Analog Computers

Reference / Paper · 1970

Analog/Hybrid — What It Was, What It Is, What It May Be

Read the PDF (12 pp) ↗

A survey article by Arthur I. Rubin of Electronic Associates, Inc. (EAI), presented at the Fall Joint Computer Conference (FJCC) 1970. It traces the evolution of analog and hybrid computers from WWII 'zeroth generation' machines (BEAC, EASE, GEDA, REAC) through successive generations defined by patch panels, improved component accuracy, and the introduction of digital elements, to the hybrid systems of 1970, and speculates on their future. Key manufacturers (EAI, Beckman/Berkeley, Reeves, Philbrick, GPS) and landmark machines are discussed alongside accuracy milestones, application domains, and the historical tension between analog flexibility and digital precision.

Manufacturer
EAI
Author
Arthur I. Rubin
Year
1970
Type
Reference / Paper
Language
English
Learning track
introduction
Pages
12
  • EAI
  • analog computer history
  • hybrid computing
  • simulation
  • electronic analog computers

← Back to the Reference Library

Analog/Hybrid — What It Was, What It Is, What It May Be

Analog/hybrid-What it was, what it is, what it may be by ARTHUR 1. RUBIN Electronic Associates, Inc. Princeton, New Jersey THE ZEROTH GENERATION for granted. Furthermore, an arbitrary function generator was also not shown. Apparently, that device, which is necessary to make analog computation capable of solving any problem, was developed later or was considered an oddball, along with the comparator (which is really represented by the dry friction element, provided that the output of the dry· friction element is Introduction The history of the analog computer goes back to antiquity, where tax maps were first reported being used for assessments and surveying. However, I shall confine this paper to the analog computer as it evolved from World War II to the present time. For those interested in the history of the analog computer, from antiquity to World War II, I refer the reader to an excellent introductory article by J. Roedel, Reference 1. The "Palimpsest" in which Roedel's history of the analog computing art is included is in itself an excellent history of analog computers in the early days d~ting from World War II to about 1954. From page 4 of the Palimpsest, I would like to show a diagram of computing devices as visualized by George Philbrick for an article in Industrial Laboratories in lVlay, 1952. Of interest to us in this diagram on the analog side, is the separation, at the bottom, between fast and slow analog which I will discuss shortly. We will'also note the presence of hybrid at the very top, and this article was written in 1952! Of course, l\1r. Philbrick's "hybrid" was reserved for the use of the analog computer first to obtain a ball-park idea of a solution, then followed by a separate digital solution to obtain a more accurate answer to the same problem. I am certain that very few people thought of this as being hybrid computation at the time. However, consider this definition in the light of later work reported by Mark Connelly (Reference 2) in his use of a "skeleton" representation of a problem on the analog in conjunction with a more accurate representation of the problem on the digital. It is interesting to observe the basic operations as defined by Roedel in Reference 1. This is shown in Figure 2. Note that the early practitioners of the analog art considered differentiation to be a basic linear element for the fast speed computers and did not show potentiometers, since the latter must have been taken Figure I-Structure of computing devices as visualized in 1952 641 642 Fall Joint Computer Conference, 1970 BASIC LINEAR COMPUTING ELEMENTS SCHEMATIC DIAGRAM MATHEMATICAL OPERATION BLOCK DIAGRAM Rf el eZ ei ~ ~A el ~Z RZ ~ .. Rf Rf eo. e l - . eZRl RZ ADDITION ~ -f ei ~ .. Rf eo_-~i R. I SCALE CHANGE Cf -i ~ D·· 1D ~.. e, 1 e o - -- f i dt CfR, INTEGRATION Rf e, ~~ e, de·1 eo.R C · _ f I dt DIFFERENTIATION NON-LINEAR OPERATIONS e. MULTIPLIER LIMIT ez -sfI DEAD ZONE OR Y FRICTION #- ~ e. e. +C, -A ABSOLUTE VALUE I +C - C " e. e. I I I ~ =lM ~ ---1" ~ -1' .. .. .. eo· Ke,el for -C< e i < C. K_, e ofor _,<-C, e,>C, for -C<ei<C, for -0-0· Con.tant 0 -, <- C, e i > C, eo· Kei for -i>O, eo·+A ei e.I eo + . -1" .. for e i < 0, Amplifier ei ~ Figure 2-Basic linear and nOll-linear analog operation and components -o--A e o- Kleil Analog/Hybrid 643 Figure 3-Pullman-Standard Car Manufacturing Company's analog computer installation used to drive a switch or a gate connected to some other computing element). There was a great deal of emphasis in those days on the solution of linear differential equations, obviously because those required the simplest computing components. Perhaps also, because one could Figure 5-Boeing analog computer, courtesy of Boeing Airplane Co. obtain check solutions to such equations with pencil and paper, and computers, being relatively new, could not yet be trusted. Hardware Figure 4-Computing equipment in a typical rack assembly The major manufacturers during this initial