KINEMATIC SYNTHESIS

OF LINKAGES
McGraw-Hill Serles in MECHANICAL ENGINEERING
ROBERT M. DRAKE, JR. & STEPHEN J. KLINE
Consulting Editors

BEGGS, /11echanisnt
CAMBEL AND JENNINGS, Gas Dynamics
CSANADY, Theory of Tnrbomachines
DURELLI, PHILLIPS, AND TSAO, Introduction to the Theoret·ical
and Experimental Analysis of Stress and Strain
ECKERT, Introduction to Heat and JIass Transfer
ECKERT ANO DRAKE, Heat and ,ifass Transfer
GROBER, ERK, AND GRIGULL, Fundamentals of Heat Transfer
HAM, CRANE, AND ROGERS, i'v/ echanics of Al achinery
HARTENBERG AND DENAVIT, Kine111al.ic S11nlhesis of /.,inJ.·ages
HARTMAN, Dyna III ic::; of illachin<'r11
HINZE, Tiirbulence
JACOBSEN AND AYRE, Eng'ineerin{/ l'il>rations
PHELAN, Fundamentals of illechanical Design
RAVEN, Automatic Control Engineering
SABERSKY, Elernents of Engineering Thennodynanzics
SCHENCK, Theories of Engineering Experimentation
SCHLICHTING, Boundary Layer Theory
SHIGLEY, Dynamic Analysis of ,ifachine::,
SHIGLEY, Kinematic .,tnalysis oj 1\fPchanist11,;
SHIGLEY, j},fechanical Eng1:neeriny Dl!si1111
SHIGLEY, Theory of 111 achines
SPALDING AND COLE, Engineering Thermodynamics
STOECKER, Refrigeration and Air Conditioning
WILCOCK AND BOOSER, Rearing Design and Application
KINEMATIC SYNTHESIS

OF LINKAGES

RICHARD S. HARTENBERG
•••
Member of ASME, VOi, 1.MECH.E.

JACQUES DEN AVIT

Member of ASME, VOi

Depa.rtment of .lfechanical Engineering

and Astronautical Sciences
/1:orthu·estern Uni1)ersity

McGraw-HIii Book Company

New York San Francisco Toronto London
Kinematic Synthesis of Linkages

Copyright@ 1964 hy i\lc(:raw-Hill, Inc. All Rights Reserved.

Printed in the l!nited States of Arnerica. This book, or pnrts thereof, ,nay not be
reproduced in any form without perrnission of th<' publishers.
Library of Congre.�s Catalog ('arr! .Vumber 64-:23:!5 l

26910
 PREFACE

In the early 19-10s, the idea of kinernatic synthesis of linkages

"'as, in a 1nanner of speaking, n1entioned for the first ti1ne in this
country. .-.\fter a decade of gestation, original articles on the
subject began to appear in A1nerica. �o,v, with the passing of
the second decade, it is appropriate to atternpt a consolidated
story of so1ne of the n1any efforts of European and .An1erican
kine1naticians.

Conventional texts give an adequate presentation of the area of

planar kine1natic analysis. In these books, first contact is estab­
lished "·ith funda1nental concepts, and kno,vn planar 1nechanisn1s
are investigated for the displace1nents, velocities, and accelera­
tions of their parts. In analysis, n1echanisn1s are put on the
psychiatrist's couch, so to speak, and are probed for their per­
fonnance traits. 1'he i1nportant and con1ple1nentary area, that
called kinen1atic synthesis-,vhere 1nechanis1ns are created to
n1eet certain 1notion specifications-is touched upon only by a
consideration of ca1ns and possibly a gli1npse of the si111pler
aspects of planar linkage synthesis.

In the design of a n1echanisn1 for a given application, a decision

1nust first be reached regarding the type of 1nechanis1n to be
ernployed, as, for exarnple, deciding bet,veen a can1 or a linkage.
rfhe number of links and connections required to give the desired
degree of freedo1n 1nust then be deter1nined. Finally, the
required d1:•mensions needed to bring about a particular n1otion
n1ust be deduced. In the broadest sense, kinernatic synthesis
thus consists of the three interrelated areas of type, nurnber,
and dini.ensional synthesis.

---- .
VI PREFACE

Of these areas, type synthesis is based largely on nonkine111atic

considerations. It depends, of course, upon the designer's kinen1atic
experience, but, beyond that, other factors, such as 1nachine-design
proble1ns and econo1nic considerations, n1ust be taken into account.
\:Vhile recognizing the proble1n of type synthesis, "·e n1ake no effort to
cope ,vith it and go directly to linkages, ,vith the avoidance of earns and
gears. 1'he present text focuses on the di1nensional synthesis of linkages,
an area a1nenable to 111athe1natical trcat1nents. Xu1nber synthesis, not
a very real proble111 for the si1nple linkages that can be devised by syn­
thetic procedures, is.given a brief discussion.
This text is devoted to the synthesis of planar and spatial linkages
and considers both basic approaches, the geo1netric and algebraic, allo,v­
ing the reader to choose ,vhichever 1nethod best suits his needs or avail­
able tools. 1'he geornetric 1nethods are quick and have the advantage
of staying close to the physical proble1n. I'hey 1nay have adequate
accuracy for 111any tasks but tend to beco111e tedious under repeated use.
:\.lgebraic 1nethods give any desired degree of accuracy and through their
functional relations indicate the relative i1nportance of any paran1eter.
1'hey also 111ake possible the study of errors in di1nensions, as those
associated \\·ith tolerance or ,vear. A.nd of course the algebraic n1ethods
lead to the use of digital con1puters, especially convenient if the sa1ne
calculations have to be repeated often for incre1nental inputs or para1neter
changes.
Recent years have seen the publication of a ,vealth of ne,,· 1naterial
relating to synthesis, both in this country and abroad. In addition,
there has been a gro,ving a\\·areness of the very considerable European
effort, especially the Gerrnan. This 1naterial is difficult to follo\\· because
of the scattered sources, terseness of presentation, and language barriers.
1'he present text is an atten1pt to present a \\·ell-balanced and intelligible
account of son1e of the geo1netric and algebraic procedures, filling in as
necessary, n1aking con1parisons, and elaborating on the i111plications to
give a ,velI-rounded picture.
'rhe bulk of the available European 1naterial see1ns to divide into
t,vo broad classes. One n1akes use of rather sophisticated gcornctry and
1nathe1natics, \\·hile, in the other, descriptions of techniques predo1ninate
over the scrutiny of principles. This book atte1npts a 1niddle-of-the­
road presentation having "rigor" \Yithout "1nortis," integrating the
European and A.1nerican contributions.
1'he book studies, co1npares, and applies various procedures for
handling the several types of proble1ns constituting the synthesis of link­
ages. The start is 1nade hy reviewing or establishing definitions, con­
cepts, and notations, ,vith recapitulation of fa1niliar n1ethods of velocity
PREFACE ..
VII

and acceleration analysis. A.n account is given of type, nun1ber, and

di1nensional synthesis. 1'his includes a discussion of the Chebyshev
polynon1ials and their application to the determination of an optin1al
spacing of accuracy points. The properties and uses of four-bar coupler
curves are discussed, including their equation, the i1nplications of their
geo1netry (cusps, sy1111netry, and crunodes), and the Roberts-Chebyshev
theore1n, ,vith its extensions to the slider-crank 1nechanis1n and six-bar
linkages. The Euler-Savary equation and the cubic of stationary
curvature are derived analytically, by using co1nplex nun1bers, the
a1nbiguities and passage to the lirnit of the direct geo1netric approach
being thus avoided. T,Yo chapters are devoted to the geon1etric n1ethods
of synthesis \Yith three and four accuracy points. Xu,nerous sa1nple
problen1s are given here; son1e deliberately contain pitfalls to sho,v what
can be done ,vhen trouble arises. In the presentation of algebraic
n1ethods of synthesis using displacen1ent equations or con1plex nu1nbers,
an atten1pt is n1ade to include the latest de,,elop1nents on synthesis ,vith
five accuracy points and the use of digital con1puters in the solution of
advanced proble1ns.
The last chapter, Synthesis of Spatial Linkages 1 is an original
contribution; this 1naterial is to be found no,vhere else. Previous ,vork
of the authors on the analysis of spatial 1nechanisrns by n1atrix algebra
is sununarized. This is follo"·ed by several sections on the synthesis of
spatial linkages 1naking use of displace1nent equations. J-Iere again a
nun1ber of exan1ples are given.
In addition to the n1aterial ,vhich properly belongs under the title
of kine1natic synthesis, co1nple1nentary 1naterial has been added to 1nake
the book 1nore readable and self-contained. The reader is introduced to
the history of kine1natics, and the develop1nent of concepts and pro­
cedures fron1 the great personalities of the past is traced. The chapter
deYoted to the construction of kine1natic n1odels is of the greatest aid in
establishing a rapport bet,veen the static dra\\·ing of one phase and the
1nany other positions a n1echanis1n \Yill assu1ne. 1-Iodels give i1n1nediate
infor1nation about displace1nent relations, unfavorable transn1ission
angles. dead points, binding, and friction. ..\ reYie,Y of con1plex nun1-
bers, detern1inants, and solutions of linear equations is included as an
appendix. l'hese tools, rust-covered for n1any readers (the voice of
experience), are resharpened in an effort to 1nake the book a self-sufficient
unit suited for self-study.
Our spiritual guide and close friend "'as the late Professor Rudolf
Beyer of ::\Iunich. His "·orks and infectious enthusias1n have been the
inspiration for this book and the source of strength for overco1ning the
n1any difficulties and discouragements inherent in the process that
viii PREFACE

creates any book, especially one dealing ,vith the development of deduc­
tive processes.
The contributions of 1nany others-friends, colleagues, and
students in over.vhehning nun1bers-cannot be ackno"·ledged indi­
vidually, 1nuch as "·e "·ould like to. \Ve ren1ain grateful to our benefac­
tors even though their kindnesses and help 111ust go unrecorded.
It is a last pleasant duty to ackno,vledge the suggestions of
the reviewers of the manuscript.

Richard S. Hartenberg
Jacques Denavit
 CONTENTS

V
Preface

Chapter 1: AN OUTLINE OF KINEMATICS TO 1900 1

Chapter 2: CONCEPTS AND NOTATIONS RELATED
TO MECHANISMS 28
2-1 MECHANISMS 28
2-2 CLASSIFICATION SYSTEMS 29
2-3 RIGID AND RESISTANT BODIES 30
2-4 MOTION--RELATIVE AND ABSOLUTE 31
2-5 CONNECTIONS, AND THE PARTICULAR MOTIONS THEY PERMIT 31
2-6 LOWER-PAIR CONNECTORS 32
2-7 HIGHER-PAIR CONNECTORS 38
2-8 FOUR-BAR LINKAGES 38
2-9 SLIDER-CRANK MECHANISMS 44
2-10 TRANSMISSION DEVIATION, AND PRESSURE ANGLES 46
2-11 PLANAR AND SPATIAL: THE MOTIONS AND THE MECHANISMS
49
2-12 KINEMATIC CHAINS 52
2-13 INVERSION 55
2-14 EXPANSION OF REVOLUTE PAIRS AND OTHER DISGUISES 57
2-15 PRISMATIC PAIR AS THE LIMIT OF A REVOLUTE PAIR
59
2-16 EQUIVALENT LINKAGES
61
2-17 SYMBOLIC NOTATIONS
64

Chapter 3: KINEMATIC MODELS 68

3-1 Tl IE MEANING OF MODELS
68
3-2 HISTORICAL SURVEY
70
3.3 PLANAR KINEMATIC MODELS
75
3.4 SPATIAL KINEMATIC MODELS
84

Chapter 4: KINEMATIC ANALYSIS OF

PLANAR MOTION 88
4-1 88
INTRODUCTION
4-2 COINCIDENT POINTS
89
4.3 91
NOTATION
4.4 LINEAR AND ANGULAR VELOCITIES
92
4.5 RELATIVE VELOCITY AND VELOCITY DIFFERENCE 97
4-6 INSTANTANEOUS CENTERS OF VELOCITY
103
X C O XT E N T S

4-7 POLES AND CENTRODES 109

4-8 ACCELERATION 117
4-9 RELATIVE ACCELERATION AND CORIOLIS ACCELERATION 120
4-10 ACCELERATION DIFFERENCE 124

Chapter 5: TYPE, NUMBER, AND

DIMENSIONAL SYNTHESIS 130
5-1 KINEMATIC SYNTHESIS 1 30
5-2 TYPE SYNTHESIS 132
5-3 NUMBER SYNTHESIS 132
5-4 DIMENSIONAL SYNTHESIS 138
5-5 SPACING OF ACCURACY POINTS 140
5-6 CHEBYSHEV POLYNOMIALS 144

Chapter 6: FOUR-BAR COUPLER-POINT CURVES 148

6-1 THE FOUR-BAR LINKAGE 148
6-2 EQUATION OF COUPLER CURVES 150
6-3 DOUBLE POINTS AND SYMMETRY 161
6-4 THE ROBERTS-CHEBYSHEV THEOREM 168
6-5 EXTENSIONS OF THE ROBERTS-CHEBYSHEV THEOREM 176
6-6 STRAIGHT-LINE MECHANISMS-APPROXIMATE AND EXACT 179
APPENDIX: ATLAS OF FOUR-BAR COUPLER CURVES 186

Chapter 7: THE EULER-SAVARY EQUATION AND THE

CUBIC OF STATIONARY CURVATURE 194
7-1 THE EULER-SAVARY EQUATION AND THE INFLECTION CIRCLE 194
7-2 THE CUBIC OF STATIONARY CURVATURE 204
7-3 EXAMPLE: FOUR-BAR LINKAGE TO REPLACE CIRCULAR GEAR
SEGMENTS FOR SMALL ROTATIONS 210

Chapter 8: GEOMETRIC METHODS OF SYNTHESIS

WITH THREE ACCURACY POINTS 215
8-1 INTRODUCTION 215
8-2 POLES OF THE FOUR-BAR LINKAGE 216
8-3 RELATIVE POLES OF THE FOUR-BAR LINKAGE 220
8-4 EXAMPLE· LOGARITHMIC-FUNCTION GENERATOR 225
B-5 EXAMPLE: SINUSOIDAL-FUNCTION GENERATOR 231
8-6 POLES OF THE SLIDER-CRANK MECHANISM 236
8-7 RELATIVE POLES OF THE SLIDER-CRANK MECHANISM 237
8-8 EXAMPLE: RECTILINEAR RECORDER MECHANISM 244

Chapter 9: GEOMETRIC METHODS OF SYNTHESIS

WITH FOUR ACCURACY POINTS 249
9-1 POLE TRIANGLES 249
9-2 FOUR POSITIONS OF A PLANE 255
9.3 PROOF OF THEOREM IV 267
9-4 EXAMPLE: SINUSOIDAL-FUNCTION GENERATOR 270
9-5 EXAMPLE: RECTILINEAR RECORDER MECHANISM 274
9-6 EXAMPLE: ACTUATION OF PRESSURE CYLINDER 279
9-7 EXAMPLE, APPROXIMATE DWELL LINKAGE 286
C O XT E X T S XI

Chapter 10: ALGEBRAIC METHODS OF SYNTHESIS

USING DISPLACEMENT EQUATIONS 295
10-1 DISPLACEMENT EQUATION OF THE FOUR-BAR LINKAGE 295
10-2 CRANK AND FOLLOWER SYNTHESIS: THREE ACCURACY POINTS 297
10-3 EXAMPLES: FOUR-BAR FUNCTION GENERATORS
WITH THREE ACCURACY POINTS 298
10-4 CRANK AND FOLLOWER SYNTHESIS: ANGULAR VELOCITIES
AND ACCELERATIONS 303
10-5 GENERALIZATION OF THE SYNTHESIS METHOD BY LINEAR EQUATIONS 305
10-6 SYNTHESIS OF THE SLIDER-CRANK MECHANISM
WITH THREE ACCURACY POINTS 305
10-7 SYNTHESIS OF THE SLIDER-CRANK MECHANISM
WITH FOUR ACCURACY POINTS 308
10-8 CRANK AND FOLLOWER SYNTHESIS: FIVE ACCURACY POINTS 311
10-9 ANALYSIS OF MECHANICAL ERRORS IN LINKAGES 315
10- 1 0 MECHANICAL ERRORS IN FOUR-BAR LINKAGES 316
10-11 GEOMETRIC INTERPRETATION OF THE ERROR DENOMINATOR 319

Chapter 11: ALGEBRAIC METHODS OF SYNTHESIS

USING COMPLEX NUMBERS 321
1 1 -1 VELOCITY AND ACCELERATION SYNTHESIS BY COMPLEX NUMBERS 321
1 1 -2 COUPLER-CURVE SYNTHESIS: FIVE ACCURACY POINTS 327
1 1 -3 REDUCTION OF THE FIRST PAIR OF COMPATIBILITY EQUATIONS 337

Chapter 12: SYNTHESIS OF SPATIAL LINKAGES 343

1 2-1 INTRODUCTION TO THE STUDY OF SPATIAL LINKAGES 343
12-2 DISPLACEMENT ANALYSIS OF THE TWO-REVOLUTE
TWO-SPHERIC-PAIR MECHANISM 344
12-3 AN OUTLINE OF THE MATRIX METHOD OF ANALYSIS 347
12-4 SYNTHESIS OF FOUR-REVOLUTE SPHERICAL MECHANISMS 3S6
1 2-5 SYNTHESIS OF TWO-REVOLUTE TWO-SPHERIC-PAIR MECHANISMS 358
12-6 FUNCTION GENERATORS FOR SYMMETRICAL FUNCTIONS 362

Appendix: MATHEMATICAL TRACTS 369

A-1 COMPLEX NUMBERS 370
A-2 LINEAR SYSTEMS OF ALGEBRAIC EQUATIONS-I: DEFINITIONS 380
A-3 DETERMINANTS 380
A-4 LINEAR SYSTEMS OF ALGEBRAIC EQUATIONS-II: RANK
AND SOLUTION 389

Problems 397
Index 431
 AN OUTLINE OF

KINEMATICS TO 1900

The history of kinematics, the story of the development of the

geometry of n1otion, is composed of evolvements in machines,
mechanisms, and n1athe1natics. The story begins "·ith the ran­
dom gro,vth of n1achines and mechanisms under the pressure of
necessity. This ,vas a period of invention and establishment
of basic forms, but there ,vas neither unity nor plan. :\Iuch
later, ,vhen existing algebraic speculations on the generation of
curves "·ere timidly applied to physical problems, the concept
of kinematics, a science, ,vas affirmed and the ancient art of
mechanisms complemented. :.\lore recently, mathematics has
been applied ,vith deliberation and firmness. Although no
general relationship linking all mechanisms has been disco,·ered,
n1odicu1ns of unity and plan have been recognized for so1ne areas.
:\lore significant is the ability to solve a ,vide variety of prohlen1s;
son1e require the use of sophisticated procedures feasible only
because of high-speed con1puters.

:\Iachine and mechanism are ,vords giving t,vo different i1nages.

The term machine is associated ,vith the use and transformation
of force, and although motion in varying degree is encountered
in a n1achine, the idea of force dominates. :\[echanism, on the
other hand, definitely conjures up the idea of motion, and while
2 KINEMATIC SYNTHESIS OF LINKAGES

forces do exist, they are relatively small and unimportant compared with
the exploitation of motion. However, machine and mechanism inhabit
the same body.
�[achines are spoken of early in history, but ,vhat ,vas considered a
n1achine t,vo thousand years ago differs considerably from our present ideas.
Fron1 earliest times man has found his o,vn po,vers inadequate for the
tasks he set himself, an1ong thern the moving of heavy ,veights. 1'his
,vas a rnajor concern of the ancients, as ,ve may judge from their imposing
monuments of large stones. Vitruvius,b a military engineer ,vriting
about 28 B.c., defined a machine as "a combination of timber fastened
together, chiefly efficacious in moving great ,veights." About a century
later, Hero of Alexandriab summarized the practice of his day by na1ning
the "five simple machines" for "moving a given ,veight by a given force"
as the lever, ,vindlass, sere\\' for po,ver, "'edge, and tackle block (pulley).
Until nearly the end of the nineteenth century it was held that these
"five mechanical po,vers" were the building blocks frorn ,vhich all more
complex assemblages ,vere constructed.
Today ,ve no longer consider a machine to be constituted of a
single body such as a lever; indeed, the use of a lever involves several
bodies, or objects capable of exerting force. Today's thoughts on
machines involve assernblages made from six basic 1nechanical com­
ponents, identified about 1875 by the German kinematician Franz
Reuleaux.b 1'hese parts are (1) the eye-bar type of link, called cranka1
in kinematics; (2) the ,vheel, including gears; (3) the cam in its many
forms; (4) the scre,v for communicating motion and force; (5) the inter­
mittent-motion devices called ratchets for want of a better name; and
(6) the tension-compression organs, or parts having "one-way rigidity,"
as belts (chains) and hydraulic lines. It is from such parts that machines
-and mechanisms-are built.
Each of these components was invented in antiquity and put to
use ,veil before the beginning of the Christian era, either singly or in
combinatio11. 1\lexandria (in Egypt, but nevertheless a Greek town)
became the greatest center of learning of the ancient ,vorld under the
Ptolemies iu the third century B.C. It was here that Greek science,
philosophy, and technology were nurtured and further developed; Euclid
wrote his imperishable "Elements" in Alexandria.
::\Iechanical components had been given application mainly in
\\·ar machinery-"thro"·ing engines" for tossing projectiles of various
sorts-man always having found money for research on and development
of armament. The mobilization of technology and science appears to
6
The superscript b indicates a biographical note at the end of the chapter.
1The name is 1nisleading but ineradicable; the resernblance to what is com­
monly called a crank today is not immediately evident. Lever is sometimes used.
AN Ot:-TLINE OF KINEMATICS TO 1900 3

be as ancient as warfare. Archimedes, ,vho ,vished to be remembered as a

mathematician (he ,vas the Xe,vton of his day), is better kno,vn as an
engineer because of the "engines" ,vith ,vhich he thwarted the Roman
siege of Syracuse (214-212 B.c.).
The labors of daily living "·ere only lightly touched by applica­
tions of mechanics. The muscle po,ver available from the bulk of the
populace gave little stin1ulus to the construction of either laborsaving
devices or production machinery except for the never-ending tasks of
raising ,vater and grinding grain. For the transmission of power, the
complexity rarely exceeded the use of more than t,vo gears to proportion
input and output shaft speeds, as bet"·een ,vater ,vheel and millstone,
noted by \'itruvius. ::\lechanisn1s ,vere more sophisticated (Fig. 1-1).
\Ve may note that po,ver and complexity ,vere not in the same device.
::\lany small and ingenious devices appeared in the temple autom­
ata designed to excite wonder in the ignorant people of pre-Christian
Alexandria. Here pneumatics as ,vell as "hard" mechanics ,vere applied
to n1echanisms that ,vould open temple doors after the kindling of
sacrificial fires, vend liquids on the dropping of a coin, produce stage
effects "'ith moving scenery, and the like. There ,vas a clever employ­
ment of speed and force relations, and in some cases regulators ,vere
fitted to control the motion.
\Vith the Roman take-over of Egypt and Alexandria at the tin1e
of Julius Caesar, Cleopatra, and Christ, science almost came to a halt.
The Ron1an Empire produced little ne,v in the ,vay of things mechanical
but accomplished nn1ch in the shre,vd exploitation of its Alexandrian
inheritance by ,vay of bigger and better machines. The Roman effort
,vent into the construction of buildings and roads: the principal Roman
invention ,vas hydraulic cement. Follo,ving the collapse of the Western
empire, the center of effort shifted to Constantinople. Before long­
it was 642-the Arabs, activated and disciplined by their prophet
�Iohammed, conquered Egypt. From this start, scientific and tech­
nological prO\\'ess ,Yas distinctly Arabian for the next centuries. The
.-\rabs not only had acted as transmitters of the Alexandrian or Greek
,vorks to the Western ,vorld but also had produced refinements and
substantial innovations of their own.
It is wrong to assume, in the light of recent studies, that Europe
,vas devoid of ideas during the "Dark A.ges," the span bet,\.·een, say,
.-\.D. 500 and 1000, during ,vhich the A.rahs rose to greatness. The
records of these centuries are indeed scanty if we look for machines and
n1echanisms. The social and economic structure, so different from the
Roman, provoked a tremendous change in agricultural methods. The
nonmechanical details are fascinating miniatures we cannot take the
tin1e to consider. A.t any rate, agricultural surplus led to rapid urbani-
4 KII'i'EMATIC SYNTHESIS OF LIN KAGES

zation. Artisans and merchants rose in the ne"· cities. The later
medievalists ,vere ready to explore 1nechanical devic;es suited to "pro­
duction" and to exploit the "natural po,vers" of ,vind and water to com­
plement (and in many cases supplant) 1nuscle po,ver. These nev.- po,vers
\\'ere to spa,vn a multitude of machines \\·it.h increasingly complex
mechanisms.
Documents of the t\velfth and thirteenth centuries containing
1nachine sketches are fe,v. Ho,vever, refinements are to be seen
(Fig. 1-2), and the designers' enterprise is limited by available materials

Gear: each tooth and hole

- ---
----- / correspond to a mile

�-:;,-..._

c:,
400-tooth
gear

0
1-Tooth
gear

,,,,Tube

Bronze, _ / �-· gear

vessel y.
L
_ o o.

Carriage wheel
of 4-ft diameter

t'lGl.RE 1-1 Schcn111tic of an odo111etC'r for a <•11.rriage, as described about 28 B.c.

by Yitruvius in "Ten Books on Architecture," translated by 1\,1. H. fvforgan,
HarYard University Press, Can1bridge, !\'lass., 1926.
Vitruvius, 1nilitary engineer under Augustus Caesar, was no innovator but
onl)· u reporter of his past and present. He ascribes this device, "which enables us,
while sit.ting in a carriage or sailing by sen, to know how many miles of the journey
we hnvr accomplished," to Ctesibius, a (:reek of proba,bly the second eentury B.c.
Vitruvius' description is Yerbal and not pictorial: the wheel dian1eter and the
numbers of teeth in the gears are S(W<"ified. One eon1plete reYolution of the
400-tooth gear represents 5,000 ft (<·ailed a mile). \\'ith each such revolution, a
small stone in a hole of the top1nost gear lined up with the tuhe leading to the bronze
ves:scl (gating device!): '·Hence, every stone. n1aking a ringing sound as it falls,
will give warning that we have gone one 1nile.'' On board ship, the sa1ne principlr
"was applied by fitting a paddle wheel to the axle" projecting through the hull.
A century later Hero described a more sophisticated form with four worm­
wheel redu<:tions between input axle and output "stone wheel"; intermediate shafts
carried dials to indicate parts of ti 1nile.
 11• ,,,......
-
....
rn,o·•• p __., d,' ... ,.,·• •■. •-t •llr. !luo •••-•

----.. ·-· ·-· ..........

.
... ,.1 ·,- t ,.• ,.. ......i ··.:,,i ,,..-· ....i....,. ·.. i.::

..
:...11,.: .
•-"r .......,.,_,J,.._.,,· ·,�11.i,.,,,_,_ "-,,,,., ....-.
"' ,_, .... ···"··- ............

..-..:� ............................ . . .•
:.,,..,,., l •n,_, • ._,•• .,. ....-:...-. ,o:-.!,. ..!i:.,�.....:.:
...,,.;,. h,;;
....,....,.... . ..,.,....,................ -r.._,..,. . ..., .....,
, , .,, ,.,,.,.,,.,, , ..... . ..... ..
:::,-::�. 7,'.�•;l,: :,t�i:�u:,�:::::';'.7,'.i;;:�I :·:::!:�i�: ;;:���;::
lt l i i

1,:"..,.,..,,.,1,,.-..i11,..,..•,.1.,, ......

hi, ,,...�:-:,:�. '�'",'..:'i:.�':/;,: ,�'lt:1t":!:..":';';;::--:::�.::�;�

1t

ll•••1ltt !hoy """1.o.io. !ltoy onut P'"l"",11 IN -· •• •ltiltil, f•r \},;,

JOO•IN. •11•n•i.h,. -,li,Ji.l.u1lt,ooo•opoltlioly,,•,;1,l,l�<1•lltltt lot,,
po.rt ol. u.. .,;...,,..,.,1t .,.,n,.,1; oor ;, th,,.. ••l•r ,_,.,f .,·Woh
., tl,, -•1 id-., ,.,.., ,,.,.10..i '" ,,nil• ,1,.,,, Do ,·i,<i', l ir...i-
""'·••1-3 T,iph&mmo,ia••mitby,o••••rnoiooitb, ..i,o,;.,.,o{tho
,.,«.a·hool'outilityb:,lh,"°"'.......,of""""""°""'°"'""';•u,pe,i,di<
"up-Hd-do><lo"""''""'- "'"PP<•!.fti,l'ythac,,...,1t,:,..i...,.di.........,.
"''"'"-1u,-,,"..,..,.-",i....,..o<1o{1t.-t,.,m._ F_x,..,.....,.·,
,
•-y,_.,. ......,..,,. ,__,,,.""-)
 .-\ :'.\' OUTLINE O F KINEMATICS TO 1 900 7

Europe of the sixteenth century possessed a number of sketch­

•
books sho\\'ing machines-and hence mechanisms. The "picture-book"
style that originated was to continue for centuries, for drawings could be
understood and transformed into hardv,are even though the text ,vas
beyond the craftsman's abilities. Some books ,vere for specialists, as
those dealing exclusively ,vith military devices or mining machinery
(Fig. 1-4). Others had a ,vider scope, the machines being classified into
those for raising "'ater, milling grain, sa·wing timber, and so on. Some
undertakings ,vere spectacular (Figs. 1-6 and 1-7).
'\Vith Leupold/ in the early eighteenth century, there is perhaps
the first reference to the idea of 1nodifying 1notion rather than just the
construction of machinery; he speaks of cranks and cams and ho,v, say,
a continuous circular motion may be transformed into a "back-and­
for,vard motion." Here was the earliest attempt at a systematic treat­
ment of mechanisms.
Two great figures appear in the eighteenth century, Euler ,, (1707-
1783) and ,vattb (1736-1819). Although their lives overlapped, there
,vas no apparent contact between the two men. Both Euler and Watt
dealt ,vith the specific and concrete, but along vastly different paths.
Euler's ":.\Iechanica sive motus scientia analytice exposita" (1736-1742)
is, to quote Lagrange, "the first great ,vork in which analysis is applied
to the science of move1nent." The concept that planar motion may in
general be described by the translation of a point and a rotation about
that point appears in "Mechanica." This thought, extended to velocity
and acceleration, is the root of analysis, graphically sho,vn by the vector­
polygon methods. The associated velocity and acceleration equations
came after Euler but because of their heritage are sometimes given the
generic title of Euler's theore1ns in the German literature. The funda­
n1ental idea of kinematic analysis thus stems fron1 Euler, as do many
other astute observations.
Watt, instrument maker and engineer, a contrast to Euler, ,vas
c�ncerned with the synthesis of movement. Generally speaking, mecha­
nism designers before \Vatt had confined their attentions to the motions
f links attached to the frame. We may call them the first and last
�hnk s; one having been given an input motion, the other produced a
transformed motion called the output. It ,vas Watt ,vho focused on
the motion of a point on the interniediate link of the four-bar n1echanism:
he�e was a ne,v kind of output ,vith the year 1784. The application of
this brilliant thought allo\ved Watt to build a double-acting steam engine;
the earlier chain connecting piston and beam (Fig. 1-8) ,vas no,v replaced
by a linkage able to transmit force in t,vo directions instead of only one
(Fig. 1-9). \Vatt had discovered coupler-point motion, although its
definition in these terms lay well in the future. It "'as a singular achieve­
ment, a pivotal point along the road of synthesis.
8 K l :-- E M .-\ T l C S Y N T H E S I S 0 1" L I N K A G E ::;

Of the t,vo discoveries, Euler's "·as to lie unnoticed by kinemati­

cians for a century or so. \Vatt's ,vas pursued in desultory fashion by
engine builders and mathematicians. Linkages ,vere devised to compete
with or supersede Watt's n1echanisn1; that certain of these linkages ,vere
actually related developed only after a generalized treatment of coupler­
point motion, kno,vu for a ti1ne as "three-bar motion," ,vas given a
century later by Samuel Roberts.b
A. further perception is to be noticed in an observation 1 of Euler's
in 1775:

1'he investigation of the motion of a rigid body may be conven­

iently separated into two parts, the one geometrical, the other
mechanical. In the first part, the transference of the body frorr1
a given position to any other position must be investigated
without respect to the causes of 1notion, and must be represented
by analytical formulae, ,vhich ,vill define the position of each
point of the body. This investigation ,vill therefore be referable
solely to geometry, or rather to stereotomy.
It is clear that by the separation of this part of the question fro1n
the other, ,vhich belongs properly to �1echanics, the determina­
tion of the motion from dynamical principles ,vill be made much
easier than if the two parts ,vere undertaken conjointly.

Euler's remarks were directed at the rigid bodies of classical

mechanics, not specifically to machine parts; yet we see here the beginning
of the separation of the general problem of dynamics into kinen1atics
and kinetics. We note in passing that his contemporaries D'Alembert
and Kant ,vere of the same mind, treating motion in a purely geometric
"Novi co1n1nentarii Aca<lemiae Petrop.," vol. XX, 1775; also in "Theoria
1

motus corporu1n," 1790. The translation is by Willis, "Principles of Mechanism,"

2d ed., p. viii, 1870.
------ - --··-··-· - --· - -- _ _ _ ____... -·· --··
t'IGURE 1-4 �line hoist of the sixteenth century. This hoist is reversible, the direction
of rotation being controlled by the water- admission valves attached to the levers C
and D. Note that the wooden axle I is of square cross section, rnaking it easy to
transmit and take off torque without keys or splines. The axle is supported by gud­
geons (iron pins of small dianieter, one of which is to be seen at the left) and is locally
reinforced by shrunk-on iron bands to prevent breakout. The use of chain was
typical, vegetable fibers being unsuited to the damp of mines. Fro1n Agricola's "De re
,netallica," Basel, 1556, translated and annotated by �Ir. and 1\1rs. Herbert C. Hoover
long before l\1r. Hoover became President of the United States. (Mining Magazine,
Lond(>n, 1912; also reprint by Dover Publications, Inc., 1Vew York, 1956.)
• ..._ ,_,,_ ....: •�• - - •-- •. O-•n
i�?;;i�[;;;;;;;,;;;;;;;,
r,e.e.. 1-li ,.,....,..,l\,iJs<,W•l<rw<><M. M J,a• o fru,., m<WU,Y by,...,.,n l!•t, fo,

wh<cl,a",l'•,.l,fJ • , t<><•·,,,,""'''"" Th<pumpru«• •" • ",..·•1, Mn,mot.,ol o

'"" •·,bo;,...·•1..1e...i to """'""'lo, Th•pun,po W f;d On1<>1"1""',· '··••h;,h ,.,..;,,1
th;,woWlo•o,•lov• ·.....k "'-,onJad. >•·"•"""'"'•,·••••"
About Jl,ti2 omo P,ltt Mori<,, ::Uo""'• o• Mo,;. i,.,,.11,,1 "• "" ,t ,,i,bri,I
lo...,.,"",in thia ,- o,nl,,--o·i....w,;v.,.,f""",JK•P, u..i.,..,.ol u, • •..,...."'
ore.... ol,i,,oJoo,Hnd••- Th,,lln, o/,10:«,..._....,,..i,,u,,....-,:!,l>uildioc,,,•nd
..,.1,1,,:1,11,... 1o,- 1h,...,o;...,.,. m,,how;n. M j,..l.ed,ro h< ,,f•i•I>,..,,..,.,."
""""" 'Tllo,•OOloolU..••..•"'••h wM,<l<ot"',.., h, ,b, fu,. ol t1611
�;:-
 A X O U T L I N E O F K I N E M .\ T I CS T O 1 900 11
(a) Profile, showing cisterns, piping. and elevations Cistern Ill
El. 533
(dimensions = feet)
c·
1stern 11
El. 345
Cistern I
El. 160
River
El. 0
_ J; J.?5 .
. - - -j - - - · - -!- -- - --- - - ---+ --- -· --- - .. - - +- - - - - - - - - - 1-j
1,000 • 2.000 3,ooo
�-600 · ·: 1.344-- --i--
- - - - 1 .976 -.·l.4-ooo
,
I I

; I
-
Petits
I
49

Chevalets - 2,020
(b) Plan, showing pumps and linkages (chevalets)
W Battery of 14 water wheels
Q Pump location - number of pumps shown by figures
=== Pipelines
Linkages to pumps

FIGURE 1-6 Sketch of the l\,farly machine, a pumping installation set to work in 1 683
near Paris.
This machine, the largest of its day, was a wonder of the world for the next
century or so. Fourteen 40-ft-diameter water wheels were exposed to 1,200 hp from
the river Seine to drive 225 pumps lifting water to a. reservoir 533 ft above the river,
establishing a substantial gravity head for the fountains, cascades, and 60-ft-high jets
at Versailles and Marly-le-Roi.
. The pumping was done in three stages. The first consisted of pumps at the
river, shoving water to hillside reservoirs at elevation 160. Pumps here were driven
by seven 600-ft linkages, and moved the water to higher reservoirs at elevation 345.
Pu1nps at this station were actuated by thirteen 2,020-ft-long linkages, depositing the
water at the 533- f t level of the aqueduct. The linkages were of the parallelogram type
and consisted in part of 64,000 ft (a.bout 12 1niles!) of iron bars. �1ost of the power
was used to overcome friction and inertia.
\\liat ma.de the machine unusual was the number of linkages, the power
handled, and the noise; the lengths of the linkages were modest even for those days,
but the power was not to be exceeded for two centuries.

manner, i.e., disregarding the causes of motion. This is ,vhat L. N. l\il.

Carnot" later called "geometric" motion.
By the close of the eighteenth century, :.\Iongeb had proposed a
course on elements of machines for the f:cole Polytechnique:

By these elements are to be understood the means by ,vhich the

directions of motion are changed; those by ,vhich the progressive
 ,2 K I N E M A T I C SYNTH E S I S OF' L I N K A G E S

motion in a right line, rotatiYe n1otion and reciprocating motion,

are made each to reproduce the others. The most complicated
machines being merely the result of a combination of some of
these elements, it is necessary that a co1nplete enumeration of
them should be dra"'n up.

'fhe suggested enu1neration ·was prepared by 1\rionge's long-time colleague

Hachette. This system classified n1echanisms by capability, i.e., the
nature of the motion trausformation a mechanism effected bet,veen input
and output. It "·as presented in chart form vvhen the course finally
materialized in 1806. The chart depicted those mechanisms able to do a
certain job, as transforming "circular continuous motion into rectilinear
alternating motion " ; various motion combinations "·ere tabulated.
This chart has been most durable, for today's compendiun1s of mechanisms
are its direct descendants. Hachette's book "1'raite elementaire des
machines" did not appear until 1 8 1 1 ; a fe,v years later Hachette's course
-and book-formed part of West Point's curriculum. It ,vould have
been the first kinematics text had it not been for "Essai sur la composition
des machines" of 1808, brought out by Lanz and Betancourt, also of the
tcole Polytechnique. 1'his text ,vas based on the :\Ionge-Hachette
scheme and, because of publication prior to Hachette's book, erroneously
receives credit for the pioneer ,vork done by Hachette.
Two more names and t,vo ne,v but related classification systen1s
intervene before the actual coining of the ,vord kinematics. 'fhe men
involved ,vere Borgnis, an Italian engineer and professor of mechanics
at the 1Tniversity of J>avia; and De Coriolis, chief engineer of roads and
- - -- - ----- -- - · · - --· ----......- --------· -· --·--------- -
FIGtlRE 1-7 l\1oving the Vatican obelisk, 1586. (From Fontana, "Della trasportatione
dell'obelisko ·vaticano," Rorne, 1590.) The picture is by courtesy of the Burntly
Library.
A 361-ton obelisk, cut in Egypt in the thirteenth century B.c., brought to Rome
about A.D. 41, and placed in the Circus of Kero, ,-..·as to become an architectural feature
of the new cathedral of St. Peter begun in the early sixteenth century. To bring the
obelisk into position in front of the cathedral required a move of 275 yd. 1lany plans
for doing this were submitted, that of Fontana being chosen. Very briefly, it involved
encasing the obelisk in oak planking, to give beam strength , and the use of a timber
tower, block and tackle, levers, rollers, etr., with 800 nH'r1 and 74 to 140 horses deployed
at the capstans (the figure shows four).
The obelisk was lifted vertically by five large levers and tackle connected to
40 capstans, after which a long rarriage was slipped under. The next step, the laying
down, was handled in part by an angular brace pin-connected to the oak casing. The
right, or sliding, end of this brace was controlled by further tackle. In this brace
we see the use of a slider-rrank mechanism, a kinematic arrangement that gave gtabil­
ity and 1naterially reduced the load on the tower tackle during the most trying phase.
14 K I N E M A T I C S Y J\' T H E S I S O F L I N K A G E S

highv,ays, but also professor in Paris, the man of the acceleration com­
ponent, who so deftly defined the product of force by distance as work.
Coriolis simp]ified Borgnis' systen1 by reducing the concept of a complete
machine to three parts or groups of parts, viz., the receivers of energy,
or recepteurs, the comrnunicaf,eurs, or parts transmitting motion, and the
operateurs, or the ,vorking parts or tools.
The physicist Ampere,1> in the course of trying to establish the
domain of his specialty, was led to defining and classifying "all human
knowledge." The consequence ,vas his "Essai sur la philosophie des
sciences" (1834). To recognize an area of mechanics different from
statics, dynamics, and molecular mechanics, he coined a ne,v term ; it
·was cineinatique, derived from the Greek ,vord for motion. Kinematics
(cinernatique was anglicized and was also adopted by the Germans as
Kinen1atik) ,vas given recognition as a science:

There exist certain considerations ,vhich if sufficiently developed

,votild constitute a complete science, but ,vhich had hitherto

Arch head

Chain

Atmospheric FlGCRE 1-8 Schematic

47ft pressure of a N ewcomen-type at­
mospheric engine circa
1775, just before Watt.
Single-acting
cylinder, The "great beam"
6 ft bore, (lever), arch heads, and
9 ft stroke chain functioned as a
simple pulley or direction
changer of applied force.
, With the piston at the
..... . Anti-backlash
weights top of the cylinder (hav­
ing gotten there because
Boiler of the weights on the
15 X 15 ft pump side), steam was
To reciprocating admitted and then con­
pump densed by a water spray.
In consequence of the
subatmospheric pressure
in the cylinder the piston
was forced down. Steam
Floor and datum was not used to force the
'1i,,nr1r1r1r1mrrur1r1r1rmilrur1rm"mr1rrurm1r1r1r/mirm1rn1, piston up.
A. 11. O U T L l :\ E O F K I :>: E M A T I C S T O 1 900 15
.-···.. A

· ...·
(a) Rod to 14 ft
condensate
pump .--i--,
Connecting
rod
Double-acting cylinder, 13 ft
19 in. bore 4 ft stroke

FIGt:RE 1-9 Schematic of \\'att's rotath·e engine, 1784, the first engine to produce
power directly on a shaft without the intervention of a water wheel fed by a
reciprocating pumping engine.
At (a) is shown the kinematic skeleton of a four-bar linkage with coupler AB.
The coupler point M describes the figure-eight-shaped coupler-point curve,
possessing two segments that are nearly straight lines; the vertical segment was
used for piston-rod guidance. In (b) the linkage is shown as part of the engine.
By means of a pantograph, another point JV' was given a motion similar to that
of ,\f.

been neglected, or have formed only the subject of memoirs or

special essays. This science ought to include all that can be
said ,vith respect to motion in its different kinds, independently
of the forces by ,vhich it is produced. . . . It should treat in
the first place of spaces passed over, and of times employed in
different motions, and of the determination of velocities accord­
ing to the different relations ,vhich may exist bet,veen those
spaces and times.

Euler's earlier idea thus received identity as ,vell as reiteration :

kin ematics was the purely geometric science of motion in the abstract.
As such, kinen1atics ,vas to develop in large and significant part separately
from mechanics and applications of mechanics. :.\Jany 1nathematical
contributions ,vere reduced to mechanisms appreciably later. Forces
\Vere left as complications to be dealt ,vith separately; "·e still do this
Under the title of kinetics of machinery.
Ampere defined the path but left the milestones to be set by
0thers. Among the first were Chaslesb and Poinsot. b They considered
geometrical bodies-not points-adding method and ideas to the concept
of the instantaneous center discovered by Johann Bernoulli in 1742.
16 K I 1' E �L\ T J C S Y I\' T H E S I S O 1'" L I .'\ K .\ <, }<; S

Across the Channel, \Villis'b text, "Principles of ::\Iechanis1n"

(1841), marked a departure from the tradition of the descriptive past.
Willis made a substantial contribution to the cause of synthesis-"·e see
this no,v- by turning to classification. It ,,·as his idea that thrre should
be a systematic approach to the design of n1echanisn1s for a given purpose.
He sought to augment e1npiricisn1-"that species of intuition ,Yhich long
familiarity ,vith a subject usually confers. . . . "
Willis proposed that "the motions of the n1achine are the principal
subject of contemplation"-an obvious acceptance of Ampere's vie,vs on
the matter. In considering logical order, Willis' ideas ·were different
from those of ::\Ionge and Hachette. He considered "that all the simple
combinations of mechanism," i.e., fron1 input to output, "1nay be dis­
tributed into three classes" :
Class A. directional relation and velocity ratio constant
Class B directional relation constant-velocity ratio varying
(�lass C directional relation changing periodically-velocity ratio
either constant or varying
He counted five ,vays of achieving the relative 1notion bet,veen input
and output links-rolling contact, sliding contact, ,vrapping connectors
(belts, chains), link"·ork, and reduplication (tackle of all sorts). His
work had a considerable influence, for in the neady thirty years bet,veen
the first and second editions of the "Principles," the classification,
nomenclature, and figures ,vere adopted by many authors.
Willis rejected Prony's analysis of the Watt curve, saying "the
algebraical equation . . . is exceedingly involved and complicated, and
of 110 use in obtaining the required practical results, ,vhich are readily
deduced by simple approximate n1ethods, . . . . " He did, ho\\·ever,
apply trigonometry to the Watt mechanism and some of its variations,
but ,vithout the generalization of Prony. The (Richard) H..obertsb four­
bar linkage ,vas likev,rise exarnined for proportions. We note, paren­
thetically, that this ·was probably the first mention of this linkage, ,vhose
actual time of origin and point of use are still to be discovered.
The second edition of Willis (1870) was substantially larger than
the first. 'fhe essential principles remained unchanged, and differential
calculus ,vas introduced sparingly, although the subject of acceleration
was left untouched. Older topics ,vere enlarged upon, as, for example,
the Hooke universal joint. The i1nplications of coupler curves escaped
Willis, for although he had a device capable of dra,ving such curves
prior to 185 I, 1 they receive no further connnents. The Peaucellier true
straight-line mechanism (1864), a linkage that caused 1nuch exciten1ent
among mathematicians, seems to have escaped \Villis' notice. But the
influence of Willis continued, as in the reflection by Woods and Stahl,
1
See Fig. 3-2.
 AN OUT L I N E O F KINEMATICS TO 1900 17
" Elementary :\'lechanisrn" (188;3 ), with a sixteenth edition in 1909, and
Robinson's "Principles of :\Iechanisrn" (1896).
Turning to one of the classics of physics, ,ve note that the first
volume of the Thomson and Tait "Treatise on Natural Philosophy"
(1867) devoted some 160 pages to kinematics, follo,ving the sense of
Ampere, i.e., dealing ,vith motion as totally independent of matter and
force.
Rankine'sb "A. :.\Ianual of :.\Iachinery and ::.\Iilhvork" (1869)
devoted 314 of its i5 78 pages to the "Geometry of l\Iachinery, or Pure
:\Jechanism." There is neither a general theory of machines nor a ne"'
systematic treat1nent of their motions, for Willis is follo,ved more or less.
The subject of pure mechanism appears as a great series of interesting
but nonrelated problems. Rankjne dealt ,vith the motion of machine
Parts by using instantaneous centers of velocity, evolving effective
methods for solving special problems.
:\fodern kinematics had its beginning ,vith Reuleaux. His no,v
classical "Theoretische I{inematik" of 1875 presented many views finding
general acceptance then that are current still. They came into English
with l(ennedy's6 able translation, called "Kine1natics of l\tlachines" 1
(1876). I(ennedy's o,vn book, "�1Iechanics of :\1achinery," published
ten years later, although dealing ·with topics beyond only kinematics,
made use of Reuleaux's concepts. Reuleaux's second book, " Lehrbuch
der I{inematik" (1900), consolidated and extended earlier notions. rfhis
book , never translated into English, also n1ade a lasting in1pression.
. Reuleaux's comprehensive and orderly vie,vs mark a high point
•n the development of kine1natics. He gave us the idea that the con­
strained motion of a 1nechanis1n or the kind of relative motion between
parts is controlled by the form of the surfaces of contact of the adjacent
Parts. lVe may think of the1n as ""'orking" surfaces of the connection.
Thus, the connection betw·een a lathe carriage and its bed is by means
of working surfaces (ways) so shaped that only translational 1notion is
Possible. Each of the ,vorking surfaces he called an element. The t,vo
contacting elements of a connection constitute a kinematic pair, the
Properties common to different kinds of pairs allo,ving a certain unity
of treatment.
In the matter of connections bet,veen rigid bodies, Reuleaux
recognized t,vo kinds· he called then1 higher and lo"'er pairs (of elements).
�ith higher pairs, tl;e t\\'O elen1ents are in contact at a point or along a
hne, as in a ball bearing or disk earn and follo,ver; the relative 1notions
of coincident points are dissimilar. Lo,ver pairs are those for ,vhich area
contact may be
visualized, as in pin connections, crossheads, ball-and-
1
Long unavail1tble, this classic has been reprinterl by Dover Publications,
1nc.,
Xew York.
18 K I N E M A T I C S Y N T H E S I S OF L I N K A G E S

socket joints, and some others; the relative motions of coincident points
of the elements, and hence of their links, are similar, and an exchange of
elements from one link to the other does not alter the relative motion
of the parts as it ,vould ,vith higher pairs.
The far-reaching impact of Reuleaux's classification system, in
which 1nechanical and hydraulic devices are arranged in the six classes
listed earlier, is reflected by the chapter headings of modern books on
kinen1atics and machine design.
Reuleaux regarded a 111echanis1n as a (kinen1atic) chain of con­
nected links (or parts), one link being fixed. Fron1 this ca1ne the in1por­
tant proposition of the kinematic inversion of the chain, viz., that a
different mechanism results every time a different link of the chain is
selected as the fixed link. He also recognized the idea of the expansion
of turning pairs.
Through hindsight we sa,v in Watt's straight-line linkage the
amorphous beginning of an ordered and advanced synthetic process.
Nearly a century later Reuleaux identified synthesis as a concept, as an
entity that might be and n1ust be pursued to guide the designer through

FIGURE 1-10 Lowering the obelisk of Luxor, 183 1 . (From Renlea11x's "Lehrlnich der
Kinematik," Brunswick, 1900.)
This obelisk, weighing 284 tons, now stands in the Place de la Concorde in
Paris, having been moved from Luxor. Like the larger Vatican obelisk, the shaft was
prepared for handling by encasing it in timber. Unlike its Italian counterpart, the
Luxor obelisk was tipped off its base directly, Fontana's vertical lift being unnecessar�·
because of the kinematic arrangement-a four-bar linkage achieved with the frame­
work ED (or link b) devised by the engineer Mimerel. The maxirnum load coming
onto the tackle was but 43 per cent of the obelisk's weight. This rig required only
JO men and ½ hr to lay the obelisk down.
 A X OUTLI X E O F KINEMATICS T O 1 90 0 19
the maze of mechanisms. His views were limited to only type synthesis:
by this is meant the determination of the type of mechanism for a given
job, e.g., reaching a decision about whether to use a linkagee-or a cam
to produce a desired motion. He clearly defined this area, ,vhich lies
ahead of the related and more recent fields of n·wniber and diniensional
synthesis, The latter is the main topic of this book.
One more contribution of Reuleaux's must be mentioned, and that
is the symbolic notation ,vith ,vhich he attempted to describe kinematic
chains by n1eans of symbols instead of pictures. His high hope that the
notation ,vould be useful in discovering ne,v mechanisms ,vas borne out
for but fe,v situations. Incomplete and thus not ,vholly satisfactory,
it "·as able to sho,v that many mechanisms of differing physical appear­
ance ,vere nevertheless kinematically identical. Recent extensions of
Reuleaux's kinematic notation do furnish a syn1bolic notation useful for a
large number of problems, as ,vill be indicated. A completely general
and all-inclusive notation is still to be developed,
Geometers and algebraicists of the 1870s became interested in
linkages as curve-dra,ving devices, not as hard,vare; their "'ork has since
been made part of the corpus of the kinematics of mechanisms, It ,vas
discovered that a link motion can be found to describe an algebraic
curve of any order. In particular, the coupler-point motion of the four­
bar linkage (then called a "three-bar motion") ,vas studied. Samuel
Robertsb sho,ved that a coupler point describes a curve of the sixth order.
A theorem important to synthesis and bearing the name of Roberts states
�he existence of three different four-bar linkages capable of drawing
identical coupler curves. It is from this theorem that the relation of
the Watt linkage to other approximate straight-line mechanisms is made
�lear. Only recently have the higher-order plane curves derivable from
! 1nkages more con1plex than the four-bar been examined ,vith kinematic
intent.
The period of the 1870s and 1880s ,vas marked by substantial
steps along the roads of analysis and synthesis. The former unmethodi­
cal (but hardly aimless) gropings gave way to the devising of radical
methods, bringing basic pre1nises into clear view. 1'hus, Aronhold in
Germany and Kennedy in England greatly extended the use of instan­
taneous centers of velocity by recognizing the theorem of three centers,
Acceleration analysis received pot3ntial aid. Kennedy, ,vho reflected
British practice, determined accelerations by graphical differentiation of
velocities calculated from instant centers. In the preface of his o,vn
book (1886), he regretted not being able "to make any use . . . of the
recently published graphic methods of Professor R. H. Smith, or the still
more recently published (in any complete form) kinematic ,vork of
Professor Burmester , , ," because of the pressure of work and ill health.
20 K I N E M A T I C ::; Y N T H E S I � O F L I N K A G E S

This "graphic method" of Sn1ith's/ a very real break-through

available in English, was presented in 1885 in a paper before the Royal
Society of Edinbnrgh. An English t-ext by Goodman 1 has this to say
on page 142:

Force and reciprocal diagrams are in common use by engineers

for finding the forces acting on various members of a structure,
but it is rare to find such diagrams used for finding the velocities
of points and bars in mechanisms. We are indebted to Professor
R. H. Smith for the n1ethod. (For fuller de.tails, readers should
refer to his o,vn treatise on the subject.)

