TDESIGN oFl ANALOG FILTERSTHE OXFORD SERIES IN EL! CTRICAL AND COMPUTER ENGINEERING Abe S. SERA, Series Editor, Electrical Engineering Allen and Holberg, CMOS Analog Circuit Design Bobrow, Elementary Linear Circuit Anclysis, 2nd Ed. Bobrow, Fundamentals of Electrical Engineering, 2nd Ed. Bums and Roberts, An Introduction to Mixed-Signal IC Test and Measurement Campbell, The Science and Engineering of Microelectronic Fabrication, 2nd Ed. Chen, Analog & Digital Control System Design Chen, Digital Signal Processing Chen, Linear System Theory and Design, 3rd Ed. Chen, System and Signal Analysis, 2nd Ed. Comer, Digital Logic and State Machine Design, 3rd Ed. Cooper and McGillem, Probabilistic Methods of Signal and System Analysis, 3rd Ed, DeCarlo and Lin, Linear Circuit Analysis, 2nd Ed. Dimitrijev, Understanding Semiconductor Devices Fortney, Principles of Electronics: Analog & Digital Franco, Electric Circuits Fundamentals Granzow, Digital Transmission Lines Guru and Hiziroglu, Electric Machinery & Transformers, 2nd Ed. Hoole and Hoole, A Modern Short Course in Engineering Electromagnetics Jones, Introduction to Optical Fiber Communication Systems Krein, Elements of Power Electronics Kuo, Digital Control Systems, 3rd Ed. Lathi, Modern Digital and Analog Communications Systems, 3rd Ed. Martin, Digital hutegrated Circuit Design McGillem and Cooper, Continuous and Discrete Signal and System Analysis, 3rd Ed. Miner, Lines and Electromagnetic Fields for Engineers Roberts and Sedra, SPICE, 2nd Ed. Roulston, An Introduction to the Physics of Semiconductor Devices Sadiku, Elements of Electromagnetics, 3rd Ed. Santina, Stubberud, and Hostetter, Digital Control System Design, 2nd Ed. Sarma, Introduction to Electrical Engineering Schaumann and Van Valkenburg, Design of Analog Filters Schwarz, Electromagnetics for Engineers Schwarz and Oldham, Electrical Engineering: An Introduction, 2nd Ed. Sedra and Smith, Microelectronic Circuits, 4th Ed. Stefani, Savant, Shahian, and Hostetter, Design of Feedback Control Systems, 3rd Ed. Van Valkenburg, Analog Filter Design Warner and Grung, Semiconductor Device Electronics Wolovich, Automatic Control Systems Yariv, Optical Electronics in Modern Communications, 5th Ed.DESIGN oF ANALOG FILTERS ne pe Rolf Schaumann Portland State University Mac E. Van Valkenburg New York Oxford OXFORD UNIVERSITY PRESS 2001Oxtord University Press Oxford New York Athens Auckland Bangkok Bogoti Buenos Aires Calcutta Cape Town Chennai Dares Salaam Delhi Florence Hong Koog. Istanbul Karachi Kuala Lumpur Madrid Melbourne Mexico City’ Mumbai Nairobi Paris So Paulo Shanghai Singapore Taipei Tokyo Toronto Warsaw and associated companies in Berlin Ibadan Copyright © 2002 by Oxford University Press, Inc. Published by Oxtont University Press. Inc. 19§ Madison Avenue, New York, New York, 10016 hupv/www.oup-usa.ore (Oxford is u registered trademark of Oxford University Press Al rights reserved, No part ofthis publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photovopsing. recording, or otherwise out the prir permission of Oxford University Press, Library of Congress Cataloging-in-Publication Data Schaumann, Rolt, 1941 Design of unalog filters / Rolf Schaumann, Mac E, Van Valkenburg. 1p. cm. — (The Oaford series in electrical and computer engineering) Rev. ed. of: Analog filters design / M.E. Vis Valkenburg. 1982. raphical references. ISBN 0-19-511877-4 (clots) [Electric filters, Active—Design and construction, 2. Analog electronic systems—Design and construction. 3, Linear integeated circuits Design and construction, 4. Operational amplitirs 1. Van Valkenburg, M. E. (Mac Elwyn). 1921— HL. Van Valkenburg. M. E. (Mac Elwya), 1921 Analog filters design. IL. Title. 1V. Series. ‘TK7S72.F5 S29 2001 621 3815324-de21 199.035428 Printing (last digit): 9 8 765432 Printed in the United States of America ‘on acid-free paperCONTENTS Preface xi 1 INTRODUCTION i 1.1 Fundamentals 1 1.2. Types of Filters and Descriptive Terminology 5 1.3. Why We Use Analog Filters. 9 1.4 Circuit Elements and Scaling 10 Problems 13, 2 OPERATIONAL AMPLIFIERS 15 2.1 Operational Amplifier Models 16 2.1.1 The Integrator Model 17 2.1.2 The Ideal Operational Amplifier 24 Opamp Slew Rate 26 ‘The Operational Amplifier with Resistive Feedback: Noninverting and Inverting Amplifiess 28 2.3.1 The Noninverting Amp! 23.2 The Inverting Amplifier 37 24 Analyzing Opamp Circuits 40 25 Block Diagrams and Feedback 43 2.6 The Voltage Follower 47 2.7 Addition and Subtraction 49 2.8 Applications of Opamp Resistor Circuits 52 Problems 59 29 3 FIRST-ORDER FILTERS: BILINEAR TRANSFER FUNCTIONS AND FREQUENCY RESPONSE 64 3.1 Bilinear Transfer Function and Its Parts 65, 3.2. Realization with Passive Elements 67 3.3. Bode Plots 78vi CONTENTS 3.4 Active Realizations 84 3.4.1 Inverting Opamp Circuits 84 3.4.2 Noninverting Opamp Circuits 91 3.4.3 Differential Opamp Circuits Allpass Filters: Phase Shaping 95 3.5 The Effect of A(s) 99 3.6 Cascade Design 104 3.7 And Now Design 107 Problems 117 SECOND-ORDER LOWPASS AND BANDPASS FILTERS 125 4.1 Design Parameters: Q and wy 125 4.2 The Second-Order Circuit 129 4.3 Frequency Response of Lowpass and Bandpass Circuits 136 44 Integrators: The Effects of A(s) 147 4.4.1 Inverting Integrators 149 44.2 Noninverting Integrators 152 4.4.3 The Effects of A(s) on the Biquad 155 45° Other Biquads 161 45.1. Sallen-Key Circuits 161 45.2 The Single-Amplifier Biquad (SAB) 170 45.3 ‘The General Impedance Converter (GIC) Circuit 178 Problems 187 SECOND-ORDER FILTERS WITH ARBITRARY TRANSMISSION ZEROS 192 5.1 By Using Summing 193 5.1.1 Phase Response of the General Biquadratic Circuit 206 5.2 By Voltage Feedforward 209 5.3 Cascade Design Revisited 229 5.3.1 Pole-Zero Pairing 233 5.3.2 Section Ordering 235 5.3.3. Gain Assignment 237 Problems 246 LOWPASS FILTERS WITH MAXIMALLY FLAT MAGNITUDE 252 6.1 The Ideal Lowpass Filter 252 6.2 Butterworth Response 254 6.3 Butterworth Pole Locations 25610 Contents vii 6.4 Lowpass Filler Specifications 261 6.5 Arbitrary Transmission Zeros 267 Problems 274 LOWPASS FILTERS WITH EQUAL-RIPPLE (CHEBYSHEV) MAGNITUDE RESPONSE 277 7.1 The Chebyshev Polynomial 277 7.2 The Chebyshev Magnitude Response 280 7.3. Location of Chebyshev Poles 284 7.4 Comparison of Maximally Flat and Equal-Ripple Responses 287 7.5 Chebyshev Filter Design 291 Problems 296 INVERSE CHEBYSHEV AND CAUER FILTERS 298 8.1 The Inverse Chebyshev Response 300 8.2 From Specifications to Pole and Zero Locations 304 8.3 Cauer Magnitude Response 310 8.4 Chebyshev Rational Functions 314 8.5 Cauer Filter Design 318 8.6 Comparison of the Classical Filter Responses 328 8.6.1 Degree 330 8.6.2 Passband Response 331 8.6.3 Stopband Response 332 8.64 Transition Band 333 8.6.5 The Q Values Required 333 8.6.6 Time Delay 335 8.6.7 Circuit Realization 338 Problems 339 FREQUENCY TRANSFORMATION 341 9.1 Lowpass-to-Highpass Transformation 343 9.2 Lowpass-to-Bandpass Transformation 352 9.3. Lowpass-to-Band-Elimination Transformation 370 9.4 Lowpass-to-Multiple Passband Transformation 381 9.5. The Foster Reactance Function 387 Problems 391 DELAY FILTERS 398 10.1. Time-Delay and Transfer Functions 398 10.2 Bessel-Thomson Response 401vit CONTENTS 11 12 13 14 10.3 Bessel Polynomials 406 10.4 Further Comparisons of Responses 409 10.5 Design of Bessel-Thomson Filters 411 10.6 Equal-Ripple Delay Response 415 10.7 Approximating an Kdeal Defay Function 422 10.8 Improving High-Frequency Attenuation Generating Gain Boosts 426 