period were the Boeing Company which made the BEAC computer, the Berkeley Scientific Computing Company which made the EASE computer (Berkeley subsequently became part of Beckman Instruments), the Goodyear Aircraft Company which made the GEDA, the IDA computer with which I am not familiar at all, the George A. Philbrick Research Company which made the GAP /R computer, and finally, there was the Reeves· Instrument Company which made the REAC computer. Some pictures of these early analog computers are shown in Figures 3 through 8. Figure 3 shows a GAP /R installation while Figure 4 shows a close-up of how those computing components were interconnected. You will note an absence of a patchboard. Can you imagine checking this one out today? Note the telephone jack panels on the Reeves computer and note also that the Berkeley and the Goodyear computers are the first ones with patch panels. These figures date from about '1952 or 1953. EAI, which was just beginning to build analog computers, does not even show. The typical size of computers in those days ranged from about 20 amplifiers up to 80 644 Fall Joint Computer Conference, 1970 need for such devices in the design of their own equipment, whether it was airplanes, electronic gear such as radars, or control systems. Philbrick, on the other hand, and possibly also the Berkeley Company, concentrated from the very beginning on the process control applications. Applications Figure 6-Goodyear GEDA computer installation, courtesy of Goodyear Aircraft Company, Akron, Ohio amplifiers, which was considered to be fairly large. One manager, in fact, was proud of the fact that he could expand his 80 amplifier installation to 160 without requiring any additional wiring. The accuracy of the components was of the order of one percent (and that applied to resistors and capacitors as well as to the electrical components). Overall solution accuracies on what was then considered medium size non-linear problems was of the order of five percent. One final point of interest is that several of these manufacturers, mainly Boeing, Goodyear, and Reeves were primarily aerospace/defense manufacturers who saw the obvious Figure 7-Berkeley EASE computer, courtesy of J. B. RAE, Los Angeles The 2nd page of the table of contents of the Palimpsest is reproduced here in Figure 9 and shows the wide variety of applications that were actively investigated in the early 1950's. You will note in parti~ular the beginnings of an analytical attack on our environmental problems in the papers on the freezing and thawing of soils as well as flood routing. The analog equipment, especially that which did not have patch panels was generally purchased for a particular problem or problem type. For example, the Pullman Car Company would buy one for solving their "transportation equipment design" problem. An aircraft manufacturer would buy a computer to study the control system of a particular airplane. There was an almost complete lack of user conveniences leading to the ridiculous situation of being able to obtain a complete, single solution to a complex set of differential equations in 5 milliseconds, but having to wait several days, at least, to change to another problem, due to the lack of a patch panel and other amenities, such as a readout system. This type of inaccessibility (to the Hnext" Figure 8-Reeves computer installation, courtesy Reeves Instrument Company Analog/Hybrid problem) has been at the root core of the ailment in the analog field and has given the analog computer the reputation of being "inflexible." This ailment is still with us, albeit to a much smaller extent, and a cure is visible on the horizon, as we shaH see later. For further information on techniques and methods that were expounded in the early years of analog computation, the reader is referred to References 3, 4 and 5. This by and large represents the first generation analog; however, since I seem to have too many generations, as we shall see later, I will term this the heroic age, or the zeroth generation. This generation coexisted with the heroic age digitals, such as the ENIAC, EDVAC, the ORDVAC, l\1ANIAC, and the UNIVAC. 645 MODELLING of PHYSICAL PROCESSES............................................................ 146 The Electro-Analogue..............................................J. M. L. Janssen and L. Ensing 147 Discontinuous Low.Frequency Del.y Line with Continuously V.ri.ble Delay ......................................................................................................J. M. L. Janssen 162 Bucket·Brigade Time Delay..............................................................G. A. Philbrick 163 Solving Process-Control Problems by Analog Computer..................................... . ............................................................................ R. J. Medkeff and H. Matthews 164 ELECTRONIC ANALOG METHODS in DETAIL.............................................. 