Analog Computers

Reference / Paper · 1948

Computing Mechanisms and Linkages

Read the PDF (371 pp) ↗

Volume 27 of the MIT Radiation Laboratory Series, authored by Antonin Svoboda and edited by Hubert M. James. Covers the mathematical design of mechanical computing mechanisms and linkage computers, including harmonic transformers, three-bar linkages, cams, integrators, resolvers, and bar-linkage combinations. Developed under wartime pressure for practical methods of designing linkage-based analog computers; originally published by McGraw-Hill in 1948.

Manufacturer
MIT Radiation Laboratory / McGraw-Hill
System
Mechanical analog computers (linkage computers)
Author
Antonin Svoboda
Year
1948
Type
Reference / Paper
Language
English
Learning track
general theory
Pages
371
  • Mechanical analog computers (linkage computers)
  • MIT Radiation Laboratory / McGraw-Hill
  • bar linkages
  • computing mechanisms
  • mechanical computation
  • function generation

← Back to the Reference Library

Computing Mechanisms and Linkages

COMPUTING AND MECHANISMS LINKAGES ii MASSACHUSETTS INSTITUTE RADIATION OF TECHNOLOGY LABORATORY SERIES Board of Editors LOUISN.RIDENOUR, Eddor-in-Ch@ GEOR~EB. COLLINS, Deput!/ Edztor-in-Chief BRITTON CHANCE, S. A. GOUDSMIT, R. G. HER~, HUBERT M. JAMWS,JULIAN K. KNIPP, JAMES L. LAWSON, LEON B. LI~FORD, CAROL G. MONTGOMERY, C. NEWTON, ALEERT M. STOISE, LouIs A. TURNER, GEORC~ E. VALLEY, JR., HERBERT H. WHEATON 1. RADAIi SYSTEM l;KCXNE~R1~~ --l~zdenou~ 2. R.4DAR .\IDS TO N.4v1GAmo-~a/L 3. R.ADAR BEAc’oNs—lioherls 4. LoRA~—Pzmce,.T1cKenzfe, and JVoodward 5. PULSE GEN ER.i’roRs<lasoe and Lebacqz 6. NIIcROwAVE 111.4 GNETR0N s—C011i71s 7. KLYSTRONS AND kIICROWAVE TRIol)Es—~amil/on, 8. PRIXCI?LES OF IVIICROWAVE CIRcrITs—,lIontgonzery, 9. h~IcROWAt-E TRANSMISSION Knipp, and Kriper Dir!ie, and Purcell CIRcrITs—h’agan 10. WAVEGUIDE HAY~Boo~-lfarcllL,itz 11. TECHXIQIE OF ~~lCROW.4VE h[EASGRE31EXTS—.~ fOntg0?lle,V 12. NIICROWAVE .\XTESNA THEORY .kND ~EslGN—Si&r 13. PRop.*G.k!rloxOF SHORT R.ADIO W’AVES-KW 14. LIICROW4VE nUPLEXERS—SnII{ Zlinand Montgomery 15. CRYST.AL Rectifiers—Torrey and ~~hif?ner hIIx13Rs-P0u71d 16. NIICROW-AVE 17. COMPONENTS HANDBOOK—~/Uc!ibum 18. ~’ACIWM TUBE .kMPLIFIERs-~a//eq and ~~a~{fna?z Hughes, McrcSichol. ,%yre, and ~f’illiams 19. ~~.kvEFoRMs-chunce, 20. ELECTRONIC TIME ~fEASLrREMENTS-ChQnce, HI(lsizer, flfclc.~icho~, and Williams Holdam, and ~fac~ae 21. ~LECTRONIC IxsTRcMExTs~reenwood, 22. CATHODE R.4Y TUBE ~ISPL.*YS—L’3071W,Starr, and Valley 23. ~rICROWATE REcEIVERS—~’Ufl l’oorhis 24. THRESHOLD SIGx.&Ls—Lawson and L:hlenbeck 25. THEORY OF SERvoMEcHAN1s.Ms-~anles, .~irhok, and Phillips 26. RADAR SC.4iSNERSAND R.iDOMES-CU(~Jy, Karellfz, and Turner 27. COMPUTING lfECHAXISMS AND LIWK.4GES—&ot)Oda 28. INDEx—Henney * CO MPIJTING MECHANISMS AND By LINKAGES AN Eciitedby OFFICE OF TONIN SVOBODA HIJB13RT SCIEA-TIFIC NATION.4L M. RESEARCH DEFENSE JAMES AND RESEARCH D13VELOPM.ENr COMMITTEE FIRST EDITION NEW McGRAW-HILL YORK AND BOOK 1948 * LONDON COMPANY, INC. .4 -=’ c“’ L. COMPUTING MECHANISMS AND LINKAGES COPYRIGHT, 1S48, EIY THE MCGRAW-HILL PRINTED 1>- THE Boo < COMPANY, l?N-ITEE ST LTES INC. OF .iMERI~A All rights reserved. This book, or parts thereof, ntay z,oi be reproduced in icn~j form dho,d permission of the pubttshers. THE M.4PLE PRESS (X)?I~PAh-Y, YORK, PA. Forezuord HE tremendous research and development effort that went into the development of radar and related techniques during World War II resulted not only in hundreds of radar sets for military (and some for possible peacetime) use but also in a great body of information and new techniques in the electronics and high-frequency fields. Because this basic material may be of great value to science and engineering, it seemed most important to publish it as soon as security permitted. The Radiation Laboratory of MIT, which operated under the supervision of the National Defense Research Committee, undertook the great The work described herein, however, is task of preparing these volumes. the collective result of work done at many laboratories, Army, Navy, university, and industrial, both in this country and in England, Canada, and other Dominions. The Radiation Laboratory, once its proposals were approved and iinances provided by the Office of Scientific Research and Development, chose Louis N. Ridenour as Editor-in-Chief to lead and direct the entire project. An editorial staff was then selected of those best qualified for this type of task. Finally the authors for the various volumes or chapters or sections were chosen from among those experts who were intimately familiar with the various fields, and who were able and willing to write the summaries of them. This entire staff agreed to remain at work at MIT for six months or more after the work of the Radiation Laboratory was complete. These volumes stand as a monument to this group. These volumes serve as a memorial to the unnamed hundreds and thousands of other scientists, engineers, and others who actually carried on the research, development, and engineering work the results of which are herein described. There were so many involved in this work and they worked so closely together even though often in widely separated laboratories that it is impossible to name or even to know those who contributed Only certain ones who wrote reports to a particular idea or development. or articles have even been mentioned. But to all those who contributed in any way to this great cooperative development enterprise, both in this country and in England, these volumes are dedicated. T m .