dialects’ of BASIC which are specific to a given processor; this makes it less portable than other high-level languages. Table 10.5 shows some typical BASIC statements and operators. These include input/output statements, program sequence control statements and arithmetic operators. Control of flow is obtained by attaching a numerical label (line number) to each statement and then executing the resulting program in an ascending sequence of these statement labels. An assignment statement would therefore read: 10 LET I = J + 50 Subroutines are called with GOSUB N statements, where N is the line number of the first statement of the subroutine. Return to the main program is then via a RETURN statement at the end of the subroutine. Program sequence is controlled by IF, THEN, ELSE, GOTO N statements. This reliance on line numbers can lead to a lack of program structure, making programs difficult to follow, and can lead to errors if the wrong line number is given. Because BASIC is an interpretive language, it is decoded and error-checked line by line during operation and so is far too slow and inefficient for the processing of
10.3 MI CR O CO N TRO LLE R AN D CO MPUTE R S O F T WAR E
267
measurement signals. C is the chosen language for fast, efficient signal processing in measurement systems by embedded microcontrollers and computers. C is a relatively low-level language operating fairly close to assembly instructions. It has a highly modular simple structure which uses functions as sub-programs; a function is a self-contained unit of code. Before any variable is used as a function, it has to be declared, i.e. given a name and type. The general term for a name is an identifier. C permits the following data types: float floating point 3.4 × 1038 to 1.2 × 10−38 int integer −32 768 to +32 767 char up to 256 characters (e.g. upper/lower case letters) The assignment operator = assigns to the variable on the left-hand side a value given by the expression on the right-hand side. Thus to convert Celsius temperature into Fahrenheit we have: temp_f = temp_c * 9.0/5.0 + 32 The basic arithmetic operators + (add), − (subtract), * (multiply) and / (divide) are available; higher-level arithmetic is provided by standard library functions. Some of these are shown in Figure 10.12(a). Relational operators pose a question to which the answer is either yes or no. These are: a>b a
[15.41]
In the steady state, dTD /dt = 0 so that [15.40] reduces to: TD =
1 PD + TS UA
[15.42]
Commonly used thermal detectors are thermopiles (Figure 15.14(a)), which consist of a large number of thermocouples in series, and bolometers (Figure 15.14(b)), which are metal or semiconductor (thermistor) resistance material in the form of thin films or flakes. In both cases the surface of the detector is blackened to maximise the absorption of incoming radiation. For a thermopile consisting of n thermocouples in series,
15.5 D E TE CT O R S AN D S I GN AL CO N D I TI O N I N G E LE ME N TS
Figure 15.14 Thermal detectors: (a) Thermopile (b) Bolometer (c) Pyroelectric.
405
with hot junction at TD °C and reference junction at TS °C (the temperature of the surroundings), the e.m.f. is given approximately by (Section 8.5) ETD ,TS ≈ na1(TD − TS )
[15.43]
From [15.42] the equilibrium temperature difference is TD − TS = (1/UA)PD, giving: ETD,TS =
A na1 D P C UA F D
[15.44]
na1 mV/W. UA The bridge circuit shown in Figure 15.14(b) gives an output voltage approximately proportional to power PD. Here R1 is a radiation-detecting bolometer at TD °C, R2 a reference bolometer at TS °C and R3, R4 fixed equal resistors. This circuit has already i.e. the detector sensitivity KD =
406 OPTICAL MEASUREMENT SYSTEMS been discussed in Section 9.1 (Figure 9.4(a)). For metal resistive elements of temperature coefficient α K−1 we have (from eqn [9.17]): VOUT ≈ Vs
R0 α (TD − TS) R3
[15.45]
provided R3 0003 R0 (resistance of bolometer at 0 °C). This gives: VOUT ≈
Vs α R0 ⋅ PD UAR3
[15.46]
indicating the system output voltage is approximately proportional to PD and independent of TS. The time constant τ of a thermal detector is minimised by using thin flakes or films which have a large area-to-volume ratio that is a small value of M/A. However, τ cannot be reduced much below a few milliseconds because of the low heat transfer coefficient U between the detector and the surrounding air. Certain materials exhibit the pyroelectric effect. These are man-made ferroelectric ceramics which also show piezoelectric properties (Section 8.7). A typical material is lithium tantalate which has a Curie temperature of 610 °C. The principle of ferroelectricity is shown in Figure 15.14 (c). The ceramic is composed of a mass of minute crystallites; provided the ceramic is below the Curie temperature, each crystallite behaves as a small electric dipole. Normally the material is unpoled, i.e. the electric dipoles are randomly orientated with respect to each other. The material can be poled, i.e. the dipoles lined up, by applying an electric field when the ceramic is just below the Curie temperature. After the material has cooled and the applied field has been removed, the dipoles remain lined up, leaving the ceramic with a residual polarisation P. The pyroelectric effect arises because the incident radiant power causes the ceramic temperature T to increase; P decreases with T according to the non-linear relation shown in Figure 15.14(c). This reduction in P causes a decrease in the alignment of the dipoles and a corresponding reduction in the positive and negative charges on opposite surfaces of the crystal. The sensor is formed by depositing metal electrodes on these surfaces. The lower electrode acquires an excess positive charge which balances the negative charge on the lower surface; the upper electrode acquires an excess negative charge which balances the positive charge on the upper surface. If ∆q is the excess charge caused by a temperature rise ∆T then: ∆q =
A dP D A∆T C dT F
[15.47]
where A is the area of the electrodes and dP/dT the slope of the P–T characteristics. The electrodes and the rectangular block of dielectric ceramic form a parallel plate capacitor CN . The ceramic can be regarded as either a charge generator ∆q in parallel with CN or a Norton current source iN in parallel with CN (Section 5.1.3) where: iN =
dq dP dT =A ⋅ dt dT dt
[15.48]
From eqn [15.42], we see that for constant power input PD, the temperature rise ∆T is constant with time. If ∆T is constant with time then dT/dt = 0 and from eqn [15.48] iN = 0; this means that the voltage across capacitance CN is zero and the
15.5 D E TE CT O R S AN D S I GN AL CO N D I TI O N I N G E LE ME N TS
407
basic sensor cannot be used for steady-state measurements of radiant power. This problem is solved by connecting the pyroelectric sensor to a charge amplifier (Section 8.7). This is an integrator with an output voltage proportional to ∫ iN dt, i.e. charge ∆q, temperature rise ∆T and incident radiant power PD. A typical pyroelectric detector of size 2.7 × 0.9 mm has a responsivity or sensitivity of 3200 V/W, noise equivalent power of 1 × 10−9 W and D* of 1.4 × 108 cm Hz1/2 W−1. When used with a standard coated silicon window, the wavelength response D(λ) is reduced to 7 to 14 µm.
15.5.2
Photon detectors Photon detectors are mostly semiconductor devices in which incident photons of radiation cause electrons to be excited from the valence band to the conduction band, thereby causing a measurable electrical effect. These detectors only respond to photons whose energy hc/λ is approximately equal to the energy gap EG between valence and conduction bands. This means that photon detectors have a narrow wavelength response D(λ) with peak wavelength λ 0 ≈ hc/EG. Photon detectors normally have higher responsivities than thermal detectors. Because they rely on atomic processes which are inherently faster than bulk heat transfer processes, photon detectors usually have shorter time constants than thermal detectors. There are two main types of photon detector: photoconductive and photovoltaic.
Photoconductive detectors Here the presence of excited electrons in the conduction band causes an increase in electrical conductivity and a decrease in electrical resistance. The resistance R of a photoconductive detector therefore decreases as the total power PD increases; the relationship is extremely non-linear and is best expressed in the logarithmic form (Figure 15.15(a)): log10 R = a − b log10 PD
[15.49]
Figure 15.15(b) shows the wavelength response D(λ) for three photoconductive detectors in common use. Cadmium sulphide has λ 0 ≈ 0.6 µm and is thus suitable for visible radiation. Lead sulphide with λ 0 ≈ 3.0 µm and indium antimonide with λ 0 ≈ 5.3 µm are suitable for infrared radiation. It is important that the detector wavelength characteristics D(λ), the source characteristics S(λ) and the transmission medium characteristics T(λ) all match each other as closely as possible. Thus a cadmium sulphide detector is ideal for measuring the temperature of a target at 4800 K (λ P ≈ 0.6 µm) while indium antimonide is ideal for a target at 550 K (λ P ≈ 5.3 µm). A typical indium antimonide detector of dimensions 6.0 mm × 0.5 mm has K = 7 × 103 V W −1, τ = 5 µs and NEP = 1.6 × 10−11 W. Photoconductive detectors must be incorporated into deflection bridge circuits to convert their output into a voltage signal.
Photovoltaic detectors[2,3,5] These are photodiodes made by forming a junction of p-type and n-type extrinsic semiconductors. Figure 15.15(c) shows typical current/voltage characteristics for a photodiode under dark and illuminated conditions. The effect of the incident radiation
408 OPTICAL MEASUREMENT SYSTEMS Figure 15.15 Photon detectors: (a) Resistance–power relation for cadmium sulphide (b) Wavelength response of photoconductive dectectors (c) Current–voltage characteristics for photodiode (d) Norton equivalent circuit for photodiode (e) Construction of pin photodiode (f ) Wavelength response of photovoltaic detectors.
is to increase the reverse saturation current by an amount iP, which is termed the photocurrent. The photocurrent is proportional to total power PD used by the detector: iP = KD PD
[15.50]
where KD A W −1 is the sensitivity or responsivity. This type of detector can therefore by represented as a Norton current source in parallel with the diode resistance RD and capacitance CD (Figure 15.15(d)). Lead–tin telluride is an example of an infrared photodiode. This operates at 77 K with a typical KD of 5 A W−1 and wavelength response D(λ) between 7 and 14 µm with peak wavelength λ 0 ≈ 10 µm (Figure 15.15(f )). Figure 15.15(e) shows the construction of a pin diode; this has an extra layer of instrinsic (undoped) i material between p and n layers. The incident photons create additional electron-hole pairs in this region with a corresponding increase in photocurrent and responsivity over a comparable pn diode. A typical silicon pin photodiode has an active area of 1 mm2, a responsivity KD ≈ 0.55 A W−1, a peak wavelength λ 0 ≈ 0.85 µm, and D(λ) as shown in Figure 15.15(f ). The wavelength response of this photodiode is therefore well matched to S(λ) for GaAlAs LED and ILD sources; also their small size makes them ideal for coupling to optical fibres. Germanium pin
15.6 MEASUREMENT SYSTEMS
409
diodes with KD ≈ 0.5 A W−1 and λ 0 ≈ 1.5 µm have a D(λ) more suited to longer wavelength applications. The principle of operation of avalanche photodiode (APD) devices differs from that of pin diodes in one major respect. A pin diode converts one photon to one electron, whereas in an APD multiplication takes place, which results in many electrons at the output for each incident photon.
15.6
Measurement systems Comprehensive reviews of optical measurement systems, especially involving optical fibre sensors, are given in references [6] and [7].
15.6.1
Modulation of intensity by source Since the power emitted by a hot body source depends critically on source temperature (eqns [15.6] and [15.10]), systems of the type shown in Figure 15.2 are commonly used for remote temperature measurement. There are two basic types of system: broadband and narrowband.
Broadband These systems use all wavelengths present in the incoming radiation and are often referred to as total radiation pyrometers. A thermal detector is therefore used, and from eqns [15.38], [15.10] and [15.6] the total power used by the detector is: CK PD = 1 F π
0002
∞
0
ε (λ , T ) T (λ ) F (λ ) dλ   C  λ5 exp  2  − 1  λT   
[15.51]
The integral can be evaluated numerically if ε (λ, T ), T(λ) and F(λ) are known. However, if we assume that ε (λ, T ), T(λ) and F(λ) have constant values ε, TM and F respectively for all λ, T, then using eqn [15.7] 1 PD = εTM K F F π
0002
∞
0
C1dλ   C  λ5 exp  2  − 1  λT   
=
1 σε TM KF FT 4 π
The corresponding detector output signal is: Detector output signal for broadband temperature system
VT = K D PD =
KD σε TM KF FT 4 π
[15.52]
The above assumptions represent a considerable oversimplification, and even if they are justified there is still the problem of uncertainties in the values of ε, TM, etc. For these reasons the system should be calibrated experimentally using a standard source
410 OPTICAL MEASUREMENT SYSTEMS of known emissivity. If the system is then used with a process source of different emissivity then a correction must be made.
Narrowband These systems use photon detectors so that only a narrow band of wavelengths is used. In some systems a narrow pass-band radiation filter G(λ) is used to further restrict the range of wavelengths incident on the detector. For these systems, the wavelength response G(λ)D(λ) of the combined filter and detector can be approximated by the ideal band-pass filter characteristics of Figure 15.16(a). Thus the expression for total power used is: 1 PD = C1K F π
0002
∞
0
ε (λ , T ) T (λ ) F (λ ) G(λ ) D(λ ) dλ   C  λ5 exp  2  − 1  λT   
[15.53]
which reduces to: PD =
Figure 15.16 Narrowband radiation temperature system (a) Wavelength characteristics (b) Evaluation of integral (c) Black body – optical fibre system.
1 C1K F π
0002
λ 0+ ∆ λ / 2
λ 0 −∆ λ / 2
ε (λ , T ) T (λ ) F (λ ) dλ   C  λ5 exp  2  − 1  λT   
[15.54]
15.6 MEASUREMENT SYSTEMS
411
The assumptions ε (λ, T ) = ε, T(λ) = TM and F(λ) = F are far more justified for a narrow band of wavelengths and give: PD =
1 π
C1ε TM K F F
0002
λ 0+ ∆ λ / 2
dλ   C  λ exp  2  − 1  λT   
λ 0 −∆ λ / 2
[15.55]
5
The integral is evaluated approximately in Figure 15.16(b) to give: PD =
≈
1 π
C1ε TM K F F
∆λ   C  λ 05 exp  2  − 1  λ 0T   
A C2 D 1 ∆λ C1εTM KF F ⋅ 5 exp − C λ 0T F π λ0
[15.56]
The approximation is valid since exp(C2 /λ 0T ) 0003 1; for example if λ 0 = 1 µm, T = 103 K, (C2 /λ 0T ) = 14.4, exp(C2 /λ 0T ) = 1.8 × 106. The detector output signal is thus: Output signal for narrowband radiation pyrometer
V (T ) = K D PD =
1
π
C1ε TM K F FK D
 C  exp  − 2  λ  λ 0T 
∆λ 5 0
[15.57]
The narrow band of wavelengths must be chosen to coincide with a ‘window’ in the transmission characteristics so that TM is reasonably large. Again the system should be calibrated experimentally using a standard source of known emissivity and a correction applied if it is subsequently used with a process source of different emissivity. The above problems of variations of source emissivity and atmospheric transmission characteristics may be solved using the system shown in Figure 15.16(c). Here a black body cavity is inserted in the process and radiation is transmitted to the radiation receiver using an optical fibre system rather than the atmosphere. The fibre system consists of a short run of special fibre, which can withstand the high process temperatures, connected to a much longer length of normal fibre.
15.6.2
Modulation of intensity by transmission medium These are systems of the type shown in Figure 15.1, where the measured variable alters the characteristics of the transmission medium and so modulates the intensity of the radiation reaching the detector. There are several ways in which this modulation can be achieved.
Intensity modulation using shutters and gratings The simplest type of shutter displacement sensor is shown in Figure 15.17(a). Figures 15.17(b), (c) and (d) show moiré fringe gratings for the digital measurement of linear and angular position. In the linear type there are two finely ruled gratings
412 OPTICAL MEASUREMENT SYSTEMS Figure 15.17 Intensity modulation using shutters and gratings: (a) Shutter modulator (b) Grating modulator (c) Linear displacement moiré fringe gratings (d) Angular displacement moiré fringe gratings (e) Digital encoder disc (f ) Binary coded disc (g) Gray coded disc.
of the same pitch P, slightly angled to each other. Relative movement of the gratings in a horizontal direction causes a series of dark fringes to move vertically across the sensor, giving an approximately sinusoidal variation in light intensity and detector output signal. In the angular case, one grating has N lines, the other N + 1 lines (N is typically between 50 and 1200); for every complete revolution of the gratings there are N complete revolutions of the fringe pattern and N cycles of the detector output signal. Thus in both cases the exact displacement can only be established by counting the number of detector output cycles; moiré gratings are therefore often referred to as incremental encoders.
15.6 MEASUREMENT SYSTEMS
413
Figure 15.18 Intensity modulation using relative displacement of fibres: (a) Longitudinal (b) Transverse (c) Angular (d) Differential.
Figures 15.17(e), (f ) and (g) show coded disc systems for the digital measurement of angular position. Here the disc is marked out in transparent and opaque segments according to the required digital code. Gray code (Figure 15.17(g)) is a cyclic code in which only one digit can change between two consecutive sectors; this is in contrast to binary code (Figure 15.17(f )), where several digits may change between consecutive sectors. If the detectors are misaligned slightly, then with a binary coded disc false codes can occur during the transition from one sector to the next. This cannot occur with Gray code; Figure 15.17(e) shows a typical eight-digit Gray coded disc. The output digital code uniquely specifies the angular position of the disc, and coded discs are often referred to as absolute encoders.
Intensity modulation using relative displacement of fibres Figure 15.18 shows four possible arrangements for the measurement of small linear or angular displacements. Here the intensity of the radiation incident on the detector is modulated according to the relative displacement of two optical fibres. The underlying principles have already been discussed in Section 15.4.2. For example in Figure 15.18(a), as x is decreased the solid angle ω of the cone of rays into the second fibre increases, causing the power incident on the detector to increase. This effect continues until the cone of rays into the second fibre corresponds to the cone of emergence of the first fibre. The arrangements shown in Figures 15.18(a), (b) and (c) are affected by fibre misalignment, vibration and source variations; the differential system shown in Figure 15.18(d) should be unaffected by source variations.
Intensity modulation using reflection Figure 15.19(a) shows a simple system for the remote measurement of angular velocity. The rotating mechanism has a reflective strip on its surface so that once in every
414 OPTICAL MEASUREMENT SYSTEMS Figure 15.19 Intensity modulation using reflection: (a) Angular velocity sensor (b) Level sensor (c) Two-fibre linear displacement sensor and characteristics (d) Three-fibre angular displacement sensor (e) Three-fibre pressure sensor.
revolution light is reflected back directly to the detector. This means that the number of detector pulses per second is equal to the revolutions of the mechanism per second. In the liquid level measurement system shown in Figure 15.19(b) a change in level ∆h causes the reflected beam to be displaced laterally by an amount proportional to ∆h. This change in lateral displacement is sensed by a two-dimensional array of detectors. Figure 15.19(c) shows a two-fibre linear displacement sensor and its characteristics. These characteristics show the non-linear relationship between the power incident on the detector and the separation x of fibres and reflecting surface. For small x, the detector power increases with x because the area of overlap between the cones of emergence and acceptance increases. For larger x, detector power reaches a maximum value and then decreases according to an inverse square law. This system is affected by vibration, changes in source intensity and reflection coefficient; there are also two values of separation for a given value of detector power.
15.6 MEASUREMENT SYSTEMS
415
A potentially more accurate sensor is the three-fibre angular displacement sensor shown in Figure 15.19(d).[8] Here there is a single source fibre and two detector fibres; changing the angle θ alters the ratio of the powers collected by the two return fibres and the ratio of detector output signals. This ratio depends only on θ and the refractive indices of the fibre core and cladding; it is independent of source intensity, reflection coefficient and fibre–reflector separation. One application of this is the three-fibre pressure sensor shown in Figure 15.19(e). Here the elastic element is a thin flat circular diaphragm clamped around the circumference (Figure 8.21). The fibre system is positioned close to a point on the diaphragm where the angular deformation is maximum.
Intensity modulation using attenuation and absorption In this type of sensor the intensity of radiation arriving at the detector is modulated either by variable attenuation in an optical fibre system or by variable absorption in gases and liquids. Figure 15.20(a) shows an optical fibre force sensor; an increase in force F causes increased microbending of the fibre. This causes rays with φ just greater than the Figure 15.20 Intensity modulation using attenuation and absorption: (a) Microbend force sensor (b) Two-fibre level detector (c) Oil-in-water detector (d) Infrared gas analyser (two-beam).
416 OPTICAL MEASUREMENT SYSTEMS critical angle θc (Figure 15.9), which would normally be totally internally reflected back into the core, to be increasingly refracted into the cladding and lost from the core. The result is that the power incident on the detector decreases as F is increased. An optical level detector is shown in Figure 15.20(b). This consists of two fibres (one inlet, one exit) cemented onto a 90° glass prism. For a low level, the prism is surrounded by air so that total internal reflection occurs at both prism faces, resulting in a high detector output signal. For a high level, the prism is surrounded by liquid so that only partial internal reflection occurs, resulting in a much lower detector output signal. The system shown in Figure 15.20(c) can be used to detect the presence of oil droplets in water. The water flows over an optical fibre core from which the cladding has been removed. The refractive index of the core is similar to that of oil, but higher than water. This means that if an oil droplet settles on the core some rays of light are refracted out of the core, causing a reduction in detector output signal. In Section 15.3 we saw that many hydrocarbon and other gases show strong absorption bands in the infrared; this means that if the percentage of the absorbing molecule in a gas mixture changes then there is a corresponding change in the transmission characteristics T(λ). Figure 15.20(d) shows the infrared gas analyser, which uses this principle. Earlier systems were broadband with IR lamps as sources and thermopile or pyroelectric thermal detectors. Later systems are narrowband with IR LED sources and photovoltaic detectors with wavelength response matched to that of the source. In both cases the source radiation is often chopped with a rotating mechanical chopper. This means that the detector output signals are a.c. so that electronic band-pass filters and phase-sensitive detectors can be used to reject electrical interference and noise. The system shown is a two-beam type, with the sample cell containing the gas to be analysed and the reference cell containing the pure gas of interest. For example, if we wish to measure the percentage of CO in flue gas then the reference cell will contain pure CO. The detector output signals are amplified and their ratio R computed; R depends on the percentage of the absorbing molecule in the sample gas, but should be independent of modifying effects such as source variations.
15.6.3
Two-wavelength systems The simple intensity-modulated systems discussed so far can be affected by environmental modifying and interfering inputs. For example, the source-modulated broadband and narrowband temperature systems are affected by changes in emissivity and transmission medium characteristics; the transmission medium-modulated systems are affected by changes in source intensity. Some of these problems can be solved using the two-wavelength system shown in Figure 15.21. Here the incoming radiation from the transmission medium is focused and split equally into two beams. One beam is passed through a narrow pass-band optical filter with centre wavelength λ1, the other through a similar filter with centre wavelength λ 2. The detectors chosen should have the same value of wavelength response D(λ) at λ1 and λ 2; either thermal detectors or photon detectors with a suitable wavelength response could be used. An electronic divider computes the ratio R between detector output signals V1 and V2. Using eqn [15.57] for a narrowband radiation pyrometer, we can derive the equation for a two-wavelength (or two-colour) pyrometer:
15.6 MEASUREMENT SYSTEMS
417
Figure 15.21 Twowavelength system.
1 R(T ) =
π V1(T ) = V2 (T ) 1 π
(C1εTM K F FK D )λ (C1εTM K F FK D )λ
 C  exp  − 2  λ  λ1T 
∆λ 1
5 1
 C  exp  − 2  5 λ2  λ 2T 
∆λ 2
[15.58]
If λ1 and λ2 are close together, so that the emissivities and transmission factors are equal, then the ratio becomes: Equation for twocolour pyrometer system
5
λ  1 C  1 R(T ) =  2  exp 2  −  T  λ 2 λ1   λ1 
[15.59]
This system does not require calibration and correction for emissivity variations. The two-wavelength principle can also be used in the infrared gas analyser shown in Figure 15.20(d). Instead of using a two-beam system, with reference and sample cells, a single beam–sample cell system is used. Here λ1 coincides with the centre of the absorption band for the component of interest and λ 2 with the centre of the band for another component.
15.6.4
Modulation of wavelength by transmission medium All of the systems discussed so far rely on the measured variable modulating the intensity or amplitude of light waves. These systems are inherently susceptible to unwanted variations in source intensity and transmission characteristics. Twochannel/wavelength systems overcome some of these problems, but wavelength modulation systems, where the wavelength λ of the light is determined by the measured variable, are completely independent of these unwanted effects. Two such systems have been developed at the National Physical Laboratory[9] and will now be described. Figure 15.22(a) shows a ‘Littrow’ diffraction grating used as a sensor to convert angular displacement θ into changes in wavelength. Broadband radiation emerging from the input fibre is first collimated into a parallel beam; this is incident at an angle θ onto a reflection grating. The grating consists of a regular array of grooves whose spacing d is comparable to the wavelength of the incident beam. The resulting diffraction pattern is complex, but only light with wavelengths λ satisfying the equation
418 OPTICAL MEASUREMENT SYSTEMS
Figure 15.22 Wavelength modulation systems: (a) Littrow grating angular displacement sensor (after Hutley[9]) (b) Fresnel zone plate used as a lens and linear displacement sensor (after Hutley[9]) (c) Wavelength/ displacement characteristics of zone plate.
2d sin θ = mλ
[15.60]
where m = 1, 2, 3, . . . etc., is reflected back along its own path to the output fibre. By limiting the range of incident wavelengths, only first-order effects with m = 1 need to be considered, giving the following unique relation between reflected wavelength λ and grating angle θ : Wavelength-encoded angular displacement sensor
λ = 2d sin θ
[15.61]
The angular range of this device is limited to between 20° and 60°. Figure 15.22(b) shows a Fresnel zone plate; it is a circular grating consisting of a number of concentric lines, the spacing between the lines decreasing with distance from the centre. The top diagram shows a parallel beam of monochromatic light incident on the zone plate. All rays are brought to a focus at a single point so that the plate acts as a lens of focal length f. However, the deviation of each ray depends on wavelength, so that different wavelengths are brought to a focus at different points; the focal length f is inversely proportional to wavelength λ. The bottom diagram shows a reflecting zone plate, used as a linear displacement sensor. The plate acts as a spherical concave mirror with a radius of curvature R inversely proportional to wavelength (Figure 15.22(c)). Broadband radiation from the input fibre is incident on the zone plate a distance x away; the radiation reflected back to the output fibre is narrowband, with centre wavelength λ such that:
15.6 MEASUREMENT SYSTEMS
Wavelength-encoded linear displacement sensor
R(λ) = x
419
[15.62]
Thus if the range of displacement to be measured is x MIN to x MAX, the required range of wavelengths in the incoming radiation is λ MIN to λ MAX where R(λ MAX ) = x MIN R(λ MIN ) = x MAX Because of the need to exclude overlapping higher orders, λ MAX cannot usually exceed 2λ MIN, with corresponding limits on the displacement range. A wavelength decoder is necessary with both of these devices to convert wavelength variations into a usable voltage signal. One possibility is to split the beam from the output fibre equally between two detectors, with very different wavelength responses D(λ) over the wavelength range of interest. The ratio of detector output voltages should vary monotonically with wavelength.
15.6.5
Interferometers We first need to discuss the phenomenon of interference, which is exhibited by any type of wave motion. Consider a monochromatic light source which is emitting light at a single angular frequency ω rad/s and corresponding wavelength λ. The light beam is then split into two beams of equal amplitude a. These beams then travel along different optical paths of lengths x1 and x2 before both arriving at a detector. The beams can be described by the wave equations (Section 16.3.1):   2π  2π  x1 , φ2 = a sin ω t − x2  φ1 = a sin ω t − λ  λ   
[15.63]
so that the resultant light beam incident on the detector is given by:    2π  2π   x1 + sin ω t − x2   φ = φ1 + φ 2 = a sin ω t − λ  λ       2π  x − x1    2π  x1 + x2   sin − ω = 2a cos   2 t     λ  2    λ  2   
[15.64]
This has amplitude: π aR = 2a cos  λ
 x 
[15.65]
where x = optical path difference x2 − x1, and intensity: Variation in resultant intensity with path difference
π I R = a R2 = 4a 2 cos 2  λ
 x 
[15.66]
Figure 15.23(a) shows the form of eqn [15.66]. We see that when x = 0, λ, 2λ, 3λ, etc., IR = 4a2, i.e. the intensity of the resultant beam corresponds to the sum (2a)
420 OPTICAL MEASUREMENT SYSTEMS Figure 15.23 Interferometers: (a) Variation in resultant intensity with path difference (b) Michelson interferometer.
of the amplitudes of the individual beams, then this is constructive interference. When x = λ /2, 3λ /2, 5λ /2, etc., i.e. the intensity of the resultant beam corresponds to IR = 0, the difference (i.e. 0) of the amplitudes of the individual beams, then this is destructive interference. Thus if path difference x is initially zero, IR is maximum and a bright fringe is observed. If x is increased to λ /2, IR is zero and a dark fringe is observed. If x is further increased to λ, IR returns to maximum and a bright fringe is again observed. Thus the variation in the intensity of the light incident on the detector, and therefore the detector output signal, with path difference x, has a period equal to λ. Figure 15.23(b) shows a Michelson interferometer which is used in the accurate measurement of displacement. Here a monochromatic beam of light from a laser source (Section 15.2.4) is split into two equal beams 1 and 2 at A using a beam splitter (in its simplest form this is a half-silvered mirror). Beam 1 travels to a fixed retro-reflector and is reflected back to the point E, so that the total optical path length x1 = ABCDE.
15.6 MEASUREMENT SYSTEMS
421
Beam 2 travels to a movable retro-reflector and is again reflected back to the point E, so that here the total optical path length x2 = AXYFE. The optical path difference x is therefore: x = x2 − x1 = AXYFE − ABCDE = (AX + XY + YF + FE) − (AB + BC + CD + DE)
[15.67]
Since XY = DE and FE = BC, then x = (AX + YF) − (AB + CD) = 2AX − 2AB
[15.68]
If displacement d is measured from the line OO′, where AO′ = AB, then AX = AB + d and x = 2(AB + d) − 2AB = 2d
[15.69]
This means that a change δ d in d results in a change δ x in x of 2δ d, so that if δ d = λ /2, i.e. half a wavelength, then δ x = λ, i.e. a full wavelength. Thus with a Michelson interferometer the detector output signal varies periodically with displacement d, the period being equal to λ /2, one half-wavelength of the laser radiation. A Michelson interferometer used as a standard for length (Section 2.4.1) would typically use a helium–neon laser source with wavelength λ = 633 nm. Thus a displacement d = 1 mm would produce: n=
d 2 × 10−3 2 × 106 = = = 3159.56 λ /2 633 × 10−9 633
cycles of detector output signal. This displacement can therefore be measured to within a resolution of ± λ /4 by counting the number of whole cycles, 3160 (rounded up) in this case. We see therefore that a basic interferometer has a very high sensitivity when used for displacement measurement. For all displacements greater than λ or λ /2 it is necessary to count the total number of cycles of intensity variation. The sensitivity can be reduced by using a source of much longer effective wavelength λ′ obtained by mixing two laser beams with much shorter wavelengths λ1 and λ 2. This is called synthetic wave or multisource heterodyne interferometry. Thus, if we add two waves with the same amplitude a but with different frequencies ω 1 and ω 2 and wavelengths λ1 and λ 2, the resultant wave is:   2π  2π  φ = a sin  ω1t − x + a sin  ω 2t − x λ1  λ 2     ω − ω  1 1 1  2 −  = 2a cos  1  t − 2π x ⋅  2  2  λ 1 λ 2     ω + ω  1 1 1  2 − + t 2 π x ⋅ × sin  1    2  2  λ 1 λ 2   
[15.70]
This is an amplitude modulated wave, with a carrier frequency of (ω1 + ω 2)/2 and a much lower amplitude modulating frequency of (ω 1 − ω 2)/2. The amplitude or envelope function:
422 OPTICAL MEASUREMENT SYSTEMS  ω − ω  1 1 1  2 − π ⋅ − 2a cos  1 t 2 x    2  2  λ 1 λ 2    is a progressive wave of the form:  2π  2a cos  ω ′t − x λ ′   with frequency ω ′ = (ω 1 − ω 2)/2 and wavelength: 2λ 1λ 2 1 1 1 1 =  −  , i.e. λ ′ = λ ′ 2  λ1 λ 2  λ 2 − λ1
[15.71]
Thus if λ 2 = 1001 nm and λ1 = 1000 nm, λ′ = 2 mm. This means that if the envelope wave with wavelength λ′ is used as a source in an interferometer, displacements of up to 1 mm can be detected within one cycle of the intensity variation. The carrier wave with wavelength (λ1 + λ 2)/2 can still be used as an alternative source if high precision measurements are required.
