JERALD G. GRAEME rE> BURR-BROWN APPLICATIONS OF OPERATIONAL AMPLIFIERS rut (> Generation Techniques sanbjuyoay uores3uag-pily) SUIT TWNOILLWYId0 40 SNOILVON Idd SURR-GROWNa BURR-BROWN APPLICATIONS OF OPERATIONAL AMPLIFIERS = 2 ot ° SUAISITGINY TWNOILWH3d0 40 SNOLVOMddV THIRD-GENERATION TECHNIQUES sto SANGiINHOSL NOLWHIN39-GHIHL The Electronics Series icGraw- Hill i SURR-BROWNAPPLICATIONS OF OPERATIONAL AMPLIFIERS Third-Generation Techniques By Jerald G. Graeme Manager, Monolithic Engineering Burr-Brown Research Corporation ‘A new reference for citcuit designers and users of operational amplifiers, this volume presents and explains those operational amplifier applica- tions which have evolved since the publication of its companion volume, Operational Amplifiors: Design and Applications. Rather than being just a collection of circuits or theoretical analyses, the book prasents numerous applications of operational amplifiers in a variety of electronic equipment: specialized amplifiers, signal controls, processors, waveform generators, and special-purpose circuits. It is a storehouse of detailed practical information, featuring numerous cirouit diagrams, circuit values, pertinent design equations, error sources, and test-based com- ments on the elficieny of the arrangements and devices. With this information, each general circuit presented can be easily modified to meet your individual needs. Focusing on both general and specitic operational amplifier applications, this useful volume © presents general information on the more subtle operational amplifier limitations, including tech- niques for error prediction and reduction © shows how operational amplifiers are used to form differential output amplifiers, instrumen- tation amplifiers, and isolation amplifiers, as well a5 how output and input characteristics can be improved © spotlights specific applications for signal con- ditioning and processing, including controlled current souress, logarithmic amplifiers, multi- pliers, comparators, multiplexers, absolute- value circuits, sample-hold citcults, and rms converters (continued on back flap) (continued from front flap) © describes signal gonerating circuits for sine waves, square waves, triangle waves, ramp trains, and pulse trains—along with techniques for amplitude and frequency control and modi: fication © outlines spacitic end-use applications such as transistor test circuits, measurement circuits, and controllers The applications presented are primarily newly evolved circuits never before published in a book and never before described in such detail. Other special coverage includes @ unique chapter on the general techniques for effectively using opera~ tional amplifiers. The book probes the more subtle characteristics and performance interrelationships of operational amplifiers to aid the user in opti- mizing the performance he derives from opera- tional amplifiers, including the means for predict- Jing and controlling thermal drifts, gain error, and noise, as well as means for overload protection, Togather with Operational Ampiifiers: Design and Applications, this book is a highly useful up-to- date reference source for all engineers, scientists, and technicians who are involved with operational amplifier applications in electronics, instrumenta- tion, process control, chemical and physical simu- lation, and other specialties ABOUT THE AUTHOR JERALD G. GRAEME, Manager, Monolithic Engi- neering, Burr-Brown Research Corporation, joined the organization in 1965 as an operational ampli- fier circuit designer. Since that time, he has de- veloped bipolar, FET, and chopper-stabilized operational amplifiers, as well as instrumentation amplifiers. A frequent contributor of circuit design material to electronic trade publications, he now directs the development of monolithic analog and analog/digital circuits. His design efforts have been focused on high performance in the areas cof low drift, low noise, and high speed and on incorporated discrete component, monolithic in- tegrated circult, and hybrid thick-film fabrication. He Is also co-author/co-editor of the previously mentioned McGraw-Hill companion volume to this book, Operational Amplifiers: Design and Applica- tions.PRESTON POLYTECHNIC LIBRARY Chorley Campus Tel. 5811 Lancaster Campus Tel. 68121 Marshall House, Preston Tel. 23111 Poulton Campus Tel. 884651 Preston Campus Tel, 51831 FOR REFERENCE ONLY This volume must not be removed from the Library without the express permission of the Librarian. WOOOO 621.381535 GRA ‘ill I 81 i WOU 30107 }00 7THE ELECTRONICS SERIES BURR-BROWN Graeme © APPLICATIONS OF OPERATIONAL AMPLIFIERS Tobey, Graeme, Hueluman * OPERATIONAL AMPLIFIERS APPLICATIONS OF OPERATIONAL AMPLIFIERS Third-Generation Techniques JERALD G. GRAEME, M.S.E.E. Manager, Monolithic Engineering Burr-Brown Research Corporation McGRAW-HILL BOOK COMPANY New York St.Louis Sen Francisco Dilsseldor! Johannesburg Kuala Lumpur London Mexico Montreal New Deihi Panama Paris S80 Paulo Singapore Sydney Tokyo Toronto‘Library of Congress Cataloging in Publication Data Graeme, Jerald G ‘Applications of operational amplifiers. Includes bibliographical references. 1. Operational amplifiers. 1. Title. ‘TK7871.58.06G7 621.9819'57'3 73-7984 ISBN 0-07-028890-1 Copyright © 1978 by Burr-Brown Research Corporation. All rights reserved. Printed in the United States of America. 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 the prior written permission of the publisher. 456789 KPKP 7987654 “The information conveyed in this book has been carefully reviewed and believed to be accurate and reliable; however, mo responsibility is as- ‘sumed for the operability of any eircult diagram or inaccuracies in calcula~ fionsor statements. Further, nothing herein conveys to the purchaser 2 license under the patent rights of any individual or organization relating 40 the subject matter described herein. ACCESSION No. 23 JAN 976 _| Came l |. OPTIMIZING OPERATIONAL AMPLIFIER PERFORMANCE. . . CONTENTS 1.1 Preserving DC Performance 1 1.1.1 Thermal transient response of input offset voltage 2 1.12. Drift effects of input offset voltage mulling 8 1.1.3 Null techniques with no drift effect 6 1.14 Prediction of input current behavior 12 1.1.5 Input guarding 14 12 Preserving AC Performance 14 1.2.1 Optimizing gain accuracy 15 1.2.2, Operational amplifier response characteristics 17 1.2.8. The significance of input capacitance 21 1.8. Optimizing Noise Performance 25 1.8.1 Signal-to-noise ratio optimization 25 18.2 Noise-figure optimization 26 14 Overload Protection 28 1.4.1 Protection from power-supply faults 28 1.42 Input protection 29 1.48 -Output protection 81vi Contents 2. AMPLIFIERS. 2.1 Increasing Input Impedance 35 2.1.1 Input buffering 36 2.12 Bootstrapping 38 2.1.8 Isolation techniques 40 22 Increasing Output Drive Capabilities 42 22.1 Output current boosting 42 22.2 Output voltage boosting 46 228 Output power boosting 49 23° Differential Output Amplifiers 50 24° Instrumentation Amplifiers 58 24.1 Simplified instrumentation amplifiers 53 24.2 Input guarding 57 24.8 Improving common-mode rejection 59 244 Allinear gain control 61 2.5. Isolation Amplifiers 63 25.1 Feedback linearized configurations 68 25.2 Modulated carrier configurations 67 3. SIGNAL CONDITIONERS. ............... 