167 Precision in High.speed Electronic Differenti.1 Analyzers................................... . .......................................... .............................. H. Bell, Jr., and V. C. Rideout 168 An.log Computer Solution of • Nonlinear Differentiai Equation ....................... . ............................................ .......................... H. G. Markey and V. C. Rideout 177 The Study of Oscill.tor Circuits by Analog Computer Methods......................... . .................................................. H. Chang. R. C. Lathrop, and V. C. Rideout 184 A Differe~ltial.An.lyzer Study of Certain Nonlinearly Damped Servomechamsms..................................................... R. C. Caldwetl and V. C. Rideout 193 Appf~tiori of .n An.los Computer to Design Problems for Transponation THE FIRST GENERATION The next generation, here termed the first, more or less coincided withthe arrival of EAI on the scene, with its establishment of the firm need for a patch panel and an integrated set-up and readout console as part and parcel of the analog computer. In other words, human factors entered into the picture, also, this generation saw the arrival of the .01 percent component, such as resistors and capacitors, which allowed linear problems to be solved more accurately than the solutions could be dispJayed on a strip chart recorder, X - Y plotter, or oscilloscope. The credit for this shift in emphasis on more accuracy and more user conveniences must go to the manufacturers who went against the ideas of some of the then old line users, who kept pointing to the problems that were being solved and observing that much of the input data was unknown perhaps even within a factor of two of the correct value. These old time analysts recognized that there was no need for obtaining very accurate solutions to such problems. However, they overlooked the crutch available to the insecure analyst if he can get a repeatable, accurate answer even though the model is not exact. This analyst then has fewer questions from his management, because when he goes back for reruns, he gets the same old answer to compare with at the same time, the solutions for the new set of parameter values. Thus, he and management both think they understand the problem. (Aside-I learned this trick early in the game. In order to convince my management and customers as to the validity or correctness of a set-up to a problem, I always went back to a "standard" solution, if a check solution was not available. And if the standard or check didn't repeat, then I would hopefully "tune-up" the equipment to produce a "replica" of the check solution. In some cases, I must confess, I may have "de-tuned" the equipment to produce the so-called "check".) EqUipment ...................................................... ............................................J. Roedel 199 ANALOG STUDIES of ELECTRIC POWER SYSTEM TRANSiENTS........ 216 Surge and Water. Hammer Problems............................................... H. M. Paynter 217 Methods and Results from MIT Studies in Unsteady Flow ........ H. M. Paynter 224 The An.log in Governor Design ...................................................... H. M. Paynter 228 Tie Line Power" Frequency Control........ .................... H. M. Paynter 229 Electronic Computer for Resolution of Steady - St.te Stability Problems and Panicularly for Automatic Control Studies................................... .I. Obradovic 233 How to Select Governor Parameters with Analog Computers ............................ . ...... ..... .................................... ............ E. C. Koenig and W. C. Shultz 237 COMPUTER TECHNIQUES in HyDROLOGy...................... ........................... Flood Routing by Admittances... ..................................... H. M. Paynter 239 240 ANALOG SOLUTION of THERMAL PROBLEMS......................... 246 Freezing and Thawing of SOils .................... H. P. Aldrich. Jr., an<l H. M. Paynter 247 The Analogue Computer and Automatic Control Applications .... R. J. Bibbero 261 Process Regulation with Analogue ControL .................................. R. J. Bibbero 264 Continuous Electric Representation of Nonlinear Functions of n.Variables ................................. ... G. A. Philbrick 266 Fir .. 1 IlrilitinK PH;; SnoolHl printinr.: 1'1'& rhirtl printing I'Jf,O Fourth printing 1965 Figure 9-A portion of the table of contents of the Palimpsest Conveniences such as a digital volt meter readout of amplifiers and all other components via push-button selectors, servo set pots as well as experiments with quarter-square multipliers and time division multipliers were introduced. The second phase lasted roughly from 1955 to 1960 and saw the rise of EAI from the position of young upstart to that of the major supplier of analog computing equipment. While EAI was rising, the period saw several companies such as Boeing, Goodyear, IDA (or perhaps Mid-Century) drop out of the industry. After these defections from the ranks of the manufacturers, the field of slow speed analogs was split amongst EAI, Berkeley, which by this time had become merged with Beckman, and Reeves Instruments. The high speed analog now had two manufacturers, the old Philbrick Co. and a newcomer to the high speed camp, the GPS Company. This period saw the 31R and the 131R and to lesser extent, the Reeves' C400 gain wide distribution. 