This reference is to Smith's "Graphics," published in 1889. 'The remarks

by l(ennedy and Goodman pertain to the method of velocity and accel­
eration polygons (or diagrams), today someti1nes erroneously called the
method of relative velocities (or accelerations). Smith also d,velt on
the utility of the velocity and acceleration in1ages. The great usefulness
of the velocity and acceleration polygons (and the images) lies in the
fact that even complicated mechanis1ns may be studied by simple,
graphical means; analytical methods applied to such mechanisms may
produce equations unsuited to hand calculations.
Burmester," :.\Iehmke, and .:\Iohr had already introduced these
concepts in Germany, 2 hut Smith's was an independent, ahnost simul­
taneous discovery, available to an unresponsive :English-speaking ,vorld.
Regrettably, Smith took his discovery uo further. In fact, I�ennedy
and Smith, both destined to live many years, applied their energies to
tasks other than kinen1atic ones. And neither did their countrymen
pursue Smith's fundamental idea. We note that Goodman referred only
to velocity diagrams, and these he used sparingly. For acceleration
studies-such as they "·ere-he continued ,vith the graphical differentia�
tion ,vhich Smith had called "practically useless." I'he Germans, 011
the other hand, equally oblivious of Smith (he is not mentioned in their
literature), continued to build upon the Burmester-l\tlehmke-1\i[ohr
foundation to fashion a very solid and distinguished edifice.
1\ctivity in A.merica ,vas static, for the ,vork was principally
based on Kennedy and even older authors such as Rankine, ,vith some
lip service to Reuleaux in a vague ,vay. A "mechanisms" course ,vas
tantamount to further exercise in mechanical drawing, for there ,vas little
1
"Mechanics Applied to Engineering," vol. I, 9th ed., 1926, reprinted 1943,
59 pages on mechanisn1s.
2 We do not n1ean to belittle the German work by not dwelling on it; it i.S

sin1ply that the English effort is so little known or appreciated that we feel it should
receive attention here.
.-\ � O lT T L I N E OF K I N E M A T I C S T O 1 900 21

in the \vay of kinematic analysis. Indeed, the teaching of acceleration

analysis in the United States may be said to have waited until the 1930s-.
This is very curious, for since the tools were available in the English
language, the language barrier cannot be invoked as the deterrent. An
accounting \Vas called for in 1942 by Professor De .Jonge ; 1 his critical
remarks set the stage for the ne"· interest and effort that ,vere to develop
in America.
We have seen lleuleaux's early concern ,vith synthesis. His
"-'Ork, s01netimes the subject of bitter argun1ents, ,vas not forgotten,
and the Gern1ans subjected t.he kinematics of mechanisms to 1nuch study.
Concepts far heyond those of Reuleaux ,vere added to the picture he had
attempted to paint; the ,vhole is no,v called synthesis of linkages.
The spirit of the ,vord synthesis implies that it is the combining
of parts to produce a ,\·holf-\; it is design, the creation of something ne"' ·
l{inematically, it is the co11 version of a motion idea into hard,vare. In
this sense, synthesis is the opposite of analysis, in which existing mecha­
nisms are examined for ,vhat they do or can cause to happen. However,
analysis and synthesis, although opposites in the broad sense, are closely
related, for the observations of analysis often serve as springboards for
synthesis.
The ,vhole problem of synthesis is taken to consist of three facets.
For purposes of discussion they may be separated, even though there
are varying degree5 of interrelation. In the course of a synthesis or a
design of a motion-converting device or mechanism, it is necessary to
reach three fundamental decisions, each kno,vn as a kind of synthesis.
rfhe first is type synthesis. flere the type of mechanism- cam,
linkage, elliptical gear train, and so on-is decided upon. It is ,vith this
Phase of synthesis that the name of Reuleaux is peculiarly associated.
The second is nuniber synthesis. In this the requisite numbers of
links and 1o,ver or higher pairs yielding constrained n1otion are estab­
lished. This is a case of guaranteeing that a mechanism has just the
right nun1ber of parts properly connected to ensure that it will "run"
as desired. Criteria of n1ovability ,vere presented by Gri.ibler,h ,vho
dealt ,vith planar situations in 1883 and 188;'>! coming to the spatial
chains in 19 1 7.
The third and last kind is diniensional synthesis, or the determi­
nation of the proportions (lengths) of the links needed to accomplish
the specified motion transformation. In Germany, Burmester was in
accord ,vith Ileuleaux's fundamental concepts and most of his nomen­
clature and definitions. }laking extensive use of mathematical prin­
ciples (mostly geon1etrical), and considering displacen1ent, velocity, and
1
A. E. R. de Jonge, \Vhat Is \Vrong with ''Kinematics" and ''.Mechanisms"?
-'tech. Eng. , vol. 64, pp. 273- 278, 747-751, April, 1942. ..
22 K I N E M A T J C S Y N T H E S I S OF L I N KA G E S

acceleration, Burmester's "Lehrbuch der Kinematik" (1888) developed

geometric methods that furthered analysis and showed the ,vay to
synthesis. The Russian mathematician Chebyshevb made signal con­
tributions to synthesis, one of ,vhich is recognized by his sharing the
theorem of the triple generation of coupler curves ,vith Samuel Roberts;
another is the four-bar linkage bearing his name. It is interesting to
note that the ,videly used Chebyshev polynomials for approximating a
desired curve ,vere developed from a concern ,vith the Watt linkage.
Our revie,,, has come to the end of the nineteenth century. This
marks a logical terminus for our survey : it is the end of the formative
period, of an era in "·hich many principles had been uncovered, analytical
methods established, and the road to synthesis opened. Subsequent
growth of these ideas, no longer nurtured only in Europe, has been of
such scope that no short overvie,v is possible. Indeed, part of the mis­
sion of this book is to discuss certain of the advances, especially those
relating to synthesis.
The ful1 story of the kinematics of mechanisms, doing justice to
the many ,vho practiced the art of mechanisms and contributed to the
science of kinematics, is yet to be written.
Provocative and ilJuminating studies, supported by extensive
bibliographies, have been made by De Jonge, 1 Freudenstein, 2 and
Ferguson. :i

BIOGRAPHICAL NOTES
Andre ?\farie Arnpere ( 1775--1836) , professor of mathematical analysis and
n1echanics at the f::cole Polytechnique, gained distinction in establishing the
relation between electricity and magnetism. His mathematical theory not only
explained existing phenomena but also predicted many new ones. A theorist
who was also an experimentalist, Ampere was honored in having the unit of
electric current strength named after himself.

Ludwig Burmester ( 1840- 1927 ) , professor of descriptive geometry and

kinematics at }Iunich, wrote on perspective, optics, groundwater flow, and other
topics that lent themselves to geometric investigation. His extraordinary
"Lehrbuch der Kinernatik" (1888) went to almost a thousand pages to form a
systematic and comprehensive treatment of theoretieal kinen1atics. It correlated
1 A. E. R. de Jonge, A Brief Account of l\ilodern Kinematics, Trans. ASJ.fE,
vol. 65, pp. 663-683, August, 1943n.
2 F. Freudenstein, Trends in the Kinematir.s of :\1echanisms, Appl. 1lftch.

Revs., vol. 12, pp. 587-590, September, 19.59.

3
E. S. Fergueon, Kinematics of l\:lecnhanisms frotn the Tilne of Watt, U.S.
1Vatl. Nluseum Bull. 228, Paper 27, pp. 185-230, 1962 (obtainable from the Super­
intendent of Documents, Government Printing Office, \\'ashington, D.C.).
 AN OUTLI�E OF K I �E M AT I C S TO 1 900 23
the known mathematical methods with Burmester's own brilliant additions, and
is still a classic.

The father of the thern1odynan1icist Sadi, L. X . .:\I. Carnot (1753-1823)

was a figure in the French ReYolution: he perforn1ed the gigantic task of raising
14 annies when the power of united Europe was launched against France in
1793. Carnot aided �Ionge in the forn1ation of the Ecole Polytechnique and
was a noteworthy contributor to 1nechanics, the art of fortification, and mathe­
matics, especially calculus and projective geometry.

Both .:\Iiehel Chasles (1793-1880) and Louis Poinsot (1777-1859) were

�raduates of the Ecole Polytechnique. They shared with .:\'loebius the most
important nineteenth-century advances made in geometrical mechanics. Poinsot
was the first to introduce the idea of a couple into mechanics. Chasles's history
of geometry is still a standard work. A collector of autographs, Chasles achieved
another distinction by buying over 27,000 "original letters"-among them 622
H •
written" by Xewton and one from Cleopatra to Julius Caesar- from a clever
forger before haling him into court.

Pafnutij Chebyshev (1821-1894), professor of mathematics at St. Peters­

burg, was one of the most important mathematicians of the nineteenth century.,
\Vithin a wide range of interests he made significant contributions to the theories
of probability, congruences and integrals, quadratic forms, prime nun1bers, the
construction of maps, and other topics. Chebyshev's consuming interest in
kinen1atics produced numerous developments in linkages, especially in their
synthesis: he is revered in Russia as the founder of modern kinematics.
(Variations in the transliteration of the name Chebyshev stem from the
nationality of the translator: one finds Tschebytschew, Tchebycheff, and Cheb�·­
shev, among others. \Ve use the Library of Congress Yersion.)

Leonhard Euler (1707-1783), versatile Swiss genius and favorite pupil of

Johann Bernoulli, received when nineteen a prize from the Paris Academy of
Sciences for a dissertation on the masting of ships. His astonishingly productiYe
�ife was spent abroad as professor of mathematics at the Academies of Science
1 n Russia ( 1727-1741 and li66-1783) and Berlin (1741-1766). He lost the sight
of one eye at the age of twenty-eight, becoming totally blind 17 years before
death.
Euler left his rnark upon a great many physi<·al and n1athematical prob­
lems. He greatly enriched rnathematics in an endless number of ways, made an
e�tensive study of optics, and, li,·ing in a period when theology and physics were
s�l)l greatly intermingled. indulged in metaphysical an<l philosophical specula­
tions. Our great debt to Euler lies in his treat1nent of n1echanics: he laid the
foundations (built upon so brilliantly by Lagrange) of analytical ("algebraical'')
mechanics. Newton's mechanics are purely geometrical, theorems being deduced
from initial assumptions by geometrical methods; Euler applied algebra to
geometry, producing the conditions of a proposition or figure.
24 K I N EMATIC SYNTHESIS O F L I N K A G E S

:.lartin Griibler ( 1 851-1935) taught at the techniC'al unh·ersities of Zurich,

Riga, and Berlin before succeeding 0. :\Iohr in the C'hair of t<>chnical rnechanics
at Dresden in 1 900. His interests in n1cchanics were varied, and included, in
addition to kinematics, stress analysis of rotating disks, turbine theor�,, structures,
a.end aerodynamics.

Hero of Alexandria is placed in the first century A.n. because of his n1ene­
tion of an astronomic·al event now ealculatect to have occurred in A.n. 62. Hero's
works became known to Europeans through an Arabian translation that trans­
mitted illustrations. 11uch of what Hero reported on was appreciably older
work, and there is no telling whi<:h parts of the works are his. It is supposed that
Hero was probably an Egyptian because of his strong bias toward application,
although his mathematical powers were ('onsicterable.

Alexander Blackie \\'illian1 KennPd!J ( 1 847-1928) was an engineer of

,nany talents. A graduate in mining, he designed compound stean1 engines for
early twin-screw ships and taught mechanics at a number of institutions before
being appointed professor of engineering at Fniversity College (London) in
1874, whose first engineering laboratory he established. With respect to the
theoren1 of three instantaneous centers he recognized Aronhold's priority but
claimed independent disC'overy. He was knighted in 1905 and became a eon­
suiting electrical engineer, holding important government positions during \Vorl<l
War I . Interests different from engineering are reflected in his ''Ypres to Verdun"
(1921) and "Petra : Its History and :i\fonuments" (1925). The latter, a definiti,·e
monograph on an ancient caravan trade center important in the time of Christ,
stemmed from explorations undertaken at the age of seventy-five.

Jacob Leupold (1674- 1 727) wrote a nine-volume series of books which

('onstituted a "display of the mechanical sciences" and which were published
during the period 1724-1739. In all, they contained some 492 copper engravings
detailing the early eighteenth-century practice in machines for many purposes,
instruments, water supply, and bridge building. As design texts, these books
were directed at "artists, craftsmen and such people who have no knowledge of
languages or other subjects, and therefore have no opportunity to . . . search
out. . . . the information from many- sources . . . needed to n1aintain, build
and use such machines" (as those displayed).

Gaspard J1onge (1746-1818) is best known to engineers as the inventor of

descriptive geometry. Not an engineer, but an inspiring tea.cher well acquainted
with the engineering science and craftsmanship of his tin1e, l\'longe developed a
rational system of drawing, to aid in the planning of fortifications and the setting
out of stonecutting and carpentry, that was treated as a military secret for nearly
thirty years, until 1794. l\,fonge was one of the principal founders of the ::f:cole
Polytechnique, the first polytechnical school in Europe, and he included Lagrange
and Fourier among its teachers. For a short time minister of marine under
Napoleon I , lVIonge was the only supporter of Robert Fulton's submarine venture.
Together with Berthollet and Fourier, }fonge directed the geodesic survey and
A X O U T L I N E O F K I N E M A T I C S T O } 900 25

description of Egypt when attached to Napoleon's Egyptian expedition. He

contributed much to the superior technical education that enabled the France
of after the Revolution to catch up with Britain in industry. As an analyst,
:'.\fonge belonged to the hriJ?;ht perio<l of Frenr.h mnthemnf.iP.i:; further illuminated
by Laplace and Legendre.

William .John l\IacQuorn Rankine (1 820-18i2), educated at Edinburgh

as a physicist, occupied the chair of civil engineering and mechanics at the Uni­
versity of Glasgow. One of the pioneers of modern thermodynamics, Rankine
wrote the first formal treatise on the subject. Rankine's short life was marked bv
prodigious energy; his works dealt with civil engineering, shipbuilding, machinery
and millwork, steam engines, and applied mechanics. His "Applied Mechanics"
was the first text in English by that name; it had a useful life of half a century
�nd is still worth looking at if for no other reason than the philosophy of the
introd uction, which someone has condensed to "Theory is the distilled essence of
practice." Rankine presented the theory of his day in such a way that it became
practice.

F'ranz Reuleaux (1829-1905), German-born, was named professor at the

Polytechnicun1 in Zurich in 1856. Here he developed machine design and, with
perceptive awareness, defined concepts fundamental to a studv of kinematics.
F'rom 1864 until his retirement in 1895 he was in Berlin: he hel�ed establish the
Technische Hochschu1e-Charlottenburg while making significant contributions
in all areas of mechanical engineering, the en1phasis of his efforts being laid on
design work and kinematics. Reuleaux traveled widely as his government's
representative to five world exhibitions during the period 1867-1881. His sharp
criticisn1s of the poor quality of his country's industrial products shook the
German economy. The creation of a uniform patent law (1877) for all the 21
states comprising the German Empire was the result of his efforts. He has been
�haracterized as the last "poly"-technologist because of inspirational pioneering
in many fields: he led in replacing empiricism with rationalism.

Richard Roberts (1789-1864), engineer of �Ianchester, began his career as

a patternmaker under John Wilkinson, the ironmaster whose cylinder boring mil1
was so essential to the success of the \Vatt steam engine. Roberts was responsible
for many innovations and inventionse; among the latter are the back-geared lathe
and planing mill (1817). Possessed of no worldly prudence and little business
acumen, he died in financial straits despite his many successe8.

Samuel Roberts (1827- 1913), F.R.S., was a barrister who turned to mathe­
matics and thus enriched the study of algebraic curves. He belonged to the
�eneration of mathematicians who investigated "link motion " ; their number
included Arthur Cayley, A. B. Ke1npe, Linguine, and Koenigs. Roberts, fond
of chess, angling, and philosophy, indulged in lathe work (a fashionable hobby
t�en called "turning") and liked to construct electrical machinery and other
Pieces of scientific apparatus. He is not to be confused with the earlier Roberts,
namely, Richard, of �lanchester.
26 K I N E M A T I C S Y N 'l' H E S I S O F L I N K A G E S

Robert Henry Smith (1825-1916), born and educated in Edinburgh,

assumed the post of organizer and professor at the Imperial University of Japap
at the age of twenty-three. He was one of the group of foreign scholars of the
1870s whose mission was to introduce modern scienC'e and establish an educational
system. His contributions are still warmly remembered by the Japanese. In
1880 he became the first professor of engineering (civil, n1echanical, and electrical)
at }Jason College, the earliest antecedent of the present Birmingham l:niversity.

Pollio 2\Iarcus VitrutJius lived in the first century a.c., functioning as archi­
tect and engineer under Octavianus, the successor to Julius Caesar. Vitruvius·
fame rests on a single book, "De Architectura," a comprehensive treatise impor­
tant to the history of art and technology because of its reporting of classical
antiquity. Architecture, to Vitruvius, included astronomy, horology, and
mechanical and military engineering.
,vith the decline of Roman power, a Latin book was for centuries of no
interest to comatose Europe. "De Architectura" was also neglected by the
Arabs. Slightly known to the medievalists, it was lifted to glory by Renaissance
scholars. A first printed edition appeared in 1486; later editors, starting in
151 1 , added their ideas of illustrations, the original figures not having survived
even in n1anuscript form.

James lVatt (1736-1819), F.R.S., someti1ne "mathematical instrument

maker to the University" of Glasgow, led an uncertain life in his earlier years,
adding to his income by making surveys and reports in connection with canal,
river, and harbor works. The separate condenser, patented in 1769 and added
to current steam engines, more than doubled the thermal efficiency; this, with
other i1nprovements, made power available to the drowned mines of coal-scarce
regions. 1'he windmill flyball governor (for regulating the gap between mill­
stones) was adapted by Watt as an engine speed regulator, giving the first
closed-loop servomechanism. "\,Vatt put numbers to the concept of horsepower
and invented the engine indicator and a letter-copying device based on the ink­
transfer process. His insistence on the use of the slide rule in his design rooms
did much to popularize what for a long tirne was known as the "Soho sliding rule."

Robert fVillis (1800-1875), engineer and archaeologist, was an ordained

deacon and priest, as were most of his academic colleagues. He served as
Jacksonian professor of natural and experimental philosophy (physics) at the
University of Cambridge from 183i until his death. ,villis' "Principles of
l\'Iechanism" (1841) was the most complete work on the subject which had yet
appeared : he was the first to present gear theory as part of a systematic treatment,
advorated the involute form with a pressure angle of 14-}0, and invented and
named the first odontograph.
Elected Fellow of the Royal Society for work done in physiology (mecha­
nism of the larynx, vowel sounds), the gifted Willis continued and extended his
Cambridge predecessor's lectures on "mechanical philosophy."
Willis' reputation as an engineer continued to grow, and touched upon
other fields, among them structures and astronomy. He accomplished much
AN OUTLINE O F KINEMATICS T O 1 900 27
original research in architectural histor�·, becoming an authority on medieval
Latin and the construction of ancient buildings.

BIBLIOGRAPHY

Armytage, W. H. G.: "A Social History of Engineering," Faber & Faber, Ltd.,
London, 1961.
Derry, T. K., and Trevor I. \Villiams: "A Short History of Technology," Oxford
University Press, New York. 1960.
Finch, James Kip: "The Story of Engineering," Anchor Book A 214, Doubleday
& Company, Inc., Garden Cit�·. New York, 1960.
Usher, Abbott Payson: "A History of .:\lechanical Inventions," rev. ed., Harvard
Unh•ersity Press, Cambridge, :\lass., 1 9.54.
White, Lynn, Jr.: ").fedieval Technology and Social Change," Oxford tJniversity
Press. Xew York, 1962.
 CONCEPTS
AND NOTATIONS
RELATED TO MECHANISMS

2-1 M E C H A N I S 1\1 S
A mechanism is a device to transform one motion into another.
If the device also transmits substantial forces, it is a machine,
which means that all machines are mechanisms in spirit. If
forces are associated ,vith the conversion of the energy of
high-temperature fluids (as steam or gas) to shaft power, then
the aggregate rnay be called an engine. At any rate, it is recog­
nized that the parts comprising the device-mechanism, machine,
or engine-must be resistant to deformation, i.e., the parts must
approximate rigid bodies. We may then say that a mechanism
is an assemblage of rigid or resistant bodies connected together
for the purpose of transforming motion.

·rwo general groups of mechanisms exist, since the motion may

go from either uniform to uniform or from uniform to nonuni­
form. Circular gears, chains, belts, and the like, comprise most
of the uniform motion converters; their many proble1ns will not
be discussed in this book. Nonuniform conversions are made
with noncircular gears, ca1ns, ratchets, and linkages, both planar
 CONCEPTS :\ �I) :-lOT.\TION'S RE'L,\TED TO MEC.H'.AN.ISMS 29
and spatial. This book will direct itself exclusively to the design of
link&ges to meet certain motion-conversion specifications.

2-2 CLASSIFICATION SYSTE'.\IS

Classification implies s01ne sort of systematic grouping of factors
that ,vill sho,v the relation of one thing to a group of things: it is a search
for order. A complete system of mechanisms ,vould sho,v the genealogy
of motion; i.e., it w·ould sho,v relations and lines of common descent
,vhen they exist. Individual mechanisms \VOtild no longer be separate
mysteries; each ,vould be part of a larger group having some,vhat similar
�haracteristics modified by personal idiosyncrasies. Further1nore, an
important property of a classification system ,vould be the aid it could
furnish a designer in finding the forms and arrangements best suited to
satisfying certain specifications. Thus far no completely unified and
general classification scheme for all mechanisms has been found, although
several attempts have been made.
In the '\\'Ork to date, two groups may be recognized: ,ve distinguish
between the functional and the structural. The functional classification
system considers the co1nplete mechanism needed to transform a given
motion into another, as the conversion of uniform rotation into reciproca­
tion. The complete mechanism is like a "black box," with provision
for accepting one kind of motion at the input and producing another
kind of motion at the output. Since it is in the nature of things to be
able to accomplish a given task in n1ore than one ,vay, there would be a
number of black boxes to choose from-their "insides," or "works,"
availability, and cost ,vould be different, but the overall effect, i.e., motion
transformation, ,vould be the same. :\Ionge's scheme belongs to this
group. "fhe shortcoming here is that no general principles exist to guide
the problem of transforming a given n1otion into another. All that can
be done is to run do,vn a list of complete mechanisms that ,vill do the
specified job and choose from among the several the one that best fits
�dditional specifications of available space and manufacture. There
is really no unity, and the system is a rnere collection of mechanisms.
The second, or structural, classification system deals v..-ith the
nature of the parts, considering them from the standpoint of their relative
motions. Willis followed this idea by regarding how the motion trans­
formation between input and output members was achieved-as by
rolling or sliding contacts, flexible connectors, barlike links, or tackle.
Reuleaux was much more intimate: he considered, not the big input­
output span, but only the immediate connection between parts. Here the
8? apes of the surfaces in contact-the ,vorking surfaces-impose par­
ticular and unique 1notion restrictions; i.e., they allo,v only a particular
type olmative motion, .. rotation about• oin�l� uig,""'tilineor trana­
lation,rotatioo about a poiot.,etc,
l'.von th,.1ghon011�oofth•kno•·noc!mficationo,yrtemooan1w,,.,.
allonee<lo,eacho•yotemohaoooomemeritotlio.tocan borxploited ...ilhopro6t
forparticularoare,aoolinlerM.

Z..J Jil(lll> A'IJ> li�IIIIITA'IT DODIF.11

A ri9id body i• a material body i11 �-hich the diotance � ...,.n
anytwopoi11U1iooin,·ariable. Thu11,ifA,H.andCa,eoth1ttpoinuola
rigid body {F�. 2-l), the three di.tan""" All, IW. AC �n =notant
no mallet how \be ri1id bodyii movodo. u fro,n pooition I to poaitioo 1
Jfoth,.,.1>01>colli11.,.,point.1A,ll,Co11oaorigidbodya,eknow",•nyother
point D maybe identifiedbyopecifyi111o1M th1tt di•tanooo DA, DB, lX'
Evoooiltl,eopoin1n1;.,.001aideotheo1ppare,ithouudui.. ofothe�d
body.itm.o.yotil!bo..,...iden,d uapointolothebodyulollgutl,etl,..,.
diata"""'" DA. DB, and OC IN! invariable, •hen the rigidbodymov.,.
fromopoaitfon I l<lpooition2,pointllmov••·itl,i<.
Whenoaorigidbodyi.omo,·e<llromoanoinitial pooition I to a final
pooait.ioo2,i1ill&.idotoohavebe<,nogiveu11dioplatt1nenl. Minotheeue
of a point , theodilplacun,ent ofoa r�idbodyod epend,onlyoottheoinitial
and6oa!pooitK>Dllaad i,oiudependent olotbe11·ayioo•·hi<hitlaoearr..d
oot;i.e.,thea,ualpathofollo,.·e<ldurinJr;thoodioplaceme111ioool,,oocoo,­
""'l""n.,._ Ma rigid body mo,-,. from &o initial to a 60&! pooitior,,
eachoolltopoiotoot.....,&1>1>rhcularpath;&t..,hoin,tantooltin>eapoint
huaoparti<:ularoveiotityandapartieularoacttleration
Tobe honeet al>oul ii,there;. o<> ..,�h thin� M & ci!lidbody, aim,,
all k,.<>ft"!I inaterialo deform ,,,.n,n•·hat. under any ,u-,_ 1,0 ln&\Wr lw�
CONCEPTS AND NOTATIONS RELATED TO MECHANISMS 31

small. If we assume, as we must at least in the beginning, that the

geometry of the body does not change as it moves-that any t,vo points
ahvays bear the same relationship to each other-then we use "rigid"
in the mathematician's sense, overlooking the inevitable small deforma­
tions that are held to acceptable values by proper cross sections of the
bodies. Realistically, we might better speak of resistant bodies, bodies
so proportioned that their deformations are acceptably small. Under
resistant bodies we may also include belts, chains, and hydraulic lines,
for these have a one-way resistance to deformation. As a matter of fact,
elastic deformations may become troublesome ,vhen machinery operates
at high speed, but they cannot be established until the kinematic work
is finished and the parts have acquired mass by having been given physi­
cal shapes and cross sections.

2-4 A-lOTION-RELATIVE AND ABSOLUTE

rfhe very fact that motion exists implies reference frames of some
sort on the moving parts. If the reference frame of one machine part
moves with respect to the reference frame of another, we speak of relative
motion. When the reference frame of one part is fixed with respect
to the earth and the motions of the other parts are referred to it, then
these particular relative motions are termed absolute motions: absolute
motion is thus a special case of relative motion, the case in ,vhich the
absolute-motion reference frame has no motion.o1 Of course, any rigid
body may be chosen· as a reference to ,vhich the motion of other bodies
is referred; the operation of an aircraft landing gear is better referred to
the fuselage than to the distant earth.

2-5 CONNECTIONS, A N D TH E PARTICULAR l\fOT IONS

THEY PER!\IJT
.\ mechanism has been defined as a number of rigid bodies so
connected that each moves with respect to another. The clue to the
nature of a mechanism lies in how the parts are connected and ,vhat kind
?f relative motion the connection allows. In kinematics, a connection
ts a joint between two members permitting a particular kind of motion.
We should note that the term connection, in fields different fro1n kine­
matics, may n1ean an immovable connection, as a structural joint or
splice, shrink fit, and the like.
Considerations based on the relative motions permitted by various
mechanical connectors lead to a recognition of three broad classes of
1
That the earth itself is in rnotion around the sun has no bearing on any
problems to be discussed here.
32 KINEMATIC SYNTHESIS OF LINKAGES

Reference
line
•
1

""'""' "'t-s
"""""'"'""'"'""''�1O
Reference
line

AA,.{,..�'-.�'-'Wii.���,�����,�,��,�w...,w...��
1
FIGl'RE 22
- Conventions for showing pair variables.

connectors. 'l' hese are the lower-pair connectors, the higher-pair con­
nectors, and the ,vrapping connectors. The last are comprised of belts
and chains ,vith their one-,vay rigidity and will not be discussed here.

2-6 LOWER-PAIR CONNECTORS

Consider t,vo Jinks in parallel planes, connected in such a ,vay that

one may turn with respect to the other. The upper sketches in Fig. 2-2
show the two links in a conventional ,vay: the small circle denotes the.
ability to turn, and the center of the circle represents the axis of rotation
about ,vhich the angle of rotation is measured. Xothing has been indi­
cated about the physical make-up of the joint: it could be made in many
,vays, as using a simple pin, a ball or roller bearing, etc. Ho,vever, no
matter how constructed, an angle such as 0 is ahvays the definitive vari­
able of motion and ,vill be called the motion variable of such a joint.
1\nother situation is that of a crosshead in its guide. Here only
(rectilinear) translation is possible, and the relative motion is described
by means of a linear displacen1ent such ass measured from some conven­
ient origin. The actual form of the crosshead and guide sections, as
laid out by the machine designer, may take any of a large number of
shapes, having in common the property of restricting the relative motion
to translation only. No matter what the form, the relative motion is
described by a single variable such as s.
With the intuitive background of these t,vo exarnples ,ve may
make a systematic approach to the problems of movable connections.
Consider a set of x, y, and z axes fixed to ground link 1 (Fig. 2-3). Link 2,
f':OXC EPTS A N D K O T -\.
. TIONS R E L A T E D T O M E C HA N I SMS 33

which is somehow suspended in space, also carries a set of rectangular

axes u, v, and w, with origin at A. 1'he two sets of axes are parallel.
If link 2 is moved to any other location, we could describe this by saying
that A's ne,v location is at x2 , y 2 , z2 : each axial displacement of A , that is,
.r2 - x, = Ax, y2 - y 1 = Ay, and z 2 - z 1 = '1.z, represents a motion of
translation. 'fhe vector sun1 of these displacements, Ax + '1.y + Az = As,
is the real displacement of A. It is convenient to keep track of the
motion of A by means of three orthogonal motions (translations), which
•nay be taken in any order.
In addition to the motion of A , there may aiso be a motion about
A, as when link 2 has turned about the point A. This turning is con­
veniently described by separate rotations about each of the axes u, v,
and w, or, ,vhat amounts to the same thing, about the x, y, and z axes.
The sequence of rotations is important for spatial mechanisms. With
regard to the motion about A, this may come either before or after the
�onsideration of the motion of A ; the order of these t\\'O operations is
immaterial.
We see from the foregoing that con1pletely to define the position
of a link requires the knowledge of six variables, three giving the transla­
tion of a point, and three giving the rotation of the link about that point.
Each of these motion variables is also said to be associated ·with a degree
of freedom, i.e., each is identified ,vith a motion of either translation or
rotation. To form a clearer picture of all this, ,ve shall examine in detail
the possible motions of link 2 ,vith respect to link 1 by observing the
nature of the connection :
1. If link 2 is permitted only a rotation about its w axis (the other
z

V
I
i
Zt
1

FIGURE 2-3 Reference frames for moving link.

34 K I N EM AT I C S Y N T H E S I S O 1'' L l N K A G ES

five possible motions being suppressed), a variable sufficient to describe

the relative motion is an angle 0, measured in a plane perpendicular to
the w axis. Following Reuleaux, ,ve designate this connection as a
revolute and give it the symbol R. The degree of freedon1 f of this con­
nection is expressed by f = I .
2. Were only translation along the w axis pern1itted (again with
the other five possible motions suppressed), the tw·o links ,vould remain
parallel to each other and the variable describing the relative motion
"vould be the perpendicular distance, say s, betv,een planes :ry and uu.
This type of motion, a rectilinear translation, is commonly associated
with a crosshead and its guide. It is a prismatic connection v,,ith the
symbol P, and degree of freedom f = 1 .
3. Were both rotation about and translation along the w axis
permitted, t,vo independent variables- one for the translation, the other
for the rotation-would be needed to describe the relative motion. Such
a connection is called cylindric: it is the motion of a shaft in a journal
bearing if there is no axial restraint. In symbolic notation ,ve ·write C ;
the degree of freedom f = 2.
4. Suppose the w axis to be threaded, as a bolt, and the corner A
tapped as a nut. As link 2 turned, it ,vould remain parallel to link 1 ,
although undergoing a translation along the w axis. Since the angle of
rotation (J and the translation s are related by the (constant) lead L of
the scre,v, there is but one variable for the relative motion of the links.
For the scre,v connection ,ve ,vrite SL ; the degree of freedom is f = 1 .
,5. Suppose that link 2 lay directly on the xy plane of link 1 (ze1 = 0)
and were allov,ed to slide on that plane. We would then recognize three
possible motions-two translations, and a rotation about the w axis.
Such a planar connection (it occurs rarely) would have the symbol F
(think of flat) and of course has three variables, ,vhich means that the
degree of freedom f = 3.
6. To suppose again, assume that there is a ball-and-socket joint
at A , thus joining links 1 and 2 with a spheric connection. \Ve imme­
diately recognize the complete suppression of any linear 1notions: the
only possible motion of link 2 with respect to link 1 is spherical 1notion,
,vhich is to say that all points of link 2 move in concentric spheres referred
to the center of the ball. The n1otion is best described by successive
rotations about the three mutually perpendicular axes: the sequence of
the rotations is important. This connection has three variables, \\'hence
f = :3. The symbol ,vill be G (for globular).
1�he foregoing six types of connections, when reduced to simple
forms of construction, are sho,vn in Fig. 2-4. The common denominator
of these connections appears to be the area contact bet,veen links. Each
of the identical surfaces of contact, the ,vorking surfaces, is called an
 CONCEP TS A N D NOTATIONS RELATED TO M E CHANISMS 35
eleinent: taken together, the two elements constitute a pair, one element
lying on one link, the second element lying on the second link. It is
apparent that the relative motion of the t'wo links is the relative motion
of the pair elements and that this relative motion is defined by the vari­
able of the connection, ,vhich ,ve now call the pair variable.
The particular forms chosen for the screw, revolute, and prismatic
pairs illustrated in Fig. 2-4 follow Reuleaux's suggestion that the revolute
and prismatic pairs may be considered as special limiting cases of the
screw pair, with the lead either zero or infinity. This observation will
be put to use in designing a more complete sy1nbolic notation to describe
mechanisms in ,vhich all connections are made by lower pairs. With SL
representing a scre,v of lead L, then the symbols for the revolute and
Prismatic pairs follo,v logically as So and S«J, i.e., sere"' pairs SL "'ith
L = 0 or L = oo , respectively.
These six element pairs, whose appearance in their simple con­
structional form is dominated by area contact, ,vere called lower pairs
b! Reuleaux ; the term came into the English language with Professor
h.ennedy's translation of 1876. It is unfortunate that this apparent
dominance of area contact has often made area contact the criterion
for lower pairs: the real concept of lower pairs lies in the particular kind
of relatil'e 1notion perrnitted the connected links; the particular motion of
each pair is defined by an obviously associated pair variable or by a
simple grouping of functionally unrelated pair variables.
A moment's reflection ,vill sho,v that identical relative motions
are possible "·ith various joint constructions having neither area contact
nor geometrically identical elements. For example, a turning connection
or revolute may be constructed with a baJI or pivot bearing, and for small
angles of rotation a knife-edge or a flexure pivot rnay even be considered:
no area contact in the first, no elements in the second. However, the
relative motion pennitted the connected parts is a rotation, defined by
an angle such as O. A. typewriter carriage moves along its ,vays on small
rollers ,vith a motion of rectilinear translation: this means a prismatic
Pair ,vithout area contact; a distance s will be the pair variable. A. ball
spline is similar. A ball bushing is a practical ,vay of constructing a
low-friction cylindric pair, just as a ball-bearing scre,v is a screw pair
of minimal friction. 1'he various constructional forms have advantages
for the machine designer, but they do not change the geometric relation
of the connected links. They are merely different ways of physically
achieving a specific kind of relative motion bet,veen the connected parts.
For all but the planar pair (symbol F) ,ve may speak of a hollow
element and a full element. 'I'hus, for the revolute pair R of Fig. 2-4a,
the bearing surface of ]ink 2 is the hollow element, ,vritten R-; and the
surface of the shaft, which lies on link 1, is the full element, written R+.
36 K I N EMATIC S Y N T H S
ESIS O F L I N KA G E S

�9
p+

s
s
2

t
�

-�
(a) (b) (c)

Zt �.

u �9
I
I
c+ I
I
I
I Y1
I
I
\
%1 \

(d) (e) (f)

FIGt:Ris �--! (a) Hevolute pa.fr (turning pair), f = 1 . The relative motion i:; rotation
about the axis and is defined by a single variable 8. (b) Prismatic pair, f = 1 . The
relative motion is translation and is defined by a single variable s. (c) Screw pair,
f = I . The relative motion is helical and is defined by either the rotation 8 or trans­
lation s related through t,,.8/21r = .is/L, where L is the lead of the screw (advance per
revolution). (d) Cylindric pair, f = 2. The relative motion is a combination of a
rotation 8 about an axis and a translation s parallel to the sarne axis; there is no rela­
tion between 8 and s. (e) Spheric pair (ball-and- socket joint), J = 3. The relative
motion is spheriral and is defined by three variables: two angles a and ef, to define the
direction 011 and the angle 8 of rotation about Ou. (j) Planar pair, f = 3. The
relative motion is planar and is defined in terms of two tran1::1lations x and y and a
rotation 0.

Since the hoUo,v and full elen1euts, ,vhen visualized as areas, not only are
geometrically identical but the hollo,v element is ""Tapped around" the
full element, the five pairs are also known as wrapping pairs. \Ve should
note that the hollo,v and full elements of the wrapping pairs may be
interchanged ,vithout affecting their relative motion. Thus, the relative
motion betv.een links I and 2 of the first five pairs of Fig. 2-4 will be the
.<

CON CEPTS A N D NOTATIONS RELATED TO MECHANISMS 37

same no matter ,vhether link 1 or link 2 is the moving link. The ,vord
wrapping must be viewed ,vith caution; inconsistent as it may seem,
usage denotes chains and belts as ,vrapping connectors, but not as wrap­
ping pairs.
i\Iachine parts as actually constructed may have interrupted,
or noncontinuous, elements (Fig. 2-5a and b). This feature does not
�hange the character of the motion but is useful to the machine designer
1n distributing loads and stresses.
The same machine part may also carry elements of different,
pairs along the same axis (Fig. 2-5c). In the figure links 3 and 4 connect
to link 2; each is a separate connection a nd must be so treated. There
are two distinct coaxial revolutes, R3 and R4. The full elements Ra+
and R4+ both lie on link 2, and the pair variables Oa and 84 are measured
from the same reference line of link 2. The difference between 04 and 0:1
is the pair variable between links 4 and 3.
For mechanical convenience an intermediate pin (Fig. 2-t,d) is
generally used to fashion a revolute connection. A.Ithough the actual

'

., /

(a) Same element R - {b) Same element R +

Referenc::.::e:-.-:--­
(J◄ line of 4
�-
--- . \O �e\- 2

Intermediary

(c)
.,_ -
•·1G uaic- :>,__v
:; rnterrupted and coax1a· I eIements.
38 K I N E M A T I C S Y N T H E S I S O F L I N KA GES

number of elements has been augmented by the introduction of the

intermediary, the intermediary does not have to be counted if its own
motion is not of interest, since its presence does not affect the relative
motion between links 2 and 3.

2-7 H I G HER-PAIR CON X ECTO RS

We have seen that the connection between parts may be considered
in terms of pair elements, the contacting surfaces of the parts. We
identified lo,ver-pair connectors as those permitting six different kinds
of specific motions.
A. second type of connection must be described for the sake of
completeness : it is what Reuleaux called the higher-pair connection. In
the higher pairs the surface elements are so shaped that only line or poini
contacts are possible between elements. Point contact is found in ball
bearings, as ,vell as bet,veen the teeth of helical gears on nonparallel
shafts. Line contact is characteristic of cams, roller bearings, and most
gears. The relative motion of the elements of higher pairs is generallY
quite complicated. 'fhe involved functional relationship bet,veen trans·
lation and rotation allows no succinct statement, and an infinite number
of higher pairs exist. Under the circumstances, higher pairs do n<>'
follo,v a simple classification scheme, as do the lower pairs. Nor are
they conveniently described by means of symbols.
I-ligher-pair connections may on occasion be replaced by a com·
bination of lo,ver pairs, to reduce unit contact pressures and allO\\' take-uP
for wear. Thus, the pin element of link 2 riding in link 4 (Fig. 2-6) is
not a very practical construction; the same relative motion between
links 2 and 4 is retained on interposing link 3. We note that the two
degrees of freedom of the higher-pair connection (translation and rot3'
tion) are maintained ,vith the substitution of the t,vo lower pairs and
that another link has been added.

2-8 FOUR-BAR LINKAGES

A versatile example of mechanism is known as the four-bar linkage
(Fig. 2-7). It consists of four rigid members : thefran1e, or fixed member,

4 3 4

FIGURE 2-6 Substitution of two lower pairs for a higher pair;

note the additional link.
 C O N CEPTS A N D N OT AT I O N S R E L A T E D T O M E C H A N ISMS 39

I
I
I
Follower
FIGt:RE 2-i Four-bar link­
age. \Vith the link propor­
_e
'
I
' Frame
I
tions shown, a continuous
rotation of the crank is con­
verted into an oscillation of
the follower.

Which is assumed stationary, and to ,vhich are pivoted the crank and
follower, whose intermediary is aptly termed coupler. These members
are connected by four revolute pairs, R 1 , R2, Ra, R s; allo"·ing relative
rotation bet,veen adjacent members; all four revolute axes are parallel.
The ,vord linkage implies that all connections in the mechanism are
lower pairs (here they all happen to be revolutes).

1.6 1.8
\.4 �...i.--�-
.S
._-..J_
,.

I
I
I
I
I
I
I
I
II
I
I
I
\ ,

' �'. 127.46

°

' /"
,
\ QB
/
,

// 60"
,
/ \
/
/
\
�IGU RE
2-8 Four-bar
linkage- as
I
0 \
generator of ·-.9
the logarithmi
c function OA.08 == 1.000 in.
Y ""' log x over
ranges of the
specified o.< 0_4A = 3.352 in.
08 B = 3.486 in.
input vari­ 0.1 0
abl x and
� the arcs of AB = 0.846 in.
oscillation. y = 10&10 .t
40 K I N E M A T I C SYNTHESIS O F L I N KAGE$

Older usage, when the four-bar linkage ,vas known as a quadric­

crank mechanism, implied that a "crank" could or could not rotate
continuously, depending on its position in the mechanism. It will be
convenient to use the ,vord crank to designate ( 1 ) the input link, whether
or not it is able to rotate completely (continuously in the same direction),
or (2) a continuously rotating link which may or may not be the input.
A.s part of an instrument, the four-bar linkage may be used for
scale conversion. Such devices are called function generators. To
convert, for example, a linear scale into a logarithmic scale, the linkage
shown in Fig. 2-8 may be used ,vith an error which is less than 0.0037 °
for a 60° range of rotation of both crank and follower.
A point on the coupler of a four-bar linkage is called a coupler
point; and its path ,vhen the crank is rotated is kno,vn as a coupler-point
curve (or coupler curve) (Fig. 2-9), and the number of such curves is
infinite. Ho,vever, by proper choice of link proportions and coupler­
point locations-this is one of the problems of synthesis-useful curves
may be found. 1\ curve's usefulness depends (1) on the particular shape
of a segment-does it, for example, approximate a straight line or a
circular arc?-or (2) on a peculiar shape of either the ,vhole curve or
parts of it. The coupler point, because of its motion characteristic, is
now the output of the linkage. The coupler curve of the four-bar linkage

. Grid \
·,
I D

I.
' --t·'
--- -�---
1.---+t-- r- .

3 \
e<
cou\l�, -· -·
\
_..;-------�--- . --··+-�-. - -· ---- 4

�-....+;:E� -- \
. · - r --
'

1 Frame

f"IGURE 2-9 Four-bar coupler-point curves. The transparent grid is part of the ·
coupler plane, link 3. The curves are traced on the plane of link 1 , the frame.
 CON C E P T S AND N OT A T I O N S R E L A T E D TO MECH A � I S M S 41

A l
Coupler curve
"Straight"/i
C2

'· (a)

·--.
C ID
I I /
AI I ./.
I I
v /
l .,,----
FIGrR
• · · F.
?
--10
· • (a) Watt
�-
hnkage, g1v 1ng approxi-
mate straight-line motio
("Parallel motion'') at C ·
n I
(b)
. . pantographs, to give
, l /
8nuila.r
outputs at G and D . �- (b)

�s of the sixth order.s1 The classical and best-kno,vn coupler-point linkage

ssthe
� - 2 "straight-line," or "parallel-motion," linkage devised by Watt in
r'ine84 to guide the upper end of a piston rod along a good-enough straight
. This was before the invention of the planer (1817); without the
laner it was impractical to make straight surfaces 4 ft long as needed
ror
the crosshead guides of the earliest double-acting engines. Linkwork
�as simp le. Watt's linkage, in. about the original proportions, is sho,vn
Fig. 2-10. The coupler point .c traces a �urv� of figu:e-eight shape.
; .
att used only the middle portion, ,vh1ch 1n this case 1s a very good
;PProximation to a straight line. The addition of a pantograph allo,ved
�r ther exploitation of the "straight" coupler-curve segment, The Watt
1 �kage is
� currently used for axle and differential suspensions of some
igh-performance cars
.
1
Sp ecial configurations of the four-bar linkage 1nay generate cou�ler curves
of the f urth
� or second order; e.g., the coupler curves of a parallelogram IJnkage are
circles .e of
' 1 ., order 2.
2
. In referring to his 1nany inventions, Watt re1narked that this was the one
of whic h he was proudest. ·
•

42 K I N E M A T I C SYNTHESIS OF LINKAGES

Coupler
curve

A
/ Drill

---,
..

=
--·'"""-- �=- �-�--
-=-=:--·- :=:
..
_ -;::-,
__ �
--:
/ � ·-:..,--,
� -'----,�

Ground
,
.-IGt"RE2 - 1 1 Portable p()Sthole borer, or setter. (From Volmer, VDI
Ber., vol. 12, 1956.)

The Watt rotative engines retained the "great beam," or "lever,''

of the Newcomen engines, which made them just as bulky (see Figs. 1-8
and 1-9) : their utility lay in the fact that they ,vere rotative, giving
power directly to a shaft. The famous "lap" 1 engine of 1787 had a
great beam of about 1.5 ft length pinned to a connecting rod over 1 3 ft
long; the engine ,vas rated at 10 hp when running at 25 strokes per minute.
Because of the one-to-one gearset of the sun-and-planet this gave a shaft
speed of 50 rpm. Engines such as this required enormous enginehouses,
,vhose cost sometimes equaled that of the engines.
Although the direct-connected engine, now called the slider-crank
type, was explored around 1800 in association ,vith early high-pressure2
The engine was used for driving the n1achinery for lapping, or P<>lishing,
1

steel ornaments. It was taken out of service in 1858 and is now in the Science
Museum, London.
2 The adjective high is always associated with an age of developn1ent, and it

is the age that sets any number. In 1800 a high-pressure engine was one dispensing
with the vacuum, working with positive steam pressures of 2 or 3 atm, and exhaustin!t
at atmospheric pressure. Around 1860 a speed of 125 rpm was excitin11: <'nough to
evoke the mistrust of old-timers accust.on1ed t.o half that, speed.
C O N C E PT S A N D
I'll O T A T l O ::-, S H E L A 'f B I> 'f O M E.1 C H :\ N I S M S 43
engines, the slow beam engines dominated the po,ver scene for many
decades. ::\1uch effort ,vas expended in searching for less bulky linkages
that would produce straight-line guidance for piston-rod ends, for engine­
building tradition demanded vertical cylinders. Of these ,ve may men­
_
tion the side-lever engines, used principally in boats and ships, and the
�tasshopper, or Evans, linkwork of sn1all stationary engines. The latter
18 of Particul
ct·irectly obvious manner,' since
ar interest it is related to the vVatt linkage in a not
as ,ve shall see later.
1\nother example of the use of an approximately straight-line
segment is the posthole borer sho,vn in Fig. 2-1 1, where point C is a
coupler point of the four-bar linkage O,1 ABO . The path of C approxi­
JJ
�ates the straight vertical segment c,r2• This device, of German design,
18 capable
of boring a vertical hole 6 ft deep.
Coupler-point curves having segments approximating circular arcs
can be invoked
to produce linkages having a d,vell or t,vo sufficiently
complete for
many practical purposes (Fig. 2-12). The figure sho,vs a

135°

2 300•

Upper dwell

o· ° 330°
60
1 C1 3

c3�
4

Lower dwell

Ov

FtoultE
2-12 Double-dwell linkage.
44 K l �EMATIC 8 Y N THES l 8 O F LINKAGES

Articles moved

'
· -----
-·- · ·

Driving crank, 2

FIGURE 2-13 Transport mechanism. (Prom C. W. Ha.rn, E. J. Crane, and W. L.

Rogers, "Mechanics of Afachin.ery," 4th Nl_, p. 439, 111cGraw-Hill Book Company, New
York, 1958.)

coupler curve having two nearly circular segments C1C2 and C3C4 of
nearly the "same" radius. T,vo bars-a dyad-are added at the coupler
point C. The second bar, link 6, is the output link: it ,vill be at d,vell
(rest) ,vhile the coupler point is traversing the "circular" arcs C1C2 and
CaC4. It must be noted that the d,vells are only as good as the approxi­
mation bet,veen the actual curv� and circular arcs; the dwells will not
be quite complete but nevertheless ,vill be adequate for many applica­
tions. If the output oscillation of link 6 is follo,ved through the complete
excursion of C around the curve, it ,,,ill be seen to have the characteristic
of the oscillating follo"\\·er of a rise-d,vell-drop-dv,ell cam.
A.s another example of the use of a ,vhole coupler curve, we may
consider the piece of n1aterials-handling equipment shovv-n in Fig. 2-13.
Here point C on the coupler of the four-bar linkage 0,1ABOn describes
the path c as crank 2 rotates through 360e °
. This motion is communi­
cated to the transport member .1 by 1neans of parallelogram linkages;
member 5 moves horizontally for the line segment (\C2 , then drops out
of the w· ay, to reappear later at the right, rii>ing nearly vertically before
moving to the left. .--\. similar mechanisn1 is used for the film transport
in some motion-picture cameras (Fig. 2-14).

2-9 SL I D ER - CR A X K M E C HAN I S �I S
.--\.enother Yersatile linkage is the slider-crank mechanism (Fig. 2-15) ,
familiar fron1 reciprocating engines and purnps. Here the translation
C O N CEPTS A N D N O T A T I O N S REL A TED TO MECHANISMS 45
0f the piston is transforn1ed into rotation of the crank shaft, or vice versa.
.e
Like the four-bar linkage, this mechanism consists of four rigid members,
fram.e (fixed member), crank, coupler (connecting rod), and follower
(slider). It differs from the four-bar linkage only i n that one revolute
pair R4 has been replaced by a prismatic pair P4, and many of the prop­
erties and applications of the four-bar linkage may be transposed to the
slider-cran k mechanism. The slider-crank mechanism is properly a
linkage, since both revolute and prismatic pairs are lo\ver pairs.
The dimensions of the slider-crank mechanism as used in recipro­
cating engines are approximately those sho,vn i n Fig. 2-15, and the path
of the center of rotation of revolute Re3 , the ,vrist pin, usually goes through
the center of the main bearings, revolute R 1. \Vhen this is the case, the
rnechanism is called a central slider crank; it is other\\1ise an offset, or
eccentric, slider crank (see Fig. 2-16).
'fhe points of the coupler of a slider-crank mechanism are, as in

Film

•·tG t·iu:
2-14 Film fe(•d . <F,·0111 "Die H11·s­
;��ischaftliche un� Angewandte Photographi:e ,"
'' ·by Kurt 1lfichel, vol. 3, Harald Weise.
Di�
Kinematographische Kamera," p. 202,
SPring I.
er- Verlay OHG, Vieuna, Hl55.) As
46 KINEMATIC SYNTH ESIS O F L I N KA G E S

Frame

Follower
(or slider)
Coupler
1-f--(or connecting
rod)

Frame
FIGrRE 2-15 Central slider-crank mechanism, the
conventional case.

the case of the four-bar linkage, also called coupler points, and their
paths as the crank is rotated are coupler-point curves, but of the fourth
order (Fig. 2-17). As ,vith the four-bar linkage, slider-crank coupler
curves may also be put to ,vork. One application is sho,vn in F'ig. 2-18.

