Analog Computers

Reference / Paper · 1956

Analog Computer Techniques

Read the PDF (141 pp) ↗

A 1956 textbook by Capt. Clarence L. Johnson (USAF, U.S. Air Force Institute of Technology) covering the theory and practical use of electronic analog computers (differential analyzers). Topics range from linear computing components, amplitude and time scaling, and synthesis of servomechanism systems, through nonlinear and function-generating techniques, repetitive analog operation, and a concluding introduction to digital differential analyzers. The book was developed from computer courses taught to undergraduate and graduate students at Wright-Patterson Air Force Base and is pitched at readers with knowledge of Ohm's law, Kirchhoff's laws, and basic differential equations.

Manufacturer
McGraw-Hill
Author
Clarence L. Johnson
Year
1956
Type
Reference / Paper
Language
English
Learning track
general theory
Pages
141
Credit
Copyright 1956 by the McGraw-Hill Book Company, Inc. Library of Congress Catalog Card Number 56-7560.
  • McGraw-Hill
  • analog computation
  • electronic differential analyzer
  • differential equations
  • servomechanism simulation

← Back to the Reference Library

Analog Computer Techniques

Ps Z > - O G@) 2) O < U G 4. [T] - AD. | m7 (-) E Z O G m Y mate ANVdWOD HOO THH-MYH DIAN NOSNHOPr “ter ui bs a . rt ANALOG COMPUTER TECHNIQUES CLARENCE L. JOHNSON CAPTAIN,U.S, AIRF FORCE io ASSISTANT PROFESSOR, DEPARTMENT OF MATHEMATICS W.3, AIR FORCE INSTITUTE OF TES. HNOSOLOGY Ati a ot ee BOOK Cc OM P Ad ANALOG COMPUTER TECHNIQUES Designed as an aid to those learning to wWse electronic analog computers a) electronic differential analyzers), this book will make the transition period from neophyte to experienced computer operator easier for the engineer. Following the introductory and gen- eral material on electronic analog corn putation, the volume consists of specific techniques for the solution of difficult or unusual problems. Wherever prac tical, techniques and principles have been presented in such a manner that an individual with a minimum knowl- edge of mathematics and electronics can readily understand them. The book treats in some detail the use of diodes and differential relays in analog computation. Special emphasis is given to function generating tech- niques including methods of represent ing functions of more than one variable. Phere is a chapter on repetitive analog computers and a chapter describing the principles of operation of the digital dif- ferential analyzer type equipment. The problems at the end of the chap- ters will be helpful to both students and those engaged in self study to insure that they have understood the material presented, A bref review of the termi nology of differential equations and a brief introduction to operator notation is included in the Appendix as an aid to the less advan el reader, ANALOG METHODS IN COMPUTATION AND SIMULATION By WALTER W. SOROKA University of California Here is an organized treatment of the more important and useful methods ol analog computation and simulation with particular reference to enginet ing and scientific applications. The book describes mechanical, electro mechanical. electrical and electronic analog components for performing basi mathematical operations and then shows how such components may ito combined into mathematical machine: for solving sets of simultaneous linear algebraic equations, polynomial equa tions. and differential equations of all kinds. The final three chapters show how dynamical analogies, finite differ ence networks, membrane analogies and conducting sheets of uniform and variable thickness may be applied to the simulation of various physical 85 lems aril to the solution of cilferential equations. Treatment is fundamental and direct The author carefully develops the theo retical basis for each approach, thu: providing the student with the tool necessary for applying these methods to actual problems. This 1s a pioneer attempt to organize classily and present a wide range ol material on analog computation in text book form. It is a valuable text and rel erence book for all scientists and eng neers at both the student and profes sional levels. CL) eee ae in| Other McGRAW-HILL Books ENGINEERING CYBERNETICS: The Science of Control By H. S. Tsren, California Institute of Technology. 