Problems 430 DELAY EQUALIZATION 432 IL] Equatization Procedures 433 11.2 Equalization with First-Order Modules 435 11.3. Equalization with Second-Order Modules 439 114 Strategies for Equalizer Design 444 Problems 447 SENSITIVITY 451 12.1 Definition of Bode Sensitivity 453 12.2 Second-Order Sections 466 12.3. High-Order Fitters 472 12.3.1 Cascade Design 474 12.3.2 LC Ladders 475 Problems 477 LC LADDER FILTERS 482 13.1 Some Properties of Lossless Ladders 483 13.2 A Synthesis Strategy 487 13.3 Tables for Other Responses 497 13.4 General Ladder Design Methods 499 13.4.1 The Twoport Parameters 500 13.4.2 Immittance Synthesis 505 13.4.3 Ladder Development 512 13.5. Frequency Transformation 518 13.6 Design of Passive Equalizers 524 roblems 529 LADDER SIMULATIONS BY ELEMENT REPLACEMENT 533 14.1 The General Impedance Converter 533 142 Optimal Design of the GIC 53615 16 17 Contents ix 14.3. Realizing Simple Ladders 540 14.4 Gorski-Popiel’s Embedding Technique 544 14.5. Bruton’s FDNR Technique 549 14.6 Creating Negative Components 558 Problems 564 OPERATIONAL SIMULATION OF LADDERS 568 15.1 Simulation of Lowpass Ladders 569 15.2. Design of General Ladders 580 15.3. All-Pole Bandpass Ladders 591 Problems 599 TRANSCONDUCTANCE-C FILTERS 603 16.1 Transconductance Cells 605 16.1.1 A Model 610 16.2 Elementary Transconductor Building Blocks 613 16.2.1 Resistors 613 16.2.2 Integrators 616 16.2.3 Amplifiers 621 16.2.4 Summers 622 16.2.5 Gyrators 623 16.3 First- and Second-Order Filters 628 16.3.1 A First-Order Section 629 16.3.2 A Second-Order Section 630 igh-Order Filters 638 16.4.1 Cascade Design 638 16.4.2 Ladder Design 638 165 Automatic Tuning 646 16.5.1 Frequency Tuning 648 165.2 Q-Toning 650 16.5.3 On-Line-Off-Line Operation 653 Problems 654 164 SWITCHED-CAPACITOR FILTERS 658 17.1 The MOS Switch 659 17.2. The Switched Capacitor 662 17.3. First-Order Building Blocks 664 17.4. Second-Order Sections 672 17.5. Sampled-Data Operation 676 17.5.1 The z-Transform 679 17.5.2 The Spectrum of a Sampled Signal 681x CONTENTS 17.6 17.7 17.8 17.9 175.3 ‘The Frequency Response for a z-Domain Transfer Function 684 Switched-Capacitor First- and Second-Order Sections 687 17.6.1 Bilinear Sections 689 17.6.2 Second-Order Sections 691 ‘The Bilinear Transformation 697 Design of Switched-Capacitor Cascade Filters 700 Design of Switched-Capacitor Ladder Filters 705 17.9.1 The y-Plane Circuit. 712 17.9.2 The z-Domain Cirevit 721 Problems 73) Index 735PREFACE Filters, and more specifically analog filters, are essential in many different systems electrical engineers are called on to design. Even signal-processing systems that appear to be entirely digital often contain one or more analog continuous-time filters internally or as interface with the analog world. Thus, their wide use makes the topic of analog filters appropriate for study not only by graduate students, but also by undergraduate students as they prepare to enter their professions. Additionally, engineers in their professional careers facing new design challenges may have to upgrade their knowledge in this field. This book was written to address all these needs. As minimum prerequisites, students should have had a course in circuits and ‘a course in mathematics that includes the Laplace transform. Beginning courses in electronics and systems are recommended as background. As the book is intended to be accessible to undergraduates, sophisticated mathematics is avoided wherever possible in favor of algebraic or intuitive derivations. In selecting the material, the decision was made to concentrate on active *inductorless” filters. Active filters can handle almost any filtering requirement that an engineer is likely to face. They are compatible with modem microelectronic systems applications, such as controls and voice and data communications, where considerations of size and weight make the use of inductors prohibitive. In many cases, active filters can even be implemented in fully integrated monolithic form, which, except at the highest frequencies, precludes the use of inductors For most applications, the operational amplifier is the device that provides active filters with the required gain. The integrated-circuit operational amplifier, the “opamp.” has pro- foundly affected the practice of analog electronic circuit design. Most electrical engineering students will have been introduced to the operational amplifier in basic circuits, electronics, or systems courses, where the opamp is often presented as an amplifier with infinite gain and bandwidth, This model has the advantage of leading to easy methods of analysis, and it often provides ready conceptual insight into the way a circuit may be designed in principle. Unfortunately, stich a simple model is inadequate for designing practical filters with exacting requirements, except at very low frequencies. To undertake the design of filters with more cricial specifications requires better opamp models. Although they make circuit analysis more difficult for the student, it was deemed important to employ a realistic opamp model in the text to ensure that the resulting circuits work predictably in actual practice. As student interest is directed increasingly at engineering practice, it is important that techniques presented in courses and textbooks lead to correctly functioning designs. In addition, a filter book must provide the student with the tools that permit assessing and explaining any deviations from the expected filter performance—even if the approach and the analysis are a little more difficult to understand by the novice.PREFACE ‘The study of filters is facilitated by access to 2 laboratory in which designs can be tested. This approach will not be available to many students due to high expenses and commitments of time. To provide the benefit of a laboratory setting during the study of filters, Electronics Workbench by Interactive Technologies, Ltd., was selected as a tool to verify the designs. This “simulated laboratory” allows experimentation with design alternatives and exploration of various design options rapidly and’ inexpensively, while gaining understanding of analog cireuits and insight into filter design. At the same time students are able to experiment with models of real commercial operational amplifiers and thereby develop confidence that their circuits and filters actually perform as predicted and designed. Almost all filter examples in this text were tested with Electronics Workbench®. Using this approach should provide the student with the assurance that the filter designs developed throughout are not just “paper exercises” but fead to circuits that work in practice. The material is organized as follows. Chapter | provides selected fundamental concepts ha common base for the study of filters. In Chapter 2, the operational amplifier is discussed and a suitable model, adequate for most active filter design procedures, is derived. The opamp is next placed into simple resistive feedback circuits, the fundamental block-diagram analysis approach is introduced, and a general method of analysis is provided that proves convenient for handling opamp circuits throughout the remainder of the book. “Experimental” examples demonstrate that the proposed models are realistic and provide Predictable and practical results. This approach is foifowed throughout the book, The design of filters, circuits designed intentionally to have a frequency-dependent transmission