8.1 Integrators and Differentiators 70 8.1.1 Specialized integrators 71 3.1.2 Integrator reset 73 8.13 Specialized differentiators 76 8.2 Controlled Current Sources 78 8.2.1 Unipolar output current sources 79 8.2.2 Improving unipolar output characteristics 84 8.2.8 Special-purpose unipolar current sources 86 8.24 Bipolar output current sources for floating louds 89 8.2.5 Bipolar output current sources for grounded loads 92 8.3 Logarithmic Amplifiers 94 8.3.1 Compensation of bulk resistance error 95 8.3.2 Improving frequency response 96 8.4 Multiplier/Dividers 97 8.4.1 Voltage-controlled resistor techniques 97 8.4.2 A logarithmic amplifier technique 100 85° Active Filters 102 8.5.1 Tunable active filters 102 8.5.2 Removing de offset 106 4, SIGNAL PROCESSORS... 4.1 Comparators 109 4.1.1 Specialized comparators 110 4.1.2 Simplified window comparators 112 42 Multiplexers and Clamps 114 4.2.1 Multiplexing amplifiers 114 4.2.2 Clamping Amplifiers 118 |. SIGNAL GENERATORS. .. 43° Absolute-value Circuits 119 4.8.1 Precision absolute-value circuits 120 4.3.2 Single amplifier configurations 125 4.3.3 Improving frequency response 129 44° Sample-hold and Peak Detector Circuits 182 4.4.1 Accuracy improvements 132 4.4.2 Speed improvements 136 4.5 RMS Converters 189 4.5.1 Computing techniques 139 4.5.2 Thermal techniques 141 5.1 Sine-wave Generators 144 5.1.1 Wienbridge oscillators 145 5.1.2 Variable frequency circuits 148 5.1.3 Variable-amplitude circuits 151 5.2 Square- and Trianglewave Generators 154 5.2.1 Single amplifier configurations 155 5.2.2 Variable circuits 158 5.3 Ramp and Pulse Generators 163 5.3.1 Single amplifier configurations 164 5.3.2 Other ramp and pulse generators 168 54 Monostable Multivibrators 171 5.5 Modulated Signal Generators 175 5.5.1 Amplitude-modulated generators 176 5.5.2 Frequency-modulated generators 177 5.5.8 Pulse width-modulated generators 180 ‘SPECIAL-PURPOSE CIRCUITS .. 6.1 Transistor Test Circuits 183 Gain measurement 184 Breakdown and leakage testing 186 Output resistance measurement 189 Noise measurement 192 Other transistor test circuits 194 6.2 Measurement Circuits 198 6.2.1 Meter circuits 198 62.2 Frequency and phase measurement circuits 200 628 Other measurement circuits 208 63 Controllers 205 638.1 Onoff controllers 205 63.2 Proportional controllers 207 638 An optimizing controller 212 64 Other Special-purpose Circuits 214 64.1 Audio circuits 214 64.2 AGC circuits 217 Glossary 221 Index 227PREFACE In less than a decade the operational amplifier made the transition from analog computer subassembly to universal analog component. It did so because of its outstanding versatility in performing electronic furic- tions. This versatile amplifier is now a basic gain element somewhat like an elegant transistor. During this transition, the operational amplifier was applied to per- form a multitude of electronic functions, and familiarity with such appli- cations greatly simplifies the task of the analog circuit designer. By adapting appropriate applications to his specific requirements, the circuit designer is able to quickly develop precise, simplified circuits. Many of the early operational amplifier applications were made available for design reference in the Burr-Brown Handbook of Operational Amplifier Applications. A more recent and expanded set was published in the McGraw-Hill/Burr-Brown book Operational Amplifiers: Design and Applications. 'G.E, Tobey, J. G. Graeme, and L. P. Huelsman, Operational Amplifiers: Design and Applications, McGraw-Hill Book Company, New York, 1971.However, circuit designers continue to envision new roles for the operational amplifier, and a new generation of applications developed as the operational amplifier technology matured. This book was writ- ten to supplement the second book cited above by providing an engineer- ing reference for the current applications, most of which were previously unpublished. Once again, an effort has been made to produce a prac- tical reference from which a circuit designer can adapt applications. Rather than being just a collection of circuits or theoretical analyses, this ‘book covers the operation, pertinent design equations, error sources, and relative merits of each circuit. With this information each general circuit presented can be modified for an individual requirement. ‘The applications of this book are divided into six chapters, beginning with general techniques for effectively using operational amplifiers, fol- lowed by general-purpose applications and then some special-purpose circuits. In Chapter 1 some of the more subtle characteristics and inter- relationships of operational amplifier performance are described to aid the user in optimizing the performance of operational amplifiers. This includes means for predicting and controlling thermal drifts, gain error, and noise, as well as means for protecting against overloads. Techniques for further improvement of operational amplifier performance are pre- sented in Chapter 2. With these techniques, it is possible to increase input impedance, isolation impedance, voltage swing, and output cur- rent, In addition, applications which produce differential outputs and true differential inputs are described. Treatment of the electronic functions performed with operational amplifiers is given in the remaining four chapters. Chapter 3 covers some advances in integrators, differentiators, current sources, logarith- mic amplifiers, multipliers, and active filters. Additional general-pur- pose circuits are described in Chapter 4 for signal processing functions such as comparators, multiplexers, clamps, absolute-value circuits, sam- ple-hold circuits, and rms-to-de converters. In Chapter 5 signal genera- tors are presented which produce sine waves, square waves, triangle waves, ramp trains, pulse trains, and timed pulses. In most cases the characteristics of the generated signal are determined by a few com- ponents external to the operational amplifiers. More specialized cir- cuits which perform some end function are covered in Chapter 6. These include circuits for transistor testing, measurement, control, audio, and automatic gain control. Following this chapter is a glossary of terms related to operational amplifiers and their applications. Preface xi 1am particularly grateful to two individuals highly skilled in the re- lated technology, Henry Koerner and Donald R. McGraw, for their review of the manuscript and the many resulting improvements. It is ‘also appropriate to acknowledge considerable related information and background derived from the Burr-Brown environment over the past ight years. Additional thanks are expressed to Fran Baker for her expeditious typing of the manuscript. also wish to thank my wife, Lola, for her accurate and attractive rendering of the illustrations. Jerald G. GraemeAPPLICATIONS OF OPERATIONAL AMPLIFIERS1 OPTIMIZING OPERATIONAL AMPLIFIER PERFORMANCE Operational amplifiers perform many electronic functions in instrumen- tation, computation, and control. The widely diverse applications of these analog building blocks demand a similarly broad range of perfor- mance characteristics. As a result, a general specification of characteris- ties overlooks performance subtleties important to various applications. Furthermore, many characteristics display complex relationships with other characteristics or application conditions. Such operational ampli- fier behavior often defies straightforward characterization. A variety of these more subtle characteristics and complex relationships are de- scribed in this chapter to aid in optimization of operational amplifier circuits such as those discussed in the following chapters. Specifically, the characteristics described here are those often encountered in opti- mizing de, ac, noise, and overload performance. 