304 pages, $7.50 This important hook aims to establish Engineering Cybernetic “ne ff new branch of enginedring science. It covers, as far as possible within the limited space, the whole field of scientific principles of control, from the simple conventional servomechanisms to the very complex con trolled and guided systems, Non-interacting controls of many variable systems, control design by perturbation or with prescribed performance, optimalizing control, noise filtering and detection, and von Neumann's theory of error contro] are some of the important topics trea tec PRINCIPLES OF NUMERICAL ANALYSIS By Ausron S. Housenotper, Oak Ridge National Laboratory International Series in Pure and Applied Mathematics, 2/4 pages, be A a 18 A senior-graduate text which develops the mathematical principles upon which many computing methods are based and in the light of which they can be assessed, Directed primarily toward digital computation the book is designed to give a unified treatment rather than a com plete catalogue of methods. Treatment is primarily theoretical, Tech niques for making estimates of errors are indicated wherever possible Funetional equations as such are not discussed, but emphasis is placed upon the methods of solving the finite systems and perhorming the nilerpolations which ore required in the digital olution (vl finris hoorwal MLNS INTRODUCTION TO NUMERICAL ANALYSIS iy F. B. Hrpepranp., Massachusetts Institute of Technology, fr ternational Series in Pure and Applied Mathematics, S20 pages so OM) Designed for workers in fields of engineering, mathematics and physi for research personnel and for courses in Numer ical Analysis offered by colleges and universities, this book provides an introductory treat ment of the fundamental processes of Numerical Analysis. These pro esses are the foundation of existent techniques associated with the effective use of modern high-speed calculating devices, Thus, the text provides a substantial treatment of the basic operations of computation approximation, interpolation, numerical differentiation and integration and the numerical solution of equations. CONTROL-SYSTEM DYNAMICS By Water R. Evans, North American Aviation, Inc., Downer California, McGraw-Hill Electrical and Electronic Engineering Series. 277 pages, $7.50 An exposition of the “Root Locus Method” invented and developed by the author, this new volume demonstrates the techniques for determin ing the response of linear control systems. The root locus, a tool to factor an aeekradt polynomial, is useful in analyzing the pertinent differential equations in feedback control systems. Developing from the simple to the complex, each solution establishes a concept which paw mits a simpler technique to be applied to the next, more complioated problem. oe “NOSNHOr > Z > Ee O G) @ O < Us Cs cal [T] : ADs <1 a Be Z ‘ Oi 33 u if) ANWdWOD NHOoOd8 THH-MY¥YeDSAN ANALOG COMPUTER TECHNIQUES Analog Computer Techniques CLARENCE L. JOHNSON Captain, U.S, Air Force Assistant Professor, Department of Mathematics U.S. Air Force Institute of Technology McGRAW-HILL BOOK COMPANY, INC. New York Toronto London 1956 ANALOG COMPUTER TECHNIQUES Copyright @ 1956 by the McGraw-Hill Book Company, Inc. Printed in the United Statea of America, All rights reserved. This book, or parts thereof, may not be reproduced in any form without permission of the publishers. Library of Congress Catalog Card Number 56-7560 iv THE MAPLE PREAA COMPANT, TOME, PA, To My Wife, Lillian and My Daughter, Tonya Sue PREFACE The electronic differential analyzer, frequently referred to as the electronic analog computer, has become an extremely important tool to the engineer, physicist, and applied mathematician in the past decade. Many of the rapid advances in the field of automation would have been impossible without this tool. The treatment of the electronic differential analyzer contained in this book was written as an aid to the computer operator. The presentation of the material is such that the average person with a knowledge of Ohm's law, Kirchhoff’s laws, and a basic knowledge of differential equations can read and understand the majority of the material presented. In a few isolated instances Laplace-transform notation is used to prove a statement. The results obtained, however, are stated in such a manner that they can be applied to the setup of computer problems without a knowledge of Laplace-transform theory. In the various stages of development, the material contained in this book has been used for the past two years in the computer courses taught to advanced undergraduate and graduate students at the U.S. Air Force Institute of Technology at Wright-Patterson Air Force Base, Ohio. The analog-computer course was included in the curricula because of the relatively large number of students undertaking independent study investigations in the fields of guidance systems and automatic control, The use of computing devices in the level of work attempted was virtually mandatory. Prior to the inauguration of the computer courses, too much of the student's effort in his thesis work was directed toward learning to use the computer and too little toward the actual investigation of his problem. Itis the author's belief that this situation has been improved by the inclusion in the curricula of the material in this book. The author further believes that the engineer, whether his specialty he electronics, mechanics, chemistry, or aerodynamics, has an increas- ing need to be familiar with both analog and digital computers. Although he may never be called upon to operate a computer, he vil Vill PREFACE should recognize the capability of a particular class of computers to solve a problem with which he may be confronted. The arrangement of the material in this book was chosen in such a manner as to permit laboratory