Conclusion This chapter has discussed the following: 1. 2. 3. 4. 5. 6.
The two general types of optical measurement system: fixed source, variable transmission medium; and variable source, fixed transmission medium. The principles and types of sources: hot body, light-emitting diodes and lasers. The principles and characteristics of optical transmission media. The geometry of coupling of detector to source and the use of focusing systems and optical fibres. The principles and characteristics of thermal and photon detectors. Measurement systems using the modulation of intensity by source and transmission medium, two-wavelength systems, wavelength modulation and interferometry.
References [1] [2] [3] [4]
Polarizers Technical Products 1988 Technical Information on Temperature Standards and Black Body Sources. keiser g 1983 Optical Fibre Communications, McGraw-Hill, Tokyo, pp. 12 – 44, 98–115, 145–67. cherin a h 1983 An Introduction to Optical Fibres, McGraw-Hill, Tokyo, pp. 270–4. houghton j and smith s d 1966 Infrared Physics, Oxford University Press, London, pp. 220–3.
PROBLEMS
[5] [6] [7] [8] [9]
423
Plessey Optoelectronics and Microwave Ltd, Application Notes on Pyroelectric and Lead Tin Telluride Infra-red Detectors. jones b e 1985 ‘Optical fibre sensors and systems for industry’, J. Phys. E.: Scientific Instruments, vol. 18, pp. 770–82. medlock r s 1986 ‘Review of modulating techniques for fibre optic sensors’, Measurement and Control, vol. 19, no. 1, pp. 4–17. hill s Optical Fibre Reflectance Transducer, Teesside Polytechnic project report, unpublished. hutley m c 1985 ‘Wavelength encoded optical fibre sensors’, N.P.L. News, no. 363.
Problems 15.1
A broadband radiation pyrometer has a thermistor bolometer detector. The radiant power incident on the thermistor in the steady state is given by: PD = 10T 4 pW where T is the temperature of the distant target in kelvin. The beam of radiation is chopped by a semicircular disc rotating at 60 r.p.m. The thermistor is incorporated into the deflection bridge circuit shown in Figure Prob. 15.1 together with an identical thermistor at 290 K, the temperature of the detector surroundings. Using the detector data given below, sketch the form of the bridge output signal when the target temperature is 1000 K. Hint: assume a square wave variation in detector power which can be regarded as a series of steps.
Figure Prob. 1.
Thermistor data Mass = 1 g Surface area = 10−4 m2 Specific heat = 102 J kg−1 K−1 Heat transfer coefficient = 104 W m−2 K−1 A 3000 D Resistance = 1.0 exp Ω where θ = thermistor temperature in kelvin C θ F 15.2
A thermocouple has an e.m.f. of 5 mV when the hot junction is at 100 °C and the cold junction is at 0 °C. A thermopile consisting of 25 such thermocouples in series is used as a detector in a chopped, broadband pyrometer. When unchopped the power in the beam of radiation incident on the hot junction of the thermopile is given by 7.5T 4 pW, where T K is the temperature of the distant target. The beam is chopped by a semicircular disc rotating at 6000 r.p.m. The reference junction of the thermopile is at the temperature of the pyrometer
424 OPTICAL MEASUREMENT SYSTEMS enclosure. Assuming the thermocouples are linear and the detector data given below, sketch the form of the output signal when T = 2000 K. Detector data Heat transfer coefficient = 2 × 104 W m−2 °C −1 Surface area = 1.6 × 10−4 m2 Time constant = 1 ms 15.3
A narrowband optical pyrometer is to be used to measure the temperature of the radiant section of a furnace in the range 1500 to 2000 K. The emissivity of the furnace wall is 0.5 at a wavelength of 0.5 µm. The transmission factor of the atmosphere between the furnace and the pyrometer is 0.90 at 0.5 µm. Calculate the range of the pyrometer output signal given the following data: Lens Diameter of aperture = 3.0 cm Focal length at 0.5 µm = 10.0 cm Transmission factor at 0.5 µm = 0.95 Photoconductive detector Diameter of detector disc = 0.5 cm Detector sensitivity = 10 Ω W−1 Centre wavelength = 0.5 µm Bandwidth = 0.05 µm Deflection bridge and amplifier Overall sensitivity = 10 V Ω−1 Temperature for bridge balance = 1500 K The Planck–Wien law for the hemispherical spectral intensity of radiation from a black body is:
W (λ , T ) =
C1   C  λ5 exp  2  − 1  λT   
where T = absolute temperature of body (K) λ = wavelength of radiation (µm) C1 = 37 400 W µm4 cm−2 C2 = 14 400 µm K 15.4
A two-colour radiation thermometer is used to measure the temperature of a body at 800 °C. The system incorporates two detectors, one measuring radiation at 0.5 µm and the other at 1.0 µm. What is the ratio of the detector output signals? State clearly all the assumptions made in the calculation. Explain the advantage of this system over broadband and narrowband systems.
15.5
An optical fibre transmission system consists of a circular LED source, a 2-metre length of optical fibre and a circular PIN diode detector. Both the source and detector are positioned 100 µm from the ends of the fibre. Detailed data for the source, a glass fibre, a polymer fibre and the detector are given below. Use this data to perform the following calculations. (a) (b) (c) (d)
The total power Ps emitted by the source in all directions (use eqn [15.23b]). The numerical aperture and maximum angle of acceptance of the glass fibre. The numerical aperture and maximum angle of acceptance of the polymer fibre. The source and detector are linked by the glass fibre; calculate the power input to the fibre, the fibre transmission factor, and power output from the fibre. (Hint: use PIM /R = KSM and POM /PIM = T .)
PROBLEMS
(e) (f )
425
The source and detector are now linked by the polymer fibre; calculate the power input to the fibre, the fibre transmission factor, and power output from the fibre. For both the glass system of (d) and the polymer system of (e), calculate the power incident on the detector and the detector output current. (Hint: use PD /POM = KMD .)
LED source Radiance = 10 W cm−2 sr−1 Diameter = 200 µm Centre wavelength = 810 nm Glass fibre Multimode step index Core diameter = 100 µm Core refractive index = 1.5 Core-cladding index difference = 0.015 Attenuation at 810 nm = 5 dB km−1 Fibre length = 2 m Polymer fibre Multimode step index Core diameter = 1.0 mm Core refractive index = 1.65 Core cladding index difference = 0.04 Attenuation at 810 nm = 500 dB km−1 Fibre length = 2 m PIN diode detector Diameter = 2.0 mm Centre wavelength = 810 nm Responsivity at 810 nm = 0.55 A W−1 15.6
A system for the measurement of linear displacement consists of a reflecting Fresnel zone plate (Figure 15.22(b)), acting as a wavelength encoder, and a pin diode photon detector, acting as a wavelength decoder. The distance between input/output fibres and the zone plate (input displacement range) can be varied between 12.5 and 25 mm. The radius of curvature R(λ) of the zone plate is given by R(λ ) = 1.25 × 104/λ where R is in mm and λ in nm. The detector has D(λ) = 9 × 10−4λ − 0.35. Calculate the wavelength output range for the encoder and the corresponding ratio of decoder output signals.
16 Ultrasonic Measurement Systems Ultrasound refers to sound waves at frequencies higher than the range of the human ear, i.e. at frequencies greater than about 18 kHz. Ultrasonic waves obey the same basic laws of wave motion as lower frequency sound waves; they have, however, the following advantages: (a)
(b)
(c) (d)
Higher frequency waves have shorter wavelengths; this means that diffraction or bending around an obstacle of given dimensions is correspondingly reduced. It is therefore easier to direct and focus a beam of ultrasound. Ultrasonic waves can pass easily through the metal walls of pipes and vessels. This means that the entire measurement system can be mounted completely external to the fluid, i.e. is non-invasive. This is extremely important with hostile fluids, such as those with corrosive, radioactive, explosive or flammable properties. There is also no possibility of blockage occurring with dirty fluids or slurries. Ultrasound can be launched into and propagated through biological tissue, making it useful for medical applications. The silence of ultrasound means that it has important military applications.
This chapter studies ultrasonic transmitters and receivers, and the principles of transmission, and examines a range of ultrasonic measurement systems.
16.1
Basic ultrasonic transmission link This forms the basis of any ultrasonic measurement system and is shown in Figure 16.1. It consists of an ultrasonic transmitter, the transmission medium and an ultrasonic receiver. The most commonly used devices for ultrasonic transmitters and receivers are piezoelectric sensing elements (Section 8.7). The piezoelectric effect is reversible, i.e. mechanical energy can be converted into electrical energy and electrical energy into mechanical. The ultrasonic transmitter uses the inverse piezoelectric effect; if a sinusoidal voltage ÄS sin ω t is applied to the transmitting crystal, then the crystal undergoes a corresponding sinusoidal deformation x (eqn [8.72]). This vibration of the crystal is transmitted to the particles at the beginning of the medium, and these are set in sinusoidal motion, causing other particles to vibrate, until eventually the disturbance is transmitted to the other end of the medium. These sinusoidal particle displacements set up an accompanying sinusoidal pressure or stress in the medium.
428 ULTRASONIC MEASUREMENT SYSTEMS Figure 16.1 Basic ultrasonic transmission link.
This is detected by the ultrasonic receiver, which is simply a force sensor using the direct piezoelectric effect. The fluctuating pressure causes a sinusoidal force F over the area of the crystal, thus producing a corresponding time-varying charge q and current i. This current produces a corresponding output voltage VOUT across a load ZL.
16.2 16.2.1
Piezoelectric ultrasonic transmitters and receivers Basic principles In Chapter 5 we introduced the concept of effort and flow variables: voltage and force are examples of effort variables; current and velocity are examples of flow variables. The interconnection of any two elements in a system is then represented by two terminals and the appropriate effort/flow pair, e.g. voltage/current at an electrical connection and force/velocity at a mechanical connection. This leads to the concept of an element being represented by a four-terminal or two-port network. Thus in Figure 5.17 a force sensor is represented as two-port network with a mechanical input port and an electrical output port. Two-port networks provide a good representation of piezoelectric transmitters and receivers; because the piezoelectric effect is reversible the same crystal can act as a transmitter (electrical input port/mechanical output port) or as a receiver (mechanical input port/electrical output port) and is said to be bilateral. Figure 16.2(a) shows transmitter and receiver two-port networks in block diagram form. In order to derive the detailed form of these circuits we need to revise the properties of piezoelectric elements (Section 8.7). Starting with the basic equations for the direct and inverse effects we have: Direct effect q = dF
[8.71]
Inverse effect x = dV
[8.72]
We also have the equation F = kx relating applied force F and deformation x of a crystal of stiffness k. Thus q = dkx and: Relationship between flow variables for piezo-crystal
i=
dq = dk x dt
[16.1]
16.2 PIEZOELECTRIC ULTRASONIC TRANSMITTERS AND RECEIVERS
429
Figure 16.2 Two-port networks for piezoelectric transmitters and receivers: (a) Block diagrams (b) Detailed form.
Similarly F/k = dV, giving: Relationship between effort variables for piezo-crystal
F = dkV
[16.2]
The crystal has electrical capacitance C and the corresponding electrical impedance is purely capacitive: ZE (s) =
∆(s) 1 = ∆j(s) Cs
[16.3]
The mechanics of the crystal can be represented most simply using single mass (inertance m), spring (stiffness k) and damping (resistance λ) elements. The corresponding mechanical circuit consists of m, 1/k and λ in series and has mechanical impedance (Section 5.2): ZM (s) =
∆,(s) k = ms + λ + ∆á(s) s
[16.4]
430 ULTRASONIC MEASUREMENT SYSTEMS
Figure 16.3 Transfer functions for transmitter and receiver systems: (a) Signal generator/transmitter/ medium (b) Medium/receiver/ external load.
Figure 16.2(b) shows the detailed form of the transmitter and receiver two-port networks. In the transmitter network, a voltage V at the input electrical port produces a force dkV which drives a velocity x through the output mechanical impedance. In the receiver network, a force F at the input mechanical port drives a velocity x through the input mechanical impedance; this produces a current dkx at the output electrical port. Figure 16.3(a) shows the input electrical port of the transmitter connected to an external signal generator and the output mechanical port to the beginning of
16.2 PIEZOELECTRIC ULTRASONIC TRANSMITTERS AND RECEIVERS
431
the transmission medium. The figure derives the overall transfer function relating crystal displacement and generator voltage. In the special case that the output impedance of the generator and the input impedance of the medium are negligible, i.e. ZG ≈ 0 and Z MIN ≈ 0, the overall transfer function simplifies to the standard secondorder form: Transmitter system transfer function
∆P d ( s) = ∆ S 1 2 2ξ s + s +1 ω n2 ωn
[16.5]
where ωn and ξ are the mechanical natural frequency and damping ratio for the crystal (see Section 8.7). Figure 16.3(b) shows the input mechanical port of the receiver connected to the end of the transmission medium and the output electrical port connected to an external load such as an oscilloscope or amplifier. The figure derives the overall transfer function relating voltage across the load VOUT to force FOUT (due to medium output pressure acting over the area of the receiver crystal). In the special case that the output impedance of the medium is negligible and the load is purely resistive, i.e. Z MOUT ≈ 0 and ZL = RL, the overall transfer function simplifies to: Receiver system transfer function
 ∆ OUT 1 d  τs   ( s) =    ∆ ,OUT C  1 + τ s  1 2 2ξ s + 1  2s + ωn   ωn
[16.6]
This is identical to the transfer function for the basic piezoelectric force measurement system discussed in Section 8.7: an ultrasonic receiver is simply a piezoelectric force sensor applied to the specific problem of measuring high frequency sinusoidal pressure variations.
16.2.2
Crystal oscillators and resonators Rather than energise the crystal with an external signal generator, the crystal is usually incorporated into a closed-loop system which oscillates at the mechanical natural frequency ωn. In order to discuss the principles of crystal oscillators we need to find the overall electrical impedance of the crystal. Figure 16.4 shows an approximate electrical equivalent circuit for the crystal. We see that the total current i drawn by the crystal is the sum of the capacitance current iC together with the current im required to drive the mechanical mass/spring/damper system. C is therefore shunted by a series electrical circuit L1, C1, R1 representing these mechanical losses. The impedance ZE of the series circuit is given by:
Figure 16.4 Equivalent electrical circuit for crystal.
ZE (s) =
∆ ∆ ∆á ∆, (s) = ∆jm ∆, ∆jm ∆á
[16.7]
432 ULTRASONIC MEASUREMENT SYSTEMS where the ratios ∆/∆, and ∆á/∆jm are both equal to 1/dk (eqns [16.1] and [16.2]), and ∆,/∆á is the crystal mechanical impedance Zm(s) (eqn [16.4]). Thus: ZE (s) =
1 A kD ms + λ + 2C (dk) sF
i.e. Equivalent electrical impedance of mechanical system
Z E (s) = L1s + R1 + where
L1 =
1 C1s
m , (dk )2
R1 =
λ , C1 = d 2k (dk )2
[16.8]
The overall electrical impedance transfer function is thus given by: 1 1 1 1 = + = Cs + H(s) 1/Cs ZE (s) R1 + L1s + 1/C1s
[16.9]
For sinusoidal signals we replace s by jω to give: 1 1 = jω C + H( jω) R1 + j(ω L1 − 1/ω C1)
[16.10]
and Overall electrical impedance of piezoelectric crystal
H ( jω ) =
ω R1C1 − j(1 − ω 2 L1C1 ) ω (C + C1 − ω 2 L1CC1 ) + jω 2CC1R1
[16.11]
Figure 16.5 gives values of all the above parameters for a typical crystal, and shows the corresponding frequency variations of impedance magnitude | H( jω)| and phase arg H( jω). From the figure we see that there are two important frequencies. The first is the natural frequency ωn = êk/úmú = 1/êL 1úCú 1ú of the mechanical system. Here we have 1 − ω n2 L 1C 1 = 0, so that the corresponding impedance magnitude |H( jωn )| is almost minimum; ωn is termed the series resonant frequency. At the second frequency:
ω1 =
C + C1 L1C1C
i.e. C + C1 − ω 12 L 1CC 1 = 0, so that the corresponding impedance |H( jω 1)| is almost maximum. ω 1 is termed the parallel resonant frequency. Below ω n and above ω 1, arg H( jω) = −90°, i.e. the crystal behaves as a capacitance. Between ω n and ω 1, arg H( jω) = +90°, i.e. the crystal is inductive.
16.2 PIEZOELECTRIC ULTRASONIC TRANSMITTERS AND RECEIVERS
Figure 16.5 Electrical impedance of typical piezoelectric crystal.
433
8 7 6 5 | H( jω) | (kΩ) 4 3 Approximate frequency response
2 1 0 +90
arg H( jω )
ωn
ω1
ω
0
−90 Crystal data fn = 31.8 kHz
k = 2 × 109 N/m
ξ = 0.01 C = 1600 pF L1 =
λ = 200 Ns/m dk = 0.4
m = 5 × 10 −2 kg d = 2 × 10−10 C/N
λ m = 312.5 mH, R1 = = 1250 Ω, C1 = d 2k = 80 pF (dk )2 (dk )2
ωn =
k = m
ω1 =
C + C1 = 2.05 × 105 rad s−1 L1C1C
H( jω n ) =
H ( jω n ) =
1 L1C1
1250
= 2.00 × 105 rad s−1
Q=
mk
λ
=
1 2ξ
= 50
1 | H( jω n) | = 1161 Ω 2
1 + 0.4 j 3 arg H( jω n) = −22°
3049 + 7512j 1 | H( jω 1) | = 8105 Ω j
2 3 arg H( jω 1) = −22°
A schematic diagram of a crystal oscillator is shown in Figure 16.6. It is a closedloop system, with a ‘maintaining amplifier’ of impedance 1/G( jω ) in the forward path, and a piezoelectric element of impedance H( jω ) in the feedback path. From Figure 16.6 we have: Figure 16.6 Schematic diagram of crystal oscillator.
i VREF − V
= G( jω ) and
V = H( jω ) i
434 ULTRASONIC MEASUREMENT SYSTEMS giving i VREF
=
G( jω) 1 + H( jω)G( jω)
[16.12]
Using the arguments of Section 9.5, we can show that continuous oscillations at frequency ω can be maintained with VREF = 0 if: |G( jω)| |H( jω)| = 1
[9.52]
and arg G( jω) + arg H( jω) = −180° Thus to maintain oscillations at ωn we require: |G( jωn)| =
1 |H( jωn)|
[16.13]
and arg G( jωn ) = −180° − arg H( jωn ) For the crystal specified by Figure 16.4 we require the maintaining amplifier to have: |G( jωn )| =
1 = 8.61 × 10−4, 1161
arg G( jωn ) = −158°
[16.14]
It is possible to adjust the frequency of oscillation slightly, by adjusting G( jω) for the maintaining amplifier.
16.2.3
Characteristics and applications Table 16.1 gives physical data for commonly used piezoelectric materials.[1] Characteristic acoustic impedance will be discussed in Section 16.3. Mechanical quality factor Q is related to damping coefficient ξ by the equation Q = 1/2ξ. Ultrasonic transmitters and receivers are normally in the form of a thin disc of piezoelectric material with metal electrodes deposited on each face. A disc of diameter 10 mm and
Table 16.1 Physical properties of piezoelectric materials.
Material
Density ρ × 103 kg m−3
Sensitivity d pC N−1
Quartz BaTiO3 PZT–5H PbNb2O6 PVDF
2.65 5.7 7.5 5.9 1.8
2.3 149 593 75 ≈10
Source: adapted in part from O’Donnell et al.[1] .
Dielectric constant ε
Longitudinal velocity of sound cL × 103 m s−1
Characteristic acoustic impedance RA × 107 kg m−2 s−1
Mechanical quality factor
4.5 1700 3400 240 13
5.65 4.39 4.0 2.7 2.2
1.5 2.5 3.0 1.6 0.4
> 25 000 < 400 < 65 <5 14
Q=
mk λ
16.2 PIEZOELECTRIC ULTRASONIC TRANSMITTERS AND RECEIVERS
Figure 16.7 Piezoelectric transmitters and receivers: (a) Vibration displacement modes (b) Performance characteristics.
435
thickness 1 mm would typically have a natural frequency fn of 5 MHz; the amplitude of the crystal deformation is around a few µm. The two usual modes of vibration are shown in Figure 16.7(a); the thickness expander mode is used in the generation and reception of longitudinal waves and the thickness shear mode in the generation of transverse waves. The performance characteristics of piezoelectric transmitters and receivers can be summarised by four graphs: Figure 16.7(b) shows the form of the graphs in a typical case.[2] Graph 1 shows a decibel plot of the transmitter frequency response. This is based on the pressure/voltage transfer function (∆
[16.15]
where A is the area of the circular face of the disc. The decibel plot is relative to 1 µbar, i.e. dB = 20 log10(|IN /1 µbar), and for a supply voltage Vs of 1 V rms. The graph shows a sharp resonance peak at fn, which is characteristic of a second-order mechanical system with low ξ or high Q.
436 ULTRASONIC MEASUREMENT SYSTEMS Graph 2 shows a decibel plot of the receiver frequency response. This is based on the voltage/pressure transfer function ∆OUT /∆
[16.16]
The decibel plot is relative to 1 volt, i.e. dB = 20 log10(ÄOUT /1 V), and for a pressure of 1 µbar. The graph again shows a sharp resonance peak at fn. Graph 3 shows how the magnitude |H( j2πf )| of the transducer electrical impedance varies with frequency (eqn [16.11] and Figure 16.5). Graph 4 shows the directional response of the transmitter, i.e. how the pressure of the sound wave (at frequency fn), launched from the transmitter, varies with angle. Ultrasonic transmission links are usually either narrowband or broadband. In narrowband systems a narrow band of sound frequencies, close to the transducer natural frequency, is transmitted over the link. Here both transmitter and receiver should have a high Q to give a sharp resonant peak. Quartz is therefore an ideal material; a composite resonator is usually set up consisting of transmitter, transmission medium and receiver. In broadband systems either a broad band of frequencies or a pulse of sound is transmitted over the link. Here both transmitter and receiver should have a low Q to give a fairly flat frequency response. In pulse systems, Q should be ideally around 0.7, i.e. ξ = 0.7, give an optimum step response and minimum pulse distortion. Here low Q materials such as PbNb2O6 or PVDF should be used with a backing layer of absorbing material to further reduce Q (Section 16.3.5).
16.3
Principles of ultrasonic transmission As mentioned earlier, ultrasonic waves obey the same transmission principles as lower frequency acoustic waves.
16.3.1
Propagation of plane waves Figure 16.8 shows the transmission medium represented as a series of layers of particles. If an acoustic plane wave is passing through the medium, then each particle in a given layer, at a given time, is displaced by the same amount. For a longitudinal plane wave, the displacement dy of a layer at position y is parallel to the direction of propagation of the wave, i.e. parallel to the Oy axis. Longitudinal plane waves are described by the partial differential equation: ∂ 2dy 1 ∂ 2d y = ∂y 2 cL2 ∂t 2
[16.17]
where dy is the displacement of layer y at time t and cL is the velocity of propagation of the longitudinal wave. For a progressive wave moving along the positive y direction, the solution to [16.17] is: Longitudinal plane wave along positive Oy direction
 y d y = • y sin ω L  t −  cL  
[16.18]
16.3 PRINCIPLES OF ULTRASONIC TRANSMISSION
437
Figure 16.8 Longitudinal plane waves: (a) Wave fronts (b) Variation of dy with y at t = t0, dy = •y sin ωL(t0 + y/cL) (c) Variation of dy with t for layer A with y = y0, dy = •y sin ω L(t + y0 /cL).
where •y is the amplitude of the layer displacement and ω L = 2π fL is the angular frequency. For a progressive wave moving along the negative y direction the solution to [16.17] is: Longitudinal plane wave along negative Oy direction
 y d y = • y sin ω L  t +  cL  
[16.19]
In each case, dy is a sinusoidal function of both time t and layer position y. Figures 16.8(a) and (b) represent ‘still’ pictures of the displacements of each layer, taken at time t = t0. Since time is now constant, dy becomes a sinusoidal function of position y only, i.e. A ωL yD dy = •y sin ωLt0 + C cL F
[16.20]
The distance between layers such as A and B, C and D, E and F, which have equal displacements, is defined to be the wavelength λL. At A and B, dy = •y, so that:
438 ULTRASONIC MEASUREMENT SYSTEMS
at A, ω Lt0 +
ω L y0 π = cL 2
at B, ω Lt0 +
ωL 5π (y0 + λL ) = cL 2
[16.21]
Subtracting gives ωLλL/cL = 2π, and since ω L = 2πfL, we have Relationship between velocity, wavelength and frequency for a longitudinal plane wave
cL = fL λL
[16.22]
Thus for a 1 MHz wave travelling with longitudinal velocity 6 × 103 m s−1 in steel, the corresponding wavelength is 6 mm. Figure 16.8(c) shows how the displacement of a given layer A, at y = y0, varies sinusoidally with time t according to: A y0 D dy = •y sin ωL t + C cL F
[16.23]
The period TL of the wave is the interval between time instants t1 and t2, t3 and t4, where the layer displacement is the same: as usual TL = 1/fL = 2π/ωL. For a transverse plane wave, the displacement dz of a layer at position y is perpendicular to the direction Oy of propagation of the wave, i.e. parallel to the Oz axis. The medium is therefore subject to time-varying shear strains. The corresponding partial differential equation is: ∂ 2dz 1 ∂ 2dz = ∂y 2 c 2T ∂t 2
[16.24]
where dz is the transverse displacement of layer y at time t, and cT is the transverse wave velocity. For a progressive wave travelling in the positive y direction, the solution to [16.24] is similar to [16.18], i.e. Transverse plane wave along positive Oy direction
 y dz = •z sin ω T  t −  cT  
[16.25]
where •z is displacement amplitude and ω T = 2πfT is the angular frequency. Using arguments similar to those above, we can show that the relationship between velocity, wavelength and frequency of a transverse wave is similar to [16.22], i.e. cT = fT λT
[16.26]
The velocity of sound in a medium depends on the appropriate elastic modulus and density ρ, according to the general equation: c=
elastic modulus ρ
[16.27]
16.3 PRINCIPLES OF ULTRASONIC TRANSMISSION
439
Gases will support compressive and tensile stresses but not shear stresses, so that only longitudinal acoustic waves can be passed through gases. The corresponding velocity is given by: Velocity of longitudinal sound waves in a gas
cL =
γP ρ
[16.28]
where P is the pressure of the gas and γ the specific heat ratio. In general, liquids collapse under the action of shear stresses, so that usually transverse waves cannot be transmitted through liquids. The velocity of longitudinal waves in liquids is given by: Velocity of longitudinal waves in a liquid
cL =
B ρ
[16.29]
where B is the bulk modulus of elasticity of the liquid. Since solids will support compressive/tensile and shear stresses, both longitudinal and transverse waves can be transmitted through solids. The corresponding velocities are given by[3] Velocity of longitudinal and transverse waves in a solid
cL =
E (1 − ν ) (1 + ν )(1 − 2ν )ρ [16.30]
cT =
s
ρ
where E is Young’s modulus, s is the shear modulus and ν is Poisson’s ratio for the solid.
16.3.2
Acoustic impedance and power In Chapter 5 we saw that mechanical force F and velocity x were an effort/flow pair, so that the product Fx represents mechanical power and the ratio F/x represents mechanical impedance. In acoustic work the particle velocity u = ∂d/∂t is the flow variable, and the accompanying pressure or stress P is the effort variable. Specific acoustic impedance is thus defined by:
Specific acoustic impedance
ZA =
P u
[16.31]
As with electrical impedance, ZA is, in general, a complex quantity of the form ZA = RA + jXA. However, for plane progressive waves, the imaginary component XA is zero,
440 ULTRASONIC MEASUREMENT SYSTEMS Table 16.2 Characteristic impedance of some common materials.
Material
Longitudinal velocity of sound cL m s−1
Density ρ kg m−3
Characteristic impedance RA = ρcL kg m−2 s−1
Steel Iron Brass Aluminium Bone Glass Biol. tissue Polystyrene Oil Water Air
6.0 × 103 5.90 × 103 3.50 × 103 6.40 × 103 ≈ 5.3 × 103 5.50 × 103 ≈ 1.5 × 103 2.35 × 103 1.40 × 103 1.50 × 103 0.34 × 103
7.80 × 103 7.90 × 103 8.60 × 103 2.70 × 103 ≈ 1.5 × 103 2.50 × 103 ≈ 1.0 × 103 1.10 × 103 0.90 × 103 1.0 × 103 1.3
4.7 × 107 4.7 × 107 3.0 × 107 1.7 × 107 ≈ 0.8 × 107 1.3 × 107 ≈ 0.15 × 107 0.25 × 107 0.13 × 107 0.15 × 107 430
Source: adapted from Blitz[3].
leaving ZA = RA. It can be shown that RA is the product of density ρ and velocity of sound c for the material, i.e. Characteristic impedance of a medium
RA = ρc
[16.32]
Table 16.2 (adapted from Blitz[3]) gives values of longitudinal velocity, density and characteristic impedance for commonly used materials. The power or intensity at any point in an acoustic field is the product Pu of through and across variables. For a plane progressive wave, we can use eqn [16.31] to give three equations for acoustic intensity: Acoustic intensity or power
W = Pu =
P2 = u 2RA RA
[16.33]
W is the rate of flow of acoustic energy, through unit area, at right angles to the direction of propagation of the wave, and has the units of W m−2.