1.1 Preserving DC Performance Other than voltage gain, the primary determinants of operational ampli- fier de performance are the input characteristics. Most notably, the 12 Applications of Operational Amplifiers input offset voltage, input bias currents, and input offset current and their respective thermal drifts are major sources of de error. Application conditions affect the associated errors, but sometimes with predictable relationships which permit avoidance of further de error. This is the case for the offset voltage thermal transient response, the offset voltage drift created by nulling the offset, the thermal drift of the input currents, and stray leakages. 1.1.1 Thermal transient response of input offset voltage ‘The thermal drift specified for input offset voltage characterizes the steady-state off- set as a function of temperature. It does not include the transient offset variations which occur during warm-up or during rapid temperature changes. Asa result, an offset voltage can easily undergo changes well above those anticipated from the specified drift. Warm-up drift is gen- erally related to the thermal drift characteristic, as the amplifier power dissipation merely raises the temperature of the amplifier above the ambient temperature. However, if this dissipation creates nonuniform heating of the amplifier, thermal gradients which produce additional drift are established. Thermal gradients also result from rapid ambient temperature changes or from thermal shock. These gradients represent temperature differentials between matched components. Most often the greatest effect of a temperature differential occurs between matched transistors, where it disturbs the delicate but precise balance of emitter- base voltage drifts or gate-source voltage drifts in differential stages. Very little difference in temperature is needed to disturb the 3 uV/°C match of a 2,000 wV/°C emitter-base voltage drifts. The resulting offset voltage transient responses under warm-up and thermal shock are illustrated in Fig. 1.1. Both responses place limita- Thermal Shock Transient Worm=up Transient ‘Thermal Drift ° Fig. 1.1 Input offset voltage transients occurring after turnon or after a rapid temperature change can add greatly 10 the changes from ordinary thermal drift Optimizing Operational Amplifier Performance 3 tions on the operating conditions of the operational amplifier if de per- formance is to be carefully preserved. To avoid errors from the warm- up transient, the circuit should not be used until the transient has passed. Depending upon the amplifier thermal mass, this transient usually lasts from 1 to 5 min in free air. However, the transient is commonly equiv- alent to an amplifier drift of only 5 to 10°C, and the associated drift may not be significant in circuits intended for wide temperature range opera- tion. The transient response to thermal shock limits the permissible rate of change in the amplifier temperature in low-drift applications. When the rate of change exceeds the rate at which the amplifier mass can change uniformly, thermal gradients which develop errors above the anticipated drift error result. The length of the thermal shock transient varies from 10 s to several minutes, depending upon the uniformity and size of the thermal mass. Monolithic integrated-circuit operational amplifiers have thermal masses which are highly uniform and very small, and they provide the shortest thermal shock transient responses. However, the thermal re- sponse characteristics of the monolithic types introduce one additional transient error. With its high thermal conductivity and low thermal mass, the monolithic chip readily conducts heat from dissipation at its output to its sensitive input. This thermal feedback introduces input offset voltage changes with load current variations accompanying signal swings. In addition to creating error, these input voltage changes can override summing junction signal changes to give an appearance of in- finite open-loop gain or negative output resistance. Careful amplifier layout avoids thermal gradients caused by thermal feedback so that the thermal shock transient will only be due to the change in chip tempera- ture. The response then resembles the warm-up drift, which can last from several seconds to several minutes. All the various thermal tran- sients are reduced by heat-sinking the amplifier. 1.1.2 Drift effects of input offset voltage nulling Almost every operation al amplifier has a provision for nulling the input offset voltage. How- ever, most of the recommended null techniques add significant offset voltage drift. Correction of de error in an operational amplifier circuit can then degrade its temperature sensitivity. A typical bipolar transistor amplifier with a 1 mV offset and 8 uV/°C drift will acquire an additional 3yV/°C drift component when its offset is nulled. This added drift term may have the same or the opposite direction as the original ampli- fier drift, so that the net offset drift can either increase or decrease. In4 Applications of Operational Amplifiers fact, if the offset voltage nulled is all due to the one stage compensated, the drift is also nulled.' However, in almost all multistage operational amplifiers no one stage accounts for all significant offset voltages. As a result, drift will often be degraded by the offset null. Offset voltage null is most commonly achieved by supplying a correc- tion current to one side of a differential stage, as in Fig. 1.2. Correction current I, unbalances the two emitter currents, as indicated, to develop ‘a change in input offset voltage of* Mr, fora=1 where re 4 transistor dynamic emitter resistance For some amount of AVos the net offset is reduced to zero. However, this amount of offset correction exists at only one temperature, since re is highly temperature-dependent. This dependence is expressed in the common equation for dynamic emitter resistance of = eae where K 9 Boltzmann's constant TA temperature in degrees Kelvin a} electron charge ‘Then, the offset voltage correction varies with temperature as expressed Fig. 1.2 Input offset voltage null is commonly achieved by unbalane- ing the currents in a differential stage. Optimizing Operational Amplifier Performance 5 by KT AVos = 2h Gi, Consideration of the junction equation