work to accompany the classroom work. For this reason the material proceeds rapidly in the early chapters into the discussion of amplitude- and time-scale factor adjust- ment and the setup of linear systems of differential equations. Non- linear components and function-generating techniques are also con- sidered in the early chapters. The later chapters consider such topics as the application of analog computers to the solution of problems other than ordinary differential equations, a more detailed description of computer components, checking computer results, and repetitive analog computers. The book is concluded with a brief introduction to the logic operation of the digital integrating differential analyzer. A brief review of the terminology of differential equations and a brief introduction to operator notation is included in the Appendix as an aid to the less advanced reader. The author is indebted to Professor C. E. Warren of The Ohio State University for his valuable suggestions during the preparation of an earlier form of the manuscript, and to Professor R. 'T. Harling of the U.S. Air Force Institute of Technology and other faculty members of The Ohio State University and the U.S. Air Force Institute of Technology who have influenced the writing of this book. The author wishes to acknowledge the able assistance of Mrs. Ada Williams for the typing of the earlier manuscript and of Mrs. Frances Borum for typing the manuscript in its present form. CLarence L. Jonnson CONTENTS Preface . 4 ww we ee el Chapter 1. Introduction -l. Historical Development . ., te Classification of Computing Equipme nt l -2. =. Problems Solvable on the Electronic Analog Computer 4. Major Components -5, Summary wheres 2. The Linear Computer Components . Introduction . . 22 The Operational Amplifers . Potentiometers , ; 24. Computer Design Differences 2-6. Concluding Remarks . Chapter 3. Time- and Amplitude-scale Factors d-1, Factors Influencing the Choice of Time Scale . 1-2. Determination of Approximate Problem Frequencies . dd, Approximate Frequencies of Higher-order Systeme t4. The Influence of Recorder Characteristics on the Choice of Time Scale . re ‘#6. Performing the Time-acale Change 4-6, Choice of Amplitude-scale Factor 7. Approximating the Magnitude of the Variables of A Problem ‘8, Summary Chapter 4. The Synthesis of Servomechanism Systems lL. Introdwetion 2. The Block-diagram Notation. . 4, Betup of a Simple Servomechanism System |, Setup of a Transfer Function f, Setup of the Factored Form of a Tranafe r Function ; i Surnmary Chapter 6, Multiplying and Resolving Servos fel. Introduction fed. Borvomultipliors OO. Division. . 2. 2. 1 6 ee ltl lll ied, Bquare Root 2. . 4 6 te thle Vil Se oe bo 12 14 17 20 22 Z4 25 at) 32 42 45 45 4u f2 66 fiz 65 ot) 74 Sf. §-6. CONTENTS Resolving Servos . . os. Choice of Rectangular or Polar Coordinates Chapter 6, Additional Computer Techniques (1. 6-2. 6-3. 6-4. 6-5. A 6-6. 6-7. Introduction Analytic-function Generation, Generalized Integration ximate Differentiation . Sean ans plifer Circuit Representation of 1 a Second-order System Computer Instability . a er ee ee Concluding Remarks . Chapter 7. The Representation of Nonlinear Phenomena 7-1. 7-2. 7-3. 7-4. 7-5. 7-6. 7-7. 7-8, Introduction . ‘ Differential Relays and Diodes—General ok Applications of Relays and Diodes to Simple Limiting Other Diode and Relay Circuits . Lo. Limiting the Output of an Integrator Representation of a Unit Impulse The Repetitive Unit Impulse. Approximation of a True Time Leg. Chapter 8. Multipliers and Function Generators 8-1. 8-2. a. 8-4. 8-5. ef. 8-7. 8-8. 8-9. 8-10. 8-11. &-12. Introduction. The Time-division Multiplier Quarter-square Multipliers. The Crossed-fields Multiplier. Other Multiplying Devices Function-generating Equipment . The Input Table _ . The Photoformer Diode-type Arbitrary-function- generating Equipment Tapped Potentiometers ces Resistive Materials as Function Generators Conclusions, Chapter 9. Miscellaneous Applications of the Electronic Analog Computer §-1. 9-2. 9-3. 9-4. 9-5. 0-6. 9-7. 9-8. Introduction ; Simultaneous Algebraic Equations ; Partial Differential Equations (Introduction) . The Solution of Eigenvalue Problems . . Replacing Partial Derivatives by Finite Differences ; Computers as Curve-fitting Devices. The Roots of Polynomial Equations Conclusions. — ss 8 Chapter 10. Analog Computer Components and Computer Control 10-1. 10-2, 10-3, Introduction. Stabilized D-C Amplifiers. kon ee Inherent Erroraof D-C Amplifiers. . «© «© «+ : 89 89 07 08 100 105 107 107 109 113 120 123 126 127 136 137 142 143 144 144 146 151 153 157 159 163 166 166 168 168 171 175 176 181 184 184 180 CONTENTS 10-4. Overload Warning Systeme 10-5. Boost .. 10-6, Operate-Reaet Systems 10-7. Potentiometer-setting Systems 10-8. Automatic Programming . Chapter 11. The Checking of Computer Results ll-1. Introduction .. 11-2, Cheeking Problem Preparation . 11-3. Partial System Checking . li-4. Checking Problem Stability . 11-5. Specialized Checking Procedures 11-6. The Role of the Digital Integrating Differential Analyser | in n Prob- lem Checking 11-7. A Dynamic Check of Computer Operation. 