16.3.3
Attenuation of a plane wave in a medium So far we have assumed that as a plane acoustic wave travels through a material, the amplitudes • and | of particle displacement and pressure variation respectively remain constant. In practice the wave becomes attenuated as it passes through the medium, so that the amplitudes decrease exponentially with distance y travelled. Thus for displacement amplitudes we have:
Displacement attenuation for a plane wave
•( y) = •(0) exp(−αd y)
[16.34]
16.3 PRINCIPLES OF ULTRASONIC TRANSMISSION
441
where •(0) is the amplitude at the beginning of the medium, •( y) is the amplitude at position y, and αd m−1 is the displacement attenuation coefficient. A similar equation relates the acoustic intensity W( y) at position y to the intensity W(0) at y = 0, i.e. Power attenuation for a plane wave
W( y) = W(0) exp(−αWy)
[16.35]
where αW m−1 is the power attenuation coefficient. Various mechanisms are responsible for this power loss. In liquids viscous friction effects are important; in solids scattering at grain boundaries and absorption at crystal lattice defects are the main loss mechanisms.
16.3.4
Reflection and refraction at a boundary between two materials Figure 16.9 shows a plane wave incident on a plane boundary separating two materials of different characteristic impedances. From the figure, we see that part of the incident wave I is reflected (R), and part is transmitted (T ). The laws relating the angles θI, θR and θT of incidence, reflection and transmission are those for any wave motion. Figure 16.9 takes no account of possible conversion of longitudinal waves into transverse waves, and vice versa, at the boundary. The ratio between intensity WR and incident intensity WI is termed the reflection coefficient αR; the ratio between transmitted intensity WT and incident intensity WI is the transmission coefficient αT , i.e.
αR = Figure 16.9 Reflection and refraction of sound at a boundary between two materials.
WR WT , αT = WI WI
[16.36]
442 ULTRASONIC MEASUREMENT SYSTEMS Table 16.3 Values of αR and αT at the boundaries of five common materials.
Steel Quartz Polystyrene Water Air
Steel
Quartz
Polystyrene
Water
Air
0.0, 1.0
0.27, 0.73 0.0, 1.0
0.81, 0.19 0.51, 0.49 0.0, 1.0
0.88, 0.12 0.67, 0.33 0.06, 0.94 0.0, 1.0
≈ 1.0, 3.7 × 10−5 ≈ 1.0, 1.1 × 10− 4 ≈ 1.0, 6.9 × 10− 4 ≈ 1.0, 1.1 × 10−3 0.0, 1.0
For normal or near-normal angles of incidence, i.e. θI ≈ θR ≈ θT ≈ 0, it can be shown that[4] Reflection and transmission coefficients for near-normal incidence
αR =
(R 2 − R1 )2 (R 2 + R1 )
2
, αT =
4R1R 2 (R 2 + R1 )2
[16.37]
We note that αR + αT = 1; this is a consequence of energy conservation, i.e. WR + WT = WI . Table 16.3 uses R values from Table 16.2 to give αR and αT values for boundaries of five important materials. We see that the αT values for quartz/steel and steel/water are reasonably large. This is of great practical value, because it means that sound waves can be launched from quartz crystal transmitters into solid metal bars, and also into liquids contained in metal pipes and vessels. The αR values for steel/air and water/air are very close to unity (α T very small); this means there is almost perfect reflection of sound waves at these boundaries. Detection of cracks and flaws in metal samples and at the liquid/gas interface in level measurement is therefore possible. We can now examine the efficiency of the simple transmission link of Section 16.1 and Figure 16.1. As an example, consider a quartz transmitter and receiver attached to opposite sides of a steel billet 10 cm thick. The ratio WR/X /WT/X of received power to transmitted power depends on attenuation losses in the steel, together with interface losses at the quartz/steel and steel/quartz boundaries. Using eqns [16.35] and [16.36] we have: WR/X = (αT)Q/S e−αW l(αT)S/Q WT/X
[16.38]
Since (αT)Q/S = (αT)S/Q = 0.73, l = 0.1 m, and assuming αW = 2.0 for steel, WR/X = (0.73)2 e−0.2 = 0.53 × 0.82 = 0.44 WT/X i.e. the efficiency of energy transmission is 44%. From Table 16.3 we see that the αT values for quartz/air and steel/air are very small. This means that it is very difficult to launch a sound wave from a piezoelectric transmitter into a gas, either directly or via a steel pipe. One way of overcoming this problem is to place a matching layer between the transmitter and the gas. If R1, R and R2 are the characteristic impedances of the transmitter, matching layer and gas respectively then:
16.3 PRINCIPLES OF ULTRASONIC TRANSMISSION
αT, M =
4R1R 4RR2 , αM,G = 2 (R1 + R) (R + R2)2
443
[16.39]
and the overall transmission coefficient between transmitter and gas is:
αT,G = αT, M × αM,G =
16R1R2R2 (R1 + R)2(R + R2)2
[16.40]
It can be shown (Problem 16.9) that this overall transmission coefficient is a maximum when: Optimum impedance of matching layer
R=
R1R 2
[16.41]
Thus in order to maximise power transfer between quartz (R1 = 1.5 × 107) and air (R2 = 430) a matching layer with R ≈ 8 × 104 is required. From Table 16.1, we see that RA ≈ 0.4 × 107 for PVDF, and from Table 16.2, RA ≈ 0.15 × 107 for biological tissue. Acoustic power can therefore be efficiently launched from PVDF transducers directly into the human body without the need for matching layers.
16.3.5
Stationary waves and resonance Consider a progressive plane wave travelling along the positive y direction through a medium with zero attenuation coefficient. If it meets the boundary with a medium of very different characteristic impedance then almost perfect reflection occurs. This means that there is now a reflected wave travelling along the negative y direction with equal amplitude to the incident wave. For longitudinal waves the incident wave is:   y d y = • y sin ω L t − cL  
   
and the reflected wave is:   y d y = • y sin ω L t + cL  
   
The resultant particle displacement is therefore:    y y   d yR = • y sin ω L  t −  + sin ω L  t +   cL  cL      Equation for stationary wave
ω y d yR = 2• y cos  L  sin(ω Lt )  cL 
[16.42]
The function 2•y cos(ωL y/cL) defines the wave amplitude or envelope. Since this is independent of t, i.e. does not move forward with time, the wave is stationary. At values of y where cos(ωL y/cL) = 0, the envelope has a minimum value of zero; these
444 ULTRASONIC MEASUREMENT SYSTEMS are nodes. At values of y where cos(ωL y/cL) = 1, the envelope has a maximum value of 2•y; these are antinodes. In practice the medium will have finite length l, and the type of stationary or standing wave set up will depend on the boundary conditions at each end of the medium, i.e. whether there are nodes or antinodes. In the special cases that the frequency of the sound wave is such that l is a whole number of either half or quarter wavelengths, then resonance is produced where •y has a large resonant value. Figure 16.10 Resonant stationary wave patterns: (a) Half wave resonance – node at each end (b) Half wave resonance – antinode at each end (c) Quarter wave resonance – node at one end, antinode at other end.
16.3 PRINCIPLES OF ULTRASONIC TRANSMISSION
445
Two types of resonance are possible. In half wave resonance (Figures 16.10(a) and (b)) there are identical boundary conditions at each end of the medium. Figure 16.10(a) shows possible resonant modes for a node at each end; these modes could occur if a medium of low characteristic impedance is bounded on both sides by a medium of high impedance, for example a gas inside a metal pipe. Figure 16.10(b) shows possible resonant modes for an antinode at each end; these modes could occur if a medium of high impedance is bounded on both sides by a medium of low impedance, for example a metal bar surrounded by gas. The other type of resonance is quarter wave resonance; here there are different boundary conditions, i.e. a node and antinode, at each end of the medium. Figure 16.10(c) shows the possible resonant modes; these modes could occur if a medium of intermediate impedance is bounded on one side by a material of high impedance and on the other by a material of low impedance. The principles discussed in this and previous sections are important in the design of efficient transmitters. As an example, consider the problem of launching a pulse of ultrasound from a piezoelectric transducer into a gas. There are three essential requirements: 1. 2. 3.
The sound wave launched from the rear face of the transducer should not enter the gas, otherwise pulse transit time measurements will be confused. The overall Q should be ideally around 0.7 to minimise pulse distortion. Maximum power should be transmitted from the front face of the transducer into the gas.
The transmitter shown in Figure 16.11 fulfils these requirements. The backing layer, with characteristic impedance similar to that of the transducer and high attenuation coefficient, accepts and absorbs the wave from the rear face. This layer also reduces the Q of the overall transmitter to a value well below that of the transducer itself. The matching layer has length l = λ /4, so that a quarter wave resonant mode is present, and impedance R = êR1úRú2ú . This ensures maximum power transfer from transducer front face to the gas.
Figure 16.11 Transmitter for launching sound pulses into a gas.
446 ULTRASONIC MEASUREMENT SYSTEMS
16.3.6
Doppler effect When a source (or transmitter) and observer (or receiver) of sound waves are in relative motion, then the frequency of the received signal differs from the frequency of the transmitted signal by an amount that depends on the relative velocities of source and observer. This shift in apparent frequency is called the Doppler effect and occurs in all types of wave motion. There are two cases to consider.[5] One is where the source is stationary and the observer is moving; the other is where the source is moving and the observer stationary. Figure 16.12(a) shows a stationary source of sound S with frequency f and a stationary observer O. If the disturbance takes time ∆t to travel from S to O, then O receives f∆t cycles in this time and the distance OS = f λ ∆t. If O moves towards S with velocity v, then in time ∆t it travels a distance v∆t to O′ (Figure 16.12(b)), and receives an additional v∆t/λ cycles of sound. The total number of cycles received in ∆t is thus ( f + v/λ)∆t and the apparent frequency is: f′=f+
A c + vD v f =f+ v=f C c F λ c
[16.43]
since 1/λ = f /c. If O moves away from S with velocity v, then ( f − v/λ)∆t cycles are received in time ∆t, so that the apparent frequency is: f′=f
A c − vD C c F
[16.44]
We can generalise these two results by the single equation: Doppler shift: fixed source, moving observer Figure 16.12 Doppler effect.
f ′ velocity of waves relative to observer = f normal wave velocity
[16.45]
16.4 EXAMPLES OF ULTRASONIC MEASUREMENT SYSTEMS
447
If S now moves towards O (stationary) with velocity v, then in time ∆t it travels a distance v∆t to S′ (Figure 16.12(c)). In this case f∆t cycles of sound occupy the distance S′O = ( f λ − v)∆t. The apparent wavelength is thus:
λ′ =
total distance ( f λ − v)∆t v = =λ− total number of cycles f∆t f
[16.46]
Since A c − vD 1 λ v = , λ′ = λ − λ = λ C c F f c c
[16.47]
If S moves away from O we have:
λ′ = λ
A c + vD C c F
[16.48]
and we can generalise these two results by the single equation: Doppler shift: moving source, fixed observer
16.4 16.4.1
λ ′ velocity of waves relative to source = λ normal wave velocity
[16.49]
Examples of ultrasonic measurement systems Pulse reflection or pulse echo systems Figure 16.13(a) shows a typical system. A piezoelectric crystal acting as a transmitter/ receiver is attached to medium 1; the characteristic impedances of media 1 and 2 are substantially different. When the crystal acts as a transmitter, an oscillator giving a sinusoidal voltage at radio frequency f is connected to the crystal for a time TW using switch A. Thus a pulse of ultrasound of width TW enters medium 1 and most of the pulse energy is reflected at the boundary of medium 1 with medium 2. The reflected pulse returns to the crystal at a time TT after the outgoing pulse. Since TT is the time for the ‘round trip’ of distance 2l then:
Transit time for pulse echo system
TT =
2l c
[16.50]
where l is the distance of the interface from the crystal and c is the velocity of sound in medium 1. The crystal now acts as a receiver and converts the reflected pulse into a voltage pulse. Switch B is now closed so that the pulse passes to the echo signal conditioning circuit where it is amplified, rectified and ‘squared up’ using a Schmitt trigger. Figure 16.13(b) shows the idealised outgoing and reflected pulse waveforms. These pulse waveforms pass to a logic circuit which detects the leading edge of the outgoing pulse and that of the first reflected pulse; the resulting output is a pulse signal
448 ULTRASONIC MEASUREMENT SYSTEMS
Figure 16.13 Pulse reflection techniques (a) Typical system (b) Idealised outgoing and reflected pulse waveforms.
16.4 EXAMPLES OF ULTRASONIC MEASUREMENT SYSTEMS
449
which is in the ‘1’ state during the transit time TT. This is used to control a counter which also receives clock pulses at frequency f from the signal generator. The total count N during the ‘on’ time is therefore N = f TT. This is transferred to a microcontroller using a parallel digital signal. The computer calculates TT from N and uses equation [16.50] with a known value of c to find l. The measurement is complicated by the creation of multiple reflections or ‘echoes’. Part of the first reflected pulse is reflected at the boundary of medium 1 and the crystal, and reflected again at the boundary of media 1 and 2 to give a second reflected pulse. The process is repeated many times, the amplitude of the reflected pulses dying away due to attenuation losses in medium 1 and reflection losses at the boundaries. Figure 16.13(b) shows these multiple reflected pulses. The following conditions should be obeyed by the pulse signal: (a)
The pulse width TW should be large compared with the period l/f of the sound wave; this ensures that there are many cycles, i.e. sufficient energy, in each pulse: TW 0002
(b)
1 f
The transit time TT should be large compared with the pulse width TW to avoid interference between outgoing and reflected pulses: TT 0002 TW
(c)
[16.51]
[16.52]
The repetition time TR between successive outgoing pulses should be large compared with transit time TT; this ensures that all reflections, following one outgoing pulse, are attenuated before the next enters the material: TR 0002 TT
[16.53]
Thus for a metal specimen with l = 0.2 m and c = 5 × 103 m s−1, TT = 80 µs. If f = 1 MHz, then the above conditions are satisfied with TW = 15 µs, TR = 1 ms. Because of the large difference in characteristic impedance between most solids and air, this method can be used for thickness measurement. Pulse reflection techniques are also commonly used for the detection of flaws in metals.[6] Here frequency f is chosen so that the sound wavelength is small compared with the size of defects it is desired to detect. The large difference in characteristic impedance between gases and liquids means that almost perfect reflection occurs at a liquid/gas interface and that these techniques are applicable to level measurement. Analysis of the interface losses (Problem 16.2) would suggest that it is better to mount the crystal at the base of the vessel, directing the waves upwards through the liquid, rather than mount the crystal at the top of the vessel, directing the waves downwards through the gas. However, most commercial ultrasonic level measurement systems[7–10] use the latter method, because it offers greater ease of installation and maintenance. Another important application of pulse reflection techniques is in the ‘imaging’ of areas of the human body.[11] Figure 16.14(a) shows, in simplified form, the various layers of tissue.[12] The characteristic impedance of these layers will be different: for example, the impedance of bone is around 0.8 × 107 whereas that of soft biological tissue is around 0.15 × 107 (Table 16.2). A piezoelectric transducer is placed on the epidermis layer, which has either a characteristic impedance close to that of soft tissue (e.g. PVDF) or a matching layer. This minimises internal reflections at the transducer/epidermis boundary and the problem of multiple echoes.
450 ULTRASONIC MEASUREMENT SYSTEMS Figure 16.14 Ultrasonic imaging of human body (adapted from Payne[12] ): (a) Layers and coordinate system (b) A-scan display (c) B-scan display.
Figure 16.14(b) shows the CRT trace obtained when the system of Figure 16.13 is used with the layer system of Figure 16.14(a). The three reflected pulses correspond to reflections at the epidermis/dermis, dermis/fat and fat/bone boundaries respectively; the time intervals between successive reflected pulses are proportional to the thickness of each layer. This trace, referred to as an A-scan display, is rather difficult to interpret; a more realistic image is obtained using a B-scan display. Here the transducer is connected to two displacement sensors which measure transducer x and y position coordinates on the body surface. The output voltage of the x sensor is applied to the X plates of the CRT and a voltage proportional to time, i.e. distance z travelled through the body, is applied to the Y plates. The brightness of the image on the screen is proportional to the transducer output voltage (Z modulation) so that a bright spot corresponds to a reflected pulse. By keeping the transducer y-coordinate fixed and adjusting the x-coordinate, an image of the body in the x–z plane is built up and stored (Figure 16.14(c)). Thus the B-scan images a ‘slice’ through the body, normal to the surface. Another alternative is the C-scan display; this is an image of the body in the x–y plane, i.e. a slice parallel to the surface of the body. This is obtained
16.4 EXAMPLES OF ULTRASONIC MEASUREMENT SYSTEMS
451
Figure 16.15 C-scan display of human foetus (by permission of Nottingham City Hospital NHS Trust and Sarah Morton).
by applying x sensor output voltage to the CRO X plates and y sensor output voltage to the Y plates and using Z modulation. Figure 16.15 shows a C-scan display of a human foetus.
16.4.2
Ultrasonic Doppler flowmeter This flowmeter uses the Doppler effect discussed in Section 16.3.6 and is shown in Figure 16.16. We see that the flowmeter is completely external to the pipe; it is thus suitable for the ‘difficult’ flowmetering problems discussed in Section 12.5 (i.e. corrosive fluids, slurries) and as a ‘clip-on’ flowmeter.[13] The transmitting crystal sends an ultrasonic wave of frequency f and velocity c into the fluid, at an angle θ relative to the direction of flow. Bubbles, solid particles or eddies in the flow stream can be regarded as ‘observers’, moving with velocity v relative to the fixed transmitter. The velocity of sound waves relative to the observer is c + v cos θ, so that from eqn [16.45] the apparent frequency f ′ seen by the observer is given by: f ′ c + v cos θ = f c
[16.54]
The particle scatters the incident sound wave in all directions, but a small proportion is back-scattered in the direction of the receiving crystal. The particle now acts as a ‘source’ moving with velocity v relative to the receiving crystal, acting as a fixed observer. The velocity of the sound waves relative to the source is c − v cos θ, so that from eqn [16.49] the apparent wavelength seen by the receiving crystal is given by:
λ″ c − v cos θ = λ′ c
[16.55]
Since λ″f ″ = λ′f ′ = c, the corresponding received frequency f ″ is: f″ c = f ′ c − v cos θ
[16.56]
452 ULTRASONIC MEASUREMENT SYSTEMS Figure 16.16 Doppler flowmeter – layout and signal processing.
Eliminating f ′ between [16.54] and [16.56] gives:  1 +  c + v cos θ  = f″ = f  f    c − v cos θ  1 −  
 v cos θ  c  v cos θ   c
[16.57]
Typically we have c ≈ 103 m s−1 and v ≈ 10 m s−1, i.e. v/c ≈ 10−2. Thus we can ignore terms involving second and higher powers of v/c. With this approximation [16.57] reduces to: Frequency shift in Doppler flowmeter
∆f = f ″ − f =
2f (cos θ )v c
[16.58]
Thus the frequency difference ∆f is proportional to fluid velocity v, and thus volume flowrate. If f = 1 MHz and θ = 30°, then ∆f = 17.3 kHz for the above values of c and v. Figure 16.16 shows a possible system for processing the electrical signal from the receiver crystal. Due to attenuation, interface and scattering losses, the receiver signal is at a low level, and is first amplified to the amplitude Ä of the transmitter drive signal. Both signals are input to an adder. The adder output signal is:
16.4 EXAMPLES OF ULTRASONIC MEASUREMENT SYSTEMS
453
VADD = Ä sin 2πf ″t + Ä sin 2π ft = 2Ä cos
2π( f ″ − f )t 2π( f + f ″)t sin 2 2
[16.59]
This is a sine wave of frequency ( f + f ″)/2 and amplitude 2Ä cos(2π∆ft/2), which is an amplitude modulated signal with a carrier frequency of ( f + f ″)/2 and a modulating frequency of ∆f /2. The signal is demodulated to give a final output sinusoidal signal whose frequency ∆f/2 is proportional to fluid velocity v.
16.4.3
Ultrasonic cross-correlation flowmeter The principle of the cross-correlation flowmeter has already been explained in Section 12.5. In general it consists of two sensors, at distance L apart, which detect random fluctuations in some property of the fluid. The sensor output signals x(t) (upstream) and y(t) (downsteam) are input to a correlator which evaluates the crosscorrelation function: Rxy (β ) = lim
T →∞
1 T
0002 x(t − β ) y(t) dt T
0
This has a maximum at β = L/v, when there is maximum similarity between the delayed upstream signal x(t − β ) and the downstream signal y(t). Figure 16.17 shows an ultrasonic cross-correlation flowmeter[14,15] consisting of two transmission links T/X(1)–R/X(1) and T/X(2)–R/X(2). The meter is completely external to the pipe and is especially suitable for two-phase flow measurements, that
Figure 16.17 Ultrasonic cross-correlation flowmeter.
454 ULTRASONIC MEASUREMENT SYSTEMS is liquids with suspended gas bubbles or solid particles. A bubble or particle has a different acoustic impedance from that of the liquid. This means that if one is present in either of the transmission links, there are two extra interfaces and corresponding reflection losses. The ultrasonic wave reaching the receiver is reduced in amplitude and changed in phase relative to the received wave with no particles present. The number of particles in either link varies randomly with time, causing the received ultrasonic waves to be randomly modulated, in both amplitude and phase. The receiver signals are first amplified and then demodulated to remove the high-frequency carrier signal. The demodulated signals, representing the random fluctuations in flow, are then amplified and input to the correlator. Each transmission link can be either energised by an external r.f. oscillator (solid line), or incorporated into a feedback oscillator system of the type described in Section 16.2 (dotted line).
16.4.4
Ultrasonic transit time flowmeter Figure 16.18 shows an ultrasonic transit time flowmeter of the ‘wetted sensor’ type where the piezoelectric transducers are in contact with the fluid. This is suitable for clean, single-phase liquids and gases. For dirty, hostile or multiphase flows, a ‘clamp on’ type, with the transducers mounted on the outside of the pipe, is preferred. Firstly, transducer B acts as a transmitter and sends a pulse of ultrasound to transducer A, acting as a receiver. The corresponding transit time from B to A is: TBA =
L c − v cos θ
[16.60]
where path length L = D/sin θ and c is the velocity of sound in the fluid. If then A acts as a transmitter and sends a pulse to B acting as a receiver, the corresponding transit time from A to B is: TAB =
Figure 16.18 Ultrasonic transit time flowmeter.
L c + v cos θ
[16.61]
REFERENCES
455
The differential transit time ∆T is therefore given by: ∆T = TBA − TAB = =
D D A L 1 − sin θ C c − v cos θ c + v cos θ F
2 D cot θ v   v2 c 1 − 2 cos 2θ  c  
[16.62]
2
As with the Doppler flowmeter the ratio v/c is typically 10−2 so that (v2/c2) cos2θ 0003 1 and eqn [16.62] reduces to: Differential transit time
∆T =
2 D cot θ c2
v
[16.63]
Thus differential transit time ∆T is proportional to fluid velocity v but is normally very small. For water in a 0.1 m diameter pipe, the transit time at θ = 45°, in the stationary fluid, is 95 µs. If the water velocity is 1 m s−1 the corresponding differential transit time is 89 ns. Thus to obtain a measurement error within 1% of reading, the maximum error in ∆T should be 0.9 ns. This high degree of precision can be obtained using a sing-around system where transducers A and B are continuously switched between transmitter and receiver modes.[16] Here the differential transit time is converted into a frequency difference ∆f = (sin 2θ /D)v which is independent of sound velocity c.
16.4.5
Ultrasonic vortex flowmeter This is described in Section 12.3.3.
Conclusion This chapter began by explaining the basic ultrasonic transmission link, which consists of a transmitter, a transmission medium and a receiver. The next section examined the detailed characteristics of ultrasonic transmitters and receivers based on piezoelectric crystals. The following section studied the principles of ultrasonic transmission including plane wave propagation, acoustic impedance and power, attenuation, reflection, refraction and stationary waves. The final section looked at examples of ultrasonic measurement systems including pulse reflection, Doppler, cross-correlation and transit time systems.
References [1] o’donnell m, busse l j and miller j g 1981 ‘Piezoelectric transducers’, Methods of Experimental Physics, vol. 19, Academic Press.
456 ULTRASONIC MEASUREMENT SYSTEMS [2] Massa Products Corporation 1983 Technical Information on Model Tr-89/B Series Transducers. [3] blitz j 1971 Ultrasonics: Methods and Applications, Butterworths, London. [4] blitz j 1967 Fundamentals of Ultrasonics, 2nd edn, Butterworths, London. [5] longhurst r s 1963 Geometrical and Physical Optics, Longman, London, pp. 112–14. [6] whittle m j 1979 ‘Advances in ultrasonic flaw detection’, CEGB Research, no. 10, November. [7] benson f w 1967 ‘Ultrasonics for liquid level control’, Measurement and Control, vol. 9, no. 11, November. [8] Robertshaw Controls Co., Industrial Instrumentation Division 1979 Publicity Material for Model 165 Ultrasonic Level System. [9] Bestobell Mobrey Ltd 1979 Instruction Manual for ‘Sensall’ Continuous Level Monitoring System. [10] Endress and Hauser (UK) Ltd, Technical Information, DU 210, DU 211 Ultrasonic Level Sensors. [11] crecraft d i 1983 ‘Ultrasonic instrumentation: principles, methods and applications’, J. Phys. E. Scientific Instruments, vol. 16, pp. 181–9. [12] payne p a 1985 ‘Medical and industrial applications of high resolution ultrasound’, J. Phys. E. Scientific Instruments, vol. 18, pp. 465–73. [13] Bestobell Mobrey Ltd, Publicity Material for Doppler Flowmeter. [14] medlock r s 1985 ‘Cross correlation flow measurement’, Measurement and Control, vol. 18, no. 8, pp. 293–7. [15] keech r p 1982 ‘The KPC multichannel correlation signal processor for velocity measurement’, Transactions of the Institute of Measurement and Control, vol. 4, no. 1, Jan.–Mar. [16] sanderson m l 1982 ‘Electromagnetic and ultrasonic flowmeters: their present states and future possibilities’, Electronics and Power, vol. 28, no. 2, pp. 161–4.
Problems 16.1
A piezoelectric crystal has an effective mass of 10−2 kg, stiffness of 1010 N m−1 and damping constant 200 Ns m−1. The electrical capacitance of the crystal is 1000 pF and the charge sensitivity is 2 × 10−10 C N−1. (a) (b) (c)
16.2
Calculate the series and parallel resonant frequencies of the crystal. Calculate the magnitude and phase of the overall electrical impedance of the crystal at the above frequencies. The crystal is incorporated into a closed-loop oscillator system which is to oscillate at the crystal series resonant frequency. Calculate the required gain and phase of the maintaining amplifier at this frequency.
Figure Prob. 2 shows a cross-section through a steel pipe through which sewage is flowing. The level of sewage in the pipe can also vary between the limits shown. Quartz crystals are attached to the outside of the pipe to form the four transmission links: T1–R1, T2–R2, T3–R3 and T4–R4. (a)
Assuming that only interface losses take place, calculate the fraction of received power to transmitted power for each link. Use the data given in Tables 16.1 and 16.2, and assume that sewage has the same transmission characteristics as water.
PROBLEMS
457
Figure Prob. 2.
(b)
Explain how these links can form the basis of systems to measure: (i) (ii)
16.3
the level of sewage; the velocity of sewage.
An open steel vessel contains liquid metal to a depth of about 0.75 m. It is proposed to measure the depth of liquid using ultrasonic pulse reflection techniques. A quartz crystal attached to the base of the vessel is to act alternatively as a transmitter and receiver. Using the data given below and in Tables 16.1 and 16.2: (a) (b)
Calculate the ‘round trip’ time TT and the fraction of received power to transmitted power. Choose suitable values for pulse width and repetition times.
Data Velocity of sound in liquid metal Density of liquid metal Power attenuation coefficient Natural frequency of quartz crystal 16.4
= 1.5 × 103 m s−1 = 5 × 103 kg m−3 = 0.1 m−1 = 1 MHz
An ultrasonic Doppler flowmeter is to be used to measure the volume flow rate of a slurry in a steel pipe of diameter 0.2 m. Two piezoelectric crystals, each having a natural frequency of 1 MHz, are positioned, a few millimetres apart, on the outside of the pipe to form an ultrasonic transmission link. The transmitting crystal directs an ultrasonic beam into the pipe so that the beam is moving in an opposite direction to the flowstream. The angle between the ultrasonic beam and the direction of flow is 60°. On average 10% of the ultrasonic power reaching each solid particle is scattered back in the direction of the receiving crystal. Assume that the slurry has the same density and sound velocity as water (Table 16.2) and a power attenuation coefficient of 1.0 m−1. (a) (b)
Find the difference between the frequencies of the transmitted and received beams when the flow rate is 1.13 × 103 m3 h−1. Estimate the ultrasonic power incident on the receiving crystal for each watt of ultrasonic power leaving the transmitting crystal. State any assumptions made in your calculation.
458 ULTRASONIC MEASUREMENT SYSTEMS
16.5
An ultrasonic transmitter is in the form of a piezoelectric disc of diameter 2.5 cm and thickness 1.0 cm. The front face of the disc is placed directly onto biological tissue; the rear face is in contact with air. A pulse launched from the centre of the disc divides into two equal pulses, each of power 1 W and width 0.5 µs, one travelling towards the front face and one towards the rear face. (a) (b)
Use the data given below to derive the form of the signal entering the tissue. Explain in detail what modifications should be made to the transducer so that the signal entering the tissue is a close approximation to a single pulse.
Data
16.6
Material
Velocity of sound (m s−1)
Density (kg m−3)
Piezoelectric Tissue Air
5 × 103 1.3 × 103 0.3 × 103
2.6 × 103 1.0 × 103 1.33
Figure Prob. 6 shows transmitting and receiving quartz crystals attached to a composite metal slab. The composite slab consists of a slab of metal A bonded to a slab of metal B; metal B contains an air-filled crack of negligible thickness in the position shown. An ultrasonic pulse of mean power 1 W and duration 5 µs is launched from the transmitting crystal into the slab. Use the data given below to discuss the main features of the signal at the receiving crystal. Estimates of times and power levels should be made.
Figure Prob. 6.
Data
16.7
Material
Velocity of sound (m s−1)
Density (kg m−3)
Power attenuation coefficient (m−1)
Metal A Metal B Quartz Air
6.0 × 103 6.0 × 103 6.0 × 103 0.30 × 103
3.0 × 103 8.0 × 103 2.5 × 103 1.33
2.0 4.0 – –
A differential transit time ultrasonic flowmeter is to measure gas velocities between 0.3 and 24 m s−1 to an accuracy within ±1%. Use the data given below to critically assess the feasibility of the flowmeter according to the following criteria.
PROBLEMS
(a) (b)
459
Ratio of received pulse power to transmitted pulse power. Explain how this ratio may be improved. Accuracy of pulse transit time measurement necessary to meet the above specification.
Data Material
Velocity of sound (m s−1)
Density (kg m−3)
Quartz Natural gas
5700 500
2600 50
Pipe diameter D = 300 mm, flowmeter length l = 600 mm. 16.8
An organic liquid containing suspended plastic pellets is flowing in a steel pipe of internal diameter 0.2 m and thickness 0.01 m. An ultrasonic transmission link is set up across the pipe; this consists of identical transmitting and receiving quartz discs as shown in Figure Prob. 8. The plastic pellets can be regarded as discs of thickness 0.01 m and diameter 0.02 m, moving parallel to the direction of flow.