reveals the associated change in thermal drift for a bipolar transistor input stage to be® d(AVos) _ AVos dT 300° where AVos is the offset correction at room temperature. From this last expression it is seen that a drift of 3.3 wV/°C is added for each millivolt of offset nulled. ‘The above results apply directly to any opera- tional amplifier with no emitter degeneration resistors in the first stage. Emitter resistors reduce the portion of offset correction that is developed on the transistors, and the change in drift above is reduced by the ratio of rz to the total resistance in the emitter circuit. A more general view of the effects of the offset compensation current results from consideration of the differential current unbalance it creates in the first stage. This result is independent of the presence of emitter resistors. For a bipolar transistor input operational amplifier it can be shown that the drift effect of current unbalance will be* ‘dVoa eae ve ) (200 V/° C)log 22 This equation is plotted in Fig. 1.3. On this graph it can be seen that only a 10 percent current unbalance changes voltage drift by 8 uV/°C. Similarly, for an FET input operational amplifier it is found that the effect on drift is expressed by® ‘dV o8) dT /. Voi = V To —2.0mv/°C where Ip; and Ipg are near the zero temperature coefficient drain cur- rent Ipz. Figure 1.4 displays this result and points out the extreme sen- sitivity of FET operational amplifier drift to current unbalance. Note that only a1 percent unbalance results in an 11V/°C change in drift. This sensitivity coupled with the high offset voltages in FET types can make the offset null and drift interaction critical. While many FET operational amplifiers use other null approaches to avoid this problem, some are severely limited by it. It is again found that removing LmV6 Applications of Operational Ampiitiers Jo Tee 43. 50 8) iyv/*C) Je 1D Fig. 1.3 Input offset voltage drift resulting from current unbalance in a bipolar transistor differential stage. of offset adds around 3 nV/° C of drift, but the offsets to be nulled are often much higher with FET input amplifiers. 1.1.3. Null techniques with no drift effect To avoid the above offset voltage drift created by offset nulling, an offset correction signal with a compensating temperature coefficient can be used, or null can be achieved by adding a de input signal. With a bipolar transistor input Optimizing Operational Amplifier Performance 7 operational amplifier it is fairly simple to supply a temperature-depen- dent offset correction signal that does not disturb offset voltage drift. This requires a correction current with a temperature coefficient oppo- site to that of r for the amplifier with no emitter degeneration resistors in the above discussion. Since the temperature coefficient of re is ap- proximately opposite to that of a forward junction voltage, the required ger ae 1 eo 0 50, w/o) (AB) «-2zmvre( — ) with oy + toatl - ‘ Fig. 1.4 Input offset voltage drift resulting from current unbalance in an FET dif- ferential stage.8 Applications of Operational Amplifiers compensation becomes fairly straightforward. A correction current de- rived from a junction voltage, such as Vpx, can accurately compensate the thermal variation of re, as both characteristics are straight-line functions of temperature. This straight-line dependence is illustrated by the com- ‘mon expressions = oe —2 mv/°C Both characteristics have temperature coefficients with magnitudes of about 0.3%/° C, and their coefficients are of opposite sign, so that their effects can be made to cancel. An offset control providing such temperature compensation with a common type of null provision is given in Fig. 1.5. The offset control derives a correction current from the emitter-base voltage of a transistor to develop the appropriate temperature coefficient. Biasing current from R; divides between R, and the transistor as controlled by their feedback loop. Current in Re develops a bias voltage for the transistor, and this voltage in turn controls the portion of the biasing current con- ducted by the transistor. At equilibrium the current in Re will be Voe/Re. If this current is much greater than the transistor base current, then the offset correction current |, is also approximately Vpg/Rs, as desired. This current is divided by the control potentiometer for offset voltage adjustment. For those amplifiers with input emitter degenera- tion resistors, the same circuit can be used if resistance is added to the emitter of the compensation transistor. The appropriate resistance value to be added is one which produces a voltage drop equal to that of the degeneration resistors in the amplifier. With this addition, the tempera- ture coefficient of the correction current is lowered to match that of the overall emitter circuit resistance. af Re 200K Vee Fig. 1.5 A temperature-compen- sating offset correction current is derived from emitter-base foed- back around a resistor. To Fig. 16 Input offset voltage null in the inverting configuration is provided by biasing the noninvert- ing input. Alternately, offset voltage null can be achieved by adding a de correc- tion voltage as an input signal, rather than by using the amplifier null provision. This approach will be illustrated for the basic operational amplifier feedback configurations with resistive feedback, although the same operation can be achieved with other types of feedback. For the inverting configuration, it is simple to null the offset voltage effect with a de input voltage, as shown in Fig. 1.6. Here the noninverting input is biased away from ground by a voltage that is equal and opposite to the offset voltage. Then the inverting input is at zero voltage level, and no input or feedback current results from Vos. ‘The analogous null approach for noninverting operational amplifier connections is not as convenient. Figure 1.7 illustrates this null tech- nique for the noninverting amplifier and the unity-gain-follower connec- tions. For the amplifier circuit, a de correction signal is supplied to the feedback network. Offset correction with the follower is achieved by passing a de current through a small resistor added in the feedback path. The voltage drop on this resistor is set to cancel the effect of Vos. How- RellRy my NONINVERTING AMPLIFIER FOLLOWER Fig. 1.7 Simple de signal nulling of offset voltage in noninverting circuits also alters gain,10 Applications of Operational Amplifiers ever, the null circuits affect the voltage gains of these noninverting cir- cuits, as demonstrated by their accompanying gain equations. While the resultant gain is closely predictable, variations of the null potentiometer have an effect on gain. The gain added to the unity-gain follower by the null circuit produces a departure from the desired gain of unity of the order of 0.03 percent. These gain errors due to the offset null circuit can be avoided by in- creasing the shunting impedance presented by the null circuit. Simply increasing resistor levels is unsatisfactory, as this lowers the available compensation current. Instead transistor current sources can be