11-8 Accuracy and Precision Soe Chapter 12. Repetitive Analog Computers 12-1. Introduction 12-2, Advantages and Disad: vantages of Rapatitive Computers 12-3. Philbrick Computer coe 12-4. Initial Conditions . 12-6. Problem Setup. . . 12-6. Trends in the Design of Repetitive Computers Chapter 13. The Digital Integrating Differential Analyzer 13-1. Historical Development 13-2, Design Features of the Newer Digital Integrating Differential Analyzers. 13-3. Operation of the Digital Integrating Differential Analyser 13-4. The Number System . tok 1-6, Communication between Integrators 1-6. The Coding Diagram . 13-7, Conclusions. Appendix A-1. The Terminology of Differential Equations A-2. Time Constants and Natural Frequencies . A-3. Operator Notation . A-4, Conditions for Stability A-§, The Equivalence of jw and the Differential Operator " Inaleg . 194 195 195 198 199 208 210 213 214 216 216 217 222 227 229 2o1 233 233 235 oT 242 242 244 247 251 252 255 257 CHAPTER 1 INTRODUCTION 1-1. Historical Development. In the period of years immediately following World War II, there has been developed a tool which has lal much influence upon the methods of analysis employed by engi- neers. This tool is the electronic analog computer. The credit for its early development, both in this country and in England, is not easily placed because of wartime security measures, If any two men can be credited with the first. published use of oper- wlional amplifiers as computer components, they are C. A. Lovell and I), B. Parkinson of Bell Telephone Laboratories. Their use of oper- ntional amplifiers was in the computer of the M9 antiaircraft-gun director built by Western Electric Company.* J. B. Russell of Columbia University noted the circuits utilized in the M9 com- puter and brought them to the attention of Ragazzini, Randall, and ussell,'} who proceeded to build the first general-purpose electronic inalog computer under contract with the National Defense Research Committee. This work led to the publication of the first article, in May, 1947, describing the operational amplifier as a computer com- ponent, G, A. Philbrick is credited by some with having independently pioneered the use of high-gain d-c amplifiers as computer components in unpublished work conducted prior to World War II. In 1947, the Reeves Instrument Corporation, under a Navy con- tract, developed a computer which was the forerunner of the present- ny REAC, At about the same time, many others began the inde- pendent development of analog computers. No attempt will be made hore to assign credit for the very rapid developments made since the publication of Ragazzini's first paper, The list of individual contribu- tora would be quite large and would require a major research effort to oneure proper recognition for all concerned. Due credit should also lo given to those men responsible for the earlier development of the * Instruction booklet prepared by the Bell Telephone Laboratories for the Weetern Electric MO antinireraft-gun director, | Neforences denoted by superscripts appear at the end of each chapter. 1 E 2 ANALOG COMPUTER TECHNIQUES d-c amplifier. Without their efforts, d-c analog computers would not be possible today. The main purpose of this brief historical review has been to impress upon the reader the relatively short time between the construction of the first d-c analog computer and its general acceptance as a tool by engineers throughout the country. Ragazzini’s paper was published in 1947, and by 1949 several computing facilities had sprung into existence using equipment similar to that described in the paper. It is the rapid development of the field of analog computation that has inspired the preparation of this book. The development of new equipments and techniques has left a gap in the literature such that each individual when learning to use electronic analog computers must, to a certain extent, travel the path of learning the hard way. There is considerable literature available, in the form of papers, that will guide the computer operator in the application of special techniques to par- ticular problems, but nowhere does there exist a treatment of the use of the analog computer sufficiently simple to serve as introductory material yet sufficiently complete to be useful to the more advanced machine operator. It is the purpose of the author to attempt to shorten the period of transition from neophyte to accomplished analog- computer operator for the engineer learning to use this new tool. 1-2. Classification of Computing Equipment. ['or many years the engineer has had available computing devices of various types to aid him in his work. The classification of the various instruments avail- able should help to clarify exactly the type of equipment to be dis- cussed extensively in the following pages. Computing equipment may be divided into two main classifications: analog and digital. The analog computer operates by representing the variables of a problem by physical quantities easily generated