Figure Prob. 8.
Using the data given in the table below, calculate the power incident on the receiver for each watt of transmitter power in the following cases: (a) (b) (c)
16.9
no pellets present in the transmission link; a single pellet completely intersecting the transmission link; n pellets completely intersecting the transmission link, all pellets being separated by the liquid phase.
Material
Velocity of sound (m s−1)
Density (kg m−3)
Power attenuation coefficient (m−1)
Steel Quartz Plastic Liquid
6.0 × 103 5.7 × 103 2.4 × 103 1.5 × 103
7.8 × 103 2.6 × 103 1.1 × 103 1.0 × 103
4.0 – 2.0 1.0
The overall transmission coefficient αT,G between a transmitter of characteristic impedance R1 and a gas of impedance R2 separated by a matching layer of impedance R is given by eqn [16.40]. By differentiating [16.40] and setting the derivative equal to zero (or any other method), show that αT,G is a maximum when R = êR1úRú2ú .
460 ULTRASONIC MEASUREMENT SYSTEMS
16.10
(a)
Explain briefly the meaning of the following terms used in ultrasonics: (i) (ii) (iii)
(b)
Attenuation Matching layer Resonant stationary wave.
An ultrasonic transmitter is in the form of a piezoelectric disc of diameter 1.0 cm and thickness 2.0 mm. The transmitter is to be used to launch 1 MHz pulses into biological tissue. Use the concepts of part (a) and the data below to detail what modifications are made to the transmitter so it meets the following specification: (i) (ii)
Maximum power for pulse transmission from the front face into the tissue Minimum power for pulse transmission from the rear face into the tissue.
Data Material
Velocity of sound (m s−1)
Density, (kg m−3)
Power attenuation coefficient (m−1)
Piezoelectric Tissue Air
5 × 103 1.3 × 103 0.3 × 103
2.6 × 103 1.0 × 103 1.33
5 × 102 – –
17 Gas Chromatography In many industrial processes, the materials passing between different stages of the process are mixtures of compounds or elements. These mixtures are often a single phase, i.e. gas, liquid or solid, but may be multiphase, e.g. gas and liquid. In order to operate many processes at maximum throughput or efficiency, it is essential to have some information on the composition of important process streams. In some cases a detailed analysis giving the fraction of each component in the mixture may be required; in others it is sufficient to know the percentage of a single key component. Important industrial examples are:
• • • •
Complete analysis of the hydrocarbon gas stream leaving a naphtha cracking furnace Percentage of carbon monoxide in boiler flue gas Acidity/alkalinity of boiler feed water Concentration of dissolved metal ions in industrial effluent discharged into rivers.
The measurement of composition is usually referred to as analysis and the corresponding systems as analytical measurement systems. The study of analytical techniques and systems is a wide subject which warrants a complete book in itself; the interested reader is referred to References 1–3. This chapter discusses the widely used technique of gas chromatography and the systems based on this technique. The use of katharometer and non-dispersive infrared systems for analysing gases has already been discussed in Chapters 14 and 15. The use of electrochemical sensors for measuring the concentration of ions in solution has been discussed in Section 8.9.
17.1
Principles and basic theory Gas chromatography is a technique for measuring the composition of a gas or volatile liquid sample by separating the sample into its constituent components. The separation is performed using two phases: a moving gas phase, and a stationary phase which is either liquid or solid. In gas–liquid chromatography the stationary phase is a thin layer of non-volatile liquid, coated onto solid particles acting only as a support. In gas–solid chromatography the solid particles themselves provide the stationary phase. A typical column for gas–liquid chromatography (GLC) is shown in Figure 17.1. The moving gas phase is called the carrier gas; this is an inert gas such as helium,
462 G A S CHROMATOGRAPHY Figure 17.1 Typical GLC column and gas flow rates in gas–liquid chromatography.
argon or nitrogen which does not react with the sample and is not dissolved by the liquid phase. A known volume of sample gas is injected into the carrier at time t0. This causes a pulse increase in the flow rate of gas into the column. The column is a long coiled tube, typically 2–3 m long and of 3 mm internal diameter, packed with small particles of solid ceramic material such as ground firebrick. The particles are coated with the liquid phase to give a large surface area for dissolving the gas. The liquid should not react with any of the components, and should have negligible vapour pressure at the column temperature. Commonly used liquids are silicones and polyglycols. The carrier gas ‘sweeps’ the sample through the packing and brings it into intimate contact with the liquid phase. A dynamic equilibrium is set up, with molecules passing from gas to liquid and from liquid to gas at equal rates. However, molecules of different components spend, on average, different amounts of time in the liquid phase. When molecules are dissolved in the liquid phase they have zero velocity down the column. The overall result is that the different components of the sample emerge or ‘elute’ from the end of the column at different times.
17.1 PRINCIPLES AND BASIC THEORY
463
Figure 17.1 shows the variation in the flow rate of gas leaving the column with time; this is called a chromatogram. Component A spends very little time in the liquid phase and elutes from the column first (at time t1) whereas component C spends a lot longer in the liquid phase and is last to emerge from the column (time t3). The average time for the ith component of the sample mixture to travel through the column is termed the retention time Ti. This is the sum of the time TMi spent in the moving gas phase and the time TSi spent in the stationary phase, i.e. Ti = TMi + TSi
[17.1]
The time TMi depends only on the length L of the column and the average carrier gas velocity ó; this is the same for all components, i.e. L ó
TMi =
for all i
[17.2]
Time TSi is usually different for each component and depends on the distribution ratio Ki of the component between the two phases, i.e. TSi TMi
Ki =
[17.3]
From [17.1]–[17.3] we have Retention time for ith component
Ti =
L ó
( Ki + 1)
[17.4]
Figure 17.2 is an idealised distance/time graph for four components, with Ki = 0, 1, 2 and 3, travelling down the column. The component with Ki = 0 spends no time in the liquid phase and the graph is a straight line of slope ó. The components with Ki = 1, 2 and 3 spend increasing time in the liquid phase and the corresponding graphs are straight lines of slope –12 ó, –13 ó and –14 ó respectively. The average carrier gas velocity ó can be found using[4] ó=
3(P 2 − 1) ó0 2(P 3 − 1)
[17.5]
where ó0 is the velocity at the column outlet, and P = PI /PO is the ratio of column inlet and outlet pressures. The molecules of the ith component do not all elute at the same time; there is a statistical distribution of times (usually normal) about the mean value Ti. The width of the distribution depends on the width of the original injection pulse, as well as molecular diffusion effects inside the column, and can be described by a standard deviation σi. An alternative measure is the base width ∆ti defined by the points of intersection of the base line and tangents drawn to the distribution at the half-peak height values (Figure 17.2). For a normal distribution ∆ti = 4σi. The resolution Rij of two adjoining components i, j is defined by: Resolution of adjacent components
Rij =
2(T j − Ti ) ∆ ti + ∆ t j
[17.6]
464 G A S CHROMATOGRAPHY Figure 17.2 Retention times and peak widths.
A resolution of 1.5 for two identical normal peaks means that Tj – Ti is equal to six standard deviations and that the peaks are almost completely separated. A commonly used measure of the separating efficiency of a column is height equivalent to a theoretical plate (HETP). This quantity arises because of the similarity between packed chromatograph columns and packed distillation columns, and the need to find the height of a packed column equivalent to one theoretical stage in a plate column. To calculate HETP for a given column of length L, we first calculate the number of theoretical plates N in the column, which is defined by A Ti D N = 16 C ∆ti F
2
[17.7]
and then HETP =
L N
[17.8]
17.2 TYPICAL GAS CHROMATOGRAPH
465
Figure 17.3 Relation between HETP and carrier gas velocity ó.
For a typical column we may have L = 1 m, Ti = 120 s and ∆ti = 10 s, giving N = 2304 and HETP ≈ 0.44 mm. HETP is not a constant for a given column, but depends on the mean carrier gas velocity ó and column temperature T °C. The relationship between HETP and ó is given by the Van Deemter equation: HETP = A +
B + Có ó
[17.9]
where A, B and C are constants for a given system at a given temperature. The individual terms represent diffusion and mass transfer effects and are plotted in Figure 17.3. We see that there is an optimum carrier gas velocity óOPT at which HETP is a minimum, i.e. the best separation of components is obtained.
17.2
Typical gas chromatograph Figure 17.4 shows a typical gas chromatograph system for composition measurement. Because column HETP depends on carrier gas velocity and column temperature, it is essential to control the flow rate of carrier gas and to place the sample injection valve, column and detector in a temperature-controlled enclosure. The sample injection valve is used to inject the sample into the carrier gas stream. Figure 17.5 shows the principle of operation of a sliding plate valve used with a gas sample. The valve consists of a plastic plate with grooved channels, sandwiched between two metal blocks. Prior to injection the valve is in position A such that the carrier gas passes directly to the column and the sample gas passes through the sample loop before being vented. This ensures that the sample loop is filled with sample gas; typically the sample loop has a volume of a few millilitres. On receiving an ‘injection’ signal from the computer the plate moves across to position B. The carrier gas now passes through the sample loop and flushes the sample volume into the column. The sample gas stream passes directly to vent. The valve is then returned to position A ready for the next injection. The operation is similar for a liquid sample, except that the volume of the sample is much smaller; this vaporises on entering the carrier gas stream.
466 G A S CHROMATOGRAPHY
Figure 17.4 Typical gas chromatograph system for composition measurement.
Figure 17.5 Injection of a gas sample with a sliding plate valve (after Pine[5]).
17.2 TYPICAL GAS CHROMATOGRAPH
467
The construction of a typical GLC column has already been described. A column for gas–solid chromatography (GSC) is similar, except that the solid particles themselves provide the active stationary phase rather than just acting as a support. Materials such as charcoal, silica gel, activated alumina and molecular sieve can adsorb gas molecules onto their surface. Many chromatograph systems incorporate more than one column, enabling sample and gas streams to be switched between columns. The purpose of the detector is to give a voltage or current output, depending on either concentration or flow rate, for each of the component gases eluting from the column. An ideal detector will give an output proportional to either component concentration or component flow rate.[1] The detector sensitivity (e.g. current/flow rate) should be high and the same for all components. The detector should also have good repeatability and fast speed of response. The katharometer (Section 14.4) has medium sensitivity, and is widely used as a detector in gas chromatograph systems for measuring compositions from several per cent down to around 1000 p.p.m. It consists of four heated, matched metal filaments arranged in a deflection bridge energised by a constant current 2i0. Pure carrier gas is passed over two of the filaments (Figure 14.7), and the gas eluting from the column over the other two. The bridge is balanced when pure carrier gas leaves the column, but for each eluting component there is an output voltage given approximately by: V KATH =D i D=
(kc − ki)xi kc[xi ki + (1 − xi)kc]
i 30 R2TF α d 0.24A
[14.52] [14.50]
where kc = thermal conductivity of carrier, ki = thermal conductivity of the eluting component, and xi = molar concentration of component i in the carrier + component mixture. From [14.52] we see that the carrier thermal conductivity must be substantially different from all the component conductivities. For small xi, V KATH ≈D i
A kc − ki D x C k 2c F i
i.e. there is an approximately linear relation between V KATH and xi. However, sensiti ivity depends on ki, i.e. sensitivity is different for different components. The flame ionisation detector has a high sensitivity for organic molecules and is used in the measurement of low compositions down to a few p.p.m. Here the gas eluting from the column is passed to a controlled oxy-hydrogen flame. The flame causes the organic molecules to be split up into charged ions, which flow to the collector electrode under the action of an applied d.c. voltage. This ionisation current is proportional to component flow rate, i.e. the number of molecules per second of component eluting from the column. The detector output signal is at a low level (see Problem 14.5, for example) and requires amplification before being input to the A/D converter. This typically (Chapter 10) accepts an input signal of 0 to +5.0 V and converts it into an 8-, 12- or 16-bit binary parallel digital signal, which is passed to the microcontroller for further processing. The microcontroller performs two main functions, which are explained more fully in the following section. These are as follows:
468 G A S CHROMATOGRAPHY
• •
17.3
Control of sequence of operations. Several operations must be performed on the system of Figure 17.4, including sample injection, column switching, control of A/D converter, signal processing and control of data presentation elements. These must be performed in the correct order and at the correct times. Signal processing. The time variation of converter input voltage is closely related to the flow rate of gas leaving the column as a series of peaks. The computer calculates the area under each peak, which is approximately proportional to the total mass of the corresponding component. The composition of the sample mixture can then be found. These data can then be presented to the observer, possibly as a complete analysis of the sample on the CRT display or as a record showing the trend of a key component.
Signal processing and operations sequencing The input voltage VIN to the A/D converter will be proportional to the detector output signal. For a flame ionisation detector, the detector output current is proportional to the number of molecules Ñi of the ith component, eluting from the column per second. In this case the converter input voltage is: V iIN = GÑi
[17.10]
where G is a constant, ideally the same for all components. Integrating [17.10], we have: ni =
1 G
0002V
IN i
1 Ai G
dt =
[17.11]
which shows that the total number of molecules ni of the ith component in the sample is proportional to the integral of converter input voltage, i.e. to the area Ai under the ith peak. Thus the total number M of molecules present in a sample of m components is given by: m
M=
1
m
∑n = G ∑A i
i 1
[17.12]
i
i 1
i.e. M is proportional to the sum of the areas under all peaks. The percentage molar concentration ci of the ith component in the sample is therefore given by: Calculation of concentrations in ideal case
ci =
ni M
× 100% =
Ai m
× 100%
∑ Ai
[17.13]
i =1
Thus in the ideal case, the composition of the sample can be found by evaluating the area under each component peak. The output voltage of the katharometer, however, is different for different components, and is approximately proportional to the concentration xi of the component in the component/carrier mixture. If Ñc is the number of molecules of carrier leaving the column per second, then: xi =
Ñi Ñi ≈ Ñc + Ñi Ñc
for small xi
[17.14]
17.3 SIGNAL PROCESSING AND OPERATIONS SEQUENCING
469
Thus for dilute mixtures, xi is proportional to Ñi provided that Ñc is a constant, so that there is tight control of carrier gas flow rate. The A/D converter input voltage for a katharometer detector is therefore given approximately by: V iIN ≈ Gi Ñi
[17.15]
where Gi is the sensitivity for the ith component, which depends on thermal conductivity ki. In this case the corresponding expression for the concentration of the ith component in the original sample is: Calculation of concentrations in practical case
ci =
(1/Gi ) Ai m
1 ∑ G Ai i =1 i
× 100% =
A*i m
∑ A*
× 100%
[17.16]
i
i =1
where the A*i are the corrected peak areas A*i = (1/Gi )Ai. The input to the computer is a sampled, coded version of the A/D converter input signal VIN. The converter has, typically, a conversion time of a few milliseconds, so that samples can be read by the computer at intervals of a few milliseconds. These sampled values may be affected by high-frequency interference or noise, and should be filtered. One simple filtering method is to calculate the average value of a group of several samples. Thus if the A/D conversion time is 5 ms an average of 40 samples is available every 0.2 s. The computer uses these filtered values to estimate the area of the peak by dividing it into a series of rectangular elements, as shown in Figure 17.6. If yj is the jth filtered value and ∆T the interval between values, then the area of the jth rectangle is yj ∆T, giving: Approximate expression for peak area
A≈
P
∑ y ∆T j
j =1
Figure 17.6 Calculation of peak area.
P
= ∆T ∑ yj j =1
[17.17]
470 G A S CHROMATOGRAPHY Figure 17.7 Flowsheet of program for evaluating corrected peak area.
where P is the number of filtered values used in the calculation. The above expression is accurate if ∆T is small compared with the peak standard deviation σ. Thus if σ = 2.5 s, a ∆T of 0.2 s is suitable, giving P = 75 for the complete peak of width 6σ. Figure 17.7 shows a flowsheet of a simple program for evaluating the area of a given peak. After initialisation, the computer reads the current filtered value yj and then evaluates the corresponding slope or gradient sj. This is given by: sj =
yj − yj−1 ∆T
[17.18]
17.3 SIGNAL PROCESSING AND OPERATIONS SEQUENCING
471
where yj−1 is the previous filtered value and ∆T the interval between values. This slope value is used to decide whether or not a peak is eluting from the column. The baseline will not always be constant with time but may drift either upwards or downwards. If the maximum slope of the baseline is +δ1 or −δ1, then the condition for a peak is: sj > +δ1
Peak:
or
sj < −δ1
[17.19]
and the condition for baseline is: Baseline: sj ≤ +δ1
or
sj ≥ −δ1
[17.20]
The integration, i.e. summation of yj values, proceeds only while the peak condition is satisfied. However, integration must also be carried out around the top of the peak where the slope is small, which means that the baseline condition rather than the peak condition is obeyed. At the baseline there is only a small difference between current and previous slopes sj and sj−1, whereas at the top of a peak the difference between these slopes is large. Thus we can differentiate between baseline and peak top using the conditions: sj ≤ +δ1
or
sj ≥ −δ1
and | sj − sj−1 | ≤ δ 2
[17.21]
Peak top: sj ≤ +δ1
or
sj ≥ −δ1
and | sj − sj−1 | > δ 2
[17.22]
Baseline:
The computer counts the total number of times I + J + K that conditions [17.19] and [17.22] are satisfied; this gives the total number of values P used in the calculation. The computer finds the first value to satisfy the condition sj > δ1, which is the first value used in the integration and is designated y1. This gives the position A of the baseline at the start of the peak. The position B of the baseline at the end at the peak is the last value to satisfy the condition sj < −δ1, that is the last value yP used in the integration. The baseline is assumed to follow the straight line AB, and the peak area Ai is calculated by subtracting the area of trapezium ABCD from eqn [17.17], i.e. P
Ai = ∆T
A P − 1D ( y1 + yP) 2 F
∑ y − ∆T C i
j 1
[17.23]
As mentioned earlier, the detector sensitivity may be different for different components i. The corrected area A*i is evaluated using A*i = Ai /Gi, where Gi is the sensitivity for the ith component. The Gi values can be found experimentally by injecting a standard mixture of known composition and comparing the areas of the corresponding voltage peaks. The computer uses eqn [17.16] to calculate component concentrations ci from the A*. i Figure 17.8(a) shows the sequence of operations for a simple system involving one column and one carrier stream. There are, however, many analyses which require multiple column and carrier operation.[5] One example is the measurement of small concentrations of benzene, ethyl benzene and paraxylene in a mixture which is mainly toluene. The resulting chromatogram, for a single column, is shown in Figure 17.8(b). We see that the peaks are not clearly resolved; the small ethyl benzene and paraxylene peaks sit high on the ‘tail’ of the large toluene peak. The problem is solved using the technique of heartcutting, which involves switching gas streams between
472 G A S CHROMATOGRAPHY Figure 17.8 Sequences of operations in gas chromatography: (a) Normal sequence (b) Chromatogram without ‘heartcutting’ (c) Chromatogram and sequence of operations with ‘heartcutting’.
two columns, as shown in Figure 17.9. Two sliding plate valves are used, one for sample injection (V1) and one for column switching (V2). The sample is injected into column 1 using V1, in the normal way. Valve V2 is initially in position C, which enables the benzene peak to pass from column 1 to column 2. As the toluene peak is about to elute from column 1, V2 is switched to position D, so that the gas leaving column 1 is vented and carrier gas 2 is passed through column 2. When most of the toluene has been vented, V2 is returned to position C. Thus the remainder of the toluene peak, together with the associated ethyl benzene and paraxylene peaks, enters column 2. These peaks can then be effectively separated in column 2 without having to use very long retention times. The resulting chromatogram is shown in Figure 17.8(c), together with the operations sequence.
REFERENCES
Figure 17.9 Heartcutting using two sliding plate valves.
473
Another switching technique involving two columns and two carrier supplies is called backflushing. This technique allows components of interest to pass through columns 1 and 2 to the detector. Components not required for the analysis, or which may harm the detector, are prevented from entering column 2 and ‘flushed’ to vent by carrier 2 flowing through column 1 in a reverse direction.
Conclusions The chapter first discussed the principles of gas chromatography and a typical chromatograph for measuring the compositions of gas mixtures. Associated signal processing and sequencing operations were then studied.
References [1] [2] [3] [4]
[5]
ewing g w 1975 Instrumental Methods of Chemical Analysis, 4th edn, McGraw-Hill, New York. strobel h a 1973 Chemical Instrumentation: A Systematic Approach, 2nd edn, Addison Wesley, New York. willard h h, merritt l l and dean j a 1974 Instrumental Methods of Analysis, 5th edn, Van Nostrand, New York. ayers b o, lloyd r j and deford d d 1961 ‘Principles of high speed gas chromatography with packed columns’, Analytical Chemistry, vol. 33, no. 8, July, pp. 986–91. pine c s f 1967 ‘Process gas chromatography’, Talanta, vol. 14, pp. 269–97.
474 G A S CHROMATOGRAPHY
Problems 17.1
A sample containing oxygen and nitrogen is injected into a helium carrier at time t = 0. The sample is swept through a column 1.0 m long packed with molecular sieve. The eluting components are detected by a katharometer detector which has an equal sensitivity for oxygen and nitrogen. The time variation in katharometer output voltage is shown in Figure Prob. 1. (a) (b) (c) (d)
Figure Prob. 1.
Assuming that the distribution ratio K for oxygen is 2.0, estimate the mean carrier velocity ó and K for nitrogen. Estimate base width ∆t for both peaks and hence find the resolution R. Estimate the number of theoretical plates N and HETP. Estimate the percentage composition of the sample (assume peaks are approximately triangular).
18 Data Acquisition and Communication Systems All of the measurement systems discussed so far have presented the measured value of a single variable to an observer; i.e. the systems were single input/single output. However, there are many applications where it is necessary to know, simultaneously, the measured values of several variables associated with a particular process, machine or situation. Examples are measurements of flow rates, levels, pressures and compositions in a distillation column, temperature measurements at different points in a nuclear reactor core, and components of velocity and acceleration for an aircraft. It would be extremely uneconomic to have several completely independent systems, and a single multi-input/multi-output data acquisition system is used. Here several elements are ‘time shared’ amongst the different measured variable inputs. This technique of time division multiplexing is discussed in the first section of this chapter, and a typical data acquisition system is described in the following section. The oil, water and gas industries are characterised by complex distribution systems involving the transfer of fluids by long pipelines from producing to consuming areas. Similarly, an electricity distribution system involves the transfer of electrical power from power stations to consumers, via a network of high voltage cables. These systems also include several items of equipment or plant, e.g. pumping stations, compressors, storage tanks and transformers, each with associated measured variables. These plant items are often located several miles from each other, in remote areas. It is essential for the effective supervision of these distribution systems that all relevant network measurement data are transmitted to a central control point. To do this a complex communications system is required. This usually consists of a master station (at the central control point) and several outstations (at the plant items). The system must be capable of transmitting large amounts of information in two directions (M/S to O/S and O/S to M/S), over long distances, in the presence of interference and noise. This chapter discusses the principles of parallel digital signalling, serial digital signalling, error detection/correction and frequency shift keying, which are used in communications systems, and concludes by describing the implementation of communications systems for measurement data with special regard to the Fieldbus standard.
476 DATA ACQUISITION AND COMMUNICATION SYSTEMS
Figure 18.1 Time division multiplexing.
18.1
Time division multiplexing Figure 18.1 shows a simple schematic diagram of a time division multiplexer, with four channels labelled 0, 1, 2 and 3. The input signal at each channel is a continuous voltage corresponding to a measured variable. The multiplexer also requires a two-bit parallel channel address signal to specify which input signal is connected to the output line. Thus if the binary address signal is 10, the switch in channel 2 is closed and input 2 is connected momentarily to the output line. The multiplexer output signal is thus a series of samples (Chapter 10) taken from different measurement signals at different times. In sequential addressing the channels are addressed in order, i.e. first 0, followed by 1, then 2 and 3, returning to channel 0 and repeating, so that the pattern of samples for the multiplexed signal is as shown in the diagram. Random addressing, whereby an observer selects a channel of interest at random, is also possible. If ∆T is the sampling interval, i.e. the time interval between samples of a given input, e.g. 0 or 1, then the corresponding sampling frequency fS = 1/∆T must satisfy the conditions for the Nyquist sampling theorem (eqn [10.1]). These require that fS be greater than or equal to 2fMAX, where fMAX is the highest significant frequency present in the power spectral density of the measurement signal. In Figure 18.1 four samples occur during ∆T, so that the number of samples per second for the multiplexed signal is 4fS. In general, for m signals, each sampled fS times per second, the number of samples per second for the multiplexed signal is:
Sample rate for m multiplexed signals
f SM = mfS
[18.1]
18.2 TYPICAL DATA ACQUISITION SYSTEM
477
Different measured variables may have frequency spectra with different maximum frequencies: thus the power spectrum of a flow measurement may extend up to 1 Hz, but that of a temperature measurement only up to 0.01 Hz. The sampling frequency of the flow measurement must therefore be 100 times that of the temperature measurement. In the multiplexed signal there will be 100 samples of the flow measurement between each temperature sample. The multiplexed signal is normally fed to a sampleand-hold device (Section 10.1). Figure 18.1 shows the sample-and-hold waveform.
18.2
Figure 18.2 Typical microcontroller based data acquisition system.
Typical data acquisition system Figure 18.2 shows a typical microcontroller-based data acquisition system.[1,2] The signal conditioning elements are necessary to convert sensor outputs to a common signal range, typically 0 to 5 V; Table 18.1 gives sensing and signal conditioning elements for different measured variables. The voltage signals are input to a 16channel time division multiplexer, and the multiplexed signal is passed to a single sample/hold device and analogue-to-digital converter (Section 10.1). In cases where all the sensors are of an identical type, for example 16 thermocouples, it is more economical to multiplex the sensor output signals. Here the multiplexed sensor signal is input to a single signal conditioning element, such as the reference junction circuit and instrumentation amplifier, before passing to the sample/hold and ADC. The ADC gives a parallel digital output signal which passes to one of the parallel input interfaces of the microcontroller. Another parallel input/output (I/O) interface provides the address and control signals necessary for the control of multiplexer, sample/hold and ADC. These are a four-bit multiplexer address signal, a sample/ hold control signal, an initiate conversion signal to the ADC, and a data valid signal from the ADC. The microcontroller performs whatever calculations (on the input data)
478 DATA ACQUISITION AND COMMUNICATION SYSTEMS Table 18.1 Typical measured variables, sensing and signal conditioning elements. Measured variable
Sensing element(s)
Signal conditioning elements
Temperature Temperature Flow rate
Thermocouple Platinum resistance detector Orifice plate
Weight Level
Strain gauge load cell Electronic D/P transmitter (4–20 mA)
Angular velocity Linear displacement Pressure
Variable reluctance tachogenerator Linear variable differential transformer (LVDT) Diaphragm + capacitance displacement sensor Piezoelectric crystal
Reference junction circuit + instrumentation amplifier Deflection bridge + instrumentation amplifier Electronic D/P transmitter (4–20 mA) + current-tovoltage converter Deflection bridge + instrumentation amplifier Current (e.g. 4 to 20 mA) to voltage (e.g. 0 to 5 V) converter Frequency-to-voltage converter A.C. amplifier + phase-sensitive demodulator + low pass filter A.C. bridge + a.c. amplifier + phase sensitive demodulator + LPF Charge amplifier
Acceleration
are necessary to establish the measured value of the variable. A common example is the solution of the non-linear equation relating thermocouple e.m.f. and temperature (Section 10.4). The computer converts the measured value from hexadecimal into binarycoded decimal form (Section 10.3). This b.c.d. data is written into a computer parallel output interface. Each decade is then separately converted into seven-segment code and presented to the observer using a seven-segment LCD display (Section 11.4). The computer also converts each decade of the b.c.d. to ASCII form (Section 10.4). The resulting ASCII code is then written into a serial and/or parallel output interface. These can transmit ASCII data in serial and/or parallel form to remote data representation devices such as a monitor, printer or host computer.
18.3
Parallel digital signals[3] Parallel digital signals were introduced in Section 10.1; one path is required for each data bit and all the bits are transmitted at the same time. Therefore, if eight data bits (one byte) are to be transmitted there are eight paths in parallel, the voltage on each path being typically 5 V for a 1 and 0 V for a 0. The total collection of parallel paths is called a data bus or data highway and is similar to the internal data bus in a computer. Since, however, an internal computer bus can only handle low power levels, it must be connected to an external data highway via a buffered interface. One commonly used parallel data highway conforms to the IEE 488/IEC 625 standard. This is a bit-parallel, byte-serial transmission system capable of a maximum transmission rate of 1 Mbyte s−1 up to a maximum transmission distance of 15 m. The standard is intended for high-speed, short-distance communication in a laboratory-type environment, where there is relatively low electrical interference. The bus comprises 16 lines: eight lines are used for data (usually 7-bit ASCII + parity check bit), three for ‘handshaking’ (see following section) and five for bus activity control. Up to 15 devices can be connected onto the bus. Each device must be able to perform at least one of the following three functions:
18.4 S E R I AL D I GI T AL S I GN ALS
• • •
479
Listener – a device capable of receiving data from other devices, e.g. a printer or monitor. Talker – a device capable of transmitting data to other devices, e.g. a counter or the data acquisition system of Figure 18.2. Controller – a device capable of managing communications on the bus by sending addresses and commands, e.g. a computer.
If the transmission distance for parallel signals is increased beyond a few metres, imperfections in the transmission line result in some of the bits in a given byte arriving out of synchronisation with the rest. Similarly the presence of external interference again results in loss of synchronisation or corruption of data. In conclusion, parallel digital signals are suitable for high-speed, short-distance communication in laboratory, environments. They are not suitable for long-distance communication in industrial environments where significant external interference may be present.
18.4 18.4.1
Serial digital signals Introduction Serial digital signals can be used to transmit data over much longer distances (typically up to around 1 km) and are therefore commonly used in telemetry systems.[3] Here all the data bits are transmitted one bit at a time in a chain along a single path. A serial digital signal is therefore a time sequence of two voltage levels, for example 0 V for a 0, 5 V for a 1 (unipolar), or −2.5 V for a 0, +2.5 V for a 1 (bipolar). The transmission path can vary from a standard twisted-pair cable to a low-loss coaxial cable or an optical fibre cable. Serial digital signalling is often referred to as pulse code modulation. Figure 18.3(a) shows the use of an eight-stage shift register to convert an 8-bit parallel digital signal into serial form. The parallel signal b7 . . . b0 is first loaded into the register and a clock signal applied. On receipt of the first clock pulse the contents of the register are shifted one place to the right, causing the least significant bit b0 to appear at the register output, i.e. the least significant bit is transmitted first. The second clock pulse causes the register contents again to be shifted one place to the right, causing the next bit b1 to be transmitted. The process is repeated until the register is empty: the most significant bit b7 is the last to be transmitted. Figure 18.3(b) shows the register used to convert a serial signal into parallel form. On receipt of the first clock pulse the least significant bit is loaded into the storage element on the extreme left of the register. The second clock pulse causes the register contents to move one place to the right, allowing the next significant bit to enter the register. After eight clock pulses the entire signal is loaded into the register, with the LSB b0 on the extreme right and the MSB b7 on the extreme left. Digital transmission links may be divided into three categories, depending on whether the communication is one-way or two-way. These categories are:
•
Simplex. One way communication from A to B where B is not capable of transmitting back to A. This may be sufficient where a remote outstation is merely sending data to a master station. However, the master station cannot acknowledge receipt of data or request retransmission of corrupted data.