646 Fall Joint Computer Conference, 1970 1958 NATIONAL SIMULATION CONFERENCE, continued Ehlers, "Standard Simulation Circuits" Gilbert, '"rhe Design of Position and Velocity Servos for Multiplying 1958 NATIONAL SIMIlLATION CONFERENCE Clymer, "Operational Analog Simulation of the Viration and Flutter of and Function Generation" Sinker, "The Card Programmed Diode Function Gener-ator" a Rectangular Multicellular Structure" Shen, ''Multiplier Circuits Utilizing Squaring Property of a Triangular Powell, "Distributed Parameter Vibration With Structural Damping and Wave" Noise Excitation" Pfeiffer, "A Four Quadrant Multiplier Using Triangular Waves, Diodes, Ladd and Wolf, "A Non-Real-Time Simulation of SAGE Tracking and BOMARC Resistors, and Operational Amplifiers" Guidance" Miller and Enger, "Liquid Transfer and Storage System Simulation by Active Element Computers" Azgapetian, "Some .Aircraft Probl_ S1aJlated by "ean8 of Z-forms" Boxer, ''Z-forms, and the Digital St.ulation of Dynamics" Ehlers, ''General Purpose DC Analog Computer with Transistor Circuitry" Pritsker, Buskirk, and Wetherbee, "Simulation to Obtain Systems Measure of Air-Duel Environment" Billinghurst and Single, "Extending the bandwidth of Precision Analog Systems" Nemerever, "A New Technique in System Perfor_nce Evaluat ion" Gilbert, ''Linear System Approximation by Differential Analyzer Simulation of Orthonormal Approximating Functions" BrllllllDer, "Solutions of Convolution Integrals by Analog Computers" Rideout, "Some Applications of a High-Speed Analog Correlator" Bekey and Whittier, ''Generalized Integration on the Analog Computer" Bruns, and Wilcher, '"rransistorized Relay Amplifier" Neshyba and Coffman, "Airborne Radar-Beacon Traffic Simulator" Munson and Rubin, Optimization by Random Search on the 'Analog Computer" Rawdin, "A Time ~ltiplexing Technique" Schwarm, "Computer Systems for Jet Transport Simulators" Heffron and Bristow, "A Method for Helicopter Rotor Performance SilllUlation" Bush and Orlando, "A Perturbation Technique for Analog Computers" 1958 NATIONAL SIMULATION CONFERENCE, continued Morrison, "APPR-I Simulator Description" The end of the period saw the introduction of the 231R computer, (See Figure 13) a machine which was to see much service in the '60s. Meilander and Hellman, "A Technique for Absolute Measurement of Analog Computer Capacitors" Gerlough, "A Comparison of Techniques for Simulating the Flow of Dis- Applications The applications of this era (the end of the first generation) perhaps are best described by scanning the list of titles of papers that were presented at the 1958 Fall National Simulation Council Conference (Figure 10). From the list of titles it is clear that the aerospace/ defense industry dominated applications, but there were a significant number of papers reporting new mathematical techniques and even applications of digital computers to the field of simulation. New hardware circuits such as the card programmed diode function generator and a quarter square multiplier were first described. Also included were descriptions of a much later transistorized analog, a computer optimization study by analog computers, as well as discrete event simulation by digital computers. crete Objects" Figure 10-1958 National Simulation Conference THE SECOND GENERATION The next generation which I must here call the second, lasted roughly from 1960 to 1965. The size of the analog computer at the upper end was getting physically larger and larger, which by virtue of the vacuum created at the small end led to the design of a small desk-top computer, which was the logical outgrowth of the transitorization of analog components. The first transitorized computers were of the small desk-top type and had a voltage range of plus or minus 10 volts. They Analog/Hybrid coexisted with their big brothers, the 100 volt vacuum tube computer, during this period. Another hardware innovation sa \,. the combining of heretofore separate fast and so called slow analog into a single machine. This occurred in both the desk-top machines as well as the large 100 volt machines. This latter turn of events, was brought about by the introduction and the widespread use of quarter square multipliers, replacing the old slow servo and time division multipliers. At the same time, manufacturers introduced the fixed diode type of function generator for such analytical functions as sines and cosines of angles and exponential functions. The above developments, in conjunction with the introduction of solid state highspeed switching made the high speed and the low speed analog in a single computer a practical reality. Two new companies were formed during this period that proved to be significant factors throughout the '60s. These were Comcor, Inc., which subsequently merged with Astrodata, and Applied Dynamics, Inc., which subsequently merged