used to supply virtually any current level from a high output resistance. For both polarities of offset null a bipolar current source is required, as illus- trated in Fig. 1.8. Here transistors Q; and Qy are current source-biased transistors having cascode collector biases from the FETs Qz and Qs. Only the difference between the two current source outputs flows in the feedback resistor Re to compensate for voltage offset, and this difference is controlled by the null potentiometer. The offset correction is the same for a voltage follower with feedback resistance. With the cascode bias the signal swing is essentially removed from the bipolar transistor collec- tors and impressed on the FET gate-drain junctions, resulting in much higher output resistance. The residual swing on the collector-base junc- tions is lower than the original signal by a factor of rag, and the resulting Va w 20 #(G4arasfllgal/28 Fig. 1.8 Offset voltage null of noninverting amplifiers without inducing offset drift or gain error is provided by an adjustable, bipolar current source. to Ry Wr “Fy os Nut Re * Rs+ Rally v-0 Re + tote? Ri Ry 2 tot Relea) Rit Re 2 Fig. 1.9 Difference amplifier offsot voltage null with the basic null circuit degrades CMR, and so a separate null amplifier may be desirable. Ry 2 10% null circuit output impedance is Zo = (grate || Cea)/2B ‘This results in output impedances of the order of 100MQ. and very small- signal shunting and gain error. ‘Application of a simple de signal will also remove offset voltage from the difference amplifier circuit, as in Fig. 1.9. Again a dc signal is added to alter the bias at the noninverting input; however, the shunting effect of the control can seriously degrade common-mode rejection (CMR). ‘A very small mismatch between the divider networks connected to the two inputs will significantly lower CMR. To preserve a CMR level near that of the operational amplifier itself, the null resistor Ry must meet the condition indicated. This condition is difficult to meet for the offset control range generally needed. Thus, a fixed resistor offset adjustment is normally made, followed by a CMR adjustment. For variable offset control with high CMR, a follower, shown as Ag, is sometimes added. Low output impedance from the follower avoids CMR disturbance. Perhaps the most precise means of input offset voltage nulling is an ‘automatic control technique, as shown in Fig. 1.10. In this case, the offset is automatically nulled on command from a periodic control signal.Fig. 110 A switched feedback integrator provides auto- imatic offset voltage nulling. Then any drift or other change in offset voltage will be continually re- moved by the null circuit. Basically the null circuit is an integrator that is switched into a feedback loop around A, when the input is grounded. In this state Ag integrates any nonzero output from Aj and supplies an input correction signal to restore a zero output level. The integrator output then holds at this correction level while A, returns to the signal- amplifying mode. During the null cycle Cy holds the most recent output voltage for monitor. An analogous offset-removing operation is per- formed in chopper stabilization.* 1.1.4 Prediction of input current behavior If the thermal drift nature of operational amplifier input currents is considered, de performance can often be significantly improved. The most obvious improvement results from the similarity of the two input bias currents. Since they are derived from matched components, the two currents tend to match, and their error effects can be made to cancel. This is commonly achieved by balancing the source resistances seen by the two operational amplifier inputs, as illustrated in Fig. 1.11. In this circuit a resistor is placed in series with the noninverting input, and this equals the parallel combina- tion of the resistors connected to the other input. As a result, the ampli fier input currents flow in equal resistances and their effects tend to cancel. Specifications of the temperature sensitivities of input bias currents and the input offset current are commonly given as an average for a Bos Current Error = Ue le2Rs eine ae Fig. 1.11 Input bias currents produce canceling error vol- tages when the resistances of the input current paths are balanced by adding Rg. specific temperature range. Generally, performance over the upper end of the temperature range will be better than that specified. To make use of this better current-drift performance, the nature of the drift should be considered. Except in amplifiers having added drift compensation, the input current drifts of bipolar transistor operational amplifiers are associated with transistor beta variations. Since the input bias currents are the base currents of the input transistors, they are inversely propor- tional to beta. For silicon transistors the beta variation with temperature produces an input bias current drift approximated by* C = —0.005/°C, T > 25°C C=—0.015/°C, T< 25°C aly ar Cly (25°C) { ‘This result is plotted in Fig, 1.12. From this curve it is obvious that op- Fig. 1.12 Input bias current drifts of most bipolar transistor opera- tional amplifiers are proportional to beta drift. % 50 7% 100 114 Applications of Operational Amplifiers eration in only the higher temperature range would be much less af- fected by current drift than operation over.the entire range. Analogous observations can be made for the input offset current. Since it is the dif- ference between the two input base currents, it is also beta-dependent and is approximated by the curve of Fig. 1.12. 1.1.5 Input guarding ~The low input currents and high input resistances of operational amplifiers permit accurate instrumentation of high resis- tance sources. However, stray leakage paths can greatly increase input currents and decrease input resistance unless the inputs are guarded. Such guarding is generally desirable if stray leakages at the nanoampere level are significant. The guard required is simply a conductive ring surrounding the input terminals and connected to a low impedance point residing at the same de and signal levels as the inputs. Leakages are absorbed by the low impedance guard ring, and leakage to the inputs is avoided by the equipotential bias of the guard and the inputs. Appropri- ate guard connections for the inverting and noninverting configurations are shown in Fig. 1.13. For the noninverting amplifier, the feedback resistors must provide a low resistance to drive the guard. However, the resistance seen by the inverting input can be matched to a higher re- sistance at the other input by adding Rs. 1.2. Preserving AC Performance Degradation of operational amplifier circuit performance at high speeds can be reduced if the sources of the degradation are identified. Gain error from bandwidth limitations can often be decreased by reducing phase compensation. Reponse speed can be optimized by choosing an Ry ow Fig. 1.13 Low impedance guard rings encircling the input terminals and tied to a voltage near that of the inputs absorb stray leakages. Optimizing Operational Amplifier Performance 15 amplifier on the basis of a variety of characteristics. Input impedance loading effects can be accurately anticipated by considering the fre- quency-dependent nature of the input capacitance. Each of these topics is discussed in this section. 