$ #,. 4 L. A. DUBRIDGE. v ----- . ..—. ... . —- . .—----- Preface HE work on linkage computers described in this volume was carried out under the pressure of war. War gives little opportunity for the advancement of abstract knowledge; all efforts must be concentrated on meeting immediate needs. In developing techniques for the design of linkage computers, the author has therefore been forced to concentrate on tinding practical methods for the design of computers rather than on developing a unified and systematic analysis of the subject. The war has thus given to this work a special character that it might not otherwise have had. The impulse to the development of the methods presented in this volume for the mathematical design of linkage computers grew out of a collaboration of the author with his friend, Dr. Vladimir Vand. That collaboration was begun in France in 1940, and was brought to a premature end by the progress of the war. Though these ideas and methods have largely been developed by the author since that time, he wishes to emphasize that credit for the initiation of the work is shared by Dr. Vand. It must be mentioned also that the techniques described in this book were for the most part developed before the author became associated with the Radiation Laboratory. The author wishes to express sincere gratitude to Dr. H. M. James, the editor of this volume, who gave the book its present form, contributing many examples and many improvements to the methods. (Sees.: 6.7,6.8, 6.15, 8.6.) The book would never have been completed in such a short time without the assistance of Miss Constance D. Boyd, who read the manuscripts, and Miss Elizabeth J. Campbell, Mrs. Kathryn G. Fowler, Miss Virginia Driscoll, and Miss Patrica J. Boland, who calculated the tables and drew nomograms. The author also wishes to thank Dr. I. Maddaus, Jr., for bibliographical research. The publishers have agreed that ten years after the date on which each volume in this series is issued, the copyright thereon shall be relinquished, and the work shall become part of the public domain. T A. SVOBODA. ., PSARA,CZECHOSLOVAmA, June, 1946. :. vii ,. 1 Contents FOREWORD BYL. A. DUBBIDGE,.. v ,,. PREFACE. vii &AP. 1. COMPUTING MECHANISMS AND LINKAGES. 1 1 INTRODUCTION 1.1. Types of Computing Mechanisms . 1.2. Surveyof the Problem of ComputerDesign 1.3. Orgrmizationof the PresentVolume I 2 5 ELEMENTARY COMPUTING MECHANISMS 1.4. 1.5. 1.6. 1.7. I.S. 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...6 . . . . . . . . . . . . . . . 12 . . . . 15 . . . . . . . . . . . . ...19 ,.23 CHAP.2. BAR-LINKAGE COMPUTERS 27 2.1. 2.2. 2.3. 2.4. 2.5. 2.6. 2.7. CHAP. 3. Additive Cells. . . . Multipliers . . . . . Resolvers, . . . . . Cams. . . . . . . . Integrators . . . . . Introduction. . . . . . . . . . . . . . . . . . . . . ...27 HietoriealNotes. . . . . . . . . . . . . . . . . . ...28 The Problemof Bar-linkage-computerDesign Characteristicsof Bar-liikage Computers. Bar Linkageswith One Degreeof Freedom Bar Linkageswith Two Degreesof Freedom. ComplexBar-linkageComputers. BASIC CONCEPTS AND TERMINOLOGY. 3.1. 3.2. 3.3. 3.4. 3.5. De6nitions . . . . . . . . . . . HomogeneousParametersand Variables An OperatorFormalism. GraphicalRepresentationof Operators The Squareand Square-rootOperators 43 . . . CHAP.4. HARMONIC TRANSFORMER LINKAGES. THE HAB~ONXC TRANSFORMED. . . . . . . 3I 32 34 37 40 43 47 49 51 54 58 . . 4-1. Definitionand Geometryof the Harmonic Transformer. 4.2. Mechanizationof a Function by a Harmonic Transformer. ix 58 58 61 x CONTENTS 4.3. 4.4. 45. 4.6. 4.7. 48. HARMONIC The Ideal Harmonic Transformerin HomogeneousParameters, Tables of Harmonic TransformerFunctions Total StructuralError of a Nonideal HarmonicTransformer Calculation of the StructuralError Function ~H~of a Nonideal Harmonic Transformer. A Study of the StructuralError Function 6H~. A Methodfor the Designof NonidealHarmonicTransformers. 62 63 67 6S 71 75 TRANSFORMERS 77 IN SERIES. 49. Two Ideal Harmonic Transformersin Series. . 410. Mechanizationof a GivenFunctionby an Ideal Double Harmonic Transformer. . . . . . . . . . . . . . . . 4.11. PreliminaryFit to a Monotonic Function. 412. PreliminaryFit to a NonmonotonicFunction 4.13. Improvementof the Fit by a Method of Successive Approxirna tions. 4.14. 4.15. . . . 79 82 89 !)1 . . m Nonideal Double Harmonic Transformers. Alternative Method for Double-harmonic-transformer CHAP.5. THE THREE-BAR LINKAGE THE 7? . . Design . ][)1 . . 