480 DATA ACQUISITION AND COMMUNICATION SYSTEMS Figure 18.3 Parallel and serial digital signals: (a) Parallel to serial conversion (b) Serial to parallel conversion.
• •
Half duplex. Transmission from A to B and from B to A but not simultaneously. With this system, after an outstation has transmitted data to a master station, the master station can then send an acknowledgement and if necessary request retransmission. Full duplex. Simultaneous transmission from A to B and from B to A. This can be done using two paths; however, using modulation techniques, it is possible to transmit in two directions along a single path.
In each of the above systems it is important that the receiver is ready to receive and identify each set of data from the transmitter. There are two ways in which this can be achieved: asynchronous transmission and synchronous transmission. In asynchronous transmission each byte or frame of data is preceded by a start bit and concluded by a stop bit (Figure 18.4(a)), so that the receiver knows exactly where the data starts and finishes. The transmission rate of serial digital signals is specified using bit rate R; this is the number of bits transferred in unit time, usually expressed in bits per second. Because of the need to check start and stop bits, the maximum transmission rate possible with asynchronous transmission is around 1200 bits s−1. This method is therefore more suited to slower transmission systems.
18.4 S E R I AL D I GI T AL S I GN ALS
481
Figure 18.4 Asynchronous transmission and ‘handshaking’: (a) Asynchronous data framing (b) Connection for asynchronous transmission using RS 232 (c) ‘Handshaking’ sequence – A transmitting to B.
For rates greater than 1200 bits s−1 synchronous transmission is used. Here a regular clocking signal is used to keep the receiver exactly in step with the transmitter. The transmitted data is preceded by a synchronising character which acts as a clocking pulse at the receiver. The receiver will then ‘clock in’ each bit of data. There are several standard methods of serial digital communication available. The choice of method depends on several criteria, including the following:
• • • •
Transmission distance Bit rate R Resistance to external interference and noise Number of multiplexed signals over a single link.
These criteria are often conflicting; for example, a high bit rate is incompatible with a long transmission distance and high noise immunity. From Section 18.4.2, we see that the bandwidth required for transmission of PCM is proportional to the bit rate R, i.e. the greater R the higher the required bandwidth. However, the available bandwidth of a given electrical cable decreases with length as the effects of resistance and capacitance increase. From Section 18.4.3, we see that for PCM affected by ‘white’ noise, the standard deviation of noise present at the PCM receiver is proportional to êRú , i.e. the greater R the greater the noise present. However, since the receiver has
482 DATA ACQUISITION AND COMMUNICATION SYSTEMS simply to decide whether a 1 or a 0 has been transmitted, this decision can be made correctly even if the pulses are severely distorted by noise. Any errors that do occur, as a result of noise and interference, can be detected by adding check bits to the serial data signal (Section 18.5). From Section 18.4.2, the bit rate R for m multiplexed signals is m times greater than for a single signal, so that high m is again incompatible with long transmission distance and high noise immunity. Finally, the amount of external noise and interference will generally increase with transmission distance. One commonly used standard for serial digital signals is the RS 232 C/V 24 interface. This specifies a 25-line connector and can be used for asynchronous and synchronous communication. In asynchronous communication only seven lines are used: two for data (transmitted and received), four for control signals and one for a common return (Figure 18.4(b)). Figure 18.4(c) shows how the control signals are used in a ‘handshaking’ sequence; this is necessary to ensure that the receiver is ready to receive data from the transmitter. RS 232 is capable of a bit rate of up to 20 kbits s−1 over short distances, i.e. up to 15 m; longer transmission distances can be used at lower bit rates. However, RS 232 is vulnerable to external interference and cable resistance/capacitance effects and is best used over short transmission distances. For higher bit rates over longer distances, RS 232 is gradually being replaced by a new standard RS 449, which is capable of a bit rate of 10 kbits s−1 over a distance of 1 km. The reason for this longer transmission distance is that, in RS 449, the transmitting and receiving data lines each have their own separate return lines (rather than sharing a common return). Each data line and return line can then form a twisted pair to give shielding from inductively coupled interference (Section 6.5.2). However, in industrial environments, where there is high external interference, RS 449 cannot be used successfully even at low bit rates. The solution here is to use current loop transmission: as explained in Section 6.3, a current transmission system has far greater immunity to series-mode interference than an equivalent voltage transmission system. Here the current in the loop is serially switched between 0 and 20 mA; 0 mA corresponds to a 0 and 20 mA to a 1. With current loop transmission the bit rate is normally limited to 4800 bits s−1; also in order to transmit data to and from a computer a converter is necessary to convert the current serial signal into computercompatible RS 232. The serial digital techniques discussed above are suitable for transmission distances up to around 1 km. Successful transmission over longer distances in the presence of high interference can be obtained by frequency modulating the serial digital signal onto a carrier (Section 18.6).
18.4.2
Transmission bandwidth Figure 18.5 shows a simple PCM transmission system. A transmitter, consisting of a sample/hold device, ADC and parallel-to-serial converter, converts an input analogue voltage into a serial digital signal, which is sent over a transmission link to a receiver. The transmission link may be cable, radio link or optical fibre. In order to estimate the bandwidth required for the transmission link it is necessary to find the extent of the frequency spectrum of the PCM signal. We first need to find the bit rate of the PCM signal; this is the number of bits per second or baud (1 baud = 1 bit s−1).
18.4 S E R I AL D I GI T AL S I GN ALS
483
Figure 18.5 Calculation of PCM transmission bandwidth.
Consider a single signal, sampled fS times per second, each sample being encoded into an n-bit code. There are fS samples per second and n bits per sample, so that the bit rate is: Bit rate for a single signal
R = nfS
[18.2]
For m multiplexed signals, each sampled fS times per second, there are mfS samples per second, so that in this case the bit rate is: Bit rate for m multiplexed signals
R = nmfS
[18.3]
Figure 18.5 shows corresponding time variations in input analogue voltage, sampleand-hold output signal and PCM signal. The graphs assume that the sample/hold device is in the SAMPLE state for an infinitely short time. This means that, for a single signal, the time in the HOLD state is equal to the sampling interval ∆T. If the ADC
484 DATA ACQUISITION AND COMMUNICATION SYSTEMS has an 8-bit encoder, i.e. n = 8, then eight bits (either 0 or 1) of information must be transmitted during this time interval ∆T. Thus the width of each bit of information in the PCM signal is ∆T/8. There are 256 possible pulse patterns during each sampling interval, but the pulse pattern corresponding to 01010101 has the shortest period and the highest frequency components. From Figure 18.5 we see that this pulse pattern is a square wave of period ∆T/4. The frequency spectrum of this square wave (Section 4.3) consists of a fundamental of frequency 4/∆T Hz, together with harmonies at frequencies 3 × 4/∆T, 5 × 4/∆T, 7 × 4/∆T, etc. If this square wave signal is transmitted over a link with bandwidth between 0 and a little over 4/∆T (Figure 18.5), then the received signal contains only the fundamental frequency 4/∆T, i.e. it is a sine wave of frequency 4/∆T Hz. The receiver can still decide correctly that the transmitted message was 01010101, so that the minimum bandwidth required for transmission of the square wave is 0 to 4/∆T, i.e. 0 to 4 fS Hz (since fS = 1/∆T ). Since this square wave has the highest frequency components of all possible pulse patterns, then the minimum bandwidth required for transmission of the PCM signal is 0 to 4 fS. Thus in the general case of a single signal, sampled at fS and encoded into an n-bit code, we have: Minimum PCM bandwidth for a single signal
PCM bandwidth = 0 to –12 nfS
[18.4]
For m multiplexed signals, each sampled at fS = 1/∆T, the time in the HOLD state is ∆T/m. This means that n bits of information must be transmitted during time ∆T/m; i.e. the width of each bit of information is now ∆T/mn. The corresponding PCM bandwidth in this case is: Minimum PCM bandwidth for m multiplexed signals
PCM bandwidth = 0 to –12 mnfS
[18.5]
From [18.2]–[18.5] we see that a single general expression for minimum PCM bandwidth is: PCM bandwidth = 0 to –12 R
[18.6]
Thus a PCM signal, derived from 16 multiplexed signals, each sampled once per second and encoded into 12 bits, has a bit rate of 16 × 12 = 192 bauds and requires a transmission link with a minimum bandwidth of 0 to 96 Hz.
18.4.3
Effect of noise on PCM signal The transmission link connecting PCM transmitter and receiver may be affected by external interference and noise as shown in Figure 18.6(a). Figure 18.6(b) shows the power spectral density φS of the PCM signal, extending effectively from 0 to –12 R Hz. The figure also shows the power spectral density φN of ‘white’ noise; here φN has a constant value of A W Hz−1 over an infinite range of frequencies. The first stage of the receiver is a low-pass filter of bandwidth between 0 and –12 R Hz. This rejects noise frequencies greater than –12 R, but noise frequencies inside the signal bandwidth, i.e. between 0 and –12 R, are allowed to pass to the comparator. The total power WN of the noise in the comparator input signal is given by the area PQRS under the noise power spectral density curve (eqn [6.23]), i.e.
18.4 S E R I AL D I GI T AL S I GN ALS
485
Figure 18.6 Effect of noise of PCM signal.
WN ≈ –12 AR watts
[18.7]
Assuming the noise signal has zero mean R, then the standard deviation σ is equal to the root mean square value yRMS. From eqn [6.33] y 2RMS = WN, so that the standard deviation of the noise present in the comparator input signal is given by:
486 DATA ACQUISITION AND COMMUNICATION SYSTEMS
AR volts 2
σ = WN =
[18.8]
The PCM signal leaving the transmitter has a value VP volts for a 1 and 0 V for a 0. The second stage of the receiver is a comparator which compares the filter outputs VF with –12 VP , i.e. one half of the original pulse amplitude. The comparator output voltage is given by: VC = VP,
if
VF > –12 VP
VC = 0,
if VF ≤ –12 VP
[18.9]
so the receiver decides that a 1 has been transmitted if VF > –12 VP, and a 0 has been transmitted if VF ≤ –12 VP. This decision is often made correctly, even if the received pulses are distorted by noise. The presence of noise, however, does mean that some decisions are made incorrectly, i.e. the receiver decides that a transmitted 1 is a 0 or a transmitted 0 is a 1. The probability of these errors occurring can be evaluated using the probability density function p( y) for the noise. Suppose that a 0 is transmitted: we can assume, then, that the comparator input signal is noise alone with standard deviation σ and zero mean value ( R = 0). The probability of a wrong decision here is the probability of the noise being greater than –12 VP, i.e. Py > –12 VP. This probability is equal to the shaded area under the p( y) curve in Figure 18.6(c), i.e. Probability of a 0 being received as a 1
Py> (1/ 2)V = P
0002
∞
p( y ) dy
[18.10]
(1/ 2)VP
If a 1 is transmitted, then the comparator input signal is noise superimposed on a d.c. voltage of VP, i.e. noise with a mean value of VP. The probability of a wrong decision in this case is the probability that the noise will be less than –12 VP. This probability is equal to the shaded area in Figure 18.6(d), i.e. Probability of a 1 being received as a 0
Py< (1/ 2)V = P
0002
(1/ 2)VP
p( y − VP ) d y
[18.11]
−∞
In the special case of noise with a normal probability density function: p( y ) =
 ( y − R )2  exp −  2σ 2   σ 2π 1
[6.14]
it can be shown[3] that both the above probabilities are equal and given by PCM error probability (error rate) for normal noise
V  1 1 P P  = − σ 2 π
0002
VP / 2 2σ
0
exp(− x 2 ) d x
[18.12]
18.5 E RRO R D E TE CT I O N AN D CO RRE CT I O N
487
Assuming that a given digit is equally likely to be 0 or 1, then [18.12] gives the probability of an error in the decoding of any digit. Figure 18.6(e) shows how error probability varies with VP /σ (signal-to-noise ratio). We see that for VP /σ greater than around 8, a small increase in VP /σ causes a very large reduction in error probability. Thus increasing VP /σ from 8 to 12 reduces the error probability from approximately 10−4 to 10−8. At VP /σ = 7, the probability of error in a single bit is approximately 10−3. This means that the probability of error in a 16-bit signal is 0.016. There is an equal probability that the error will occur in any of the digits in the PCM signal. If the error occurs in the least significant bit (LSB) then the resulting measurement error will be small; if the error occurs in the most significant bit (MSB) then the measurement error will be 50% of full scale.
18.5
Error detection and correction For the reasons given above, it is important that any errors occurring during the decoding of a noise-affected PCM signal are detected, and in some cases corrected. This is achieved by the use of redundancy. Redundancy here means the addition of extra check digits to the information digits containing the measurement data. Thus each complete code word consists of n digits made up of k information (measurement) digits and r = n − k check digits as shown below. Such a code word is referred to as an (n, k) code and has a redundancy of (r/n) × 100%.
18.5.1
Single parity check bit system The simplest error detection system uses a single check bit, i.e. r = 1. The check bit is chosen using the concept of parity.[4] A complete code word has even parity if the total number of 1s is even, and odd parity if the total number of 1s is odd. Thus in an even parity check system the check digit is set so that the total number of 1s in the complete code word is even; in an odd parity check system the check digit is set so that the total number of 1s is odd. Examples are given below: Information bits
Even parity code word
Odd parity code word
1011 1000 0101 1111
10111 10001 01010 11110
10110 10000 01011 11111
The parity check bit is added to the information bits at the PCM transmitter using modulo 2 addition. This process is characterised by the rules: 0 ⊕ 0 = 0,
0 ⊕ 1 = 1,
1 ⊕ 0 = 1,
1⊕1=0
488 DATA ACQUISITION AND COMMUNICATION SYSTEMS and can be implemented by either an exclusive-or logic gate or a read only memory. The transmitter performs modulo 2 addition on the information bits. Thus in the above example: 1⊕0⊕1⊕1=1 1⊕0⊕0⊕0=1 0⊕1⊕0⊕1=0 1⊕1⊕1⊕1=0 In an even parity system the check bit is the result of modulo 2 addition of the information bits; in an odd parity system the check bit is the inverse of this result. The PCM receiver checks the parity of the complete received code word for correctness. Thus in an even parity system, a received code word with even parity is deemed to be correct, one with odd parity incorrect. This checking is performed by modulo 2 addition of all the digits in the code word. For an even parity system the result of addition is zero if there is no error. This simple system has several limitations. It only detects the presence of an odd number of errors, e.g. 1 or 3 in the above example. An even number of errors, e.g. 2 or 4, gives the correct code word parity and goes undetected. Even if an error is detected, this system cannot decide which bit or bits are in error and therefore cannot correct the code word.
18.5.2
Practical error detecting systems In industrial telemetry systems the amount of random noise present is often small; i.e. there is usually a high signal-to-noise ratio VP /σ and consequently a low probability of errors. Occasionally, however, large interference voltages lasting a short time occur; these voltage transients are often caused by switching electrical equipment on or off. In this situation it is obviously important to detect as many error combinations as possible, but it is not worth attempting to correct errors. If the receiver detects an error it simply requests a retransmission of the code word. Since substantial interference and corresponding ‘single burst’ errors occur infrequently, interruptions to normal operation due to requests for retransmission also occur infrequently. By using several check bits, each checking the parity of a different combination of information bits, it is possible to detect practically all error combinations. A typical arrangement used in an industrial telemetry system[5] is shown below:
Information bits b11
b10
X X X X
X X X
b9
b8
X X X
X X
b7
b6
X X
X
X
b5
Check bits b4
b3
b2
X
X X X X
X X X
X
X
X
b1
b0
X
c3
c2
c1
c0
X X
X X
X X
X X
18.5 E RRO R D E TE CT I O N AN D CO RRE CT I O N
489
Here there are 12 information bits and four parity check bits, i.e. k = 12, r = 4 and n = 16. For example, the transmitter sets c3 so that the combination b11 b9 b7 b5 b3 b1 c3 has even parity. The complete scheme is represented by: 5 c 3 = b11 ⊕ b9 ⊕ b7 ⊕ b5 ⊕ b3 ⊕ b1 4 c 2 = b11 ⊕ b10 ⊕ b7 ⊕ b6 ⊕ b3 ⊕ b2 6 [18.13] c1 = b11 ⊕ b10 ⊕ b9 ⊕ b8 ⊕ b3 ⊕ b2 ⊕ b1 ⊕ b0 4 c 0 = b11 ⊕ b10 ⊕ b9 ⊕ b8 ⊕ b7 ⊕ b6 ⊕ b5 ⊕ b4 ⊕ b3 ⊕ b2 ⊕ b1 ⊕ b0 7 A similar system with 16 information bits and five check bits, generated using the polynomial 1 + x 2 + x 5, is described in Ref. 6.
18.5.3
Error correction systems Suppose we require to correct a single incorrect digit in a complete code word of n digits. The check digits must contain the following information: either that the complete code word is correct or the position of the incorrect digit. This means that the check digits must be able to signal (n + 1) possible situations; the number of check bits r required is therefore given by: 2r = n + 1 i.e. r = log2(n + 1) = 3.33 log10(n + 1)
[18.14]
Thus three check digits will correct a seven-digit code word, four check digits a 15-digit code word and five check digits a 31-digit code word. However, it is possible to have more than one error in a code word, so that the check bits must signal more than (n + 1) possible situations. For example, in order to correct both single errors and double errors in adjacent digits, (2n + 1) situations must be signalled, and r = log2(2n + 1). One practical error-correcting method is to send the information digits in groups or blocks, i.e. n groups each containing m digits. The information digits are arranged in a n × m matrix form and parity check bits added to each row and column at the transmitter. A 4 × 4 system is shown below: Odd parity check on rows
1 4 2 4 3
0 1 1 1
0 0 0 1 0 0 1 1 0 1 1 0 1 0 0 1
0
0
0
1
0
Transmitted
Odd parity check on columns
0 1 1 1
1 2 3 4 1* 0 0 11 0 0 1 12 0 1 1 03 1 0 1* 1 4
0
0
0 1
0
Received
Each column and row in the received code is then checked for parity. In the examples shown, rows 1 and 4 and columns 1 and 3 have even parity, so that the incorrect digits are in the positions marked. This method is useful for information in b.c.d. code, since here the digits are arranged in groups of four, one group for each decade.
490 DATA ACQUISITION AND COMMUNICATION SYSTEMS
18.6
Frequency shift keying In the previous section we saw that the frequency spectrum of the PCM signal extends, effectively, from 0 to –12 R Hz. This means that part of the PCM spectrum may coincide with the spectrum of interference voltages, usually at 50 Hz, due to nearby power circuits (Figure 6.13(b)). Also the bandwidths of practical transmission links do not normally extend down to 0 Hz: for example, a British Telecom landline may have a bandwidth between 300 and 3300 Hz, and a VHF radio link bandwidth between 107.9 and 108.1 MHz. These two problems can be solved by modulating the PCM signal onto a carrier signal, whose frequency lies within the bandwidth of the transmission link. The spectrum of the signal is now shifted up to the carrier frequency, away from the interference spectrum, so that the latter can be rejected by a band pass filter (Figure 6.13(d)). Two types of modulation were discussed in Chapter 9. In amplitude modulation (AM) the modulating signal alters the amplitude of a sinusoidal carrier; in frequency modulation (FM) the modulating signal alters the frequency of the carrier. It will be shown later in this section that FM requires a greater bandwidth than AM. However, provided the change in carrier frequency is sufficiently large, an FM receiver is better at improving a given signal-to-noise ratio than an AM receiver. For this reason, in most telemetry systems, the PCM signal is frequency modulated onto a carrier; this is called frequency shift keying (FSK).
18.6.1
FSK transmitters and receivers The voltage controlled oscillator (VCO) is the basis of both FSK transmitters and receivers. The principle of feedback oscillators was discussed in Section 9.5; we saw that the frequency of an electrical oscillator depends on the inductance and capacitance of an L–C–R circuit. The frequency of oscillation of a VCO is determined by the magnitude of the input voltage. Thus if V and f are corresponding values of input voltage and oscillation frequency, we have:
Frequency of voltage controlled oscillator
f = fC + kV
[18.15]
where fC is the frequency at zero voltage (unmodulated carrier frequency), and k Hz V−1 is the VCO sensitivity. The corresponding VCO output signal is: VVCO = Ä sin 2π ( fC + kV )t
[18.16]
If a PCM signal is input to a VCO (Figure 18.7(a)), the VCO output is an FSK signal; this has two frequencies, f1 for a 1 and f0 for a 0. For bipolar PCM with V = +VP for a 1 and V = −VP for a 0, the FSK frequencies are: f1 = fC + kVP = fC + D, for a 1 f0 = fC − kVP = fC − D, for a 0
[18.17]
where D = kVP is the maximum frequency deviation of the carrier. A typical FSK transmitter, suitable for a British Telecom landline, has fC = 1080, f0 = 960, f1 = 1200, i.e. D = 120 Hz.[7]
18.5 FREQUENCY SHIFT KEYING
491
Figure 18.7 FSK transmitter and receiver: (a) Transmitter (b) Receiver.
Figure 18.7(b) shows an FSK receiver which converts an incoming FSK signal back to PCM. The first stage of the receiver is a band pass filter which rejects all noise and interference outside the FSK bandwidth. The second stage is a phase locked loop (PLL), which consists of a VCO, a multiplier and a low pass filter in a closed loop system. The multiplier detects any difference in phase between the input signal and the VCO signal. Suppose that the input signal is Ä sin 2π( fC + D)t (corresponding to a 1) and initially VOUT = 0, so that the VCO output signal is Ä sin 2π fC t. The multiplier output signal contains sum and difference frequencies, i.e. 2fC + D and D; the LPF removes the 2fC + D component so that now VOUT is a low amplitude signal of frequency D. This causes the frequency of the VCO output signal to increase until it is equal to that of the input signal, i.e. until fC + kVOUT = fC + D. At the same time the frequency of VOUT falls from D to zero as the amplitude of VOUT increases. When the system settles out, VOUT is equal to a d.c. voltage of magnitude +D/k (for a 1). Similarly, when the input frequency is fC − D, the system settles out with fC − D = fC + kVOUT and VOUT a d.c. voltage of magnitude −D/k (for a 0). In both cases the VCO frequency is said to ‘lock’ onto the input frequency.
18.6.2
Bandwidth of FSK signal In Section 9.3 we saw that if a single sine wave of frequency fi is amplitude modulated onto a sinusoidal carrier of frequency fS, then the spectrum of the AM signal
492 DATA ACQUISITION AND COMMUNICATION SYSTEMS consists of two lines at frequencies fS − fi and fS + fi. This means that if a random signal, with spectrum between 0 and fM, is amplitude modulated onto fS, the spectrum of the AM signal lies between fS − fM and fS + fM . If a single sine wave Äi sin 2π fi t is frequency modulated onto a sinusoidal carrier using a VCO, then from eqn [18.15] the instantaneous frequency of the FM signal is: f = fC + kÄi sin 2π fi t = fC + D sin 2π fi t
[18.18]
Here fC is the unmodulated carrier frequency and D = kÄi is the maximum deviation of f from fC. The resulting FM signal is given by: VFM = Ä sin 2π( fC + D sin 2π fi t)t
[18.19]
From Figure 18.8 we see that the spectrum of this FM signal is wider and more complex than the corresponding AM spectrum. In FM there are several lines, symmetrically arranged about fC, the number and relative amplitudes of the lines depending on the modulation index D/fi.[8] For a random modulating signal containing frequencies between 0 and fM, the FM spectrum consists of a large number of lines. In this case, the number of lines and the width of the spectrum depend on the appropriate modulation index D/fM. For very small and very large values of D/fM, i.e. D/fM 0002 1 and D/fM 0003 1, the frequency spectrum of the FM signal extends approximately from fC − (D + fM) to fC + (D + fM).[8] For D/fM ≈ 1, the FM spectrum extends approximately from fC − (D + 2fM) to fC + (D + 2 fM).
Figure 18.8 Comparison of AM and FM spectra.
18.7 CO MMUN I CAT I O N S Y S T E MS F O R ME AS UR E ME N T
493
Summarising these results we have: Approximate FM bandwidth
Approximate FM bandwidth = fC − ( D + f M )
D D 0002 1, 00031 fM fM D to fC + ( D + 2 f M ), ≈1 fM
to fC + ( D + f M ),
= fC − ( D + 2 f M )
[18.20]
In FSK the modulating signal is PCM. The PCM signal has a frequency spectrum effectively between 0 and –12 R (Section 18.4.2); here fM = –12 R, where R is the PCM bit rate. Equations [18.20] can therefore be used to give corresponding expressions for the bandwidth necessary to transmit an FSK signal, i.e. Approximate FSK bandwidth
Approximate FSK bandwidth = fC − ( D + 12 R) = f C − ( D + R)
to fC + ( D + 12 R), to fC + ( D + R),
2D 2D 0002 1, 00031 R R 2D ≈1 R
[18.21]
Thus an FSK signal with fC = 1080 Hz, D = 120 Hz and R = 200 baud has 2D/R = 1.2, and requires a bandwidth approximately between 760 and 1400 Hz.
18.7 18.7.1
Communication systems for measurement Introduction In the introduction to this chapter, we saw that in many industrial situations it will be necessary to transmit measurement data from transducers/transmitters (outstations, O/S), located at different items of plant equipment, to signal processing and data presentation elements (master stations, M/S), located in a central control room. The distances between individual plant items and between the items and the control room may be up to a few kilometres. A communications system is therefore required which is capable of transmitting large amounts of information in two directions (M/S to O/S and O/S to M/S), over long distances in the presence of external interference and noise. Such systems are often referred to as telemetry systems. In Section 9.4.3 we discussed intelligent or smart transmitters. These transmitters incorporate a microcontroller which is used not only to calculate the measured value of the variable but also to control a digital communications module. Since this module can both transmit and receive digital information, an intelligent transmitter can therefore act as an outstation in a telemetry system. The main problem here is that there are a large number of digital communication standards in current use. This means that transmitting/receiving equipment made by one manufacturer may not be
494 DATA ACQUISITION AND COMMUNICATION SYSTEMS compatible with that made by another manufacturer. There is a need therefore for an agreed signal standard for the digital communication of measurement data, just as a 4–20 mA current loop has been used as an analogue transmission standard.
18.7.2
Reference Model for Open Systems Interconnection (OSI) In 1983 the International Standards Organisation (ISO) approved the Reference Model for Open Systems Interconnection (OSI) as an international communication standard. The aim of the standard is to allow open communication between equipment from different vendors. It will be many years before this aim is achieved but the OSI model will provide the basis of the aim. In the meantime the model can act as a bridge between two different proprietory systems which previously could not communicate. The OSI model is an abstract concept consisting of seven layers (Figure 18.9(a)). Each layer represents a group of related functions or tasks. OSI protocols are used to define these functions but do not define how these functions are implemented. The main purpose of the model is to provide a structure whereby vendor independent systems can be implemented. The functions of the seven layers can be summarised as follows:
• • • Figure 18.9 Layered protocol models: (a) ISO OSI model (b) Fieldbus model.
Application Layer (7) This is the highest-order layer in the model and its purpose is to ensure that a user application program can both receive and transmit data. Presentation Layer (6) This layer ensures that an application correctly interprets the data being communicated by translating any differences in representation of data. Session Layer (5) This layer is responsible for controlling communication sessions between applications programs.
18.7 CO MMUN I CAT I O N S Y S T E MS F O R ME AS UR E ME N T
• • • • 18.7.3
495
Transport Layer (4) This layer provides those layers above it with a reliable data-transfer mechanism which will be independent of any particular network implementation. Network Layer (3) This layer provides the actual communication service to the transport layer; it controls communications functions such as routing, relaying and data link connection. Data Link Layer (2) This layer controls the transfer of data between two physical layers and provides for (ideally) error-free sequential transmission of data over a link in a network. Physical Layer (1) This layer defines the actual signalling method employed, together with the type of transmission medium, e.g. copper wire, fibre optic, radio waves, etc.
Fieldbus Work is currently in progress to produce a single standard for two-way digital communication of measurement and control data between intelligent sensors/transmitters located in the field and a computer-based master station located in a control room. This standard is referred to as ‘Fieldbus’ and is particularly, but not exclusively, applicable to the process industries. The standard is based on a simplified three-layer version of the seven-layer OSI model (Figure 18.9(b)): the physical layer (1), the data link layer (2) and the application layer (7). Detailed discussion has taken place over many years with the aim of producing an agreed standard for each of the three layers. These standards are now in the process of being finalised. We now discuss how each of the layers can be implemented.
Physical layer[9]
Figure 18.10 Network topologies: (a) Star (b) Ring (c) Multi-drop.