with Reliance Electric. These companies helped to fill the partial vacuum created by the withdrawal of GEDA from the field during the second generation and the subsequent rise and fall· of several other companies. During the same period (1960 to 1965), the Reeves Instruments Company, which had been in the business of manufacturing analog computers from the very beginning, more or less indicated that it was finished with this activity. The second generation equipment brought to fruition the concept of patchable, parallel, digital logic as an integral part of the analog computer. At the same time, it should be noted that the first conference on the use of combined analog and digital computers in a single problem simultaneously, (the first conference on essentially hybrid applications) was held in 1960, Reference 12, which is why I chose 1960 as a key date for the second generation. Hybrid computation evolved along several different paths. In one path a "stand-alone digital," was placed next to a "stand-alone analog" with communication between the two allowed via logic lines, control lines, an AD (analog to digital) converter with a multiplexer, and several D / A (digital to analog) converters. Along another path a "logic computer" was developed by Electronic Associates which consisted of a number of parallel, patchable circulating (memory) delay lines, "and" gates, "or" gates, flipflops, four bit registers, shift registers, and one shots (mono-stables) to the point where, if a programmer were clever enough, he could devise and patch together his own special purpose digital computer. The logic computer ultimately got reduced to a reasonably small, manageable complement of logic functions (see Figure 14) on later analog 647 Index: Volume 4 "'Ntol Comput~r 'ot~lt~en. OillUl1 Computer Control of. W I eMM/e,.nd. •. McChH.. .. Counter·Current CrysrlUIUlton Process, A".'ol Simul,lion of .II. W. L God'r~.nd It. D. ~nh.m .. Differential Equ.tions. The OE5-', A New 0,.,1,1 '.n '.n Computer for Solvina. L Le'lfin.... . Apr Dilir.' Differenli., Amlyzer.nd .ts Apphc.tion IS. Hybrid Compuli"1 System (I~nl. TM '.r.II~, O. A. R.eich.rdt, M. W Hoyr..nd W T LH . Feb Equlpotenti.! lines of. Potentill' f.eld, Plottlnl the. w. R. Lym.. . ...................... Feb. 81 Fotly-Step Qu.ntinr. H. ~tA:s .nd C. M. C.lUwn . . Mlly )01 26 HysteresiS, Use of tOBie.' EI~ts to Simul,te. C W. Schulte.. . ................... JuM 164 264 104 Imperfect Differenti'tor. A Note on the Simul'tlon of the. A. Bridgem.n .. ..................... J.n. Muhiplier, A four-Qu.dr.m. W H. Alliston ....... ,.n. Reticle Ev.lu.tion Studl~. A Mflhod for G~'.tin. 21 S~mple·~nd·Hold CircUli, Hllh Sentitivity, Di~~!i~~~i.~ ~r~~r;:r~':.~u~;dtoUt ~nd Simul~ted T~rleu Use'd In, M. O. C.mp ........ Apr. L. T.v~rnin; . . J.n Hybrid Computer, MAD.l~ ... ~ Prolnm for DiSiul Simul~hon of ~. V. C .idroot and L T~vernjn; .. Hybrid Compul~lion, A Discunion of I«klTYn' .. ApptOalch 10, It E. lord lind R. A. Ne5blf June Hybrid C~mputins System EIetlK"flf, TM Parlillef Dilit~r Differenti.. l An.tlyzer .. nd .1.. Apphe.. llon ~s ii, O. A. Reichilrdt, M. W. Hoyt. ~nd W T Lff . . Feb. Hybrid Sisnil' Processor, A Pu.lleI/Sequentilll, Stored Prolrilm,}. D. Crilndine.nd T C H ••~n . . Jan l.suerre functions, Function GeMr.hon on An.log Computer UsinS <d'neralized. C H Bur,in. Mlr. Measuring Statistics from QUilntlzed Dilla, Hybrtd Computer Techntqufl for, C. A Korn . Apr. Nonstahonary R.ndom Processes, Generation of, R. E Sry.n.. . ... J.n. Numeriul M~lhod for Simu',Ition, A N~, M.Fcwvlef.. Mily Pontry'gin's MilJumum Principle to Solve OMDimensiON' Oplimlz.tion Problems, With .nd without Constraints, on iln Ite,.tive ANlol Comp'ltt"'r, Usinl, H. L. Steinmetz .. tUM Pseudo-R.ndom Noi~, FJCPt"'rimt"'nls Usinl. It L H,Impton.. Punched-C.,d Prolrammer, MACON: MAtriX CONnector, L. /. K.mm.. 223 C. C. Willem5 ................. . '.n APACHE. It 0 'renn.n .. nd R. N. Lin~rpf 1-42 Let's fiCe .t, It 0 'renn.n.nd It. N. linftJ.,'~t ... Apt. 226 PACTOLUS Prolr.m for S,mul.tion of Antt"'nN Servos, U~ of th~, N. B. Kurek ............... . J~ne 16S 229 An.l08 D.tIi Procflsinl with. Hybrid Computer. W. C. McClintock... , ..... , ............... June 169 Anlilos Ori~nted 'nput L..n8u'It"', A Compilet with .n, M. L. Sf~;n. /. Ros~,.nd O. S. P.tlc~f ......... Mar. 159 382 Hum.n Intefilchon In S""U Groups, Computer Simul.tion of, J. T. Cull.horn and /. E. Cull.horn . . 'an 50 . . Apt 246 Input liIn.u.... A Compiler with A",'08 Oriented. M L. St~in, /. Rmt"', and D. B. P..tk~t .... Mar 159 . .Mlly 332 R,Indom Proces~, Cenefiltion of Nonst.tion"y, R. E. Sry.n.. . ... Jan. River Kit.bmi, Flood Simul.tor for the. K Otobil. K. Shib.ir.ni, .nd H. Kuw.t.. . .. Feb 42 Notw eene,.tor. A Hybrid ANlol-Dilital ps.udo·a.ndom, R. L. T. H.mpton ... 86 p.,.m~t.r Optimizer for ASTRAC II, A Hybrid .n Simul.lion, A New Numeric.l Method for. M. E. Fowler .. ....... Mly Simulation of a Chemic.l Feedt"'t System, An.lrsis .Ind. J. C. Vogt .