107 5.1. FundamentalEquationsfor the Three-barLinkage. 5.2. Classificationof Three-barLinkages 5.3. SingularCasesof Three-bar Linkages, 1 ~2 5.4. 117 The Problem NOMOGRAPHIC 55. 5.6. 5.7. 5.8. of Designing Three-bar . kfETHOD . . . . . . . . . . . . . . . . . . . . . 118 Analytic Basis of the homographic The Nomographic Chart . . . . Calculation of the Function Linkage .,.,.......,,, Complete Linkages. Representation Nomogram, ,., . Method. . . . . . . . Generated of Three-bar-linkage . . . . . . . . . . of the Design Problem for . . . . . . . . . . . . . . . Functions . . Restatement Method, 5.10. 5,11. Survey of the Nomographic Method . Adjustment of ba and a, for Fixed AX,, AX,, b, 512. Alternative Methods for Overlay Construction. Choice of Best Value of b, for Given AX,, AX,. An Example of the Nomographic 118 . . . ...120 by a Given Three-bar . . . . . . . . ...122 5.9 513. 5.14. 107 108 . . . by the . . ..1’25 the Nomographic . . . . . . ...127 128 132 . . . . . . Method. THE GEOMETRIC METHonFORTHR~E-BLR LINKAGEDESIGN. . . 136 137 . 139 . . 145 5.15. Statementof the Problemfor the Geometric Method. . 146 5.16. Solutionof a SimplifiedProblem. . . . . . 147 5.17. Solution of the Basic Problem. . 151 5.18. Improvementof the Solutionby SuccessiveApproximations. 154 5.19. An Application of the Geometric Method: Mechanization of the Imgarithmic Function. . . . . . . . . . . . . 156 k xi CONTENTS CHAP. 6. LINKAGE COMBINATIONS DOM . . . . . . . . . . . . . . OF Two COMBINATION AGE . . . . . . WITH ONE . DEGREE OF FRf3E- . 166 HARMONIC TRANSFORMERS WITH A THREE-BAR I.lNK- . . . . . . of the Problem. ., 6.1. Statement 6.2. Factorization 6.3. 6.4. Example: Example: Factoring the Given Function Design of the Three-bar-linkage 166 .,,., . of the Given 166 Function . Harmonic 168 171 174 Component 6.5. Redesign of the Terminal 6.6. Example: Redesign of the Terminal Harmonic Transformers 6.7. Example: Assembly of the Linkage Combination, 186 187 Transformers 193 THREE-BAR LINKAGES IN SERIES. 6.8. CHAP. 7. The Double FINAL 7.1. 8. Roles of Graphical 9. 195 OF LINKAGE and Numerical arameters . . . . . 73. Use of the Gauging Parameter 7.4. 7.5. Small Variations Large Variations . CONSTANTS Methods in Linkage . . . . . . . . Linkage Constants. . . in Adjusting of Dimensional of Dimensional 199 Design. . . 199 ..2o2 201 Constants Constants 205 205 7.6. Method 7.7. Application of the Gauging-parameter Linkage . . . . . . . . . . . . . Method to the Three-bar . . . . . . . . ...207 7.8. Application Method 7.9. Linkage. An Example,...,.. The Eccentric Linkage as a Corrective of Least Squares LINKAGES 8.1. 82. 8.3. Cu. Linkage ADJUSTMENT 72.GaugingP CHAP. Three-bar 195 206 of the Gauging-parameter WITH TWO 217 OF FREEDOM 223 Function. 223 226 228 of Grid Structures. 232 DEGREES Analysis of the Design Problem Possible Grid Generators for a Given The Concept of Grid Structure 8,4. Topological 8.5. The Significance Transformation 8.6. Choice 8.7. Uw of Grid Structures of Ideal Grid Structure of a Nonideal BARLINKAGE to the Three-bar . .,........209 Device. Grid Generator. in Linkage MULTIPLIERS 233 . Design. . . . . . 238 . . 243 . .250 9.1. The Star Grid Generator,, 9.2. A Method 9.3. 9.4. Grid Generators for Multiplication. A Topological Transformation of the Grid Structure 9.5. Improvement 9.6. 9.7. Design of Transformer Linkages. Analytic Adjustment of Linkage Multiplier Constants 271 277 9.8. Alternative the Error of a Grid Generator 281 for the Design .250 of Star Grid Generators Ideal Grid Structure.....,.. . . of the Star Grid Generator Method for Gauging with Almost . . . . . . ...251 of a Divider. for Multiplication. 256 258 264 I ] I CONTENTS xii 10. BAR-LINKAGE FUNCTION GENERATORS WITH TWO DEGREES OF FREEDOM . . . . . . . . . . . . . . . . . . ...284 ~H,4P. 1 , I Summaryof the Design Procedure. 10.1. 10.2. Example: 10.3. Example: 10.4. in Vacuum Curve Tracing Function i I 4 First Approximate in Vacuum Improving . . . . . . . . . . . the Mechanization .,,..... and Transformer . . . . of the Ballistic . . . . . . Mechanization of the Ballistic . Linkages . 284 ...286 Function for Noncircular Scales .292 295 1 > APPENDIX A. TABLESOF HARMONIC TRANSFORMER FUNCTIONS APPENDIX B. . . . 301 PROPERTIES OFTHETHREE-BAR-LINKAGE NOMOGRAM INDEX . . . . . . . . . . . . . . . . . . . . . . . . 333 . 353 1 CHAPTER 1 INTRODUCTION 1.1. Types of Computing Mechanisms.