The first decision to be made is the network topology, i.e. the geometry of the interconnection between the outstations and the master station. Figure 18.10(a) shows three classic network topologies: star, ring and multi-drop. In any given link in the star arrangement there is only communication between the master station and a single outstation. This means that less multiplexing but more wiring is required; also if the master station fails the whole network fails. The ring arrangement also has limited reliability. The network will only function if all the stations are working; if any station fails then all stations beyond the failed station are also inaccessible. In the multi-drop arrangement several outstations and master stations share a common transmission path. Some method of multiplexing is therefore required to separate the signals from each outstation. In time division multiplexing (TDM, Section 18.1) the
496 DATA ACQUISITION AND COMMUNICATION SYSTEMS signals are separated in time so that, for example, O/S 1 transmits to M/S during a given time slot, O/S 2 transmits to M/S during the next time slot and so on. The alternative is frequency division multiplexing (FDM) where the signals are separated in frequency by modulating them onto different carrier frequencies (Sections 9.3 and 18.6). However, FDM requires more complex hardware so that TDM is usually preferred. A protocol will be necessary to arbitrate between the different outstations and avoid data collisions. The multi-drop arrangement has high reliability: the network will still function if any outstation fails, and if a suitable protocol is used it will still function, to a limited extent, if the master station fails. The multi-drop arrangement is therefore the preferred topology for Fieldbus. The next decision to be made concerns the interconnection medium to be used to implement the bus arrangement. This could be copper wire, optical fibre or a radio link. It is generally extremely difficult to implement the bus arrangement using optical fibres and there are problems in making reliable fibre/fibre connections in hostile environments. If the distances between the outstations and master station are many kilometres then radio links may be required. However, in many industrial situations these distances are only a few kilometres at most; this means that copper wire is feasible. Since it is also the cheapest practical way of implementing the bus arrangement, it is preferred for Fieldbus. In order to minimise the effects of inductive and capacitive coupling to interference sources (Section 6.5), screened twisted-pair wires should be used. The final decision in the physical layer is the signalling method to be used. There are two methods of transmitting digital data: either in parallel form (Section 18.3) or in serial form (Section 18.4). We saw in Section 18.3 that parallel digital signals are suitable for high-speed, short-distance (up to 15 m) communication in laboratory environments where there is low-level electrical interference. Serial digital signals (pulse code modulation) are more suited to longer transmission links where significant electrical interference is present. If the serial signal is not modulated onto a carrier so that the signal bandwidth is determined entirely by the bit rate R (eqn [18.6]), this is referred to as base-band transmission. Figures 18.11(a) and (b) show two base-band transmission methods: non-return to zero (NRZ) and bi-phase Manchester. NRZ is characterised by one voltage level (e.g. 0 V) corresponding to binary 0 and another (e.g. 5 V) corresponding to binary 1. The major disadvantage of this method is that a long sequence of 1s or 0s results in the voltage on the line remaining constant. This problem is overcome in the bi-phase method where one phase, e.g. , corresponds to a 0 and another phase, e.g. , corresponds to a 1; this guarantees a change in voltage at least once per bit period. The noise immunity of the serial signal can be increased by modulating it onto a carrier signal; this is carrier band transmission. Figure 18.11(c) shows frequency shift keying (FSK, Section 18.6). Here there are two distinct frequencies: the lower frequency corresponds to binary one and the higher frequency to binary zero. One possible implementation of the physical layer is provided by the HART protocol which is marketed by Fisher-Rosemount.[10] This is compatible with the smart transmitters discussed in Section 9.4.3. A multi-drop bus topology is used which can accommodate up to 15 smart devices with one power source. The interconnection medium is a single shielded twisted pair (maximum length 3 km) or multiple twisted pairs with an overall shield (maximum length 1.5 km). The signalling method is FSK based on the Bell 202 communications standard which uses 1200 Hz to represent binary 1 and 2200 Hz to represent 0; the bit rate is 1200 bits/s. This FSK signal is
18.7 CO MMUN I CAT I O N S Y S T E MS F O R ME AS UR E ME N T
497
Figure 18.11 Serial digital signalling: (a) Non-return to zero (NRZ) (b) Bi-phase Manchester (c) Frequency shift keying (FSK).
superimposed on the normal 4–20 mA analogue signal in a current loop (Figure 9.23). The amplitude of the FSK signal is 0.5 mA, so that if the normal analogue current is 12 mA, the minimum and maximum instantaneous currents are 11.5 and 12.5 mA. The average value of loop current is therefore unchanged at 12 mA. The Bell 202 communications system also allows smart devices to be directly connected to leased telephone lines. This allows the device to communicate with a central control point many kilometres away. Since the transmitter power is also isolated from the communications in leased line applications, any number of devices can be networked.
Data Link layer[9] As a result of the above discussions, the physical layer will be a single multidrop bus which is shared between several intelligent devices using time division multiplexing. The data link layer is concerned with the transfer of data between these devices. The first requirement is for a method of controlling the bus which specifies the time interval during which a given device will transmit and receive information. The simplest method is to have a single fixed central master station which controls the times at which each of the outstations transmits and receives. This method has limited reliability, however, since if the master station fails then the whole network fails even though all of the outstations may be working correctly. The reliability is increased if two or more stations are designated to be masters. Another method is decentralised mastership; here each of the stations connected to the bus takes it in turn to be the master for a given period of time. This can be achieved using token passing. The station holding the token is responsible for initiating all communications on the network for a given period of time. At the end of the period it must pass on the token to the next station. In order that data is correctly routed between stations, addressing information must be added to the data. Thus if a given station is transmitting data to another station, then the data should be preceded by an address code which uniquely defines the source and destination addresses. An 8-bit address code could specify up to 16 source
498 DATA ACQUISITION AND COMMUNICATION SYSTEMS Figure 18.12 Typical frame format.
addresses and up to 16 destination addresses. There is also a requirement for control or command instruction codes where, for example, a given device is required either to receive (read) incoming data or to transmit (write) data to another device. Finally, there is a need for any errors present in the transmitted information to be detected and ideally corrected. This is done by adding additional parity check bits (Section 18.5). The total information transmitted from one outstation to another now consists of different types, one type following another in time. It is essential therefore that the data is arranged in packets or frames. Each frame should have a format which clearly separates and distinguishes between different types of data, for example addressing codes and check codes. Figure 18.12 shows a typical frame format; it commences with a beginning of message or START flag code, followed by an address field, a command field, an information field and a parity check field, and concludes with an end-of-message or STOP flag code.
Application layer The role of the application layer can be summarised as follows: 1.
2.
It enables the master station to obtain measurement data from the data link layer in a form suitable for further processing in a user applications program. An example is a master station obtaining volume flow rate data Q and density data ρ from smart field transmitters in a form suitable for an applications program to calculate mass flow rate Ç = ρQ. It enables the master station to output data from an applications program to the data link layer in a form suitable for transmission to outstations. An example is an applications program in the master station changing the input range of a smart field differential pressure transmitter from 0 to 2.5 m of water to 0 to 3.0 m of water.
The application layer therefore specifies a set of control or command instructions which occur as a corresponding code in the data frame (Figure 18.12). Three necessary basic instructions are:[10]
• • • 18.7.4
READ, e.g. Master station reads measured value; Transmitter reads input range data WRITE, e.g. Master station specifies identification number and input range of transmitter; Transmitter outputs measured value CHANGE, e.g. Master station changes transmitter input range.
Fieldbus development Figure 18.13 shows a possible future development of Fieldbus.[11] It consists of a low-speed bus and a high-speed bus linked by a bridge. The low-speed bus can be implemented using the HART protocol described earlier. This uses multi-drop bus topology, with up to 15 smart transmitters (Section 9.4.3)
18.7 CO MMUN I CAT I O N S Y S T E MS F O R ME AS UR E ME N T
499
Figure 18.13 Development of Fieldbus.
Figure 18.14 Frame format for HART protocol.[10 ]
or other field devices connected to the bus. The interconnection medium is twisted pairs and the signalling method is FSK, based on the Bell 202 modem standard; the bit rate is 1200 bits/s. Figure 18.14 shows the general format of a message frame in the HART protocol. Each field of information can consist of several bytes; each byte of information consists of a start bit, eight data bits, an odd parity bit and a stop bit.[10 ] There are three classes of commands in the HART command set. Universal commands are recognised and implemented by all field devices. Common-practice commands provide functions implemented by most, if not all, field devices. Device-specific commands represent functions that are unique to each field device. Typical commands are ‘read measured variables’, ‘identify device’ and ‘read device information’. The high-speed bus can be implemented using an Ethernet network. Ethernet is the most widely installed local area network (LAN) and is specified in the IEEE 802.3 standard. The most commonly installed Ethernet systems are called 10BASE-T and use ordinary telephone twisted-pair wire. Bit transmission rates of up to 10 Mbits/s are obtained with this standard; the signalling method is bi-phase Manchester. Other allowed transmission media at 10 Mbits/s are thin wire coaxial cable (up to 185 m), thick wire coaxial cable (up to 500 m), broadband coaxial cable (up to 3500 m) and optical fibre. For Fieldbus development it is proposed to use fast Ethernet, based on the 100BASE-T standard. This uses twisted-pair wires and operates at 100 Mbits/s. A variety of intelligent devices, including computers and controllers, can be connected on to the high-speed bus using multi-drop topology.
500 DATA ACQUISITION AND COMMUNICATION SYSTEMS
Conclusion This chapter commenced by discussing the principles of time division multiplexing and its use in data acquisition systems. This was followed by a discussion of the principles of a range of techniques which form the basis of communications systems. These are parallel digital signalling, serial digital signalling, error detection/ correction and frequency shift keying. The chapter concluded by studying how communication systems for measurement data can be implemented, making special reference to the emerging Fieldbus standard.
References [1] Mowlem Microsystems 1985 Technical Information on ADU Autonomous Data Acquisition Unit. [2] RDP Electronics 1986 Technical Information on Translog 500 Data Acquisition System. [3] matthews p r 1982–83 ‘Communications in process control’, Measurement and Control, Nov. and Dec. 1982, Jan. 1983. [4] carlson a b 1975 Communication Systems, 2nd edn, McGraw-Hill–Kogakusha International Student Edition, pp. 410–17. [5] Kent Process Control 1985 Technical Information on Kent P4000 Telemetry Systems. [6] Serck Controls 1977 Product Data Sheet on a Telemetry Drive Module PDS 10/9. [7] A.T.S. (Telemetry) Ltd 1987 Technical Information on Type 1100 F.S.K. Data Modules. [8] carlson a b 1975 ibid., pp. 225–37. [9] atkinson j k 1987 ‘Communications protocols in instrumentation’, J. Phys. E: Scientific Instrumentation, vol. 20, pp. 484–91. [10] bowden r 1996 HART Field Communications Protocol, Fisher-Rosemount Technical Publication. [11] Fisher-Rosemount 2002 The Basics of Fieldbus – Technical Data Sheet.
Problems 18.1
Sixteen analogue input voltages, each with a frequency spectrum between 0 and 5 Hz, are input to a time division multiplexer. The multiplexed signal passes to a serial digital (PCM) transmitter consisting of a sample/hold device, 12-bit binary ADC and a parallel-to-serial converter. The PCM signal is transmitted to a distant receiver over a link affected by ‘white’ noise with a power spectral density of 0.2 mW Hz−1. (a) (b) (c) (d) (e)
Suggest a suitable sampling frequency for each input signal. What is the corresponding number of samples per second for the multiplexed signal? What is the maximum length of time the sample/hold device can spend in the hold state? What is the maximum percentage quantisation error for the ADC? Find the bit rate and minimum transmission bandwidth for the PCM signal.
PROBLEMS
(f )
18.2
501
The first stage of the PCM receiver is a low-pass filter with a bandwidth ‘matched’ to the frequency spectrum of the PCM signal. Estimate the standard deviation of the noise present at the filter output.
A PCM transmitter sends out a 12-bit serial digital signal with 5 V corresponding to a 1 and 0 V corresponding to a 0. The signal passes over a transmission link, affected by random noise, to a PCM receiver consisting of a low-pass filter and a comparator. The comparator input signal is the PCM signal with noise superimposed on it; the probability density p(y) of this noise is the triangular function shown in Figure Prob. 2. The comparator decides that a 1 has been transmitted if the comparator input signal is greater than 2.5 V, and that a 0 has been transmitted if the comparator input signal is less than or equal to 2.5 V. (a) (b) (c) (d) (e)
Calculate p(0) such that p(y) is normalised. Find are probability that y is greater than +1.0 V. Find the probability that a 0 is received as a 1. Find the probability that a 1 is received as a 0. What is the probability that a single error occurs in a complete code word?
Figure Prob. 2.
18.3
Ten measurement signals are input to a multiplexer so that each one is sampled twice per second. The multiplexed signal is input to a serial digital transmitter incorporating a 10-bit ADC. The resulting PCM signal is converted into FSK such that 720 Hz corresponds to a 1 and 480 Hz corresponds to a 0. Estimate the bandwidth required by the FSK signal.
18.4
(a)
Random noise is characterised by a Gaussian probability density function of standard deviation σ = 1.0 V and mean value R = 0 V. Using the probability values given below, calculate the probability that the noise voltage: (i) (ii) (iii)
(b)
exceeds +0.5 V lies between −1.0 and +1.0 V is less than −1.5 V.
Information in binary form is sent with a 0 represented by 0 V and a 1 represented by 5 V. During transmission the above random noise is added to it. The receiver decides that a 0 has been transmitted if the total input voltage is less than 2.5 V and that a 1 has been transmitted if the total input voltage is greater than 2.5 V. Use the probability values below to estimate the probable number of errors if 1600 bits of information are transmitted. Py > ä+0.5σ = 0.3085,
Py > ä+ σ = 0.1587
Py > ä+1.5σ = 0.0668,
Py > ä+2.5σ = 0.0062
19 The Intelligent Multivariable Measurement System In Section 9.4 we introduced the concept of the intelligent transmitter which has the capability of calculating an estimate of the measured value of a variable. The concept of computer estimation of measured value was also explained in Section 3.3. This chapter discusses the principles and implementation of intelligent multivariable measurement systems, which have the ability to estimate the measured values of a number of process variables. This type of system will require several sensors, usually one for each measured variable. The output of a given sensor may also be affected by some of the other process variables, so that inverse sensor models (Section 3.3) are required which take account of this. The system should also have the ability to calculate estimates of the values of process variables which are not measured from estimates of variables which are measured, using process models. This is the concept of the virtual instrument discussed in Section 10.3. This chapter first discusses the structure of a typical intelligent multivariable system and the elements which are present. It then considers the types of modelling methods which are used in these systems.
19.1
The structure of an intelligent multivariable system Figure 19.1 shows the general structure of an intelligent multivariable measurement system. The purpose of the system is to present the observer with a set of measured values of the process variables. We now discuss the elements which are present in the system.
19.1.1
The process The process could be a chemical plant, power station, steel mill, oil refinery, car, ship or aircraft, for example. In all cases there may be a number of information variables which describe the process. Examples of these are given in Table 1.1. In this chapter we define {P} to be the set of n process information variables or simply process variables. Thus the set of process variables for a gas pipeline could be:
504 TH E INTELLIGENT MULTIVARIABLE M E AS URE ME N T S Y S TE M
Figure 19.1 The intelligent multivariable measurement system.
{P} = {volume flow rate, temperature, pressure, density, mass flow rate, enthalpy flow rate} We can then define {PT} as the set of n numbers which are the true values of these process variables. However, it may be impossible, impractical or uneconomic to measure all n process variables. We define {I} as the set of m measured variables (m ≤ n). {I} is therefore a subset of {P}; in the gas pipeline example above we could have: {I} = {volume flow rate, temperature, pressure} The system should therefore have the ability to calculate the n − m unmeasured variables from the m measured variables. In the example given, mass flow rate, density and enthalpy flow rate are calculated from volume flow rate, temperature and pressure. We can define {I T} as the set of m numbers which are the true values of the measured variables.
19.1.2
The sensor array Assuming that one sensor is required for each measured variable, an array of m sensors is required. The principles and characteristics of a range of sensing elements in widespread current use are discussed in Chapter 8. Each sensor will require associated signal conditioning elements, such as bridges, amplifiers and resonators; these are discussed in Chapter 9. Furthermore, analogue-to-digital and frequency-todigital converters (Section 10.1) are required to give parallel or serial digital output signals. {U} is the set of m uncompensated digital sensor output signals. {U} is passed to a signal processing element; this is a computer with software to implement two types of model.
19.1.3
Inverse sensor models This model is shown in Figure 19.2. Here, for a given sensor i, with an uncompensated output Ui, the model calculates an estimate I i′ of the ith measured variable Ii. {I′} is the set of m numbers which are estimates of the measured values. We can say that the model maps the set {U} into the set {I′}. However, an individual estimate I i′ may also depend on estimates of other measured variables {I}. In general, the
19.1 THE S T R UCTURE O F AN I N TE LLI GE N T MULTI VAR I ABLE S Y S T E M
505
Figure 19.2 The inverse sensor model.
model therefore provides a function f which maps the sets {U} and {I ′} into the set {I′}, i.e. f: {{U}, {I′}} → {I′}
[19.1]
In the example of Figure 19.2, I 1′ depends on U1 and I 3′ , I 2′ depends on U2 and I 1′, and I 3′ depends on U3 and I 2′ . In many cases this mapping can be represented by m inverse sensor equations of the form (eqn [3.20]): I′ = K′U + N′(U ) + a′ + K M ′ I′MU + K′I I′I
[19.2]
which describe non-linear characteristics, a single modifying input I M ′ and a single interfering input I I′. If eqn [19.2] can be applied to the three variables I 1′, I 2′ and I 3′ in Figure 19.2 then we have: I 1′ = K′1U1 + N′1(U1) + a′1 + K′M13U1I 3′ I′2 = K′2U2 + N 2′ (U2) + a′2 + K′I21I′1
[19.3]
I′3 = K′3U3 + N 3′ (U3) + a′3 + K′M32U3I′2 + K′I32I′2 Here I′3 acts as a modifying input on I 1′, I 1′ as an interfering input on I 2′ , and I′2 as both a modifying and an interfering input on I′3. In general, a given sensor may be affected by several modifying and interfering variables. The three eqns [19.3] can be solved to give I 1′, I′2 and I′3.
19.1.4
Process models This model is shown in Figure 19.3. Here the model uses the set {I′} of estimates of the m measured variables to calculate the set {P ′} of measured values of the n process variables. We can say that the model provides a function g which maps the set {I′} into the set {P′}, i.e.
506 TH E INTELLIGENT MULTIVARIABLE M E AS URE ME N T S Y S TE M Figure 19.3 The process model.
g: {I ′} → {P′}
[19.4]
If we consider the relationship between individual members I′i and P′i of the sets, then we have: P′i = I′i
for
i = 1, . . . , m
P′i = gi{I′} for i = m + 1, . . . , n
[19.5]
where gi is the ith function of the set {I′}. In some cases this mapping may be implemented by a number of simple equations. If we return to the gas pipeline example of Section 10.1.1 we have the set of six process variables: {P} = {volume flow rate Q, absolute temperature θ, absolute pressure P, density ρ, mass flow rate Ç, enthalpy flow rate À} and the set of three measured variables: {I} = {Q, θ, P} The three non-measured process variables ρ, Ç and À can be calculated from Q, θ and P using:
ρ=
P Rθ
[19.6]
Ç = ρQ À = CP Çθ where R is the gas constant and CP the specific heat at constant pressure. The set of measured values {P′} are then transferred to the data presentation element for display to the observer.
19.2 MO D E LLI N G ME T H O D S F O R MULTI VAR I ABLE S Y S T E MS
19.2 19.2.1
507
Modelling methods for multivariable systems The general multivariable modelling problem In the previous section we saw the need for both inverse sensor and process models. This section examines different modelling methods that can be used. Figure 19.4 shows the general multivariable modelling problem. Given a set {x} of p input variables and a set { y} of r output variables, we need to find a function f which maps the set {x} into the set { y}, i.e. a function f such that: f: {x} → {y}
[19.7]
In many cases this mapping may be represented by a number of equations. In Section 19.2.2 we show how the basic equations of physics and chemistry can be used in modelling. Section 19.2.3 explains the derivation and applications of regression equations using experimental data. Finally Section 19.2.4 shows how artificial neural networks can be used to implement the mapping in situations where suitable equations cannot be found.
19.2.2
Physical and chemical model equations Table 19.1 lists some physical and chemical equations which can be used to model sensors. It should be noted that in some cases the fundamental equation has to be corrected for practical use by introducing correction factors whose values are found from experimental data. The equation for an ion-selective electrode can be used in a measurement system to measure the concentration or activity a of dissolved sodium (Na+), potassium (K+) and calcium (Ca2+) ions in a sample of water. An array of three ion-selective electrodes is required, one for each ion. However, the e.m.f. of each electrode depends not only
Figure 19.4 The general multivariable model.
508 TH E INTELLIGENT MULTIVARIABLE M E AS URE ME N T S Y S TE M Table 19.1 Physical and chemical equations used in sensor models. Elastic
x=
1 F k
Accelerometer
x=
ma k
εε 0 A d
Differential pressure flowmeter
Ç = CD Eε
n2 0002
Vortex flowmeter
f=
Coriolis flowmeter
T = 4lrω Ç
E = bmωr sin(mωr t)
Thermal sensor
TD =
1 PD + Ts UA
Electromagnetic – flowmeter
E = Blv
Density transmitter
ρ=
A B + +C 2 f n fn
Thermoelectric
ET1,0 = ET1,T3 + ET3,0
pH electrode
E = E0 − 0.0592 pH
Piezoelectric
q = dF
Ion-selective electrode
E = E0 +
Resistive – strain gauge
R = R0(1 + Ge)
Resistive – thermistor
R = K exp
Capacitive
C=
Inductive
L=
Hall effect
V=−
Electromagnetic – tachogenerator
A βD C θF
A RH D IB C t F
π 4
d 2 2ρ 1 ∆ P
Sv d
Rθ loge(aX + KX/Y aY) F
on the activity a X of the selected ion X but also on the activities aY, aZ of the other ions Y and Z. The sensor array can be described by the three equations: ENa+ = E0Na+ +
Rθ loge(aNa+ + KNa+/K+ aK+ + KNa+/Ca2+ aCa2+) F
EK+ = E0K+ +
Rθ loge(aK+ + KK+/Na+ aNa+ + KK+/Ca2+ aCa2+) F
ECa2+ = E0Ca2+ +
[19.8]
Rθ loge(aCa2+ + KCa2+/Na+ aNa+ + KCa2+/K+ aK+) F
If the e.m.f. values are measured and the values of the cross-sensitivity constants KX/Y are known, then eqn [19.6] can be solved to give aNa+, aK+ and aCa2+. Table 19.2 lists some physical and chemical equations which are used to model processes. These equations can be used to calculate process variables which are not measured from variables which are measured. Simple examples are the calculation of velocity from acceleration and the calculation of current from voltage. From Table 19.2 we see that we can calculate the volume flow rate Q of fluid flowing through a pipe by measuring the velocity v at an area element dA and integrating over the total cross-sectional area AT of the pipe to give: Q=
0002
AT
v dA 0
[19.9]
19.2 MO D E LLI N G ME T H O D S F O R MULTI VAR I ABLE S Y S T E MS
Table 19.2 Physical and chemical equations used in process models.
F m
Mass
M = ρV
x = ∫ y dt
Volume
V = ∫ Q dt
x = ∫ x dt
Volume flow rate
Q = ∫ v dA
Gas density
ρ=
P Rθ
Liquid density
ρ=
ρ0 1 + α (T − T0)
Liquid viscosity
η = η15 exp  E
Acceleration
y=
Velocity Displacement
Mechanical power W = Fx Energy
E = ∫ W dt
Electric current
i=
Charge
q = ∫ i dt
Electrical power
W = iV
V R
509

1 1    −  R0  T 288  
Gas enthalpy flow rate À = Cp Çθ
If we then divide AT into 12 area elements Ai, i = 1, . . . , 12, and measure the velocity vi at each area element, then eqn [19.9] becomes: 12
Q=
∑v A i
[19.10]
i
i 1
The vi can then be measured by an array of 12 vortex shedding sensors.[1] Each sensor i is located at area element Ai and detects vortices shed downstream from a bluff body. If fi is the frequency of vortex shedding measured by the ith sensor, then from Table 19.1 we have: fi =
S vi or d
d vi = · fi S
[19.11]
where S is the Strouhal number and d the width of the bluff body. From eqns [19.10] and [19.11] we have: Q=
d S
12
∑
fi Ai
[19.12]
i 1
so that Q can be found from the measured fi if the constants Ai, S and d are known.
19.2.3
Multivariable regression equations There are many situations where an equation with a well-defined form does not exist to describe a given physical or chemical effect. One example is the e.m.f. of a thermocouple ET,0 (Section 8.5) with the measured junction at T °C and the reference junction at 0 °C; here the non-linear relation between ET,0 and T has to be described by a polynomial or power series of the form: ET,0 = a1T + a2T 2 + a3T 3 + a4T 4 + . . .
[19.13]
510 TH E INTELLIGENT MULTIVARIABLE M E AS URE ME N T S Y S TE M The coefficients a1, a2, a3, etc., are found from experimental data values of ET,0 and T using regression analysis. In the multivariable modelling problem of Section 19.2.1 it is even more likely that some of the equations will have to be found using regression techniques. We consider a model which has a single output variable y which depends on p input variables {x1, . . . , xp} and we wish to establish a regression equation of the form: y = f(x1, x2, . . . , xp)
[19.14]
To do this we need q sets of experimental data; in general q will be much greater than p. The input data is defined by the matrix:  x11   x21  X =  M  xk1 M   xq1
x12 K x1 p   x22 K x2 p  M M   xk 2 K xkp  M M  xq 2 K xqp 
[19.15]
The output data is defined by the column vector:  y1     y2  Y =  M  y  k M  yq 
[19.16]
Y = f(X)
[19.17]
and
As an example we consider y = f(x1, x2)
[19.18]
where y is a function of two independent variables x1 and x2, and we need to find an equation of the form: y = b0 + b1 x1 + b2 x2 + b11 x 12 + b22 x 22 + b12 x1 x2
[19.19]
where there are six coefficients to be found. We use q sets of data (q 0002 6); the kth set is {yk , x1k , x2k} where k = 1, . . . , q. From eqn [19.19] the kth estimate of y is: 2 yk,est = b0 + b1 x1k + b2 x2k + b11 x 1k + b22 x 22k + b12 x1k x2k
[19.20]
Since the kth observed value of y is yk , the kth value of error is: ek = yk − yk,est
[19.21]
and the kth value of square error is: ek2 = ( yk − yk,est )2
[19.22]
19.2 MO D E LLI N G ME T H O D S F O R MULTI VAR I ABLE S Y S T E MS
511
The sum of square errors SSE is therefore: q
∑ e 2k =
SSE =
k 1
q
∑ (y
k
− yk,est)2
k 1
q
∑ (y
=
k
2 2 − b0 − b1 x1k − b2 x2k − b11 x 1k − b22 x 2k − b12 x1k x2k )2
[19.23]
k 1
The coefficients b0, b1, b2, b11, b22 and b12 are chosen so that SSE has a minimum value; this occurs when: ∂SSE ∂SSE ∂SSE ∂SSE ∂SSE ∂SSE = = = = = =0 ∂b0 ∂b1 ∂b2 ∂b11 ∂b22 ∂b12
[19.24]
These conditions give rise to a set of six simultaneous equations which can be expressed in matrix form: Ab = g
Figure 19.5 Regression matrix equations.
  q    q  ∑ x1k  k =1   q  ∑ x2k  k =1 A=  q  ∑ x 12k  k =1   q 2  ∑ x 2k  k =1   q  ∑ x1k x 2 k  k =1
[19.25]
q
∑ x1k k =1 q
∑ x 12k k =1
q
∑ x2k k =1
q
∑ x 12k k =1
q
q
∑ x1k x 2 k ∑ x 13k
q
∑ x 22 k k =1
q
∑ x1k x 2 k k =1
q
q
k =1
k =1
∑ x1k x 22 k ∑ x 12k x 2 k
k =1
k =1
q
q
q
q
k =1
k =1
k =1
k =1
q
q
q
q
q
k =1
k =1
∑ x1k x 2 k ∑ x 22 k ∑ x 13k k =1
∑ x 12k x 2 k ∑ x 32 k
∑ x 12k x 2 k ∑ x 41k
q
∑ x1k x 22 k k =1
∑ x 12k x 22 k ∑ x 13k x 2 k
k =1
k =1
q
q
q
q
k =1
k =1
k =1
k =1
q
q
q
q
q
k =1
k =1
k =1
k =1
k =1
∑ x1k x 22 k ∑ x 32 k
∑ x 12k x 22 k ∑ x 24 k
q
∑ x1k x 23 k k =1
∑ x 12k x 2 k ∑ x1k x 22 k ∑ x 13k x 2 k ∑ x1k x 23 k ∑ x 12k x 22 k G H H H H H H H H b= H H H H H H H H I
J b0 K K K b1 K K K b2 K K K b11 K K K b22 K K K b12 K L
 q ∑  k =1   q ∑  k =1   q ∑  k =1 g=  q ∑  k =1   q ∑  k =1   q ∑  k =1
      x1k yk     x 2 k yk     x12k yk    2  x 2 k yk    x1k x 2 k yk   yk
                      
512 TH E INTELLIGENT MULTIVARIABLE M E AS URE ME N T S Y S TE M where the matrix A and the column vectors b and g are defined in Figure 19.5.[2] The vector of coefficients b is then given by: b = A−1g where A−1 is the inverse of A. The regression equation [19.19] has a similar form to the inverse sensor equation [19.2]. Comparing the equations, if we set: y = I′,
x1 = U, x2 = I′M = I′I
b0 + b1x1 + b11x 21 = a + K′U + N′(U) b2 = K′I ,
b12 = K′M ,
[19.26]
b22 = 0
then this regression equation can be used as an inverse sensor equation. A more extensive version of eqn [19.19] has been used to model a resonant silicon pressure sensor (Section 9.5.2) and an associated temperature sensor.[3] Here the inverse model equation for the pressure sensor is of the form: 3
y=
3
∑ ∑k x x
i j ij 1 2
[19.27]
i 0 j 0
where:
y = pressure in millibar x1 = output frequency in Hz x2 = temperature sensor output in millivolts kij = calibration constant for a given sensor.
Sixteen calibration constants are therefore required. Here temperature is both a modifying and an interfering input, and the temperature sensor is a forward-biased diode. The voltage drop across the diode depends non-linearly on temperature.
19.2.4
Artificial neural networks In complex multivariable processes and sensor arrays, well-defined physical and chemical equations may not exist to provide a sufficiently accurate model of the system. Furthermore, it may be impossible to predict the form of a suitable multivariable regression equation; this will be especially so if a large number of variables are required to represent the system accurately. In these situations artificial neural networks can be used as a modelling technique. These are empirical models which approximate the behaviour of neurons in the human brain. Figure 19.6 shows a typical artificial neural network. It consists of three layers: the input layer, the hidden layer and the output layer. In this example there are four input variables, x1, x2, x3 and x4, i.e. p = 4, and a single output variable y, i.e. r = 1. Beginning with the input layer, each of the inputs xs is multiplied by the input weighting wIs to give four weighted inputs wIs xs, s = 1, . . . , 4. There are four summing elements in the input layer. Each summer s has five inputs, the four weighted inputs wIs xs and an input bias bIs. The output of each summer is therefore: 4
gs = bIs +
∑w
x,
Is s
s = 1, . . . , 4
[19.28]
s 1
The four gs signals are passed to the hidden layer. Each gs is input to a function block S. The form of this function is:
19.2 MO D E LLI N G ME T H O D S F O R MULTI VAR I ABLE S Y S T E MS
513
Figure 19.6 Example of artificial neural network.
hs(gs) =
Figure 19.7 The balanced sigmoid function.
1 − exp(−gs) 1 + exp(−gs)
[19.29]
and it is referred to as a balanced sigmoid or hyberbolic tangent. It has the S-shaped form shown in Figure 19.7. This function has the advantages of being within the range −1 to +1, being able to approximate almost any other function and having derivatives that can be conveniently calculated. It would, for example, be easier to calculate derivatives of the function h(g) = 1 + g + g 2 + g 3 + g4 but this would be less flexible as an approximation to other functions. The four signals hs are then passed to the output layer. Here each is multipled by an output weighting wos to give four signals wos hs. These are input to a summing element, together with an output bias bo. The output y of the summer is therefore: 4
y = bo +
∑w
h
[19.30]
os s
s 1
and this is the final output of the network. We see that the network of Figure 19.6 involves eight weights w and five biases b. The values of w and b are chosen so that for q sets of experimental input data {x1k , x2k , x3k , x4k},
k = 1, . . . , q
the corresponding network output values yk,est are as close as possible to the observed values yk. This process is called training the network, and one method of achieving this is back propagation. Here the errors ek = yk − yk,est are calculated and the sum of square errors, SSE, is found. The partial derivatives of SSE with respect to w and b are then calculated and the values of w and b adjusted according to the magnitude and sign of these derivatives. This process is continued iteratively until values of w and b are found which minimise SSE.