•............. ..... . ....... Apr Simulation of. Spinning Rigid Body. Quaternion P.rameters in the. E. E. L. Mirchell.nd A. E. Roger5 ......... , ..................... JUM State Vari.ble Techniques irr An.'08 Compuler Programming, AppliCiltion of. /. L. Holmmond, Ir . . Mily St.tistics from Qu.ntized D.ta. Hybrid·Computer Techniques for Meuuring, C. A. Korn..... . ... Apr. 312 104 35 324 2S6 .n LNrnlnl Control Systems. A Comput~r Simul.ted On-LIM Experinwnlln. /. O. Hill, C. J. McMurtry, .nd K S Fu ................................ ·.Feb . .... Mlr 179 Ana'OI·0illt.l, a. A. Mitcht"'lI, Jt .............. June 199 P.rti.1 Differential <>per.tors, A New Method of Ana'yzin. tht"' TruncAtion f"o, in the Finit~ Represent.tion Difference of, R. Vichrwveuky ... Mar. 190 '~do· ••ndom Noi~ Genero1tor, A Hybrid 390 229 ANto1,Oi1it.1. R. L T. H.mpron. , ............ Mar. Tfilnsistors in Curr~t-Arylo. Computinl. t. P. Kefloot. . ..................... ,..... .. May Trunation Error in the Finite Itepresentlltion Difference of P.rtl.' t)iff~ti.1 Oper.tors, A New Method of Ana'yzinl the. R. Vichnfovets'y ...... . Mi:r. 119 339 190 SIMULATION Figure ll-Index to Vol. 4 SIMULATION, June, 1965 computers and became fully integrated with the analog computer instead of being a separate device. The largest analog computing consoles had upwards of 200 amplifiers in them. One example is the Applied Dynamics 256 (Figure 15) which had 256 amplifiers. There was a movement towards more "committed" amplifiers such as the class 0 type quarter square multiplier with many of the largest, most sophisticated users trending towards the class 0 resolver as well. This meant the computer was easier to use, but it became more expensive. The applications of analog during the period 19601965 are more difficult to characterize since by this time the well known industry-wide Spring and Fall Joint Computer Conferences, which are sponsored by AFIPS, had replaced the old National Simulation Conferences. To help plug the information on applications gap, a significant event occurred during the second generation, which was the launching and the publication of the new journal SIMULATION by Simulation Councils, Inc., under the editorship of John McLeod. The index of articles published in Volume 4 (June, 1965) is shown in Figure 11, as an example of the type of applications 648 Fall Joint Computer Conference, 1970 that were being done on these bigger, better and more powerful systems. It may be remarked in passing that even during this second generation period, indeed throughout the history of the analog, the analog has been used very much as it was originally used when there was no patchboard on the analog console. This method of use consists of committing the analog to a single problem, of very high priority, and tieing it up full time doing the same job over and over and over again, as exemplified by the.typical hardware or man-in-the-Ioop simulator. Very often when the project that required the simulator was completed or nowadays we would say cancelled, there was no further use or need for the analog computer, since no one else had been able to get at the machine during the "fat" days. Those analysts who had short duration, small problems, which can be considered to be ideal candidates for the analog computer, especially during the development or the "model" stage of the problem, were forced to go against their own wishes to the, by then, widely .available large, fast, digital computer of the 7090 class. These small, repetitive, studies went to digital not because the machine was fast, not because the digital was cheaper, not because it was better, not because it was more accurate, but simply because it was available! THE THIRD GENERATION The third generation has shown itself to be in existence from roughly 196,5 to the present time, 1970. The major hardware characteristic of this generation is the complete transitorization of the analog computer, for both the large scale 100 volt machine and the small scale 10 volt machine. A new scale machine evolved in between these two extremes, called the medium scale. A major hardware feature is the integral design of digital logic as part and parcel of most analog consoles, ~ 1962 1964 Typical Computer 231R (EAI) recorder now compatible with accuracy and re- continued peatability of computer. 231RV (EAI Electronic mode control of integrators, time- Beckman More useful bandwidth. scale selection (6 decades) via push buttons, more accurate multipliers and sinusoid generators. Digital logic control capability - for the first time analog has a full 10KC bandwidth in all components, Variable breakpoint and polarity, card-programmed function generators. This allows instant set-up of DFGs (takes only one hour to turn around a problem). (Mostly pot set t ings time.) !!!!: Typical Computer 1951 ClOO (Reeves) 1966 Ci-5000 ADI-4 analog cOlllputer - all gates (reset, hold, operate, duction .of removable patchboard. EAI 8800 are electronic) bandwidth up to and beyond 100 KC, EAl 680 1954 3lR (EAJ) 20 amplifier computer, expandable to 60; more l3lR (EAJ) C400 (Reeves) reliability e.timated as 60,000 hours KrBF for _plif1ers vs. 5,000 hours measured on 231R-V. accurate servo multipliers; integrated slaving 1956 Fully tranSistorized, more accurate, more reliable 20 amplifier computer; servo multipliers intro- system; .01% capacitors; .01% resistors, both C_puters can have 300-400 a ..pl1fier. in one temperature controlled. console. Analog directly controllable by .... 