2-10 TRA N' S �1 1 S S I O X , D EV I ATION , AN D

P R E S S r R E A X G LES
It "'ould be useful to have a measure, criterion, or index of ho"·
,vell a mechanisn1 n1ight "run" , vhile it is still in kinematic skeleton
form on the dra,ving board. "Run" is a term that more formally means
the effectiveness "·ith ,vhich 1notion is imparted to the output link; it
implies smooth operatiou, in "·hich a maxi1num force component is
available to produce a torque or a force, ,vhatever the case might be,
in an output member. Generally speaking, torque and force are not
at all compatible ,vith only the kinematics and statics of a given situation.
As is well kno,vn, the magnitude of the dynamic forces may be several
times as large as the static forces and may in addition possess quite
different directions. Even a test on a kinematic model ,vill check out
only an approximation of the static forces and will tell nothing about the
dynamic forces.
Ho,vever, some evaluation of the state of affairs is better than

R1
" ··- _ _ _
F • - -�-- FIGURE 2-16 Offset, or eccen­
tric, slider-crank mechanism.
C O N CE P T S A N l> N O T A T I O N S R E L A· T E D T O M E C H A N I S M S 47

1 Frame

FIGURE 2-17 Slider-crank coupler-point curves. The transparent grid is part

of the coupler plane, linke�- The curYes are traced on the plane of link 1 , the frame.

none. Alt 1 defined the aptness of motion transference from the driving
link (not the input link of the mechanism) to the output link in terms
of the transmission angle: the transmission angle 'Y is the smaller angle
bet,veen the direction of the velocity difference vector vs.1 of the driving
link and the direction of the absolute velocity vector vs of the output
link, both taken at the point of connection. This is sho,vn in Fig. 2-19a
for a four-bar linkage. Since the velocity vectors are perpendicular to
their respective links, the transmission angle is also given by the angle
between link centerlines, b. Clearly the optimum value of 'Y is 90° ; the
recommended tolerance is about + 50°. A.lt 2 recognized that this
kinematically determined transmission angle does not reflect the action
of gravity or dynamic forces. Thus, in a single-cylinder piston engine
the transmission angle (,vhich is here bet,veen the connecting rod and
crank) becomes zero at the dead-center positions, requiring the dynamic
action of a flyv,heel to further the motion. Linkages with more than
four bars have peculiar difficulties.
1 H. Alt, Werkstaltstech., vol. 26, pp. 61-64, 1932.
1
H. Alt, Getriebetechnik, VD/ Tagungsheft, vol. 1, p. 197, 1953.
 48 K I NEMA T I C S Y N T H E S I S O F L I N K A GE S

Long
straw

Short
straw
t'IGIIRE 2-18 Straw packer making use of the coupler curves of a slider-
crank mechanism. The crank is link 2; the coupler is link 3, with coupler
points C1, C2, and Cl. (After Kurt Rauh, "Pral,tische Getrfrhelehre," 2d rev.
ed., vol. 1 , fig. 2!)9, Springer-Verlag OHO, Berlin, 1951.)

Another approach ,,..·as taken by A. Bock, 1 ,vho suggested working

,vith the directions of the static force and velocity at the point of con­
nection, terming the angle bet,veen the directions the deviation angle o.
The deviation angle is sho,vn in f,'ig. 2-19b; its optimun1 value is 0°.
When the driving link (in this case link 3) is a t,vo-force member,
i' + o = 90° . 'fhis relation fails when the driving link has more than
two forces acting on it, as may be seen from Fig. 2-20 (adapted from
Bock). Nerge 2 inclines to,vard Bock's vie,v, although he does not name
the angle. The pressure angle of a disk cam with roller follo,ver- the
angle between the common normal at the point of contact and the
follower motion of the roller center-is recognized as the deviation
angle (Fig. 2-2 I).
1 A. Bock, VDI Ber., vol. 29, p . 158, 1958.
2
G. �erge, V DI Ber., vol. 29, p. 157, 1958.

,,/· (vu)
;('.
"'t/
/ o _..
/4
__.. .:.-t -
-y

(b)

2-1!1 (a) Trans1nission nnJ?;len,, of a four-bar linkage; (b) transmission angle

i·1ct·Ri-; 1
and deviation angle 6 of a four-bar linkage.
C ONCEPTS A N D NOTAT IONS R E LATED TO M E C' H .-\ X I SMS 49
Second output
<vcB > C 6

Second input
)
__.!,,=JI•==- --. - � - {Ve)

First output --...J

First input
_____
'
\
,

\I >
'

FIGURE 2-20 Transn1ission 1

;;:
and deviation angles of a
six-link mechanism.

2-1 1 PLANAR AND S PATI AL :

T H E l\.l OTIONS AND THE l\•I ECHANISMS

Planar and spatial 1notions of bodies are distinguished from each

other by noting the motions of all particles of the bodies. A body is
said to have planar motion if all its particles move in parallel planes, i.e.,
when the true paths of all its particles can be represented on a single
plane parallel to the planes of the moving particles. 1\ body rotating
about a .ti.red axis, for example, has planar motion, and any plane per­
pendicular to the axis may be considered as the plane of motion, for the
true paths of all particles can be projected into this plane. Other bodies
may be referred to this plane, provided that their motions are a com­
bination of rotations about axes that are parallel to the fixed axis and
translations along axes perpendicular to the fixed axis. A. mechanism
whose links have planar motions all parallel to the same plane is called a

\(V9)

¼\�=B�------=-=-=-�
n
Normal 0�

I
,I
I

FIGURE 2-21 Disk earn, in which \

pressure and deviation angles are
identical.
50 K I N E M A T I C S Y N T H E S I S OF L I N K A G E S

planar mechanism: the true paths of all particles of all links may be
shown in one plane, "the plane of the paper." The four-bar linkage,
the slider-crank mechanism, gears on parallel shafts, the disk cam ,vith
reciprocating foHo,ver, and so on, are typical examples.
A. body has spatial motion if all its particles do not move in parallel
planes. A scre,v turning in its nut, and hence also moving axiaHy, has
spatial motion, since the angle of rotation and the axial translation can­
not be depicted on the same plane: any particle of the screw describes a
helical path in space. A mechanism which is not planar is said to be
spatial. A spatial mechanism may contain but one link ,vith spatial
motion (as a screw); or it may have a number of links ,vhose planar
motions are not parallel to a common plane. Among the characteristics
of spatial mechanisms are the presence of nonparallel axes of rotation
and cylindric and ball-and -socket joints. The Hooke universal joint is
a familiar spatial linkage; it is also a representative of the special case of
spherical mechanisms.
The Hooke coupling is commonly called a universal joint because
of its ability to transmit motion between t,vo intersecting but noncol­
linear shafts. It should be remarked that there are a number of universal
joints and that the Hooke type is but one of the lot. In continental
Europe it is known as the Cardan (also Kardan) joint. As it happens,
neither Cardan nor Hooke invented it; Hooke's name is associated with
it since he put it to use in the seventeenth century.
A recognizable Hooke joint is shown in Fig. 2-22a. The t,vo
shafts misaligned by an angle a are represented by the revolutes R 1 and
R 2• The central cross 4 carries the revolutes Re4 and R a, whose axes are
at right angles. Furthermore, the axis of R 1 is perpendicular to that
of R 4 , and the axes of R3 and R 2 are also perpendicular. Lastly, all four
revolute axes intersect at a common and fixed point O; it is this mutual
intersection of aH revolute axes at a fixed point that declares this spatial
mechanism to be also a spherical mechanism. We may go one step
further and remark that the Hooke joint is itself a special case of a spheri­
cal mechanism by reason of the three right angles.
The Hooke joint is shown in one schematic form in Fig. 2-22b.
We recognize that all particles of link 4 (no matter ,vhat its physical shape
might be) move on spherical surfaces centered at the fixed point 0, that
is, all particles of link 4 move on concentric spheres ,vhose center is the
fixed point 0. Such a motion is specificaJly called spherical, to distin­
guish it fron1 less ,,·ell 1·egulated spatial 1notions that "·ould occur if
revolute axes did not intersect at a common point. Links I and 3,
considered individually, have planar ,notion; the path of any particle
is a circle lying in a plane perpendicular to the particle's axis of rotation.
Ho\\'ever, since the axes of R 1 and R2 possess a common point at 0, ,ve
C O N C E P T S AND N O TA T I ONS REL A T E D 'l' O M E C H A S
NISMS 51

2,Frame

Hooke Joint (a, b)

(a)

FIGURE 2-22 Four-revolute spatial 1nechanisms.

can also imagine the particles of links 1 and 3 to move on spheres centered
at 0. The simplest case of a spherical mechanism ,vould involve two
bevel gears; the simplest spatial mechanisn1 ,vould be a ,vorm-and-wheel
or two crossed helical gears.
The Hooke joint is a spherical four-bar linkage ; like the planar
four-bar, it has four revolute connections. The difference bet,veen the
two lies in the orientation of the revolute axes. In a spherical four-bar,
52 K I N E �I A T I C S Y N T H E S I S O F L I N K A G E S

the definitive parameters are the four angles bet,veen axes; in a planar
four-bar, the parameters are the four link lengths. One other four­
revolute linkage, a spatial mechanism also, exists: it is the Bennett
mechanisn1 (Fig. 2-22c). In this, the opposite links hav<> the same
lengths and th� san1e angles of t"·ist, but the lengths and the t,vists are
related.
For other and more co1nplicated spatial n1echanisn1s, see Chap. 12.

2-12 K I X E ).( A T I C CHA 1 X S

A. material body "·ith t,\·o or more kine1natic elements is called a

link. Each element represents a place of contact ,vith, or connection to.
another link. A link carrying t,vo elements is a binary link; if there are
three elements, it is a ternary link, if four, a quaternary link; and so on.
A. planar four-bar linkage, a spherical four-bar, and a Bennett mechanism
are each composed of four binary links. A cam in contact ,vith only a
single follo\\'er ,vould also be a binary link.
The bell crank, link 2 in Fig. 2-23a, is a ternary link, for it con­
nects, or "contacts," links 1 , 3, and 4. Link 3 of the transport mecha­
nism of Fig. 2-13, reproduced here as Fig. 2-23b, is a quaternary link,

Revolute
4 element
2
\
/
/
/

Cam surface,
1 element of
higher pair
(a) (b) (c)

., Cam surface, element

,------ --.;,,----,of higher pair

(e) revolute pair

element

(d)

I<'IOt:RE 2-23 Exa1nples of links.

 C O N C E P T S .-\ N D XOTAT I O N S R E L A T E D T O .M E C' H .-\ N I S !\I S 53

1
"
1
�

0 8

T + T
1 1

8
@+
C1
3 03C2
7

FIGt:RE2-24 Example of quintary link. The wheel, link 2 (which includes B), con­
nects with links 1 , 3, 4, 5, and 6. Note that C1 and C2 are coupler points of a parallel­
ogram linkage and describe circles of crank radius.

connecting as it does ,vith the (four) links 2, 4, 5, and 6. Other examples

of links, with higher-pair elements, are sho,vn in Fig. 2-23c, d, and e.
The locomotive ,vheel of Fig. 2-24 presents an interesting situa­
tion : it not only is a quaternary link, or quintary when wheel-rail con­
tact is considered, but also sho,vs the construction of a short-throw crank.
Wheel, axle, crankpin A , and "eccentric link" B constitute link 2. The
axle is in contact ,vith the fran1e (link I ) ; the crankpin carries the con­
necting rod (link 4) and side, or parallel, rod (link 3). In addition, the
eccentric link B connects to the eccentric rod (link 5). This last connect­
ing point, C, describes a circle of radius e about the center of the ,vheel.
Although the above definition of a link is very general and includes
the possibility of one-way-rigid links such as bands, ropes, belts, and
fluids, the present text ,vill be concerned only with rigid links.
A kinematic chain is an assemblage of parts, or links, connected
by pairs. Geometric considerations sometimes preclude motion of tt:e
chain after closure, in ,vhich case the chain is called a structure, and thi:s
may be statically determinate or indeterminate (Fig. 2-25). A chain is
closed ,vhen all pairs are complete because of mated, or connected, ele­
ments, as in Fig. 2-26. Incomplete pairs indicate an open chain (Fig.
2-27). A siniple- closed chain is composed of only binary links, each link
connecting to but two others, as in Fig. 2-26a and b. Com.pound-closed
chains contain ternary and higher-order links, each connecting t') more

FIGURE 2-25 Nonmovable

chains or structures. Statically determinate Statically indeterminate
 54 K I .!\' EM A T I C S Y N T H E S I S OF L I N K A GE S

R, ���� Rs

2 R2 4
R 1 0----------<l R31..:
V :.i�.� �,��JiM
!,tttlmit
ti§ii�¾����-,, ��
R4 R1 1 R1
Simple-closed Compound-closed
(a) (b) (C)

nouRE 2-26 l\lovable closed chains.

than two other links, as in Fig. 2-26c, ,vhere each of the ternary links
1 and 3 connects to three other links.
With the aid of the notion of kinematic chain, a mechanism, con­
sidered earlier as a motion-transfvrming device, may now be given a
new, more accurate, and perhaps more restricted definitione: a mechanism
1s a movable closed kinen1atic chain ,Yith one of its links stationary. 1
1
Although we have gone from chain to mechanism by selecting a fixed link,
no input link-the source of the motion to be transformed-has been designated.
The Germans have a word for a mec-hanism in whic-h the driving link has been selectede:
it is Getriebe and has caused much confusion. The German Ketle means chain, and
.lfechanismus is the equivalent of mechanism, but there is no acc-epted translation for
Getriebe, althou�h drive and train have been used. In consequen<'e the untranslatable
Getriebe appears as mechanism when used as a noun. In adjective forn1, e.g., Getr·itbe­
/ehre. (/,ehre = t. heory or scienc-e of), it is translated as kinemat:cs or 1nechanisn1s.

R 10-------.....ie::::::p
�
,

FIGURE 2-27 Open chains.

CONCEPTS A N D �OTATIONS RELATED T O MECH A NISMS 55

2
4

FIGl"RE 2-28 Four-revolute planar chain meeting

Grashof condition (see Sec. 3-3). 1

Thus, in Fig. 2-26, the chain at (a) becomes a mechanism, a four-bar

linkage, \vhen one of its links (link 1, for example) is made stationary to
form the frame of the mechanism. The chain at (b) gives a slider-crank
mechanism ,vhen link 1 is 1nade the frame.
. s noted earlier, a mechanism (either planar or spatial) in ,vhich
�
all connections are lower pairs is called a linkage. Linkage and lo,ver­
pair mechanism are therefore taken to be synonymous.

2-13 I :'.'l' V E RSIOX

.A n1echa11isn1 is derived fron1 a closed kinematic chain by making

one of its links stationarya: by choosing different links as the stationary
link or frame, the same closed chain ,vill yield as many distincta1 mecha­
nisms as it has links. The chain of Fig. 2-28, for example, yields four
different four-bar linkages, as shown in Fig. 2-29. The four-bar link­
age a, sometimes called the crank-and-rocker mechanism, gives an
oscillation of the follo,ver 2 for a continuous rotation of the crank 4.
A.t b, a rotation of link 3 ,vith constant angular velocity gives link 1 a
continuous rotation ,vith variable angular velocity. This is the drag-link
mechanism; double-crank mechanism is also descriptive. At c, the
situation is similar to that of a, but with different motion characteristics.
At d, links 1 and ;3 can rotate only through angles less than :360a° : this
is the double-rocker n1echanism.
The process of fixing different links of a chain to create differe11t
n1echanisms is called kine,natic inversion. The four-bar mechanisms
a, b, c, and d (Fig. 2-29) are the four inversions of the four-bar chain
(Fig. 2-28).
The mechanism shown in Fig. 2 -30a, called the Scotch yoke, con-
sists of four rigid links connected by two revolute pairs R1 and R2 and
two prismatic pairs P3 and P,. Crank rotation at constant angular
velocity gives the yoke a translation that is a sinusoidal function of time.
An in version in ,vhich link 2, the crank, is chosen as the fixed link yields
the mechanism of J?ig. 2-30b. This, kno,vn as the Oldham coupling,
transmits rotation between two parallel shafts ,vith an angular velocity
Distinct refers to the input-output relations of links attached to the fran1e ;
1

the relative motions of all links remain the same.

56 K I NEMATIC SYNTH E S I S O F L I N KAGES

R4 3

/
4

' · - .. , __ ----
1
- -
R1 R2

(a) Crank-rocker (b) Double crank, also draglink

2
R2 R3

(d) Double-rocker

(c) Crank-rocker

FIGURE 2-29 Inversions of the four-bar linkage.

ratio of unity. Another inversion, in which link 4 (yoke) is made the

fixed link, yields the elliptic trarnmel (Fig. 2-30c) ; here the center point C
of link 2 traces a circle, ,vith all other points describing ellipses. A
fourth and last inversion, in ,vhich link 3 (block) is fixed, gives a Scotch
yoke different from the first.
As a further example of inversion ,ve discuss the slider-crank
mechanism (Fig. 2-3la). When crank 2 is fixed, mechanism b gives
link ;3 a continuous rotation of variable angular velocity ,vhen link I
rotates at constant angular velocity. This motion transformation is
similar to that of the drag-link inversion of the four-bar linkage; it has
been applied to many machine tools, ,vhere it is called the Whitworth
quick-return mechanism. With connecting rod 3 made the fixed link,
mechanism c has found application in oscillating-cylinder steam enginess1
For example, the Great Eastern, built in 1858, had a four-cylinder oscil­
1

lating engine (74-in. bore, 14-ft stroke) developing 3,410 hp at 11 rpm with stearn at
24 psig. This engine, sitting low in the hull; drove overhead crankshafts to which
56-ft.-diarneter paddle-· wheels were directly connected. In addition, there was a.
conventional four-cylinder horizontally opposed engine of 4,890 hp (and 39 rp1n)
for a 24-ft screw. If all this failed, six masts could spread l½ acres of sail!
C O N C EPTS AND NOTATIONS RELATED T O M E C H A N I SMS 57

and pumps. lVIechanism d finds use as a hand pump: link 1 is oriented

in the vertical, and link 2 is extended to form the pump handle.
Among the earliest aircraft power plants ,vas the rotary engine.
Spectacularly successful in 1910, its useful life extended well into World
War I. This engine (Fig. 2-32) v.,as the same inversion as the Whitworth
mechanism: the crankshaft (link 2) was bolted to the fuselage to become
the fixed link. The propeller was bolted to the crankcase (link 1), and
this assembly, complete with cylinders, revolved.

2-14 E X PAN SION OF REVOLUTE PAIRS A N D

OTH E R D I S GUISES
'foo often the physical shape of the connection betv,een links is
such that the true character and function of the connection are not
immediately apparent. The reason for the disguise may stem from
practical design considerations such as strength requirements, ease of
manufacture, or space limitation, all of ,vhich obscure the nature of the
kinematic elements of the connection, although the relative motions of
the links remain unaffected. The situation is simply that the center

�
3
·:•.$-·.:.•·•.,x...::,
wmP :
•.
:< R2

P. 3
�
i\R1
4,yoke �
s,.......................:.............. I

(a) Scotch yoke (b) Oldham coupling

-

(c) Elliptic trammel

FIGURE 2-30 Inversions of the Scotch yoke.

58 K I N EMA'l' l C S Y N T H ESIS O F L I N K A G E S

,/

Stroke •I
(a) I

2
(b)

FIGURE 2-�1 The slider�rank chain inversions.

Cylinder and
crankcase 3

--/e�

Fuselage

Propeller bolted to
crankcase, link 1
I
FIGURE 2-a2 Rotary aircraft engine, a slider-crank inversion with link 2 fixed.
C O N CEPTS AND NOTATIONS RELATED TO MECH A N I SM S 59

FIGt'RE 2-3a Expansion of revolute pair.

of what is kinematically a revolute pair is not directly discernible. A

case i n point is the common eccentric (Fig. 2-33b)e; examination sho,vs
it to be an oversized, or "expanded," crankpin, evident on a comparison
with the slider-crank mechanism of Fig. 2-33a. Inspection identifies
the center A of revolute R 2 in Fig. 2-33b and finds the crank of length OA A .
Revolute 2 i n this expanded form is knO'wn as an "eccentric," and the
shaft constituting revolute R 1 may now be continuous instead of inter­
rupted by a crank. Obviously the physical diameters of the revolutes
R 1 and R2 are of no kinematic importance: ,Yhat is significant is that the
centers OA and A are the same distance apart in the t,vo mechanisms.
The trick lies in discovering the disguised center of the "misproportioned"
revolute R2. ,,ariations in form and size of a turning pair that do not
alter the relative motion bet,veen connected members constitute an
expansion of a revolute pair.
Further examples may be found. A four-bar linkage is shown
i n Fig. 2-34a in its familiar form; at b revolute pair R3 has been expanded,
as was R2 in the slider crank, and the coupler has now become block �­
For a complete revolution of crank 2 block 3 traverses only the small arc
E 1E2 of the element of revolute Ra on member 4. The motion of block 3
,vould still be described by means of an angle referred to B. Link 4 may
be given a still different physical form as at c ,vithout altering the relative
motions of the links. It should be observed that the "curved slider"
is a form of revolute: the motion of block 3 in its guide is a rotation
defined by an angle, not by a linear distance. We note that the space
requirements of forms a, b, and c are quite different.

2-15 PRISl\1 AT I C P A I R AS THE LIM IT

OF A R E V O LUTE PAIR

Although differing in appearance from the "pin connection" of

Fig. 2-34a, the "curved slider" of Fig. 2-34c remains a revolute pair as
long as its radius of curvature is finite. The center of curvature B is
part of the moving plane 4 from which the physical shape of link 4 has
been cut. If, how·ever, the radius of curvature of a revolute pair becomes
60 K I N EM A T I C S Y N T H E S I S O F LINKAGES

infinite, i.e., when its center of rotation goes to infinity, then, and only
then, does the revolute pair become a prismatic pair, i.e., the pair variable
changes from an angle to a linear distance.
1'his transition from a revolute to a prismatic pair is sho,vn in
Fig. 2-35. By expansion of the revolute pair R4, the four-bar linkage
shown at a takes the form b. Suppose now that the center of rotation
08 of the revolute R4 is moved do,vn along the vertical by increasing
the lengths of members 1 and 4 as sho,vn at c. By expansion of the
revolute R 4 , this ne,v four-bar linkage takes the form d, in which the
radius of curvature of curved slider R� is greater than at b. ::\loving
the center On farther down simply increases the radius of curvature of R 4.
At the limit, ,vhen On is at infinity on the vertical and members 1 and 4
have become infinitely long as atse, the four-bar linkage becomes a slider­
crank mechanism f. The curved slider has now become straight, yield­
ing a prismatic pair. 1'hus, a prismatic pair may be considered as a
revolute pair whose center is at infinity in the direction perpendicular
to the generatrix. Having arrived at this stage, On may be located at
infinity, either "up" or "down."
In a previous section, the prismatic pair was considered as a
limiting case of a screw pair with an infinite lead. The prismatic pair
is here considered as a revolute pair ,vith its center of rotation at infinity.
These two interpretations of the prismatic pair, ho,vever, should not be

Ra, B

- B
1 I
I
I

�
OB

(b)

(c)

FIGPRE 2-34 Expansio11 of reYolute pair.

CONCEPTS A N D NOTA T I O N S REL A T E D TO M E C H .-\ N I S M S 67

R2 R2 Ra
3 3 R,
R3

---- 4
'
I
1
--..._....._..._ -- -- R, --- --- --
I
I

(a) (b) +o
OB B

R2 R2 R3
3 3
R3 R,

''
� � I
', '--
'' 4 '--
''
I
I
'-
' '" I
I
,, ' ,, I
(C)
Oa ; R, (d)
¥B I

R2 Ra
3
Ra
(p'
I
�
1 !
4 I
I
I
I

'
(e) (f) I
!
l O8 at infinity

FIGl.RE 2-35
Prisrnatic pa.ir as a limiting case of revolute pair
with center at infinity.

considered as conflicting, for each vie,v may serve a different purpose ; it

is a case of the end justifying the means- making use of the most con­
venient and legitimate argument for differing purposes. As noted
before, the first interpretation is convenient in setting up a symbolic
notation for lower-pair mechanisms. The second interpretation is put to
use in the synthesis of planar mechanisms, allo,ving many of the proper­
ties of the four-bar linkage, when carried through the limit process sho,vn
in Fig. 2-35, also to become properties of the slider-crank mechanism.

2-16 E Q U I VALENT L I N KA G E S

The complete kinematic analysis of a mechanis1n includes, a1nong

other things, the determination of velocities and accelerations. Diffi­
culties may be encountered when links are connected by a higher pair,
 .,

62 K I N EM A T I C SYNTHESIS OF L I N K A G E S

as shown in Fig. 2-36a, where the relative motion between the two profiles
consists in rolling coupled in uncertain fashion with sliding. On pro­
ceeding in the usual manner with vector equations (see Chap. 4), it
would be necessary to apply the Coriolis theorem and to know the curva­
ture of the path traced by a point of one link with respect to the other.
If no easily recognized path is found, it may be difficult, or at the very
least tedious, to establish the desired path curvature.
The equivalent linkage replaces the higher pair with properly
disposed lo,ver pairs. These will, for the instantaneous phase under
consideration, give correct values of velocities and accelerations. Let

1 n. common normal
I

2
Al A
OA
'
1

'----- jp Additional
'I link
I

B! B

n, common normal
/

I
I/4
I 3

Os
(c) (d)

FIGURE 2-36 Equivalent linkages.

CONCEPTS AND NOTATIONS RELAT E D TO MECHANISMS 63
n,common
· / normal
Additional link
/
A •/ A
\
2 '\ O,I
�

(a) (b)
FIGCRE 2-:37 Carn with constant curvature profile a.nd its invariant equivalent
linkage.

�4 and B be the centers of curvature of the profiles of the higher pair

at their point of contact. Because of the properties of the center of
curvature, the distance A.B ,vill remain constant for three infinitesimally
close positions of the cam and follo,ver, and for these three positions the
same correlation in displacements bet,veen links 2 and a may be obtained
by means of the four-bar linkage OA ABOn shown at b. 1'his linkage is
called an equivalent linkage of the higher-pair mechanism sho,vn at a.
1'his equivalence, however, is valid only for three infinitesimally
close positions of the mechanism. When the links are rotated through a
finite angle, as at c, the dimensions of the equivalent linkage d are differ­
ent. In other ,vords, an equivalent linkage is generally valid only for
a given instant or phase; it does not ordinarily apply to a complete cycle.
In general, tw·o mechanisms are said to be equivalent if they give
the same correlation between three infinitesimally close positions of their
input and output 1ne1nbers. The velocities and accelerations of the
points of a moving link are defined by considering three infinitesimally
close positions of the link. Such positions are preserved in equivalent
linkages, ,vhich may therefore be used to evaluate velocities and accelera­
tions of more complicated mechanisms. Equivalent linkages, ho,vever,
cannot be used to evaluate displacements or the time rate of change of
acceleration, son1etimes called the jerk, or pulse, because these quantities
can be defined only in terms of more than three infinitesimaHy close
positions of a link.
In some few instances an equivalent linkage may be found that
,vill duplicate the motion transformation bet,veen input and output links
of a given mechanism throughout the motion cycle. Consider, for
example, the cam mechanism shown in Fig. 2-37a, the cam profile being
circular, with center at A . The roller has its center at B. The distance
A.B is now constant throughout the cycle, ,vhence the equivalent linkage
OAABOB shown at b is valid for all instants (positions or phases). The
motion transformation between links 2 and 4 of both mechanisn1s is the
same throughout the complete cycle.
64 KINEMATIC S Y NT H E S I S OF LINKAGES

2-17 SYMBOLIC NOTATIONS

Language, either ,vritten or spoken, is a means of communicating

an idea or thought. In written form, it involves the use of signs no,v so
highly conventionalized that their original pictorial origins are not
evident. Any system of signs may be called a language, e.g., the language
of mathematics, in which the symbols stand for operations whose word
descriptions are tedious and unhappy. :\Iathematical notation is the
most po,verful shorthand known: not only is it precise, but, more impor­
tant, it is manipulative. Physics, chemistry, and electronics also have
highly developed notations by means of ,vhich symbols other than only
the mathematical allow the statement of ideas and complex situations
in a compact manner.
No real symbolic notation i::-- widely used in kinematics. Prac­
tically all that exists is a stylized representation of connections, useful
for drawing skeleton diagrams. These diagrams sort out the basic
geometry of a problem and are therefore indispensable, but they lead
no further, for they do not represent a shorthand of the whole mechanism
-they are only a quick picture of it, and no manipulation is possible.
Notwithstanding the controversial nature of the subject, the
authors have been led to reconsider the matter of symbolic notations
and devise an extension of Reuleaux's system to supplement the use of
skeleton diagrams. Of this new symbolic notation, particularly con­
venient when applied to lower-pair mechanisms (both planar and spatial),
an elementary form I will be presented in this section. It provides a
concise qualitative method for identifying linkages by the number and
nature of their pairs. This elementary form will be elaborated upon in
Chap. 12 to make it quantitative by including the geometric relations
between the pair elements of each link. The quantitative form of the
symbolic notation allows an interpretation in terms_ of matrix algebra
to make it manipulative; it is particularly useful for the study of spatial
linkages. However, the complete form is not necessary for the planar
studies forming the major parts of this book.
1. Symbols for lower pairs Each of the six lo,ver pairs described
in Sec. 2-6 will be represented by its symbol:
Spheric pair G (think of Globe-sphere)
Planar pair F (think of Flat)
Cylinder pair C
Screw pair with lead L SL
Revolute pair R
Prismatic pair P
This form is in the spirit of what Reuleaux called the "contracted formulae"
1

(Kennedy, "Reuleaux' Kinematics of Machinery," p. 24)3).

CONCEPTS A N D NOTATIONS RELATED T O M E C H A N I S M S 65

As already remarked in Sec. 2-6, a revolute pair may be considered

as a screw pair with a lead equal to zero; similarly, a prismatic pair is a
screw pair with an infinite lead. The symbols for the revolute and
prismatic pairs may therefore be written respectively as So and S«> with­
out the introduction of any new symbols. However, because the revolute
and prismatic pairs are very common, it is convenient to give them their
particular symbols of R and P.
2. Synibolic description of simple-closed chains A simple-closed
chain consists of only binary links; i.e., each link i1, connected to two,
and only two, other links. The chain is completely described if each
connection is designated by its proper symbol, the connections being
labeled in sequence, clock,vise or counterclockwise. One connection is
arbitrarily chosen as the starting point and the others noted in sequence
,vhile going around the loop in either sense. Thus, the symbolic descrip­
tions of the chains shov,n in Fig. 2-26 might be written
For a :
or etc.

etc.
3. Symbolic description of compound-closed chains A compound­
closed chain consists of a combination of simple-closed chains, possible
because some of the links connect to more than two other links, i.e.,
some of the links are ternary, quaternary, etc. Such a chain is completely
described if each connection is designated by its proper symbol (as in
the case of simple-closed chains) and enough simple-closed chains are
described to include all the connections (pairs). This method is similar
to the analysis of electrical circuits, in w·hich the voltage equation is
written for each independent loop of the circuit. Examination of the
compound-closed chain of Fig. 2-26 discloses three different simple­
closed chains:
and
Any t,vo of the above chains, ho,vever, are sufficient to describe the <
compound-closed chain, since they include all the pairs. rfhe compound- \
closed chain of Fig. 2-26 may therefore be written as
R1R2RaR4
R 1R2RsRs R;
Two other combinations of simple-closed chains would also serve.
The utility of the symbolic notation is sho,vn in rather striking
fashion by Fig. 2-38. At first glance, and perhaps even at the second,
66 K INEM A T I C S Y N T H E S I S O F L I N K A G E S

l
I

-----+- --+- - . O-=--

R1

(b) Rapson's slide, a marine steering gear:

the rudder post is at R1 , and block ·4
is actuated by chains or wire ropes _
(a) Davis automobile steering gear

FIGURE 2-38 Two kinematically identical mechanisms R,PtRJ'4 •

l
Rapson's slide and the Davis steering arrsngement evidence no kinship.
Identification and ordering of the kinematic pairs sho,v the two mecha­
nisms to be the same chain. This is all nicely obscured by the different
shapes of the corresponding pieces of hardware (in part due to the
exchange of hollow and full elements of the prismatic pairs) and, of
course, the unrelated areas of application. Reuleaux, who was the first
to appreciate the need for and the niceties of a symbolic notation by
creating one, applied it to showing the kinematic similarities existing
in the many rotary steam engines known to him. 1
�Iechanisms with higher pairs are subject to symbolic representa­
tion only in terms of their equivalent linkages. For certain cases, such
as that of l◄'ig. 2-37, a single invariant equivalent linkage may be found
that will duplicate the motion transformation between input and output
links of the mechanism for the complete cycle. Here the symbolic
representation of the cam mechanism is given by that of its invariant
equivalent linkage, a planar four-bar.
The foregoing symbolic descriptions are qualitative rather than
quantitative: they identify the nature and number of the pairs involved
in a mechanism and the order in which the pairs appear but give no
information about angles and distances bet,veen pair axes. Such a
description does not differentiate, for example, between a planar fdur-bar
linkage, spherical four-bar, Hooke joint, and Bennett mechanism, for
they are all written R,R 2RaR4; further qualification must be supplied
1 Ibid., pp. 342-384.
C O N C E PTS A N D N O T A T I O N S R E L A T E D TO M E C H A N ISMS 67

by an adjective. An idea of how the necessary quantitative information

is added to the symbolic notation will be given in Chap. 12; until then it
must be conveyed in words as part of the context.

BIBLIOGRAPHY

Kennedy, Alex. B. W.: "Reuleaux' Kinematics of l\fachinery," :\[acmillan &

Co., Ltd., London, 1876. Reprinted by Dover Publications, Inc., New
·York, 1963.
---: "The 1Iechanics of Nlachinery," ::\Iacmillan & Co., Ltd., London, 1886.
Reuleaux, F.: "Theoretische Kinematik," Friedrich Vieweg und Sohn, Bruns­
wick, Germany, 1875.
---: "Lehrbuch der Kinematik, " Friedrich Vieweg und Sohn, Brunswick,
Germany, 1900.
\Villis, Robert : "Principles of ::\Iechanism," 1st ed., Parker, London, 1841.
---: "Principles of �Iechanism," 2d ed., Longmans, Green & Co., Ltd.,
London, 1870.
 KINEMATIC MODELS

3-1 THE :MEANING OF �lODELS

The term model evokes many interpretations if considered in
detail. In the large sense of the word, however, there is the
implication that a model-,vhatever its form-establishes com­
munication and furnishes an understanding transcending the
,vritten word. :\1odel has nothing to do ,vith size: it has to do
only ,vith perceptive form.

l\!Iodels of varied kinds have been used for millennia. Depend­

ing on their purpose, models may express ideas of design, confirm
or negate an assumption or hypothesis, guide analytical reasoning
away from a false path, obtain fundamental data, and so on.

In the study of models we shall be principally concerned ,vith

hardware types suitable for kinematics. These represent a
language of a very special kind, quite distinct from the conven­
tional symbolic notations that have been devised for convenience
and manipulative procedures. lVIodels are a visual language.
The figures of a geometer are models: with them the relations
of points, lines, angles, and planes are kept track of. But ordi­
nary geometry is static, one diagram being sufficient. Kine­
matics is geometry in motion: there is a ne,v diagram for each
instant. Displacement, the first aspect of motion, demands a
movable geometric figure capable of displaying (1) the changing
 KINEMATIC MODELS 69
relations of angles and line lengths, (2) the position changes of points,
and (3) the curves traced by the points. It is one thing to show a
mechanism in a given position, and quite another to show several phases,
separating each phase in the welter of lines. In fact, the time spent in
making a model may very ,veil be less than that required to pursue graph­
ically a motion through the several phases judged to be critical or of
interest. Good and bad transmission angles will discover themselves,
unexpected change points may appear, and the like. A model certainly
furnishes the quickest way to the display of coupler-point curves.
While Lagrange's epoch-making "Niechanique analytique" was
further distinguished by a total lack of figures, most persons require
diagrams and sketches to define the geometrical relations according to
which their work proceeds. Kinematic models able to demonstrate the
displacement relations (among other things) consequent to any position
are the kinematic equivalents of the geometer's diagrams and the mechan­
icist's figures.
With a spatial mechanism the drafting problem may verge on
the unspeakable because of the maze of construction lines and the need
for auxiliary views. Spatial models, more difficult to construct than
their planar counterparts, are nevertheless well worth the trouble, for
situations defying the imagination are put on display.
Kinematic models fall into three categories. The first group is
composed of models that the instruct-or demonstrates to the class: all
members of the class may or may not follow the motions as they come
and go. The models of the second group are those which are available
for personal manipulation and study ("fussing"). The third and most
re,varding category embraces all models, no matter ho,v crude or rude,
made by the interested individual. This group, and ,ve denote it as
the "do-it-yourself" variety, is the most valuable of the lot. The virtue
and merit come from the fact that the man who builds his own models­
who nurses them along and makes them work-has lived ,vith the mecha­
nism and learned its moods. Because of this intimacy the model builder
learns much more than the model viewer or model fusser.
Fortunately, a great number of planar models of even considerable
complication can be made on a desk top from materials and office equip­
ment at hand: no machine-shop facilities are required. Cardboard strips,
thumbtacks, scissors, knife, adhesive, etc., are the materiel and tools.
Spatial models, having links subjected to bending and torsion and con­
nections such as ball-and-socket joints and cylindrical pairs, require
metal 1 and somewhat more in the ,vay of a "shop" facility, but no high­
grade shop ,vith elaborate tools. 1'he model maker, in using whatever
1
It is possible to fold torsionally and flexurally stiff links from cardboard,
but rnetal strips, if available, are more satisfactory.
70 KINEMATIC SYNTHESIS OF LINKAGES

material is convenient, will have to supply ingenuity, patience, and time.

However, after gaining a little experience, the time expended on a model
will be reduced and a simple model produced with little effort. Another
,vay of putting this is to say that not only can standard situations be
investigated-those presented in a book or paper, say-but different pro­
portions of the links (lengths) may also be looked at. All this is much
more satisfying and instructive than being bound to a prepared model
of invariant proportions (e.g., the first and second categories mentioned
earlier) or trying to folio"' link motions on a drawing. This is not to
say that drawing-board work can be dispensed with; on the contrary,
the drawings become much more meaningful. 1Iodels augment draw­
ings; they do not supplant them.
It is of these do-it-yourself models that we shall speak after having
considered something of the gen:-ral background of kinematic model
,vork.

3-2 HISTORICAL SURVEY

Geometric models of various sorts have an ancient history. For

example, Archimedes (third century B.c.) ,veighed models of parabolic
segments before developing the theorem that the parabolic area is two­
thirds of the area of the circumscribing parallelogram. Archimedes is
also kno,vn to have made a planetarium of some kind in ,vhich the motions
of the sun, moon, and planets were displayed. We note that this early
kinematic model was concerned with showing the relative motions of
points. :\lore elaborate devices of the same nature, but operating by
wheehvork and stemming from the eighteenth century, are known as
.
orrer1es.
Nonkinematic models of the past and present are legion. Archi­
tects and builders of medieval times made models of their cathedral
structures to plan the erection of the stonework. For centuries ships
were built by the scaling up of dimensions taken from hand-carved hulls.
In the elaborate "Admiralty models," details such as interior framing,
decks, and most other structural features were included. For the
seventeenth century one may read that it had become the "custom of
the times" to develop large-scale undertakings from the miniature. To
do more than mention the models used by all phases of the aircraft
industry would be to belabor the point.
An early, perhaps first, discussion of kinematic models stems
from Robert Willis (1800-1875), the kinematician of Cambridge.
Willis was the successor to the Rev. William Farish, a remarkable man.
Farish (1759-1837) occupied the chair of chemistry before being elected
to the office of Jacksonian professor of natural and experimental philoso-
 KINEMATIC MODELS 71

phy in the year 1813. We read in Willis (1851) that Farish "soon after
commenced a Course of Lectures on Arts and l\1anufactures, which he
repeated yearly until his death in 1837. The plan of this Course included
the exhibition of almost all the more important machines which were
then in use in the manufactures of Britain. 1'his led him to conceive
the possibility of devising a system of mechanical apparatus consisting
of the separate parts of which machines are made, so adapted to each
other, that they might admit to being put together at pleasure in the
form of any machine that might be required." 1 Willis goes on to say
that by 1837, when he himself became Jacksonian professor, "it appeared
. . . that his [Farish's] idea of a Protean mechanism "·as capable of
being carried out in a different and more complete manner, so as to be
of greater practical utility, and of a more extensive application to philo­
sophical apparatus in general."
Few details are known about the Farish "system," but the pieces
were chiefly made of metal, some of the steel shafting being of ¾-in.
octagon section. Willis said:

The entire plan . . . had the merit of great simplicity and

ingenuity, but the machines ,vere apt to appear somewhat
embarrassed and complicated by the variety of clamps and
junctions of the framework. It is to be regretted that the
ingenious inventor of this mechanism did not draw up a detailed
account of his system. rfhe general appearance of his models
may be gathered from the sketch of an optical grinding engine
appended to his Paper on Isometrical Perspective. . ..

This is Fig. 3-1. In this paper J.'arish proposed the no,v ,videly used
isometric perspective of 30°, derived from sighting do,vn the diagonal
of a cube. He coined the name isometric, since the three perpendicular
dimensions of the cube were equal.
Building on the experience of his predecessor, and taking advan­
tage "of the facilities which the improved state of machine-making
afforded," Willis devised a rather complete system, or kit, of parts, 2
giving all dimensions, naming the pieces, and so on. He recommended.
that each setup, once achieved and in ·operating order, be sketched in
isometric projection for the record.
The various parts of the WilJis kit were also substantial, since
1 These lectures became part of the curriculum designed to liberalize educa­
tion in the classics!
2 It seems that these kit parts have not been preserved. Such n1odels as are
ascribed to Willis-and a few still exist at Cambridge-are made of n1a.terials and
components quite different from those described so glowingly in 1851.
d,e appo,.""U"""', int.-n&d for J..,,u,..,.hall 11"'1 hdore lar�o audien""",
noodeJ �euorou• prupo;tiou• for vi.il;i]ity. The shafting dian,el<er
var;.,dlrom-lto l in.;theitshl<>!nndlramework ii11 lc11Ktho up t,<> !Oft)
�-e,,,, o{ lt- by 2t,-in. wood; ""lit iron wa, ullod for br1>Cket.s, etc., the
oonneclion, of \lie /ramin1t bein� ma<l� witl, J-in. bolt�
Willi,foundithathi,cxtendodi,y<tem.,compli,l,eJ!ourpurpooc-o
(l)thoi conotrncfa"'ofi 1110,,lel,i ofrne<lurni,u'"'uchudi..,u"""'1iut,;,
"l'rinciplc,a of Mechaniom" (18·11); (2) the oo,..t.ruet>On of oompMe
machi,,..;(3)"the,m,.trnct>Oua11da"AollCm<ntoliappan,t1"fo,E.J>Cri-
1111mta! Philoooplty iu �•neral '; (4)"tho \rW of new oomLinatio,,. •n,I
origino.t,_,.rch."
KINEMATIC MODELS 73

We see from the Farish-Willis efforts that the first formal kine­
matic models were assembled as needed from a kit of parts and taken
down •after use; no permanent display of models ,vas maintained.
Some idea of Willis' system of components may be gained from
Fig. 3-2. Here, in \Villis' fig. 47, parts have been assembled into "an
arrangement for the exhibition of trial of various regulators for mecha-

0 2
Scale

1 I
49

FIGURE 32- Mode.is built from Willis' kit of components. Their identification is
taken from \Villis (1851): fig. 45, parallel-motion curve machine; fig. 46, friction
engine; fig. 47, equatorial clock; figs. 48, 49, hook bolts; fig. 50, detail of pendulum.
74 KINEM ATIC SYNTHESIS O F LINKA G E S

nism moved by weight"; and "fig. 46 is a machine to elucidate the laws

of friction."
The eye-catching upper-left figure must also be described by
Willis:

Figure 4:) is a machine to describe the curves that belong to

parallel motions 1 under various proportions of the radius rods
and link, and different positions of the describing point upon the
latter. . . . A parallel motion consists of two radius rods con­
nected by a link: a pencil is attached to this link, and ,vhen the
rods are made to turn about their fixed centres of motion, the
pencil describes a curve, \\·hich, under certain proportions
bet"·een the lengths of rods, link, etc., possesses the property
that a part of its length hs so nearly rectilinear, that in practice
it may be employed as if it ,vere a true straight line.

Willis has just described what is today called the four-bar linkage
and the couple1·-point curves that it can draw. We note his emphasis
on the "nearly rectilinear" portions of the curve ; today we speak of
approxi1nate straight-line segments. In Willis' time there was a greater
interest in such parallel-motion linkages because of the steam engine.
As noted earlier, Watt had shown the way in 1 784 when he devised a
four-bar linkage whose coupler point guided the upper end of his piston
rod along an approximate straight line, good enough for his purpose.
In Watt's day, long before the invention of the planer (1817), pin-con­
nected linkages ,vere readily made, while sliding crossheads ,vere not.
1 Willis (1841 ed., p. 399) explains:

A parallel n1otion is a tenn so1new hat a\\·kwardly applied to a combination

of jointed rods, the purpose of which is to cause a point to describe a straight
line by communicating to it sin1ultaneously two or more motions in circular
arcs, the deviations of these motions from rectilinearity being n1ade as
nearly as possible to counteract each other.
The rectilinear motion so produced is not strictly accurate, but by properly
proportioning the parts of the contrivance, the errors are rendered so slight
that they 1nay be neglected.

Kempe (p. 49; see Bibliography) says:

I have been more than once asked to get rid of the_ objectionable term
"parallel motion." I do not know how it came to be employed, and it
certainly does not express what is intended. The apparatus does not give
"parallel motion," but approximate "rectilinear motion." The expression,
however, has now become crystallized, and I for one cannot undertake to
find a solvent.
 � k•
i �nN µ,·e rioe to • mombor of "po.r l J
;::::;i;:;�;�':!n ;';::\::,::
Wet rnw,tt-nizet.,....hcr,;.....,ttmodelt oe<i,;1y. ltti,ttha\1>1
n..,1,.nx (18211-1t00..",) in ll<rlin and •her lhe 1801.lo. Reuleo,ux'• �·ere
alltpcrmanenttmodel,,>tur<lilytoonJ!t,ucl<'<ltofi,.,.,to.,dtb,....,co.eht•hn-..-.
;11,; A p&rticulu kincm&tic situation (1.ig. 3-�). Tht �uif""'"' R.•,lin
eollc<tion of ,..,..,ral hundred model• u, d..tmy«I in Wwld Wor 11
Smaller collediona of Heulco.ux model, "ill �xiot; the prin,ipo.l "'"' i, ,t
Co,,ocll Uroiver>itytOJtdtnurubentZOOtil,•n"
The new T..,hnical !lni,•c..,j\y of Rtrlin h,H not auemµted W
,..,..,,..,.,te it. loot collttti,-,,,_ ln,tea<l, plonor moddo, indi1·iduol!y
mount<.<! on po.neb perhapo 7 by 14 i11., i.. ,.� l,een ,...,mhlod in groupo
of 21 in wall cabinets. �uch amoyo a.-. brou�ht ;.,w ,notio,, hy • plWl
button, eo,cb po.11el'• mod,!, b<in.ll dri,·<11 by • ,i.ngle eloctri<, moto,.