or controlled such as shaft rotaitons or electrical voltages. The representation is such as to give continuous correspondence to the variables of the problem being studied. A digital computer counts and obeys logic rules exactly. A major difference between analog and digital computers is the accu- racy attainable. Accuracy of a digital computer can be extended by simply carrying more significant figures, whereas the accuracy of an analog device is limited primarily by the accuracy of individual com- ponents and of the measuring device. There are many types of analog computers in use today. These may be subdivided into two broad categories: general-purpose and special-purpose, Figure 1-1 shows a few of the analog computers and the class to which each belongs, The special-purpose computers will not be considered further in this work, The general-purpose com- INTRODUCTION 3 puters can be further subdivided into two classes: the direct- or physical-analog computers, and the indirect- or mathematical-analog computers. The network analyzer, which belongs in the direct-analog class, is the oldest of the general-purpose analog computers. Its operation is based upon the analogy of the mechanical behavior of dashpots, springs, and masses to the corresponding electrical behavior of resistors, condensers, and inductors. There are several reasons for the network analyzer’s not having achieved the widespread use that the electronic analog computer has Analog | | General Special purpose purpose | Planimeters Direct Indirect Computing bombsights Gun directors. Link trainers Scale models Equivalent D-c electronic Wind tunnels circuits analog computers Polynomial evaluators Network Mechanical ; analyzers differential F-86D flight simulator analyzers Digital integrating differential analyzers Fra. 1-1, Classification of analog-computing equipment. realized. First, the cost of a large analyzer installation is in the same order of magnitude as that of the large-scale digital computer and the mechanical differential analyzer ($100,000 to $500,000). Second, although the equivalent electrical networks for mechanical systems may be easily derived, the actual simulation is much more difficult bocause of the absence of perfect components. The resistance associ- ated with inductors, the leakage of capacitors, and the inductance of reaiators all introduce errors into the system. It is the consideration of such factors that makes the simulation of many systems more diffi- cult than on the electronic analog computer. Despite its disadvantages, there are computational tasks for which ihe physical-analog computer remains unexcelled. An example is the analysis of electrical power-distribution systems, Several of the major 4 ANALOG COMPUTER TECHNIQUES network analyzers in the country are kept busy a large portion of the time on this one type of problem. Other problems such as the dynamic analysis of structural problems, e¢.g., an aircraft wing, can be handled well on this type of computer. New network analyzers will probably continue to be built by power companies, aircraft: manufacturers, and universities to supplement the capacity of existing facilities. The mathematical-analog computers include the mechanical differ- ential analyzers, the digital integrating differential analyzers,* and the electronic differential analyzers. The mechanical differential analyzers are capable of accuracies up to five significant figures and usually con- sist of ball-and-disk integrators having mechanical or electromechani- cal couplings. They are at present in a very unenviable position of being squeezed by the accuracy of the digital computer on one hand and the speed and ease of operation of the electronic differential ana- lyzer on the other hand, Since the cost of a mechanical differential analyzer 1s of the same order of magnitude as that of the large-scale digital computer, it is very doubtful whether another large computer of this type will ever be constructed. The digital integrating differential analyzer is a newcomer in the computation field. The existence of such devices is widely known, but few people, other than those actively engaged in their operation, understand their basic principles of operation, An attempt will be made in Chap. 13 to familiarize the reader with this type of equip- ment in order that he may be more aware of its capabilities and limitations. The electronic differential analyzer is the device that will be treated most extensively in the subsequent pages. Other than in Chap. 13, the major effort will be to present the limitations and capabilities of this class of equipment. Any future reference to analog computers will mean specifically the electronic differential analyzer. 