11 digHal c_puter. Integrated readout; human engineered for faster, hybrid co.putation. easier programmer use; electronic time division Easy ... ting with digital for Self-contained patchboard _ digHal logic (much, IlUch larger than in 23IR-V). multipliers; mechanical digital voltmeter; tube Card-set DFGs progr_ble from a standard IBM diode function generators. card. 1959 231R (EAJ) 100 amplifier computer; modular concept patchboard; significant improvements in amplifier bandwidths providing faster response and switching ti~es; compressed time capability (some jobs can be run ~ Typical Computer 1968 to ADl - Various Digital pots for microsecond (electronic gate) Present EAI - Various setup - or millisecond (reed relay setup), as fast as 10:1 real time instead of all at real large scale use of MDACs in hybrid interface, time); faster potentiometer readout; electronic software developed for autaaatic setup and digital voltmeter; solid state diodes in function checkout of hybrid analog computers. Direct generator; repetitive operation capability. digital/analog function generator (more accurate 1962 Improved More accurate 1/4 square multiplier; electronic than card set diode function generator) com- 231R (EAI) sinusoidal generator; point storage via trans is- pletely controllable from digital computer. tor circuit; card-set function generators; Mark 200 Figure 12-History of analog computer evolution since 1951 Analog/Hybrid 649 Figure 14-Digital expansion system by EAI (allows parallel patchable digital logic expansion to 10 volt systems, in a . self-contained desk top frame Figure 13-231R computer, courtesy Electronic Associates, Inc. small and large, which has certainly made pure analog computation, if we include this digital logic, more powerful than it has ever been. Another hardware feature is the complete flexibility of the multi-time scale integration capability of the analog, wherein one can have a choice of fast, slow or in-between speeds of solution as well as the flexibility of using any integrating capacitor as an integrator gain. The most versatile machines have. a choice of 6 capacitors, giving the programmer a five-decade range of integrator gains or time scales. Examples of this class of computer are the Applied Dynamics AD/4 (Figure 16), the Electronic Associates, Inc. 8800 (Figure 17) .and the Comcor Ci-5000 (Figure 18). Note the two patchboards in each, one for digital logic, and one for analog components. This· period also saw a more intimate ·tie-in of the analog computer with a digital computer due to the development of such true hybrid devices as the lVIDAC (multiplying D/A) and the "digital attenuator" or "digital potentiometer." So widely accepted has the hybrid aspect of analog computation become that it appears that close to half of the larger consoles that are being sold at the present time are going into hybrid systems. This in turn has led to the need, and the development of software specifically designed to aid the hybrid programmer and operator. The large systems have grown larger and larger and now are truly prodigious, consisting of 300, 400, even 500 amplifiers in a Figure 15-Applied Dynamics large scale 256 amplifier computer 650 Fall Joint Computer Conference, 1970 Figure 16-Applied Dynamics AD /4 analog computer single console. At the low end of the scale, the 10 volt desk-top computers have grown larger and larger until they are no longer desk-top and now are fully grown consoles consisting of several hundred amplifiers, as exemplified by the EAI 680 computer shown in Figure 19. The solid state revolution, which only overtook analog in the third generation has led to the concept of the class 0 type component or "blackbox" use of the analog components to help minimize patching and to make it easier for the more casual user of the machine to program, patch, and obtain solutions by himself. Another reason for this trend is that the solid state amplifiers are obviously less costly and more reliable than their vacuum tube predecessors. Analog speeds of solution which could be too fast to be absorbed by hUrhans, or recorded by devices, even back in the early 50's, are even faster. Present day bandwidth ranges from a minimum of 100 KHz to over 1 MHz. Some of the other important equipment improvements are quarter square multiplier accuracy of close to 0.01 percent and arbitrary function generation performed by a true hybrid device, the digitally controlled function generator (DCFG), which eliminates >spurious·· drifts, non-repeatability;, and difficulty in setup of the old diode function generator. These, together with the new digital potentiometer, a good hybrid interface with good software, and a well integrated