3...J PLA�'AJI X J N E )IA'l'J(' >I O O � L O

A ■ eon b o i n...�ne<l, the probl<m, •t�i,d&r,t upon thot.,,,.,.tn,ctioo
oftplanort modolat an,t l...,t t,yingt thont th_t dt_tialt modelo. The
materiab,iandi\-00.loipr...,uti noi difficulti... Theii....i<linkiotookiUli�­
bollrd; th� model, illnstratcd �-""' ma,lo of .trip, nr pi- cut from tho
t..h of p,1per pad._ The ehoi� of ot""k J�pe,"l.o upon th� ""J.llir,,d
,tiff,__ A phototrimmer i, a oonnnient but not ioJU!peo..,ble aJjuncl
toi..,;_,,... Thu,nbta<kotore uacd for tumih� cor,,,<'<tio,,. l><'hff!<n link.,:
riJi,,. in a ofrai�hl olot (u,..,hMt.tl �uide), a thumbt""k fouction• al..,
"' c,-h...J. D,pe,,ding npon the <ituotio,i. thumhta<k point• may
be up or <lo�·o, Wheo a point otidu, up, the com,-ior, may lo, ,.,we
morei _u,.., byi p...,,,.ingi ai cireulari puncl,i11gi (orrlinaryi holei punch)of
11111,oilai folderi ,tocki o,·e,thei l>()int, clippiu�ithei haurdou1e.\"'len1th,
andi..,.,mingtheipunchil,Kitoi thci rcmaiJ1tk-rioli thei pointi ,..jthi a dabiof
quick-<l,yinK cem�o,\ (,11eh .. Dueo)
Although it ,.- ;n l,e in,ttneti,·� to l,uild .om• rno,kl, ;,. tl, e form
<JI a kinematie chain (Fig. 3--1), moot intereot e,,uter-. in 1Dtthani,mo,
i.e.,i el1&i11,i ,.·ith .-..-,i fo.-di link. Thl.oi li11k i,i fi,.,Ji l,y "'"''"tine ,ti on a

""'""3--4 r.,.,.i,,,i,�.;"'"'imn,uat,tl,-(1,..holiro,,d,1..," il) I , < , + ,

11Mo"G,..oo(ilou,-bo,"'ii,,.i••h;,!,i tl,-,hDrln1ilio< N,,_t,..i,..m,.i.t,!J,i•·i1h...,,-tw
ito.,.;pbo.--.. '11) l + • > P + t.ionly<lou_,_k.,...,l,•oio- ,-.;hl,. c l l l )
+ ..io..,11 raoo of ch•o. . ,..,.., . (II') l + • • P + t. .,... Ill
:,. ;�.:;,:;-i..
 KINEMATIC MODELS 77
"breadboard" or base of some sort. A very convenient base is a piece
of ½-in.-thick insulating board, cut to 8½- by 1 1-in. size. If this is not
available, several thicknesses of carton stock wiJI also serve. Such
materials accept thumbtacks readily and also permit their easy retrac­
tion. A piece of paper, on which the frame points and other pertinent
data have been entered, may be stapled over the base so that coupler
curves may be drawn on it, etc.
The previous remarks give some idea of what may be involved
i n the construction of planar models. The problems incident to various
situations-sired by necessity and born out of trouble-cannot be dis­
cussed in detail, but solutions will present themselves. There will be
annoyances, such as buckling links, the need for reconnection ·in order
to have one link pass over another, sliders that stick or are too loose, and
so on. The need for some debugging should always be anticipated. A
little patience and a light touch \\>·ill help. The model will be a pretty
thing to see when it runs; not the least of the pleasure is to note the
amount of quantitative information that is furnished, in addition to the
basic qualitative data.
The four-bar linkage and the study of some of its moods are ideal
for first models. The four-bar is perhaps the most widely used linkage,
for despite its apparent simplicity- only three moving parts- it manages
to conceal a great sophistication. To start with, there are of course as
many mechanisms as there are links, a different link being fixed to define
each case. In addition, the link proportions (lengths) or parameters
control the extent of the relative motion bet\veen any t,vo connected
links. We may have a particular interest in the motions of the two links
connected to the frame. Confining our attention to the four-bar linkage,
we shall call a link able to rotate continuously in the same direction a
crank; if it can only oscillate, a rocker. The remaining moving link is
the coupler.
Possible combinations of frame-connected links yield crank­
rocker, double-crank, and double-rocker mechanisms: we shall call these
different kinds. Criteria for the various situations may be studied with
the aid of the Grashof relation.
Let us identify the longest link as l, the shortest as s, and the
remaining two as p and q. The fo1lowing relations, stated without proof,
are valid:
1 . A Grashof four-bar is one in \Vhich the sum of the lengths
of the longest and shortest links is less than the sum of the
lengths of the other two links, that is, l + s < p + q. Then
a, b. Two different crank-rocker n1echanisms are possible. In
each case, the shortest link will be the crank, the frame
being either adjacent link.
78 KINEM A T I C S Y N T H ESIS O F L I N K A G E S

c. One double-crank (drag-link) 1nechanism results when

the shortest link is the frame.
d. One double-rocker 1nechanisn1 is formed when the link
opposite the shortest is the frame.
2. If l + s > p + q, then only double-rocker mechanisms result ;
there will be four, depending on which link is n1ade the frame.
3. If l + s = p + q, the four possible 1nechanisn1s are like those
of (1), all, however, suffering fron1 a condition known as change
point. At the change point the centerlines of all links become
collinear1 whence the cranks n1ay change direction of rotation
unless given guidance.
4. a. The parallelogram linkage is a special case of (3); all four
mechanisms are double cranks if they are controlled through
the change points.
b. The deltoid linkage is the other possible special case of (3).
In this, two equal short links are connected to two equal
longer links, and in that order. With a long side as frame,
a crank rocker is possible ; the frame as short side may give a
double-crank mechanism, in which the short rotating link
makes two turns to the longer link's one. t The uncertainties
of these mechanisms are related to getting past the change
points.
::\[echanisms to the foregoing proportions may be investigated on
the drawing board. Greater satisfaction is derived from a set of models,
for which some convenient dimensions are suggested:
I Links 1.5, 3, 4, and 5 in. long
II Links 2.5, 3, 4, and 5 in. long
III Links 2, 3, 4, and 5 in. long
Linkages made to the above dimensions are shown in Fig. 3-4.
In addition, there is a parallelogram linkage. 'fhe links may be con­
nected in any order, for the relations of l + s and p + q are invariant
with respect to the link sequence. The models sho,vn can be held in
the hand, although it may be desirable to thumbtack the link chosen as
fixed to the breadboard.
Another example involves the most spectacular feature of the
four-bar linkage, the coupler-point curve, familiarly called the coupler
curve. A coupler point is any point of the coupler plane. We recall
that any actual coupler is only a discrete part of the plane-as much as
may be useful or necessary. Only two points of the coupler plane have
rotational motion ; all other points are said to have planar motion,
treated as a combination of translation and rotation. The coupler
1
This is what the Germans call a Galloway mechanism, after the Engli!'Jh
engineer who patented it in 1844.
 eurv.,,,arnilraoo<l hyioi roupl..-poi,o( on tho p1ane ufilhefrome link; in
our � th;, i• (he link tac�eJ to the brudbrurd. Th°"" ou"·.., of
U,e filth on:ler in the �neroJ """"• may h•se v,ry ouriou• ,r,.,_pe,, in<Je,,d
J'mtioD11 of \b-, oun·""---""'°'" o,I partioulor ■hap,,-.,,u, oft.,,, be
o�ploital to � u... .. we •h•ll 11tt. Sp«;,.J fou,..bar linhge oon­
fiiluroti<>Da 'fill l{ive fourth- •nd ...,ond--ord<r eurv""
Theibeot-kno,miex.ompleio!theiuoeiofithcirouplerieurveiioiini the
meochaniamdevioed byiJo,.,..,.Wott.theioo--e&lk,dparalklimotioo,ialroa.dy
,.,.,.•• ._, ,.,,..,.t_ ,.;,.w. 1.. •l •rt"-"'- -••- - • tro1.,. ,.-•.,,
/1) ..,•.•_ ..
,.... --· --- . ,_ ........ ... .........., ...., ti

�� ;·;::•.... __._;��t),.;�, �:·���!,.;":)

C-plw-,..;., ,,.,..• -r �100 Jan ..,,,.,n, • .,, ..i,uti•�
0;,..•••i•- lf ..Ja.. ,..,..,.t,...iaroip,opiti«,o.io ,i,-io,,i.,.,.f, onio••
hnir1& •" AJJ•nin•N ,,...11 ill do• •IJLII Matioa ,._. l,o .,.;-1
fl,o ;•• -, 1,. ,l,·rnn•ratool b7 pn•rwi•• o -,1... -1 ., ia., a ,,. .,1,
of ll in., u, • /-11�-..·•r .t I ia., •l"'•in1 tl,o loo& n,· o ., in. opon on ,1,.
N• -..·itl, ,,., ,,. .,l ,. 11,. 1..rt. T-..·o -it."" ,,n,., •• -•· a ,trip
. iJ ia. 1...i. ., a ot,;p � lf io. 1.... 0..i .., ,1 .. ir ,._. .- 11,,
U,ooili-k •-•i•• om,k ••• ••�lor H<I •• - ol • •• tM •-•
hot••" folloo·u oN •npl•. 1'ho t-..·o frw .... ofi•. ., o oknlt l.­
N llio •1<Wo of tlio liokof0 - ••nnootal •itO lloo poio.t ol a ,...,.ii.
rl,o ,-ooil poinl io tt.oo o -pl« J•Jint -..-h- nrn -W k,o t-..i
(fii.Hl,il.fll
 rhc uprer portion of tho """"' rontain, • n<'tlcii· cir<ular are
.,.,,tK>t,. '11,"e"eo,Mr" ,,thi, •n: U!e,k1<rluinNl ,,ilh & po.ir ofedivider,,
the•n,eleu,;tht.,;,ll!e,-.,.lrict.d toe•OO<lt t in
Thi,cent.,- i,eLo�led R, •"d • li11k ioemade to tbi,elen,;th. On,•
ond of tho link i, ,tuck o,·cr • tac� inoertod al the �pier point; the
uthe, euJ i,, �;,.,.,J tu (l,e boo.rd al ff . Thi,e,,,... • £vc...J,ar li..hge,
movableeafterea luhiot, o,·erethe a,o lehgth, and LinJin� toeeithereoide
 """,.;,"', •""•p/,,,j,al
""�.. u r...,,,pi.,, o1,,..,w ...,1,••;,,... ..;i�,.,,.,,....,. ,1
...... Tho U,puue.,. U, All -•e,.,.,.,,.,_..,.,..., at U,, knob. Th, 11,W .,.
or,-in.ebnu;.. tod. (o)eAne/K:NC-........ n..,.,,,.,,.,_..,....i ......i.u..
.,.1e-illa,.,., 8uo of t- •f hy 8f ia. (•JeAaeRIJGR-i........ Thcou>pul
7
� --:;::e' �°'.:-" c.!e,e:.!::e"'e �•� ·���;:•••;.,,,, Thoou,,.,_;,
i!;

of it Theemo,,.bility ' (lheom,cally irni-:-ihlo}eaco,u,,.efrometheptay

i n ihe conn«tio,,. ""d deform•t""' of th• linko, .;.,.., th• aro io no\
trulydmilar
A ,i:lth link, My 4 in.eJon1,0l p,-,�. Thelack at R i,r,,·en,od
andeonee<end of thoe,o,-..elink i,,eroi,.,,.,,Meto1l,o"'Ji""'liuk. Th•oih<r
•11deofeth• now 1;.,k io pinnodtoelh• 1,.,.roteote,.,,m,eoonl'enion\ locatio,,;
it.emotion.,illeh<,heeootputeoflhe n...,haniom, Onemo,ingethemttha
niltm th�b a full oyrle, the '"''P"' link •·ill t,,, """" to ha,� ou appro•i
• Mo..bS;iy;, d- i• i,,,e. �
84 K I � EM . .\T I C S Y N T H E S I S O 1'' L I N KA GE S

mate dwell while the coupler moves over the approximate circular arc
and the crank rotates continuously. This dwell period, like the approxi­
mate straight-line motion, may be good enough for many purposes.
Since four-bar coupler points generally describe sixth-order curves,
their curves will possess neither truly straight lines nor truly circular
arcs-but they may be satisfactory approximations. A different d\\rell
mechanism may be constructed by utilizing the "straight" portion of
the curve.
· Other examples of planar mechanisms are shown in Fig. 3-7. The
method of coping with sliders is indicated, and it must be admitted that
the simple-harmonic-motion device has an action that leaves something
to be desired. The Atkinson engine model (upper left) unravels the
mystery of how four piston stror.'38 per crankshaft revolution may be
achieved by having the piston rod driving along a coupler-point curve.
Other situations that also seem confusing when seen as only drawings,
such as the Nordberg radial-engine linkage (upper right), and even the
conventional radial engine (coupler curve again!), are clarified by models.
A moving link of each of the last two models was made of l\1ylar to permit
a view of the curves traced by points of the moving link.
The last three models ,vere constructed on pieces of manila folder
cut to 8½ by 11 in.; their format allows them to be carried in notebooks
and stored in file folders.

3-4 S P AT I A L K I N E M A T I C M O D E L S

To put it some,vhat inelegantly, spatial linkages are those taking

motion around a corner, whereas planar linkages merely transform
motion in a plane. In leaving the plane and adding an effective third
linear dimension to the links, the forces at the connections of the no,v
nonflat links give rise to bending moments and torques. These moments
and torques require that the links be stiffer to maintain the geometrical
truth of the connections. Generally speaking, cardboard links will no
longer do unless they are formed to give substantial moduli of inertia;
and neither is the cardboard convenient for t,visted links or the fashion­
ing of cylinqric and spheric pairs. In consequence, it is usually best to
use metal, whose choice of form-strip, rod, or tube-depends on the
method of fabrication and the nature of the connections between links.
Spatial mechanisms of less than seven links will contain cylindric and
spheric connections, both more difficult to make than the turning and
sliding pairs of planar mechanisms, and metal links lend themselves to
this sort of thing. rfhe models of Fig. 3-8 sho,v some solutions to various
problems.
It is evident that such spatial models do require more of a shop
 than ioeootJodefo,eth,,e..,.,;truotiot,e<J( pla""' model, lluteitei,e,I..,
obviOW1eUialelho-,yeoliop,quipm.n> i.e1ninimaJ.
The b.oeoi -.. cut from packin,:......, lumt,,,,-. "l'l,e ,h&ftin,: ;1
l,ruingeN>d ; etheeflate,t,ipoe<J(eth,,eoylin,hicepoi,.ehapPf"letoebeebt-.
It wa, M!lveni,n, to mok� ,o,ne luteo.u,l(S ,-:ia 1h,...,i.,.1 N>d ond. and
,appodholeo. Theb.11"of.theeophe,K>llejoint!earenwlcolepia,t.ieet,,-,.d,
kolOWll.e,.. Pop-it, availablee., ,,..u "-"P'-"'""· l'op,, dipo ool,WNJe..,
theecanlof.,,,ntho-1,;eu. TI,oi..putohof,-.,.e,_,�;,,.J toe,no\ion
of ,x�yerototioo, bye1..-oot,.pk,lefiltin,:intoegrooveo in theeohaf"'- Sere,r
eynoeouppo,tin•etheeoutputelinkoep,er,nitethee,imultanevw,et,anolati<>n
,ud rototiOfl of. the N>d "°""t,tutin• "" elen,...,t ol the <}·liudrN". output
�"
Some •i-tiol .,...., J...,J,mo ..-hooo linU ,,.,, "''"""'"e<I by only
t,irning ""i,._e111cb .. tbeett<nnett.e11,aro, ,wl f".oldl><,•,em..oy bee,...<1o
d,e•r,e:-iu.eoh<,e1-m,taie,t•i l'• ou• .. hlvetw°"""'e•"doon""'t<,,lhyeme<al
eyelet. (fip. :J..D ,.,d ;1-10). Eyel<I kit, and ..,1ia11 <ool (, ,,,.,1., name
;, Driu)em1yeheefouo.deiuehal'"0..-,,., .....,..e.,,,1eBOtionoecoonteroeoledion;
''°""· Thoe"''" of. t!, c turnin.g po.;,,. ha,-.. b<en emiohU.,o,! by oolderm•
wi...,. into ,he ew,!ct,, n,, ,.,min• poi,. of •�h«icol linka� �110,· ,

' . .... 't,,µJe; ... -.. .e • +·, ,.,.1,e- ,..e.... o1 , , 2 w1 ,. ·e,

"""""eo/eJOe..,d W". (R;,1,1) A ('-,ldt..,, 5n<>I' -h>oiao,, Ak,mmome,t,;,­
'-' f-"- "ldth.n-•• ...J....,.,, .... ;-. ,,..1. - ,;..-,,: 1,,- ot-.t gj b,- Jj i•.
llimpki eardboardi oonotructioni WJ&S.i 12-llior>di lZ-13). !tiUli .-e!l,ioft<,r
loyin1i outi lbei e,entnr.li � ol tboi linkagei thoti d,:6,,.i ,b,oi _..,,._i to
,utthoi-=toni f..,.io..Jhill#themi.-ithi mul<in�itapo(noti...llopha,,.
tape). Thi,,provitk,,oilreern"'->On,i.-ithootith,biaoiolj..t oifold.ian,t
ollo,.olayi n,i: iniwi- l-Oicallioutthoi._..,. ol rott.tioo.
Spherical ;o,,,1-1i ofi higher ,,uality than tho t-dond p,oper-<lip
typo an, a,-ailoble ;,. • '""'""Y of oi-, padeo. onplar motioo, ar>d rod
•nda. Th,,i, - ;,, o!,own ini Fi1.i IZ-lt
Fo,i w,ryi -Wi n>Odob• ..,.h ... tb<.i net<k,di fo,i quaotitati,·e
veri6<a1N>D.i rom....,..Wlyi o,-.ilahloi ..,.,,pon,nt.oi mayi olt<,ni bei .-li to
l(OOO od, ..ntqe. � ntiliation of po,ru onch _.. i-r,"-11 ouppo,to.
bnri np. olwti "l[, ..,.urot<ly divided dWo, and ,..,ro� 1..__ only wch
thi np .. aanbforil<>eoli manufO<'tur,,i(fl,11. IZ-16)
Tho fo«g<>in� n,ot<,,..,lo a"d motho<h ..ei.t;,,_ l-0 model "°"""<U¢­
ti<wn haV<1 '-n -•W ui..._,iona. C".,.,_,,inlyi oth,,tima1er..,ba11<I
t<,ChniqUMi willi 1P,..,i «,ua.llyi •liBfacto,y.i o,i ,--.,,i hM.,..,,i ,_,lta. Tho
pri1><ip,o.Jiid.ot. Ul l-Oibuildmodebondnoteth,imotl0fleha.-.ct<,riat1,-th,3·
.,.,,apo,W,,olid!:1D0081nil�

.,.,... i-,0i .........i ou....,i _h.......i........i.. ...., .,,.....,...i,:,1,_i !lloo ot

_ ,, i,,- 1 i..
 KINEMA T I C M O DELS 87

BIBLIOGRAPHY

Cundy, H. :.\Iartyn, and A. P. Rollett: "1Iathematical :.\fodels," Oxford Uni­

versitr Press, Kew York, 1954, 1961.
Farish, William: On Isometrical Perspective (1820), Cambridge Phil. Soc. Trans.,
vol. 1 , 1822.
Grashof, F.: "Theoretische :i\Iaschinenlehre," YOI. 2, Voss, Hamburg, 1883.
Hartenberg. R. S.: Die )Iodellsprache in der Getriebetechnik, VD/ Ber.,
vol. 29, 1958.
Kempe, A.: "How to Draw a Straight Line," 2\Iacmillan & Co., Ltd., London,
1 877. Also in E. W. Hobson et al., "Squaring the Circle," Chelsea Pub­
lishing Company, New York, 1953.
\Villis, Robert: "A System of Apparatus for the Vse of Lecturers in :.\Iechanital
Philosophy, Especially in Those Branches "Thich Are Connected with
::\lechanism," John "\Veale.. London, 1851.
---: "Principles of �lechanism, " 1st ed., Parker, London, 1841.
--: "Principles of Mechanism," 2d ed., Longmans, Green & Co., Ltd.,
London, 1870.
 KI N·E MATIC ANALYS IS
OF PLANAR MOTION

4-1 INTRODUCTION
Kinematics, the study of geometry in motion, covers two broad
and interrelated areas that are subject to separate study. An
existing or specified mechanism or hypothetical situation may
be investigated for ,vhatever characteristics and properties it
possesses; this is formally called analysis. The inverse of such a
procedure is synthesis: in this, a mechanism is created meeting
the specification of certain desirable characteristics and prop­
erties. lvlany of the operations forming a synthesis are directly
related to or make use of procedures stemming from analysis.

This chapter presents some generally useful aspects of velocity

and acceleration analysis for planar mechanisms, based on
geometrical (graphical) procedures. N' o previous acquaintance
with kinematic analysis is assumed; this allows a treatment in
closed form, but due attention is given to details and concepts
deemed important even to the mature reader ,vho may have
fallen into careless habits. Sophisticated procedures and special
methods for particular types of mechanisms are not a direct
part of the n1ateria] to be discussed.
KINEMATIC ANALYSIS OF PLANAR MOTION 89

4-2 COINCIDENT POINTS

The properties of coincident points are best discussed by con­
sidering a mechanism to consist of as many superposed planes as there
are links. Each plane is of infinite extent, and any physical link or
machine member is then only a portion of a plane. The superposed
planes of two mechanisms are visualized in Fig. 4-1.
The relative motion of one plane with respect to its neighbor is
dictated by the nature of the connection or contact between planes.
Thus, a revolute or pin connection ,vill allow only rotation, as for
example at On, B, and C in Fig. 4-Ia. A prismatic pair or rectilinear
slide, as bet,veen links I and 4 in Fig. 4-la, limits the relative motion to
rectilinear translation. Turning to the cam mechanism (Fig. 4-Ib), we
find a motion between link 2 (cam) and 3 (follower) that is a combination
of rolling and sliding. The latter would also occur if two links were
connected by a pin riding in a curved slot.
Consider the pin located at On (Fig. 4-la). Whether it is large

.. --
Straight
1 guide between
1 and 4

(a) The four planes of a slider-crank mechanism

Straight
guide between
1 and 3
(b) The three planes of a cam and follower mechanism

FIGURE 4-1 Coincident points.

90 KINEMATIC SYNTHESIS OF LINKAGES

or small, roller bearing or not will have nothing to do with its kinematic
function of limiting motion to only rotation: kinematically it is only the
axis about which rotation between links 1 and 2 takes place that is
important. This axis pierces both planes at OB; and this point, common
to both planes, is identified equally well as On1 or 0B2, points on links 1
and 2 at On, but clearly coincident or superposed. Each of these coin­
cident points is firmly attached to its own link at the axis of rotation,
superposition being always maintained. We shall call these permanent
coincident points permanent centers.
Consider next point E, shown on link 3 (Fig. 4-la). We shall
call it E3• Directly under it at this moment-or this position of the
mechanism-lies point E2 of link 2, and under this E 1 of link 1. These
three superposed points are fixed to their links at the locs,tions at which
we see them; with motion of the hnks, they retain their positions on their
respective links, but not with respect to each other, each going with its
link. �'lotion of link 2 will cause separation of the coincident points.
Each no longer superposed point will trace a curve on each of the planes
different from its own. Sometimes the shapes of the curves are obvious,
but not always We see that E 1 will trace a circular arc (about Os) on
plane 2; E2 will trace a circular arc (about OB ) on plane 1. These t,vo
curves are traced in opposite directions; i.e., if link 2 is rotating counter­
clockwise, E i's trace on plane 2 ,vill be developing in the opposite sense.
E2 will also trace a circular arc on plane 3. Ea will trace a circular
arc on plane 2 and some sort of curve on plane 1. The shape of this
curve-actually a coupler-point curve-is usually difficult to imagine. It
is of the fourth order, since it is derived from a slider-crank mechanism
(the corresponding curve of a four-bar linkage is of the sixth order).
Point E1 \\'ill trace a different fourth-order curve on plane 3.
It should be remarked that ,vhile our sketch does not attempt to
show it, there is also a point E4 busy tracing curves. Thus, E\ makes a
straight-line trace on plane 1, a circular arc on plane 3, and some kind of
curve on plane 2. And, of course, points E,, E2, and Ea leave their
traces on plane 4.
As drawn, Fig. 4-la evidences three coincident points at C.
Points C3 and C4 of the pin axis are identical in their actions \\'ith respect
to plane 1, tracing straight coincident lines on it. 11ea1nvhile C 1 is
tracing a straight line on plane 4 and a fourth-order curve on plane :1.
The motion between cam profile and follower point (D of Fig.
4-lb) is a combination of roll and slide, as ,ve have noted earlier. ,.\n
observer stationed at D 3 would notice the cam rolling about Da and
simultaneously sliding by it; i.e., the point D2 (fixed to the cam profile)
would trace some curve on the observer's plane 3. Da's trace on plane 2
is the cam profile itself. Here, too, the originally superposed points
move a.way from each other.
KINEMATIC ANALYSIS OF PLANAR MOTION 91
Coincident points of moving systems are thus seen to be of two
kinds: (1) those which remain permanently coincident, occurring only
along a permanent axis of rotation; and (2) those which separate on
motion, having been coincident only instantaneously, as at the moment
of a particular configuration.

4-3 NOTATION
The discussion of physical events is dependent upon a notation
of some sort, preferably associative to convey meaning. In the absence
of a universally accepted notation, we shall frequently use the symbols
noted belo,v and shall also improvise as necessary. The context will
serve to distinguish the meanings of symbols having several qualities.

a total acceleration (vector)

a"' a' normal and tangential components (vectors)
or' a6 radial and transverse components (vectors)
acor Coriolis component (vector)
a= lal magnitude of acceleration (scalar)
OcB acceleration difference of points C and B, defined
by ac = OB + OcB
A, B, C', ... points of a linkage, usually joints
0 point of zero motion
P, Q, R, points on Jinks other than joints
PQ vector representing the directed distance PQ
V total velocity (vector)
yr , v' radial and transverse components (vectors)
v = lvl magnitude of velocity (scalar)
Vea velocity difference between points C and B,
defined by Ve = Vo + Ven
I, 2, 3, Jinks comprising the chain or mechanism
ti ("tetta"), angle defining the -/J line of velocity
distribution
!) position vector, variable
p = IPI magnitude of the position vector
r radius of a circle, fixed
OBB, BC, . . . distances
8 angle of position vector with respect to reference
line; p and 8 are polar coordinates defining
the position of a point
x, y rectangular position coordinates
"' angular velocity
0: angular acceleration
92 KINEMATIC SYNTHESIS OF LINKAGES

Consistent units are implied, ,vith radians for angular measure

unless otherwise specified.

4-4 LINEAR AND ANGrLAR VELOCITIES

Afotion refers to an event requiring a certain amount of time for

its completion. lVIotion includes at least three distinctly different yet
related characteristics-displacement, velocity, and acceleration. To
these may be added higher derivatives, such as jerk and those follo,ving.
To describe the motion of anything, we need a reference system to which
we refer the motion or from which we observe it. In kinen1atics, any
arbitrary reference system will do, but, of the lot, one may be much more
convenient than the others. The motions of the various links of a
mechanism are generally referreu to a link called the frame, and the
motions with respect to it, i.e., relative to it, are called absolute motions.
i\fotions referred to a link other than the frame are called relative motions.
We see that absolute motion is merely a special case of relative motion.
When only planar motion is considered, as in this chapter, we
may speak of a reference plane. .Follo,ving the selection of a convenient
origin O and rectangular axes Ox and Oy fixed in the reference plane, the
position of a point may be defined by cartesian (x, y) or polar (p, 0)
coordinates; the angular position of a line may be given by an angle
such as 8 bet,veen the axis Ox and the line.
Displace,nent is defined as the difference bet,veen the position
coordinates of the final and initial locations. Dependence on only
position coordinates makes displacement a vector quantity; it is the
most direct route from here to there. Distance, on the other hand,
depends upon the actual route or path; it is not a vector, but only a
magnitude. For example, a man (point) driving from Chicago to
Ne,v York suffers a displacement expressible by the differences in the
latitudes and longitudes of the two cities. The distance traveled depends
upon the highways chosen and ho,v often the man got lost and had to
retrace or correct his route. Had the man been able to go "as the cro"·
flies," his displacement and distance ,vould have been the same. On
the man's completion of a round trip, his displacement v,as zero (same
position coordinates after as before), but the distance ,-vas the odometer
reading of his car.
The sa1ne ideas apply to the angular motion of a line. If a line
rotates directly to 270t°, then both angular displacement and distance
will be 270t°. If the line oscillates on the way to the 270t° mark, the
angular distance will be greater than the 270t° displacen1ent. When the
line rotates continuously in the same direction, the angular displacement
wilJ be zero on passing its original position; the angular distance is of
KINEMA T I C ANALYSIS OF PLANAR MOTION 93

course 360t°. With continuous rotation the angular distance mounts

by 360t° for each revolution, while the displacement can never exceed 360°.
linear velocity is defined as the time rate of change of position of a
point. Since this change of position or displacement is a vector quantity,
velocity is also a vector quantity, having the sense of the displacement
being taken on.
Angular velocity is the time rate of change of angular position
of a line; it is a positive or negative number for planar mechanisms.
Speed is the time rate of covering distance; it is not a vector, but
only a number, for it has all the varied directions of the path. Our
Chicago- Ne,v 1�ork driver ,vould have had an average velocity given
by dividing the displacement by the trip time, whereas his average speed
would have been the result of dividing the actual distance by the time.
The speedometer sho"·ed the instantaneous speed, which \Vas also the
rnagnitude of the instantaneous velocity. Angular speed is usually
given in rpm; ,vhen there is a designation of direction, as "1,800 rpm,
clockwise when facing the output shaft," we speak of angular velocity.
We consider the details of linear velocity-or what happens ,vhen
a point of any plane changes its position-,vith the aid of Fig. 4-2. The
plane of the paper is the reference plane on which the point B of the
moving link k traces the path s shown; the point changes its location
from Si to s1 in a time interval D.t.
We are at liberty to select an arbitrary origin and reference line
for the polar coordinates p and 8. We shall consider the path s to be
traced by the head of the bound vector e as the ti1ne t varies. The vectors
e(t) and e(t + D.t) for positions s, and sr will also be called position vectors.
For analytical reasons they must of course be functions of time t and
must possess at least a second derivative. We also establish two sets of
rectangular directions at Si- They are the r (radial) and 8 (transverse)
directions, related to the reference coordinate system; the t (tangential) and
n (normal) directions related to the path.
The finite vector change D.e (t) = e (t + D.t) - e (t) corresponding
to a small but finite time interval D.t is directed along the chord sis1. If
We form D.e(t) /D.t, we recognize this as the average velocity vector vs.av,
also lying along the chord. This says that ,ve get from s. to sr by pro­
ceeding along the chord D.e(t) at a constant (i.e., average) rate (speed
or magnitude of velocity) numerically equal to ID.e(t) I/D.t.
As the time interval D.t approaches zero, we have
. D.e(t)
hm -- = vs(t)
At-+0 D.t

the velocity of the point B at the instant t defining the position Si- The
Vector vn(t) becomes tangent to the curve at Si. To find the magnitude
94 K I NEMATIC SYNTHESIS OF L I N K A G ES

• .:
-..... ........

B,�,
I \'- .
s,
_·.
--·-- _ _. � p(t), chord
. . ·'. ·. · I • .• 1/ )· �
. I
\,,,, . \ ..
,� �-
'
I ._ ' /Radial, r
(' /
.. \\
.. .....'--/............
, :: , ' /// _ . . ..-- Normal, n
OB = jp(I + .1 t)I I - . .---
oc = l p <t>I / '- Si

I
I '' ''
I \_ ''Transverse,/1
I
4& ,
<
I
/� p(t)
-...:: ,4 \Tangential, t

' Path s of B on
�6' reference plane
' 11
\
0 - -- --- - - - Reference line

FIGURE 4-2 Velocity of a point of a plane moYing with respect to a.

reference plane.

of VB(t), we note that A,;,(t) ¢ 0 for small At > 0. We may then write
As -
- As IAe(t) I As
At IAe(t) I At IA,;,(t) I
In this, IA,;,(t)I is the length of the chord, and the ratio As/lA,;,(t)I = arc/
chord tends to 1, whence in the limit
ds d,;,(t)
IvB(t) I = =
dt dt I

1'he velocity vector VB thus has the magnitude lvBI = ds/dt, where s is
the arc length along the path.
Inspection of the figure shows that vs has no component in the
direction n of the normal to the path, since it is directed along the tan-
K I N E M A T I C .\ '.ll .-\ L Y S I S O F P L A N A R M O T I O N' 95

gent, lying ·wholly in it. However, components are seen to exist for the
radial rtand transverse O directions. We shall investigate these.
We take the vector fle(t) to be composed of two other vectors,
namely, fle(t) = s,-C + Cs1. On dividing by flt,
fle(t) s,-C Cs,
-----'- flt- = - +-
flt
= VB' aV
flt
That is, VB,av is composed of two velocity vectors each having a different
direction. This situation is maintained to the limit,
. fle(t) . s,-C . Cs,
hm -- = hm - + hm - = vn
.l.1-+0 flt .l.1-0 flt .11-0 flt

The vector s,C/flt has a direction perpendicular to the bisector of the

angle s,Os, = fl8. and its magnitude is
sic = i; 2p(t) sin (flB/2) I = J
(t)
fl()
flt At I i e At
the sine of a small angle being equal to itself (expressed in radians).
With decreasing At, AO also decreases, and, in the limit At - 0, s1 coin­
cides with s;, and the bisector of the angle coincides ,vith Os,. The
vector s,C/flt thus becomes perpendicular to the position vector e(t), and
its magnitude is le(t) fl()/Atl = le d8/dtl. This component of the velocity
is denoted as v,/.
The second component vector Cs.,/At has the direction of the line
Os,; its magnitude is the increment in magnitude (i.e., length) of the
Position vector during the time At, namely, le(t + At) I - le(t) I. As flt
is taken smaller aud smaller, Os1 approaches Os,, ,vhence the vector
Cs,/At lies in the direction of the position vector e(t) of location s.; its
magnitude is the rate of change of length of the position vector, or dp/di.
This radial component of velocity is ,vritten vn P.
In summary, the velocity vn of a point at a given instant is
composed of t,vo rectangular components,

,vhere the vectors v8 and v8P are as follo,vs:

8

Transverse component vn8

:vlagnitude: p dO/dt
Direction: perpendicular to position vector e
Radial co1nponent v8 P

1Iagnitude : dp/dt
Direction: along position vector e
96 K I N E M A T I C S Y N T H E S I S 0 1'� L I N K A G E S

As t:i.t approaches zero, the direction of the vector t:i.e approaches

the tangent to the path, ,vhich means that there is never a normal com­
ponent, namely, vB = vB and vB = 0.
1 n

It is important to note again that the velocity of a moving point

is with respect to a plane (not to a point in the plane); the existence of a
velocity requires a path. For example, an aircraft's velocity is reckoned
with respect to the earth, not ,vith respect to a point such as a city.
.. A particular case in ,vhich a moving point describes a circular
arc about an already specified point of the reference plane occurs so often
that special comment is necessary (Fig. 4-3a). Since the origin has been
arbitrarily (but by specification) placed at the center of the circular path,

on reference plane
_,,,.-,,- n. normal to
path ; also
radial r

Reference plane

(a)

(b) Link k pinned to link 1 at 08

4-3 Point of moving plane k describing circular­

1''1GURE
arc path on reference plane I .
K I N EMATIC ANALYSIS O F PLANAR MOTION 97

P =r= const, whence the velocity reduces to

dfJk
v»� = vn/ = r-
dt

As before, VBi lies on the tangent to the circle (path). The r and fJ
directions, although identifiable. are usually suppressed in favor of the
n and t directions, with ,vhich they coincide.
1'hus far ,ve have discussed the motion of a point ,vith respect
to a reference plane in general terms. We had no mechanism for guiding
the point on the path. ,ve may nov,r consider a specific and very com­
mon type of guidance, namely, that occasioned by a pin connection.
The physical situation is sketched in Fig. 4-3b, in ,vhich plane k
(link k) is pin-connected to the reference plane 1 (link 1 ) at On. Link
point Bk then describes a circular arc on plane l from whose reference
system x1y1 the (Jk of the moving link 1nay be measured.o. Then

We remark again that the above equation is a consequence of two cir­

cumstances, (1) the circular path, i.e., a link of constant length, and (2)
the fortuitous specification of the origin On. But even with all this, the
velocity of Bk is ,vith respect to the reference plane; we observe that On
Was convenient for calculation, the velocity being defined by one vector
component instead of the two that would have resulted from the specifi­
cation of any other origin.

4-5 RELATIVE VELOCITY AND V E L O C I T Y D I F F E REN C E

The concept of relative velocity is best understood by considering
a plane 2 moving with respect to a reference plane 1 . A point 1 called B
movi ng over plane 2 has an absolute velocity referred to plane 1, denoted
by Vn (Fig. 4-4). Seen by an observer moving with plane 2, the point B
has a relative velocity ,vith respect to plane 2 denoted by VB/2• Finally,
the point B2 coincident ,vith B at the instant considered, but moving
,vith plane 2, has an absolute velocity VB , 2

The relation connecting the three velocities-the absolute velocity

vn, the relative velocity vBn, and the absolute velocity of the coincident
Point vn2-is obtained by considering small displacements as sho\vn in
�ig. 4-5. Here, plane 2 moves from position 2 to position 2' during the
tnn e interval tlt. At the beginning of the interval, B 2 and B are coinci-
1 This point B may be either (1) a point of another plane or (2) a point
defined by a certain geometry, e.g., the intersection of two moving lines.
98 KINEMATIC SYNTHESIS O F L I N KAGES

----

\_ Path of B on plane 2

Plane 1
----mrmm1rrf)rmrm-,rml7T71!1,rm71,,mm,,11/J//17lJJll11llll

FIGCRE 4-4 Relati,·e

. velocity.

dent ; during the interval flt, B2 moves with plane 2 to position B;.
Point B, however, moves ,vith respect to plane 2 and at the end of the
interval will occupy position B' distinct from B�. Now, from the
definition of velocity,
BB' . 828� B�B'
vn = lim - = 11m (-- + -- )
.....o flt
.:U .:U-+O flt flt
. 828� . B�B'
- Ilffi - - + 1lffi - -
�-o /J.t A!-+0 flt

·
. 2
,;,:
\ /

\
I
)
I
/
/

........ _ --- __ .,.,,

Plane 1

r,
{flTT!II!/fl17fll! rtTlTfTl rrTflTT1lTIT1lIII!T777Tf
FIGURE 4-5 Proof of relative-velocity theorem.
KINEMAT I C ANALY S I S O F PLANAR MOTION 99

Plane l

FIGURE 4-ti Velo<'ity difference of two points on the same link.

But the first limit in the second equation above is simply the velocity
of point B2, and the second limit is the velocity of B as seen by an observer
moving with plane 2, i.e., the relative velocity of B, whence
VB = VB, + VB/2 (4-1)
The above equation thus relates the absolute velocities of t,vo coincident
points B 2 and B moving independently of one another. The relative
velocity vB12 is the velocity that point B ,vould appear to have to any
observer moving with plane 2.
Quite a different situation ,vill 110\\' be considered (Fig. 4-6)t: again
plane 2 moves ,vith respect to the reference plane I, but no,v plane 2 con­
tains two distinct points A and B with absolute velocities vA and vB • 'fhe
i•elocity difference bet,veen these two points is defined as vBA = vB - vA,
from which
VB = VA + VBA (4-2)
The velocity difference v BA will now be related to the distance AB
and the angular velocity w 2 of plane 2. Again, velocities ,vill be con­
sidered as limits of vanishingly small displacements, as sho,vn in Fig. 4-7.
The displacement of plane 2 during the time interval flt is considered to
consist of two steps: a translation from AB to A'B", follo,ved by a rota­
tion about A' from A'B" to A'B'. Then,
BB' BB" B"8'
vB = lim - = lim (- - + -- )
At-o flt At-o flt flt
AA' 8"B'
- lim - + li1n --
At-+O llt At-+0 flt
since 88" = AA'. The first limit in the last equation above is v A, the
JOO KINEMATIC SYNTHESIS OF LINKAGES

I
/

/�
FIGURE 4-7 Determination of the
Bisector of d (J _j velocity difference of A and B.

velocity of point A ; the second limit may therefore be identified ·with the
velocity difference of the two points,
B"B'
VnA = lim --
.1t-+O At
The vector B"B' has a direction perpendicular to the bisector of the angle
B"A'B' = AO and a magnitude B"B' = (AB) AO, AO being a small angle.
As At approaches zero, the bisector approaches the direction AB, and
B"B/At = AB/(AO/At) approaches AB/(dO/dt). Noting that AO is the
angle of rotation of plane 2 during the time interval At, we conclude that
the velocity difference Vs,1 is perpendicular to A.B and has a magnitude
equal to (AB)w2.
Because of the similarity between Eqs. (4-1) and (4-2), the
velocity difference is often called the "relative velocity" of B ,vith respect
to A . It should be noted, ho,vever, that the two situations described
by these equations are different, and it is misleading to identify relative
velocity (vn12) and velocity difference (v8,1) only on the basis of a simi­
larity in equations. We have already noticed that both velocity and
displacement are vectors and can be referred only to a plane (for planar
motion) in which a reference direction such as Ox has been defined. A
velocity cannot be referred to a point, since its direction cannot be
specified. The difference between relative velocity and velocity differ­
ence may be further explained in terms of an example. ...4.. person (point
B) pacing the interior of a railroad car (plane 2) riding through the
night ,vould be able, by simple observation, to determine his velocity
with respect to the car (vB12), for this is independent of the motion of
the car. It would be impossible for him, however, to establish the
velocity difference between the t,vo ends of the car, for this velocity
KINEMATIC ANALYSIS OF PLANAR MOTION 101

difference is the difference of two (absolute) velocities referred to the

earth, ,vhich the person cannot see; the velocity difference depends on
the motion of the car with respect to the earth.
Velocity analysis-the determination of the linear velocities of
points and the angular velocities of links of mechanisms-is conveniently
handled by the method of vector polygons. To establish the funda­
mentals of velocity analysis based on vector polygons, we shall examine
the velocity condition of the four-bar linkage of Fig. 4-8a, the angular
velocity w2 of link 2 being known. This will involve finding the linear
velocities of the several points, the angular velocities of links 3 and 4,
and the use of the velocity image.
Suppose that in our awkwardness we attempt first the velocity
of D. We would write the equation vv = VB + VvB and proceed to
evaluate the three terms as best we can. Each of the vector terms has
two endo,vments, magnitude and direction, each of which may be treated
as an "unknown" in the same sense as the unknowns of algebraic equa­
tions. Algebraic equations are scalars, and one such equation cannot
be solved if the number of unknowns exceeds one. A vector equation
(such as we have) may be viewed as consisting of two scalar equations,
Whence the limit of unkno,vns is two per vector equation. In algebra,

I Oc
(a)
Direc vDc

\\ o·1rec
(b) v08
FIGURE 4-8 Velocity analysis of the four-bar linkage.
1 02 KINEMATIC SYNTHESIS O F LINKAGES

as many simultaneous equations as there are unknowns are needed for

solution; with vector equations, only half as many equations as unknowns
are needed.
To return to the evaluation of the vectors, we shall examine each
in turn for (I) direction and (2) magnitude and shall hope that not more
than two unknowns exist for the entire equation:
Term v D both magnitude and direction are unknown
Term Vs completely known in magnitude (OBB)w2 and direc-
tion
Term v vB only direction is kno\\·n
1�his information is conveniently summarized on the equation itself by a
system of checks (kno,vns) and 1·'s (unkno,vns) as follo\\·s :
ZtXt vv VZI

Vn = Vn + VoB (4-3 )
With three unkno,vns the situation is hopeless, ,-.·hence another tactic­
another point-must be looked at, with hope for better luck.
Turning to the alternative point C, ,ve ,Yrite ve = VB + VeB•
Termwise examination allO'ws the equation to be marked as
xiv vv vzt
Ve = Vn + Ven (4-4)

Here x1 and x2 indicate unkno,vn velocity magnitudes. But since the

unknowns total only t,vo, the vector equation can be solved. We choose
.an origin 0.. and from it lay off the completely known VB to some con­
venient scale (Fig. 4-8b). The equation instructs us to add ve8 to VB;
all that can be done no,v is to lay off the direction of VcB- Only ve is
left; it is an absolute velocity and hence originates from 0,., going out
to meet the "other side" of the equation. On laying off the direction
of Ve the line intersects the Ven direction line. Since the tail of the
vector Ven must start from the head of VB-the cue comes from the
equation-the sense (as different from the direction) and magnitude of
each of the partially known vectors are established; i.e., vectors veB and
Ve are completely defined. The angular velocity of link 3 is given by
wa = VcB/BC. Similarly, w 4 = vc/OcC. The senses of the angular
velocities are deduced from those of the linear velocities.
With Ve at hand, the velocity vv becomes approachable, for we
may involve Vo in two equations, which, after being checked off, show
an acceptable four unknowns:
vv vx,
VD = Vy + VDB
..;..;
VD = Ve + VDC
K I NEMATIC ANALYSIS OF PLANAR MOTION 1 03

For clarity, the first vector polygon has been redrawn (Fig. 4-8c)
and the solution of the last two equations added. Common to both
equations, VD is shown by the broken-line vector Ov D'. The shaded
figure B'C'D' is known as the velocity image of link 3; it is geometrically
similar to the figure BCD that is link 3. 1'he line B'C' stands at right
angles to the line BC, the image having been rotated 90t° in the sense of w8 •
Note also that D' is on the side of B'C' corresponding to the location of D
with respect to BC.
The velocity image provides a great convenience, for the velocity
of any other point of link 3, such as E, may be found without recourse
to further equations. It is necessary only to locate the corresponding
E' on the image, and the vector O,.E' (not shown) will be v E-
It follows now, of course, that the velocity image of a link may be
drawn directly after establishing the velocity of two points of the link,
e.g., the points B and C. Had ,ve done this, there would have been no
need for the simultaneous solution of the two vector equations relating
point D to points B and C. However, there are situations in which the
procedure of simultaneous solution is necessary.

4-6 INSTAN'fANEOUS CENTERS O F VELO C I T Y

By instantaneous center is understood a pair of coincident points
having either zero relative velocity or acceleration at the moment of
observation. The instantaneous centers of velocity and acceleration
do not occur at the same point except in trivial situations. That two
Points of different planes have momentarily equal velocities or accelera­
tions provides a means for taking the known motion of one plane into
another.
While the position of the velocity center is in many cases intuitive
or is found without too much difficulty, the same cannot be said for the
acceleration center. As an analytical device, the instantaneous center
of velocity has wide application, and the following remarks will be
directed to the velocity center, sometimes abbreviated simply to IC.
Consider a plane 2 moving with respect to a reference plane I,
and assume that the velocities of t,vo points A 2 and B2 are known, as in
Fig. 4-9a. Note that, while the velocity of one point, say A 2, could be
chosen arbitrarily, the velocity of the other point B2 has to be such that
the velocity difference v8,A, = VB, - v.-1, is perpendicular to the line
4 2B2. Having chosen an origin 0.,, ,ve can dra,v a velocity diagram for
the motion of plane 2 with respect to plane I, as sho,vn in Fig. 4-9b.
�ow the velocity of any point C 2 of plane 2 may be found by locating
its image c; on the velocity diagra1n, similar triangles being used as
sho,vn in the preceding section.
104 KINEMATIC S Y N T H N
E SIS OF L I N K A G E S

Since there is a one-to-one correspondence between the points

of plane 2 and their images on the velocity diagram, we may also solve
the reverse problem by the same method and find which point C2 of
plane 2 has a specified velocity Vc2 • The specified velocity locates the
image c; such that O,C; = vc2 and a triangle A 2B2C2 similar to A ;B;c;
,

is constructed over the line segment Ae2B 2 to locate C2. In particular,

we may find by this method the point 02 of plane 2 having zero velocity
at the instant considered. This point is the instantaneous center of
velocities of plane 2 ,vith respect to plane 1 ; its image must be the origin
0., of the velocity diagram. The instantaneous center 0 2 is therefore
located as shown in Fig. 4-9, such that triangles A 2B 202 and A;B;o" are
similar. Note that this means that 02 is the intersection of the lines
A 2u and B 2v, respectively peroendicular to the velocities at points A 2
and B 2• Note further, by comparing plane 2 with the velocity polygon,
that the velocity of any point such as C 2 is perpendicular to the radius
vector 02C2 from the IC to the point, and the magnitude of the velocity
is Ve, = w2(02C2).
In a mechanism consisting of a number of links, an IC relates
to two links; with a mechanism of n links the total number N of possible
ICs is the number of combinations of n things taken two at a time, or
N = n(n - 1)/2. Thus a four- link mechanism will have an N of 6; for
a six-link mechanism, N = 15; an eight-link mechanism will have
N = 28; and so on.
Clearly some sort of bookkeeping system is needed when the

u
\

Plane 2

,,/
,,..-•''/

o,,

(b)
Plane 1

(a)
I-'IGURE4-9 Existence and properties of the instantaneous c-enter of two
planes in relative motion.
KINEMATIC ANALYSIS OF PLANAR MOTION 105
✓ ✓ ✓
12 13 14 15 16
23 ✓ 24 25 26
34 ✓ 35 36
I
I
45 ✓ 46
(a}
56✓ I 15
1 2 I
I
I
3
I
6 I
I
I 16 at co
C
5 4
(b}

56
6
;O
'- c
I
'-. '-. '-
I
'-/13
I --;:
I '-

16 at co
FIGURE 4-10 Determination of ICs of a six-link mecha­
nism.

number of links exceeds four. A six-link mechanism (Fig. 4-10) will

serve as example. The !Cs may be tabulated as in Fig. 4-lOa, each link
with every other, and checked off as they are found. However, the circle
�iagrain of Fig. 4-lOb has greater utility. In this, each dot represents a
li'nk, and a line between any two dots represents the IC of the two links
(dots) so connected. Starting with the dots ordered around a circle
(the latter for the sake of neatness), the obvious ICs are entered in the
�iagram by dra,ving full lines between appropriate dots.
·. · · · The obvious ICs are those associated with permanent connections
between then.liriks. The term permanent means pin connections and
aiso hicludes the infinitely distant IC of a slider, located on a normal
t�: .t4e direction of rectilinear translation.n. This idea is based on the
.
Philosophy that such translation is regarded as the limiting case of . rota-r
tion- about an infinitely distant permanent center of rotation,. a center
conimon to then.t\VO links bearing the kinematic elements of the· sliding
Pair. · . .
·
106 KINEMATIC SYNTHESIS OF LINKAGES

In our example (Fig. 4-10) the obvious or directly available ICs

are 12, • 23, 34, 4,5, 14, 56, and 16, the last being involved with translation.
The normal on which IC 16 lies is drawn through IC 56 because of the
implication of the theorem of three centers, to be stated below. Actually,
any line parallel to it will point to IC 16.
The remaining centers are found from deductions based on the
theorem of three centers:e1 The /Cs of any three links (connected or not)
having planar 1notion lie on the same straight line. (We leave the proof
as an exercise.) It is in the application of this theorem that the circle
diagram is more aggressive than the tabular array in suggesting what
to do next. Since a point such as an IC is defined by the intersection
of at least two lines, an unknown IC is specified if it is a line in the circle
diagram that is the common siJe of two triangles. Thus, IC 13 (shown
as a broken line in Fig. 4-1Ob) is rendered by the intersection of the line
12-23 with the line 14-34. Similarly, IC 15 is at the intersection of
14-45 and 56-16. The remaining I Cs are found in like fashion.
We may now distinguish between absolute and relative ICs. Abso­
lute I Cs are those carrying the frame numeral or digit; they are the top
row of the tabular array, Fig. 4-l0a, characterized by the zero velocity
of their coincident points. All other ICs are called relative; their coinci­
dent points have velocities different from zero, being in motion with
respect to the frame. The significance of these ICs is that the coincident
points on the two Jinks involved in a relative IC have equal velocities,
for example, V(23>. = V 23>a, V a4> = v(34)., etc.
c c a

It will be noticed that permanent ICs are distributed among both

the absolute and the relative !Cs. The use of the IC as a medium for
determining both velocity distribution in a given link and n1otion trans­
mission between links will be discussed in terms of a four-bar linkage
(Fig. 4-lla). A line drawn from 13 to the tip of V2a (which is also VB,)
defines an angle t'J 3• (The symbol .,J is the cursive O and may be pro­
nounced "tetta" to distinguish it frorn theta.) This .,J3 line gives the
velocity distribution along any radial line of link 3 emanating from 13,
whence the velocity of any other point, such as vc. = va., may be found
by duplicating .,J3 on a radial line from 13 passing through the point Ca.
Also, w3 = V23/(e13-23) = v3./(13-34), and w4 = V 3 4/(14-34).
Action at a relative center such as IC 24 may be visualized as in
Fig. 4-1 lb. IC 24, as remarked, lies at the intersection of Jines 23-34
and 12-14. The sketch shows planes 2 and 4 extended to cover the

• Read "IC one-two." If the smaller digit is always placed first, possible
dualities will be avoideda.
1
This is known as the Aron hold theorem in Germany, and sometimes as the
Kennedy theorem in the United States, after the independent discoverers, Aronhold,
1872, and Kennedy, 1886.
KINEMATIC ANA LYSIS O F PLANAR MOTIO:S 107

center 24; each plane, of course, retains its absolute rotation about its
own center, as 2 about IC 12, etc. From VB, - vn the rJ2 line for link 2
1s found and then set off on the radial line 14-24. The perpendicular
to it at IC 24 defines v< 24> ,; but since this is at the center 24, the coincident
point of plane 4 has a velocity precisely equal to that of its mate on plane
2, namely, vc 24> - vc24J, - V24. In this manner a known velocity of
t

plane 2 is transferred to plane 4.

The tJ4 line may now be established and then reconstructed with

(a)

/
/

(b)
PtooaE 4-11 Velocity distributions and {} lines of a four-bar linkage. (a) Cou-
Pier and frame, absolute IC 13; (b) crank and follower, relative IC 24.
108 KINEMATIC S Y N T H ESIS OF LINKAGES

respect to line 14-34 to define vc, = v3 4. From this w4 = V34/(l4-34).

We should note that we ,vere able to get from Jink 2 to lipk 4 without
involving the intermediate link 3.
The case of IC 24 is typical of all relative IC situations. The
action at such an IC can ahvays be visualized (if the need arises) by
considering the pertinent planes to be extended far enough to cover the
center, after ,vhich the {} lines are brought into play.
Relative centers allo,v one to go directly from one link to any
other, provided that the IC of the t,vo links is at hand. For example,
if ,ve had an eight-link mechanism ,ve could make the direct transition
from link 2 to link 8 via IC 28. However, IC 28 might be troublesome
to find: it could be the last to disclose its location [N = (8 X 7)/2 = 28].
Four-bar and slider-cr�11k linkages were at one time called indirect
contact mechanisms, since the coupler link 3 is the intermediary of links
2 and 4, one of "'hich is the input link, ,vith the other the output. :\1echa­
nisms in ·which the input and output links are in immediate connection
on curved or camlike surfaces, as in Fig. 4-12a, are now often called
"direct-contact" mechanisms. Other examples include meshed gear
teeth and many conventional cam and follower setups.
For the purposes of this discussion, the plane of link 2 has been
extended beyond the limits of the profile it carries, and link 3 lies over it.
At the point of contact, B, the t,vo curves have a common tangent t and
normal n. The point B is comprised of two coincident points B 2 and Ba­
The path that Ba ,vould describe on plane 2 is not a simple curve; as
the figure sho,vs, it does not coincide ,vith the profile of link 2, and its
shape and curvature cannot be found ,vithout some trouble. Similarly,
B2's path on plane 3 is not an obvious curve.
As links 2 and 3 move while remaining in contact along their
profiles, the instantaneous contact point B runs along the profiles and
thus moves with respect to both links 2 and 3. By application of the
theorem of relative velocities [Eq. (4-1)), the velocity of this point may
be written in two different ,vays,
and
from which VB/2 - VB/3 = Vn, - Vs, = VB,/2 = -VB,/3

Since B moves along the profiles of links 2 and 3, both v 812 and v n13 are
along the common tangent t and from the last equalities above we con­
clude that the relative velocities VB,; 2 and vn,13 are also along this tangent.
With vB,/2 along the common tangent t, the IC 23 must lie on the common
normal n, and with the aid of the theorem of three centers we can now
locate 23 as the intersection of 12-13 ,vith the normal n as shown in
Fig. 4-12b. The figure also shows the t'J lines of the velocity distribution
of links 2 and .3 with vc2a>, = Vc23>,:
KINEMATIC ANALYSIS OF P L A N A R MOTION 109

13

(a)

Y(23)2= V(23) 3
/ 3

(b)

ICs of direct-contact mechanisrns.

4-7 P O L E S A N D C E N TR O D E S
The instantaneous center of velocity, IC, of a plane 2 1nov1ng
With respect to a reference plane 1 ,vas defined as the point of plane 2
having zero velocity at the instant considered. The location and prop­
erties of this point "'ere deduced by velocity nnages. The purpose of
1 10 KINEMATIC SYNTHESIS O F LINKAGES

the present section is to reconsider instantaneous centers of velocity

from a different point of view, which will be useful for further develop­
ments in Chap. 8. This new approach is based on the awareness that
every displacement of a moving plane over a reference plane is equivalent
to either a unique translation or a unique rotation. For a rotation in
\vhich a finite angle AO is involved, the center of rotation is called a pole.
A succession of positions of a body may then be likened to a succession
of finite rotations about properly selected poles. The use of the pole
furnishes a powerful tool for synthesis as developed in Chaps. 8 and 9.
Let AB of Fig. 4-13 represent a line in a moving link. Successive
positions are shown as A 1B1 and A 2B2. In Fig. 4-13a the link is moved
from position I to position 2 by turning about the pole P12. This pole

B,

I I
A, \
\ I A2
\ I I
\
I�I 812 ,,..,,.-B
\ I · \, ,,.- 2

\
\
.,I
'.--IJ12--:-----..... /
�
I /
/
/

\ I I ,,,,.
\' III ,,.-- ,,.--/
1.
(a) ��,,.-
P12 (b)

FIGURE 4-13 The poles of finite displacements.