514 TH E INTELLIGENT MULTIVARIABLE M E AS URE ME N T S Y S TE M One application of neural networks is to model an array of semiconductor sensors used to measure small concentrations of pollutant gases in air.[4] Here there is an array of six sensors; each sensor is formed by depositing a thin film of a given metal phthalocyanine onto a planar configuration of thick film platinum electrodes (Section 8.1.4). Metals such as Zn, Pb, Ag, Co and Cu are used to measure the concentrations of gases such as NO2, Cl2, H2S, HCl, NH3 and water vapour. The resistance of a given metal sensor can depend on the concentration of all the gases present in the mixture; these relationships can also be non-linear. These cross-sensitivity and non-linear effects can be modelled by a neural network which has six inputs, corresponding to the six sensor resistances, and six outputs, corresponding to the six gas concentrations. A process modelling application of neural networks is the prediction of the concentration of NOX components present in the exhaust gas from an industrial boiler.[5] Here 10 input variables were identified as being necessary for an accurate model of the process; these included fuel flow rate, air flow rate, stack gas recirculation flow rate, humidity, temperatures and pressures. A 10-input, single-output neural network gave predicted NOX concentrations within approximately 1% of values obtained with an NOX analyser.
Conclusion This chapter has first discussed the structure of intelligent multivariable measurement systems and then shown the need for inverse sensor and process models. It went on to discuss multivariable modelling methods based on physical/chemical equations, multivariable regression and artificial neural networks.
References [1] [2] [3] [4] [5]
mudd j and bentley j p 2002 ‘The development of a multi-channel vortex flowmeter using a twelve sensor array’, Measurement and Control, vol. 35, no. 10, pp. 296–8. walpole r e and myers r h 1978, Probability and Statistics for Engineers and Scientists, 2nd edn, Collier MacMillan, pp. 314–25. frost d (Druck Ltd) 1999 ‘Resonant silicon pressure transducers’, Sensor and Transducer Conf., NEC Birmingham. jefferey p d et al. 1998 ‘Thick film chemical sensor array allows flexibility in specificity’, Sensor and Transducer Conf., NEC Birmingham. hayes r l et al. ‘Using neural networks to monitor NOX on an industrial boiler’, Advances in Instrumentation and Control, vol. 51, part 1, pp. 259–71.
Answers to Numerical Problems Chapter 2 1
6.06 µV °C −1, 3.61 × 10−3 µV °C −2, 2.59 × 10−6 µV °C −3
2
β = 2946 K, α = 1.86 × 10−4 kΩ, 3.64 kΩ
3
(a) +25.9% (b) 0, 53.3 mV cm−1 V −1 (c) 19.3 mV cm−1
4
13.2%
5
(b) 208.6 Hz, 0.6 Hz
6
RT = 100(1 + 3.908 × 10−3T − 5.82 × 10−7T 2)
7
(a) a = 4.0 mA, K = 1.6 mA bar−1, KI = +0.4 mA °C −1, KM = 0.4 mA bar −1 V −1 (b) 18.0 mA
8
O = 2.0 × 10 −3 I + 1.0
9
O = 8.0 × 10 −4 I + 4.0
10
+ 0.2 V, + 4%
11
− 0.5 mV, −2.5%
12
(a) E = 54.78T
13
KI = 0, KM = 0.005 V kN −1 °C −1
14
KM = 0, KI = 0.02 V °C −1
15
KM = + 6 × 10−6 mA Pa−1 K−1, KI = + 0.02 mA °C −1
16
(a) 19.6 mV, 0.392% (b) 76.3 µV, 0.00153%
17
1.0%
(b) −210 µV, − 0.77%; −109 µV, − 0.40%
Chapter 3 1
+ = − 0.425 °C, σE = 1.93 °C
2
(a) 120.7 Pa (b) −500 Pa
3
(a) (i) 4.95 V (ii) 4.97 V
4
(b) Increase of 10−6 rad V −1
5
(a) + = +5.0 °C, σE = 2.6 °C
6
(a) + = + 0.08 m/s, σE = 0.35 m/s
7
+ = − 0.67 K, σE = 4.87 K
8
− 0.5 kN
9
+ 0.32 m
516 A NS WERS TO NUMERICAL PROBLEMS
Chapter 4 1
−29.4, −10.8, − 0.5, +24.1, + 0.7 °C
2
(a) 5 × 10−3 m N−1, 20 rad s−1, 0.3 (b) 1 cm (c) 1.0 + 0.5 [1 − e−6t(cos 19t + 0.32 sin 19t)] cm
3
50[1.07 sin(10t − 3°) − 1.00 sin 10t] + –503– [2.16 sin(30t − 19°) − 1.00 sin 30t] + –50–5 [1.62 sin(50t − 156°) − 1.00 sin 50t] N
4
(a) 0 to 0.1 rad s−1 (b) 0 to 0.033 rad s−1 (c) 0 to 0.33 rad s−1
5
(a) 10 rad/s, 7.0,
6
(a) ωn = 10 rad/s, ξ = 0.1, V(t) = 0.49[1 − e−t (cos 10t + 0.1 sin 10t)]
7
(a) 50[0.734 sin(10t + 40°) − 1.00 sin 10t] + –503– [1.39 sin(30t − 2°) − 1.00 sin 30t]
0.1 10−2s 2 + 1.4s + 1
(b) Approx.
1 10−4s 2 + 1.4 × 10−2s + 1
+ –50–5 [2.44 sin(50t − 79°) − 1.00 sin 50t] 9
0.707 0.555 sin(0.2t − 45°) + sin(0.3t − 56.3°)] 2 3 (ii) E(t) = 10[0.894 sin(0.1t − 26.5°) − sin 0.1t] + –102– [0.707 sin(0.2t − 45°) − sin 0.2t] + –10–3 [0.555 sin(0.3t − 56.3°) − sin 0.3t] − –10–4 sin 0.4t
(a) (i) TM (t) = 10[0.894 sin(0.1t − 26.5°) +
Chapter 5 1
(a) 2.92 × 108 Ω, 0.15 pH mV −1 (b) −20.5%
2
(a) 2.0 V cm−1 (b) 500 Ω, 50 V
3
−56 Pa
4
0.1s 2 + 10s + 1000 5.1s 2 + 30s + 1100
5
9.09 mV
6
4.5 V, 10.5 kΩ
7
(a) 2.0 V, 1.6 kΩ, 1.72 V (b) 5.0 V, 2.5 kΩ, 4.0 V (c) 8.0 V, 1.6 kΩ, 6.9 V
8
(a) 10 V (b) ETh = 7 V, RTh = 21 Ω
9
(a) 0.5 cm (b) ETh = 5.0 V, RTh = 2.5 kΩ
(c) 399 Ω (c) 4.0 V
Chapter 6 1
(a) 0.15 V, 1.0 V
2
+1.0, +0.6, +0.2, − 0.2, − 0.6, −1.0, − 0.6, − 0.2, + 0.2, + 0.6, +1.0
3
(a) 10−4 W, 10−2 V, 10−2 V (b) −20 dB (d) Increased to +10 dB (e) Increased to +30 dB
4
(a) 3.14 mV, 100 V (b) 4.15 mV, 15.85 mV
5
(a) 1.5 mW (b) 8.5 mW (c) −7.5 dB (d) 55 mV (e) 300 Hz (f ) 92 mV
AN S WE R S TO N UME RI CAL PRO BLE MS
Chapter 7 1
(a) MDT = 6.2 h (b) MTBF = 10 074 h (c) λ = 0.87 yr −1 (d) A = 0.99940
2
(a) 0.62 (b) 0.24 (c) 0.31
3
TLOC = £19 000 for system (1), TLOC = £15 350 for system (2)
Chapter 8 1
(a) 3.91 × 10 −3 °C −1, −5.85 × 10 −7 °C −2 (b) + 0.76%
2
R1 = R3 = 120.0025 Ω, R2 = R4 = 119.9975 Ω
3
88.5, 55.3, 22.1 pF
4
7.6, 3.4 mH
5
521 mH, 5.6 mH
6
3.46 V, 367 Hz; 34.6 V, 3670 Hz
7
(a) −1.07%, − 0.65% (b) 51.8 µV °C −1, 8.68 × 10 −3 µV °C −2 (c) 248 °C
8
(a) 20 N m−1, 0.51 N s m−1 (b) 0 to 1.24 cm (c) 1333 Ω
9
− 0.25 to +0.25 rad
10
(a) 1.2 mm (b) 0 to 0.17 mm
11
(a) G(s) = 0.02
D 5.4 × 1010 A 10−3s D A −3 2 C 1 + 10 s F C s + 4.65 × 103s + 5.4 × 1010 F
(c) G(s) = 0.002
A 0.1s D A 5.4 × 1010 D 2 3 10 C 1 + 0.1s F C s + 4.65 × 10 s + 5.4 × 10 F
12
(a) 100 g (b) (ii) 1 V
13
(a) 625 N m−1 (b) 1 mm (c) 3.40 to 7.58 mH
14
62.5 °C
15
60 Ω
16
12 320 Ω
17
−5 × 10−4, +2 × 10−4
18
119.88 Ω, 120.048 Ω
19
169.9 pF
20
2.78 pF
21
120.126 Ω
22
(a) − 4.87% (b) 255 °C
23
201 °C
24
2 kHz
25
0 to 5 mm
26
60 mV
Chapter 9 1
(a) R2 = 100 Ω, R3 = 5770 Ω, R4 = 5770 Ω (b) R2 = 100 Ω, R3 = 6000 Ω, R4 = 6000 Ω (c) R2 = 100 Ω, R3 = 5000 Ω, R4 = 6000 Ω
2
R2 = 10 Ω, R3 = 1650 Ω, R4 = 1650 Ω
517
518 A NS WERS TO NUMERICAL PROBLEMS 3
(a) (i) 0 to 1.0 V (approx.) (ii) −1.5% (b) 0 to 0.6 V (approx.)
4
R2 = 1000 Ω, R3 = 264 Ω, R4 = 2370 Ω, Vs = 2.40 V
5
(b) 9.1 mV
6
2584
7
0 to 44.2 mV
8
(a) 64.2 nF (b) 0.178 V (c) 0.3%
9
(a) Since fn = 32 Hz, ξ = 0.7, |G( jω)| = 1 up to 10 Hz (c) +0.2 sin 2000 πt and −0.2 sin 2000πt
10
1.9 kΩ
11
RIN = 10 kΩ, RF = 1 MΩ, CIN = 0.159 µF, CF = 0.159 nF
12
(a) 4.8 × 103 N/A, 19.3 N
13
81.6 kHz at 1 mm to 122.4 kHz at 3 mm
14
|G( jωn )| = 20 and arg G( jωn ) = −90°, between 2.60 and 4.56 kHz
15
(b) L = 4.7 mH (c) At fn = 100 kHz, |G( jωn)| = 0.0295, arg G( jωn) = −90° At fn = 120 kHz, |G( jωn )| = 0.0354, arg G( jωn) = −90°
17
(a) 105 to 106 Pa (b) |G( jωn )| = 200, arg G( jωn) = −90° between 1.0 and 3.0 kHz
18
(a) 0.1 V (b) 0.01 V
19
(a) 10 kΩ (b) 25.6 V
20
(a) 5 × 10−4
(b) 1.0, 0.2
(b) 15 mV
Chapter 10 1
(a) (i) ±0.0122%
(a) (ii) 000111000010, 100001101010 (b) 1C2, 86A
(c) (i) ±0.05005% (c) (ii) 0001 0001 0000, 0101 0010 0101 2
(a) 85.3 Ω
3
(a) 0 to 1/τ rad s−1 (d) 0 to π/4τ rad s−1
4
32
(b) 1.97 V
Chapter 11 1
(a) 7.5 rad V −1, 1.0 Hz, 35.6
2
(a) 0 to 15 cm
(b) approx. 10 kΩ in series, 0.15 rad V −1
Chapter 12 1
(a) 10 m s−1 (b) 0.883 kg s−1 (c) 1.5 × 105
2
(a) Re = 1.2 × 10 at max. flow (b) 7.68 cm
3
(a) 0.14 m (b) 2.86 × 105 Pa
4
(a) Re = 2.7 × 106 at max. flow (b) 0.146 m (c) 0.135 m
5
23.8 mV and 4.3 Hz at min. flow, 499 mV and 90.2 Hz at max. flow
6
(a) 69.9 to 699 Hz (b) 104
7
432 pulses m
−3
8
τ = 150 ms,
9
(a) 7200 kg h−1 (b) 9.4 × 105
10
(d) 5.84 cm
5
45 to 450 µs
(c) 14.3 m3
1 = 10 ms fc (d) 7610 kg h−1
AN S WE R S TO N UME RI CAL PRO BLE MS
519
Chapter 13 1
(b) (i) 0.200 to 1.00 bar (b) (ii) 0.202 to 1.007 bar
2
EMAX = 1.9 µJ safe to use with hydrogen–air
Chapter 14 1
EOUT = (3.93 + 6.55 ê v )1/2
2
(a) 4.27 + 0.33 ê v
3
τv = 2.7 s, i.e. bandwidth 0 to 0.06 Hz, therefore unsuitable 1 (a) G(s) = 4 2 6.4 × 10 s + 1068s + 1 |G( jω)| = 0.144 at ω = 2π × 10−3, therefore cannot follow variations 1 (b) G(s) = 2 1600s + 93s + 1 |G( jω)| = 0.91 at ω = 2π × 10−3, can follow variations more closely
4
5
(b) τv = 4 ms, therefore unsuitable
(a) 0 to 10.6 mV
Chapter 15
Figure Soln Prob. 15.1.
1
See Figure Soln Prob. 15.1
2
Period TP = 10 ms, maximum TD − Ts = 37.5 °C, thermocouple constant a1 = 5 × 10−2 mV °C −1, output voltage range = 0 to 46.9 mV
3
0 to 5.97 mV
4
2.10 × 104
5
(a) 10 mW (b) 0.26, 15.1° (c) 0.47, 27.8° (d) 170 µW, 0.998, 169.7 µW (e) 2.2 mW, 0.794, 1.75 mW (f ) Glass 169.7 µW, 93.3 µA; Polymer 1.75 mW, 961 µA
6
500 to 1000 nm, 5.5
520 A NS WERS TO NUMERICAL PROBLEMS
Chapter 16 1
(a) ωn = 1.00 × 106 rad s−1, ω 1 = 1.18 × 106 rad s−1 (b) 50 Ω, −2°52′ at ωn, 14.1 kΩ, −3°27′ at ω 1 (c) 0.02 A V −1, −177°
2
(a) 7.7 × 10−3, 7.7 × 10−3, 7.3 × 10−10, 7.3 × 10−10
3
(a) 1 ms, 0.103 (b) Tw = 30 µs, TR = 10 ms for example
4
(a) 6.7 kHz (b) 6 × 10−4 W
5
(a) See Figure Soln Prob. 16.5(a)
6
See Figure Soln Prob. 16.6
7
(a) 4.67 × 10−7
8
(a) 5.8 mW (b) 4.96 mW (c) 5.8 e−0.01n(0.93)2n mW
(b) 115 µs at 24 m s−1, 1.5 µs at 0.3 m s−1
Figure Soln Prob. 16.5(a).
Figure Soln Prob. 16.6.
Chapter 17 1
(a) 2.63 × 10−2 m s−1, 2.26 (b) 6 s, 7 s, 1.54 (c) average N ≈ 5400, 1.85 × 10−4 m (d) 18% O2, 82% N2
Chapter 18 1
(a) 15 samples s−1 for example (b) 240 samples s−1 (c) 4.17 ms (d) ±0.0122% (e) 2880 baud, 0 to 1440 Hz (f ) 0.54 V
2
(a) 0.333 (b) 0.222 (c) 1.39 × 10−2 (d) 1.39 × 10−2 (e) 0.167
3
280 to 920 Hz
4
(a) 0.309, 0.683, 0.067 (b) 10
Index
absolute encoders, 412 a.c. amplifier, 219–20, 224–7 a.c. carrier system, 224–7 a.c. loading, 80 accelerometers negative feedback type, 72–3 piezoelectric type, 187–8 principle, 177–8 strain gauge type, 180–1 accuracy, 35–47 acoustic impedance, 434–5, 439–40 acoustic matching, 443–5 acoustic power, 439–40 across (effort) variables, 84–93 active sensing elements, 149–50 address bus, 261–3 address code, 261–5 address map, 261–5 address register, 261–5 address signal, 261–5 addressing, 261–5 aliasing, 249 alphanumeric displays, 289–92 ampere, 23–7 amplifiers a.c., 219–20 buffer, 216–17 charge, 185–6 chopper stabilised, 224–7 differential, 216–19 ideal operational, 214–15 instrumentation, 223–4 inverting, 215–16 non-inverting, 216 operational, 214–27 practical limitations, 221–3 relay, 356–7 strain gauge, 218–19 voltage adder, 220–1 voltage follower, 216–17 amplitude modulation (AM), 224–7 amplitude ratio, 61–5 analogue chart recorders, 304–6 analogue filter, 117–18, 219–20
analogue to digital conversion, 256–60 analogue to digital converters (ADC), 256–60 dual slope, 256–8 flash, 259–60 successive approximation, 257–9 anemometer (constant temperature), 374–8 angular accelerometer, 178 angular velocity measurement system, 270–2 angular velocity sensor, 170–2, 414 application layer, 494–8 argument of complex number, 62–5 arithmetic/logic unit, 262–3 artificial neural networks, 512–14 ASCII code, 272–3, 307–9 assembly language, 264–5 attenuation of ultrasonic wave, 440–1 autocorrelation, 104–7, 121 availability, 129 averaging, 119–21 backlash, 13 balanced amplitude modulation, 224–7 balanced bridge, 207–14 band limited white noise, 104 band pass filter, 117–18 band stop filter, 117–18 bandwidth, 71 base units, 23 BASIC language, 266 bathtub curve, 131 baud, 479–84 bellows, 178–82 bilateral transducers, 92–3 binary coded decimal (bcd), 251–2 binary codes, 251–4 binary counter, 254–6 biphase Manchester, 496–7 bit, 251–4 bit rate, 479–84 black body, 389–90 block diagram symbols, 7 bolometer, 404–6 Bourdon tube, 181–2 bridges, see deflection bridges, 205–14
522 INDEX buffer amplifier, 216–17 byte, 251– 4, 262–3 C language, 267– 8 calibration: static, 28–31 cantilever, 179–80 capacitance: electrical, fluidic thermal, 84–7 capacitive coupling, 114–16 capacitive sensing elements, 160–5 capsule, 181–2 carrier gas, 462–7 carrier signal, 224–7, 490–3 cathode ray tube (CRT), 295–9 displays, 295–9 character displays, 289–92 characteristic acoustic impedance, 439–40 charge amplifier, 185–6 chart recorders, 304–6 CHEMFET, 194–6 chemical equations, 507–9 check bits (digits), 487–9 chopped radiation systems, 416 chromatography, gas, 461–79 clock pulses (signals), 254–6, 479–80 closed-loop recorders, 304–306 closed-loop systems, 42–3, 72–3, 228 –30, 235– 6, 304–6, 374 – 8 code word, 251– 4 codes, 251– 4 common mode interference, 108–13, 116–17 common mode rejection ratio, 116–17, 221–2 communication systems, 475–501 comparator, 258, 485 compensation, 41–7 dynamic, 70–3 compensation leads (thermocouple), 174–6 compliance, 85 composition measurement, 159–60, 190–6, 461–73 compressible fluids, 314–20 compressive stress, strain, 156–8 computer software, 264–9 system, 260–3 conduction of heat, 367 cone of acceptance, 399–401 convection, 367–8 convolution integral, 346 Coriolis mass flowmeter, 340–2 correlation, 104 –7, 121, 344 –7 corrugated diaphragm, 181–2 cost penalty function, 141–4 counter, 254–9 critical damping, 59–65 cross-correlation flowmeter, 344–7 crystal oscillator, 431–4 crystal: piezoelectric, 182–8, 428–36 cumulative power function, 102–4 cumulative probability distribution function, 100–1
Curie temperature, 405–7 current source, 82–4 current transmission, 82–4, 108–9 current transmitters, 228–35 Dall tube, 324–8 damped angular frequency, 60–1 damping force, 56–7, 177–82 damping ratio, 56–65 data acquisition systems, 477–8 data bus, 260–3 data link layer, 494–8 data presentation elements cathode ray tube (CRT), 295–9 chart recorders, 304–6 electroluminescence display (EL), 302–4 laser printers, 307–9 light emitting diode (LED), 292–5 liquid crystal displays (LCD), 299–302 paperless recorder, 306–7 pointer-scale indicator, 287–9 decibel, 71, 108–9, 435–6 decoder: 7 segment, 291–2 deflection bridges capacitive differential displacement sensor, 212–13 capacitive level sensor, 212–13 inductive differential displacement sensor, 212–14 resistive katharometer, 378–81 platinum resistance detector, 209 strain gauge, 208–12 temperature difference, 210–11 thermistor, 209–10 Thevenin equivalent circuit, 205–6 demodulation – frequency, 491 demodulation – phase sensitive amplitude, 226–7 denary numbers, 251–4 density transducer, 239–40 derived units, 24–5 deterministic signal, 97–8 diaphragm, 181–2 dielectric, 160–5 differential amplifier, 216–19 differential capacitance displacement sensor, 161–2 differential equations, 51–7 differential pressure flowmeters, 321–9 differential pressure (D/P) transmitters applications, 231–2 closed loop-electronic, 228–30 open loop-electronic, 230–3 pneumatic-torque balance, 357–61 smart, 233–5 vibrating plate resonator, 239–40 differential reluctance displacement sensor, 166–8 differential transformer, 168–70 digital codes, 251–4 digital communications, 475–99 digital displays, 289–304
INDEX
digital filter, 275–82 digital filtering, 275–82 digital printers, 307–9 digital signals parallel, 478–9 serial, 479–90 digital to analogue converter (DAC), 256–7 diodes, 292–5, 407–9 direct model, 15–17, 44–5 discharge coefficient, 323–9 displacement sensors, 149–70, 411–13, 417–22 displays, 287–304 Doppler effect, 446–7 Doppler flowmeter, 451–3 drift (amplifier), 221–2 dual slope ADC, 256–8 dummy leads, 209, 245 dynamic characteristics of first order elements, 51–5, 58–9, 61–3 of second order elements, 56–7, 59–61, 63–5 dynamic compensation, 70–3 dynamic errors, 65–70 earth loops, see multiple earths effort variable, 84–93 elastic modulus, 156 elastic sensing elements, 177–82 electrical oscillators, 236–8 electrochemical sensing elements, 190–6 electroluminescence displays, 302–6 electromagnetic coupling, 111, 114 electromagnetic flowmeter, 343–4 electromagnetic radiation, 385 electromagnetic sensing elements, 170–2, 343–4 electromagnetic shielding, 114 electromechanical oscillators, see resonators electronic transmitters, 228–35 electrostatic coupling, 112 electrostatic deflection (in CRT), 295–6 electrostatic screening and shielding, 114–16 emissivity, 390–1 encoders, 412–13 encoding, 251– 4 environmental effects, 11–13 error bands, 14–15 error detection and correction (PCM), 487–9 error probability function measurement, 36–41 PCM, 484 –7 error reduction techniques dynamic, 70–3 steady state, 41–7 errors in measurement dynamic, 65–70 steady state, 35–41 errors in PCM transmission, 484–90 Ethernet, 499 estimation of measured value, 44–7
even parity, 487–9 exclusive OR, 488 expansibility factor, 323–6 extension leads, 174–6 failure rate data, 135–8 definition, 126–31 function, 131–2 models, 135–8 farad, 23–6 Faraday’s law, 170 feedback accelerometer, 72–3 constant temperature anemometer, 374–8 differential pressure transmitter, 228–30 in dynamic compensation, 72–3 in static error reduction, 42–4 oscillators and resonators, 235–40 pneumatic transmitters, 357–61 field effect transistors (FET), 194–6 Fieldbus, 495–9 filtering, 117–18 filters, frequency response, 117–18 first-order differential equation, 275–82 first-order elements, 51–5 sinusoidal response, 61–3 step response, 58–9 flapper/nozzle, 353–6 flash ADC, 259–60 flat screen displays, 299–304 floating point, 253–4 flow measurement systems, 313–47 flow variable, 84–93 flowmeters Coriolis, 340–2 cross-correlation, 344–7, 453–4 differential pressure, 321–9 electromagnetic, 343–4 inferential mass, 339–40 turbine, 330–2 ultrasonic Doppler, 451–3 ultrasonic transit time, 454–5 vortex, 332–7 fluid mechanics, 313–19 fluid velocity sensor hot wire and film sensors, 374–8 pitot tube, 319–21 flux (magnetic), 165–8 foil strain gauge, 157–8 force balance systems, 228–30, 357–61 force sensing elements, 177–81, 182–7 Fourier analysis, 67–70 Fourier transform, 106–7, 277 frame, 498–9 frame format, 498–9 frequency demodulation, 491 frequency modulation (FM), 490–3
523
524 INDEX frequency response of amplifiers, 219–23 of first and second order systems, 61–5 frequency shift keying (FSK), 490–3 frequency signals, 235–40 frequency to digital conversion, 254–6 Fresnel zone plate, 417–19 fundamental frequency, 67–70 fundamental interval, 153 full-scale deflection (FSD), 10–11 gain (amplifiers) closed loop, 215–19 d.c. open loop, 214 gain: bandwidth product, 222–3 gain and phase conditions, 70, 236 gallium aluminium arsenide (GaAlAs), 391–3 gallium arsenide phosphide (GaAsP), 295, 391–3 gallium phosphide (GaP), 295, 391–3 gas sensors, 159–60, 193–4 gauge factor, 157–8 gauge pressure, 189–232 Gaussian probability density function, 17–20 gears, 13 graphic displays, 292, 297–304 Hall effect sensors, 196–7 HART protocol, 496–500 harmonics, 67–70 heat balance equation, 51–2, 369–71, 404 heat flow rate, 51–2, 369–71, 400 heat transfer coefficient, 367–8 heat transfer effects, 367–81 henry, 24 hexadecimal code, 253 operand, 264–5 to decimal conversion, 272 high-level language, 265–9 high pass filter, 117–18 histogram, 30–1, 100–1 hum, 110 humidity sensors, 154–5, 161–3 hysteresis, 13 ideal straight line, 9–10 identification of characteristics dynamic, 58–65 static, 28–31 impact pressure, 319–20 impedance acoustic, 439–40 general definition, 84–7 impulse response, 345–6 incremental encoders, 412 index register, 262–3 indicators, pointer scale, 287–9 inductive sensing elements, 165–70
inertance, 85 information bits (digits), 487 infrared detectors, 403–9 infrared radiation, 385–93 initiate conversion, 258–9 injection laser diode (ILD), 392–3 input impedance, 77–80, 214 input offset voltage, 214, 221 input/output interface, 261 instruction, 264–8 instruction decoder, 262–3 instruction register, 262–3 instrumentation amplifier, 223–4 integrator, 185–6, 269, 279–80 intelligent systems, 503–14 intelligent transmitters, 233–5 interference – effect on measurement circuit, 107–17 interference signals common mode, 110–13 series mode, 110–13 interfering input, 11–12 interferometers, 419–22 International Practical Temperature Scale (IPTS), 27–8 intrinsic safety, 362–3 inverse element models, 44–6, 504–5 inverting amplifier, 215 ion selective electrodes, 190–3 j operator, 62 Johnson noise, 110 junction, thermoelectric, 172–6 junction diode, 292–5, 407–9 katharometer, 378–81 kelvin, degree, 23 kilogram, 23 kinetic energy (fluids), 317–18 lag, first order, 52–5 laminar flow, 315–16 Laplace transform definition 52–3 tables, 52–3 lasers, 392–3 laser printers, 307–9 lease significant bit (LSB), 251–4 least squares fit, 28–9, 509–12 level measurement, see liquid level light-emitting diode (LED), 292–5, 391–2 linear differential equations, 52, 55, 56–7 linear systems frequency response, 63 principle of superposition, 68 linear variable differential transformer (LVDT), 168–70 linearity, 9–11 liquid crystal displays (LCD), 299–302 liquid level measurement, 161–3, 231–2, 456–7 live zero, 228
INDEX
load cells, 179–80 loading electrical, 77–84 generalised, 84–93 logarithmic amplitude ratio (decibel), 71, 108–9 longitudinal wave (sound), 435–6 look-up table, 272–4 low pass filter, 117–18 magnetic circuit, 165–8 magnetic flux, 165 magnetic reluctance, 165 magnetomotive force (mmf ), 165 magnitude of complex number, 62–3 mass, seismic, 177–8, 180–1, 187–8 mass flow rate, 317, 326, 339–42 mean down time (MDT), 127–9 mean time between failures (MTBF), 129 mean time to fail (MTTF), 126–7 mean value of element output, 19–20 of probability density function, 17–20, 101 of random signal, 99 of system error, 36–41 measured value, 3–4 measured variables, 3 memory random access (RAM), 261–3 read only (ROM), 261–3 metre, 23–6 Michelson interferometer, 420–2 microcontroller in chromatography, 466–73 in error reduction, 44–7 in flow measurement, 339–40 in speed measurement, 270–2 software, 264–5 system, 263 microprocessor, 261–3 mineral insulated thermocouple, 176 modifying input, 11–13 models process, 505–6 element, 504–5 modelling methods, 507–14 modulation amplitude (AM), 224–7 frequency (FM), 235, 490–3 modulo 2 addition, 487–9 modulus of elasticity, 156 moment of inertia, 177–8 monitors, 295–304 most significant bit (MSB), 251–3 motors, 304–6 moving coil indicator, 287–9 multiple earths, 111–16 multiplexer, 477–8 multiplexing, time division, 477–8, 483, 495–6
525
multiplier, 7, 264–72, 281–2 multivariable modelling , 507–14 systems, 503–7 narrow band radiation thermometer, 410–11 ultrasonic link, 436 National Measurement System, 23–8 National Physical Laboratory (NPL), 23–8 natural frequency, 56–7 negative feedback, 42–3, 72–3, 228–30, 235–6, 304–6, 374–8 noise effect on measurement circuit, 107–9 methods of reduction, 113–21 sources, 110 statistical quantities, 98–107 non-linearity definition, 10–11 methods of compensation, 41–7 of deflection bridges, 206–12 of loaded potentiometer, 80–2 non-return to zero (NRZ), 496–7 normal distribution, see Gaussian probability density function Norton equivalent circuits, 82–4 nozzle (flow), 324–7 nozzle/flapper, 353–6 number systems, 251–4 Nusselt number, 367–8 Nyquist sampling theorem, 247–9 observation period, 97–8 odd parity, 487–9 offset voltage, 214, 221 opcode, 264–5 open-loop dynamic compensation, 72 open-loop transmitter, 230–3 opposing environmental inputs, 41–2 optical fibres, 395–8, 413–15 optical measurement systems, 385–422 optical radiation, 385 optical sources, 387–93 optimum damping ratio, 60, 72 orifice plate, 322–9 oscillators crystal, 433–4 electrical, 236–8 electromechanical, 238–40 voltage controlled (VCO), 490–1 oscilloscope, cathode ray, 295–9 OSI model, 494–5 overshoot, 60–1 paperless recorders, 306–7 parallel digital signals, 254–60, 478–9 parallel impedances, 82–4
526 INDEX parallel reliability, 133–5 parallel to serial conversion, 479–80 parity check digits, 487–9 partial fractions, 59–62 passive axis, 157 periodic signals, 67–70 permeability, 165–8 permittivity, 160 persistence, 296 pH electrode, 191–2 phase difference, 61–5 phase locked loop (PLL), 491 phase-sensitive demodulator (PSD), 226–7 phosphor, 295–9, 302– 4 phosphorescence decay, 295–6 photon detectors, 407–9 photoconductive, 407–9 photovoltaic, 407–9 physical equations, 507–9 piezoelectric effect, 182–3 sensing elements, 182–8, 428–36 pitot-static tube, 319–21 pixel matrix, 292–3 Planck’s Law, 389–90 platinum resistance sensor, 152–5 pneumatic displacement sensor, 353–6 measurement systems, 353–62 relay amplifier, 356–7 torque-balance transmitters, 357–61 pointer–scale indicator, 287–9 Poisson’s ratio, 156–7 polarisation, 299–302 polynomial, 11, 28 –9, 509–12 potential energy (fluid), 317–18 potentiometer displacement sensor, 149–52 power acoustic, 439–40 cumulative function, 102–4 power spectral density function, 102–4 power spectrum, 102–4 Prandtl number, 367–8 pressure absolute, 189, 232 differential, 189, 232 gauge, 189, 232 pressure energy, 318 pressure sensing elements, 160–4, 177–82 pressure tappings, 322–7 primary sensing elements, 149–50 printers, 307–9 probability for random signal, 100–1 of failure, 125–6 probability density function definition, 14 –15, 100 –1 for measurement system error, 36–7
for random signal, 100–1 for repeatability, 17–19 for tolerance, 19–20 process, 3, 503–4 process models, 505–6 program, 264–9 pulse code modulation (PCM), 479–87 pulse echo system, 447–51 pyroelectric detectors, 405–7 pyrometer thermal radiation, 409–11, 416–17 quantisation, 249–51 quantum level, 249–51 quartz, 434–6 R-2R ladder network, 256–7 radiance, 387 radiation sources, 387–93 radiation (thermal) measurement systems, 409–11, 416–17 random signals introduction, 97–8 statistical characteristics, 98–107 RAM, 261–3 range, 9 raster display, 296–8 recorders chart, 304–6 paperless, 306–7 redundancy in PCM error detection, 487–9 to improve reliability, 133–5 reference junction, 172–6 reference junction compensation circuit, 174–5 refresh displays, 295–9, 302–4 register address, 262–3 data, 262–3 instruction, 262–3 shift, 479–80 regression analysis, 28–9, 509–12 regression matrix equations, 509–12 relative permeability, 165–8 relay, pneumatic, 356–7 reliability data, 135–8 design and maintenance, 139–40 fundamental principles, 125–35 reluctance displacement sensors, 166–8 reluctance (magnetic), 165 repair time, 127–30, 139–40 repeatability definition, 17–19 measurement of, 30–1 resistive deflection bridges, 206–12 resistive sensing elements, 149–60 resolution, 13–14, 249–51 resonance, 64, 235–9, 435–6, 443–5
INDEX
resonant frequency, 64, 235–9, 435–6, 443–5 resonators vibrating plate, 239 vibrating tube, 239 response sinusoidal input, 61–5 step input, 58–61 Reynolds number, 315–16 rise time, 301 ROM, 261–3 root mean square (r.m.s.) value, 99 rotational mechanical systems, 177–8 sample and hold device, 249–50 sampling, 247–9 scaled variables, 273–5 scales design of, 287–9 Schmitt trigger, 254–6, 270–1 screening, 114–16 second-order elements, 56–7 sinusoidal response, 63–5 step response, 59–61 seismic mass, 177–8, 180–1, 187–8 semiconductor diodes, 292–5, 407–9 photon detectors, 407–9 strain gauges, 158 temperature detectors, 153–6 sensor array, 159–60, 514 sensing elements (sensors), 149–97 sensitivity (steady state), 11–12 serial digital signalling, 479–87 shear modulus of elasticity, 156, 181 shear stress–strain, 156, 313–14 shielding, 114–16 shift register, 479–80 shot noise, 110 SI units, 23–8 sidebands, 225– 6, 492 signal autocorrelation function, 104–7, 121 averaging, 119–21 conditioning, 205–40 deterministic, 97–8 filtering, 117–18 modulation, 118 –19, 224 –7, 235, 490 –3 processing, 247–82 random, 98–107 signal conditioning, 205–40 signal processing, 247–82 signal-to-noise ratio, 107–9 sinusoidal response, 61–5 span, 9 spectral density, 102–4 spectrum of a signal, 102–4 speed measurement system, 270–2 speed sensing elements, 170–2, 413–14
spring, 177 standard deviation from calibration experiment, 30–1 of error distribution, 36–40 of random signal, 99 of repeatability distribution, 17–19 of tolerance distribution, 19–20 standards, 21–8 static characteristics statistical, 17–20 systematic, 9–17 static (steady-state) calibration, 21–31 statistical characteristics of element, 17–20 statistical representation of random signal, 98–107 steady-state characteristics, see static characteristics steady-state compensation, 41–7 Stefan–Boltzmann constant, 389 step response, 58–61 stiffness, 177–82 Stolz equation, 325 storage elements, 261–3 strain definition, 156–7 in elastic elements, 156–8, 177–82 types, 156 strain gauge, 156–8, 179–81 stress, 156 subtraction, 7, 216–18, 264–8 successive approximation ADC, 257–9 summing amplifier, 220–1 superposition principle, 68 tachogenerator, variable reluctance, 170–2 telemetry, 493–9 temperature coefficient of resistance, 152–5 temperature measurement, thermal radiation, 409–11, 417 temperature sensing elements resistive, 152–5 thermoelectric, 172–6 temperature standards, 27–8 tensile strain, 156 thermal capacitance, 84–7 thermal detectors, 368–81, 404–7 thermal noise, 110 thermal resistance, 84–7 thermal sensing elements dynamics of, 369–71 fluid velocity detectors, 371–8 thermal conductivity detectors, 378–81 thermal power detectors, 404–7 thermistor, 153–4 thermocouple, 172–6 thermoelectric sensing elements, 172–6 thermopile, 404–5 Thévenin equivalent circuits, 77–82, 205–6 Thévenin impedance, 77–82, 205–6 Thévenin theorem, 77
527
528 INDEX thickness measurement (ultrasonic), 447–9 through variables, 84–94 time constant electrical, mechanical, fluidic elements, 55 thermal elements, 52 time delay measurement, 344–7 time division multiplexing (TDM), 477–8, 483, 495–6 time standards, 23–6 torque-balance transmitters, 228–30, 357–60 torque sensing element, 180–1 total lifetime operating cost (TLOC), 141–4 traceability, 22–3 transducer, see density transducer and sensing elements transfer function definition, 54 first-order, 54–5 second-order, 56–7 transformer linear variable differential, 168–70 transient response, see step response transit time flowmeters, 454–5 translational mechanical system, 84–7 transmission bandwidth FSK, 493 PCM, 484 transmission characteristics optical lens materials, 400 optical transmission media, 393–8 ultrasonic transmission medium, 436–45 transmission of data, 475–500 transmitters current, 228–35 pneumatic, 357–61 smart, 233–5 ultrasonic, 428–36 transverse wave (ultrasonics), 438–9 triple point of water, 23 true value (of measurement variable), 22 turbine flowmeter, 330–2 turbulent flow, 315–16 two-phase flow measurement, 342, 344–7 two port networks, 87–93, 428–31
units SI base, 23 SI derived, 24–8 unreliability, 126
ultrasonic imaging, 447–51 measurement systems, 427–55 transmission link, 427–8 transmission principles, 436–55 transmitters and receivers, 428–31 unavailability, 130 United Kingdom Accreditation Service (UKAS), 22
Y deflection (in CRT), 295–9, 450–1 Young’s modulus, 156
value measured and true, 3–4 variables, measured, 3 velocity measurement system, 270–2 velocity of approach factor, 323–6 velocity of fluid pitot tube sensor, 319–21 thermal sensor, 371–8 velocity sensing elements, 170–1, 413–14 vena contracta, 322–3 Venturi tube, 322–7 virtual instrument concept, 503–7 software, 268–9 viscosity, 313–14 viscous damping, 56–7, 177–82 visual (video) display unit (VDU), 295–9 volt, standard of, 27 voltage balance recorder, 304–6 voltage controlled oscillator (VCO), 490–1 voltage follower, 216–17 voltage summer, 220–1 volume flow rate, 316–17 vortex flowmeters, 332–8 vortex shedding, 332–3 wave, acoustic plane, 436–43 waveform deterministic, 97–8 periodic, 67–70 random, 97–8 wavelength, 436–8 weighting resistors, 256–7 Wheatstone bridge, 206–7 white noise, 104 Wiener–Khinchin relationship, 106–7 X deflection (in CRT), 295–9, 450–1
Z modulation (in CRT), 295–9, 450–1 Z transform, 275–81 zener barrier, 362–3 zero intercept, 9–10 zirconia, 193–4

Measurement Systems • A system can be defined as an arrangement of parts within some boundary which work together to provide some form of output from a specified input or inputs.