1.2.1 Optimizing gain accuracy With its high-gain feedback loop an op- rational amplifier establishes closed-loop characteristics which are primarily determined by the feedback elements themselves. ‘The degree to which these elements fix closed-loop characteristics is determined by the feedback-loop gain. In general, closed-loop gain characteristics deviate from those predicted from the feedback elements by a fractional error € equal to the reciprocal of the loop gain," « = 1/A,, As a result, gain accuracy is optimized by maximizing loop gain. To permit this a very high open-loop gain is common for an operational amplifier, and heavy feedback or high loop gain is then possible. However, the open- Joop gain developed at direct current typically extends only to about 10 Hz with amplifiers having unity-gain phase compensation. Gain accura- cy above this frequency is, then, increasingly more difficult to attain, especially when high closed-loop gains further reduce the feedback-loop gain. However, for the higher closed-loop gain applications, unity-gain phase compensation is not needed, and reduced phase compensation can be used to improve the bandwidth of the open-loop gain. Tailoring the phase compensation to each application improves gain accuracy in this way. The effect of this broadbanding is represented by the Bode dia- grams of Fig, 1.14. For closed-loop gains greater than unity the response is tailored as in Fig, 1.14a to ensure stability at a gain level A; but not lower. A similar approach increases gain at intermediate frequencies with a modified unity-gain phase compensation, as in Fig. 1.14b. In the latter case, response characteristics are préserved for any gain by con- trolling the phase shift for a 45° minimum phase margin for all gain levels of unity or above. One further means of increasing gain at high frequencies is the feedforward technique described elsewhere.* To tailor phase compensation for the above results, both theoretical and experimental response evaluations are useful. Response errors are minimized by evaluating loop gain, peaking, overshoot, slewing rate, and settling time, where appropriate. As a basic guide in theoretical evaluation of phase compensation, a phase margin of 45° generally op- timizes overall response. With this phase margin a gain peak of 2.5 4B4 ata Initial Response —--— Broodbandes Response Fig. 1.14 Higher ac gain is attained by modifying phase compensation for (a) closed- loop gains greater than unity and (b) a closed-loop gain of unity. and a square-wave overshoot of 15 percent without ringing are common- ly attained. For experimental selection of phase compensation the pre- ceding overshoot characteristic serves as a useful guide. By observing the small-signal square-wave response of the operational amplifier, the appropriate phase compensation can be chosen with generally good re- sults. Then the other response characteristics are evaluated by testing the closed-loop response for peaking, the open-loop response for gain bandwidth, and the large-signal square-wave response for slewing rate and settling time. Improvement of any one of these characteristics can then be attempted by monitoring the characteristic while making further adjustments to the phase compensation. Further details of phase com- pensation criteria and techniques are presented elsewhere.® Capacitance loading can greatly affect the response characteristics of an operational amplifier and should be included in phase compensa- tion selection. As illustrated in Fig. 1.15, the load capacitance C,, forms a low-pass filter with the amplifier output resistance. The resulting re- sponse pole increases the slope of the amplifier response, as indicated ‘on the open-loop response curve, and adds phase shift in the feedback loop. Peaking, ringing, or even oscillation can be produced by this effect at low gain levels unless further phase compensation correction is made. However, the internal phase compensation of many operational ampli- Optimizing Operational Amplifier Performance 17 ted rial Open Looe iesponse a= Ettect of Fig. 1.15 Capacitive loading typically shifts the last response pole back in fre- ‘quency by its effect upon output resistance. jens i ssi icati se cases additional phase fiers is not accessible for modification. In these cases ad ‘ compensation ean be added in the amplifier feedback loop, as shown in Fig 1.16. The added components, Rs and G;, decouple the amplifier from Cz, as represented by the pole shift and added zero of the compen- sated Bode plot. 1.2.2 Operational amplifier response characteristics For high-speed ap- plications a fast operational amplifier is obviously desirable. How- ever, an operational amplifier may be fast in any of five ways. ‘The am- plifier appropriate for a given application is chosen by considering the relevant response characteristics, which include bandwidth, full-power response, slewing rate, overload recovery, and settling time. For opera- $M sen Loop Response with Co Compensation crmyend cy 1/e. Fig. 1.18 Settling time t, defines the time at which measurements can be made within an error ¢ following a step change in signal.20 Applications of Operational Amplifiers Second, the open-loop response of the operational amplifier should very closely approximate a single-pole response to permit fast settling.® This is identified by an accurate, continuous —6 dB per octave response slope. Such a single-pole response is readily modeled by a simple RC response to provide a prediction and theoretical limit of settling time. ‘This yields 0:367 tog /e i where e is the fractional error and fis the gain-bandwidth product of the feedback loop or the closed-loop bandwidth. ‘To minimize settling time this gain-bandwidth product should be maximized. ‘The maximum gain- bandwidth product attainable with an operational amplifier equals its unity-gain cross-over frequency f,, and this results for a voltage follower. As the closed-loop gain is increased, loop gain decreases, and f, decreases from f, in a manner which can be related by the feedback factor* B. ‘The feedback factor represents the fraction of the output which is fed back to the input and defines f, as f= Bi seed, : = Aq Mninverting eal ee B= [qa __ inverting where Acy; is the ideal closed-loop gain predicted from the feedback elements for an ideal amplifier. Equally important to fast operational amplifier response are the char- acteristics of feedback elements and other circuit conditions. Power supplies must be bypassed close to the amplifier terminals by capacitors having good high-frequency characteristics. Capacitance loading and stray capacitances should be minimized in component selection, compo- nent placement, and shielding. To further reduce time constants, resis- tance levels should be as low as can be driven by the operational ampli- fier output. Figure 1.19 shows the sources of two of these limiting time constants. First, the stray capacitance Cz of the resistor must charge and discharge through R,, and the associated time constant limits settling time. In addition, the charging rate limits both slewing rate and settling time. Shield capacitance C, produces similar limitations when charged Fig. 119 Low feedback resistances and a driven shield reduce the damping effects of component, shield, and stray capacitances. through high source resistance Rc. However, the effect of C, can be greatly reduced by driving the shield as shown. In this manner the shield is driven by a signal very nearly equal to the input signal to reduce the signal on C, by a factor equaling the open-loop gain of the amplifier. However, C, must now also charge and discharge through the feedback elements. 