characteristic, starts in Chapter 3. The discussion begins with simple first-order blocks and illustrates the procedures by Which practi filters of higher order can be obtained as cascade connections of low-order blocks. This treatment is then essentiaily repeated for diquads, ic., filter sections of second order, All-pole functions are considered in Chapter 4, and functions with finite transmission zeros in Chapter 5. The presentation is restricted to only those circuits that have proven themselves in practice. Along the way, integrators, the main building blocks for all active filters, receive careful consideration. The consequences of the frequency dependence of the opamp gain, A(s), on the integrator and on the second-order sections are analyzed. The discussion of the effects of A(s) is contained in separate subsections. and can, at the discretion of the instructor, be glossed over. The material is included in the text for the benefit of students or engineers who need to desiga filters with exact parameters and understand their behavior. Next, Chapters 6 through 8 present the classical lowpass approximations, the Butterworth, Chebyshey, inverse Chebyshev, or Cauer functions that have maximally flat, equal-ripple, or elliptic magnitude characteristics. A comparison of the procedures allows the student to assess the distinct properties and the different efficiencies of these functions in approximating a given specification. Together with the frequency transformation discussed in Chapter 9, it will enable the student to design nearly all types of magnitude transmission requirements. The necessary background required for the design of delay filters is presented in Chapter 10, together with a comparison of the delay realized by different filter types. Chapter 1] addresses the design of delay equalizers. Although alluded to earlier in the book, the discussion and the mathematics of the sensitivity of filter performance to component tolerances are left until Chapter 12. An earlier presentation would have delayed the development of actual filter circuits, and would have unnecessarily detracted from design, Nevertheless, students should comprehend the meaning of sensitivity, and have methods of sensitivity analysis in their repertory of design tools. ItPreface enables them to compare different approaches and to understand the reasons why certain circuit structures and design methods should be avoided if the design is to perform well in practice. In the design of filters of high order, many of the best performing circuits are based on simulating passive LC ladders. A discussion of LC ladders is included in Chapter 13 because it is considered helpful for an understanding of the origin of the LC prototypes. Conceptually, LC ladder design procedures are relatively easy, but the mathematical details tend to be formidable. The discussion is necessarily brief and concise, but is kept sufficiently general to enable the careful reader to design complete lossless ladders. Other readers may just wish to understand the .C ladder concept and the origin of useful topologies, and then consult any of the numerous design tables available in the literature and proceed with an active simulation, ‘These simulation methods are based on two different approaches. Chapter 14 describes how the troublesome component, the inductor, can be replaced by an electronic circuit whose input impedance is inductive so that the complete filter can be realized in miniaturized form. The method discussed in Chapter 15 foilows a different strategy: recognizing that capacitors and inductors in the L.C ladder perform integration, their operation is simulated replacing inductors and capacitors by integrators. In the final two chapters, the problem of designing an active filter in fully integrated, mono- lithic form on an integrated circuit chip is addressed. This is a topic of wide-ranging current industrial interest in which considerable research is performed. The discussion is divided into two parts. In Chapter 16, the gq —C methods considered that uses transconductance amplifiers to permit integrating continuous-time filters.at the high frequencies required in modern commu- nication systems. For these applications, opamps are often ruled out because of their relatively small bandwidth, The gin —C method leads to very simple methodical design approaches and commercially successful high-frequency filters. The techniques presented may also be used to design discrete filters, but are aimed mainly at an implementations in integrated-circuit form. In Chapter 17, switched-capacitor filters are discussed. Switched-capacitor filters are sampled- data circuits, that is, they are analog filters but no longer operate in the continuous-time domain. ‘These integrated filters have been proven very successful at low to medium frequencies in commercial practice, but their analysis and design entail a new and different mathematical treatment that the engineer must understand to be able to develop successful products. ‘The book was planned as a revision of the 1982 monograph Analog Filter Design by Mac E, Van Valkenburg. Sadly, Profesor Van Valkenburg passed away in 1997 just as the first plans for the project were being formed, and so could, unfortunately, not participate in the revision. Thaye attempted to follow Mac Van Valkenburg’s proven approach and have maintained the overall sequence of topics as far as possible. However considerable rewriting and shifis in emphasis were necessary, new subjects had to be included, and some topics needed more emphasis and others less, to impart a more modem flavor and to reflect the developments in the field over the past two decades. Tam indebted to numerous students on whom much of the material was tested, and to my colleagues and friends in the circuits and systems area whose research discussions helped clarify many concepts and ideas presented in this text. It was a pleasure to work with Peter Gordon and the staff of Oxford University Press during the production of the book. Finally. express my gratitude to my wife Blanka for her assistance in proofrea: and, in particular, for her patience and encouragement, not to mention the innumerable cups of espresso delivered during the preparation of this book. Rolf Schauman Portland, ORINTRODUCTION 1.1. + FUNDAMENTALS 1.2. + TYPES OF FILTERS AND DESCRIPTIVE TERMINOLOGY 1.3.» WHY WE USE ANALOG FILTERS. 1.4 + CIRCUIT ELEMENTS AND SCALING PROBLEMS The subject of this book is a special class of electrical circuits, commonly referred to as analog active filters. Analog filters make use of resistors R, capacitors C, and inductors L. In addition we use active devices to obvain gain. In this book we deal with miniaturized filters, which for the most part means that inductors cannot be used because, except at the highest frequencies (hundreds of megahertz, 10° Hz, even gigahertz, 10° Hz), their size cannot be reduced to a level compatible with modern integrated electronics. We shall see in the course of our study that inductors can be avoided if we have access to gain. In that case, the only passive components we need are resistors and capacitors, and gain is provided by operational amplifiers (opamps) or operational transconductance amplifiers (OTAs). Such filters are generally referred to as active filters, sometimes more specifically as analog active filters to distinguish them from digital filters. Signals in analog active filters are normally continuous functions of time, sometimes sampled, whereas in digital filters signals are digitized, that is, they are represented by ordered sets of numbers. In modern communication systems, both analog signals and digital signals must be processed. Often both analog and digital circuits and filters must be implemented together on the same integrated circuit chip for so-called mixed- ‘mode signal processing. This book deals with the design of practical analog active filters and in this introductory chapter we provide some background material that will be useful in the studies to follow. 