environment Output/s Inputs System boundary
• The boundary divides the system from the environment and the system interacts with the environment by means of signals crossing the boundary from the environment to the system, I.e. inputs , and signals crossing the boundary from the system to the environment, I.e. outputs
• A useful way of representing a system is as a block diagram. Ex. – Motor system – Amplifier system
Interconnected system Output from CD player
Output from amp
Input to amp
Input to speaker Output
Input CD player
Amplifier Sound
Electrical signals
Bigger electrical signals
In drawing a system as a series of interconnected blocks, it is necessary to recognize that the lines drawn to connect boxes indicate a flow of information in the direction indicated by the arrow and not necessarily physical connections.
• The purpose of an instrumentation system used for making measurements is to give the user a numerical value corresponding to the variable being measured. • An instrumentation system for making measurements has an input of the true value of the variable being measured and an output of the measured value.
Importance of effective measurement in process industry Failure to measure effectively the level of liquid in bottom of the tower lead to --- Fire --- Explosion
Instrument
 Typical components of instrument  A Sensor: (measures a physical quantity and converts it into a signal)  A Modifier (Change the type of signal)  A Display unit (transmitting arrangement )
Perhaps the best advice for engineering students is that “instruments are always incorrect”!!!!.
Functional Elements of an Instrument
Process/ Measured medium
Primary Sensing Element
Data Transmission Element
Variable Conversion Element
Variable Manipulation Element
Data Presentation Element Observer
Functional Elements of an Instrument (Cont‘d)
Typical Example:
• An instrument contain various parts that perform prescribed functions in converting a variable quantity or condition into corresponding indication. a) Primary sensing element: of an instrument receives energy from the measured medium and produces an output depending in someway on the value of the measured quantity. b) Variable Conversion element: merely converts the output signal of the primary sensing element (voltage or displacement) into a more suitable variable or condition useful to the function of the instrument c) Variable manipulating element: manipulates the signal represented by some physical variables to perform the intended task of an instrument. d) Data Transmission Element: transmits the data from one element to another, as simple as shaft and bearing assembly or as complicated as telemetry system e) Data Presentation Element: performs the translation function such as simple indication of a pointer moving over a scale.
Functional Elements of an Instrument (Cont‘d)
Fluid Temperature Measured Quantity
Process/ Measured medium
Tubing
Temperature Tube
Primary Sensing Element
Variable Conversion Element
Pressure
Data Transmission Element
Pressure
Variable Conversion Element
Motion
Variable Manipulation Element
Motion
Data Presentation Element Observer
Spiral Bourdon Tube
Linkage Gear
Scale & Pointer
Performance Characteristics of Instruments
 Static characteristics  Dynamic characteristics
• The performance characteristics of an instrument are very necessary for choosing the most suitable instrument for specific measuring task. • Static Characteristics: considered for instruments used to measure an unvarying process conditions • Dynamic Characteristics: for measuring quantities that fluctuates with time. •
The relative importance of each issue depends upon the specific application; for example, one application might require excellent accuracy, while another might require only moderate accuracy, but high reliability. • Generally, we find that the greater the requirements for good performance, the higher the cost for purchase and maintenance. Therefore, we must find the proper balance of performance and cost, rather than always specify the best performing sensor
Performance Characteristics of Instruments (Cont‘d)
 Static characteristics Static characteristics of an instrument includes;  Accuracy  Precision  Repeatability  Range  Resolution 
Others ( Sensitivity , Dead zone etc.)
Elements of an instrumentation system • 1. Sensor – element which is effectively in contact with the process for which a variable is being measured and gives an output which depends in some way on the value of the variable and which can be used by the rest of the measurement system to give a value to it.
Sensor : Thermocouple Output: e.m.f. Input: temperature
• 2. Signal Processor • This element takes the output from the sensor and converts it into a form which is suitable for display or onward transmission in some control system. • The term signal conditioner is used for an element which converts the output of a sensor into a suitable form for further processing
• Thus in the case of resistance thermometer there might be a signal conditioner, a Wheatstone bridge, which transforms the resistance change into a voltage change, then an amplifier to make the voltage bi enough for display
Input: Resistance
Wheatstone bridge
Out Amplifier Voltage change
Larger voltage change
• 3. Data presentation • This presents the measured value in a form which enables an observer to recognize it. Input:
Output: Display
Signal from system
Signal in observable form
• Alternatively, the signal may be recorded, e.. On the paper of a chart recorder or perhaps on magnetic disc, or transmitted to some other systems such as control system. Display
Input Sensor
True value of variable
Signal processor
Record Transmit Output measurement value of variable
• Transducers are defined as an element that converts a change in some physical variable into a related change in some other physical variable. • It is generally used for an element that converts a change in some physical variable into an electrical signal change. • A measurement system may use transducers, in addition to the sensor, in other parts of the system to convert signals in one form to another form.
Sensor
A
Temperature signal
Signal Processor
B Resistance change
Data presentation
C Current change
Measurement of pointer across a scale
Static Characteristics
Static characteristics of an instrument includes; 1. Accuracy Accuracy is the degree of conformity of the measured value with the accepted standard or ideal value, which we can take as the Accuracy is usually expressed in engineering units or as a percentage of the sensor range, for example: true physical variable.  Thermocouple temperature sensor with accuracy of  1.5 K.
 Orifice flow meters with accuracy of 3% of maximum flow range
• Conditions: – the same measurement procedure – the same observer – the same measuring instrument, used under the same conditions – the same location – repetition over a short period of time.
Precision (repeatability) and accuracy (deviation
Not precise Not accurate
Precise Not accurate
Not precise Accurate
Precise Accurate
Accuracy is a consequence of systematic errors (or bad calibration) accuracy and precision may depends on time (drift)
Resolution and accuracy
• Resolution expresses how many different levels can b distinguished • It is not related to accuracy
2.1.3.1 Analog mechanical position potentiometer capacitive balanced transformer (LVDT) (linear or sin/cos encoder) strain gauges piezo-electric
+cheap, -wear, bad resolution +cheap, -bad resolution +reliable, robust - small displacements +reliable, very small displacements +extremely small displacements
Static Characteristics • Accuracy is the extent to which the value indicated by a measurement system or element might be wrong. • Accuracy is often expressed as a percentage of the full range output or full-scale deflection (f.s.d.)
• For example, a digital thermometer may have an accuracy of ±0.1oC. • Is quoted in its specification as : – Full scale accuracy – better than 2%
• For example, a system might have an accuracy of ±1% of f.s.d. • If the full-scale deflection is, say, 10A, then the accuracy is ±0.1 A.
• Accuracy is the indicator of how close the value given by a measurement system can be expected to be to the true value. • The accuracy is a summation of all the possible errors that are likely to occur, as well as the accuracy to which the system or element has been calibrated.
• Accuracy is needed for some variables, such as product quality, but it is not required for others such as level in a large storage tank.
Static Characteristics
Static characteristics of an instrument includes; 2. Precision 
Precision is the degree of exactness for which an instrument is designed or intended to perform.

It is composed of two characteristics; 1. Conformity 2. Number of significant figures
• The term precision is used to describe the degree of freedom of a measurement system from random errors. • Thus, a high precision measurement instrument will give only a small spread of readings if repeated readings are taken of the same quantity.
• A low precision measurement system will give a large spread of readings. • For example consider the following two sets of readings obtained for repeated measurements of the same quantity by two different instruments: • 20.1mm,20.2mm,20.1mm, 20.0mm,20.1mm,20.1mm,20.0mm • 19.9mm,20.3mm,20.0mm,20.5mm,20.2mm,19.8mm,20.3mm
• The results of the measurement give values scattered about some value. The first set of results shows a smaller spread of readings than the second and indicates a higher degree of precision for the instrument used for the first set.
Static Characteristics
Static characteristics of an instrument includes; 3. Repeatability
The closeness of agreement among a number of consecutive measurements of the same variable (value) under the same operating conditions, approaching in the same direction. The term “approaching in the same direction” means that the variable is increasing (decreasing) to the value for all replications of the experiment.
• The terms repeatability and reproducibility are ways of talking about precision ins specific contexts. • The term repeatability is used for the ability of a measurement system to give the same value for repeated measurements of the same value of a variable.
• Common cause of lack of repeatability are random fluctuations in the environment,e.g. changes in temperature and humidity. • The error arising from repeatability is usually expressed as a percentage of the full range output.
• For example, a pressure sensor might be quoted as having a repeatability of ±0.1% of full range. Thus with a range of 20kPa this would be an error of ± 20Pa. • The term reproducibility is used to describe the ability of a system to give the same output when used with a constant input with the system or elements of the system being disconnected from its input and then reinstalled. The resulting error is usually expressed as a percentage of the full range output.
• High precision does not mean high accuracy. A high precision instruments could have low accuracy.
Static Characteristics
Static characteristics of an instrument includes; 4. Reproducibility
• The closeness of agreement among a number of consecutive measurements of the same variable (value) under the same operating conditions over a period of time, approaching from both directions. The period of time is “long”, so that changes occurring over longer times of plant operation are included.
Reproducibility includes hysteresis, dead band, drift and repeatability.
Drift
• Gradual change in instruments measurements. OR • Measure of difference in repeatability. • Under laboratory conditions drift of an element can be determined by one of two ways; 1. Point drift 2. Calibration drift
Point Drift
• By maintaining exact operating and load conditions , monitoring of output variations for a fixed input signals as a function of time is called point drift. • Used for stable process conditions
Calibration Drift
• By maintaining input signals, operating conditions, a load approximately constant comparison of calibration curves at the beginning and at specified intervals of time is called Calibration drift. • Used for varying process conditions
Dead Zone
• Dead zone is the largest range of values of a measured variable to which the instrument does not respond. • This is sometimes called dead spot and hysteresis.
Backlash  Backlash or mechanical hysteresis is defined as that lost motion or free play which is inherent in mechanical elements such as gears, linkages or other mechanical transmission devices that are not rigidly connected.
Static Characteristics
Static characteristics of an instrument includes; 5. Range/Span  Range represents the minimum and maximum values which can be determined by an instrument or equipment.  Difference between upper and lower range is known as Span.  Span can be the same for two different range instruments.
If a chemical reactor typically operates at 300 C, the engineer might select a range of 250-350 C. Since the reactor will be started up from ambient temperature occasionally, an additional sensor should be provided with a range of -50 to 400 C.
Static Characteristics
Static characteristics of an instrument includes; 5. Linearity This is the closeness to a straight line of the relationship between the true process variable and the measurement. Lack of linearity does not necessarily degrade sensor performance. If the nonlinearity can be modeled and an appropriate correction applied to the measurement before it is used for monitoring and control, the effect of the non-linearity can be eliminated. Linearity is usually reported as non-linearity, which is the maximum of the deviation between the calibration curve and a straight line positioned so that the maximum deviation is minimized
• Typical examples of compensating calculations are the square root applied to the orifice flow sensor and the polynomial compensation for a thermocouple temperature sensor. • The engineer should not assume that a compensation for non-linearity has been applied, especially when taking values from a history database, which does not contain details of the measurement technology.
Static Characteristics
Static characteristics of an instrument includes;
6. Reliability Reliability is the probability that a device will adequately perform (as specified) for a period of time under specified operating conditions. Some sensors are required for safety or product quality, and therefore, they should be very reliable.
If sensor reliability is very important, the engineer can provide duplicate sensors, so that a single failure does not require a process shutdown
• Probability = number of occurrences of the event/ total number of trials • A high reliability system will have a low failure rate. • Failure rate is the number of times during some period of time that the system fails to meet the required level of performance, i.e. • Failure rate= number of failures / number of systems observed x time observed
• A failure rate of 0.4 per year means that in one year, if ten systems are observed, 4 will fail to meet the require level of performance. • If 100 systems are observed, 40 will fail to meet the required level of performance. • Failure rate is affected by environmental conditions.
• For example, the failure rate for a temperature measurement system used in hot, dusty, humid, corrosive conditions might be 1.2 per year, while for the same system used in dry, cool, non-corrosive environment it might be 0.3 per year.
• Reliability is affected by maintenance and consistency with process environment. Also, some sensors are protected from contact with corrosive process environment by a cover or sheath (e.g., a thermowell for a thermocouple), and some sensors require a sample to be extracted from the process (e.g., a chromatograph).
Calibration
Assigning standard values to an equipment is calibration.
The process for determination by measurement or comparison with a standard of correct value scale reading on a meter or other measuring instrument.
Error • The term error is used for the difference between the result of the measurement and the true value of the quantity being measured, i.e. • Error = measured value – true value
• Thus if the measured value is 10.1 when the true value is 10.0, the error is +0.1. • If the measured value is 9.9 when the true value is 10.0, the error is –0.1.
•
Errors can arise in number of ways:
1. Hysteresis error 2. Non-linearity error 3. Insertion error
Actual relationship Assumed relationship Increasing Non
error error
Instrument Reading
Decreasing Value measured
Actual relationship Assumed relationship
Non
error error
Measured Value
True Value
• Hysteresis error- is used for the difference in outputs given from the same value of quantity being measured according to whether that value has been reached by a continuously increasing change or a continuously decreasing change.
• Thus, you might obtain a different value from a thermometer used to measure the same temperature of a liquid if it is reached by the liquid warming up to the measured temperature or it is reached by the liquid cooling down to the measured temperature.
• Non-linearity error is used for the error that occurs as a result of assuming a linear relationship between the input and output over the working range, i.e. a graph of output plotted against input is assumed to give a straight line.
• Few systems or elements, however, have a truly linear relationship and thus errors occur as a result of this assumption of linearity.
• Linearity error is usually expressed as a percentage error of full range or full scale output.
Insertion error • When a cold thermometer is put in to a hot liquid to measure its temperature, the presence of the cold thermometer in the hot liquid changes the temperature of the liquid. • The liquid cools and so the thermometer ends up measuring a lower temperature than that which existed before the thermometer was introduced.
• The act of attempting to make the measurement has modified the temperature being measured. • This effect is called LOADING and the consequence as an insertion error. • If we want this modification to be small, then the thermometer should have a small heat capacity compared with that of the liquid. A small heat capacity means that very little heat is needed to change its temperature.
• Example of loading is when an ammeter is inserted into a circuit to make a measurement of the circuit current, it changes the resistance of the circuit and so changes the current being measured. • Ex. Two voltmeters are available, one with a resistance of 1 K ohm and the other 1 Mohm. Which instrument should be selected if the indicated value is to be closed to the voltage value that existed across a 2 Kohm resistor before the voltmeter was connected across it.
• The 1Mohm voltmeter should be chosen. This is because when it is in parallel with the 2Kohm, less current will flow through it than if the 1 Kohm voltmeter had been used and so the current through the resistor will be closer to its original value. Hence the indicated voltage will be closer to the value that existed before the voltmeter was connected into the circuit.
Static Error
•
Static errors are generally of three types;
1. Mistake or gross error (human mistakes) 2. Systematic errors (instrumental or environmental errors) 3. Random or accidental errors (unknown)
Range • The range of variable of system is the limits between which the input can vary. • For example, a resistance thermometer sensor might be quoted as having a range of –200 to +800oC. • A meter might have dual ranges i.e. 0 to 4 and 0 to 20 • The range of variable of an instrument is also sometimes called its SPAN.
• The term dead band or dead space is used if there is a range of input values for which there is no output. For example, bearing friction in a flow meter using a rotor might mean that there is no output until the input has reached a particular flow rate threshold Output reading
Dead space
Input of variable being measured
Sensitivity • Indicates how much the output of an instrument system or system element changes when the quantity being measured changes by a given amount, i.e. the ratio out/input. • For example, a thermocouple might have a sensitivity of 20uV/oC and so give an output of 20uV for each change in temperature.
• Thus, if we take a series of readings of the output of an instrument for a number of different inputs and plot a graph of output against input, the sensitivity is the slope of the graph. • The term is also frequently used to indicate the sensitivity to inputs other than that being measured, i.e. environmental changes.
• Thus a pressure measurement sensor might be quoted as having a temperature sensitivity of ±0.1% of the reading per oC change in temperature. • Example: A spring balance has its deflection measured for a number of loads and gave the following results. Determine its sensitivity. • Load in Kg 0 1 2 3 4 • Deflection in mm 0 10 20 30 40
• Fig. 1-20 shows the graph of output against input. The graph has a slope of 10mm/kg and so this is the sensitivity.
Stability • The stability of a system is its ability to give the same output when used to measure a constant input over a period of time. • The term drift is often used to describe the change in output that occurs over time.
• The drift may be expressed as a percentage of the full range output. • The term zero drift is used for the changes that occur in output when there is zero input.
Performance Characteristics
 Dynamic characteristics Dynamic Characteristics of an instrument includes;
1. 2. 3. 4.
Speed of response Fidelity Lag Drift
• Speed of Response: is the rapidity with which an instrument responds to the change in the measured quantity. • Fidelity: degree to which an instrument indicate the changes in the measured variables without dynamic error. • Lag : is a retardation or delay in the response of an instrument to the changes in the measured quantity • Drift: is a undesired change or gradual variation in output over a period of time that is unrelated to changes in output, operating conditions or load.
Other Issues
1.
Maintenance
2.
Consistency with process environment
3.
Safety
4.
Cost
Other Issues
Consistency with process environment •Direct contact – Sensors such as orifice plates and level floats have direct contact with process fluids.
Sheath protection – Sensors such as thermocouples and pressure diaphragms have a sheath between the process fluid and the sensor element
•Sample extraction – When the process environment is very hostile or the sensor is delicate and performs a complex physiochemical transformation on the process material, a sample can be extracted.
• Most sensors will function properly for specific process conditions. For example, many flow sensors function for a single phase, but not for multi-phase fluid flow, whether vapor-liquid or slurry. The engineer must observe the limitations for each sensor.
• Some sensors can have direct contact with the process materials, while others must be protected. Three general categories are given in the following. • Sensors in direct contact must not be degraded by the process material.
• The sheath usually slows the sensor response. • Samples must represent the fluid in the process.
Other Issues
Location of Measurement Displays
Local display
Local panel display
Centralized control room
Remote monitoring
• The measurement is displayed for observation by plant personnel. Typically, the display uses analog principles, which means that the display presents the measurement as a position in a graphical format, which could, for example, be the height of a slide bar or the position of a pointer. Often, the value is displayed as a line on a trend plot that provides the values for some time in the past. In addition, the measurement can be displayed as a digital number to provide more accuracy for calibration. Finally, measurements that are transmitted to a digital control system can be stored in a historical database for later recall and for use in calculating important parameters useful in monitoring process behavior, for example, reactor yields or heat transfer coefficients.
• ------ A sensor can display the measurement at the point where the sensor is located. This information can be used by the people when monitoring or working on the equipment. A measurement that has only local display involves the lowest cost, because the cost of transmission and interfacing to a digital system are not required. Note that no history of these measurements is available unless people record the values periodically.
• ---- Some equipment is operated from a local panel, where sensors associated with a unit are collected. This enables a person to startup, shutdown and maintain the unit locally. This must be provided for units that require manual actions at the process during normal operation (loading feed materials, cleaning filters, etc.) or during startup and shutdown. Usually, the values displayed at a local panel are also displayed at a centralized control room.
• ------ Many processes are operated from a centralized control room that can be located a significant distance (e.g., hundreds of meters) from the process. The measurement must be converted to a signal (usually electronic) for transmission and be converted to a digital number when interfaced with the control system. A centralized control system facilitates the analysis and control of the integrated plant.
• ----- In a few cases, processes can be operated without a human operator at the location. In these situations, the measurements are transmitted by radio frequency signals to a centralized location where a person can monitor the behavior of many plants. Typical examples are remote oil production sites and small, safe chemical plants, such as air separation units.
The Smart Sensor Revolution
Digital conversion and transmission
Diagnostics
Signal conditioning
Configuration
• Currently, sensor technology is experiencing a dramatic change. While the basic physics and chemistry of sensors are not changing, sensors are being enhanced by the addition of microprocessors at the location of the sensor. This change makes the following features possible that were not available with older technologies. • ------ The “signal” from the sensor is no longer simply a single value representing the measured value. The sensor can transmit additional information, including diagnostics and corrected estimates of a variable based on multiple sensors, e.g., orifice pressures and density. All values can be transmitted digitally, which allows many sensor values to be sent by the same cabling, which reduces the cost of an individual cable for each measurement, as required with analog transmission.
• ---- - The sensor can provide sophisticated diagnostics of its performance and warn when a measurement might be unreliable. • ----- The sensor can identify unusual signal characteristics and eliminate noise or “spikes” according to methods defined by the engineer. • ------ The range of a sensor can be changed quickly to accommodate changes in process operating conditions.
Principle measurements desired in Industry
Principle measurements desired in industry (a) Temperature (b) Pressure (c) Level
(d) Flow (e) Others ( Composition, pH etc.)
You have two challenges
What variables should be measured? What sensor should be specified for each measurement?
Reactor with feed-effluent heat exchange
Calibration • Calibration should be carried out using equipment which can be traceable back to national standards with a separate calibration record kept for each measurement instrument.
National Standards
Traceability Chain
Calibration center standard
In-company standards Process instruments
• Determine the sensitivity of the instruments that gave the following readings: • (a) • Load kg 0 2 4 6 8 • Deflection mm 0 18 36 54 72 • (b) • Temp oC 0 10 20 30 40 • Voltage mV 0 0.59 1.19 1.80 2.42 • [c] • Load N 0 1 2 3 4 • Charge pC 0 3 6 9 12
• Calibration of a voltmeter gave the following data. Determine the maximum hysteresis error as a percentage of the full-scale range: • Increasing input: • Standard mV 0 1.0 2.0 3.0 4.0 • Voltmeter mV 0 1.0 1.9 2.9 4.0 • Decreasing input: • Standard mV 4.0 3.0 2.0 1.0 0 • Voltmeter mV 4.0 3.0 2.1 1.1 0
• (a) 9mmkg • (b)0.060 mV/oC • [c] 3pC/N
• 5%