—Computing mechanisms may be divided into two distinct types: arithmetical computing machines, familiar to the layman through their common use in business offices, and continuously acting computing mechanisms and linkages that range in complexity from simple cams and levers to enormously complex devices for the direction of naval and antiaircraft gunfire. The arithmetical computing machines accept inputs in numerical form, usually on a keyboard, and with these numbers perform the simple and arithmetical operations of addition, subtraction, multiplication, division-usually by the iteration of addition and subtraction in counting devices. The results are finally presented to the operator, again in numerical form. In their simplest forms these machines have the virtue of applicability in a wide variety of computations, including those requiring very high accuracy. By elaboration of these devices, as by the introduction of punched-tape control, their possibilities for automatic operation can be greatly increased. Characteristic of their operation, however, is their production of numerical results by calculations in discrete steps, involving delays which are always appreciable and may be very large if the required calculation is of complex form. Continuously acting computing mechanisms are less flexible and have less potential accuracy, but their applicability to the instantaneous or to the continuous solution of specific problems-even quite complex ones— makes them of great practical importance. They may serve as mere indicators of the solutions of a problem, and require further action by human agency for the completion of their function (speedometer, slide rule); or they may themselves produce a mechanical action functionally related to other mechanical actions (mechanical governors, automatic gunsight). Continuously acting computers fall into two main classes: function generators and differential-equation solvers. Function generators produce mechanical actions—usually displacements or shaft rotations—that are detite functions of many independent variables, themselves introduced into the mechanism as mechanical actions. Simple examples of such mechanisms are gear differentials, two- and three-dimensional cams, slide multipliers and dividers, linkage computers, and mechanized nomograms. Computers of the second class generate solutions of some definite 1 2 INTRODUCTION [SEC.1.2 differential or integrodifferential equation—often an equation that involves functions continuously determined by variable external cirElementary devices of this type are the integrators, comcumstances. ponent solvers, speedometers, andplanimeters. From these elementary devices one can build up complicated mechanisms that perform elaborate calculations. We may mention their application in gunsights, bombsights, automatic pilots (for airplanes, submarines, ships, and torpedoes), compensators for gyroscopic compasses, tide predictors, and other robots of varied types. The present volume will deal only with the problem of designing continuously acting computing mechanisms. 1.2. Survey of the Problem of Computer Design.-There is no set rule or law for the guidance of a designer of complex mechanical computers. He must weigh against each other many diverse factors in the problem: the accuracy required; the cost, weight, volume, and shape of the computer; its inertia and delay in action; the forces required to operate it; its resistance to shock, wear, and changes in weather conditions. He must consider how long it will take to design the computer, how easily it can be built, how easily it can be operated by a crew, whether suitable sources of power will be available, and so on. The complexity of the theoretical and practical problems is so great that two designers working on a given problem will never arrive at precisely the same solution. For practical reasons, a designer should be asked to find a computer that meets certain specified tolerances, rather than the best possible computer for a given use. He should know what will be the maximum tolerated error of the computer, the maximum cost, weight, and volume occupied, the maximum number of operators in the crew, the maximum number of servomechanisms allowed, and so on. Tolerances provide a convenient means for controlling the development of the computer, and— if established in a practical way—they permit some freedom of choice by the designer. Choice of Approach to the Design Problem.—The type of computer to be built is sometimes indicated in the specifications. If not, the first task of the designer is to decide whether the computer is to be mechanical, electrical, optical, or a combination of these. At the same time that this important decision is made, the designer must weigh in his mind the path that his thinking will follow. There are two principal methods for designing a computer: the constructive method and the analytic. The constructive method makes use of a small-scale model of the real system with which the computer is to deal. For example, a constructive antiaircraft fire-control computer might determine the elements of the lead triangle by maintaining within itself and measuring the elements of a small model of this triangle. SEC.1.2] SURVEY OF THE PROBLEM OF COMPUTER DESIGN 3 In using the analytic method, the designer concentrates on the analytic A relation between variables, relations between the variables involved. such as (1) can be given mechanical expression in terms of displacements or shaft rotations, without regard to the nature of the quantity represented by the variables x, y, and z. For example, one may possess two devices that generate output displacements