1.1 FUNDAMENTALS Figure 1.1 shows a circuit with a voltage source v)(t) connected to the source (excitation) terminals 1-1’. The situation in Fig. 1.la shows input and output as differential floating voltages. This is preferred for implementation in integrated circuit form for noise immunity, lower nonlinearities, and improved dynamic range. However, for most circuits in this book we find that input and output have a common terminal, Fig. 1.1b. This common terminal is 12 INTRODUCTION normally ground where v(¢) = 0 and to which all signals are referenced. The output voltage v(t) is observed of measured at the output (response) terminals 2-2’. If such a nvo-port circuit is operating in sinusoidal steady state, the two voltages v(t) = Vj cos(wt +0) and v2(t) = Vs cos (wr + 62) ab may be represented by phasors [ile =vi2e, and Va) et <= Va0 a2 ‘More generally, the input and output quantities may be represented by their Laplace transforms Vj(s) and V2(s) where s = 0 + ja is the frequency variable. V; and V> are complex numbers that along the jeo-axis are expressed in terms of their magnitude and their phase, Y= Vo)|,_,,, =|Vi Gey] e* and Vy = Vaisy = |Vo(jo)|e* (1.3) s=jo 0 because |T(jw)| > 1. The same equation (1.10) is used to define loss, [T(j@) < 1, but now @ < 0 because |T(jo)| < 1. Note that o as defined is a measure of gain, that is « > 0 is gain, whereas “negative gain,” a < 0, implies loss. This definition is consistent with the way in which most instruments (e.g., network analyzers") display the measured gain or attenuation, but it is not used consistently in the literature on filters. We will not attach a sign to « but let the context make clear whether we mean loss (attenuation) or gain. When we speak of an amplifier providing a dB gain, it is clear that the output is larger than the input. Conversely, when a filter or a transmission channel provides @ dB loss, obviously its output is smaller than its input. Only if confusion can arise will we attach a sign to @ The unit of gain and attenuation is the decibel (4B), with dB counted positive for gain and negative for attenuation. « = 0 dB when |7(jeo)| = { and input and output signals have the same magnitude, We may solve Eq. (1.10) for |7(jw)| by dividing a by 20 and taking the antilogarithm: IT (jeo)| = 10/29 = 19.5a109 oan Table 1.1 provides the corresponding values of a {dB] and |7(jw)|. It is sometimes helpful to have these values in mind when designing filters, For example, a loss of « = 80 dB means the filter’s output voltage is attenuated to 1/10,000th of the input voltage, and a gain of 80 4B means the output is 10,000 times a large as the input. At the —3 dB point the output is attenuated to 70.7% of the input magnitude. This is also refested to as the “half-power point” because power is proportional to the square of the voltage and 0.707? = 0.5. Other data for Table 1.1 can be filled by use of Eq, (1.10) and Eq, (1.1). For example, an amplifier with gain = 43 dB amplifies a signal approximately 141 times, and a filter that reduces a signal to 0.4% of the input magnitude has loss or attenuation of a ~ 48 dB. You may also wish to remember that =! 4B implies approximately a 10% decrease in the value of || ~2 dB implies approximately a 20% decrease in the value of |7'| 3 UB implies approximately a 30% decrease in the value of |] —6 dB implies approximately a 50% decrease in the value of |7| You can verify from Eq, (1.11) that each additional increase of loss by a = 6 dB reduces the value of |7} by a further 50%. Also note that each increase in gain by 20 dB or attenuation by 20 dB increases or reduces, respectively, the magnitude of a transfer function by a factor of 10. " A network analyzer will be the main (simulated) instrument that we use throughout the text to “measure” the gain, attenuation, or phase of the circuits we design. The network analyzer display will give readers confidence that their circuits behave as designed.1.2 Types of Filters and Descriptive Terminology 5 TABLE 1.1 Size of Attenuation or Gain and Its Values in dB Loss Gain a [6B] Ire a [eB] [Tuer 100 10s 100 10 60 10 60 10° -20 on 20 10 6 0501 6 1.995 3 0.707 3 Laid 1 0x91 ' Lim Ou 0.989 ot Lou 1.2 TYPES OF FILTERS AND DESCRIPTIVE TERMINOLOGY Filters are classified according to the functions they are to perform. Over the frequency range of interest we define passbands and stophands. In the ideal case, a passband is the range of frequencies of the filter where |7| = 1 and a = 0, that is, signals are transmitted from input to output without attenuation or gain. In a stopband |7'| = 0 and a = —oo, which means that transmission is blocked completely. The patterns of passbands and stopbands that give rise to the most common filters are shown in Fig. 1.2. They are defined as follows: 1. A lowpass filter characteristic is one in which the passband extends from @ = 0 to © = @., Where we is known as the cutoff frequency (Fig. 1.2a). w A highpass filter is the complement of the lowpass filter in that the frequency range from ()to we is the stopband and from «, to infinity is the passband (Fig. 1.2b). 3. A bandpass filter is one in which the frequencies extending from « to ay are pas while signals at all other frequencies are stopped (Fig. 1.3c). 4, The bandstop filter is the complement of the bandpass filter where signal components at the frequencies from «; to a are stopped and all others are passed (Fig. 1.2d). These filters are also sometimes referred to as notch filters because of the “notch” in their transmission characteristi d, There will be other kinds of filters introduced as our study progresses, such as filters specifically designed for a desired delay or filters with multiple passbands and stopbands, but filtering action can usually be visualized in terms of the basic four types of filters in Fig. 1.2. In practice, it is not possible to realize the ideal transfer functions shown by solid lines in Fig. 1.2 with real filters consisting of a finite number of elements. We shail see throughout cour study of filters that for real circuits the transfer functions defined in Eq, (1.4) are always described by real rational functions of the complex frequency s. A real rational function is a ratio of polynomials in s as shown in Eq, (1.12). NGS) _ bys +B a8" | + 2. + bis + bo Ts DOS) ays" + apis" +... tars + ay (12 where the coefficients aj,f = 1,....n, and bj, j = 1,...,m, are real numbers. The coefficient a, in Eg. (1.12) can arbitrarily be set to unity, a, = 1, by dividing numerator and denominator by d,. The numerator coefficients bj can be positive, negative, or zero,6 INTRODUCTION Figure 1.2. The four basic types IT lowpass IU highpass of ideal filter functions. Solid Jines: ideal function; dashed lines: real filter funetions where the magnitude is a continuous function of «. (© but the denominator