KINEMATIC A N A LYSIS OF P L A N A R MOTION 111

was defined by the intersection of the midnormals a and b, that is, the
J?erpendicular bisectors of the distances A1Ao2 and B1B2. The A and B
rays have been rotated through the finite angle A812, the increment in
angle between 81 and 82. The same situation, but without construction
lines and by now superfluous identification, is shown in Fig. 4-13b. Note
that A8 12, the finite angle of rotation between positions 1 and 2, is written
simply as 812 . It is understood that the pole ,vas found from the mid­
normals' intersection.
In Fig. 4-13c a third position AoaBa is shown, the pole P23 having
been determined from the midnormals to A2Aa and B2Ba. For con­
venience, the pole P12 is also labeled C12. Suppose the line P12P23 to be
rotated by an amount - 812 , thus defining point C2a, and that C23B 1 is
drawn. The line C12C23 may now be considered to be part of the moving
plane containing AB. When A 1B 1 is swung into A 2B2 about P 12, C12C23
swings down to become coincident with P12 P2ai figures C1 2Ao1B 1C2a and
P 12A 2B2P23 are identical, and so are, of course, the triangles P 1 2A 1B 1
and P12A 2B 2• With the move to position 3, the triangles P23A 2 B2 and
P2:1AoaBa are identical.
The foregoing allows us to take a different view of things. Al­
though we are really concerned ,vith moving A B from position 1 (A 1B 1 )
to position 2 (A 2B2), and although we have no idea of the mechanism
that is to do this, we begin to see how it might be accomplished: the clue
lies with the lines C12C 23 and P 1 2P2a- P 12P n was established from the
poles; these poles are fixed in the reference plane, and so is the line
P 1 2P2a. C12C23 ,vas found by a rotation of - 8 1 2 ; it is a line now attached
to the moving plane carrying AB. We are thus able to imagine going
from position 1 to position 2 by rolling the line C 12C2 a of the moving
plane into congruence with the fixed line P12P2 3. We still have the
moving plane rotating, but we think now of two lines-one moving, 011e
fixed-providing the action by coming into congruence and effecting the
position change of AB by a rolling action.
Figure 4-14 shows the moving plane in six successive positions;
the poles ,vere determined from the intersections of the appropriate mid­
normals. By joining the poles we obtain the beginnings of a polygon;
it is incomplete, since AB (or the moving plane) does not return to its
original position. This polygon is called the fixed polygon, for it is in
the reference plane.
Another polygon, but one that moves because it is attached to
the moving plane AB, must next be determined. The point C23 is
located as before; and, as we have seen, when C 12C23 swings down to
coincide with P 2P23, the link moves into position A 2B 2• The line
C2aC34 has the length of P23Pa• and makes an angle (23 + 823) with
1

C12C23, Note that the 823 added to 23 is of opposite sense to the finite
1 12 KIN E MATIC S Y N T H E S I S O J.t' LINN
KAGES

----- ----- A,

Jo'IGURE 4-14 Poles, fixed and moving polygons for finite rotations.

rotation 823 of the link between positions 2 and 3. This process is con­
tinued; thus (}34C �r, = Pa4P 4 5, and the angle with its predecessor line is
(34 + (}34), and so on. The moving polygon is created in this manner;
,vhen it rolls around the corners of the fixed polygon, the link AB ,vill
assume the successive positions. This is easily demonstrated by drawing
A 1B 1 and the moving polygon on a transparent overlay and rolling the
latter's polygon on the fixed polygon. We remark that these polygons
are independent of the actual method of moving the link A B ; all that
they manage to do is to position the link for a succession of finite rotations.
KINEMATIC ANALYSIS OF PLANAR MOTION J J3

Ordinarily the motion of a body is continuous. The six positions

of AB may be likened to six flashlight photographs of a continuously
moving link, the finite angles 118 = fJ corresponding to finite times 11t
between successive photographs. With our poles ,ve have been able to
account for the positions at the beginning and end of any interval.
Were we to think of photographing the procedure at shorter and shorter
intervals, we should have to provide more poles, until at the end, when
11t betw·een pictures approached zero, the infinity of individual poles
\\'Ould have to be called instantaneous centers. The polygons would
no longer have corners because of the infinitesimal spacing of the defining
points; instead there would be smooth curves with the special name of
centrode. The centrodes are actually the loci of the instantaneous centers
and will be identified by the symbol 1r (lowercase p of the Greek alphabet;
think of path). 1'hese curves, or centrodes, retain the roll property
of the polygons from which they sprang. By rolling the centrode
attached to the body about the centrode fixed to the reference plane, the
motion of the link is faithfully represented. Any point on either is the
instantaneous center for any virtual motion. We shall distinguish
between the curves by calling them the moving and fixed centrodes;
they are also known as the body (moving) and space (fixed) centrodes.
Figure 4-15 sho\\-'S a line AB of a moving plane 2 somehow guided
continuously along the t,vo paths sho,vn. The velocity directions (but

Path of A

-
Path of B
(v)
\Bz
Link 2
\1
\ /
Ba
(v)

�
\ \ I
�

\ I
�

\ I
To locate Ji:
�
I
. AA 2 B2J2 =A A1 B1 2 "" �Ii-- I
I'

",,
--
--� \ .,;;.1
.
To locate 1'3 :
1 :i
.
A AaBala == AA1B I 11 '/:-J
IV
,r2 , moving
centrode,

Fixed
centrode, ff'1
j'//}"'/
link 2

link 1
,,, J3

FIGURE 4-15 Determination of moving centrode.

114 K I N E MATIC SYNTH ESIS OF LINKAGES

not their magnitudes) are indicated by (v) at A and B for successive posi­
tions ; these directions are both tangents. The intersections of the nor­
mals from the pair of velocities at each of the three positions define the
instantaneous centers for each, namely, / 1, l2, and J 3. A curve drawn
through these points is then the locus of the instantaneous centers fixed
to the frame link 1 : this is the fixed centrode, labeled 1r1 for the link
(plane) to which it is attached.
It is considerably more trouble to establish the moving centrode
1r2. The use of a transparent overlay on ,vhich the line AB has been
drawn ,vill prove convenient. Upon laying this over A 2B 2 (position 2),
12 is marked on the overlay. The overlay's line AB is then laid over
A 1B 1, and 1; is located by pricking through the overlay's point I2• By
repetition of this operation at A aB 3 , at ,vhich Ia is picked up, and from
,vhich the overlay is moved back to A 1B 1, 1; is found. We may say
that the triangle .4 ,.B,. l,. must ahvays be returned to A .B 1 to locate 1:,,
the point 011 the moving centrode. The arc lengths, such as /�/� of the
tn()ving centrode 1r2, are of course equal to the corresponding arc lengths
I 11 2, just as the mating side of the former polygons were of equal length.
Without this equality of the vis-a-vis arcs no rolling would be possible.
It ,vill be instructive to make an overlay having on it A 1B 1 and the mov­
ing centrode 1r2 ; when 1r2 is rolled over 1r1, the various positions of AB
,vill be indicated.
A four-bar linkage (double-crank) is shown in Fig. 4-16, with
portions of the fixed 1r 1 and moving 1r3 centrodes. Successive positions
of the coupler are indicated by the nurnber of primes on each J ; thus, I"'
means position 3. We may no,v imagine removing the actual physical
guides (links 2 and 4) between the coupler (link 3) and the frame (link 1)
and achieving the proper displacement of the coupler by rolling 1ra on 1r1-
'fhe proper velocity is dependent on supplying the rolling motion with
the proper angular velocity. The two centrodes amount to connecting
the planest:� and 1 by a kinematic pair having but one degree of freedom,
viz., roll. The elements of this pair are the centrodes 1ra and 1r1.
In like manner, but by the u'3e of IC 24, two centrodes 1r2 and ,r4
could be developed. Since both centrodes are attached to the moving
links 2 and 4, they are denoted as relative centrodes.
Centrodes simulating any planar motion other than translation
rnay be found. With translation, the centrodes move to infinity. In
the double-crank linkage of Fig. 4-16, the cranks never become parallel,
whence the centrodes 1r3 and 1r1 ,vill be closed curves. 1�his is not true
for crank-rocker mechanisms, for the crank and rocker become parallel
at two phases, IC 13 going to infinity for each.
F .:,r some situations the centrodes are simple curves. When a
wheel rolls along a rail, the IC is at the point of contact: the moving
centrode is the rim of the wheel, and the rail is the fixed centrode (Fig.
KIN EMATIC A N A LYSIS OF PLANAR MOTION 115

2
B

ir-3. moving I

\ fixed, 'lrI �
�
:::J

["
�
Moving link 3

FIGURE 4-16 Centrodes of the coupler (3) with respect to the

frame (1) of a four-bar linkage.

4-17a). With tv,•o circular gears in contact (Fig. 4-17b), the relative IC
is always at the point of contact (pitch point) of the pitch circles ; the
two relative centrodes are the pitch circles. Also, as sho,vn in Fig.
4-17c, displaying a crossed parallelogram linkage, both the relative
centrodes are ellipses.
The direction of the common tangent (or the common normal)
to a pair of centrodes at a given instant will have a particular significance
in the studies of path curvature to be undertaken in Chap. 7. One
method to determine these directions, without actually dra,ving the
centrodes themselves, will now be presented as an application of the
Velocity analysis. The direction of the common tangent is along the IC
Velocity, 1 v1, whence our objective becomes the determination of that
Velocity.
Again, for the sake of simplicity, the example chosen is that of
the four-bar linkage OA ABOs, sho,vn in Fig. 4-18a. In order to apply
1 See page 196.
116 KINEMATIC SYNTHESIS OF LIN K A GES

the method of velocity diagrams, ,ve "materialize" the point I, IC of

link 3 with respect to 1, as the center of a pin riding between two forks
as shown, so that for all positions of the links this point is at the IC.
Assume for convenience an angular velocity wa = 1 rad/sec counter­
clockwise. With the assumed dimensions ,ve have

ll,t = IAwa = 2.33 fps Vs = IBw 3 = 1.74 fps

1'hen, using the theorem of relative velocities [Eq. (4-1)],

Vr = V112 + Vr, and Vr = YJ/4 + Vr,

v:e1 v'v vz, vv
or V112 ··l- v12 = V111 + vr,

The velocities v11 and vr, may be found by using t'J lines and the velocities
vA and v B as shown, so that both their direction and magnitude may
be checked as known. The directions of v1I2 and vrI4 are along the forks
of links 2 and 4, so that their directions may be checked. The remaining
two unknowns, the magnitudes of v,,2 and v,n, are found by drawing

r3 , moving
/

1r2, moving

,rl' fixed

I 1
(b)
(a)

r4 , moving

(c)

FIGURE 4-17 Situations for which centrodes are simple curves.

KINEMATIC A N ALYSIS OF PLANAR MOTION 117

>-
Centrode tangent _,, ·'­
"-.. .

IA = 2.33 ft
IB = 1.74ft

1 (a) l

FIGURE 4-18 Determination of IC velocity. 0.�011

= 3 ft, O..tA = 0.9 ft, A B = 2 ft, OnB = 1.3 ft.

the velocity diagram shown in Fig. 4-18b. An IC velocity of magnitude

V1 = 8.5 fps
is found. Its direction is that of the common tangent to the centrodes
at the instant considered.

4-8 ACCELERATION
The concept of acceleration is more difficult than the motion
aspects discussed so far. Position, displacement, path, velocity, and
8Peed are all qualities we can observe directly. We can judge the
118 K I N E MATIC SYNTHESIS OF LINKAGES

Hodograph

(a)

Oh
FIGURE 4-19 Definition of acceleration and hodogr11ph.

implications of these qualities because of our considerable experience

with them. For example, we are experts at integrating velocity vectors
and are alive only because of our successes in doing so. The avoidance
of collision between two moving bodies depends on an integration ,vhich
it is hoped will produce different position coordinates of the moving
bodies at all times, as in driving a car or crossing a street.
Acceleration is much more subtle and complex in its nature than
the velocity change from v,hich it comes. It may be said that our per­
sonal experience has been gained indirectly, for ,ve observe the effect of
acceleration, force, rather than the acceleration itself. For the layman
there are no instruments providing numerical values of acceleration, in
the sense that there are devices, as speedometers, that display informa­
tion about distance and speed.
Consider (Fig. 4-19a) the path traced by a moving point occupying
positions B and B' at instants t and t' = t + At. The velocities of the
point at the instants considered are v 8 and v� = vu + Av, ,vhere AY
is the velocity increment during the time interval At (Fig. 4-19b). The
average acceleration during this time interval At is
Av
.
OR av =
-
At
and the instantaneous acceleration, or simply acceleration at time t, is
. Av
= hmo­
AJ--o At
KINEMATIC A N A LYSIS O F PLANAR MOTION 1 19
The acceleration is therefore a vector representing the instantaneous
rate of change of velocity; its sense is that of the velocity being acquired.
A useful aid to the understanding of acceleration is the hodograph.
The path traced by a moving point is shown in Fig. 4-19a, with several
positions corresponding to times t, t' = t + flt, etc., and their correspond­
ing velocities. If these and all intermediate velocities are laid off from
an origin Oh as shown in Fig. 4-19b, the curve h joining the tips of the
velocity vectors is kno,vn as the hodograph. We may think of the
hodograph as being generated by the tip of a rotating and stretching
velocity vector. The chord between the tips of any two velocity yectors
of the hodograph is the velocity increment flv for the associated time
interval flt, and the "velocity" with ,vhich the hodograph is being
traversed is the acceleration of point B at the instant considered. The
acceleration of B is therefore at all times tangent to the hodograph at
the corresponding point.
1�he velocity increment Av may be resolved into the sum of t,vo
components, flvn and flv1, as sho,vn in Fig. 4-l 9c. On considering these
two components separately, the acceleration is given by
flyl flyn
an = lim - + lim -
&->0 flt &->0 flt

The direction of flvn is perpendicular to the bisector of the angle

AO between v and v', and its magnitude is Av,. = v flO, flO being assumed
to be a small angle. The angle flO also appears bet,;veen the normals n
and n' to the path at positions B and B' (see Fig. 4-19a). For a small
interval flt the distance BB' will be small, and the normals n and n' will
intersect at R, close to the center of curvature of the path, so that
RBt� RB' = p. Then considering triangle RBB', flO = BB'/p = v flt/p
and Av,. = v 2 flt/p. As flt approaches zero, flvn/flt becon1es perpendicular
to v, that is, norinal to the path, and its n1agnitude becomes v 2/P,
where p is the radius of curvature of the path at B. This vector, called
the nornial acceleration and denoted as OB.., always points toward the
center of curvature of the path and is zero when the center of curvature
is at infinity (rectilinear path, or inflection point).
The direction of �v' is along the velocity vector v', and its mag­
nitude is the change in magnitude flv of the velocity during the time
inte rval flt. As At approaches zero, the direction of flv1 approaches that
of v ; that is, it becomes tangent to the path, and its magnitude is dv/dt,
the rate of change of the speed of the point along the path. This last
Vector gives the tangential acceleration, denoted as aB'-
In summary, we have shown that the acceleration of a moving
Point is the sum of two· i·ectangular components,
120 KINEMATIC SYNTHESIS OF LINKAGES

o,.
r
Path of Hodograph
point B of B

FIGURE 4-20 Hodograph of a point moving at constant speed

around a circular path, Iva! = v (const), a" = v 2/r.

,vhere the vectors aBn and aB' are as follo,vs:

Normal acceleration ann
:\Iagnitude: v2/p
Direction: normal to path, pointing to,vard center of curvature
Tangential acceleration aB1
:.\!Iagnitude: dv/dt
Direction: tangential to path, pointing in direction of increasing
speed
'l'he hodographs of points having particular motions are interesting
to note. If a point is moving along a straight line with constant speed,
its hodograph is just a point, for its acceleration is zero. If a point
moves around a circle of radius r with constant speed v, its hodograph
is also a circle, but of radius v (Fig. 4-20). The acceleration of the point
is then only the normal ,vith a magnitude a" = v2/r.

4-9 R E L ATIVE A C C E L E R A TI O N A N D
C O R I O LIS A C C E L E R A T I O N

In dealing ,vith accelerations we shall follow the same pattern

as ,vith velocities and shall establish the distinction between relative
acceleration and acceleration difference. Relative acceleration will be
considered in this section, together with its necessary complement, the
Coriolis acceleration component. Acceleration difference ,vill be con­
sidered in the next section.
As in the study of relative velocity, consider a plane 2 moving
with respect to a reference plane I (Fig. 4-21). A point• called B moves
,vith respect to plane 2 with a relative velocity v1112 and a relative accelera­
tion a H12. This relative acceleration is that ,vhich ,vould be measured
As in the case of relative velocity, this point B may be either (1) a point
1

of another plane or (2) a point defined by a certain geometry, e.g .. the intersection of
two moving lines.
K I N E M A T I C ANALYSIS OF PLANAR M O T I O N 121
by an observer stationed on plane 2. It is related to the velocity of B
with respect to plane 2 and the curvature of the path traced by B on
plane 2. Point B, however, also has an absolute velocity vn and an
absolute acceleration an referred to plane I. Finally, the point B2 fixed
in plane 2 and coincident with B at the instant considered has velocity
vB, and acceleration a.B, ,vith respect to plane I. From the theorem on
relative velocity we have that

and one might expect something analogous to hold for acceleration.

This, ho,vever, is not the case, for \vhenever plane 2 has an angular
velocity w2 different from zero,
an = a n12 + as, + acor (4-5)
,vhere a cor is the Coriolis acceleration component. There is nothing
directly obvious about this component, also known as the compound
supplementary acceleration. The term supplementary is descriptive,
for the Coriolis component results from changes of t,vo linear velocity
vectors occasioned by the rotation w2 of the path that B traces upon
plane 2. The Coriolis acceleration has a magnitude 2vs 12w2 ; its direction
1s always normal to the path of the relative motion, and its sense is that
of vs,2 after that vector has been turned 90t° in the sense of the angular
velocity. Different situations are shown in Fig. 4-22.

Direc acor

Plane l

FIGURE 4-21 Relative acceleration and Coriolis acceleration.

122 KINEMATIC SYNTHESIS OF L I N K A G E S

��--+-Path of B
on plane 2
B �

\, -
... -· �•• - ♦--

PIGURE 4-22 Direction of Coriolis acceleration for different

cases.

To prove the Coriolis theorem [Eq. (4-,5 )) in its complete generality

(even within the realm of planar motion) by the direct methods of the
present chapter would be tedious and serve no useful purpose ,vith its
complexity . 1 The proof that follows is based upon a special situation
in which the center of rotation is fixed. As implied above, the general
case, in which the center of rotation itself is moving, will ans,ver to the
Coriolis theorem as developed here.
The situation shown in Fig. 4-2:3a represents a plane 2 rotating
about O» ,vith respect to the fixed plane I with an angular velocity w�­
The point2 B of block 3 is guided by the slot in plane 2 and moves with
respect to this plane with a velocity v 81 2 and an acceleration a 0;2- Posi­
tion vectors and velocity components are shown in Fig. 4-2:�b for two
positions of the slider corresponding to times t and t + 6.t. It is clear
that both the velocity components (radial and transverse) have under­
gone changes of magnitude and direction. These changes are best seen
by considering separately ,,vhat has happened to each velocity component.
The radial component is treated in Fig. 4-23c, the increment
!:,,.vBr = DF being resolved into t,vo components, a radial DE and a trans­
verse EF. �fhe diagram in Fig. 4-23d shows the resolution of the incre­
ment ll.v»6 = GJ into its components, a radial HJ and a transverse GH.
1
See Prob. 4-9 for a comph!te (planar) proof using complex number notation-
2 This point B is as.sociated with plane 3 and should therefore be identified
as Ba. However, the subscript 3 has been suppressed to reduce unwieldiness.
K I N E MATIC A N A LYSIS OF PLANAR MOTION 123
'fhe total change of velocity in the radial direction IS given by
DE - J H. The radial component of acceleration is then
. -
1Im JH
At->0 '1.t

By analogy with similar vector situations we may write

d d02
aBr = - VBr - VB8 --
dt dt
With vs' = dp/dt, v8 9 = p(d82/dt), and d02/dt - w2, this magnitude of
the radial component of the acceleration is
d2p 2
a n' = dt2 ··- pw2 = a 1112 - pw2
2
(4-6)

Link l(fixed)

8direc.
(a)
\
\
/ \

''\
'\
/ \
/
/

o,
v•'
B = v8
l + Avl
B
(c) Radial components of velocity at B
\6direc
'H
,✓"\

, dP
BVs = -
dt
o,
(d) Transverse components of velocity at B

(b)

F"IG UBE 4-23 Proof of Coriolis theorem in a special case.

124 K I N EMATIC S Y N T H E S I S O F L I N K AG E S

For this radial, or slide-path, direction we recognize aB12 as the

magnitude of the relative linear acceleration of B with respect to link 2 ;
it is a measure of the speed change of the sliding velocity that would be
present even if link 2 were not rotating. The pw2 term is the normal
2

acceleration of the coincident point B2, obviously a consequence of

rotation of link 2.
As for the transverse direction, the total velocity change is given
by EF + GH, ,vhence

or

In the () direction, ,vhich is perpendicular to the slide path, the

component pa2 is the angular acceleration of point B2. The first of the
vB1 2w 2 terms is a result of the direction change of the radial velocity
component ; the second VB 12w2 has to do ,vith the changing magnitude
of the transverse velocity component. Gathering the t,vo like terms
into one, ,ve realize that this supplementary term is compounded from a
sliding velocity along a rotating path. We write
(4-i)
Assembling the acceleration components to form as, we have
0B = 0B' + 08
11

= 0B 2 + 0B/2 +O cor

where ,ve have

Coriolis acceleration acor

l\Iagnitude: 2vs 12w2

Direction: rotated 90t° from vB12 in same sense as w2
It ,viii be observed that the Coriolis acceleration "·as derived
from the EF and G H components (Fig. 4-23c and d). A.s remarked,
these components are in the () direction, whence they are normal to the
path (and V1112).

4-10 ACCELER A T I O N D I FFERE N CE

'fwo points A and B of a moving plane 2 are sho,vn in Fig. 4-24.
The reference plane is 1, and the angular velocity and acceleration of
plane 2 ,vith respect to plane 1 at the instant considered are w2 and a2-
KINE MATIC ANALYSIS 01<' PLANAR M O T I O N 1 25

Plane 1

FIGURE 4-24 Acceleration difference of two points on the same moving plane.

With OA and os representing the accelerations of points A and B, the

acceleration difference of these two points is defined as

OBA = OB - 0A
from ,vhich OH = O,t + OB,1 (4-8)
The remarks made concerning the concepts of relative velocity
and velocity difference apply equally well to the relative acceleration
and acceleration difference. It is inconsistent to call asA a relative
acceleration, as is often done, because an acceleration cannot be referred
to a point but must be referred to a plane (in the case of planar motion).
Furthermore, since Eq. (4-8) deals with the difference of two absolute
accelerations, no component such as the Coriolis can be distinguished.
We emphasize that A and B are points of the same moving plane 2 ; they
are not points of two different planes, as in the preceding section.
The acceleration difference will now be related to the distance A B
and the angular motion of plane 2 as described by w2 and 0:2. Again,
�ccelerations are considered in terms of vanishingly small velocity
increments. During a small time interval Ll.t, points A. and B are assumed
�o move to positions A ' and B' as show·n in Fig. 4-25a. The diagrams
1 n Fig. 4-25b and c show the situations so far as velocities are concerned
at the beginning and end of the time interval ; or in equation form we have

VsA = vs - VA and v�A = VB' ,

But v� = VA + ilVA v� = VB + AVB , - VA
+ AVBA
v»A = VsA
126 KINEMATIC SYNT H E S I S O F L I N K A G E S

vii'

--�·• ---

o.,
B

(a)

FIGURE 4-25 Determination of the acceleration difference of .4 and B.

where AvA, AvB, and AvBA are the increments in the velocity of A, the
velocity of B, and the velocity difference between A and B during the
time interval At. Then

1'he acceleration difference, which may be ,vritten as

. Avo Av.1
OBA - hm - - Jim - -
.a.i-o At .a.t-+O At
AVBA
is then OBA - lim
.a.t-+O At

rfhe velocity differences VBA and v�A are shown in Fig. 4-2,3d, ,vhich also
shows the increment Av8A as the sum of two components (AYBA)n and
(AvBA)'. The angle AB shown in that diagram is also the angle of rotation
of plane 2 during At. Bypassing intermediate steps which are now
familiar, ,ve state the final result: the acceleration differen.c·� is the sum
of two rectangular components,

OBA = OBA n + 0BA

1
KINE M A T I C ANALY S I S OF PLANAR M O T I O N 127

where the vectors OBA and OsA' are as follo,vs (Fig. 4-24) :
n

Normal acceleration difference OBA n

}\:Iagnitude : vnA 2/AB
Direction : along AB, pointing from B toward A
Tangential acceleration difference OBA'
l\Jagnitude: (AB)a2
Direction: rotated 90t° from AB; in the same direction as a2
As an example, we shall investigate the acceleration condition of
the slider-crank mechanism of Fig. 4-26, assuming the angular velocity
w2 and angular acceleration a2 of link 2 to be kno""n.

(b)

(a)

Oirec ak

(d)
FtouaE 4-26 Acceleration analysis of the slider-crank mechanism.
·128 KINEMATIC SYNTHESIS OF LINKAGES

It ,vill be convenient to employ scales conforming to the relation

ka = k.,2/k•. * The velocity polygon (Fig. 4-26b) is then constructed
from the single equation Ve = VB + VeB and the velocity image. With
point D' at hand, the velocity differences v»B and vnc are directly
available.
Full possession of the velocity vectors allo,vs the graphical deter­
mination of all a" components at points B, C, and D, as shown in Fig.
4-26a. Only aB' needs external calculation, as by slide rule.
The diagram in Fig. 4-26c shows the vector polygon of the equation

v:r.1 vv vv VV v:r.:
ae = OB". + OB
1
+ acB" + ac,/

in ,vhich the magnitude of ac and the direction of acn1 are the unknowns.
After completion of the polygon, the absolute angular acceleration of
link 3 is calculated from aa = acn'/BC (clockwise).
The acceleration of D can be established from the simultaneous
solution of two equations in four unkno,vns,

.ri:ra vv vv vza
av - aB + Ona" + OnB'
ZlZ% VV VV VZt
CD = ae + ave" + anc'
For clarity, the first vector polygon has been redrawn in Fig. 4-26d and
the solutions of the last two equations added, defining the homologous
point D". The vector OaD", common to both equations, is the accelera­
tion of D; it is not shown because of the already many lines. The
shaded figure B"C"D" is kno,vn as the acceleration image of link 3 ; it is
geometrically similar to the figure BCD that is link 3" Note that D"
is on the same side of B"C" as D is ,vith respect to .BC. The acceleration
image has been turned through an angle 1r - /3 ,vith respect to its link
in the sense of a3 ; {, = arctan (a3/w3 ). 2

In use, the performance of the acceleration image is similar to

that of the velocity image: the simultaneous solution of the t,vo vector
equations for on could have been avoided. Thus, the acceleration of
point G is given by the vector OaG" of the acceleration diagram (Fig.
4-26d), since G" is the homologous point of G.

* The scale factors are defined in consistent units, e.g.:

I in. on drawing = a distance of k, ft on real mechanism.
1 in. of velocity vector = k., fps.
1 in. of acceleration vector = k0 fps 2•
KINEMATIC A N ALYSIS OF PLANAR M OT I O N 1 29

BIBLIOGRAPHY

Beggs, J. S.: "l'viechanism," 1:lcGraw-Hill Book Company, New York, 1955.

Cowie, A.: "Kinematics and the Design of �Iechanisms," International Text­
book Company; Scranton, Pa., 1961.
Ham, C. W., E. J. Crane, and W. L. Rogers: ":Vlechanics of l'vfachinery," 4th ed.,
::\fcGraw-Hill Book Company, New York, 1958.
Hrones, J. A., and G. L. Nelson : "Analysis of the Four Bar Linkage," Massa­
chusetts Institute of Technology Press, Cambridge, Nlass., 1951.
Kennedy, A.. B. W. : "The ::\Iechanics of }Iachinery," :Vlacmillan & Co., Ltd.,
London, 1886.
Smith, Robert H . : A New Graphic Analysis of the Kinematics of �·[echanisms,
Trans. Roy. Soc. Edinburgh, vol. 32, pp. 507-517, 1882-1885. Also in
"Graphics," pp. 114-162, Longmans, Green & Co., Ltd., London, 1 889.
 TYPE, NUMBER, AND
DIMENSIONAL SYNTHESIS

5-1 KINEMATIC SYNTHESIS

The study of motions in machines may be considered from the

two different points of view generally identified as kinematic
analysis and kinematic synthesis. Kinematic analysis is the
determination of the motion inherent in a given machine or
mechanism. Formerly displacement analysis was of paramount
interest, and it still may be. However, increases in rotational
speeds have made a knowledge of velocity and acceleration
characteristics critical factors in the design of the many elements
comprising the complete machine. Inertia forces deriving from
the accelerations may be several times as large as the static
forces. In consequence, cross-section dimensions of links and
bearing selection are contingent upon acceleration magnitudes
and directions. Kinematic synthesis is the reverse problem:
it is the determination of mechanisms that are to fulfill certain
motion specifications. Synthesis is the very ·fundamental of
design, for it represents the creation of new hard,vare to meet
particular needs in motion-displacement, velocity, or accelera­
tion-singly or in combination. Some typical examples of kine­
matic synthesis are given in the following:
TYPE, NUMBER, AND DIMENSIONAL SYNTHESIS 131

Cz

p ,
FIGCRE 5-1 Point C is guided FIGURE 5-2 C',orrelation
along curve p to successively of the angular positions
occupy positions C1 , C2, C3, c•. of two cranks.

Guiding a point along a specified curve. Point C (Fig. 5-1) is to

be guided along the path p while successively occupying the positions C1 ,
C2, C3, C4 • An early application of this problem is found in Watt's design
of his double-acting steam engine (Fig. 2-10). Other applications are
found, for example, in the posthole borer (Fig. 2-11) and the baler
(Fig. 2-18).
Correlation of the angular posi,tions of two cranks. Two cranks a
and b (Fig. 5-2) are to have a specified relationship between their crank
angles <I> and y;. An application of this problem is found in the loga­
rithmic-scale converter (Fig. 2-8), in which a linearly divided scale in
<I> presents "1 as a linearly displayed logarithmic function.
Correlation of the angular posi,tion of a crank with the positions of a
point along a curve. Point C (Fig. 5-3) is to be guided along the curve p
and to occupy the successive positions C 11 C21 C31 C4 as the crank occupies
the positions a 1, a2, a3, a4. An application of this problem is found in
the film-feed mechanism (Fig. 2-14), ,vhere point C must move from
C1 to C3 through C2 for a 180° rotation of the crank and return to C 1
through C 4 during the remaining 180a° of crank rotation.
Correlation of the angular posi,tions of three cranks. Three cranks
a, b, and c (Fig. 5-4) are to have a specified relationship bet,veen the
crank angles q,1 and <J,2 and the crank angle y;. An application of this
problem is found in a fire-control computer giving the elevation angle
as a function of range and relative altitude.
As remarked in Chap. 1, the overall problem of synthesis may be

a4
aa
C2

p
('.·•,
W/uHA� d
(p I

(',
1/,
1/, ,
1'IGURE 5-3 Correlation of the angular FIGURE 5-4 Correlation of the angu­
P<>sitions of a crank with the positions of lar positions of three cranks.
a point along a curve.
132 KINEMATIC SYNTHESIS OF LINKAGES

approached in three sometimes interrelated phases. It is necessary to

reach decisions on (I) the form or type of mechanism, (2) the number
of links and the nature of the connections needed to permit the required
movability, and (3) the proportions (lengths) of the links necessary to
accomplish the specified motion transformation.
The first phase is called type synthesis. Here the choice of the
kind of links or constructional units is determined, as linkwork, gears,
cams, belts, etc. The second phase, called number synthesis, deals with
the number of links and the number of pairs of a given type required to
obtain a given number of degrees of freedom, i.e., a given number of
independent inputs to the mechanism. The third phase is called dimen­
sional synthesis. This last ohase of the problem, ,vhich is relatively
straightforward in most cases of cam and gear mechanisms, presents
challenging problems in the case of linkages, and it is to these problems
that the greater part of this book is addressed.

5-2 TYPE SYNTHESIS

The selection of the type of mechanism needed to accomplish a
given purpose depends to a great extent on considerations of usage,
materials available, manufacturing processes, etc., ,vhich lie outside the
field of kinematics. With so 1nany factors, there can be no scheme
whereby a mechanism may be uniquely determined on naming desired
motion specifications. It will be necessary to settle for a line-up of
possible combinations that could do the job, from which the "best one"
for the particular application in view is chosen. For example, many
quite different steam-engine configurations have existed side by side,
each arrangement having been dictated by non.kinematic requirements
such as location of rotating shaft, headroom, ease of maintenance, overall
size, critical material, manufacturing facilities, and so on, to name only
some features the designer must consider.
For type synthesis, the classification of mechanisms, and in par­
ticular Reuleaux's six groups and their extensions, are often helpful for a
systematic consideration of various possibilities, such as the choice
in the use of a cam or a linkage. It may be impossible, however,
to reach a decision as to which is the optimum solution until the best
possible linkage design is compared with the best possible cam design.

5-3 NUMBER SYNTHESIS

A collection of connected links must meet certain requirements in

order to be called a mechanism. If links are connected in a manner
resulting in a configuration such that relative motion of the links is
impossible, the assembly is called a structure. Three bars, pin-con­
nected to form a triangle, represent a ,vell-known structural unit. Four
TYPE, NUMBER, AND DIMENSIONAL SYNTHESIS 133

pin-connected bars of proper relative proportions constitute the familiar

chain known as the four-bar linkage, said to be movable if relative motion
of the links is possible. Further, after fixing one link, designating
another as the input link and still another as the output, the motion
of the output (or for that matter the relative motions bet,veen any t,vo
links) depends on only (1) the link lengths (or, more properly, relative
lengths) as parameters and (2) one variable, usually the position of the
input link. All motions repeat themselves identically with each cycle.
Such a mechanism is thus not only movable but is also said to be con­
strained; i.e., an exact duplication of the motions may be counted upon
each time the input link assumes the same position.
When a projected mechanism (or kinematic chain) has more than
four links, the existence of constrained movability may not be imme­
diately apparent and some sort of movability criterion is desirable.
The most obvious external characteristics of a kinematic chain are
the number of parts and the number of connections. Movability studies
based on only these two factors-the number of links and the number of
joints-have acquired the name of number syntheS'is. The oldest and
still useful, although incomplete, estimate of movability is known as the
Gri.ibler criterion. Number synthesis is presently based on a choice of
parameters limited to the above two, but the situation is not that simple.
Link lengths and special configurations of axis directions and locations,
positions of instantaneous centers of velocity, complexity of connections,
and perhaps other factors contribute to movability or lack of it. The
interaction of all factors has not been summarized in one comprehensive
and all-revealing relation. As we shall see, a simple relation exists based
on a count of links and joints but exceptions have to be noted when
particular geometries are present in the chain.
The number of degrees of freedom of a system is the number of
independent variables that must be specified to define completely the
condition of the system. In the case of kinematic chains, it is the num­
ber F of independent pair variables needed to completely define the
relative positions of all links. For example, the truss shown in Fig.
5-5a has zero degrees of freedom (F = 0); here the relative positions of
the links result from their lengths, and no pair variable can be specified.

...---------��-
,..______,/1>2

(a) Truss. F = 0 (locked) (b) Four-bar chain, F = l (c) Five-bar chain, F =2

(constrained)

FIGURE 5-5 Exan1ples of kinen1atic chains .with F = 0, 1, and 2.

 · · - - -,, - ·-----·. �·. - ··-···-----

134 KINEMATIC SYNTHESIS OF LINKAGES

The number of degrees of freedom of the four-bar chain of Fig. 5-5b is

unity (F = I); here one variable such as tJ, is needed to define the relative
positions of all links. A five-bar chain (Fig. 5-5c) has two degrees of
freedom (F = 2), for two angles such as t/> 1 and <1>2 are needed to define
the relative positions of all links.
A kinematic chain is said to be 1novable when its number of degrees
of freedom is one or greater (F > 1); it is otherwise locked (F < 1). If
the number of degrees of freedom is equal to unity (F = 1), the chain
is said to be constrained. �lost mechanisms used in machinery have
constrained motion; ho,vever, in computer or control mechanisms, which
may have several inputs, the number of degrees of freedom will be the
same as the number of inpG.t.s.
The four-bar linkage has been subjected to many studies, including
movability! One such is that of Grashof (1883), whose criterion for
the identification of a four-bar linkage possessing at least one link able
to rotate continuously was discussed in Chap. 3. Grashof, it will be
remembered, was concerned with the kind of mechanism-crank rocker,
double crank, or double rocker-and noted that the link proportions, in
addition to the fixing of a particular link, played a dominant role in
establishing the kind of mechanism.
Grashof began his deductive study with a closed four-bar chain,
which he took to consist of the links a, b (coupler), c, and d (frame).
Furthermore, he specified c > a, d � a, and O < b < a + c + d and
excluded the special cases in which two links were of the same length.
In words, he started from a general and movable four-bar mechanism.
If the lengths a, c, and d are given, and if b is allowed to increase
from its lower limit of zero, or if b is allowed to decrease from its upper
limit of a + c + d, the movability of the mechanism increases gradually
in the sense that links a and c will oscillate over increasing arcs. The
mechanism, locked at the limits b = 0 and b = a + c + d, becomes a
double rocker ,vith the onset of each of the indicated changes in b. From
here Grashof went on to sho,v the relations needed to delineate the three
kinds of mechanisms inherent in the four-bar linkage. With even simple
chains the influence of link lengths is seen to be a powerful one.
Another approach, and one suited to planar chains of more than
four links, is that taken by Griibler, also in 1883. In this the develop­
ment proceeds from the number of degrees of freedom allowed by the
kinematic pairs connecting successive links and leads to the degree of
freed om of the chain. The deductions appear as a rule expressed in the
form of an equation which may be called a criterion of movability. The
planar criterion was rediscovered in 1928. 1 The Grubler criterion for
spatial linkages bears the date of 1917.
1
\V..J. \\Talker, The Relation between Kinematic Pairs and Links in a Mecha­
nism, Phil. Mag., ser. 7, vol. 6, no. 37, p. 631, October, 1928.
TYPE, NUMB E R , A N D D IM E N S I O N A L S Y N T H E S I S 135

Reference link 1
y

FIGURE 5-6 Rigid link 2 in O '------- �--------

Xp
planar motion with respect
to reference link 1 .

Grubler's criterion for planar mechanisms of lower pairs (1883)

incorporated the number of links n and the number of joints j into an
equation valid for constrained motion; that is, F = l for the chain. The
Grubler equation is 2j - 3n + 4 = 0. It is possible, however, to present
a criterion that predicts more than just the fulfillment of the single
condition F = 1. Accordingly, we depart from Griibler's approach
although following it in spirit.
Consider two links in planar motion (Fig. 5-6). Link 1, ,vith
coordinate system Oxy, is taken as the reference plane. The moving
plane, or link 2, carries a line u containing a point P. The position of
link 2 ,vith respect to link 1 is determined from the specification of three
variables, the coordinates Xp and y P of the point P and the inclination (J
of the line u. These three variables are further identified as the degrees
of freedom of a body in planar motion; for such a body fmax = 3. If Xp
and YP are made invariant by a revolute connection at P, two restraints r
to motion are imposed on link 2 and its degree of freedom drops to f = l;
its position is now decreed by only a single variable, here 6. A revolute
connection is thus characterized by either f = l or r = 2.
Suppose that we have a closed chain of n links, connected by j
revolute pairs, that is to have a degree of freedom of F. Each link
initially possesses three degrees of freedom before connection to any
other link, whence the total-degrees-of-freedom number 3n. On choosing
one link as a reference for all others, i.e., fixing one ]ink to create a mecha­
nism, n - l moving links remain, and the degrees of freedom no,v total
0nly 3(n - 1). Each revolute connection means the loss of two degrees
of freedom; with j connections there is a loss of 2j degrees of freedom.
We may summarize our situation by ,vriting

F = 3 (n - 1) - 2j (5-1)
136 KINEMATIC SYN T H ESIS O F LINKAGES

A mechanism with constrained motion has F = l, ,vhence

1 = 3(n - 1) - 2j
Rearranging the terms produces
2j - 3n + 4 = 0 (5-2)
This last is the classical form of the Grtibler criterion. 1 We remark that
it involves only the number of links n and the number of revolute con­
nections j and pays no attention to link dimensions or other geometric
features.
Further discussion of the movability criterion is helped by con­
sidering Eq. (5-1). This e1uation is especially useful in that it has a
form that can be remembered easily, even derived on the spot. Having
two unknown quantities, we are forced to solve for them by comparing
trial and error ,vith experience. Application of this equation to various
closed kinematic chains wi11 allo,v certain deductions:
1 . T,vo links, t,vo joints: t,vo overlapped links
n = 2 j = 2 F = 3(2 - 1) - 2 X 2 = - 1
We may call this a statically indeterminate structure.
2. Three links, three joints : truss
n=3 j= 3 F = 3(3 - 1) - 2 X 3 = 0
We recognize this as a statically determinate structure.
3. Four links, four joints as a movable four-bar chain
n=4 j = 4 F = 3(4 - 1 ) - 2 X 4 = 1
This is constrained motion, and one input is required. However, we note
that the criterion fails if one link is as long as the sum of the other three
(Grashof), for the linkage is then immovable and statically indeterminate.
4. Five links, five joints
n = 5 j = ,5 F = 3(5 - 1) - 2 X 5 = 2
This is unconstrained motion unless two inputs are provided. Also,
the criterion is again insensitive to the case in which one link's length
is equal to that of the sum of the other link lengths (extension of Grashof),
such a link.age being immovable and statically indeterminate.
Two additional examples might be considered. In Fig. 5-7a one
diagonal link, number 5, has been added to a four-bar chain, forming t,vo
1\Vriting in both Civilingenie11r, vol. 29, p. 187, 1883, and "Encyklopadie der
mathematischen Wissenschaften," vol. IV, p. 127, 1901, Griibler recognized that both
Chebyshev (1869) and Sylvester (I87.:t) were aware of a relation equivalent to our
Eq. (5-2).
TYPE, NUMBER, A N D DIMENSIONAL SYNTHESIS 13 7
B

(a) F = 0 (b) F = - 1

FIGURK 5-7 Locked and statically indeterminate chains.

immovable trusses. Clearly, n = 5. The number of joints is six, how­

ever, for the revolutes at A and B serve also as connections of and to
link 5 and hence must be counted for this service. Applying the criterion,
F = 3(5 - 1) - 2 X 6 = 0
and we agree that the linkage represents a statically determinate truss.
The addition of a second diagonal, link 6 (Fig. 5-7b), means n = 6,
j = 8, and
F = 3(6 - 1) -. 2 X 8 = - 1
The structure is statically indeterminate.
Grlibler noted also that prismatic pairs, like revolutes, possess
one degree of freedom and hence might be included as joints. In Fig. 5-8
the moving link 2 is connected to the reference plane (link 1) by a pris­
matic guide enforcing the direction of motion 8 between the links. This
guide could be anywhere; two physical possibilities are shown. How­
ever, some fixed point such as P is needed as reference from which to
measure the translation s. This s is then the pair variable or degree of

Xp Xp,

Reference link 1

FIGURE 5-8 Rit?:id link 2 connected to reference link l by a

prismatic pair.·
1 38 KINEMA T I C SYNTHESIS O F L I N K A G E S

FIGL"RE 5-9 Higher-pair connection

between links 1 and 2 allowing simul­
taneous rolling and sliding between
elements. Origins 01 and 02 are fixed
Link 1 on their respective elements; the pair
variables are w1 and w2, whence f = 2.

freedo1n characteristic of the pris1natic pair; ,ve may also say that two
restraints are offered.
Prismatic joints need to be regarded ,vith more circumspection
than revolute connections when used in the Griibler criterion, for geo­
metric singularities make themselves felt much sooner. In general,
several parallel prismatic connections should be viewed with suspicion,
and they certainly should not be on the same link. /\ closed three-link
chain with nonparallel prismatic connectors is constrained and movable
(F = 1), while a four-link affair evidences two degrees of freedom, even
with nonparallel connections. Further discussion of prismatic pairs
would be some,vhat academic, for it is unlikely that they will be the sole
constituents of chains found in practice, and they will certainly not
appear in profusion in the synthesis problems to be considered.
Higher pairs found in planar mechanisms may also be included in
the Griibler criterion if the t,vo degrees of freedom they provide are
recognized. The higher pair of Fig. 5-9 allows both roll and slide between
pair elements, whence f = 2 and r = 1 . It would be counted as two
joints, and another (binary) link would have to be included in the count,
since a higher pair is the equivalent of two lower pairs, or, when replaced
· by an equivalent linkage, picks up another link (see Fig. 5-10).

5-4 DIMENSIONAL SYNTHESIS

By dimensional synthesis we understand the determination of the
dimensions of parts-lengths and angles-necessary to create a mecha­
nism that will effect a desired motion transformation. Laying out a
cam to meet certain specifications is dimensional synthesis. The caI11
provides a functional relationship between shaft rotation and follower
displacement, velocity and acceleration : the cam is a function generator .
A linkage is another type of function generator, but the paths to the
goal are not as direct as with cams.
In considering dimensional synthesis, it is recognized that it has
two aspects, called approximate and exact. We shall meet this distinc-
T Y P E , NUMBER, AND DIMENSION AL SYNT H E S I S 1 39

tion in terms of the Watt approximate straight-line mechanism and the

Hart and Peaucellier devices able to produce a true straight line (Sec.
6-6). Were we to ask a Watt linkage to generate the function· y = mx,
a true straight line, it could do so only approximately in the sense that
the coupler point's trajectory would be a wavy line intersecting the true
line every now and then. The two curves, that generated mechanically
a.nd the one desired, have in common only their points of intersection:
these are aptly called the accuracy (or precision) points of the generated
function. For a given length of line we should like to have (1) as many
accuracy points as possible and (2) a minimum deviation (or error)
between the curves.
Exact synthesis is limited, since few arbitrary functions can be
handled. For example, it takes at least a six-bar linkage such as the

.,

= 4, i = 4, F = 1
1 {=2
n

Normal

3 3
F=2

2 I- -·+---
C2
�----,,--
1 1
n = 4, j = 4, F = l

Joints to be counted are in circles

n = 8 + 1 + 1 = 10; j = 13; F = 1
1

FIGl'RE 5-10 Higher pairs and the Grtibler criterion.

140 K I N E M A T I C SYNTHESIS O F L I N KAGES

Hart to generate a true straight line. On the other side of the coin. there
is a notable exception: only a trivial mechanism is required to describe a
circle, while a four-bar is needed to produce circular arcs that are orily
approximate. Nevertheless, exact, or precision, synthesis is limited to
certain "nice" functions, whereas approximate synthesis can do a job
within a limited range on almost any function.
The concern in this text is with approximate synthesis. There
are t,vo approaches-the geometric, or graphic, and the analytic, or
algebraic. Geometry is, of course, analytical, but the term analytic has
become associated ,vith algebraic methods of computation in comparison
with graphical constructions. Speaking in rather broad terms, it may
be said that the geometric methods have been developed by the German
school, with emphasis on planar linkages. These methods are con­
sidered in Chaps. 6 to 9. The Russian effort, once heavily geometric,
has now a strong bias to,vard the algebraic attack, ,veil suited to spatial
mechanisms. In America-and one can only speak of recent years-the
principal developments have been in algebraic treatments, with the
inclusion of spatial linkages. The algebraic methods, ,vhose rise roughly
parallels the coming of age of digital computers, are considered in Chaps.
10 to 12.
The geometric methods can furnish, with reasonable accuracy,
quick and dependable solutions to a number of problems. They give
direct feeling for mechanical details ,vhich ,vill be important in reducing
a given solution to hard,vare and which may be obtained on the dra,ving
board without making use of the sometimes unfamiliar or unavailable
techniques of automatic computation. In the solution of some problems,
as in the case of spat,ial mechanisms or ,vhen the requirements for accu­
racy are more demanding, geometric methods may, however, become
cumbersome, lengthy, and undependable. For such problems, analytic
methods, dependent on the use of automatic computation techniques,
may yield practical and econo1nical solutions.

5-5 SPACING OF ACCURACY POINTS

It ,vas pointed out earlier that, when a mechanism is designed to
generate a given function or trace a given curve, it is not possible in
general to obtain a mathematically exact solution but that the mecha­
nism fits the function or curve at only a finite number of points, the
accuracy points. 1'he number of these accuracy points is equal to the
number of fixed parameters that may be used in the synthesis and varies
in general bet,veen three and six. The problem considered in this sec­
tion is that of spacing the accuracy points ,vithin the interval of function
generation to minimize the errors between accuracy points.
TYPE, NUMBER, A N D DIMENSIONAL SYNTHESIS 141

R(x)

,_____ Interval = 2h
L_ __- h --- ---+- - - +tr---�
d
- R(xJ
dx
=0

a+h

FIOt:RE 5-11 Error curve with three accuracy points.

Consider the function f(x) to be approximated in a given interval

of variation of x by means of a mechanism which generates the function
F(x; q1, . . . , qn) = F(x; qk), where q1, . . . , qn are the values of theen
design parameters in the mechanism. The function f(x) describes the
desired feature of the motion; this is to be approximated by, say, the
path of a coupler point or a relation between crank and follower rotations
in a four-bar linkage. The function F(x; qk) defines the feature of the
motion as it is actually performed by the mechanism ,vith design param­
eters set to q 1, • • • , q " . The difference between these two functions
is the structural error,
R(x) = /(1:) - F(x; q,)
The general appearance of the structural error ,vhen plotted against x is
sho,vn in Fig. 5-11, ,vhere the center of the interval has been taken at
x = a and its width as 2h, thus,
a-h<x <a+h
The error is zero at each accuracy point (points a1, a2, and a 3) and
reaches a series of maxima and minima between accuracy points.
If the magnitude of the error R(x) is to be kept to a minimum
throughout the whole interval of variation of x, the optimum spacing
of accuracy points will be that for ,vhich the error curve takes the form
shown in Fig. 5-12, ,vhere all maxima and minima as well as the values
of the error at the extremities of the interval are of the same magnitude.
The exact spacing of accuracy points corresponding to this yptimum
situation depends, of course, on the function to fie generated as well 'Bs
on the mechanism, and its exact determination is a complex problem. It
Inay be shO\\'n, hO\\'ever, that the structural error R(x) may be expressed
as
R(x) = KPn (x) (5-3 )
With Pn(x) = (x - a,)(.t - a2) · · · (x - an)
142 KINEMATIC SYNTHESIS O F LINKAGES

R(x)

-- - --- Interval = 2h--- - --'t

k--- - h -- -1-- - + h ------

FIGl'RE 5-12 Optimum error curve.

,vhere a 1 , a2, . . . , an are the accuracy points and K may be considered

constant as a first approximation. Within this first approximation, the
problem of optimum spacing is reduced to finding a polynomial P,. (x)
of degree n and leading coefficient (the coefficient of .t") equal to unity
that deviates least from zero. The solution to this problem is given by
the Chebyshev polynomials (see Sec. 5-6),
h2
T2 (x) - (x - a) 2 -
2 when n =2
31 2
T3(X) - (x - a) 3
- (x - a) when n =3
or in generaP Tn (x) = ��" 1 cos (n arccos x � a)
The accuracy points are the roots of the corresponding Chebyshev poly­
nomial, and their values may be conveniently found as followse: From
the center of the interval (a - h, a + h) taken along an axis x, draw a
circle of radius equal to one-half this interval, and inscribe a regular
polygon of 2n sides in it in such a way that t,vo of its sides are perpen­
dicular to the x axis. The projections of the vertices of this polygon
onto the x axis determine the accuracy points. The construction for
n = 4 is shown in Fig. 5-13.
For some problems it is desirable to minimize the first derivative
of the structural error with respect to x rather than minimize the
structural error itself. Such a problem would be the synthesis of a
mechanism in which the velocity ratio bet,veen input and output links
is to vary in a prescribed manner throughout the range of operation.
Examination of Fig. 5-11 sho,vs that the derivative of the error, i.e., the
slope of the error curve, generally becomes zero for one value of x between
t,vo accuracy points and that it is comparatively large at the accuracy
points themselves. This shows that, if a function is generated ,vith n
1 The hardly obvious general expression is developed in Sec. 5-6.
'r Y P E , N U MBE R , A N D D I M E N S I O N A L S Y N T H E S I S 1 43
accuracy points, its derivative will ordinarily be exact at only n - 1
points, whence less accuracy must be expected in the design of a mecha­
nism for prescribed velocity relations than in a design for prescribed
displacement re]ations.
A. first approximation to optimum spacing of accuracy points for
the velocity case may be obtained by taking the derivative of the struc­
tural error as expressed in Eq. (5-3),
d d
dx
R(x) = K
dx
Pn(x)

and remarking that dR/dx deviates least from zero when dPn/dx itself
deviates least from zero, K being assumed constant. The polynomial
Pn(x) must therefore be of degree n and have a leading coefficient equal
to unity, and its derivative must deviate least from zero. Thus dPn/dx
is taken proportional to the Chebyshev polynomial Tn-1, and
P,. (x) = b[fTn -1(x) dx + CJ
,vhere b is a constant of proportionality and C a constant of integration.
The determination of the accuracy points and the meaning of the
constants will be demonstrated by an example. Let it be required to
establish four accuracy points a1, a2, aa, a4 for the interval - 1 < x < + l ,
for which (within the usual first approximation) the derivative of the error
is minimized. Since n = 4, h = 1, a = 0, we find
4
P4(x) = b[ f (x 3
- ¾x) dx + C] = b (� - : x2 + C)
The value of b must be found such that. the coefficient of x• in
P4(x) is unity, whence b = 4 and
P 4 (x) = x' - ¾,r2 + 4C
><---- Interval == 2h ----...i

---h - +-- -+ h --,�

------- - ~

o ___
--,-:---.__ _...,.--,--.-----%
a- 04 0 1 02 01 o
1 3
\
j
I
I
I I
I
FIGURE 5-13 Determination I I
of four accuracy points with I I
Chebyshev spacing.
144 K I N E M ATIC SYNTHESIS O F LINKAGES

3 x2
x•- 2
0.5

-1.0 +LO
I

-- -x
(
; 4' = 0.28
·- - .... - -
-0.5
-0.6
A

FIGURE 5-14 Determination of four accuracy points to minimize the

derivative error according to tl,-, first approximation of Chebyshev.