xy and x/y, respectively, given input displacements z and y. Combining these with a third device for adding their output displacements, one can then produce a computer that, given input displacements x and y, generates a final output displacement z having continuously the value specified by Eq. (1). The computer is then a “mechanization” of Eq. (1), rather than a model of any special system involving variables x, y, and z thus related. Computers designed by analytic methods consist of units (“cells”) that mechanize fairly simple relations, so connected as to provide a mechanization of a more complex equation or system of equations. For any given problem a great variety of designs is possible. This variety arises in part from the possible choice among mechanical cells mechanizing a given elementary relation, and in part from the variety of ways in which the relation between a given set of variables can be given analytic expression. Thus, each of the equations Z=; (yj+ l), (2a) 2=X () (2b) y+~, Y Z’7J= Z(yz + 1), (2C) [all equivalent to Eq. (l)] suggests a different method of connecting This flexibility in analytic mechanical cells into a complete computer. design methods makes it possible to arrive at designs that are in general more satisfactory mechanically than those obtained by constructive methods. In the present volume we shall be concerned entirely with mechanical computers designed by the analytic method. Block Diagram of the CompuLer.—To each formulation of the problem in analytic terms there corresponds a block diagram of the computer. in thk diagram each analytic relation between variables is represented by a square or similar symbol, from which emerge lines representing the variables involved; a line representing a variable common to two relations will connect the corresponding squares in the diagram. In mechanical terms, each square then represents an elementary computer that estab- 4 INTRODUCTION [SEC. 1.2 lishes a specified relation between the variables, and the connecting lines represent the necessary connections between these elementary computers. By examination of block diagrams the designer will be able to see the principal virtues of each computing scheme: the complexity of the system, the workhg range of variables, the accuracy required of individual components, and so on. On this basis he can make at least a tentative selection of the block diagram to be used. Selection oj Components jor the Computer.-Knowing the accuracy and mechanical properties required of each computing element, the designer can select the elementary computers from which the complete device is to be built. As an example of the diverse factors to be borne in mind, let us suppose that it is required to provide a mechanical motion proportional to the product of two variables, X1 and X2. A slide multiplier of average size will allow an error of from 0.1 per cent to 0.5 per cent of the whole range of the variable; this error will depend on the quality of the construction A linkage multiplier -on the backlash and the elasticity of the system. will have an error of some 0.3 per cent due to its structure, practically no error from backlash, and a slight error due to elasticity of the system if the unit is well designed; the space required by a linkage multiplier is small, but its error cannot be reduced by increasing its size. If these devices do not promise sufficient accuracy, the designer must use multipliers based on other principles. It is possible to perform multiplication by use of two of the precision squaring devices illustrated in Fig. 1.23, by connecting these in the way suggested by the equation X1X2 = +(X1 + x2)* – *(X1 – X2)2. (3) The error of such a multiplier may be as low as 0.01 per cent, but the About the same accuracy is attainsystem has an appreciable inertia. able by a multiplier based on the differential formula for multiplication, d(xlxz) = xl dxz + x2 dxl; (4) this employs two integrators, and is commonly used when two quantities This scheme is useful only are to be multiplied in a differential analyzer. when it is possible to allow a slow change in a constant added to the product XIX~a change which will result from slippage in the integrators, negligible for a single multiplication but accumulating with repetition of the operation. From this discussion it should be evident that there is no “best” multiplier. Similarly, other components of a computer must be selected with due regard for their special characteristics and the demands to be made upon them. Mathematical Design oj the fi@em.—From the block diagram one should proceed to the mechanical design of a system through an inter- SEC.1.3] ORGANIZATION OF THE PRESE>T VOLUME 5 mediate step-that of establishing the “mathematical design” of the system. The mathematical design ignores the dimensions not essential to the nature of the computation to be carried out-diameters of shafts, dimensions of ball bearings, dimensions of the fram~but specifies the dimensions of levers measured between pivots and joints, the size of friction wheels, tentative gear diameters and gear ratios. The properties of this design should be studied carefully, because this usually leads to a change in some detail of the design, and sometimes even to choice of a new block diagram. Final Steps in the Design.