1.2.3 The significance of input capacitance The differential input capac- itance of an operational amplifier is frequency-dependent rather than a constant. Asa result, the capacitance value which should be considered is not immediately apparent. Similarly, the exact effect of input capaci- tance in operational amplifier circuits is not always obvious. To examine the significance of this capacitance, consider Fig, 1.20, where the capaci- tance C; shunts the differential input resistance Rj. Due to this shunting, the amplifier input presents a heavier load to the feedback network, increasing the associated gain error. In addition, the input capacitance introduces phase shift in the feedback loop and degrades the frequency stability and transient response. The degree to which C; produces these effects can be readily foreseen, but only if the appropriate value of the frequency-dependent C; is known. ‘The appropriate value to consider for C; can be found by examining the frequency-dependence of the capacitance, as depicted in Fig, 1.21. As shown, C; is a Miller-effect capacitance which is multiplied by the first stage gain A, to produce its large, low-frequency value Cy. As the gainRP Fi, 1.20, Diferontial input capacitance shunts the inpat resistance, producing gain error and feedback phase shift. of the stage drops with frequency, so does its multiplying effect on collec- tor-base capacitance C,. The input capacitance then reduces to its small, high-frequency value Ci, which is formed with C, and emitter-base capacitance C,,. An analogous situation exists for FET input operational amplifiers with the gate-drain and gate-source capacitances. ‘To compute the effects of C; the exact expression describing its behav- ior could be used. Fortunately, however, it is generally possible to ap- proximate C; with one of its end values Cy, or Cy. The transitional Straight Line << “hpproximatin Actual yume Fig. 1.21 The differential input ‘capacitance varies with frequency ‘over a wide range and has a com stant reactance in the transition region. Ciett-an gee Se Optimizing Operational Amplifier Performance 23. values of C; between f, and fy need not be considered, as the reactance of C; in this range is constant (see Fig. 1.21). With a constant reactance, C) appears resistive and cannot present increased shunting or further phase shift. The constant reactance results when the decrease in reac- tance of C. is counteracted by the decrease in gain A,. To decide which end value of C; to use in calculating its effects, the significance of each can be considered. Typically Cu, is much larger than Cy: 500 pF as compared to 3 pF. However, 500 pF does not present a very heavy shunt at low frequencies. This low-frequency value of C; exists only up to fy, and this frequency is most often the first open-loop response pole around 10 Hz. Above this frequency Cy, produces no additional shunt- ing or phase shift. Thus, if the 500 pF level Cy, has no significant effect before its pole at f,, it may be neglected. ‘Then the 8 pF level Cy will approximate C. To demonstrate the relative significance of each capacitance level, Fig, 1.22 shows their shunting effects on the open-loop differential input y* 0% 10% 10 1 10 100 tk 10k 100K ftH2? Fig. 122 The differential input resistances of FET and bipolar transistor operational amplifiers are far different, but input capacitance shunting cequalizes the input inpedances above 10 Hz.24 Applications of Operational Amplifiers resistance. The input impedance of a typical bipolar transistor opera- tional amplifier is not significantly affected by the large, low-frequency Cy. For most such amplifiers the input capacitance can therefore be approximated by the small Cyy. However, the 10'°Q open-loop input resistance of an FET input amplifier is heavily shunted by the Miller- effect capacitance Cx. For Cy, of 500 pF, the 10Q input impedance provided by FETs exists only up to 8 X 10” ‘ Hz. Beyond 10 Hz. the FET input impedance is the same as that of the bipolar transistor input. Then for FET input amplifiers, Cy, must be considered when high feed- back or source impedances rely on the high input impedance to avoid loading error. Fora given application the loading error is evaluated con- sidering the closed-loop input impedance, which equals the impedance of Fig. 1.22 multiplied by the loop gain, The falloff of the loop gain increases the rolloff of the closed-loop input impedance. When it is necessary to avoid the high Miller-effect input capacitance, an operational amplifier having a cascode input stage can be used. Such a stage is presented in Fig. 1.23. This stage preserves the high FET vo Fig. 1.23. Differential cascode biasing removes output voltage swing from the input transistors to reduce differ- ential input capacitance. Optimizing Operational Amplitier Performance 25 input resistance up to about 0.5 Hz by eliminating Miller-effect multipli- cation of the gate-drain capacitances. The voltage swing on these capac- itances is not multiplied by the stage gain and Miller effect is avoided. 1.3. Optimizing Noise Performance ‘The effects of the input noise voltage and currents of an operational amplifier combine to limit the signal sensitivity of the amplifier. To optimize noise performance, circuit conditions are adjusted for a maxi- ‘mum signal-to-noise ratio. Noise figure is generally a poor parameter for such adjustments on operational amplifier circuits, as will be de- scribed. 