The velocity error, which depends on the derivative dP4/dx, is independ­

ent of the constant term 4C related to the displacement error. This
constant term may therefore be adjusted to reduce the displacement
error. The variable portion of Pn(x), that is, x' - -:}x2 is plotted against
x, as shown in Fig. ,5-14, and the accuracy points a 1 , a 2, a 3 , a4 are
found by the intersection of this plot ,vith a parallel to the x axis equi­
distant from the maxima and minima of the plot. Thus,
a1 = 1.12
a2 = 0.4i
aa = 0.4i
a4 = 1.12
The constant term 4C represents the maximum displacement error.
It is noted that two of the accuracy points, a 1 and a4, lie outside the
interval of generation. This is not peculiar to the example but will
occur in all spacing of accuracy points for ,vhich the derivative of the
error is minimized. It is to be expected, since the slope of the error
curve (Fig. ,5-14) becomes larger for values of x beyond the minima at
A and B.

5-6 C H E B Y S H E V P O L Y N O .M I A L S 1
The use of Chebyshev polynomials for the choice of accuracy
points as presented in the preceding section ,vould give the optimum
choice only if the factor K in Eq. (f>-3) were precisely a constant. This
factor, ho,vever, is truly a constant only for the case in ,vhich the function
f(x) is a polynomial of degree n and is to be approximated by the function
F(x; q"), the latter being also a polynomial but of (lesser) degree n - 1.
Here n is the number of accuracy points, or the number of design param-
1 This section may be omitted on first reading.
 · · · - · . ,- .

TYPE, N U MBER, A N D D I MENSIONAL S Y N T HESIS 145

eters q1, . . . , qn . This would be the case of a parabola being approxi­
mated by a straight line, a cubic approximated by a parabola, a fourth­
order curve approximated by a cubic, etc. Although the above situation
is scarcely the case in dimensional synthesis, the developments to be
given here will help in an understanding of the significance and limita­
tions of Chebyshev spacing of accuracy points.
Consider the problem of approximating a polynomial of degree n,
f(x) = A ..+1Xn + A ..xn-l + " . . + A2x + A l
in the interval (a - h, a + h) by means of another polynomial of degree
n - 1,
F(x; qk) = qflxn- l + qn-1Xn-2 + " " " + q2x + q.
where the coefficients q 1 , . . • , q.. play the role of the design parameters
in the more general problem of dimensional synthesis. Since there are n
coefficients in F(x; q1c), it will be possible to determine these coefficients
such that F(x; q1c) = f(x) for n values of x denoted as a 1 , . • • , a,., the
accuracy points. The coefficients q1, . . . , q,. of the approximating
polynomial must then satisfy the set of equations
+ q2a1 + q1 = f(a1)
+ q2a2 + qi = f(a2)
. . . . . . . . . . . . . . . . .. . . . . . .. . .
q,.a,. n-1 + q.,_.a,. n-9- +
and a solution of this system will yield a polynomial F(x; q,.) which
coincides ·with the desired polynomial f(x) for n discrete values of x,
that is, at the accuracy points. For values of x in the interval a -- h,
a + h other than the accuracy points, F(x; q,.) \\'ill deviate from f(x), or
·f(x) = F(x; q1c) + R(x)
where R(;r) is the deviation term. This term, the difference between two
polynomials of degree no higher than n, is also a polynomial of degree n.
Since /(x) = F(x; q,.) for x = a1, . . . , a n, the term R(x) must also be
zero for x = a 1, . . • , a,, and may be ,vritten in the form
R(x) = K(x - a 1 )(x - a2) · (x - a,.)
or R(x) = KP(x)
Where P(x) = (x - a 1 ) (.r - a2)
is a polynomial of degree n with leading coefficient (coefficient of xn)
equal to unity. The constant K, which in this case equals ..-1. ,.+ 1 , depends
0n the function f(x) but not at all on the accuracy points a 1, . . • , a.,..
It appears therefore that in the present case Eq. (5-3) holds exactly.
'I'he remainder of the problem consists in finding that polynomial of
146 KINEMATIC S Y N T H E S I S O F LINKAG E S

degree n with leading coefficient equal to unity deviating least from zero
in the interval a - h, a + h. 'fhis is known as a Chebyshev polynomial.
The polynomial P(x) of degree n has n zeros at x = a 1, . . . , an
and has a leading coefficient equal to unity. Its graphical representation
would exhibit a maximum or a minimum between each of the zeros, and
the problem is to reduce to a minimum the magnitude of the largest maxi­
mum or m1n1mum. It is reasonable to assume (as Chebyshev did) that
this situation will occur when all maxima and minima are equal, but this
succession of equal maxima and minima suggests that such a polynomial,
by suitable changes of variables, could be represented by trigonometric
functions. The interval of approximation a - h, a + h is first reduced
to the interval - 1 < z < + 1 by the change of variable
x- a = zh
Consider now the function cos n8, ,vhere cos 8 = z. This func­
tion has the desired features of P(x), that is, n zeros in the interval
a - h, a + h or - 1 < z < +1 with equal maxima and minima between
all successive zeros. This function cos n8 may be expanded in terms
of cos 8 = z; this is trivial for n = 0 and n = 1 . For n = 2,
cos 28 = 2 cos2 8 - 1 = 2z2 - 1
and for n > 2, the expansion may be deduced from the recurrence formula
cos n8 = 2 cos (n - 1 ) 8 cos 8 - cos (n - 2)8
derived from elementary trigonometric identities. 'fhus,
cos 38 = 4 cos3 8 - 3 cos 8 = 4z3 - 3z
cos 48 = 8 cos4 8 - 8 cos2 8 + 1 = 8z' - 8z2 + 1
Note that in these polynomials the leading coefficients are zn- 1 • Since
in the polynomial P(x) the leading coefficient must be unity, P(.r) must be

( - a)
of the form
Tn (Z) = 2 1-n cos (n arccos z)
X
or T,. (x) = hn2 1-n cos n arccos
h
and the zeros must correspond to
x a
-
n arccos -- - -
h
(2k - 1 ),r
- - or x = a + h cos - (2k--
- l),r
·· - -
2 2n
The polynomials T,., introduced by Chebyshev in a memoir (1853)
·devoted primarily to approximate straight-line mechanisms� are called
Chebyshev polynomials. It may be shown1 that of all polynomiah; of
· R. Courant and D. Hilbert, ''Methods of Matbe1natical Physics," vol. I,
1

pp. 88- 90, In�rscieiice Publishers, Inc., New York, 1953.

 - · - ·-·--�

TYPE, NUMB E R , A N D DIMENSIONA L SYNTHESIS 1 47

same degree and the same leading coefficient they deviate least from zero.
If theon accuracy points a1, . . . , an of the interval are set equal to the
roots of the polynomial Tn, that is,
(2k_o- l)1r
a,.• = a + h cos 2n
k = 1, . . . , n

then /�(.r) = T,. (x) and the deviation R(x) for the resulting approximating
polynomial F(x; q,,) will be smaller than for any othero·distribution of
accuracy points.
Although the functions f(x) and F(x; q,,,) in an actual case of
dimensional synthesis would not be polynomials of degree n and n - 1,
respectively, these functions may be expanded in Taylor series. As a
first approximation, the terms of order higher than n in f(x) and higher
than n - l in F(x; qk) may be neglected for the purpose of locating the
accuracy points. The problem would then reduce to the case considered
above, and a Chebyshev spacing \.vould result. The Chebyshev spacing
therefore appears as a convenient first choice, but it should not be
expected to yield optimum design. A spacing of accuracy points reduc­
ing the structural error to a minimum can be obtained only by repeated
trials: starting \.vith a Chebyshev spacing, a dimensional synthesis is
carried out and followed by an analysis of the mechanism to compare
the functions f(x) and F(x; q,.). The structural error is plotted as a func­
tion of x, and the accuracy points are moved close together where the
maximum structural error occurs. The process is then repeated until
all maxima and minima of the structural error between accuracy points
and at the ends of the interval have been equalized. An example of this
procedure, which requires the use of automatic computation to be feasi­
ble, is discussed in Sec. 10-8, where the synthesis of four-bar linkage
with five accuracy points is considered.

BIBLIOGRAPHY
Chebyshev, P. L.: Theorie des mechanismes connus sous le nom de parallelo­
gran1mes (1853), from "Oeuvres de P. L. Tchebychef," vol. 1, Markoff et
Sonin, St. Petersburg, 1899. Also in "Ylodern .Yiathematical Classics:
Analysis," ed. by Richard Bellman, S730, Dover Publications, Inc., New
York, 1961.
Grashof, F.: "Theoretische Maschinenlehre," vol. 2, Voss, Hamburg, 1883.
Grubler, �1artin: "Getriebelehre," Springer-Verlag OHG, Berlin, 1917.
Kennedy, Alex. B. W.: "Kinematics of Machinery," }facmillan & Co., Ltd.,
London, 1876.
 FOUR-BAR
COUPLER-POINT CURVES

6-1 THE FOUR-BAR LINKAGE

Link,vork, in its early applications, consisted mainly of revolute­
connected members and ,vas widely used for converting the con­
tinuous rotation of a ,vater wheel into a reciprocating motion
suited to piston pumps (Fig. 6-1). The piston-cylinder com­
bination at the end of the line represents a prismatic pair, of
course, but ahead of this there are only the revolute connections
generally associated ,Yith linkwork. Agricola's arrangements
show wheel and pump--power source and point of work-fairly
close together. Such compactness did not al,vays prevail; link­
works of magnificent proportions were also part of the past. A
linkwork is a means of power transmission as ,vell as being a
motion transformer. Before the introduction of rope transmis­
sions and the now universal electric wire, linkwork ,vas employed
for long-distance transmission of power. Gigantic linkages,
principally for mine pumping operations, connected water wheels
at the riverbank to pumps high up on the hillside. One such
installation (1713) in Germany ,vas 3 km long.

Such linkages consisted in the main of ,vhat ,ve call four-bar

linkages, i.e., planar four-revolute mechanisms, and terminated
in a slider-crank mechanism with a prismatic pair.
,roe ... ,._, Linl.....i.b<t•..,., ...... .i-1-.1,.,.,.P", """""'h
""ntw,,. [F,- ........i.••··o.,,...,,_,,....,•"H_,,,....i.,;.,.(1912).]
150 KINEMATIC SYNTHESIS OF LINKAGES

four-bar linkage was used in a new· \\·ay, for the significant motion output
With Watt's invention of the "straight-line motion" (I 784) the

was not that of the follower but that of the coupler: Watt had found a
coupler point describing a curve of special usefulness.
The first analytical investigation of a coupler curve, the curve of
the Watt mechanism, was undertaken by Prony,1 who examined Watt's
"straight-line motion" for deviations (I 796). Samuel Roberts showed
(1876) that the "three-bar curve" 2 -today we call it the coupler curve
of the four-bar-is an algebraic curve of the sixth order; i.e., a straight
line will cut it in not more than six points. Cayley and others shovved
further properties of the curve. 1'heir interest lay in exploring linkages
hypothetically able to g�nerate specific algebraic curves of any order:
the applications to mechanisms ,vere to be made later.

6-2 EQUATION OF COUPLER CURVES 3

Derivation

The equation of the coupler-point curve for a four-bar linkage may be

obtained by analytic geometry. 1'he derivation presented follO'\\'S that
of Samuel Roberts, with only slight changes in notation. The equation
will be written in cartesian coordinates, with the x axis along the line of
centers OA.On and they axis perpendicular to that line at OA (Fig. o 2). -
Let (x', y'), (x", y"), and (x, y) be, respectively, the coordinates of points
A, B, and coupler point M; then

=x -k y' = y - b sin 8
=
x'
x" = x - a cos (8 + 'Y)
cos 8
and y" y - a sin (8 + -y)
Since A and B describe circles (or arcs of circles) about centers OA and
0a, respectively,
x'2 + y'2 = r2 and (x" _ p) 2 + y "2 = 82

1
Gaspard Fran90is Prony (1755-1839), engineer, was an associate of the
famou� bridge builder Perronet and became his successor as director of the Ecole des
Ponts et Chaussees. Also professor of mathematics at the Ecole Polytechnique,
Prony wrote textbooks on mechanics and hydraulics but is perhaps best remembered
for the friction brake, or absorption dynamometer.
% Only the moving links were counted and called bars. In recent years, onlY
Svoboda ("Computing Mechanisms and Linkages," Massachusetts Institute of
Technology Radiation Laboratory Series, vol. 27, McGraw-Hill Book Company, New
York, 1948) has used the term "three-bar."
s This section may be omitted at first reading.
FOUR-BAR COUPLER-POINT CURVES 151

y
M(x,y)

FIGURE 6-2 Coordinate system

and notations used to derive �"' - -
-------
p �:W., ----x
equation of coupler curve. 0A'

Substituting the values of :c', y ' and x", y" into the last t,vo equations
yields
(x - b cos 0) 2 + (y - b sin 0)2 = r 2
and (x - a cos (0 + -y) - p] 2 + fy - a sin (0 + -y)]2 = s2
,vhich, by application of trigonometric identities and ordering of terms,
become
x cos O + y Sill fJ =
. x2 + y2 + b2 _ r2
2b
and
[ (x - p) cos 'Y + y sin -y] cos (J - [ (x - p) sin 'Y - y cos -y] sin 0
_ (x _ p)2
- + y2 + a2 _ 82
2a
The equation of the coupler-point curve 1nay now be obtained by
elimination of O between the last two equations. Solving these equations
for cos O and sin 8 and substituting the values obtained into the identity
cos 2 8 + sin2 8 = I yields the general four-bar coupler-curve equation
I sin a((x - p) sin 'Y - y cos -y](x 2 + y 2 + b2 - r 2)
+ y sin f3[(x - p)2 + y 2 + a2 - s2)} 2
+ {sin a[(x - p) cos 'Y + y sin -y](x 2 + y 2 + b 2 - r2)
- x sin f3[(x - p) 2 + y 2 + a2 - s2])2
= 4k2 sin 2 a sin 2 fJ sin2 -y[x(x - p) - .IJ - py cot -y]2 (6-1)
In this, k is the constant of the sine law applied to the triangle ABAf,

k - _!!:_ _ - ---�-- - C
sin a sin /3 sin 'Y
152 KINEMATIC SYNTHESIS O F LINKAGES

'"fhis equation is of the sixth degree and because of its properties

also bears the forn1idable title of tricircular sextic. One of its properties
has been n1entioned: a straight line will intersect it in no n1ore than six
points. Its further features may be .studied either geon1etrically or
algebraically. The geo1netric exarnination begins ,vith Sec. 6-3. An
introduction to the algebraic study of coupler curves will be given by
considering some properties deduced by Roberts and others. As a guide
to an understanding of the 1nethods of analytic geometry, second-order
curves (conic sections) will be exa1nined en route.

Circle of Foci
On setting
L = sin a[(x - p) sin -y - y cos -y) M = y sin /3
N = sin a[(x - p) cos -y + y sin -y] P = - x sin /3
</, = x2 + y2 + b2 _ ,2 "1 = (x - p)2 + .1/,. + a 2 - s2
Eq. (6-1) takes the forn1
(Lq, + My;) 2 + (N<I> + Pf) 2 - 4k 2 (LP - N1Jf) 2 = 0 (6-la)
Note that the equation
LP - lvM = 0
or :r(x - p) + y2 - PY cot 'Y = 0
represents a circleot passing through O.,i and OB (Fig. 6-3). For reasons
that will appear later, this circle is called the circle of singular foci.

Multiple Points
A n1ultiple point of a curve, as, for exan1ple, a cusp or a crunode (see
Sec. 6-3), is a point ,vhere the curve has several tangents. We propose
to show that the coupler curve has n1ultiple points at each of its inter­
sections with the circle of foci. When a curve is defined by an equation
of the fornt
/?(.r, y) = 0
its tangent 1nay be found by equating to zero the differential of the
function F(x, y),
iJF dx + cy
iJF dy = 0
ex
1
In rectangular coordinates every equation of the form
.r2 + yi + Dx + Ey + F = 0
represents a circle.
FOUR-BAR CO UP LEH.-PO 1 �T CU HVES 153
y

p sin /3 sin a
sin'Y

,,-�-Circle
of foci

p
psin/Jcosa
sin'Y

FIGURE 6-:3 Triangle of the singular foci OAOnOc similar to coupler

triangle ABM.

Since dx and dy are infinitesin1al changes in the coordinates x and y along

the curve, they also define the slope of the tangent as
dy aF/ax
dx = aF/ay ·

expression niust be indetern1inate, which n1eans that both aF/ax and

At a multiple point, where the curve has several tangents, the above

iJF/ay must be zero to satisfy the indeterminancy.

The left n1en1ber of Eq. (6-la) is the function F(x, y) corresponding
to the coupler curve, whence

� = 2(L<t> + M,f) <Jax (L<t> + Af,f) + 2(Nrt, + Pf) a� (N<I> + Pf)

- 8k 2 (LP - Nll1) _!___ (LP - NA,f)
ax
A similar expression is forn1ed for iJF/ay on replacing :r by y. Since
the points of intersection of two curves are found by considering their
equations as sin1ultaneous, the intersection of the coupler curve and the
circle of foci is given by the pair of equations

(Lq, + M,f) 2 + (N<t> + P,f) 2 - 4k 2(LP - NM)= 0 LP - NM= 0

Since the left n1e1nber of the second equation is zero, each tern1 of the
154 KINEMATIC SYNTHESIS 01" LINKAGES

first equation is like"•ise zero, and

L<J, + Mi/I =0 N </, + Pi/I = 0
'fhe last two quantities serve to n1ake fJ/?/fJx and fJF/fJy zero also, thus
satisfying the requirement of an indeterminate slope at the point of
intersection of the t,vo curves and establishing the presence of a n1ultiple
point.

Imaginary Points
Further properties of the curve may be deduced fron1 its equation by
considering in addition to real points of the plane, whose coordinates
x and y are real nun1bers, in1aginary points having co1nplex nurnbers as
coordinates. Complex nu1nbers of the fonn z = x + iy are used in this
text to represent real points of coordinates x and y, in which both x and y
a.re real. 1'he situation in this section is different, because x and y are
then1selves complex: such points are called in1aginary. As re111arked,
these points have no geon1etric or physical meaning, for they cannot,
with their four coordinates, represent real points in a plane. They are
useful, however, because of their analytic resemblance to real points.
Although no imaginary point exists in a n1aterial plane (the plane is
already con1pletely "filled" with real points), consideration of such points
is someti111es helpful in the study of curves by n1eans of their equations.
As in the consideration of n-di111ensional spaces with n larger than 3,
it often turns out to be convenient and suggestive to think in geometric
language about quantities having only analytic meaning.
Thus, although the coupler curve is a closed curve ,vhich does not
extend to infinity, wc shall be able to speak of its imaginary points at
infinity and detern1ine its asyn1ptotes at those points. The asyn1ptotes,
as may be expected, turn out to be i1naginary lines, but three of their
intersections are real and very significant points. It is fron1 a considera­
tion of these points that Roberts deduced for the first time ,vhat we call
the Roberts-Chebyshev theoren1, that the san1e coupler curve may be
generated by three different four-bar linkages.

Second-order Curves
i\.n asy111ptote of a curve is a straight line such that a point, tracing a
curve and receding to infinity, approaches indefinitely near to the straight.
line. An asyn1ptote may also be considered as a tangent to a curve at a
point an infinite distance fron1 the origin.
Before considering the sixth-order curve of primary interest, the
matter of points at infinity and asymptotes of a curve will be reviewed
F O U R-BAR C O U P C
L ER-POINT CURVES 155

in terms of the more familiar second-order curves ,vhose general equa­

tion is
Ax2 + By2 + Cxy + Dx + Ey + F = O (6-2)
As n1ay be recalled, the above equation-depending on the values of the
coefficients A , B, and C-represents an ellipse, a parabola, or a hyperbola.
Such curves intersect any straight line in t'\\·o points which may be real
or i1naginary, at finite distances or at infinity. For points at infinity on
such curves, the coordinates x and y are infinite, whence the first three
terms of highest power are so large that the last three terms may be
neglected. Points at infinity then lie on the curve defined by the equation
Ax2 + By2 + Cxy = 0 (6-3)
The directions t of the points at infinity on a curve are defined from the
equation of a straight line passing through the origin, y = x/t or t = x/y,
in which t = cot <J, (Fig. 6-4). Since the coordinates of the intersection
between line and curve must satisfy the equations of both, we have the
pair of equations
.Ax2 + By 2 + Cxy = 0 X = ty

Elimination of x and y produces

At2 + Ct + B =0 (6-4)
from ,vhich

t'
= -c + vc 2 - 4A B t" =
-
____
c - -
v
1c2 - 4AB
-=--
2A --:-----
2A

y
t'
I
I

r------ �
Directions of /
points at /
infinity
. I
t"- ) I

-- -- ----------j--- .{_ ,J," ,J,' "'

---
---
I
I o -- ._
--..
/....- I --. --.C --
/
I
I
I
I
I
FIGURE 6-4 Dircetions of points at infinity of a second­
order curve.
156 KINEMATIC S Y N T H E S I S OF L I N K AGES

Each value of t indicates the direction of a point at infinity, i.e., the

angles ,t,' and ,t," (Fig. 6-4).
The species of conics n1ay be distinguished by the number of real
directions in ,vhich lines passing through the origin n1eet the curve at
infinity. The discri1ninant of the expression for t serves for the testing :
1 . If C2 - 4AB < 0, the directions t' and t" are in1aginary and
the curve is an ellipse. Points at infinity on the curve are imaginary,
since they lie on lines having in1aginary directions. As an example,
consider the circle (a special case of the ellipse) given by the equation
x2 + y2 = R2

Following the procedure produces

t2 +1 = 0 ,vhence t' = +1·, t" = -i

which is to say that the circle has no real points at infinity (they would
have to lie on real lines, i.e., be lines having a real direction). For
analytical purposes, the presence of a direction is tantan1ount to the
existence of so1ne kind of point at infinity. If the direction is found to
be real, then a real point exists at infinity on the curve; if the direction
is found to be either + i or - i, then the i,naginary point at infinity is
called a cyclic point. 1 The circle is then said to have t,vo points at
infinity, one in each of the directions i and -i, that is, the two cyclic
points.
2. If C2 - 4A. B = 0, the directions are real and equal, t' = t",
and the curve is a parabola. The t,vo real points are coincident at

3. If C2 - 4AB > 0, the directions are real and distinct, t' � t",
infinity and are a double point.

and the curve is a hyperbola. There are tv.o real points at infinity
because of the t,vo branches.

Asymptotes
An asymptote to a curve is its tangent at a point at infinity; since second­
order curves have t,vo points at infinity, there ,vill be an asyn1ptote for
each direction. To find these asyn1ptotes, it is convenient to transform
the x, y coordinates of a point P on the curve into coordinates related
to the directions of the points at infinity, i.e.! into coordinates related
to the oblique axes defined by t' and t". These new coordinates (Fig.
6-5) ·will be called X and Y and ·will be expressed in tern1s of ratios of
perpendicular distances measured fro1n P to the ne,v axes t', t" to per­
pendicular distances measured fro1n an invariant point Q(t 0) of the old,
or x, y, system, viz.,
1 A cyclic point is also known as a circular point at infinity.
FOUR-BAR COUPLER-POINT CURVES 157

X =oPMo= PM and = PN PN · ·
QH I sin <t,' y QK -:-- 1 sin <t,"

The change in coordinates between the systems is then

X = x sin <J,' - y cos <t,' = x _ t'Y

sin <J,'
x sin <t," - y cos <t,"
and Y = . </,11··-- = X - t"y
SIU

The equation of the second-order curves [Eq. (6-2)] becon1es

- X Y[2A t't" + 2B + C(t' + t")] - X(t' - t")(Dt" + E)

+ Y(t' - t") (Dt' + E) + F(t' - t") = 0 (6-5)

where t' and t" are the familiar solutions of Eq. (6-4).
T,vo conditions apply to the asyn1ptotes :
1 . They must be parallel to the directions t' and t" of the points
at infinity.
2. They must intersect the curve at two points at infinity. (Recall
that an asymptote is a tangent at infinity and that furthermore a tangent
is the limit of a secant whose two points of intersection with the curve
have become coincident.)
Now, a line parallel to the t" axis has, in terms of the new coordi­
nates, an equation of the form Y = const, where the constant determines
the distance fron1 the axis t". Such a line ,vill intersect the curve at
infinity as well as at a point P (Fig. 6-.5 ). The X coordinate of that

Line parallel to /
t" axis intersects I
curve at infinity /
I P(x,y), second intersection
..____2 / of line and curve
-------- I I

Curve
_ M�

(���:::��- .-:---t_______--J
- --_ ___
1- :;;/..,/1�-::::-__ ---
.
f'------ �--·---f+----__ ---------
�
I, '-, !
----- ---
---_
, o --- ---
I '- ; Q(l,0)
__________;;::,...::-----f--.l.J�---------%

---d.. --
__
i

--
I N
I K - -- --
1
I
I
I
FIGURE 6-5 Relations of oblique axes t' and t", curve and asymptote.
158 KIN EMATIC S Y NTHESIS O F LINKAGES

second point is given by Eq. (6-5) \vhen Y is replaced by the constant

of the line. Since property 2 above requires that the asymptote intersect
the curve at two points at infinity, P n1ust also be at infinity and its
coordinate X must be infinite. With X infinite, the terins independent
of X in Eq. (6-5) are trivial and the coefficient of X n1ust be zero, giving
- Y[2:-lt't" + 2B + C(t' + t")] - (t' - t")(Dt" + E) =0
On returning to the rectangular coordinates x and y

= Dt" + E
X - t"y
vc 2 - 4AB

This is the equation of the asymptote in the t" direction.

The equation of the asyn1ptote parallel to the t' direction is found
in a similar rnanner by equating to zero the coefficient of Y in Eq. (6-5);
this gives
X - tIy = - ---;-::;Dt'
==
2
+=
E ··-::
vC - 4AB
As an exan1ple, consider the hyperbola
x2 - 2y2 + 2x + 2 = 0
The directions of its points at infinity are found fro1n t2 - 2 = 0, that is,
t' = y2 and t" = -v2. The asy1nptotes are then given by
X - y y2 = - 1
and x + y v2 = 1, that is, by the real lines
and

As a further exan1ple, consider the circle x2 + y2 - R 2 = O, for

which we already know that its points at infinity, the cyclic points, lie
in the directions t' = i and t" = -i. Upon rewriting the equation
of this circle in terms of X = x - iy and Y = x + iy, it becomes
X Y - R2 = 0. The coefficient of X is Y, which must be zero to satisfy

tion t" = -i is
the equation with X infinite, whence the asyn1ptote parallel to the direc­

Y=O or X + iy = 0

Similarly, the asymptote parallel to the direction t' = i is found to be

X =0 or X - iy =0
Both these asymptotes are imaginary and possess one real point, the
origin of coordinates at which they intersect.
FOUR-BAR COUPC
LER-POINT CURVES 15 9

Asymptotes of the Four-bar Coupler Curve

Exarnination of the curve equation (6-1) shows its terms of highest degree
to be 1:6 + y 6• In consequence the significant equation for points at
infinity is x6 + y 6 = 0. As ,vith the second-order curves, this equation
is taken with y = x/t, giving
l6 +1 =0 or (t 3 - i) (t 3 + i) = 0 (6-6)
as the equation the directions of the points at infinity must satisfy.
This equation of the sixth degree has six solutions-all in1aginary
-and the coupler curve therefore has six imaginary points at infinity
in the directions
ti = t
.
and li t
= - i.
each being a triple solution of Eq. (6-6). The cyclic points are therefore
triple points of the coupler curve. The asymptotes, also imaginary,
must be parallel to the directions i and -·i, and since each cyclic poh1.t
is triple, there will be a total of six asyn1ptotes forming two sets of three
parallel irnaginary lines. The determination of the asymptotes follows
the rnethod described for secon d -order curves.
In order to carry out the computations more conveniently, Eq.
(6-1) is first rewritten, use being made of the identity
(1 2 + V2 = (U + iV)(U - iV)
with U and V as the parentheses appearing in the left-hand member of
Eq. (6-1). 1'hus, after some algebraic n1anipulations,
{ sin a(x2 + y 2 + b2 - r2) (ix - y - ip) e-i,.
+ sin .B[(x - p) 2 + y 2 + a2 - s2) (y - ix) }
X {sin a(x2 + y2 + b2 - r2) ( - ix - y + ip)e•-r
+ sin .B[(x - p) 2 + y 2 + a2 - s2] (y + ix) }
= 4k 2 sin2 a sin2 fJ sin2 -y[x(x - p) + y2 - py cot -y] 2
1'he substitution of X for x + iy and Y for x - iy in this equation yields
I - sin a(XY + b2 - r2)(X - p)e-,,.
+ sin .B[(X - p) (Y - p) + a2 - s2 ]X}
X {sin a(XY + b2 - r2) ( Y - p)e',.
- sin .B[(X - p)(Y - p) + a2 - s2] Y l
= 4ik2 sin2 a sin2 .8 sin2 'Y
2
X { XY - � (X ( l - i cot -y) + Y(l + i cot -y)] } (6-7)

The highest powers of X and Y in this equation are X3 and Y3 ; the

asymptotes are therefore obtained by equating their coefficients to zero.
160 KINEMAT I C SYNTHESIS OF L INKAGES

The coefficient of X 3 is
[ - sin aYe-i-r + sin i3(Y - p)][sin aY(Y - p)eh - sin i3Y(Y - p)]
The requirement that this coefficient be zero will be met if any one of the
three equations
Y=O Y=p sin a Ye-i-r = sin .6( Y - p)
is satisfied. When written 1n terms of cartesian coordinates, these
equations become
= =
x-P
X - iy
sin fJ cos a
sin 'Y
0
- i
.( .
X -

Y- p
p - iy 0
sin fJ sin a)
sin 'Y
= 0
(6-8)

They represent a set of three parallel asyn1ptotes corresponding to the

triple point at infinity in the direction t' = i. (Parallel lines intersect

t" = -i is obtained by equating to zero the coefficient of Yn3 in Eq.

at infinity.) Another set of three parallel asyn1ptotes in the direction

(6-7), i.e.,
[ - sin a Y(X - p)e- i-r + sin i3(X - p)X][sin aXei-r - sin .6(X - p)] = 0
This require1nent will be satisfied if one of the following equations holds:
X =0 X = p sin aXei-r = sin (3(X - p)
These transforrned into cartesian coordinates are
+ iy = 0 X - p + iy = 0
_ sin � sin a) =
X
_ sin� cos a + (6-9)
x p i( y p 0
sin 'Y sin "Y

Singular Foci
For a curve passing through the cyclic points- as the four-bar coupler
curve does-the points of intersection of asymptotes of the curve in the
direction of the cyclic points are called singular foci. 1 Since the coupler
curve has two sets of three parallel asyn1ptotes of this type, it has a total
of nine singular foci. Examination of Eqs. (6-8) and (6-9) shows that
three of these intersections are real; i.e., there are three real singular foci.
They are the origin OA; the point On (x = b, y = 0), and a third point
Oc of coordinates

Xoc = P
sin fJ cos a
SID 'Y
Yoe = P
sin fJ sin a
•
Sill 'Y

1E. N. Laguerre, Sur !es courbes planes algebriques (1865), fron, "Oeuvres de
Laguerre," vol. II, Gauthier-Villars, Paris, 1905.
FOUR-BAR COUPLER-POINT C U RV E S 161
which also lies on the circle of foci defined earlier in this section.- · It
may be further observed (Fig. 6-3) that the angles at OA , 08, ·and Oc
of the triangle O AOnOc are respectively equal to the angles at A , B, and
M of the triangle ABM: the triangles OAOBOc and ABM are therefore
sin1ilar.
In summary: Starting fro1n the coupler-curve equation, a series
of n1anipulations identified the points OA , 0 n, and Oc with unique prop­
erties of a sixth-degree equation, viz., that they constitute what are
called the three real singular foci of the curve. Repeatin� ahnost exactly
the words of Roberts, we are led to ren1ark that, since the singular foci
are sin1ilarly related to the coupler curve, we might have taken as fixed
centers the focus OA and the third focus Oc, and by n1eans of links of
suitable lengths we should obtain the same coupler curve. In like m.an­
ner ,ve might have taken as centers Oa and Oc. We conclude, then,
that the coupler curve can be described in three different ways by four-bar
linkages having fixed centers at any two of the singular foci and couplers
forming triangles that are similar to the triangle of the three foci.
This problem of the triple generation of coupler curves will be
reconsidered in Sec. 6-4 by geon1etric means, and a complete determina­
tion of the three four-bar linkages will be given there.

6-3 DOUBLE POINTS A N D SYMMETRY

Coupler curves for 1 0 four-bar linkages are shown in Figs. 6-31
to 6-40 of the appendix of this chapter.n1 'fhe four-bar linkages generat­
ing these curves fulfill the Grashof condition and are of the crank-rocker
type. The crank at the left has unit length in each figure. The lengths
of the coupler A , the rocker B, and the frame C are given in n1ultiples of
the unit crank length. Coupler points are indicated by sn1all circles;
they are spaced at unit intervals on a rectangular grid carried by the
coupler, giving the coupler points convenient coordinates with respect
to the coupler. The length of each dash corresponds to 10° of crank
rotation (in the original there were twice as n1any dashes, each worth 5°).
Fron1 the length of the dashes a fair idea of the linear velocity of a coupler
point may be gained ; the esti1nate of the linear acceleration will be
rougher.
Coupler curves have a variety of shapes, as inspection of the
figures will show. These figures, selected for their possible utilization
in problems of this text, do not show all possible features because of the
arbitrary disposition of the points on the coupler grid and the ratios of
link lengths. In general, coupler curves may possess double points
1Redrawn from "Analysis of the Four Bar Linkage" by Hrones and Nelson
by permission of the Massachusetts Institute of Technology Press.
162 KINEMA'r I C SYNTHESIS O F L I N K A G E S

(cusps and crunodes) a�. well as syn1n1etry about an axis. The properties
of the curves, difficult to study algebraically because of the unwieldy
equation, will be examined geon1etrically. This chapter devotes itself
to geometric features other than curvature ; this very in1portant topic
is the subject of Chap. 7.
A double point is a point on a curve at ,vhich the curve has t,vo
tangents. A double point may be of two types: a crunode, at which the
tangents are distinct, the curve crossing itself; and a cusp, at which the
tangents are coincident, the curve being tangent to itself.

Cusps
The most familiar example of the cusp is derived fro111 the curve traced
by a point on the periphery of a rolling ,vheel (Fig. 6-6a). The curve is
the common cycloid, one of the special cases of the trochoid. We should
recognize, before going further, that P is a point on the n1oving centrode
11' m and that I, a point on the fixed centrode 1r1, was the instantaneous
center of velocity for the moment that P and I were coincident. It is
quite evident that P came down to I, stopped, and moved off in a direc­
tion opposite to that of approach. The velocity of P, although zero at
an instant, experienced no discontinuity. We note that a cusp is a curve
property associated with a point on a moving centrode and with the
relative motion of centrodes. Thus, if a coupler point happens to lie on
the moving centrode of the coupler, a cusp will develop at the position
where point and instant center are coincident; furthermore, the tangent

•
p

(a) Cycloid with a cusp

at I where the curve is
I tangent to itself
I

(b) Cusp at which both branches

lie on the same side of
the coincident tangents

FIGURE 6-6 The cusp, a. double point with coincident tangents.

FOUR-BAR COUPLER-POINT CURVES 163

1r,

r
5 4 3 (fixed 11
..----'----;__ centrode) 10
6 9 12
C1
2 11
7
5 10
4 Di
6
OA 2
7 11

8 10
9

9
(a) Four-bar linkage OA ABOB (in position 1) and fixed centrode 1r1

1r"' (moving centrode) D

(b) Plane of coupler showing the moving centrode r"'

I<'IGURE 6-7 Cusps of coupler curves.

of the cusp will be normal to the fixed centrode. The cusp may also take
the form shown in Fig. 6-6b, in which both branches of the curve lie on
the same side of the common tangent ; this tangent is also normal to the
fixed centrode.
We may see the action in a four-bar fron1 Fig. 6-7. The linkage
is shown in its entirety in Fig. 6-7a; the coupler link is AB, located in
164 K I N E M A T I C SYNTHESIS OF LINKAGES

6-8 Prolate cycloid

1:--IouRE
with a (symmetrical) crunodc
at Q.

position 1. Figure 6-7b shows the isolated coupler, a portion of its plane
with the n1oving centrode sketched in, and four coupler points-C, D, E,
and F-located on the moving centrode. The curves that these points
trace on the fixed plane are sho,vn in Fig. 6-7a; each coupler curve shows
a cusp for the instant at which the point on 11" touches 1r 1. For the instant
,n

depicted, the 1noving and fixed centrodes are in contact at C1 , whence

the cusp there. With rotation of the crank, the centrodes roll, and cusps
are formed by the other coupler points at appropriate positions.

Crunode
The crunode is a 1nore obvious fonn of double point than the cusp; as
noted earlier, the curve crosses itself and therefore has t,vo distinct
tangents. A si1nple exan1ple again derives fro1n a special case of the
trochoid, specifically the prolate cycloid (Fig. 6-8). With regard to the
four-bar, it will be seen that a crunode is related to the circle of singular
foci corresponding to the coupler point .
A..
• coupler curve with t,vo crunodes is shown in Fig. 6-9. For

FIGURE 6-9 Coupler curve with

double points Q and Q'.
F O U R-BAR COUPLER-POINT CURVES 165
M,Q

FIGURE 6--10 Two positions of

a four-bar corresponding to a
double point at Q.

the crunode Q (Fig. 6-10), there rnust be two positions of the coupler A B
such as Ao1B 1 and Ao2B 2 for which the coupler point Jf assumes the san1e
position 1 Q on the fixed plane. Considering the quadrilateral OAA 1 QA 2,
in which OA .4 1 = 0,1 A 2 and A 1 Q = A 2Q, it is clear that OA Q bisects
the angle A 1 QA 2 = 2 {3. Similarly, OBQ is the bisector of the angle
B1 QB 2 = 2 -y. However, since the coupler is rigid, 2 {J = 2 -y, or fJ = -y;
the vertex angle of A1 QB1 is 2{J + a; for A 2QB 2 it is a + 2 -y. Since
/3 = 'Y, the angle OA QOs = {3 + a + 'Y is therefore the same as the vertex

OA QOs = A1QB1 = A 2QB2

angle of the coupler,

Consider now a triangle OA OsOc sirnilar to ABM constructed with

OA 08 as base (Fig. 6-11); the last equalities then yield
· oA QOs = OAOc0 8
which implies that the point Q must necessarily belong to the circle
passing through points O.� , Os, Oc. It 1nay be recalled that this point Oc
1This assumption is valid provided that the IC he uniquely defined and be
distinct from Q. See Bricard, vol. II, p. 308.

Oc
/ Circle of foci

Q'

FIGURE 6-11 The double points

of a coupler curve, if any, lie on
the circle of foci.
166 K I N E M A T I C SYNTH E S I S OF L I N K A G E S

is the third singular focus of the coupler curve (Sec. 6-2) and the circle
OAO BOc the circle of foci whence a coupler curve has crunodes at each
of its intersections with the corresponding circle of foci. If the curve
does not. intersect the circle of foci, then it has no double points.

Symmetry

Coupler curves which are sy1nn1etrical about an axis may be generated

by a four-bar linkage with a coupler base .4B and follo,ver of equal length,
AB = OBB (Fig. 6-12). The coupler point generating a syn1metrical
curve must then lie anywhere on the circle centered at B and passing
through A.
Since BOB = BA = B!vl, the above circle also passes through O n
and the inscribed angle AOnll1 satisfies the relation
ABA1 B
A OB M = = = const
2 2
Consider now the linkage in two positions OA A 1B 1 OB and OA.A 2B 2OB for
which points A 1 and .4 2 are syn1n1etrical with respect to the line of fixed
centers O A On (Fig. 6-13). For these positions, triangles OBAe1 B1 and
OBA2B 2 are equal, since corresponding sides are equal, whence fl 1 = fJ 2•
Now, the isosceles triangles OBB 1 llf I and OBB 2M2 are equal,
() ,:JI 1 = 0 nllf2

-��---B

t, F
'- '- \
A \

'- ,..._ \
B
'- ,,�
"\
"\
o,.

FIGURE 6-1 :! Sy1nn1etrical coupler curve.

 __ ......---•·-·
........._ - --

1'' 0UR-BAR COUPL ER-P OINT CURVES 167

C
..........._
�

M,C-'-" ,, 1'
·' · " : ·e
. {
.
c+> ; ,#�i .,- .,-)/'/
_e_ ...,,_"".:' -� 8 /\\ /

FIGURE6-13 Two positions of a four-bar corresponding to

symmetrical points ftf, and M2 on coupler curve.

and the midnorn1al c to llf1M2 passes through OB and bisects the angle
M10BM2, whence

Since the angles OA O aA 1 and OAOsA 2 are also equal (A 1 syn1n1etric to

A 2 with respect to O A.08),
fJ
,,. + a = -
2
+o

Adding the last two equations yields

=
')' 2-
fJ

The n1idnorrnal c to A,fiJ.lf 2 therefore n1akes a constant angle with the

line of fixed centers OA Os , whence it is an axis of synunetry for the
coupler curve generated by point 111. It 1nay further be noted that
symmetric points on the coupler curve correspond to syn1n1etric positions
of the crank with respect to the line of centers OAOs .
The foregoing situation was a sufficient condition for syn1n1etry,
but not a necessary one. For exan1ple, the Watt straight-line mecha-
168 KINEMATIC SYNTHESIS 0 1'' LIN KAGES

nism, in which the coupler point lies on the coupler line AB, traces a
coupler curve symmetrical with respect to the line of centers O A OB. *
Note that the line of centers is also the circle of foci and that the curve
has a crunode at its intersection ,vith the line of centers.

6-4 T H E R O B E R T S - C H E B Y S H E V T H E O R E l\-1 1

A ren1arkable property of the planar four-bar linkage is found in

the Roberts-Chebyshev theoren1n: Three different planar four-bar linkages
will trace identical coupler curves. The develop1nent of this theoren1 is
presented in modern forn1, and extensions related to six-bar and slider­
crank mechanisms are given.

Coupler Curves and Cognate Linkages

Kinematically equivalent four-bar linkages-usually called equivalent

linkages- are comn1only used for velocity and acceleration analyses of
planar direct-contact mechanisms such as cams and noncircular gears.
The equivalent linkage is really an analog of the direct-contact mecha­
nis1n, and it "·ill be reme1nbered that the dimensions of the linkage change
with time, i.e., ,vith the position of the n1echanisrn: the equivalent linkage
might ,vell be called an instantaneous linkage.
When the designer's concern is only with the curve traced by a
coupler point of a planar four-bar, then other planar linkages tracing an
identical coupler-point curve may be found by the application of the
Roberts-Chebyshev theoren1. For want of a na1ne, these linkages­
related through their conunon coupler curve-will be called cognate
linkages. It is emphasized here that these linkages do not look alike:
their relation stems only from the identical coupler curves they trace.
However, in contrast to the equivalent linkages, the dimensions of the
cognate linkages do not change with time, and a cognate linkage may
therefore be substituted for the entire cycle of the 1notion of its related
linkage : this provides a linkage whose space requiren1ents n1ay be 1nore
favorable than that of the (original) linkage being replaced. The
velocity and acceleration characteristics of the cognate linkages will not
in general be identical link for link.

• See Prob. 6-4.

1 1\,1uch of the 1naterial of this and the following section appeared
in Machine
Design, Apr. J 6, 1959, and is reprinted by courtesy of the Penton Publishing Company,
Cleveland.
l<' OUR-BAR COUPLERC
-POINT CURVES 169

The Roberts-Chebyshev Theorem

Roberts 1 and Chebyshev were n1athe1naticians of considerable stature
during the latter part of the nineteenth century. Both were members
of the Royal Society, and both studied, an1ong many other things, coupler
curves of four-bar linkages.
The French and Gern1an literature speak of the "Roberts' theo­
ren1," while the dual name appears in the Russian. It seen1s proper to
use both names: Roberts announced his discovery in 1875, Chebyshev
in 1878. Except for the final result, there is no resemblance between the
two developments, for the approaches are as different as they can be.
To explore the theore1n, three different planar four-bar linkages
will trace identical coupler curves, we shall consider a four-bar linkage and
see ,vhat can happen to it (Fig. 6-14). The given four-bar is OA ABOn,
shown in Fig. 6-14a (drawn in solid lines) and carrying the coupler point
M, which traces a planar curve (not illustrated). On the left., the dash­
line parallelogra1n involving O..tAn1 and .i\1 A 1 is added, and the triangle
A1MC1 is constructed si1nilar to triangle ABM-note where the angles
are. On the right, a sin1ilar construction of a parallelogran1 brings us
to the point C 2. A third parallelogra1n is then constructed to bring us
to the point Oc, which ,ve shall assume for the ti1ne being to be a fran1e
point.
We can distinguish a total of three four-bar linkages mutually
connected at their coupler points M :
OAABOn (solid) the given linkage
o....4 1C10c link OAOc not shown
OBB2C20c link OBOc not shown
We also recognize, intuitively because of the parallelograins, that
the entire complex of 10 links is movable. (The IO-bar linkage is, in
fact, overclosed-there are 2 n1ore links than necessary; thus, 2 links
such as 0,1 A 1 and 08B2 could be ren1oved, to leave the 8 re1naining links
as a constrained n1echanism.)
It is thus clear that each of the three four-bar linkages traces,
through the identical coupler point ilf, identical coupler curves. This
allows the replacen1ent of the given linkage by either of the other t,vo,
whose space require1nents are different because of link lengths and one
frame-joint location, Oc. These are then cognate linkages, the relation
stemming frorn the comn1011 coupler-point curve.
For the theore1n to be true, it is necessary that the fran1e joint Oc
be truly fixed, as are o... and 08 : we 1nust justify our earlier assun1ption.
1
San1uel Roberts (1827-1913). Another Roberts, Richard (1789-1864), has
given his name to the Roberts ''straight-line'' motion, which he described prior to 1841.
170 KINEMATIC SYNTHC
ESIS OF LINKAGES

0
-- - - �
w.;.;��L.

C10-- -- - / 1\ \ \
�.\
� \
(!,,.,,, ')' \

'

,/I
'B2

,\ \
:wr;-"'�
.
\ '

,
OA Ji ' Os
,

(a) Given linkage and curve of coupler-point M (b) Construction for cognate linkages

;zr;�
Oc

'
/ \ ",, '\

A 1 s1/
I
,.
£4,�
"'
;.,, �
•:• "'
'
\
I
I '��
2

I \ \
/; I ', \
.�:.<t:
I

(c) Lefthand cognate linkage and

� � (d) Righthandcognate linkage and
�k:.
curve of coupler-point M curve of coupler-point M
FIG1;RE 6-14 Three cognate four-bars and their identical coupler curves-the
Roberts-Chebyshev theorem.

To show this by purely geon1etric construction is tedious, but it engaged

the attention of mathematicians such as Cayley, Clifford, and Kleiber,
among others. The simplest demonstration seen1s to be that of Ja. B.
Schor (1941) 1naking use of complex nun1bers: it is given by Bloch and
is somewhat as follows. 1 In Fig. 6-15a the point Oc will be fixed if
z = OAOce'6 = const (6-10)
The complex-number proof goes back to Hart in the "Messenger of Mathe­
1

matics," p. 32, 1883 (quaternion proof of the triple generation of three-bar motion).
FOUR-BAR COUPC
L ER-POINT CU RVES 171

This n1eans that OAOc and the angle � ,vill have to be expressed in terms
of the invariant dimensions of the first linkage OAABOa and must be
independent of the angular displacements t/>1, f/>2, and q,3 of this linkage.
These angles are measured counterclockwise fro1n the x axis, for con­
venience laid through the line 0.1.0 s ; t/> 1 and t/>a also appear in the cognate
linkages.

D.�\ --� (x)

\\ \·
--�\��o�s---� x
l :.-
(a) Sketch for showing that
joint Oc is a frame-point

(b) The over-closed ten-bar linkage

1
FIGURE 6-15 Cognate four-bar linkages.
172 KINEMATIC SYNTHESIS OF LINKAGES

Starting fron1 0A, and taking advantage of the parallelogran1s,

we may write
z = O,tA 1eH..t�a) + A 1 C1 ei(<1>,+a) + C1 Ocei<<1>,+a) (6-11)
Fron1 t.he parallelogran1s and similar triangles, the first factors of the
above equation may be found in terms of the link lengths of the given
linkage,
AM
(a)
AB

A 1C 1 =AB
M
;:: OA A (b)
MC2 AM
=
},f B2 AB
or since 1l1C2 = C 1 Oc and MB2 = BOB,
AM
C1Oc = BOB (c)
AB
With a hint fron1 Eqs. (b) and (c), Eq. (a) may be put into a more con­
venient forn1,
OAA t = - AB
AM
(a')
AB
Substituting Eqs. (a'), (b), and (c) in Eq. (6-11), ordering the terms, and
factoring, ,ve get

(6-12)

The terrns in the parentheses are recognized as a vector sun1 equivalent

to OAOB, or
AM
Z = - - OAOne•a (6- 13 )
AB
On comparing Eq. (6-10) with Eq. (6-13), it is seen that (1) OAOc =
(A1"1ifIAB)OAOB = const and (2) o = a = const.
This shows not only that Oc is a fixed point but also that the
triangle OA OcOB is sin1ilar to the original triangle of the coupler, AME.
From this last information, we may in1mediately locate the point Oc:
it is merely necessary to construct on the fran1e link 0 4_ 08 a triangle
similar to AMB; the upper vertex is Oc.
The location of 00 was part of Roberts' demonstration. Cayley
suggested the plan of Fig. 6-16; it is a simple way of determining the link
lengths of the cognate mechanisn1s. In this construction the given
F O U R - B A R COUPLER-POINT CURVES 173

linkage is pulled out straight, as it ,vere, and the parallel lines are drawn
to define the cognate linkages.
The four-bar linkage invoked so far for den1onstration purposes
had the coupler point M lying to one side of the line AB. The theorem
still applies when the coupler point lies on the line AB, either between
A and B or beyond A or B. The deter1nination of the cognate linkages
requires a bit more care now, for all the links of Fig. 6-16 will lie on top
of each other, the cognate couplers having becon1e "lines."
As Jf moves closer to the line AB, it is apparent that the general
geon1etry is preserved as the links approach collinearity with each other
and the line OAOB ; OAOs will be divided in the same ratio as M divides
AB. Note that C1 and C2 will divide A 1M and MB2 sirnilarly.
As an example of the case where M lies between A and B, consider
the linkage of Fig. 6-17, which would be a Watt linkage if the coupler
point M lay at the midpoint of AB. Referring to Fig. 6-17b, ,ve then
note the following:
1. The frame point Oc lies on the line OAOs and divides it in the
same ratio as J.lf divides the line AB.
2. 0AA 1 is parallel to AM, and llfA 1 is parallel to O A ,4., thus
defining A 1 .
3. 08B 2 is parallel to MB, and 1lfB2 1s parallel to 088, thus
defining B 2-
4. C1 will divide the line A 1M in the sa1ne ratio as 1°111 divided AB,
allowing the link C1 0c to be drawn to complete the left-hand linkage.
5. C2 will divide the line 1lfB2 in the san1e ratio as M divided AB,
allowing the link C20c to be drawn to con1plete the right-hand linkage.

F I G U R E 6-16 Plan for determining the

lengths of cognate links.
174 KINEMATIC S Y N T H E SIS OF LINKAGES

B B

(a) Given four-bar fb) Cognate construction

(c) First cognate (d) Second cognate

FIGURE 6-17 Cognate four-bars, M between A and B.

For the case in ,vhich M lies on an extension of AB, the method

is the san1e as before, except that the line division is now external, follow­
ing the position of M.
Cayley sho,ved that the coupler point M and the instantaneous
centers of coupler and frarne of each of the three linkages are collinear
at all times and that this line is the norn\al to the coupler curve (see
Fig. 6-18).

Velocity Relations in Cognate Linkages

Returning to Fig. 6-15b, assu1ne that the angular velocity of link o.. A
of the given linkage OA A BOJJ is w2• The angular velocities of the other
n1oving links AB and ORB, deter1nined in so1ne convenient ,vay, would
be wa and w4. Since OA A is parallel to .4 1 M, OcC 2 parallel to C1M, and
the angle A ill:fC I fixed, the angular velocity of link OcC 2 of the right-hand
cognate linkage is also w2. Si1nilar considerations are applicable to other
l<' O U R - B A R C O U PLEH.-POINT C U RVES 175
links of the cognate linkages, and the following table of velocity equiva­
lences results :
LEFT COGNATE GIVEN RIGHT COGNATE

These velocity relations 1nust be taken into account ,vhen the

point JI;/ is to be driven along the coupler curve C ,vith prescribed veloci­
ties. For example, if the desired motion M is obtained fron1 the linkage
OAAB08, with OA A driven at a constant angular velocity w2, the same
motion of M (curve and velocities) will be obtained by using the right­
hand cognate linkage OsB1C20c and driving the link OcC2 at the same
constant angular velocity w2• If the left-hand cognate linkage OAA 1 C10c
must be used, then it will have to be driven at a variable angular velocity
corresponding to the angular velocity of either AB or 08B when OA ,4.
of the given linkage is driven at the constant angular velocity w2•

Historical Note
There is a salient contrast between the developn1ent and staten1ent of
the theoren1 as given here and the original works of Roberts and Cheby­
shev. Pararnetric equations for the coordinates of the coupler point of
the Watt linkage had been derived by the French engineer Prony (1755-

FIGURE 6-18 Collinearity of

�he three coupler-and-frame
instantaneous centers and
Point M.
176 K I N EMATIC S Y :-; T H E S I S O F LINKAGES

1839). · · Eighty years later, Roberts' analytical investigation showed the

curve to be of the sixth order and demonstrated the presence of the third
singular focus of the coupler curve, i.e., the point Oc which had with
respect to the coupler curve all the properties of the fixed centers OA
and O8 . He deduced that the san1e curve could therefore be generated
by three different four-bars.
Chebyshev had studied four-bar linkages to generate approximate
straight-line segments, arriving at the particular form kno,vn as the
Chebyshev straight-line n1otion. By geon1etric perception, and reasoning
fron1 similar triangles, he derived the theorem for this sin1ple case, before
generalizing it to all coupler-point situations.