-From the mathematical design of the The elements system one can proceed to the design of a working model. of this model should be accessible rather than massed together, inexpensive, and quick to manufacture. If the performance of the working model is found to be satisfactog, the first model can be designed. Here the The parts of ingenuity of the designer must be used to the maximum. the mechanism must be arranged compactly to decrease space requirements, weight, and the effects of elasticity and thermal expansion, but they should not be massed in such a way that assembly is difficult, or Sometimes division of the whole computer repair or servicing impossible. into several independent parts is advisable. Finally, the computer can be built and tested against specifications. 1.3. Organization of the Present Volume.—It is not possible to discuss in one volume all elements of the problem of computer design. This book will deal principally with bar-linkage computers—specifically, with the mathematical design of elements for such computers. Bar linkages are mechanically very satisfactory, and computers built from them have many important virtues, but the mathematical design of these systems is relatively difficult and is not widely understood. There are few standard bar-linkage elements for computers; it is usually necessary to design the components of the computer, and not merely to organize standard It is hoped that the design methods elements into a complex assembly. to be described here will lead to their more general use. Bar linkages can be used in combination with the standard computing mechanisms. For this reason, and for the contrast with the bar linkages which are to be discussed later, this volume begins with a brief survey of some more or less standard elements of mechanical computers. Chapter 2 is devoted to a general discussion of bar linkages. Chapter 3 establishes terminology and describes graphical procedures of which extensive use will be made. Chapters 4, 5, and 6 discuss, in order of their increasing complexity, bar linkages with one degree of freedom—generators of functions of one independent variable, Chapter 7 indicates some mathematical methods of importance in bar-linkage design. Finally, Chaps. 8, 9, and 10 develop methods for the design of bar-linkage gener- [SEC.14 INTRODUCTION 6 ators of functions of two independent linkages have very striking advantages. ELEMENTARY variables—a COMPUTING field in which bar MECHANISMS The remainder of this chapter will gi~’e a brief survey of elementary Discomputing mechanisms, or “cells,” of more or less standard type. cussion of bar-linkage cells will be deferred to Chap. 2. 1.4. Additive Cells.-” Additive” or “linear” cells establish linear relations between mechanical motions of the cell, usually shaft rotations or slide displacements. If these are described by parameters X1, X2, X3, the cell will compute X3= Q. X,+ Q’. X,+C. (5) Here Q, Q, and C are constants depending on the design of the cell and the choice of the zero positions from which Xl, XZ, and X3 are measured. By FIG. 1 1.—Bevel-gear clifferential. proper choice of the zero positions, C can always be made to vanish; in what follows it will be assumed that this has been done. The bevel-gear deferential (Fig. 1.1) is a well-known linear CCI1f’[Jr The parameter Xl is the rot:~which all three parameters are rotations. tion of the shaft S1 from a predetermined zero position, X1 = O; the positive direction of rotation is indicated by symbols representing the hewl and tail of an arrow with this direction. The parameter X2 is the rotation of the shaft SJ from a similar zero position; Xj is the rotation from its zero position of the cage C carrying the planetary bevel gears G. The zero positions are not indicated in the figure. The equation of the bevel-gear differential is X, = 0.5X, + 0.5XZ. (6) To derive this it is convenient to consider the value of X2 corresponding to given values of X, and Xs. Let us consider the differential to be originally in the position X1 = X2 = X3 = O. The parameters Xl and X3 can then be given their assigned values in two steps, the first a rotation of both the SEC,14] ADDITIVE CELLS 7 shaft S1 and the cage C through the angle X3, and the second a rotation of the shaft S’1through an additional angle Xl – Xi. In the first step the differential moves as a unit; the shaft Sz is rotated through the angle X3. In the second step, the cage is stationary and the movement of the shaft S1 is transmitted to the shaft SZ with its sense of rotation reversed; the rotation through angle Xl — X3 of the shaft SI causes rotation through Xt – Xl of the shaft SZ. The total rotation of the shaft SZ is then It is, X2 = Xt + (Xt – X,), from which Eq. (2) follows immediately. of course, essential that all rotations be taken as positive in the same sense. It is remarkable that Eq. (6) is independent of the ratio of the bevel gearing of the clifferential; the essential characteristic of this type of X3 FIG.1.2.