coefficient a; must all be positive, a) > 0. = 0.1.0... — 1. If these restrictions are violated, the circuit will oscillate and the transfer function cannot be realized with positive elements. Also, to be realizable with a finite number of real components, the degree n of the denominator polynomial D(s) must be larger than, or at least equal to, the degree m of the numerator polynomial N(s), n = m . The magnitude of Eq. (1.12), when evaluated on the je-axis, |T(j)|, is a continuous function of frequency that cannot realize the abrupt behavior depicted by the solid lines in Fig. 1.2, Rather, realistic filter haracteristics that correspond to the four basic types are shown by the dashed lines in Fig. 1.2. We will see later that the sharpness of the transition from passband to stopband, as well as the shape and width of the passband, can be controlled in the design of the filters. In the chapters to follow we will switch from the linear magnitude or gain characte! |T(je)| to the logarithmic attenuation cl teristic @(w) depending on which representati makes the point more clearly. The attenuation characteristics corresponding to those of F 1.2 are shown by the dashed lines in Fig. 1.3. The (wo quantities are related, of course, by Eq. (1.10). When specifying the desired attenuation requirements for a filter to be designed. it is customary to plot attenuation as positive values of e as in Fig. 1.3. Often, frequency is plotted on logarithmic coordinates; then the curve, attenuation in dB versus logarithmic frequency, is known as a Bode plot, and the asymptotic slope is measured in multiples of 6 dB per octave1.2 Types of Filters and Descriptive Terminology 7 (a factor two in frequency) or 20 dB per decade (a factor 10 in frequency). This property will be examined in greater detail in Chapter 3. Since it is impossible to realize filters with the solid-ine characteristics in Fig. 1.2, having abrupt changes from passbands to stopbands and stopbands to passbands, we must learn to cope with the gradual transitions of the realistic filter transfer functions illustrated by the dashed lines in Figs. 1.2 and 1.3. The latter are used to approximate the ideal characteristics. The way this is accomplished is illustrated in Fig. 1.3. We will specify certain boundaries (shaded 1.3) and determine the required characteristics not exactly, but only within certain tolerances in terms of attenuation @ for the passbands and stopbands as follows: 1. Ina passband the attenuation is always less than a maximum attenuation designated as ox 2. Ina stopband the attenuation is always larger than a minimum valve designated as Gin 3, Bands between passbands and stopbands so defined are known as transition bands. In terms of Fig. 1.3a we see that the passband extends from @ = 0 to @ = wp, the range a, dB] (wo) Transition Ke Pass = a, dB «© @ Figure 1.3 Practical filter attenuation specifi functions a(w) (dashed lines). ions to be met by continuous8 INTRODUCTION of frequencies from @p to w, is the transition band, and aif frequencies greater than w, constitute the stopband. In this figure, as well as in Fig. 1.3b, we have used the subscripts p and 5 to indicate the edges of the passbands and stopbands. The same concept applies to the bandpass and bandstop cases shown in Figs. 1.3¢ and d. Here there are two transition bands. In later chapters, we will use the shaded attenuation characteristics in Fig. 1.3 as the filter specifications. In terms of Fig. 1.3a the design problem will be as follows: Given the four quantities oma» Onin, @p, aNd 6g, find an attenuation specification that satisfies the four requirements. The form of the solution, the attenuation a(w), is indicated in Fig. 1.3 by the dashed lines. Mathematically, the solution is found by a process called approx- imation, it determines the transfer function T(s) in Eq. (1.12) such that a(w) defined in Eq, (1.10) meets the requirements, For the lowpass in Fig. 1.3a this means a(w) < max in @ < wp, and @(w) > G@mig in @ > es. The complementary description is true for the highpass in Fig, 1.3b. Similarly, for the bandpass specification in Fig. 1.3¢ we require Qe) = ain in S Oy ANd > Wy, aNd A(@) S max iM wy SO S Cy, Where wg and @y, are the lower and upper comers of the two stopbands, and wp, and @py are the comers of the passband, Analogously, the complementary requirements hold for the bandstop specifications in Fig. 1.3d, The approximation problem will be studied in detail in later chapters. Another factor must be considered in design. The attenuation curves shown thus farhave a inimum value of a = 0 dB, corresponding to {7 (jco)fmax = | in Fig. 1.2. Since we are usi active filters, this need not necessarily be the case, because the active elements may provide gain. If it is necessary to meet the specifications exactly, it will be necessary to provide a circuit to reduce the so-called insertion gain, the gain provided by inserting the filter between input and output. If the circuit provides excess attenuation as normally found in passive filters, this so-called insertion loss must be overcome by additional amplification. These two conditions are illustrated in Fig. 1.4 in which the characteristic bandpass curve is shifted up or down, but the shape is not changed. In filter design, such frequency-independent insertion loss or gain is usually not important. me a. dB Insertion gain OE ® @ tb) Insertion toss Figure 1.4 Insertion gain and loss in a bandpass characteristic.1.3. Why We Use Analog Filters 9 1.3, WHY WE USE ANALOG FILTERS The basic concepts of the electric filter were developed in 1915 independently by Wagner in Germany and Campbell in the United States. In the years since that invention, filter theory and implementation techniques have been developed to a high degree of perfection. Implementing economically an active filter, a filler thai uses gain, became possible with the invention of the vacuum tube, and the development of feedback theary by Black, Bode, and others in the early 1930s. The present era of wide use of high-quality low-cost discrete analog active filters is due to the development of the inexpensive monolithic operational amplifier (apamp) by Widlar in 1967. In the recent trend to place ever more complicated complete systems on a monolithic integrated circuit (IC), filter designers felt pressure to devise techniques that allow the integration of analog filters onto the IC along with digital circuitry. The solution was found in the more recent developments of switched-capacitor filters for low to medium frequencies* by Fried in 1972 and by many other researchers in industry and at universities throughout the 1970s. Finally, the use of transconductance amplifiers led to integrated filters at high frequencies (useful from the audio range up to tens to hundreds of megahert7). Although increasingly many filtering applications are now handled with digital signal processing techniques and digital filters, generally there remains the question of whether to choose an analog or a digital filter for a particular application. In practice, there are a number of situations in which analog continuous-time filters are either a necessity or provide a more economical solution. Among these are interface circuits. They connect the real-world analog signals to the digital signal processor and provide bandlimiting before the signals can be sampled for further processing with sampled-data or digital techniques, and