6-5 EX T E N S I O N S O F T H E
R O B E R T S - C H E B Y S H EV T H E O R EM

a. Cognate Slider-crank Mechanisms

A. 1nodification of the Roberts-Chebyshev theore1n applies to the slider­
crank n1echanisn1n: two different planar slider-crank mechanisms will trace
identical coupler curves.
If a sliding pair replaces one of the turning pairs of a four-bar
linkage, a slider-crank mechanisn1 results (Fig. 6-19a). The center On
is at infinity, with the link B0 8 now infinitely long. On "straightening"
this linkage (Fig. 6-19b), we can draw only a portion of the schematic,
for B08 now extends to infinity to the right (arguing by analogy). Under
these circumstances the linkage "above" OB vanishes, for practical pur­
poses, and Oc finds itself at infinity. Beyond what n1ight have been C2
lies Oc -at infinity-whence C1 n1ust also be a slider. However, the
cognate linkage, the slider-crank OAA 1C 1, sharing the coupler point M
with the given linkage, has been defined.
The construction of the cognate slider crank may be followed from
Fig. 6-l 9c; a parallelogram and a sin1ilar triangle are added as shown.
The point 13 is the instantaneous center of the given coupler and frame,
35 is the transfer center, and 1 5 applies to the cognate coupler. The
course of the ne,v slider is then along the perpendicular to 15-.-'>7 drawn
through t-, 7.
It is still necessary to justify that OACi is a straight line of constant
inclination. Considering Fig. 6-19c, we note that the triangles C 10AA 1
and OABA are si1nilar, whence a1 = a. Now �1 = o + a1, and

but /j1 = {:J or o + a1 = () + a, ,vhich with a1 = a means that

o = () = const
FOUR-BAR COU PLER-POINT CURVES 177

B
l
(a) Given slider-crank mechanism
,
\
and curve of coupler-point M (d)The cognate slider­
crank mechanism

A
(b) Plan
B
fl
08 , oo

.//' .
/ '
17, oo
/ ,/ 14,co

35� /

15 /

4
X 0�
(c) Construction for cognate slider-crank mechanism wuur#HH.
FIGURE 6-19 Cognate slider-crank mechanisms.

Hence the point C 1 (the cognate slider) follows a straight line of constant
slope. The point C1 could thus be used to trace a straight line-no
cognate slider is needed for this.
As with the four-bar linkage, the instantaneous centers of the
coupler and frame of each of the slider-crank 1nechanisn1s are collinear
with coupler point 1.lf at all times; this line is also the nonnal to the
coupler curve.
The velocity relations are the following:

W2 = W5 = W6
W3

vc, _ O.-iA1 _ AM
VB AB AB
1 78 K I N E M A T I C S Y N T H E S I S O F L I N KAGES

(a) Given four-bar

\---_ Antiparallelogram
I
\
-,
-, '
',
\
\ ''-
I
\
I '
I \
I
\' '
\
M
\
•I /
\

I
,·,._...,.--...:c0A ) I ---/-,I
\ • :J/
/}::, -,-,-\ •
" ....
\� / / \ I A,�:
,, ..,_, \ I�f///2,"' ·
,--- ' k
- '.l,•t1f·�:...
- - 1{ .. .-�
. .. :"::!, !Ji;,:�
I
C- --"

VilfJJ;
\
, , At?:ff#!1t:?1lf> ·
"""

____ ,,,...,,
B1 /
/

.......
(b) First cognate six-bar

B /
,,,--.... ,'
/ \ Antiparallelogram
\
I
I \
I \
/ I
/ I
'
M ,I

I t--/�
I

'
'
I \'
I I
I \
I \ ____J!l��
--1--\ /'Os / Os
��i..---
I \ ��---�,
L/...,... I ,
: \
I ' /
/ / /
/
\ /
C i
I \
/
\ / \ /

r#_!!P'�-- -- --- - B
',-�B1
\
I / /.
� ··is'•' _J:f -- '
,I
I
1

Bz _./
/

(c) Second cognate six-bar

F'IOURE 6-20 Four- and six-bar cognate linkages, }\[ off the line AB.
1'' 0UR-BAR CO UPLER-POINT CURVES 179

b. Six-bar Linkage Cognate to a Four-bar Linkage

An extension of the Roberts-Chehyshev theore,n is as follows : The
coupler-point curve of a planar four-bar linkage is also described by the joint
of the dyad of a proper six-bar linkage.
The course of this extension ,nay be seen fron1 Fig. 6-20b. Here
OAABO8 is the given four-bar linkage carrying a coupler point ill. The
procedure for finding the six-bar linkage n1ay be seen to be as follows :
1. Fron1 0A and J.lf construct a parallelogran1 locating C.
2. Fro1n O8 and .Al construct a parallelogra1n locating B1 •
a. 'fhe links },r/C and .AfBi are the dyad whose joint is at M.
4. Construct the triangle A 1OBB1 sin1ilar to A ilfB : it is reversed
and upside do,vn ,vith respect to A 'fl;/B.
5. Connect C and A 1, for1ning a third parallelogra1n OA O 8A 1C.
An identical dyad is fonned when the other frame point O8 is used
(Fig. 6-20c).
1'he situation in which the coupler point !IJ lies on the line AB,
either between A and B or beyond A. or B, follows the san1e pattern.
If the fran1e parallelograrns go into the antiparallelogran1 con­
figuration at the change points, then con1pletely different coupler-point
curves will be traced.
It is apparent that the Roberts-Chebyshev theore1n is useful in
those proble1ns of synthesis involving the coupler-point curve, for the
linkages that are related through the curve are identified. l'he choice
of the cognate 1nechanisn1 is dependent on space require1nents (link
din1ensions and fran1e-point locations), velocity and acceleration con­
siderations, and the value of the transrnission angle during certain phases
of the 1notion.

6-6 S T R A I G H T- L I N E 1\:1 E C H A N I S l\-1 S - A P P ROX ll\,J A'rE

A N D E XACT
The Watt linkage and "great bean1" gave engines of the early
nineteenth century vast bulk, and other, n1ore con1pact "parallel n1otions"
of only pin-connected men1bers ,vere sought. Of course, the Watt link­
age was studied for optimun1 proportions, but different four-bar linkages
Were devised, an1ong then1 the Evans (United States), or grasshopper
(Fig. 6-21a); the Roberts1 (Fig. 6-21b)a; and the Chebyshev (Fig. 6-21c).
Nun1erous sin1ple planar linkages possessing the geometric rela­
tions necessary for generating true straight lines followed. We shall
discuss two, the Peaucellier and the Hart.
1 This Richard Roberts, engineer, is not to be confused with the 1nathen1a­

tician Samuel Roberts; see Sec. 6-4.

180 KINEMATIC SYNTC
HESIS 0 1" LINKAGES

(a) Evans, or "grasshopper"

(b) Roberts (c) Chebyshev

:nGURE 6-21 The hest-known "straight-line" motions other than ,vatt's.

The true straight-line pin-connected linkages involve a geo1netric

relation kno,vn as inversion (not the kine1natic inversion, i.e., the succes­
sive fixing of links of a chain to create a variety of 1nechanisn1s). If
two points P and Q, restrained to move along a straight line passing
through a fixed point 0, n1aintain the relation
0P X OQ =k= const of inversion
then:
1. The points P and Q are said to be inversely related.
2. When the straight line rotates about 0, the curves traced by
P and Q are said to be the inverse of each other. In particular:
FOUR-BAR C O U P LER-POINT CURVES 181

a. If one point, either P or Q, traces a circle not passing through

0, the other point will also describe a circle ; i.e., the inverse
of a circle is a circle..
b. If one point traces a circle passing through 0, the other
point's circle will have an infinitely great radius; i.e., it
"·ill describe a straight line.
Mechanical devices realizing these conditions are kno,vn as
inversors, of ,vhich the earliest and best kno,vn, although not quite the
simplest, is that of Peaucellier (1864).
Let us first establish point 2b. In :Fig. 6-22, 0 and Op are fixed
points, with OpP = OOP, so that P describes a circle passing through 0.
Let Q be a point on OP satisfying the relation

OP X OQ =k
and consider the perpendicular Q1' to 00p passing through the point Q.
We shall sho,v that the distance OT depends only on the constant k and
the diameter of the circle so that, as P describes the circle, Q ,vill describe
the fixed straight line TQ. On drawing the line PS ,ve realize that the
triangles OPS and OTQ are sirnilar, whence
OP
OS
=
OT
OQ
or OP X OQ = OS X OT = k

and OT = !!:_
OS

Consider now the Peaucellier cell, the six-bar chain of Fig. 6-23a.
It is forn1ed by connecting a rhon1bus of equal sides s with t,vo bars of
equal length l; the figure is syn1n1etrical about the 1nedian m. Three
Points, seen to be ahvays collinear regardless of the configuration, have
been labeled 0, P, and Q. In Fig. 6-2:3b we now add t,vo n1ore links (the
frame and OPP, with OpP = 00p), displace m, and draw the line BDC,

I
I
I
0 T

FIGURE 6-22 Proof of proposition 2b.

1 82 KINEMATIC S YNTHESIS OF LINKA GES

, I ,,.- nt
\ I ,,

\:·" I
-' D ,,-, / ()
\

\\ I
I
2::-/-�\:--cc I
��Op- - - \I _Ji__
IT

I
I
' ,____ ,...., //
/
C I
(a) Peaucellier cell (b) Peaucellier inversor in which Q draws
the straight line QT perpendicular
to the base link OOp

FIGURE 6-23 Peaucellier straight-line mechanisnt.

which ,vill always be perpendicular toem. Then

OP X OQ = (OD - PD) (OD + PD)
= OD2 - PD2 = (OB2 - BD 2) - (PB2 BD2)
= OB2 - PB2 = l2 - s2 = const
-

The Peaucellier mechanism, since it satisfies the condition

OP X OQ = const
is thus an inversor.
Next we shall consider point 2a. In Fig. 6-24 0 and Op are again
fixed points, with OO P = d and O PP = r, so that P describes a circle
(which this time does not pass through 0, since r � d). Let P' be the
other intersection of OP with the circle and Q be a point on OP satisfying
the relation of inversion
OP X OQ =k
Draw PS and P'S', and consider the triangles OPS and 0S'P'. These
triangles are similar: their angles at O are equal; their angles at P and S'
are also equal, since they are inscribed in the sa1ne circle and subtend
the same chord P'S; the sa1ne holds for their angles at P' and S. Then

= OP'
OS OP
OS'
or OP X OP' = OS X OS' = (d + r)(d - r) = d2 - r2 = const
On dividing the relation of inversion by the above, we find the ratio
OQ k
OP' - d2 - r2
F O_U R - B A R COUPLER-PO INT CURVES , 183

When P describes the circle centered at Op, P' describes the same circle
and, because of the last relation, Q n1ust describe a curve similar to this
circle, with center of similitude at O : this curve is also a circle, and its
radius is
R =r k
dz - r2
In the Peaucellier cell, k = l2 - s2, whence
z2 - s2
R = r d2
- r2

The cent.er OQ of the circle traced by Q corresponds to the center Op in

the same similitude, whence
OOQ l2 - s 2
bop - d2 - r2
With the proportions sho,vn in Fig. 6-25, d < r and l > s, thee·ratio
of similitude is negative, and 0Q is therefore to the left of 0, that is, in
the opposite direction fro1n Op.
Peaucellier, captain of engineers in the French Arn1y at the time
of his invention (he rose to general), was n1otivated in his search by the
approxin1ate straight-line n1otions of his ti1ne and their use in engines.
However, the day of the big-beamed engines was drawing to a close, the
con1pact direct-connected or slider-crank engines supplanting their bulky
ancestors. In consequence, engine applications dwindled. The Peaucel­
lier straight-line mechanisn1 is, nevertheless, of n1ore than passing interest,
for it was the first device able to generate-create-a straight line in the

.\
Q

\
I

'\' '\
\I
'
I

\
\ I
s· d r
OQ 0 Op s 'I
II
P'

FIGURE 6-24 Proof of proposition 2a,

184 K I N E M A TIC S Y N T H E S I S OF LINKAGES

FIGURE 6-25 Peaucellier in­

versor for drawing circular
arcs of large radii.

sa1ne sense that a compass generates a circle ; using a straight-edge, one

merely copies an existing line.
The instrument's ability to generate circular arcs, especially flat
arcs beyond the practical range of a tran1mel bar, gave it some applica­
tion in manufacturing machinery and in the drafting roorn. As a co1n­
pass, the pivot point OP is made adjustable, allowing the distance d to
be selected for a desired radius.
A n1odern application of the Peaucellier linkage is the autofocusing
mechanism of a photographic enlarger. The optical problem involves
the proper spacing of negative, lens, and paper to maintain a continuously
sharp image over the range of magnification; the problem is summariied
by the equation xx' = I' (x and x' are the distances, f the focal length of

of the form OP X OQ = k.
the lens). This equation can be n1echanized by an inversor, since it is

Figure 6-26a, showing the optical syste1n, assumes that both

Gauss points are at the center of the lens, justifiable with a normal
photographic objective (but not a telephoto lens). Focal points are
indicated by F, with f the focal length. The film must be n1oved a dis­
tance x' while the paper moves the distance x, both relative to the lens.
l\fagnification is given by m = h/h' = u/v = x/.f. The Newton equa­
tion for the conjugate distances x and x' has already been noted as
x'x = r.
The rearrangen1ent of Fig. 6-26b groups the variables x and x'
about the lens and leads directly to the hard,vare sche1natic (Fig. 6-26c) .
Here the fa1niliar points of a 1nodified Peaucellier cell are indicated ;

sliders. We may no,v write Of> X OQ = k = l2 - s2 • Physically, P

P, Q, and 0' are constrained to n1ove along the san1e straight line by

is at the wrong spot, but OP = QO' = x', and OQ = x. ConsequentlY

FOUR-B A R COUPLER-POINT C U RVES 185

Negative
! h'
h
*x· T t 1 Negative

•F II
I.
I :
I
!
-+
! I
V
x· I
Lens
--+- -,-
- �

I. //
I
f

I
I
/f' %
u

I
I
I
I
pe Paper
�--
h_ 7 1!!1!
Pa i!l!!l!!l!iil!i
r !i!iiiii!i��-�-� ·········�·······•� _____ _
(a) (b) (c)
FIGURE 6-26 Autofocus enlarger employing Peaucellier mechanism.

the mechanization of the inversion relation is a direct analog of the

optical requirement, and l2 - s2 = /2.
Another mathematically correct straight-line mechanisn1, but of
only six links, n1ay be derived from the contraparallelogram chain of
four pin-connected links (Fig. 6-27), in which EB = CD and BD = EC.
Four points such as 0, P, Q, and 0', lying on a line m parallel
to BC (and hence also to ED), will divide the distances between the pin
connections in the san1e proportion. Furthern1 ore, the points will con­
tinue to remain in line when the chain is deformed. These points are
also related by inversion; thus, OP X OQ = const = O'Q X 0'P.
The inversion relation may be established after constructing CC

-----% ----�

B - - - - - - - -- - - - - C --

I
I I

E �------- Y
-
- ---------- -•-
D'
----
C'
---�ID / I

FIGURE 6-27 Contraparallelogram chain of Ha.rt.

186 K C NEMATIC S Y N THESIS O F L I N K A G E S·

FIGURE 6-28 Hart•� 1n-

versor.

perpendicular to ED and CD' parallel to BE. Fro1n si1nilar triangles

OP
= OE
.e. -r BE and
OQ
.
y- =
OB
BE;
!J
Q
B
'fhen OJ> X OQ = X (? xy = xy X const
BE BE ·
Now x = J.,;C' - C'D', and y = EC' + C'D', whence
xy = EC' 2 -C'D' 2
= (EC2 - CC'2) - (CD 2 - CC'2)
= EC2 - CD2 = 12 - s2 = const
and therefore OP X OQ = const for all configurations.
We recognize that, ,vhen O is n1ade a fixed point and P is guided
along a circle passing through 0, Q will trace a straight line (Fig. 6-28).
Point Op 1nay be chosen at will, subject to the above restriction. ,fhe.e
path of Q will be perpendicular to the base line 00p. This is the six­
bar mechanism of H. Hart (1875).

APPENDIX: ATLAS OF FOUR-BAR COUPLER CURVES

The curves sho,vn in this appendix are a selection fro1n the atlas
of Hrones and Xelson, as 1nentioned in Sec. 6-3. This atlas contains

B Follower

FIGURE 6-29 Notations used in

Hrones and Nelson's four-bar
� ----C ---- ,-....i coupler curve atlas.
F O U R - B A R C O UPLER-POINT CURVES 187
C C c·

4
A=3
• • A=4
3 • •

:t .�L
3 '.i� "' •
2 • 2 •
l��'--'---'-----=
1 2 3 4 B 2 3 4 B 1 2 3 4 B
FIGURE 6-30 Determination of the linkages to be considered in this appendix.

approxin1ately 7,300 curves dra,vn to large scale (730 pages, 1 1 by 17 in.)

and constitutes a very practical tool for the designer, who, by pag­
ing through, may find a shape and configuration suitable for a given
application.
The four-bar linkages considered here (as well as in the original)
are of the crank-rocker type, i.e., having continuous rotation of the crank,
,vith oscillation of the follower. With the notation shown in Fig. 6-29
the link lengths n1ust therefore satisfy the conditions
C<A+B-1
C > IA - Bi + 1
A, B, C > 1
In order to detern1ine what co1nbination<., of values of A, B, and C
are con1patible ,vith these conditions, consider the diagra1ns of Fig. 6-30,
-- --- ........ -- c.--
-.., --- ---
- ----- ---------
-
l<J'/
/
/

/ ✓/
\
\.�
,,..-r
).,.-
/'

'>--<>-
--- \
L\
-�-------
__ ---o._ ..... ............
....."..:::._
- ...__ -- . -.....,_
/ / ,-":"- ----- '- .._ - �
/' I \
.
I I /
.,,I--r
----- � - .....

I '' // -r--- ......_ '-,- -- �

--- -. '"-�...
\I
I/ I {_ _ _ _ ,- \ .-�
---- � -__....,,
; _ _ _ ' I
/
,,
/' .
-1..

I
�-- ----- " -� -- --
I :
-?
✓ ,.. - - �

I I { "" \
I r \ - - ----r / -,--\\ : ---------- --- \\
........._,_

' \ /- ---

\ I I \'\ /"t->-/ --:::.�--

/ -i', ...:....--- r: --
I -._
I - ..._,_
\ \f ---. / -� ✓ ',� 'I, �

, __ i' 1 \ / f ""
'-...

\ \1 •. \ � \, \ \
�
-
" - ---- -',
- '<-
�-- - ..... ___ \
"
\

\
\ ..__ " I
' /, " I \ ........ ___.___\ I
',
,., ',\

,.,
/' --- , ...._...I
\ ,_j

\\_ ',
\
\
', ..,_--.:
� /
I \
\
,,,

\ "- I - - \-- _:)

\\
'-...

"\ '-....____ //
'-...
---.......
- ---- , \

FIGURE 6-31 A = 2, B = 2, C = 2.
 K I N E M A T I C S Y N T H E S I S OF L I N K A G E S
- -- ------
188

�---- ---.... ..._-�-

_-_ .......
.--
,__..C
......_,____
..
----"?r- ----
.......
/ , .-
/
. .............
.::►.::-
---
---
�---- --\
✓ \ - '\ -- ,
,,,,,-r
,.,..-
:: -...

(.__ .--r
/ ·\ ,,,. ►-- �- ---

/ ..{
-- ,/ .....__ "-
/
- -- I -/
/
r /
/ I ---.....
l----. I I
/
/-r-..- / \','= //--- .....__ '\\
I /
I
I ,,,., ......1'I I '- - ,.,.f_.
I I(
.._
,\) I
/ I '-
\ 'k- 0-1
✓

I\ II I\ - ' "'\ --1--------.:.

-
/�./
\
/ I 1 \

\ I ,,, \ ,,.,,-/ ....,--.e::>,::;.::::---..._ // - : .

✓

/ ✓- ,.,. , - ---...... '

--

.- / I " J

\ I / ,, / L>, \ :;,, "" �:'P -......,.......,·:;:::· = J

� - ......_ ,}, \ .
.•/

\\ I ,-r '\_) \.f "" _ . _.. --,, -.::.

--

1/ .. _ .
'\I >-. I' . __=_ -•s;\ _
-l / ' I\ \
\

\"'"'" ____...........,.__\------ � I
�--- - -
✓

......_
",

/
J \
\
j

"'
------ \ ..,,,,
'-.._ I
--- /

FIGURE 6-32 A = 2, B = 3, C = 3.

-- -- -..... " ---

-- -
/
/ , -'<(
\ '
...... ...._ __
�---
' - --........._
✓

/ /
/ � -- - .....

/ / /
'.if /
\
f
\
/,,. )I
.,.v -���-==_, 'I
:::-...,__
..._
, -.......
' '-

----h.- - - /
/
...L
_,,.. / -- ...._ I , ,\ - ..._ ',

// - ---
II , I

",, I
-....i
-,,,L
I I
I �1\ _,,..- 7 - ..._
✓

II I \ ' ,, \\ ',
I ,,.,,,. / _:..-,--- ...... ,

\ I .,..- -\-
' . _ /. \
,,_- -- - - L
\I I/
'\

"-L..- -- - - --\

\ti"- ,,.{7'
I -·
'-� -- \-, - - -- "--...- I
\ "
\
If
I ':<�
/ ......._'�--r,....._""" ✓L
- ....... --
-----
a�- " - - ...._
.........
...._

l\ \ - - --2'"," '=
,.._ _.,\
\ \. "- ', '\

\
\ l \

\
I
I \ '',
\ --
/\. '
\ '-
\ ,,_ - _ .::;__
I -
\ " " \. / \ -----i-------·
\ \\ \
'. •.
/'
'
'" ",...1.. ___/
\ -----'><'
, .---
\

"- I
'- I
....._ __ - -- ✓

f'JGURE 6-33 ,{ = :3, B = 2, C - 3.

FOUU-BAR COUPLER POINT CURVES 189

FIGl'RE 6-34 A = 3, B = 3, C = 2.

FIGURE 6-35 A = 3, B = 3, C = 3.
190 KINEMATIC SYNTHESIS OF LINKAGES

FIGURE 6-36 A = 3, B - 3, C = 4.

•·1ouRE 6 - a i ,-1 = 4 , B = 2 , c = 4.
FOUR-B A R COUPLER-POINT CURVES 191

.l<'JGURE o-38 A = 4' B = 3, C ,.. 3•

,, /
.,,,,,.- ---- �--
........
. --�---..
' -
-' - ........
'- - - -....._
/'

)•
/ ..-
,,
\v,,.,--
-

/�- /✓
/ \ � ��
/
\ /' -- --�-
I I I,\' /\ : \ '- ..._e '
---+- __ __,�
I I I I
f I ...,_'t--- '\ '\.
I
//
.--1--
I - ...... I I \ I / .,,/
I / I 'I -� -
"' I ,,,. .,,, -,:::-�- �-.........-- -
..._ '
I ✓ I -- - -

t ,,..--�- / / \ ' --- -\ ' ..... '

\
-- .
\ I r-. _�_
\'..., / _.J/_ __.,, '\ /' ''-
\
I .' I / ( \''-- '\

---= - __',,
-- ---- ,,..._ _____,,
/ ,':::,.., --
·--- -.. ' I I
';fl
� ' -- - - ..... _ I

\ r , ____ ,,,,f \ I f,..._ -", ,,- '>. ,, :-;,'"'.; ---

I
// \ I \ .,,.---_
,,,
\ \ ,' - '<"- -1- -- \ ,,_ -- -"--
I '- ,-
;'\ \ \ \. , \\
�\
\ '-, \
\ // \\ \\''
I
\
I
\
-1--_____
\
)
---- -
' ,,,,
I
\
\
' -- --�/
I
", ' I
,
' l
_-,
I
_____
/I
-/

I
\
"' j,.. ____.,,.
'- I
'-
'-..._ __ /
/

FIGURE o-39 A = 4, B ==' 3, C = 4.

192 K I N E M A TI C S Y N THESIS O F LINKAGES

------- - - --
- --- --
.......

'
/ y�
.-
/ / \. ·""- - - - - ,,,.� - - - - ....
'�� ........
/ \ ,,,-," \ ,,,. ..... ...--·\
I ) ,......
)

1c,___ -;:/
/

/ _L_ 1/I \ /
✓ \\ ///
I/
\ /
-
/_' -- - - ,,,

,
..,,
I / /. .,- 7I -.... ' '\ �I,
'ii\ /, . / --:.-- '\ ,,...- --;<'
✓ ,,..,.. -
-- J.Jr----
11 I
,,, I ',, �-- > - - - - - --
/ I I '.';
i --�'- >�:::: � "
I / \ r '
-✓-- I
/ ,, ____ .,

\ _,,.,- -.........-- -:.... �

\ \ \I / ,·
'{
t------,/
_)-

=::::. - ---
\ / _, /
I

'' I f>' -----�

" ,:_.,-,,. ,, '' / ,, ,, ' ' ', _ __
--. ,.,
/' ,, �� -.,- -

r '-__ _ _ .,,f.-t
. ' \ / ' /T _..

\/,r,
\ '\.'

1 1 ,.}- - ·
·y------�,,,
-
'v
I \ l /\ '\
\ "-
'
\ '- /
I I -----r
I -- -
\
\
\
'-
...... _ - �".....
/

I
I-- - --1 -- -
' I __ .,,.
/

/
, I
\ ' ---
' --, ---
I /

'-, ---
---- .,, ,, . /

FIGURE 6-40 A. = 4, B = :3, C = 5.

where B and C are plotted on the horizontal and vertical axes and A is at
a 45° angle. The three diagrams correspond to A = 2, 3, and 4 ; larger
values of A are not considered. Each point in the plane of the diagrams
corresponds to a four-bar linkage, and values satisfying the above condi­
tions are located in the "rectangles." Limiting B to values no larger
than 3 (a follower no longer than three times the crank length) and taking
unit incre1nents for B and C gives a total of 10 linkages, shown as dots,
for which coupler curves are drawn in Figs. 6-31 to 6-40. The curves of
a nun1ber of coupler points are shown for each linkage, and each dash
corresponds to 10° of crank rotation, thus giving a representation of the
coupler-point velocity. l\1uch more detailed inforn1ation is given in the
original atlas of Hrones and Nelson, but only at the expense of a rather
overwhelming effect. The curves presented here will be sufficient for
preliminary designs, which may later be refined by the geon1etric or
analytical methods of synthesis presented in later chapters.

BIBLIOGRAPHY

Beyer, Rudolf: "Technische Kinematik," Johann Ambrosius Barth, l\lunich,

1931. Also, J. W. Edwards Publisher, Incorporated, Ann Arbor, l\1ich.,
1948.
--- : "Kinematische Getriebesynthese," Springer-Verlag OHO, Berlin, 1953-
English translation by H. Kuenzel, "The Kinematic Synthesi:- of l'v1ech­
anisms," McGraw-Hill Book Company, New York, 1963.
F O U R - B A R COUPLER-POINT CURVES 19 3

Bloch, S.eS.: "Angenaeherte Synthese von ::VIechanismen," translation from Rus-

sian, VEB Verlag Techni.k, Berlin, 1951.
Bricard, R.. : "L�ons de cinematique," vol. IL Gauthier-Vi liars, Paris, 1927.
Cayley, A.: On Three-bar lVIotion, Proc. London JIath. Soc., vol. 7, 1876.
Chebyshev, P. L.: Les plus simples systemes de tiges articulees (1878), from
"Oeuvres de P. L. Tchehychef," vol. 2, :VIarkoff et Sonin, St. Petersburg,
1907. Reprint, Chelsea Publishing Company, New York, 1 962.
Hartenberg, R. S., and J. Denavit: The Fecund Four-bar, Trans. Fifth Conj. on
Jt,fechanisms, Penton Publishing Co., Cleveland, 1958. Also, Cognate
Linkages, ltfachine Design, vol. 31, 1959.
Primrose, E. ,J. F.: "Pla.ne Algebraic Curves," The :.\'1acmilla.n Company, New
York, 1955.
Prony, G. F. de: "Nouvelle architecture hydraulique," vol. II, Paris, 1796 .
Roberts. S. · On Three-bar :.\1(otion in Plane Space, Proc. London Jfath. Soc.,
vol. 7, 1876.
 T HE EULER-SAVARY
EQUATION AND THE CUBIC
OF STATIONARY CURVATURE

7-1 THE Et;LER-SAVARY EQUATION AND THE

INFLECTION CIRCLE
The discussions of coupler curves have so far dealt with certain
particularities of a curve, such as how to achieve (or avoid)
double points and sy1nmetry. A further in1portant characteristic
is the determination of the center of curvature at points on the
curve by a direct n1ethod. By this is n1eant a procedure that does
not depend on first establishing the velocity and norn1al accelera­
tion component of the point, after which the radius of curvn.ture
and its center 1nay be found. Other things of interest include
being able to discover or predict coupler points tracing approxi­
n1ate straight-line or circular-arc segn1ents, which the designer
1nay exploit in the arrangen1ent of the rnechanis1n.

What is known as the Euler-Savary equation gives the radius of

curvature and the center of curvature of a coupler curve in rather
direct fashion. In the course of the developn1ent other welco1ne
information is gathered. The so-called inflection circle shows the
_location of coupler points whose curves have an infinite radius of
EULER-SAVARY EQUATION 195

curvature. Crudely put, the inflection circle gives the location of nearly
flat segments on the coupler curves. Actually the radius of curvature
is infinite at only one point along the curve, but the flatness associated
with a large radius of curvature 1nay extend for a useful distance to either
side of this point.
The curve called the cubic of stationary cu1 i•ature, 1 to be studied in
the next section, indicates the location of coupler points that will trace
segments of approxiinate circular arcs; the radii and extent of arc vary
from arc to arc, i.e., point to point.
It should be rerr1arked that the Euler-Savary equation and the
cubic of stationary curvature are not restricted to four-bar linkages but
apply to planar motion in general. The applications in this chapter,
however, are oriented to the four-bar coupler curve.
As we have seen, the planar motion of a link, such as that of a
coupler guided by crank and follower, 1nay for analytical purposes be
replaced by the rolling n1otion of a n1oving centrode against a fixed
centrode. 1'he clue to having the coupler or a coupler point foilow a
desired motion with a certain degree of approximation lies in a considera­
tion of what is happening at the point of contact of the centrodes.
1
This is the German Kreisungspunktkurve. Neither the original nor its
literal translation, circling-point curve, is directly helpful in conveying the significance
of the curve. Calling it the cubic of stationary curvature, as Professor Hall has done,
gives it a reasonable identification, since the curve is of the third degree.

1
Path normal n

y1, lmag

�+-----r-------x1, Real
01

FIGURE 7-1 Two planes rolling on their centrodes; definition of nota­

tion used for Euler-Savary equation.
 196 KINEMATIC SYNTHESIS OF LINKAGES

Consider a plane 2 moving with respect to a fixed plane 1 (Fig.

7-1), and let 1r1 and 1r2 be, respectively, the fixed and moving centrodes
of the motion. At the instant considered, the two centrodes are in con­
tact at the instantaneous center of velocity, I; as plane 2 moves ,vith
respect to plane 1, 1r2 rolls over 1r1 with an angular velocity w. The point
of contact I between the two curves-the instantaneous center I-shifts
along 1r1 with the velocity vr, identified in Sec. 4-7 as the IC velocity.a1
The concept of IC velocity of the instantaneous center I n1ay be
further clarified by considering two positions of the moving centrode,
1r2 at time t and 1r; at time t' = t + At (Fig. 7-2). A.t time t the instantan­
eous center is I, the point of contact between 1r1 and 1r2; at time t' a new
point I' is the instantaneous cente:. while the point I has moved to I 2•
The IC velocity is the limit of the ratio 11'/At as At goes to zero. Thus,
the IC velocity is not the velocity of any given point (or material particle)
but instead expresses how the instantaneous center shifts along the fixed
centrode ?r1.
Returning to Fig. 7-1, a point A of the moving plane 2 traces on
plane 1 a path CA, whose center of curvature OA is located on the line IA.
What is known as the Euler-Savary equation will establish a relation
among the positions of A, I, and OA on this line; in other words, it gives
the curvature of paths generated by points of the moving plane.
The Euler-Savary equation may be derived in several ways, but
the 1nethod based on the use of the norn1al component aAn of the total
acceleration of A reveals n1any of the physical realities and is presented
in the follO'wing.
1 The instantaneous center is sometin1es called pole; its velocity is then known
as pole velocity, which of course is the same as our v,. However, we wish to main­
tain the distinction between an instantaneous center (infinitesimal rotation) and a
pole (finite rotation). The IC velocity is also known as the displacement velocity
of the instantaneous center.

FIGURE 7-2 Successive in­

stantaneous centers / and
/' used to define IC velocity.
EULER-SAVARY EQUATION 197

�;�: �ti

---- Path
normaln

---
v,

J<'IGURE 7-3 Components of the acceleration of point A,

The total acceleration of A may be written as

0A = CA
1
+ CA" (7-1)
2
VA
in which laA"I = --
AOA

Here vA is the n1agnitude of the velocity of A, found by referral to the

instantaneous center /, that is, VA = !Aw; and AOA is the radius of
curvature of the path CA at A. An evaluation of aA' does not concern us.
In Fig. 7-1, the centrodes 11"1 and 1r2 are in contact at instant tat the
instantaneous center /. The rectangular axes 01x1 and 01y1, fixed in
plane 1, are respectively considered as real and in1aginary axes, allowing
vectors in planes 1 and 2 to be expressed as con1plex nun1bers. If w is
the angular velocity of plane 2, positive ,vhen counterclockwise, the
velocity of A is

The acceleration is obtained by differentiation of the velocity,

(7-2)

In this equation, ia(0 1 A - 0 1 1) = ialA, with a = dw/ dt, positive

counterclockwise. Also, (d/dt) (01A) = vA; and (d/dt)(O i l) = v1, the
 198 KINEMATIC .SYNTHESIS OF LINKAGES

IC velocity. 1 Since V.t = iwlA, we may assen1ble Eq. (7-2) in the forn1

C.t = ialA - w 2
(1A + i :) /J-· .,., ,
•/! C ,I - 4.>
' Ar i'
rJ
' ...
--·-i>
·-·
., ·-- �
- II ,('
,1... c,_ __ ,'j. l
J

,/-
Turning to F'ig. 7-3, it will be convenient to introduce- a---p9.int K ,, ·
-·- - ·
such that IK = -iv,/w, whence 1 r ,tl-
··•' I · ""�..._., ,.-
1 I'
11 __' ,-,,dJ
,,-,. ' /'
V.
1 �1tJ1
--·---
� ___.:: I

CA = ialA - w2 (1A - IK) ----- ---- - -

✓

)
I - C �I
.
.
- . \
) l' - I
A.�· J ·- , .,�
�t·:f,-· '
.

,.,_
-

Since IA - IK - KA, the acceleration n1ay be ,vritten as 2 w 1.

.).

,;-, :,✓

C.t = ialA - w2KA (7-3)

According to this expression, the acceleration is the sum of two
vectors. Vector ia(IA) is perpendicular to IA, whence it lies completely
in the path tangent of the curve; it contributes only to the component
aA' (not shown). Vector -w2 (KA) lies along KA and is directed toward
K; by its projections it contributes to both aA' and aA" (not shown).
We note with the aid of Fig. 7-4 that a unique situation exists if
the angle IAK is 90°, for then -w2 (KA) lies completely along the path
tangent; i.e., it has no co1nponent in the normal direction: for this case
1
We assume that v, is finite, i.e., that the centrodes are of different curvature
at their point of contact. For a discussion of curvature problems, see 0. Bottema, On
Instantaneous Invariants, Proc. Yale Conj. Jfechani8ms, Shoestring Press, New Haven,
Conn., 1961; also G. R. Veldkamp, "Curvature Theory in Plane Kinematics," J. B.
Wolters, Groningen, Netherlands, 1963.
2 This equation is valid for all points of the 1noving plane, including the IC

(point /). Note that, in Eq. (7-2), (d/dt)(OAA) � (d/dt)(OAI), even though I and
A may coincide at the instant considered.

Path normal n
/

Path
tangent I

PIGURE 7-4 Inflection circle and

shape of point paths on or near it.
 EULER-SAVARY EQUATION 199

... : ;.·
Path normal n

., .

·'

----11"2

FIGURE 7-5 Derivation of o,.

the Euler-Savary equation.

aA
n
=
0. The locus of all points A whose angle IAK is 90° is a circle of
diameter IK: this is the inflection circle. For all points of this circle
0A" = 0, and the radius of curvature of their paths is infinite (Fig. 7-4).
The point K, the intersection of the normal at I and the inflection circle,
is son1eti1nes called the inflection pole.
Consider novv a point A not on the intlection circle (Fig. 7-.5 ). 'fhe
normal acceleration aA" is a vector quantity having n1agnitude. direction,
and sense. The direction is always along the path norn1al n. If this
normal is oriented fron1 I to A, then the magnitude and sense of a,4." 1nay
be defined in terms of a real number a..4" (with 1nagnitude and sign).
Thus'
aA
n
= VA2
AOA
will be positive if AOA is p�sitive, i.e., if it has the san1e sense a.f: IA. If
A. 0A is in opposite sense from IA, aA" will be negative. In the figure, AOA
?as a sense opposite to that of IA., ,vhence the norinal aeceleration aA"
1s negative.

However, aA" is also the projection of - w 2 (KA) onto the path

normal n, namely,
(IA) 2
aA" = w2 -- = ProjrA [ - w2 (KA)] = - w2 ProjrA (IA - IK)
AOA
200 K I N E M A T I C S Y N T H E S I S O F L l !li K .-\ G F. S

The second intersection JA of the line J.4 ,vith the inflection circle is also

�t�2 = -
the projection of K onto IA. We may then write

w2 w2 (IA - IJA ) = - w2 (J AA)

(IA) 2
or - - J AA
AOA
The negative sign is renloved by recognizing that AOA - - 0Ail, whence
(IA) 2
OA A
= J AA

Rearranging this as
(7-4)

we have the Euler-Savary equation.

The Euler-Savary equation relates three directed quantities lying
on the path nor1nal n. The rule of sign introduced earlier reduces to
having OAA and .l AA ahvays laid off in the san1e sense along the line IA.
rrhus, when .lA has been established, the sense of .lAA gives the sense of
OA A. The converse is also true.
In Fig. 7-6 a n1oving link has been sketched, carrying a point A
inside the inflection circle. Fron1 considerations not displayed, the

Moving link

Inflection
circle

i-o Summary of the Euler-Savary equation. JA•·1

and OAA must always read in the same direction, toward A.
FIGt;Ri,;
E U L E R-SAVARY E Q U A T I O N 201

FIGURE 7-7 Center of curvature Oc of

coupler curve found from Euler-Savary
equation.

instantaneous center I and the inflection circle ,vere established, whence

JA was located. The radius OAA is seen to be directed down and to the
left, for both 0,111 and JAA must be laid off in the sa1ne direction.
An alternative forn1 of the Euler-Savary equation is given by
I I I
· - - (7-5)
IA JOA IJA
The earlier rule of sign must be observed.
'fhe practical application of the Euler-Savary equation allows
exploitation of the properties of the inflection circle. The inflection
circle for a n1oving system is found from the known 1notion of two coupler
Points, after which the radius of curvature of any other coupler point
rnay be found.
Example Find the center of curvature of the coupler curve traced
by the point C of the four-bar linkage sho\vn in Fig 7-7.
The steps are as followsa:
1. Locate the instantaneous center I fron1 the intersection of OAA
and OBB. This is one point of the inflection circle.
2. OA is the center of curvature of the path of the coupler point A.
A second point of the inflection circle, JA, which ,vill lie on the line OAA
(or its extension ) , is established fro1n the Euler-Savary equation,a·
(IA) 2
J AA = 0AA
= 5.3 1n.
. 2 02 K I N E M A T I C SYNTHESIS OF LIN KAGES

3. Similarly, J B, a third point of the inflection circle, n1ay be found,

(IB) 2
JBB = 0BB
= 4.8 1n.

4. With three points of the inflection circle known, the circle itself
may be drawn; its center G is the intersection of the perpendicular bisec­
tors of IJA and IJB•
5. The line IC is extended to locate Jc on the inflection circle.
The center of curvature Oc lies on the line IC; its position is determined
from a third application of the Euler-Savary equation, viz.,

(If) 2
O cC = .fcC = 2.3 in.

it being borne in n1ind that OcC and JcC must have the same sense.

Geometric Construction for the Points on the Path Normal n

The Euler-Savary equation involves the four points I, �4, JA, and VA
lying on the path nonnal n. Knowledge of any three allows the calcula­
tion of the fourth fron1 the equation. However, the fourth point may be
found fro1n a geometric construction, with the avoidance of 1neasuring
and con1putation. The n1ethod is based on ,vriting the Euler-Savary in
the form

it being recognized that IA is the rnean proportional between OA A

and J AA,
Suppose that the points I, A, and .J A are given (Fig. 7-8). Dra,v
an arbitrary line Au (other than AJAl) through A and another arbitrary
Jine Iv (also different fro1n AJAI) through I; these lines intersect at L.
Draw through J A a parallel to IL intersecting Au at K, and draw through
L a parallel to IK; this last line intersects A.J AI at OA·
This construction may be justified by considering the sin1ilar tri­
angles AJAK and AIL, from which

J AA KA
IA = LA-
and the similar triangles IAK and OA A L, from which

IA KA
OA A - LA
EULER-SAVARY EQUATION 203

FlGURE 7-8 Geometric construction of the

center of curvature O,.. when the points A, JA,
and I are known.

Combined, these relations yield

or

,vhich is identical to the Euler-Savary equation.

The reader will observe that this construction may start from any
three of the four points /, A , JA, OA . 'fhus, for example, ifa/, JA, and OA
are given, point A can be obtained by drawing successively OAL and /K
Parallel to one another, then IL and JAK parallel to one another, and
taking the intersection of LK with OAIJA to locate A.

Historical Note

Problems of curvature related to the develop1nents of the present section

have been studied by a number of mathen1aticians during the course of
the eighteenth and nineteenth centuries, but their works differ in many
respects. First of all, two related but distinct problems were considered:
( I) the path curvature, or curvature of the paths traced on the fixed plane
by points of the moving plane-this is the proble111 that we have studied
here; (2) the envelope curvature, or curvature of the envelopes on the fixed
Plane of curves in the 1noving plane, as in Prob. 7-7. The 1nethods either
Were purely geon1etric or used kine1natic considerations, as we have.
Finally, the degree of generality of the proble1ns considered varied, some
being restricted to only cycloidal motion with its circular centrodes.
204 K I N E M A T I C S Y N T H ESIS OF LINKAGES

The earliest effort seems to have been that of L'Hospital (1696),

who discussed the problems of path as \\'ell as envelope curvature and
derived the formula
I 1 1 1 1
( ) (7-6)
102 101 - lih Jf!1 sin 1"
(see notation in Prob. 7-7). His 1nethod was purely geo1netric and was
lin1ited to circular centrodes. The inflection circle was discussed by
De la Hire (1706), who also used a geometric-but different- approach.
The inflection circle was rediscovered in 1853 by Bresse, this time by
use of kinen1atic considerations. Bresse also gave a derivation of Eq.
(7-4) and discussed the laws of poili::,s having zero tangential acceleration
(see Prob. 7-7). Euler's contribution of 1765 is part of a n1emoir con­
cerned with gear-tooth profiles. He considered the problern of envelopes
in the case of circular centrodes (circular gears) ; his n1ethod, based on
calculus and geon1etry, may be extended to the general case. Reconsider­
ing the proble1n of tooth profiles, Savary, son1ewhere in the decade
1831-1841, rederived Eq. (7-6) for envelopes and gave a number of geo­
metric constructions applying to various situations.
We note finally that extensions of the Euler-Savary equation to
spherical and spatial n1otions have been developed by Garnier.

7-2 T H E C l,'. B I C O F S T A T I O N A R Y c r R V A T r R E
The path that a point of a 1noving plane 2 traces on a fixed plane 1
,vill generally have a changing curvature. In Fig. 7-9, the radius of
curvature of the path CA at point A is A0A = p. An instant later, when
the point has moved a distance �s along the path to A', the radius of
curvature is A '0� = p'. With �P = p' - p, the rate of change in radius
of curvature with respect to the displacen1ent s along the path is
dp
= lim �P
ds a-o �8

I
IA'

I
I
IP' = P+tlP
I
I
I
• 0'.4

J<'IOURE7-9 Curve with changing radius of

curvature.
EULER-SAVARY E Q U A T I O N 205

(T

-
"'
FIGURE 7-10 Notation used in de­
riving the cubic of stationary
curvature.

When this derivative is zero, the path CA is said to have a stationary

curvature. This does not 1nean that the curvature is constant, since
higher-order derivatives will, in general, be different fron1 zero, but it
means a higher order of contact 1- at least of the fourth order-between
CA and its circle of curvature, or osculating circle. At a given instant only
certain points of the n1oving plane have paths ,vith stationary curvature,
and their locus is found to be a cubic, the cubic of stationary curvature,
which will now be detern1ined.
Let the motion of plane 2 with respect to plane 1 be defined by the
fixed and moving centrodes 1r1 and 1r2 (Fig. 7-10). A.t the instant con­
sidered, the angular velocity of plane 2 is w and the JC velocity is v1 ;
together they yield the vector v1/w as shov,1n. Let k be the inflection
circle at the instant considered; its dian1eter is IK. Since I K = -iv1/w
(Sec. 7-1), this vector I K is always rotated 90° clockwise fron1 v1/w. The
center of curvature O.t of C.t is detern1ined by the Euler-Savary equation
as

valid in' sign as well as magnitude when an orient.ation is chosen on the

line IA.. With OAA = p, IA = r, IK = D (dia111eter of inflection circle,
always positive), ,ve find that

J AA = IA - IJA = r + D sin 1/;

1
Wben two curves meeting at a point also have a common tangent, the
measure of how close they lie together in the neighborhood of the point is called the
order of contact.
206 KINEMATIC SYNTHESIS O F LINKAGES

where i/t is the counterclockwise angle fron1 vr/w to the oriented line IA.
The radius of curvature may then be expressed as
r2
p - -·- · ·-·· - ····· (7-7)
r +
D sin if;
If u defines the position of the instantaneous center I along the n1oving
centrode -ir2, n1easured from an arbitrary origin P2, so that its value
increases with time, then
dp dp du
ds = du ds
Since d<T/ds is finite, except for the instantaneous center itself, dp/ds will
be zero if dp/du is zero, whence points of stationary curvature are charac­
terized by the relation
dp
du
=0
Differentiating Eq. (7-7) v,rith respect to <1 yields

dp
=
2r � (r + D sin i/t) - r2 (� + D cos if; <!Ji + � sin YI) (7-8)
du (r + D sin Y1)2
The derivatives ·dr/du and di/t/du may be evaluated by considering the
vector IA expressed in con1plex-nun1ber fonn with respect to a set of
axes 02x2, 02.112 moving with plane 2. Let ": = vr/lvrl be a unit vector in
the direction of vr; then
IA = r":eit
or 0� - 021 = r-ee;;,
Taking the derivative of this expression with respect to u yields
d(02A) d(02I) dr ..,. d": ..,. . ,J; ._, d
--- - - - =- 'ee•r + r - e'" + 1,T": - e•r
du du du du d<T
Since the axes 02X2Y 2 move with the plane 2 and A is a point of that plane,
d(02A)/d<1 = 0. Furthermore, d(0 21)/du = ":, and d-e/d<1 = i":/R2, where
R 2 is the radius of curvature of the moving centrode ; this radius is positive
or negative depending on the convexity of the centrode T2. It is negative
in Fig. 7-10. �,laking these substitutions in the above equation and divid­
ing by -e yields

- 1 = dr
- ( +-
ir
R2
d,J; i '
+ ir. - )e·· r

YI =
d<T du
.
or - cos i/1
.
+ i sin
dr
+ ir. ( I + dY1)
du R2 du
EULER-SAVARY EQUATION 207
Equating real and imaginary parts gives the desired derivatives,
dr = - cos i/t and di/t sin i/t 1
du du = r - R2
Equation (7-8) n1ay no,v be written as
dp 3r2D sin y,, cos YI 1 1 1 1 1 1 dD 1 1
( (7-9)
du = (r + D sin i/;) 2 [ sin i/t 3 (R 2 - D ) - cos y,, 3 du D) - r ]

and points satisfying the equation

1
1
M sin YI +
N cos i/t
_ !r = o (7-10)
in ,vhich

and

,vill be poitlts of stationary curvature.

In arriving at Eq. (7-10), however, we note that the terins preced­
ing the brackets of Eq. (7-9) have been ignored. In certain situations the
factor r2 sin YI cos i/t n1ay yield points of stationary curvature not recog­
nized by Eq. (7-10). Although Eq. (7-10) is commonly used to deter­
mine such points, the co1nplete equation to be regarded is

r,,. sin f cos

(
i/t l',f s�n VI + N c�s YI - �) = 0 (7- 1 1 )

On reducing Eq. (7-1 1 ) to cartesian coordinates by use of the rela­

tions cos 1/; = x/r and sin y; = y/r, a third- degree equation results,

(x z (y) -
X + N
+ yz) M xy = 0 (7-12)

Its curve is appropriately called the cubic of stationary curvature. We

1nay note that Eq. (7-10) will yield the san1e cubic equation, but only
after 1n ultiplication by x and y. Equation (7-12) may therefore define
Points of stationary curvature not predicted by Eq. (7-10). Thus, in
the exan1ple of Sec. 7-3, for which 1 /N = 0, the curve has t,vo branches;
one is a circle (second degree) , the other a straight line (first degree) : the
spirit of a cubic is therefore 1naintained even with particqlarization.
The straight branch satisfies Eqs. (7-1 1 ) and (7-12), but not Eq. (7-10).
Example Construct the cubic of stationary curvature of the
coupler of the four-bar linkage in Fig. 7-11 for the position shown.
The steps are as follows :
1 . Locate the instantaneous center I from the intersection of
o....A and 08B.
208 K I N E M A T I C SYN'f H ESIS O F L I N K A G E S

Centrode
at I j
tangent

,,.---+:
'
�.,••<> ••

it,
/
"1,..
stationary
Cubic of I
curvature

"1= 45°
� = - 3.22 in.
Asymptote

i-11 Cubic of stationary curvature of the coupler of a

four-bar linkage with respect to the frame. 0,-.A 1.28 in.,
FIGURE
=
OsB 2.54 in., o,., oR = 3.80 in., AB = 3.82 in., .,, = 90 .
= °

2. Deterinine the direction of the com1non tangent IT to the

centrodes (a) by construction of the inflection circle (to reduced scale if
necessary) and 90° rotation of its diameter IK, or (b) by construction of
the IC velocity vr by using the method of Sec. 4-7.
:\1ethod a is used in the present example. The inflection circle
passes through the instantaneous center I and two other points JA and Js,
which may be located along the lines O,.A and O8B as done in the example
E U L E R-SAVA R Y E Q U A T I O N 209

of the last section. The Euler-Savary equation gives

(IA) 2
(IB)
2
JAA = OAA
=
23.8 1n. JBB = OsB = - 4·0 i.n.
IJA = IA - JAA = - 18.27 in. IJs = IB - .lsB = 7.18 in.
Since these are large distances, a practical difficulty arises, which,
however, may be resolved by construction of the inflection circle to
reduced scale. To a scale of one-tenth the original, these distances are
IJ� = - 1.83 in. IJ; = 0.72 in.
When laid off from I (in the proper sense) on IA and IB, these distances
define the points J� and J�. The diarneter of the reduced inflection
circle is then I K', and the conunon tangent to the centrodes is J T,
rotated 90° from IK'.
The instantaneous center I and the centrode tangent IT are the
references with respect to which the cubic of stationary curvature is
defined in polar coordinates by Eq. (7-10).
3. Points A and B are on the cubic, since their paths are true
circles,
For A : 'YA = 150° TA = 5.53 in.
For B: 'YB = 191 °30 TB = 3.18 in.

The equation of the cubic must be satisfied for these values of i/, and T ;
this gives a set of two equations to determine the unknowns M and N,
1 1 1 1 1 1
M (0.5) N(
..,...._
.866)
_ -0- -
= 5.5:3 M (0-
��.20)
-
N(0:98)
= 3.18
From these relations
1 1
= -0.023 and
N
= -0.196
M
The equation of the cubic is then
0.023 0.196
- - = -
1
sin i/t cos if; T

4. Note that, for r - oo ,

0.023 0.196
sin i/t
= cos i/1
or tan it, = -0.117, whence i/t = - 6°40'. The cubic therefore has an
asy1nptote 1naking an angle of - 6°40' with the centrode tangent at /.
The construction is completed by making a table of values of r for values
of t/t between O and 180°. Values of I/; between 180 and 360° would 1nerely
repeat the points of the cubic already found by varying iJ, from O to 180°.
210 KINEMATIC SYNTHESIS OF LINKAGES

Table 7. J VALUES OF r AND VI FOR THE

CONSTRUCTION OF THE CUBIC OF
STATIONARY CURVATURE (FIG. 7-t t )
VI r "' r
oo 0 105° 1 . 36
15°
-3. 44 120° 2 . 73
30° - 3 . 69 135° 4 . 06
5°
- 3 . 22 150° 5 . 56
4
600 -2.39 165° 8 . 35
75° - 1 . 28 173°20' 00

900 0 180° 0

These values are given in Table 7-1, and the cubic itself is shown 1n
Fig. 7-11.
A nun1ber of points have been selected on the cubic of stationary
curvature, and the curves traced by these coupler points are dra,vn in
Fig. 7-12. These curves show what is 1neant by stationary curvaturea:
in all cases (except perhaps at the instantaneous center I, where the
coupler curve 1nust have a cusp) the intersection of the coupler curve with
the cubic is a point of maximum or mini1nu1n curvature. The coupler­
point curvature, however, so1netimes varies quite rapidly in the vicinity
of the stationary point; the designer must use the cubic of stationary
curvature with caution, i.e., must examine the excursions of the points
with care.
We may su1nn1arize the n1ost obvious characteristics of the cubic
of stationary curvature ,vith the aid of Figs. 7-11 and 7-12:
1. 'fhe curve is a property of the rnoving plane (e.g., the coupler)
but is drawn on the fixed plane and is unique to the phase (position) of
the linkage.
2. There is a crunode at l; the t,vo tangents are the centrode
nonnal and centrode tangent at I.
3. Coupler points "on the cubic," unless prejudiced by being close
to the fixed centrode, will describe approxin1ate circular arcs.