—Cylindrica1-gear clifferential. differential is that the gearing of the cage transmits the relative motion of the shaft S1 to the shaft SZ in the ratio 1 to 1, but with reversed sense. It is not necessary to use bevel gears in the cage to obtain this result; cylindrical gears can accomplish the same purpose. A cylindrical-gear differential is shown in Fig. 1.2. This differential is equivalent to the It is common bevel-gear differential, except in its mechanical features. flatter, and easier to construct in large numbers, but there is one more gear mesh than in the common type; there may be more backlash and more friction. It should be noted, however, that bevel gears are subject to axial as well as radial forces in their bearings, and that these may also increase friction. The spur-gear differential shown in Fig. 1.3 has only two gear meshes, and is quite flat. The planetary gears G in their cage C do not invert the motion of the shaft SI when transmitting it to the shaft S3, but can be made to transmit it at a ratio different from 1. The eauation of this 8 INTRODUCTION [SEC.1.4 differential is X, = QX, + (1 – Q) X,. To prove this relation we can use the same method as before. begin by considering the differential in the zero position, (7) Let us X1=X,=X3=0. We wish to find the value of X3 corresponding to given X, and X2. We introduce the angles Xl and X2 in two steps, first turning both the shaft FIG. 1.3.—Spur-gear clifferential. FIG.1.4.—Differential with axially displaced spiral gear. S, and the cage C through the angle X2, and then the shaft S1 through an In the first step the differential is turned as a rigid additional Xl - X,. ln the second body; the shaft S, is also turned through the angle X2. step the shaft S’s is turned through Q(xl — X2); its total motion is X, = X, + Q(X, – X2), in agreement with Eq. (7). If we make Q = Q’ = 0.5 by proper choice of the gear ratios, we can The fact obtain a differential equivalent to the” bevel-gear differential. that the free choice of Q gives to this differential a larger field of applicability does not necessarily mean that this differential should be preferred SEC.1.4] ADDITIVE CELLS 9 tothosewith Q = 0.5; it inconvenient tousedifferentials with Q =0.5 as prefabricated standard elements. A deferential m“th axially displaced spiral gear is shown in Fig. 1.4. The parameter,, which measures the axial displacement of the spiral gear andthepin PZ, invariable only ~\-ithinfinite limits. The mechanical structure of this differential is, however, much simpler than that of the differentials already mentioned, for which all parameters can change without limitation. The equation of this differential is X3=X1*9X2, (8) where nis the number of threads perinch along the axis of the spiral gear on the shaft Sz and m is the number of teeth on the gear with whichit FIG.1.5.—Differential worm gearing. meshes. Thehelical mgleofthe gears should beat least 450 for smooth action and small backlash. Thediferential worm gearing shown in Fig. 1.5 is used for the same purpose asthe preceding differential, especially if the range of values of X, corresponds to a large fraction of a revolution of the shaft S, or even to several revolutions of this shaft. The equation of this d”%-entialis x,= ++ X1+ +X2 (radians) (9) where t is the number of teeth of the worm gear, m is the multiplicity of the threads of the worm, and R is the radius of the worm gear. The sign in Eqs. (8) and (9) depends on the sense of the threads of the spiral or worm gear. The screw differential shown in Fig. 1.6 combines an axial translation Xl of a screw with a translation X2 of the nut N with respect to the screw; x3 = xl + x,. (lo) 10 [SEC.1.4 INTRODUCTION To obtain the first translation, the pin P on which the screw turns is displaced by X1. Therotation of thescrew comes fromthegear G,which meshes with a cylindrical rack C and slides along it. The real input IWILI x, i I * ‘“ ‘c-k’ -+E=Eaii FIG.1.6.—Screwdifferential. parameter of the differential is not X,, but the angle X, through which the rack is turned. The equation of the differential is then X3 = x, * kx4. (11) The sign depends on the sense of the screw; k is a constant determined by the gear ratio, the number of threads per inch on the screw, and their multipli~ity. All three parameters of this differential have constructive limits. The belt deferential (Fig. 1.7) makes use of the inextensibility of a belting on several pulleys. In practice, chains, strings, and special cables are used as belts. The equation of the belt differential is X3 = c – 0.5XI – 0.5X2, (12) where C is a constant depending on the choice of zero points of the parameters. .++ The tension in the belt must not fall below zero at any time; if it does, the belt FIG.1.7.