reconstruction back to the analog world, Also, filtering requirements at very high frequencies where ultrafast sampling and digital circuitry may not be realistic and economical (see Fig. 1.5) may require analog techniques. Analog active filters always use gain and capacitors. In practical discrete active filters, resistors are also used and gain is obtained from the opamp. In integrated active filters we obtain gain by making use of opamps or transconductance amplifiers (also referred to as operational transconductance amplifiers, known as OTAS), and we utilize capacitors, resistors, and, at the highest frequencies, integrated inductors. To be able to decide which components to use, and whether to use an active filter in preference to a filter assembled entirely out of passive components, we must consider Factors such as the folloy The technology desired for the system implementation Availability of de supplies for the’active devices, and power consumption Cost. The range of frequency of operation. The sensitivity to parameter changes and stability. Weight and size of the implemented circuit es Noise and dynamic range of the realized filter tthe time of this writing, commercial switched-capacitor fillers are limited to the range trom audio frequencies up to a few hundred kilohert10 1.4 CIRCUIT ELEMENTS AND SCALING INTRODUCTION Discrete analog ative RC filters Switched-capacitor active Hers Totegated analog active ilters Passive LC filters Distributed (waveguide) filters L 1 L 1 L L 1 L 1 L 1 L ee fe 1 ke MH 1 GHe Frequency. Hz Figure 15 Choice of fiker ¢y) ¢ as a function of the operating frequency range. ‘The meaning of several of these criteria will become clear as we progress in our dis active filters. Some puidelines for a possible choice of filter type can be obtained from Fig. 1.5 as a function of the desired frequency range of operation. The range of LC filters is limited at the low end by the bulk of the inductors and for high frequencies by parasitic and distributed effects. We see that compared to passive LC filters, discrete opamp-based active filters can realize filters for lower frequencies, but not for higher frequencies, whereas integrated analog filters, dependi ign and the type of devices used, can span the range from low audio frequencies to the gigahertz range. Switched-capacitor filters will be seen to be limited in their application range from about 10 Hz to about t Miz by impractical element sizes and by the bandwidth of the active devices. Microwave filters cover the highest frequency range by relying on distributed elements and waveguide designs. The indicated limitations of active filters depend, of course, on the active devices used: opamps or OTAs. The limits may be temporary and will change as technology advances and faster active devices become available, If sensitivity to component variations and fabrication tolerances is important, passive LC filters often have an advantage. We will consider this aspect in Chapter 13. Although still used in large numbers, their design is not compatible with modem fully integrated systems. As our discussion progresses, we will learn that to address this difficulty many methods have been developed to simulate the performance of LC filters with active circuitry. Finally, active filters require, of course, power supplies. The power supply voltages range anywhere from } V to 5 ¥, with typical designs at the time of this writing at or below 3.3 V. You will have learned in basic electronics courses that as the power supply voltage for biasing the active elements shrinks, so will the linear range over which the active devices can be used. Consequently, the usable linear signal level becomes smaller with reduced power supply voltages. Since active devices generate noise, which limits the smallest signals that can be processed, dynamic range becomes a serious concern for the designer. Dynamic rang defined as the difference between the largest undistorted signal and the noise level. It wil] become clear &s our study progresses that filter design is primarily a frequency-domain matter and that we seldom make reference to time-domain quantities, such as rise time or1.4 Circuit Elements and Scaling 11 overshoot. Design specifications or physical measurements are made in terms of frequency j in Hertz. However, it turns out to be much more convenient to use radian frequencies cw in rad/s rather than f. We will follow this practice and use « as long as possible and only convert to f im the last step. Experience in design will show the advantages of this choice We will make extensive use of both magnitude and frequency scaling, and also of normalized element values as well as normalized values of frequency. This has several reasons. It avoids the need to use very small or very large component values, such as pF (10! F) capacitors and M@ (10° @) resistors. It permits us to design filters whose critical specifications are on the frequency axis “in the neighborhood” of « = | rad/s. Further, it permits us to deal with only dimensionless specifications and components without having to be concerned with units, such as Hz, @, F, or H. Finally, the most important reason is that much of the work of filter designers is based on the use of design tables. In these tables so-called “prototype” lowpass transfer functions are assumed to have a passband along the normalized frequency « in 0 < @ < wy = 1 anda stopband in 1 < @, < w < oo. See Fig. 1.34. In addition, these prototype filters are designed with normalized dimensionless elements from which the real physical components are obtained by denarmalization. The relationships between the physical elements R, L, and C and their normalized represemations Ry, Ly, and Cy are ws RSC, (1.W3a, b, e) In Eq, (1.13), Rs is an arbitrary scaling resistor (in Ohms) that normalizes the impedance level and aos is the radian frequency (in rad/s) that scales and normalizes the frequency axis, such that «feos = 1, usually, at the passband corner. These expressions, as well as their i Rs 1 c es osRs, and R= Rn (L14a, b, are easy to remember by noting that R. L, and C have units of , 11, and F whereas Ry, Ln and C, are dimensionless numbers. Thus, the scaling factors ws and Rs do not only change the numerical values of the elements or frequency parameters, they can be seen to remove or restore the units depending on the direction of the transformation. As an example, assume that a prototype filter was designed, and design tables indicate that Ry = 1, Ly = 3.239, and Cy, = 1.455 are the required normalized components. IF the impedance level is selected as 1,200 @ and the frequency was normalized by ws = 10.8 Mrad/s = 10,800,000 rad/s, we compute the real inductor value from Eq. (1.14a) as 1,200 2 —— 3.239 = 360 pelt 10.8 x 10° MHz Similarly, we find for the other components R = 1.2 kQ and C = 112 pF. We still point out that ail components with physical units (@, 8, H, F, s, Hz) are scaled, but dimensionless parameters, such as gain, are not, Thus, in the above example where Rs = 1.200 @ was chosen as the resistor to scale the impedance level, a transconductance of value gm = 245 1S is normalized to win = RSSm = 1200 2 x 245 pS = 0.294 but an amplifier gain of value K = 45 dB keeps its value K in the normalized circuit12 INTRODUCTION TABLE 1.2 Typical Component Values in Discrete and Integrated Reaiizations Discrete Integrated Tolerances Lave 10-40% absolute 0.11% for ratios