1.3.1 Signal-to-noise ratio optimization The noise characteristies of op- rational amplifiers are commonly modeled by equivalent input noise voltage and current sources, as in Fig. 1.24.) With this separation of noise sources from the amplifier, the signal sensitivity of the amplifier is immediately apparent regardless of the application. The effects of the equivalent noise sources combine to produce a total equivalent input noise which determines the signal-to-noise ratio of the amplifier. The noise currents igi flow in the source resistances, creating a noise voltage which combines in a mean-square sense with e,; to produce this total noise. Fig. 1.24 The combined effects of all noise sources in an operational amplifier are represented by equi- valent input noise generators.26 Applications of Operational Amplifiers ratio, and therefore voltage amplifier noise performance, will be opti- mized. This desired minimum results when source resistance is zero. When the input signal is a current rather than a voltage, an analogous expression can be written which indicates a signal-to-noise ratio maxi- mum for infinite source resistance. However, voltage inputs are more common, and the relative noise performance of different amplifiers can be compared by plotting e,; for each one, This is done in Fig. 1.25 for the basic types of operational amplifiers. From these curves it is seen that bipolar transistor input operational amplifiers commonly provide the lowest noise at low source resistance, while FET input types are best for very high source resistance, 1.3.2. Noise-figure optimization While noise figure is a more common figure of merit than signal-to-noise ratio for noise performance optimiza- tion, it applies in a special class of circuits which does not include most operational amplifier circuits." Noise figure is also commonly displayed asa function of source resistance, as in Fig. 1.26. Once again the graphi- cal display indicates the relative merits of different amplifier types at different source resistance levels. However, the curves also suggest that noise performance for a given amplifier is best at that level of Rg which ‘minimizes the noise figure. From this feature comes the common recom- mendation that source resistance be adjusted to minimize noise figure. @ nit(a¥, ems )(DC 10 1kH2) CHOPPER STABILIZED 400 50: BIPOLAR INPUT a 1.04 05 — source a RESISTANCE on, = (OHMS) wo ik 10K = 100K «1M 10M_—t00M Fig. 1.25 The type of operational amplifier providing best noise performance for a given source resistance ean be found using epit as a figure of merit Optimizing Operational Amplifier Performance 27 NF. (dB) (OC totKHz) BIPOLAR CHOPPER- Fit. 1.26 Noise figure curves also INPUT STABILIZED splay the relative merits of the ‘tls types of amplifiers, ad the ‘curves suggest, unreliably, an op- timum source resistance for best noise perform ° 5 es 400 1k 10k 400k 4M 10M 400M Rg (OHMS) ‘This isin contradiction with the signal-to-noise ratio results above, which indicated that zero source resistance minimized e,;, for maximum signal sensitivity. At zero source resistance the noise-figure curves reach infin- ity, suggesting very poor performance. However, adding source resis tance to provide minimum noise figure can only increase the noise volt- age developed by input noise currents. This contradiction commonly results from an overgeneralization of the performance indicated by the noise figure. In general the noise figure does provide a relative comparison between two amplifiers in a given cir- cuit, but it does not necessarily give a reliable comparison between two circuit conditions for the same amplifier. Noise-figure comparison of performance for various source resistances is, then, not always reliable. This practice came from the r-f and audio circuits, in which transformer coupling is common, as represented in Fig, 1.27. When the turns ratio is Fig. 1.27 With a transformer-coupled input, a turns ratio varie tion which increases source resistance, and thereby adds noise, also increases the transformed signal, so that signal-to-noise ratio might be increased.28 Aplications of Operational Amplifiers changed to increase the transformed source resistance, the signal is also inereased. This results in maximum signal-to-noise ratio and optimum noise performance at some nonzero value of resistance. In this case the optimum source resistance is the one which minimizes the noise figure, and the noise figure is here a direct measure of noise performance. However, this result of this specific cireuit form cannot be applied more generally to operational amplifiers. 1.4 Overload Protection” Voltage and current overloads are common causes of operational ampli. fier failures. While most operational amplifiers incorporate some over load protection circuitry, few can withstand all the different overload conditions which can damage an amplifier. Some of these overload con- ditions are obvious, such as input breakdown under excessive input volt- ages or output overheating under short circuit. Most operational ampli- fiers are protected to some degree against the latter two conditions, However, other overload conditions which may be peculiar to certain applications are less evident. These include the effects of voltages main- tained by capacitors when the power supplies are turned off, and the forward-biasing of integrated-circuit substrate junctions. A voltage re- tained at an amplifier input by a capacitor or other source can forward- bias and destroy a substrate junction when the negative supply voltage becomes less negative than the input voltage. Overload protection cir- cuits for power supply faults, input overloads, and output overloads are described in this section. 1.4.1 Protection from power-supply faults The most common power- supply faults in operational amplifier circuits are supply reversals and voltage transients. Damage from these overloads is prevented by the circuits of Fig. 1.28. To protect against damage from voltage reversal, a diode is added in series with each power-supply terminal to block reverse current flow. This protection also prevents forward-bias of an inte- grated-circuit substrate junction, since a reverse-biased diode will now disconnect the negative supply. However, for the latter protection alone resistors can be added in series with the inputs to limit the substrate cur- rent flow to a few milliamperes. ‘Transient protection is provided by the zener diode clamps and the tes pe SUPPLY REVERSAL SUPPLY TRANSIENT PROTECTION PROTECTION Fig. 1.28 Damage from power-supply faults is prevented by using diodes to block current under a voltage reversal and by clamping the supply terminals to remove transients. voltage-absorbing FET current sources shown. The zener diodes are chosen for ON voltages greater than the normal supply voltages but less than the maximum supply ratings of the amplifier. Under these condi- tions the zener diodes will be OFF under normal supply voltages, and they will clamp the impressed voltages under supply transients. The current source-connected FETs are chosen for Ipss levels above the nor- mal current drains of the amplifier. Below the Ipss level the FETs are below pinchoff and appear as small resistances in series with the supply lines, Under a transient the zener diodes turn to ON to clamp the supply terminal voltages, and their current drains raise the FET currents to Ipgs. Now the FETs are in pinchoff and appear as high impedance cur- rent sources to support the excess voltages. As long as the transients do not create voltage breakdown in the FETs, the transient currents are limited to Ipss- 1.4.2. Input protection Overload conditions at operational amplifier in- puts are essentially those of excessive common-mode and differential voltages. Either condition can induce a voltage breakdown that will damage or destroy the input transistors. Because of the precise match- ing needed between the input transistors, only minor damage from breakdown can significantly degrade de input characteristics. Such