—Belt differential. will sag and the equation of the differential will not hold. To obtain positive action in the direction of increasing X3, it is necessary to preload the belt by putting a load on the output pulley— ior instance, by a spring that can exert a force large enough to produce the desired action. The maximum driving force required for thk dMerential will then be about twice the force necessary to operate it without preloadlng. The loop-belt di~erential (Fig. 1.8) has the belting in the form of a loop with length independent of the position of the pulleys, The belt can then SEC-1.4] MULTIPLIERS 11 be preloaded (turnbuckle B) without adding to the driving force of the differential, except by the increased friction in the bearings. Belt differentials are sometimes used to add a large number of parameters; they are easily combined in batteries, as indicated schematically in Fig. 1.9. In such an arrangement the parameter XT may have so large a range that it is impractical to use a slide as the It is better output terminal. FIG.1S.-Loop-belt clifferential. practice to use a drum (dashed line in Fig. 1.9) on which the belt is wound on. and at the same time wound off. To prevent slippage, the belt should make many turns on the drum and be fastened to it; a chain on chain sprockets may also be used as the belt. The above enumeration does not exhaust the possibilities for linear mechanical cells; there are many variants the use of which may be dictated by special circumstances. As a rule, when a differential is used in a computing mechanism, two of its members (the input terminals) are moved by external forces; this results in movement of a third member (the output terminal) which is in turn required to furnish an appreciable force. If differentials were frictionless, any two of their three terminals could be used as input terminals. In reality, only a few of the differentials described here have FIG. 1.9.—Loop-belt differential for complete interchangeability of the the evaluation of terminals. For instance, with the X1= C–X1–2XZ+2X, –2X, + 2X, – 2X6. screw clifferential [Fig. 1.6) it is impossible to have Xd as the output parameter if the helical angle of the screw is so low that self-locking of the nut on the screw occurs; it is possible to use Xl as an output parameter, and, of course, also X,. With the chfferential worm gearing of Fig. 1“5, X, is an impracticable output parameter. 12 INTRODUCTION 106. Multipliers.-Multipliers three parameters a relation are computers that establish between RX3 = X, .X,, ( [SEC.1.5 (13) where R is a constant that depends on the type of multiplier and on its dimensions. The action of the slide multiplier shown in Fig. 1“10 is based on the proportionality of the sides of two similar triangles. These are triangles with horizontal bases, and vertices at the central pin shown in the figure: Iv;,:—’ .- FIG.1.10.—Slidemultiplier. the first has a base of length R and altitude Xl, the second a base of length X2 and altitude X8. Thus R x, T, = X’ (14a) RX3 = X,X2. (14b) or The figure gives a schematic rather than a practical design; the lengths of the sliding surfaces as shown are not great enough to prevent self-locking These lengths determine the in all possible positions of the mechanism. space requirements for multipliers of this type; they must be relatively large in two directions. It is difficult to make this type of multiplier precise. The pins in slots, as shown in the figure, are mechanically inadequate, and roller slides on rails must be used. One can not achieve the same end by increasing the dimensions of the multiplier because the SEC.1.5] MULTIPLIERS 13 elasticity of parts comes into play, not only when the parts are operating in a computer, but also when they are being machined. The slide multiplier shown in Fig. 1.11 saves space in one direction. There are fewer sliding contacts, and the slides are easier to construct. Fm. 1.11.—SlidemultiplierwithinputsXI, XI - Xz. . I?m. 1.12.—Intmwction nomogram for multiplication z; = z j . z*. This device cannot multiply Xl and X2 directly to compute RX3 = X1X2; the input terminals must be given translations of Xl and X, – X2. The difference is easy to obtain if the parameters are generated as shaft revolutions before entering the multiplier; screws can then be used instead 14 INTRODUCTION [SEC.1‘5 of the slides shown in the figure, and the required difference can be formed by a gear differential. Nomographic Multipliers.—A multiplier that is structurally related to a nomogram for multiplication will be called a “nomographic multiplier. ” Such multipliers can be derived from intersection or alignment nomograms; the examples to be given here are related to intersection nomograms. FIG.1.13.—An intersection nomogram for multiplication, obtained from the nomogram in Fig. 1.12 by a projective transformation. Figure 1.12 shows an intersection nomogram for multiplication in an unusual form, the full significance of which will be made clear in the latter part of this book. Thk represents the formula Xi = XiX