Resistors Preferred rang 1-100 ke Process dependent: values Lower limit 008-242 with 10% to 30% tolerances Upper limit 100-500 k2 in the range of 50 2-1 KO. Capacitors Readily walizable 5 pP uF 05-5 pF Practical 05 p10 uF 02-10 pF Marzinally practical 0.2 pP-S00 uF 0.1-50 pF Indvetors Really waizable 1 H-20 mit Real inductors with large losses Practical 0.1 11-50 mH ‘of the over of 10 nl ar fess Marginally practical 100 oH It will become clear in the chapters to follow that ordinarily there is no unique solution to the design problem. One of the decisions that the designer has to make is that of element size. Making appropriate choices will become easier with experience, and selecting a suitable impedance normalization factor, Rs, will help. Table 1.2 serves in guiding the selection Whether an element value is conveniently realizable depends on the chosen technology: here we distinguish between discrete filter designs and filters to be implemented on integrated nits, Note that in integrated circuits, ratios of like components can be very accurate with careful layout and processing, but untuned absolute values of components can have very large tolerances. This is the reason why the parameters of integrated-circuit opamps are not predictable with any accuracy or reliability and why in the design of active filters it is very important to make the filter independent of the opamp parameters. Practical sizes of resistors and capacitors are limited by the available silicon area on an IC chip. Integrated inductors are very small and at the same time very lossy. Simulated inductors can be larger and less lossy, but they add noise; it is not difficult to implement a simulated inductor in the range of many mH or even H. ‘The design of active filters generally requires accurate components. Typically. resistors with 1 or 5% tolerances are used in discrete circuits, more rarely in less critical applications 10 oF 20% resistors will suffice. On the other hand, capacitors with 10 or even 20% tolerances are more readily available and are preferred to save cost. As a rule, suitable capacitor values are preselected, such as 0.1 or 0.01 jLF, because fewer standard capacitor values are available for the filter designer to choose from. It makes little sense to compute a capacitor for a specified filter to three or more digits and then find out that no company manufactures that capacitor. The resistors needed for the filter are determined from these predetermined values and a specified frequency. For example, frequency is set by an RC product as fo = 1/(RC); then, for fy = 12 KHz. and choosing C = 9.01 uF we find R = 1/(2m foC) = 1.326 k@. The next closest 1% resistor can be chosen for the design. If the resulting tolerances of the RC product are too large, the resistor must be trimmed. We should note that the fact that components with at best 1% tolerances are available to the filter designer does xor mean that the computations leading, to the element values can be carried out lo only two or three significant digits. The numerical mathematics in filter design is as a rule very ill conditioned, especially for high-orderPROBLEMS 13, so many digits should be retained in the calculations to achieve valid results. The problem is that many intermediate results involve small differences of two relatively large numbers. For instance, suppose a step in the algebra calls for the difference of 1.324495 — 1.323122 = 0.001373 Being mislead by the available 1% components, a designer may choose to carry only three digits, 1.32 — 1.32 = 0.00, clearly a meaningless result, Even computing to four digits, 1.324 — 1.323 = 0.001, leaves only one digit, which has a 38% error. Let us emphasize, therefore, that computations in filter design must be carried out with 7 to 10 or, for high-order filters, even more digits. We shall throughout this yext carry out all computations to the required accuracy, but keeping practice in mind, use element values to only two or three digits. If circuit performance calls for higher a curacy tuning will be a sumed in our designs PROBLEMS 1.1 The input voltage of a filter is v;(¢) = V2 cos(eor + (8) otgax = 3dB in | MHz < f < 2.4 MHz; 2.68) and its output voltage is v2(t) = V2 - 5.34 min = 75 4B in f $730 kHz and nin = 48 12 13 14 1s costeat + 4.87). At the applied frequency «, deter Bin f > 7.8 MHz rine the gain in dB and the phase shift in degrees 4.6 The transfer function of filter is specified to equal implemented by the filter Atthe frequency f = 12 kHz, a filter is designed Tis) 2(s' 9.32) to attenuate the input signal by 78 dB. Find the 57+ 1.3228 + 0.9763? + 0.1505 + 1 amplitude of the output signal if the 12-KH2 input has an amplitude of 1 The Fequency i normalized by f= 18 AE A wide-band input signal of amplitude 100 mV is eee ne ee ee me ae " of attenuation increase in dB per decade at high applied to the filter. In the stopband, the remaining frequencies. At which frequencies isthe attenuation signal components at the filter's output must be no intite? ea larger than 45 ¢V. Determine the required stopband : js ° aluentation eof te filter in al 1.7 According to a design table, the normatized com- ponents of a passive LC filter are L, = 1.2547, fan amplifier has 35 dB gain at f. = 100 MHz and Te = O9BTS. Ly = 0.8765, C, == 25632. Co shift the phase by —42°, determine the output signal 1.3764, and Re = Ry -— 1, The impedaace level Gelivered if the input is vin (1) = 2.4 costes +45 is normalized by Ro = 300 © and the normalizing Identify the filter type (lowpass, bandpass, etc.) e- frequency is fo = 10.8 MHz. Find the values of the scribed by the following attenuation specifications denormalized components. and calculate the wialths of the transition band(s). 1.8 The normalized components of an active filter were (2) emag = 0.01 AB in f < 3.4 KHZ; agin = 45 computed to be Ry = 1.243, Ry = Ry = 1.677, GB in 96 kHz = f < 00 Ry = 6.888, ard Cy 0.768; the amplifier (6) cinny = 0.01 dB in 12.5k812 < f < 24 KE; gain is required to be K = 1,93. The normalizing Ctnin = 45 UB in f <7 KHz and f > 40 kHz Heuer i fo= 360 kt Groowe the impedance level such that the filter can be built-xith C = (oe na ee AF capacitors and determine the remaining ekements Gon in fs TkHeand f 2 40 Kz of the circuit, including the final value of amplifier (@). edgin = 60.0B in f < 24 KHZ; cus = 0.5 UB sain, in f = 40 kHz 1.9 Calculate the rate of attenuation increase in dB/ (anus = 0.1 dBin f < 360 KHz; nin in 600 kHz < f 804B octave and in dB/decade as f approaches zero and infinity in the function14 INTRODUCTION Te) 357 (s? +36) © sh 2.3449% + 1.82457 + 1.2675 + 0.987 1.10 An engineer is asked to build a filter to realize the transfer function 3s° 4257 ~ 0.85 + 1 $2,559 1.257 +3 +3.91)5 + 0.6 Ts) ‘The engineer objects that the function is not realiz~ able. List all items that are wrong with the funetion as stated,= OPERATIONAL AMPLIFIERS 2.1 + OPERATIONAL AMPLIFIER MODELS 2.1.1 The Integrator Model 2.1.2 The Ideal Operational Amplifier OPAMP SLEW RATE THE OPERATIONAL AMPLIFIER WITH RESISTIVE FEEDBACK: NONINVERTING AND INVERTING AMPLIFIERS 2.3.1 The Noninverting Amplifier 2.3.2 The Inverting Amplifier 2.4 + ANALYZING OPAMP CIRCUITS 2.5 + BLOCK DIAGRAMS AND FEEDBACK 2.2 23 2.6 * THEVOLTAGE FOLLOWER 2.7 + ADDITION AND SUBTRACTION 2.8 + APPLICATIONS OF OPAMP RESISTOR CIRCUITS PROBLEMS In this chapter we introduce the operational amplifier (opamp), the main device used to