1-3. Problems Solvable on the Electronic Analog Computer. There is a considerable overlap of fields of usefulness of analog and digital computers. In general, a large-scale digital computer can do any job which can be accomplished on an analog computer, but many prob- lems can be handled adequately on an analog computer far more rapidly and easily than by any other means, The correct choice of *The digital integrating differential analyzer ia, by the strictest definition, a digital computer. The technique of problem preparation ia, however, very similar to that used for the other mathematical analog computers, For that reason the author has arbitrarily listed the computer as an analog device, Bome authorities prefer to list the digital integrating differential analyzer in a category sparate from other computers, i-c., aa a digital analog compuler, INTRODUCTION 5 computer is important to the economical solution of problems. Fae- tors to consider in making the choice are the accuracy required and the nature of the problem. In general, if more than four-significant- figure accuracy is required, the electronic analog computer cannot satisfy the requirements. Few analog computer laboratories can do that well. The type of problems best adapted to solution on an electronic analog computer are those involving systems of simultaneous differ- ential equations, linear or nonlinear, with constant or nonconstant cocficients. Fortunately, the complexity of problem setup is increased only slightly for nonlinear problems and problems involving non- constant coefhcients. Some problems, other than those which belong in the category of ordinary differential equations, can be satisfactorily handled on an analog computer. These will be discussed in later chapters. Analog-computer results are normally presented graphically as a continuous plot of the variable quantities. In many instances this method of presentation is most convenient for engineering use. The graphical representation of results has another important aspect. It is relatively easy for the engineer, as he operates the analog computer, to visualize the results as the actual dynamic response of the physical svatem under investigation. Thinking of the analog computer as just a mathematical device is to be discouraged as much as possible. The usefulness of the computer will be greatly enhanced if the operator views it as a tool to help him think in terms of the physical system. 1-4. Major Components. The major types of components of an electronic analog computer are relatively few in number. First, and the most important components, are d-c amplifiers or operational amplifiers. It is these amplifiers that become summers and integrators upon the addition of proper feedback and input impedances. Second, il is necessary to be able to set coefficients in a problem. As will be shown later, this may be accomplished either by the use of potenti- ometers or by adjusting the ratio of feedback and input impedances applied to the operational amplifiers. Third, it is necessary to have a set of controls capable of starting and stopping the computation. In addition, it is desirable to have a control position to perform the funetion of holding the problem solution at any point in the solution. ‘These control operations are performed by a system of relays. Usu- ally provision is made in the control system for the automatic appli- eation of the initial conditions of the problem while the computer controls are in the nearer position, A knowledge of the actual oper- ation of the operate-reset relays is unimportant for the solution of " i] ANALOG COMPUTER TECHNIQUES simple routine problems but is very important to the operator as the problem complexity increases. If problems other than those involving linear differential equations with constant coefficients are to be solved, units capable of multiply- ing variable quantities must be provided. As problem complexity increases, there arises more and more frequently the need for arbitrary- function-generating equipment capable of generating functions not easily represented mathematically. Similarly, more complex problems often require the representation of nonlinear phenomena. This is accomplished by the introduction of diodes or relays into the com- puting circuits. The basic components mentioned in the preceding paragraphs, together with suitable recording equipment, make up the major por- tion of the equipment used in the solution of problems on analog com- puters. Each of these components will be treated extensively in suc- ceeding chapters. 1-5. Summary. It is desirable to emphasize to the beginner that the understanding of the manner in which an analog computer per- forms its operations is extremely important. This is true, perhaps even toa greater extent than for any other computational aid. With- out an understanding of the actual operation and limitations of the equipment the operator can never rise above the level of knob twister. An effort will be made throughout the remainder of this book to present the fundamental principles of analog computation in a manner that will allow the reader to attain that understanding easily. REFERENCE 1, Ragazzini, J. R., R. H. Randall, and F. A. Russell: Analysia of Problems in Dynamics by Electronic Circuita, Proc. [RE, May, 1947, pp. 444-452, CHAPTER 2 THE LINEAR COMPUTER COMPONENTS 2-1. Introduction. Theoretically, there are two logical philosophies along which an