system design, make it theoretically possible to setup and checkout an analog computer in a few seconcls. Some persons have been lmown to state the opinion that an analog computer of today is not much different than one of 10 years ago. A reading of this paper should dispel such a notion. To make clear the advances that have been made in the analog field, from post W orId War II to the present time, I have summarized in Figure 12 the major hardware improvements by year of general availability showing the typical computers incorporating the named improvements. It is obvious that these improvements have come at more frequent intervals than analog computer generations as I have defined them, and shows that major improvements have come along in the analog field at an average spacing of about 23-1 years. This interval of time is, interestingly enough, approximately equal to the half-life of a "generation" of analog computers. This fact might lead to the conclusion that one generation of computers cannot survive (or absorb) two sets of major hardware improvements, but that the manufacturers have been reasonably successful in extending the life of a generation of their computers through .at least one significant hardware evolution. Perhaps it is the ability to extend the life of a "generation" of analog computers, because of the nature of the organization of analog computers (parallel building blocks) which has led to the inaccurate observation that "analog computers of today are not much differen.t than they were 5 or 10 years ago." ANALOG/HYBRID TODAY We have now come to the point in analog/hybrid developments where not only do we have more raw computing speed than it is possible to take full advantage of, for solutions, but we also have more speed in terms of setup and checkout than we have customers Figure 17-680 lOV computer with display wing Analog/Hybrid who understand this type of computation. Or to put it another way, we've reached the stage in evolution where we can get a customer on, get his answers for him, and get him off, far faster than is justifiable based on the fact that we have a highly serial, slow input, mainly the input from a single man, to a very fast parallel console. We have almost reached the stage, as a matter of fact, where the slow recorders on the outputs from the analog are one of the limiting output factors. We've reached the point where we can make many, many solutions in a very short time. In other words, we are production oriented in terms of solution speed. At the same time, we have retained all of our man-machine interactive capabilities which everyone says is desirable in the engineering use of computers, but which obviously work against production. In fact, production capabilities are so great that I have estimated that for every hour of production running on our modern hybrid systems, the amount of post run data reduction of the results by a large fast, stand alone digital computer operating in a batch mode would be at least two and possibly as high as five hours depending on how much analysis is desired, or more realistically, how much the project can afford. The application of hybrid equipment is still heavily oriented toward the aerospace-defense industry where most of the large systems are installed. The chemical process industries have maintained some interest in these systems over the years, but not at an increasing rate. The education field has interest in the small and medium size systems. Nuclear and power technology have shown signs of increasing awareness of the ':lap ability of hybrid systems for their highly complex Figure 18-Comcor Ci-5000 analog computer 651 Figure 19-8800 100V transitorized computer with display wing design, control, and training studies. Other popular applications are as an on-line testing device, such as measuring the amount of pollutants in an automobile engine exhaust (Reference 6); measuring the roundness of tires (Reference 7) in acting as an on-line predictor or / ad;iptor-controller for a wide variety of processes (Reference 8), and for helping to control the quality of s~el (Reference 9). , So what is the hybrid/analog system of today? It is a highly efficient fast production device when the user or man is not allowed to intervene and interfere with its operation. This is in direct contradiction to its other main feature, that is, its ease of man-machine communication which almost cries out for man's intervention. I would say that the analog/hybrid computer exhibits schizophrenic characteristics which may explain why not too many people understand it. It is almost impossible for a device to be responsive to man's intervention and at the same time to be highly productive. At leaSt not the way the hybrid systems are configured today. It is this paradox that limits the expansion of· the analog/hybrid field. The analog hardware today is far more reliable than its early beginnings. The MTBF for a transistorized amplifier is somewhere between 30,000 hours and 60,000 hours. The high quality, chopperless amplifier, a recent devel