provide gain in the design of active filters. The other gain device is the transconductor or operational transconductance amplifier (OTA). It is used mainly in fikecs for very high frequencies. We will delay the discussion of OTAS and OTA circuits until Chapter 16. Since the performance of active filters depends critically on the opamp, it is very important that we fully understand the behavior of opamps to be able to undertake successful filter designs. Opamps are relatively complicated electronic circuits, consisting of transistors, resistors, and capacitors. As such, the signal amplification of gain they provide must be expected to be a function of frequency. To analyze the behavior of each filter with complete transistor-level electronic circuits for the opamps is too cumbersome and prevents us from gaining the needed insights into filter behavior. We will, therefore, in the next section develop a simple model, the integrator model, that will prove adequate to represent the behavior of the electxanie circuit for the vast majority of applications studied in this book. We will also define an even simpler model, the ideal ‘epamp, that will allow us to develop circuits very rapidly and to gain preliminary insight into their behavior under carefully observed conditions, specifically, a limited frequency sange. Experience has shown that even very simple models often allow us to predict acircuit’s behavior 1s16 OPERATIONAL AMPLIFIERS with an accuracy that is sufficient for judging the circuit's usefulness. We will, therefore, base many of the designs in this book on ideal opamp models, but we must remaict aware of the limitations imposed by this model. The ultimate test of a model's validity is its success in describing and predicting circuit performance in practice, in the laboratory. Opamps are rarely used alone; instead they are used in combination with other circuit elements that provide feedback, determine gain, bandwidth, and so on. The circuit elements used in active filters in addition to opamps are capacitors and resistors. Their values and interconnections are selected ize the specified frequency response of the filter. The simplest element used in the feedback network is the resistor, and opamp-resistor networks will be studied in this chapter. 2.1 OPERATIONAL AMPLIFIER MODELS Opamps are differential amplifiers, familiar in modem electronics (Gray and Meyer, 1993; Sedra and Smith, 1998). They differ from ordinary amplifiers by having two inputs. Their operation is such that the output voltage v, is the difference of the two input voltages multiplied by an overall gain factor. In terms of the voltages defined in Fig. 2.1, we have volt) = Afe+ (4) — VC) Qb where A is the gain of the opamp, Throughout our study, voltages will be measured relative to a 0-V ground reference, In general, the voltages can be arbitrary functions of time. Also, we will normally assume that the voltages (2nd ull signals) are sinusoidal and we shall use capitaf letters for our work, implying phasors or Laplace transforms. An important property of differential amplifiers is that signals, that are common 10 both iaputs, are not amplified: the amplifier rejects them. For example, if v..(r) and v(t) have the same additive noise n, this noise is not transmitted to the output v9(!), volt Allo) +n] = fv) +} = Afes(t) = vy) 22) resulting in a cleaner signal, Signal components that are common to both inputs, such as 2 in Eq. (2.2), are called common-mode signals and are rejected by the opamp. On the other hand, if the two inputs have no components in common, they are labeled differential-mode signals and each is multiplied by the gain A Differential amplifiers of very high gain were developed in the early [940s by George Philbrick and otters. These were intended for use in analog simulation, radar, and control systems applications. Early units employed vacuum tubes and were both bulky and expensive ‘The trend toward extremely small and inexpensive opamps began in the 1960s, when Philbrick, Burr Brown, and other companies developed modular solid-state units. The modern monolithic opamp, early versions of which were the LM 10) and the ;cA 709, was designed by Widlar in 1967. Dual and quad opamps (two and four units on an IC chip) followed in the early 1970s. The typical dimensions of an opamp on a silicon chip, depending on technology, are 1.2 x 2.0 mmm, Currently, a Jarge number of opamps and complete active filters can be implemented on the same IC chip. A simplified circuit diagram of an JC realization of a bipolar opamp is shown in Fig 2.24, with pin connections for a single-opamp package indicated in Fig. 2.2b. In actual use, extemal components, most notably power supplies, are connected to the package as indicated in Fig. 2.2c. Since in the design of active filters, we are normally concerned only with the2.1. Operational Amplifier Models 17 %O——4 —Ampiifiee 7 Xo] Amplifier vO | fr? Gain a Gain A fl ra @ » Figure 2 Operational amplifier. All voltages are referred to the Q-V ground terminals, Inputs and output uscally have a common ground terminal as indicated in b. signal paths, power supply connections are usually not shown explicitly in circuit diagrams. In this book we shall follow the convention of not showing power supply connections, but ‘we must remember that the electronic circuit in the opamp requires de power supplies to become active. Power supply voltages are found over wide ranges. They can be symmetrical as shown in Fig. 2.2c, such as £15 or £5 V, so that the signals are symmetrical around the 0 V ground reference. Alternatively, the power supplies can be single ended, such as 5, 3.3 V, and as low as 1.5 V, with one power supply terminal at ground. This is required, for example, in portable equipment with battery operation. In that case, the signals are floating around a nonzero de voltage, normally one-half the power supply voltage. Generally, the trend today is toward lower power supply voltages to save power and permit ever smaller device sizes. Similar circuit configurations as shown in Fig, 2.2a exist for other bipolar, complementary metal-oxide semiconductor (CMOS). or biCMOS designs (Johns and Martin, 1998). As is indicated in the figure, they all consist of a differential inpat stage gy, that provides a current i= SmlV4(1) — v (1) (2.3) to an inverting high-gain voltage amplification stage with gain —A}. The output is obtained from a unity-gain buffer stage to drive external loads and any off-chip circuitry. The opamp is completed by the necessary bias circuitry and a compensation capacitor, C1. as shown in Fig. 2.2a, 2.1.1 The Integrator Model An understanding of the opamp’s behavior can be gained from the simple block diagram of the opamp circuit, excluding the bias circuitry and, as we agreed, excluding the power supply. tt contains the sections identified in Fig. 2.2a as shown in Fig. 2.3. We have included the resistor R to represent the large but finite output resistance of the transconductance stage gin. Without having to concern ourselves with the details of the operation of the transconductance, gq. and the amplifier stage, —Az, we can analyze the behavior of the opamp as follows: The current 1 is given by 1 = 8m (V4 —V-) 4) At the input of the amplifier stage —Ap, it divides into the resistor current Vi/R and the capacitor current sC) Ve = sCilVi — (—A2Vj)). Thus we have trom Kirchhoff’s curren law Bm Ve — VIF VILI/R +5C, + Ar = 0 (25)