analog computer for use in solving differential equa- tions may be developed, The first of these might be based upon repeated differentiation, and the second upon a process of repeated integration. From mathematical considerations, both systems are | 2 ——— iz ; © anf Sed i=l La (a) eo "~~ 2 ei (5) &o e, = Ke; © al (ec) Fie, 2-1, Symbolism frequently used to represent the computer components neces- sary to solve linear differential equations with constant coefficients, All voltages e and ¢, referred to in the diagram are varying d-c voltages measured with reapect to a common ground. The components shown are (a) an integrating amplifier, (b) a summing or inverting amplifier, and (c) a potentiometer, adequate. From an engineering approach, however, the process of differentiation has a serious drawback, Differentiation is a ‘noise’’- amplifying process, and all electronic gear, to a greater or lesser extent, produces random noise. This noise, however slight, results in a much higher noise level if differentiation is used. The second possibility, that of repeated integration, offers no such difficulty; the noise is amoothed, since integration is an averaging process. IJtepeated inte- gration ia the basis of the electronic differential analyzers as we know them today. 7 bt ANALOG COMPUTER TECHNIQUES In order to solve a system of linear differential equations with con- stant coefficients, the following types of equipment must be available: 1. Devices capable of performing the process of integration 9, Devices capable of summing several quantities 3. Devices capable of multiplying a quantity by a constant 4, Devices capable of multiplying by the constant —1 The symbolism of Fig. 2-1 may be adopted to represent the above devices, Note that the ability to perform sign inversions and to sum two or more quantities has been given to the integrating amplifier and to the summing or inverting amplifier. Assuming that the above devices are available, they may be interconnected in such a manner as to produce the solution of any linear differential equation with con- stant coefficients. The method of connection is illustrated by the following example. Example 2-1 a OF ay SF + ase = f(b) (2-1) Equation (2-1) may be rewritten in the form dr dr dt = ~Oi — gt + fit) (2-2) Then, assuming that d*z/df* is a known quantity, it may be integrated to give dx/dt and this in turn may be integrated to produce z. From Eq. (2-2) above, it a d*x _de . (a2) ar Ne a ~agx tit) de dt * a, 2 dz + Fie. 2-2. Computer circuit showing the component interconnections necessary to aolve Eig. (2-1). can be seen that, having z and dz/dt, it is only necessary to multiply each by an appropriate constant and sum them together with f(f) in order to get dx jdt, the quantity that was assumed to have been known, The block diagram of Fig. 2-2 illustrates the method. First examination of the block diagram may lead one to believe that he is reach- ing down and lifting himself by the bootstraps. This ia not the case, however, as the system operates much as any closed-loop servo system. The rate of change of THE LINEAR COMPUTER COMPONENTS 9 a quantity depends on the magnitude of the quantity and the time history of itself and ite derivatives. It is very important that one reason out the flow of informa- tion in the above closed-loop system. 2-2. The Operational Amplifiers. In the above illustrative exam- ple, the magnitude of the constants a, and a; must be less than unity, ca | K, " Ke Tu &,* - ft >. Kyejdt tq ——————__ f,.. ne : (6) Fig, 2-3. More complete symbolism representing (a) the integrating amplifiers- (}) summing amplifers, The symbols show the gain associated with each input of the amplifiers, since a potentiometer is only able to multiply by a constant less than unity. This need not be the case in general, however, as the integra- tors and summing amplifiers have the ability to multiply by constants other than unity. The transfer functions of an integrating amplifier and of a summing amplifier are, therefore, more completely repre- ty sented as in Fig. 2-3. Until now, it has been assumed i High-gain that the processes of integration « 4 | Se amplifier &s and summation can be performed. =A It might be well at this point to Fic, 2-4. Block diagram of an opera- show one means by which this may tional amplifier showing the high-gain , id A J detailed bl t d-c amplifier and the feedback and in- e one. more detal . oc put impedances zy anc 2,. diagram of the above operational amplifiers is given in Fig. 2-4. In the figure — A represents the gain of the amplifier, and z, and 2; are the feedback and input impedances, respectively. If the amplifier is assumed to draw no current at its input grid, then from Fig. 2-4, Eqs. (2-3) to (2-6) can be written hi ta (2-3) i= — “ (2-4) pe ot % (2-5) ot ¢ = —Ae, (2-6) 10 ANALOG COMPUTER TECHNIQUES From Eqs. (2-3) to (2-5), a zy Substituting Eq. (2-6) in Eq. (2-7) gives e+ eo/A _ _ Go/A + & (2-8) mi zy Simplifying Eq. (2-8) gives @ Aa ej zp +i + Az, (2-9) Multiplying numerator and denominator by 1/ A and factoring 2 from the denominator gives the final form of the transfer function @ ow _ 2-10 ei z 1 + 1/A(z;/2% + 1) 2-20) If A is sufficiently high, a good approximation for Eq. (2-10) is fe 1 ei Fi (2-11) An alternate and simpler derivation of Eq. (2-11) can be made by assuming that the voltage e, is equal to zero in Fig. 2-4. This assump- tion is valid to a good approximation providing the gain A of the amplifier is sufficiently high. Suppose the amplifier output ¢ is restricted to remain within some finite region of voltage, usually +100 volts. Since e, = — Ae, it is apparent that e, is approximately equal to zero for A >> 1. Further, since the amplifier can be assumed to draw no current, ty = Ta and hHeEe- == -— 1 z, zy Rearranging and solving for ¢,//e; again gives fe, — Fi . as (2-11) The derivation of this important result by the latter method has the disadvantage of not emphasizing the nature of the approximations made in the derivation. For the present, however, it is not necessary to discuss the second-order effects of the operational amplifier, ‘These effects are, of course, important in the design of the amplifiers and in attempting to understand the limitations of the equipment, They THE LINEAR COMPUTER COMPONENTS ll are, however, less important to the beginner learning to use the analog computer. In later derivations of transfer functions of computer circuits, the derivations will frequently depend upon the approximation Bo ts i A =O l'or most analog-computing equipment the approximation is very good. The value of amplifier gain ranges from 2 * 10° for some repetitive equipment to over 10* for at least one of the better commercially avail- able computers. In Eq. (2-11), if zy and z; are both 1-megohm resistors, Ce _ Fr ej zy —] (2-12) Similarly, if 2; is 0.1 megohm and z, is 1.0 megohm, the transfer func- tion is e& ew 1X10 | a 4 O1X10° — Kquations (2-12) and (2-13) show that, to change the gain of a sum- ming amplifier, it is only necessary to vary the size of the input resistor. (Normally, the feedback resistor is held constant at 1 megohm.) If zy is a condenser, then* 10 (2-13) 1 1 1 = 0 = 56 (2-14) cx I/pC_ and rim = R ~~ pRO (2-15) if = 1% 10 farad and R = 1 & 10° ohma, fo 1 __i1 « 10°X10p pp 10) or é = —Je dt (2-17) Similarly, it may be noted that the gain of an integrator can be varied by changing the size of the input resistor, just as in the case of the summing amplifier. A differentiator could be as easily formed by letting z; be a con- denser and z,; be a resistor: then &, gy R — a =— — e4 a0 pl * The equivalence of jw and the operator p is demonstrated in Bee. A-5. A brief introduction to operator notation is included in Bec, A-d. —pRC (2-18) 12 ANALOG COMPUTER TECHNIQUES Differentiation is seldom used in the solution of problems on analog computers, as the noise amplification produced by differentiation is very undesirable. At times it is preferable to rewrite a set of differ- ential equations completely rather than to differentiate. If differenti- ation cannot be avoided (a very rare situation), then an approximate differentiation may be used to keep the noise at a usably low level. A circuit for producing approximate derivatives will be discussed in a later section. So far, the transfer function of the integrator and of the summing amplifier have been developed without showing that each has the iy — fa iy iy =—-r ; iy Fy 4 Th -A &s * i * * Fa Ca Fa Fia, 2-5, Block diagram of an amplifier having a feedback impedance z; and several input impedances z,, %,.. . , % connected in parallel to the grid input of the amplifier, ability to sum several functions at its input. This may be most easily demonstrated by considering the block diagram of Fig. 2-5. In Fig. 2-5 the single input has been replaced by several inputs a & ».., x. Since the gain of the amplifier is very large, e, may be set equal to zero, since lle 0 (2-19) Ba a=— 7 The current drawn by the amplifier may be neglected as before; therefore te Fito bin = Oe (2-20) Replacing i., i, . . . , i, by equivalent expressions, Eq. (2-20) becomes fo ee. Ee _ oo - . + rs T + :. zy (2-21) me — Fe, He... Hy oR e -! or e, = 2. fa os &, :. En zy » x (2-22) 2-3. Potentiometers. Potentiometers are frequently used in analog- computer setups to perform multiplication by a constant loss than 1, THE LINEAR COMPUTER COMPONENTS 13 Very often these potentiometers are 10-turn helical wire-wound types of high resolution and excellent linearity (usually from 1 to 0.05 per cent of full scale). By means of a vernier dial, parameters of the problems may be accurately and conveniently set. A loading correc- tion will have to be applied, in most cases, to compensate for the load- ing on the potentiometer, The potentiometers most commonly used in analog computation vary in total resistance from 10,000 to 100,000 ohms. The lower limit e Ry NAG . “ m1 e] ry High-gain “ . ad ¢z | amplifier (a) = (5) Fie. 2-6. Block diagram of a potentiometer when used as the input to an amplifier: (a) schematic representation; (b) detailed circuit connections; (c) equivalent. cir- oult representation from which the p