7-3 E X A M PLE : F O lJ R -B A R L I N K A G E T O R E P L A C E
C I R C U L A R G E A R S E G 1\-f E N T S FOR
S M A L L ROTATIONSe1
'fhe cubic of stationary curvature will be used in this section to
design a four-bar linkage to replace a pair of circular gear seg1nents (Fig.
1 Professor Hall worked this problern in a somewhat different manner; see
Inflection Circle and Polode Curvature, Trans. Fifth llfechanisms Conf., 1958; also ,
Linkage Design Technique, Machine Design, vol. a1, p. 44, 1959.
EULER-SAVARY E Q U A T I O N 211
7-13a). The gears have an angular velocity ratio of -i and operate over
a 30° range of rotation of the smaller gear. The procedure used here will
yield a linkage (Fig. 7-13b) for which the velocity ratio and its first two
derivatives are correct in one position.
Consider an inversion of the mechanisn1 of Fig. 7-13a in which
gear 1 is n1ade stationary ; the 1notion of gear 2 is then constrained by the
rolling contact between the pitch circles which constitute the fixed and
1noving centrodes 1r1 and 1r2 (Fig. 7-14). The centers of these centrodes
are, respectively, 0A and O B, and the inflection circle may be found by
application of the Euler-Savary equation,
(JOB) 2
KOB = · oA0--B = -0.8 in.

I
I
p - 1� <----........_ (::.:__r-:._:---
11
Cubic of stationary
curvature ,,,e-_ ___ \\
.._ �
Y
I//
II \\\ \ "\;-
L- �\
f r = - 3.22eln.
\i

f-1
1 :.
I
,,�1r--
\
\
A
\
\
\
\

FIGURE 7-12 Coupler curves traced by points on the cubic of

stationary curvature.
212 K l � E .\I.-\ T I (' 8 Y � T H E 8 l S O I? L I N K A G E 8

The diameter of the inflection circle is thus D = 1.2 in., and the vector
vr /w points down (Fig. 7-14).
The four-bar linkage OA,4BO B will be obtained by taking point B
on the cubic of stationary curvature of plane 2 with respect to plane 1.
Point A is then the center of curvature of the path of B. The equation of
the cubic of stationary curvature in polar coordinates ,vas found in the
last section as
1 1 _ 1
+ = 0
M sin y; N cos 1/1 r
I/; is 1neasured counterclockwise fro1n the vector vr/w, and

A� = ! (�2 - b)
Since the centrodes in the present case are circles, the diameter of
the inflection circle is constant and 1/ N = 0 : the cuhie of stationary
curvature thus reduces to a circle 1 of equation
r = M sin 1/1
The diameter 111 of this circle 1nay be found in tern1s of the known diame­
ter of the inflection circle, D 1 and the radius of curvature of the 1noving
1
\.Ye may observe that, with 1/J,,,: = }(dD/d<r)(1/D) = 0, Eq. (7- 1 1 ) shows
that all points of the line OAOB for which ,/, = 90 or 270° are points of stationary
curvature. \lire see that in the present example the cubic of stationary curvature
consists of the above circle and the line OAOB. Only tht> c>ircle, however, is of interest
in the design prohlE>m at hand.

---- 2 in.---i---- - 3 in. - ---

--- - ----- 5 in.----- --�
(a)
B

FIGURE 7-13 Four-bar

linkage to replace circu­
A lar gear segments.
E U L E R - S A V A R Y EQU A 'J' I O N 213
Inflection circle 1"1

----
Cubic of stationary
curvature (reduced
· to a circle)

7-14 Synthesis of a four-bar linkage O,.ABOB to replace a pair of

circular gear segments.
FIGCJRE

centrode, R2. The radius R 2, 1neasured in the direction of the unit norn1al
n, itself rotated 90° counterclockwise frorn v1/w, is R2 = - 2 in. 1 and
D = 1.2 in., \vhence

M = - 2.25 in.

The cubic of stationary curvature is the circle of dian1eter / L sho-wn in

Fig. 7-14. Point B 1nay be taken anywhere along this circle.
Choosing B so that the line IB 1nakes a 30° angle ·with OA Ou
as shownI
IB = - 2.25 sin 60° = - 1.9;j in.
Point A is then found by application of the Euler-Savary equation,

= (IB) 2 (IB) 2 3.8

AB
JB
= - 1.2 sin 60°
_
-0. 91
-4.18 in.

Which co1npletes the deterrriination of the linkage.

1�he calculated errors for 13° and - 17° fro1n the position used in
the design are in this case 0.4 and 0.9 percent of the total rotation.

BIBLIOGRAPHY

AUievi, L.: "Cinematica della biella piana," R. Tipographia Francesce Giannini

and Figli, Naples, 1895.
Bresse, C.: l\iemoire sur un theoreme nouveau concernant Jes 1nouvements plans,
J. ecole polytech. (Paris), vol. 20, 1853.
214 K I N E M A T I C SYNTHESIS OF LINKAGES

Euler, L . : Supplementum d e figura dentium rotarum (1765), Novi Comment.

Acad. Sci. Imp. Petropolitanae, vol. 1 1 , St. Pet-ersburg, 1767.
Garnier, Rene: "Cours de cinematiqne.'' 3d rev. eel., vol. II. G1n1t.hier-Villars,
Paris, 1956.
Hall, A. S., Jr.: "Kinematics and Linkage Design," Prentice-Hall, Inc ... Engle­
wood Cliffs, N.J., 1961.
Hire, P. de la: Traite des roulettes . . . , Jfem. math. phys .e. . . acad. roy.
sci. (Paris), 1706.
Hirschhorn, Jeremy: "Kinen1atics and Dynamics of Plane ::\I echanisnis,''
McGraw-Hill Book Company, New York, 1962.
L'Hospital, G. F. A.: "Analyse des infiniment petits . . . ," Paris, 1 696.
::VIueller, R.: "Einfuehrung in die theoretische Kinematik," Springer-Verlag
OHG, Berlin, 1932.
---: Papers on Geonietrical Theory of nfotion Applied to Approximate
Straight Line 11otion, translated by D. Tesar, Kansas State Univ. Bull.,
vol. 46, no. 6, .June, 1962.
Wolford, J. C.: An Analytical l\fethod for Locating the Burmester Points for
Five Infinitesimally Separated Positions of the Coupler Plane of a Four­
bar Mechanism, ASilfE J. Appl. ilfech., ser. E, vol. 27, no. 1 , lVIarch,
1960.
 GEOMETRIC METHODS
OF SYNTHESIS WITH
THREE ACCURACY POINTS

8-1 INTRODl:-CTION
The geon1etric 1nethods for the synthesis of planar linkages for
various duties antedate the n1ore precise algebraic attacks. They
still occupy a n1ost i1nportant place an1ong the available proce­
dures, for they are relatively fast in producing ans,vers and, since
they 1naintain touch with physical reality to a n1uch greater
degree than do the algebraic 1nethods, are often more readily
understood. .�lso, their degree of accuracy is adequate for many
situations and they are valuable adjuncts to the algebraic
1nethods, for their approxi1nate solutions serve as useful guides
in directing the course of equations or in reaching certain decisions.

The relatively long develop1nent period of the geo1netric 1nethods

has led to n1any techniques hand-tailored to particular types of
proble1ns. It is the intent of this chapter and the next to dwell
on the general and avoid the particular, i.e., to present general
concepts that 1nay be applied to a wide variety of problems and
abstain frorn peculiarly specialized techniques.
216 KINEMATIC SYNTHESIS Ol<' LINKAGES

The synthesis of linkages involves rnoving a link from a first

position to several others, son1eti1nes under specifications of velocity
and acceleration. A sequence of problen1s will be used to display basic
geometric concepts. The solutions of the first and si1npler problems of
this chapter will establish relations to be considered further in the next
and will be applied to n1ore con1plex situations.

8-2 POLES OF THE FOUR-BAR LINKAGE

Problem 1 Arrange for the transfer of a link AB fron1 position
A 1B 1 to a second position A2B2.
This proble1n may be solved in different ways, as shown in Fig.
8-1. For example, the 1nidnormals a12 and b u to the distances A 1,'1 2 and
B 1B 2 will intersect at P12. Links P12A1 and P12B1, connected as shown
in Fig. 8-la, will allow AB to assu1ne its two positions. This solution­
one may think of ,41P12B1 as a solid triangle pivoted about P1 r-is trivial
but recalls that all lines associated ,vith the plane of link AB undergo the
same rotation; i.e., the angles through ,vhich they turn are equal in n1agni­
tude and sense of rotation. The half angles .1 1 P12a12 and B1P12b 12 are
also equal.
The transfer of the link .4B 1nay also be carried out by a four-bar
linkage with AB as coupler. The centers of the fixed revolutes OA and Os
may be chosen any,vhere along a 2 and b 12, respectively, the other two
1

revolute centers being located at A and ·Bon the n1oving link (Fig. 8-lb).
Note that this solution to the problen1 involves two independent choices
with revolute centers at A and B on the moving link. For each center
0A chosen on a 2 there are an infinity of solutions corresponding to different
1

choices of OB along b 1 2. But since there are also an infinity of choices pos­
sible for OA along a12, it n1ay be said that the present proble1n has an
infinity to the square number of solutions, denoted as <:r> 2•
If either OA or On is chosen at infinity along its 1nidnor1nal (Fig.
g..1c and d), the corresponding four-bar linkages change into slider-crank
n1echanisn1s. The nun1ber of solutions for each situation is infinite because
of the infinite choices that are available for locating the fixed revolute
(OA or OB, as the case n1ay be).
If both OA and O8 are chosen at infinity along their 1nidnor111als
(Fig. 8-le), a PPRR n1echanis1n results and the solution is unique.
In the foregoing a link AB, that is, a portion of a plane ,vith ele­
ments (connections) at A and at B, ,vas caused to rnove fro1n one position
to another. If the specification is altered to n1oving a plane containing
the line AB frorn one position to another, the nurnber of solutions for
each of the previous situations is multiplied by <:r> 4. This addition co1nes
from having to choose two points C and D of the plane as locations for the
 217
POINTS
HREE ACCURACY
HODS-T
GEOME'l'RlC MET

(b) Solutions: a)
z

unique
(a) Solution:

B,.

(c) Solutions: a)
I (d) Solutions: 110

\o.. at 110

(e) Solution: unique

;/08 at a)
OA at co\

two positions of a. link.

FIGURE 8-1 Problem 1,
218 KINEMATIC SYNTHESIS OF LINKAGES

revolute elements; each point has two coordinates, whence four choices
must be made.

DEFINITION The point P12, the center for the finite rotation of AB
from A1 B1 to A2B!!, is the pole of the rotation 612-

The pole P12 is a property of the finite displacement of AB fro1n

A1B1 to A2B2 and is independent of how the link actually n1oves between
these two positions. However, if A2B2 is n1ade to approach A 1B 1, so
that A1B1 and A2B 2 becon1e two infinitesin1ally close positions, the pole
P12 then becornes the familiar instantaneous center of rotation of link AB
at the instant considered. Figure 8-lb depicts potentially useful angular
relations that should be noted, na1nely, A 1P 12A2 = B1P1 2B2 = 812 and
their half angles A1P12a12 = B1P12b12 = 612/2. This observation leads
to the following theorem for four-bar linkages:

rfHEOREM I When viewed from a pole of rotation, the coupler and

frame are seen under angles that either are equal or differ from each other
by 180a°
; this is true for the two positions defining the pole. Similarly,
the crank and follower are seen fron1 the pole under angles that either
are equal or differ from each other by 180° for the t,vo positions.

A situation in which the angles are equal is shown in Fig. 8-lb;

angles differing by 180° are shown in Fig. 8-2.
Problem 2 Design a four-bar linkage to transfer a link AB
through three specified positions A1B1, A2B 2, A3B 3 (Fig. 8-3).

I
/b12

A20--'---------o
B1

FIGURE 8-2 To Theorem I.

GEOMETRIC METHODS-THREE ACCURACY POINTS 219

FIGURE 8-3 Problem 2, three positions of a link.

The centers of the fixed revolutes OA and O s are uniquely defined

here as the intersections of a12, a23 and b12, b2a, respectively (Fig. 8-3).
If A and B are the revolute centers of the coupler, this problem has a
unique solution.
Problem 3 Design a four-bar linkage of frame OAO n, in which
a given position q, and a given angular velocity w2 of the crank produce a
specified position i/1 and a specified angular velocity W4 of the follower.
The centers of the fixed revolutes being OA and OB (Fig. 8-4),
the specified positions of the crank and follower center lines are defined
by angles q, and i/1, and the revolute centers of the coupler n1ust lie on
these centerlines. The velocity of A is perpendicular to OAA ,vith a

(8-1)
magnitude

Si1nilarly, the velocity of B is perpendicular to OsB with a magnitude

Vs = (OBB)w4 (8-2)
and the instantaneous center of rotation of the couple AB, at this instant,
is the intersection/ of OAA and O BB. But since I is the instantaneous
center of AB, AB rotates at this instant about I with an angular velocity
'->a; thus
anJ VB = (JB)wa

or
220

Substituting the values of and from Eqs. (8-1) and (8-2) into this
KINEMATIC SYNTHESIS OF LINKAGES

last equation yields

VA Vo

(OAA)w2
(OoB)w.1
IA
IB
or (8-3)
IB IA W4
OBB
The solution may be summarized as follo,vs:
OAA W2

1. Lay out the fra1ne and construct the instantaneous center I

corresponding to ct, and if,,.
2. Choose A arbitrarily along 011.I.
3. Evaluate the ratio lB/OoB from Eq. (8-3).
4. Construct point B along Ool so that it divides the segn1ent
0ol in the above ratio.

In n1oving a specified coupler fron1 one position to another, ,ve

8-3 RELATIVE POLES OF THE FOl:R-BAR LINKAGE

considered the coupler fro1n the vantage of a point called the pole P12,
detern1ined fro1n the 1nidnonnals of the two coupler positions (Fig. 8-lb).
This pole, conunon to fran1e and coupler, gave no relation between the
swing angles (or angular displacements) of the crank and follower. 1'he
crank (input) and follower displacen1ents <J, 2 and "1 12 for a given coupler
displacement are identified in Fig. 8-5a. To correlate <1>12 and "112 through
the coupler niotion, we shall consider the follo,ver 1notion with respect
1

to the crank. We do this by 1neans of a kinematic inversion: ,ve shall

t'IGURE 8-4 Problen1 3, spec1-

fieation of angular velocities
i,nd positions of links 2 and 4.
 G E O M E T R I C M E T H O D S-T H R E E A C C U R A C Y P O I N TS 221

Oa Oa
(a) (b)

FIGURE 8-5 Four-bil-r linkage displacernents needing correlati0n.

assume the crank fixed, thus becon1ing observers on it, and shall ren1en1ber
that in any displacen1ent of a linkage the relative ,notions of all links
remain the san1e, regardless of which link is fixed.
The linkage with the crank OA A i fixed, in a position ready for
displacen1ent, is sho,vn in Fig. 8-5b. On a linkage displacement consistent
with that of Fig. 8-5a, the follower displacen1ent ,vith respect to the crank
(Fig. 8-6) is seen to be the result of two separate rotations - c/> 1 2 and i/; 12-
It is our purpose to combine these two follower rotations into a single
equivalent rotation and find the unique point, called relative pole 1 R1 2,
about which the single rotation takes place.
The displacernent of the follower fro1n O BB 1 to O�B; is the result
of two rotations (Fig. 8-6)a:
1 . A rotation about OA of angle � c/>12, from OBB 1 to O�B�. Note
the negative signa: if OAA i rotates clock,vise ,vith respect to OA OB, then
OAOB rotates counterclock,vise ,vith respect to OAA 1.
2. A rotation around On of angle i/112 fro1n O�B; to O�B�.
The angle of rotation from OsB i to O�B; is thus i/112 - c/>1 2. The
relative pole R1 2 is the intersection of the rnidnonnals b�2 of B 1 B; and c�2
1 A second relative pole exists for a displacement of the crank with respect to
the follower. \Ve shall not consider it here, for we shall not neerl it. Its construction
is similar to that of the pole which we are discussing.

,?7R 12
Bi
L
//(relative pole)

/
z

O.t Oa
FIGURE 8-6 Definition of the relative ·pole.
222

of 080�, and the triangle OAO»R12 yields a + /3 = -· <J,12/2. Since

K I N E M A T I C S Y N T H E S I S O F L I N K A G ES

fJ = (t/1 1 2 - </,1 2)/2, the half angle of rotation from O»B1 to O�B�, we may
write a = - </,12/2 - (t/11 2 - </,1 2)/2 = -t/11 2/2 and R12 is the intersec­
tion of the lines c;2 and z, 1naking, respectively, the ang;les - </,1 2/2 and
-t/;1 2 /2 ,vith OA OB,
Theore1n I also applies to the inverted n1echanis1n and its relative
pole R 1 2 : viewed fro1n R12, OAA I and OeB1 appear under equal angles,
and OAOn and Ao1B1 are also seen under equal angles.

<P12 and "1

Select convenient fran1e points 01.. and On, and dra,v two lines 1naking the
Construction of the Relative Pole for Specified 12

angles - q,1 2/2 and -t/;12/2, respectively, with OAO/J. Their intersection
is R 1 2• Xote again the negative signs: if t/,12 is clockwise. then - q,1 2/2 is
counterclockwise from OAOB• Similarly, if t/;12 is clockwise, then - t/; 1 2/2
is counterclockwise.
Problem 4 Design a four-bar linkage in which a given angular
displacen1ent t/,12 of the crank produces a given angular displacement i/tu
of the follower (Fig. 8-7).

1 . Assume a convenient frame OAO», and construct the relative

Solution

pole R1 2 corresponding to <1>12 and 1/,-12.

2. Choose point A 1 arbitrarily.
3. Draw a line R1 21t such that
in magnitude and direction
4. Choose any point B I on R12u.
By Theore1n I, the desired linkage is OA A 1B10B. The arbitrary choices­
A 1 and B1-give the problem oo 3 solutions. The rnethod fails if
</,u = 0, or 1/112 = 0, or t/,12 = 1/112.

follower

FIGURE 8-7 Solution o f

Prob. 4.
G E O M E T R I C M E T H O D S-'l' H R E E A C C U R A C Y P O I NT S 223

Oc

(a) (b)
F'IGURE 8-8 :\lodification of timing of cam mechanism.

The foregoing 1nethod nuiy be applied to modifying the follower

dwell of an existing cam n1echanism ,vithout changing ca1ns. Assume
that the cam follower (Fig. 8-8a) has a d,vell for 60° of camshaft rotation
(input). It is desired to have the follower at dwell for 90° of input rota­
tion. The solution involves slowing the present cam by making it the
output link 08B of a four-bar linkage (Fig. 8-8b), proportioning the link­
age such that the 60° of constant cam radius corresponds to 90° of input
rotation, now applied to the crank OAA. 1'he cam link 0 8B must be
able to rotate continuously; this condition den1ands that the four-bar be
a drag-link (double-crank) configuration. Accordingly the Grashof rule
l + s < p + q must be observed, with the frame OAOs serving as the
shortest link.
The detennination of this linkage is shown in Fig. 8-9. After
constructing the relative pole R 12, point A 1 is chosen so that OA.41 > 0.AOs;
Ru-u is constructed as in step 3, Prob. 4, and B1 is chosen along R 1 2u
such that
O.t A 1 + 0.AOB < A1B1 + OnB1
and A 1B 1 < OAAo1
The desired linkage is OA A 1 B10B, also sho,vn in Fig. 8-9.

= 90•

\/
c/>12

0 'd
OA Os

nouaE 8-9 !Vlodification of tiruing of ctuu mechanil:!m; construction of linkage.

224 KINE M A T I C S Y N T H ESIS OF LINKAGES

----- 1/,-13

✓---�
1/,-1-• i
, ,/,
¥'0 3
I/ I
I
-\
I I
----0---L
-1.
Os

FIGURE 8-10 Solution of Prob. 5.

Problem 5 Design a four-bar linkage in ,vhich two successive

clockwise angular displacen1ents 4>12 and 4>2a of the crank produce, respec­
tively, two successive clockwise angular displace1nents i/112 and i/;23 of the
follower (Fig. 8-10).
Solution Let (/,13 = '1>12 + q,23 and yt13 = t/t12 + t/t2a.
1. Assume a convenient frame OA Os, and construct relative poles
R12 and R13.
2. Choose Ao1 arbitrarily.
3. Draw the lines R12u and R 1av such that
= A1R12u = OA.R120s
= A 1R13V = OAR130B
/32

4. The intersection of R1 2u and R13v is B1 .

/33

The linkage, chosen from an1ong the available oo 2 solutions (choice

of point .41), is a double-rocker 1nechanism. �'he choice of A 1, ,vhile good
for displaying the construction, was unfortunate fro1n an operational
standpoint. An exa1ninatio11 of the linkage, best afforded by a n1odel,
will uncover three shortco1nings: (1) starting fron1 position OA·'' 1 , the
input link n1ust first be turned counterclock,vise, and then n1ust have its
direction of rotation reversed; (:2) the linkage ,vill go through a dead
point ,vhen fully extended to the lefto; (:3) it ,vill be necessary to disconnect
the linkage ,vhen it is fully extended to the right, 1 nove the links, and
reconnect the1n before going to the final position. 1'hat the linkage does
meet its specifications in a formal manner is small consolation for the
roundabout way in which this is accornplished. The proble1n of hov,r to
remedy this situation will be enlarged upon in Sec. 8--5.
G E OM E T R I C METHODs�·THREE A C C UR AC Y POSI N T S 225

8-4 EX A M PL E i: LOG A R I T H l\i l C - F U N CTIOK GEN E R A TOR

If the output and input variables of a linkage are proportionally
related to the variables of a speczfi('.d function such as
y = .f(x) or z = g(x, y)
the linkage is called a function generator. The linkage for z = g(x, y)
,vill obviously require t,vo inputs, one for each of the independent varia­
bles x and y. In ·what follows, ,ve shall not consider such double-input
function generatorsa: our attention ,vill be directed to the simpler situation
represented by y = f(x) requiring only a single input.
The principle of a single-input function generator of the four-bar
type is shown in Fig. 8-11. The independent variable x is to be repre­
sented n1echanically by the rotation <J, of the crank OAA, or input, with
the follower O BB rotation YI displaying the dependent variable y. The
discrete relations between x and <J,, y and YI are usually made linear, but
they need not be.
A.s an exan1ple, we shall design a four-bar linkage to generate the
function y = log x in the interval 1 < x < 2. The independent variable
ranges from x = 1 = x. to x = 2 = x1, or ,ve n1ay say that the range
Ax = x1 - X in general terms. The range of motion or angular sweep
8

of the x pointer (link OAA) corresponding to ax will be designated

Aq, = tJ,1 - q,,; this range is arbitrary and ,vill be taken as 60a°
counter­
clockwise. On the assun1ption of linear relationships, any value of x
within the interval of generation is related to its q, value by
"' - "'• = -
a"' = r q, - q,, - X - X a

X - X, ax "
or
aq, ax
The dependent variable ranges bet,veen y. = log x, = 0 and

%•Scale, linear y-scale, linear

y = f(X)
�
Y,

q>I

OA
Crank: input
Os Follower: output

FIGURE 8-11 Principle of four-bar linkage function generator y = j(x).

226

y1 = log x1 = 0.3010, and the range is ll.y = y, - y�. The corresponding

KINEMATIC SYNTHESIS O F LINKAGES

range of motion of the y pointer is /l.,J; = ,J,,1 - ,J;,; the actual value of t:,.i/1
is arbitrary and is chosen here as 90° counterclockwise. The variables
y and 1" are related within the interval of generation through the linear
relation
i/1· ····-
- ,;,,• - a,;,,
- -r or
y - y.
= ·Y
·-
- Ys
-
Ay II Ay

In this exa,nple, the four-bar linkage will be designed to give an

exact value of the logarithn1ic function at only three points of the y-x
curve, corresponding to three values of x at the accuracy points. Let
xi, x2, x3 be the accuracy points; tb.ey are chosen with Chebyshev spacing
in the interval 1 < x < 2, that is (Fig. 8-12),
1.5 - 0.5 cos 30° - 1 .067
1.5
Xi -
1.5 + 0.5 cos 30°
Xs =
X:i - - 1 .933

The corresponding values of y are

Y• log 1.067 = 0.0282
Y2 log 1.5 = 0. 1761
=
y3 log 1.933 = 0.2862
=
=
The change in x from the first to the second accuracy point is
X12 = X2 - Xi = 1.,5 - 1.067 = 0.433
and from the first to the third accuracy point
Xu = Xa - X1 = 1.933 - 1.067 = 0.866

· ' ,·
1.2 1.4 1.6 1.8 2.0

FIGuRi,;8-1:.l Three accuracy points with Chebyshev spac­

ing in the interval 1 < x < 2.
GEOS
M E T R I C M E T H O SDS- T H R E E A C C U R A C Y PO INTS 227
y

- "1t
�� -
Yt = 0.301 -,-------.....-----r-::ao
1/123
1/13

t
33• 0.20
1/1 AV' = Y2

90• AY =
Vl12= 2
44• 0.301 0.10

"'1 .,,
Y1
L._-.:.
lL-Jl--------.J.::..5______._--=.0
l;.;: 2.;.;;
.,,, ..,__---+-y, = 0 --=-=--
} = X, X1 X2 X3 "t : 2.0
t1- t, = s.s·
+-- - --Ax = x1 - x, = 1
---- A</, = </, - <J,,
1
= 60°-- -,M

FIGURE 8-1� Function JI = log x, I <x < Z to be generated.

The corresponding changes in tf, are therefore

.C2 - Xi
- A t/, 0.433
t/,2 - t/,1 l 60 26
0

'Pl3 = tp3 - 'P l = 52

¢12 = = = =
Ax
°
Similarly, the change in y fro1n the first to the second accuracy point is
y � - Yi = 0.1761 - 0.0282 = 0.1479
and from the first to the third accuracy point
Ya - Y• = 0.2862 - 0.0282 = 0.2580
The corresponding changes in y; are

¥112 = i/12 - i/11 = Y2 ti- Y1 Ay; = 0.1

_
479 0
0 3010 9 = 44
0

°
y
"11a = I/ta - i/11 = 77
All pertinent data are asse1nbled in Fig. 8-13.
The problern is nov,' reduced to the ter1ns of Prob. 5, Sec. 8-3, and
the solution proceeds as shown in Fig. 8-14 . A frame length OA 08 of
4 in. was chosen and the relative poles R12 and R13 constructed. An
angle "11 = 90t° and a crank length OAA = 3 in. were selected, giving the
point A 1 as shown. 1'he lines Ruu and Ruv were drawn to give B1 at their
228 KINEMATIC S Y N T H E S I S O F L I N K A G ES

intersection. The desired linkage is OAA1B 10B in the first position.

Measurements from Fig. 8-14 yield
OnB1 = 1.68 in. OAOB = 4 in. A 1B1 = 5.8 in.
"11 = 32
°

From Table 8-1,

</,1 - "'· + 4 = 90
°
or <I>. = 86°
'fl = If• + 8.5 = 32 if;. = 23.5°
°
or
The linkage was next redrawn and fitted with </, and if; scales
(Fig. 8-1.5 ). A cursory exan1ination sho,vs that the follower rotations
1/112 and 1/123 will be produced as the input rotations </,12 and </,2a are imposed.
Consideration of the link lengthi:, shows the device to be a double-rocker
linkage.
The perfonnance of a linkage is gauged by how accurately it gener­
ates the specified function (Table 8-1). In Fig. 8-1.5, 1 1 values of <J, were
chosen at 6° intervals ; the corresponding values of if; were constructed
and converted to Ymeeh• The error is the difference between y = log x
and Ymech• The n1axin1un1 value of the error in y ,vas found to have a

3 in.

4>1 == 90•

--<b13

FIGURE 8-14 Synthesis of function generator y = log x, l < x < 2, three accuracy
points.
G E O M E T R I C M E T H O DS-T H R E E A C C U R A C Y P O I N TS 229

y = log x, I < x < 2, THREE ACCURACY . POINTS

Tobie 8-1 TABULATION OF VARIABLES INVOLVED IN FUNCTION GENERATOR

VARIABLES

X 11 = log x ,J,, DEO ,/,, DEO

INITIAL VALUE x, = 1 Y• = 0 .,,, "'·

YI = 0.3010 ,P/

= 60° = 90°
X/ = 2

= 1 ily = YI - 11•
nN.u, VALUE Y,/

RANGE OF
Ax = xi - x, = 0.3010 A</> = ,pf - <f,, A,J, = Y,/ - Y'•
VARIATION

= = Y'• + 8.5
+ (x1 - x,)r, ,J,1 = Y'• + (111 - 11,)r
FIRST
<1>1 = </>, v
ACCURACY
POINT
XI 1.067 YI = 0.0282
= .,,, + 40 °

= = ,J,, + 52.5°
8ECOND
ACCURACY X2 1.50 YZ = 0.1761 <In = .,,. + 30 °
Y,2
POINT

= = 0.2862
THJRD
ACCURACY x, I.933 II• .,,, = .,,. + 56 °
Y,2 "' Y,• + 85.5 °

POINT

.,,,. = .,,. - = 26° = 44o

VARIATION
FROM FIR8T X: - XI = 0.433 1/2 - t/1 = 0.1479 <f,1 Y,U = Y,2 - ,f,t
TO SECOND

= 52°
VARIATION
F'ROM FIRST XJ - ZI "' 0.866 YI - 1/1 = 0.2580 <l>u = <I>• - <1>1 Y,13 = Y,, - ,f,1 = 77
°

TO THlRO

n1agnitude of 0.003, which corresponds to 0.003/0.30 = 0.01, or I percent

of the total variation in y.
In the design of this function generator, as in n1ost problen1s of
kinematic synthesis, errors of three types may be present, viz., graphical,
mechanical, and structural.
Graphical error con1es fro1n the inevitable inaccuracies of drafting
(Figs. 8-14 and 8-15). It is the result of the accu1nulation of sn1all errors
rnade in laying out angles, draw·ing perpendicular or parallel lines, or
taking the intersection of t\\·o lines. The graphical error is spread at
random over the whole range of operation of the 1nechanisn1 and does not
vanish at the accuracy points. It is a function of the designer's skill and
judgment and the scale to which the graphical construction is carried out.
Mechanical error results fro1n imperfect 1nachining of the parts,
play in the joints, and deforn1ations due to loading. The evaluation of
230 KINEMATIC S Y N T H ESIS O F LINKAGES

.. t
q, - 4'a (deg)

12 6 0

1)3 76 69 60.5 52
�

(p l = 90• (.x J = 1.067)

FIGt:RE8-15 Angular relations in function generator y = log x, 1 < x <. 2. The

mechanism is shown at the first accuracy point.

the mechanical error in terrns of the tolerance under which a mechanisn1

is built is considered in Chap. 10.
Structural error, as not�d in Sec. 5-5, is the difference between the
mechanically developed function Ymech and the desired function y = .f(x),
with no other errors present. The structural error is easily recognized,
for it vanishes at the accuracy points. It n1ay be decreased by n1odifying
the distribution of accuracy points or increasing their nun1ber.
The error appearing in Table 8-2 is the result of both structural

Table 8-2 ERROR IN LOG-FUNCTION GENERATOR, THREE ACCURACY POINTS

X "' - q,., i/1 - if,, log x Ymech Ymech - log X

DEG DEG

1.0 0 -1 0 -0. 003 - 0 . 003

1.1 6 12.5 0 . 041 +0.042 0.001
0 . 079
I .3
1.2 12 0 . 079 0
34
23.5
18 0. 114 0 . 1 14 0
1.4 24 43 0 . 146 0 . 144 - 0 . 002
1.5 30 52 0. 176 0. 178 0 . 002
1.6 36 60 . 5 0 . 204 0. 202 -0.002
] .7 42 0. 230 0 . 230 0
1.8
69
48 76 0. 255 0 . 254 -0.001
1 .9 54 83 0 . 279 0.278 -0.001
2.0 60 90 0 . 301 0.301 0
G E O M E T RIC M E T H O DS-THR E E A C CURACY P O I N T S 231
and graphical errors, and since its distribution hears no relation to the
accuracy points, it may be assigned mostly to graphical error. The use
of a larger number of accuracy points (as shown in the following chapter)
would therefore be of little use in i111proving the accuracy of the present
linkage. In other cases, ho,vever, a larger structural error is present, and
the use of a greater number of accuracy points is then justified. A case
of this type will be considered in the next section.
The satisfactory performance of the linkage may be ascribed to
several factors, including luck. For one thing, the function possessed
no violent changes within the interval of generation. Next, the several

frame length OAOs, and crank length 0.1A 1-happened to produce a

arbitrary choices-ranges of 1notion of crank and follower, starting angles,

n1echanisn1 with well-proportioned links and good transmission angle

throughout the desired interval in x. Just how to recognize, and hence
avoid, an unfortunate choice cannot be reduced to definite rules. When
circumstances combine to produce a poor linkage, the only rr.1nedy is to
base a redesign on different arbitrary choices. This is done in the example
of the follo,ving section.

8-5 EXAM P LE : S I N U S O I iD AL-FUNCTION GEN ERATOR

The problem considered here is the design of a four-bar linkage to
generate the function y = sin x in the interval 0 < x < 90°. Three
accuracy points are taken in this interval with Chebyshev spacing, viz.,
X1= 45 - 45(0.866) = 6°
X2 = 45
°

X3 = 45 + 45(0.866) = 84
°

The corresponding values of y are

Yi = sin 6a° = 0.1045

Y2 = sin 45° = 0.7071
!J:i = sin 84a = 0.9945
°

The ranges of variation of q, and it, are chosen as

Aef, = °
60a both counterclockwise

With a linear relationship between x and </, and y and t we have

= x2 - Xi
Cup = 26° 'P 3 = Xa - X1 Aq, = 520
</,12
Ax
l
ax
"112 = Y2 - Y1 t,.iy = 54c ¥'ta = Ya - Yi ti.it, = goo
t,. y t,.y
232 K I N E M A T I C S Y N T H E S I S OF L I N K A G E S

--
<Pia
2
--
<P12
--1/112
2

FIGL'RE 8-16 Construction for function generator y = sin x, 0 < x < 90°, three
accuracy points, first trial.

The linkage design follows the familiar pattern. Starting frorn

two arbitrary base points OA and 08 , the relative poles R 1 2 and Ria are
constructed, and a point Ae1 is chosen arbitrarily, with B 1 given by the
intersection of R12u and R13v (Fig. 8-16). The corresponding linkage, in
position 1, is redrawn in Fig. 8-17 as OAA 1 B;Os. The point B�' is another
possible position of Be1 corresponding to the same link lengths and input
angle </>, but a different output angle if;. The two points B� and B�' cor­
respond to tv,o different ways in which the four-bar linkage OA ABOs
may be closed ; the second linkage is OAA 1B�' 0 n. Both linkages are
double-rocker types.
The linkage OAA 1 B�OB must be checked for proper correlation of
crank and follower for the known angular increments. This n1ay be done
on the board or, better yet, with the aid of a cardboard model. To check
on the board, the positions Ae2 and Ae3 are constructed such that
and
and the points B; and B� are located. On con1paring the angles B�OBB;
and B�O BB� with ¥1 1 2 and if; 13, it is found that, ,vhile B�OBB; = i/,,12,
B�OBB� � ¥1 1 �- If the second linkage OAA 1B;'0B is investigated by
entering the successive positions of B�', nainely, B�' and B;', on the draw­
ing, it is found that B�08B;' = i/;1 3• 1'his circun1stance is due to the
fact that one of the accuracy points, the third, is located on a branch of
the output-input curve different fron1 that containing the first two �ccu­
racy points. The first two accuracy points are con1patible with the first
G E O M E TR I C M E T H O D S - T H R E E A C C U R A C Y P O INTS 233
linkage, the third with the second linkage. In going from the first branch
to the second, the linkage OAA1B�Os, ,vhich started in the open position,
assumed a crossed position before opening out again. This method of
synthesis does not differentiate between the two branches of the output­
input curve. No systematic way is known to the authors for avoiding
the trouble that was encountered. The linkage that was found, however,
was not unique, but prejudiced by the several arbitrary choices ; a change
in the ranges of the variation of the angles q, and y; is perhaps the first
remedy to apply.
The most satisfactory way to verify a linkage design is to make a
cardboard model and observe it. The form of the output-input curve of
linkage OA A 1 B�08 (Fig. 8-18) ,vas quickly established with the help of
a model. The three accuracy points are indicated by the circles at (1),
(2), and (3); the little figures indicate successive phases, the crossed posi­
tions, dead points, and transmission angles being quite evident.
In summary, the foregoing design is impossible, and a new atten1pt
must be made.
In planning a design, a consideration of the function to be gener­
ated often gives helpful hints as to what type of solution should be
attempted. Since the function y = sin x has a n1aximum for x = 90e°,

B"3

FIGURE 8-17 Function generat-0r, y = sin x, 0 $ x $ 90° , three

accuracy points, first trial.
 l/1", output
234 KINEMATIC SYNTHESIS O F LINKAGES

120

q,•, input
-80 -40
80

\
0,4. <iY
nrm;;,r,m?t:\

-80

FIGCRE 8-18 Output-input relation for function generator y = sin x,

0 <x< 90 , first attempt. Accuracy points at (l ), (2), (3).
°

it is evident here that a four-bar linkage of the crank-rocker type will be

better suited as a solution than a double rocker, since it has a maximu1n
of YI for certain values of q,. In a crank-and-rocker n1echanism, the crank
length OAA is sn1aller than the frarne length OAOB. This condition n1ay
therefore be used as a guide to choose the point A 1 in the synthesis of the
linkage. A rather large variation of q,, with a s1naller variation of i/t,
seen1s advisable, for the angle of swing of the rocker is never very large in
a crank-rocker n1echanis1n having good force-trans1nission qualities.
G E O M E T R I C M E T H O D S - T H R E E A C C U R A CY P O I N T S 235

FIGURE 8 - 1 9 Construction for function generator, Y = sin .r, 0 < x < HOC, three
accuracy points, second attempt.

rfhe ne,v design ,vill assume A<J, = 120o°, Ai/I - 60o°, fron1 ,vhich
</>12 = 52o°° </>1a= 104o°
= 36o V'13 = 53 .5
°
1"12
The new construction is shown in Fig. 8-19; upon taking .4 1 as shown,
with OAA 1 < OAOe, the resulting linkage is then OAA1B10B, redrawn in
Fig. 8-20. A rough check of this linkage shows that all three accuracy

1/; - 1/1, (deg)

q, - tp, (deg)

72

84

96

Jo'Jot·RE8-20 Function generator y = sin x, 0 < z < 90°, with

three accuracy points, final linkage.
236

= sin x, 0 < x < 90°,

KI N E M AT I C S Y N T H ESIS O F LINKAGES

Table 8-3 ERROR IN FUNCTION GENERATOR, y

THREE ACCURACY POINTS

x, <f, - <J,., YI - y;,, S10 X Ymeeh Ymech - Sin X

DEG DEG DEG

0 0 1 0 0 . 017 0.017
9 12 9 0. 156 0 . 150 -0.006

27
18 24 17.5 0. 309 0. 292 -0.017
36 26 0. 454 0 . 434 - 0 . 020
36 48 33 . 5 0 . 588 0 . 558 - 0 . 030
45 60 43 0. 707 0 . 717 0 . 010
54 72 49.5 0. 809 0 . 825 0 . 016
63 84 55 0 . 891 0.917 0 . 026
72 96 58 0 . 951 0 . 967 0.016
81 108 ,59 . 5 0 . 988 0 . 992 0 . 004
90 120 59 .5 1 . 000 0 . 992 - 0 . 008

points are on the sa1ne branch of the input-output function, and an analye­
sis sin1ilar to that of Sec. 8-4 yields the results shown in Table 8-3.
The error Ymech - sin x follo,vs in the present case a n1ore regular
pattern than in the case of the logarith1nic function. An error curve
would show that the error, vanishing at the accuracy points, 1nay 1nostly
be attributed to structural error. This indicates that the accuracy of
the present solution has not reached the li1nit i1nposed by graphical error,
as was the case for the logarith1nic function of the last exa1nple, but that
the addition of a fourth accuracy point would further reduce the error.
Such a synthesis ,vill be considered in Sec. 9-4.

8-6 POLES O F T H E S L I DER-CRAN K M EC H AK IS h-1

Problem 1 will be reconsidered here, with particular attention
given to the slider-crank solutions, Fig. 8-lc and d reappearing again as
Figs. 8-21 and 8-22. In these solutions, P 1 2 is a pole of the coupler A B
of a slider-crank mechanis1n. The ideas of Theoren1 I, relating to the
pole angles of a four-bar linkage, ren1ain true ,vhen one of the fixed revolute
centers goes to infinity. This theore1n may therefore be extended to
apply to a slider-crank n1echanism, with the understanding that one of the
fixed revolute centers is now at infinity in a direction perpendicular to
the translation of the slider.
In Fig. 8-21a, the revolute center OB at infinity along the midnorn1al
b12 is suggested by an arrow at the lower end of the n1idnormal be 12• The
 G E O M E T R I C M E T H ODS-T HR E E A C C UR A C Y POINTS 237

frame angle a is shown in Fig. 8-2la; it is related to the coupler angle by

(8-4)
However, the choice of Os at the lower end of b12 was arbitrary, since the
prismatic pair could have been obtained equally well from a revolute pair
having its center at infinity at the upper end of b12, as shown in Fig. 8-21b.
The frame angle {3 is directly equal to the coupler angle,
(8-5)
It may be noted that both Eqs. (8-4) and (8-5) are consistent with Theo­
rem I. Sirnilarly, the crank and follower angles of Fig. 8-2la are related by
A 1 P12 0,1 = B1P120s + 180°
For Fig. 8-21b, the crank and follo,ver angles are directly equal,
A 1 P1 20A = B1P1 20B
The application of Theore1n I to the second slider-crank solution
of Prob. l is shown in Fig. 8-22, where OA n1ay be chosen at either end of
the midnorn1al a12.

8-7 R E L A T I V E P O L E S OF THE S L I D E R-C RANK :M E CH A N IS M

On being vie,ved fro1n the crank, the follower 1notion of a four-bar
linkage ,vas seen to be a consequence of t,vo separate rotations, the crank

/
/

(a}

FIGt.:RE 8-21 Extension of Theorem I to the slider-crank mechanism, 08 at infinity.

238 KINE M A T I C S Y N T HES I S 0 1'' LI N K A GES

""\l °' atm

Bi B,

;•..
0 12

/"
l

A1

L
I
z
A2 ;

Os

a,
\

A 10, at m

FIGl"RE 8-22 Extension of Theorem I to the slider-crank n1echanism, 0A at infinity.

7r
,_r �

J!
\

I
.I .
i .V

· -·-;·>. i :
! I

'· 8n

j
;__ _ �
bu
··- 1·
.....S. 8
�-
\
�·
·.
/·/! 8�3

'' ____2_,
\

I
I :
-- r..
/ • I

B i '--- -.
t ------ I i
---I
. .._ --......_

Ba·...

FIGURE 8-23 Composition

of a translation T 2 and a
__ _ _h_
1

rotation 823 into a single

0A L rotation 8 13 = 8 23 about the
relative pole Ru.
G E O M E T R I C METHODS-THREE A C C URACY P O I N T S 239
¢12 and the follower t/112 (Sec. 8-3). These two rotations were replaced
by a single rotation (equal to the net angle of the two parts) about a
unique point, the relative pole R12. In this manner it was possible to
correlate the crank and follower rotations.
The follower of a slider-crank 1nechanis111 is the slider whose
motion is translation. Again there is the need to correlate crank and
follower motions; in the slider crank the two separate n1otions are the
crank rotation q, 12 and the slider translation S12- The con1position of
the rotation and the translation leads to a single rotation of the same mag­
nitude as its part, referred to a unique point again called a relative pole.
This composition, and the determination of the relative pole, will be con­
sidered first in general terms.
Consider that the moving plane II is sliding with respect to the
fixed plane (the paper) and that it assun1es three positions II1, II2, Ila
(Fig. 8-23 ) . The displacement fro1n II1 to II2 is a translation defined by
the vector T 1 2• The displacement from II2 to Ila is a rotation 023 about a
pole OA in the n1oving plane; notice that the center of rotation OA trans­
lates. l\1idnormals to both 0� 0.11. = T12 and B1Ba intersect at a pole R1 3•
The plane II can be revolved from position II 1 to Ila by a single rotation 813
about Rt1 3 . It is evident fron1 the figure that the angles marked 81 3 are
not only equal but also equal to 82a ; that is, the single angle of rotation is
also equal to its component, 8u = 82a-
The figure also indicates ho,v Ru may be located. 1'he line O.11.z
(or, si1nply, the z line) is erected perpendicular to the direction of transla­
tion. The line u, n1aking an angle ()23/2 with the z line, is then laid off;
the line v, the m.id normal of the translation, has its foot at L, which is
- T12/2 fro1n OA· This construction is summarized in Fig. 8-24.

--
Li--:--""'iOA
T1 2
2
\o
,.
FIGURE8-24 Construction of the relative pole Ril , given
Tu and 8u about O,. .
240 K I N E M A T I C SY N T H E S I S O F L I N K A G E S

- B 1_ _ _ _ _ _ _ ""9B2

- ----t-
o... I
r . ,. _,
l---n,,""'°
,(-

t---- S1z _ j
__

FIGURE 8-25 Displacements of a slider crank in need of

correlation.

A displacement of slider-crank mechanism from a position OAA 1B 1

to another OA A. 2B2 is shown in Fig. 8-25, where the crank rotation is t/>12
and the slider translation s12. This same displacement, as seen with
respect to the crank OAA 1, is shown in Fig. 8-26, where the slider is first
translated by the vector s12, B i moving to B2, and then rotated about
0A by the angle - (/)12, B 2 moving to B;. These two displacements of
the slider n1ay therefore be combined into a single rotation of angle - t/>12
about a center, now called relative pole R 1 2.
The process of inversion, i.e., fixing the crank OAA and letting

., z

�- ---·
"

\_ c/>12
2

--- S 12 ---- - ;.,-i

FIGURE 8-26 Determination of a relative pole Ru for slider crank; notation is

like that of Fig . 8-23.
 G E O M E TR I C M E THOSDS-THR E E A C C UR A C Y PO INTS 241
V

ll
I
\. I
i

R12
I
'P12

V0,4

FIGURE 8-27 Construc­ L

tion of the relative pole •12
of a slider-crank mecha­ _ 81 2
nism. 2

the follower rnove, has been invoked for Fig . 8-26; the notation is geared
to that of Fig. 8-23. The construction is shown in summary form in
Fig. 8-27.

Construction of the Relative Pole

Given OA, center of rotation of the crank; "112, rotation of the crank ;
and s12, corresponding rectilinear translation of the slider (Fig. 8-27).
1. Construct OAz perpendicular to s12 and the angle

zOAu = -2 <J,12

2. Construct OA L = - s12/2 and Lv parallel to OAz.

The relative pole R12 is the intersection of OAu and Lv.
Problem 6 Design a slider-crank mechanism in which a given
angular displace1nent q,12 of the crank produces a given linear displace­
ment s 12 of the follower (Fig. 8-28).
Solution
1. A.ssun1e a convenient center OA and direction of translation of
the slider OAs ; construct the relative pole R12 of the crank and follower
corresponding to <f,12 and s12.
2. Choose point A 1 arbitrarily.
3. Draw a line R1 2u such that
{3 = A1R12u = OAR12Os
(OB is at infinity in a direction perpendicular to the translation.)
4. Choose any point Bi on R12u.
'fhe desired linkage is OAA1B1.
242 KI N E M A TI C SYNTHESIS O F L I N KAGES

Os(oo)

\,
\ �1 2
z

OA

,.
---- u

13

.,
FIGt:RE 8-28 Solution of Prob. 6.

The solution of the above problen1 may be applied to modify the

1notion obtained from a ca1n n1echanis1n. The cam 1nechanisn1 of Fig.
8-29a is designed to give a ao0 oscillation of the follo,ver. It is desirable,
however, to 1nodify this 1nechanis1n and, by using the same can1, to
obtain a reciprocating 1notion w·ith a 1-in. stroke. A solution of this
problem, shown in· Fig. 8-29b, 1nakes use of a slider-crank mechanism
to convert the oscillation of the can1 follower into reciprocating motion
of the slider. The 1nechanis1n, however, 1nust be designed to give a
1-in. translation of the slider for a 30° rotation of its crank, which is also
the ca1n follower. The detern1ination of this 1nechanisn1 is shown in
Fig. 8-30. After constructing the relative pole R12, A. 1 is conveniently
chosen on the can1 follower at the center of the roller, R12u is constructed,
and B 1 is chosen along R12u to give a mechanisn1 having good force-trans­
n1ission characteristics.
Problem 7 Design a slider-crank n1echanis1n in which two suc­
cessive angular displacen1ents <J,o1 2 and </,23 of the crank produce, respec­
tively, two successive linear displace1nents s12 and 82a of the follower,
shown in Fig. 8-31.
Solution Let '1>1a = </,12 + (/,23 and 813 = s12 + 823.
1. Assu1ne a center OA and direction of translation OAS to con­
struct relative poles R12 and Ri a.
G E O METRIC METHODS-THREE ACCURACY POISTS 243

Slider

Coupler

-...... ----�--o- A,
-
OA o�- - ._ . _ _ _

(a) (b)

F'IGURE 8-29 Slider-crank mechanism to modify the motion obtained from a cam

cf> 12

112

w;})pn. ..._______
OA
</,l2 \

i 1 f- - :�- -- -�
--...........
2

--
1'
.===-- - ---1
::-,-----'- �:---.
-=R�12:::--'"---
---_-- ----- V
08 atoo

FIGURE 8-30 Synthesis of slider-crank mechanisn1 to modify cam motion.

244 KINEMATIC R Y N TH E S I S O F L I N K A G E S

cl>12 I

0.\
812

,
2
--
2
--i

--

09(C10)

FIGURE 8-31 Solution of Prob. 7.

2. Choose A 1 arbitrarily.
3. Draw lines R 12u12 and Re1 3u1 3 such that
and
4. The intersection of R 12u12 and Re1 au1 3 is B1.
The desired linkage is OAA 1 B 1 • 1'here are oo 2 solutions depending
on the choice of A 1.

8-8 EX A I\1 P L E : RECTI L I N E A R R E C O R D E R M E C H AN 1 S 1\1

The curvilinear coordinate chart paper of so1ne n1echanical record­
ing instrum.ents is necessary because the displacen1ent of the pen is along
a circular arc, not a straight line. The difference between sinusoids in
rectilinear and curvilinear coordinates is shown in Fig. 8-32, demonstrat­
ing how the distortion confuses the interpretation of the data. lVIechani­
cal recorders for rectilinear coordinates n1ay be designed by having the
pen travel on a proper portion of a four-bar coupler-point curve. The
principle of this application is shown in Fig. 8-33, where the coupler point
B (pen) must be n1ade to cover equal distances along a straight path
for equal rotations of the crank OAA within the range of the instrument.
The synthesis of this linkage will be carried out in three steps.
First, a slider-crank mechanism OA AB is designed such that equal crank
 G E O M E T R I C M E THODS-THREE A C C U R :\ C Y P O I N T S 245
displacements, within a specified range, will produce equal displacements
of the slider. Second, a coupler point C following an approximate circular
arc centered at Oc is found and a link OcC added. Third, the prismatic
pair guiding B along the prescribed straight line is re1noved ; the point B
,vill follo-w this line approxirnately, and B is the coupler point of the four­
bar linkage.
In the specific case considered here, the range of rotation of the
crank is taken equal to 90e°, for ,vhich the total displacement of the pen
is to be 4.5 in. Three accuracy points are chosen with Chebyshev
spacing,
<P12 = 39e
°
<f,13 = 78°
s12 = 1.73 in. 81 3 = 3.46 in.

In Fig. 8-34, 0A is the center of rotation of the crank and OA.1: the direc­
tion of translation of the slider. The relative poles R12 and R 13 are con­
structed, point A., is chosen arbitrarily, and the point B, found fro1n the
intersection of 1,.1 2 and u, 3 • The linkage OA A1B1 is then redrawn (Fig.
8-35) and its positions OA A 2B2 and OA A3Ba, corresponding to the other
two accuracy points, constructed.

,,
T7 I
l
I
I I
I
I !

i
!
I
I
I I

O r----;----,::-1f'---+----+4
2- --
1r
'
FlOURF: 8-:32 Comparison of the I

same sine function . Upper curvr

\
drawn in rectilinear c-oordinates : \
'\
lower <:urve drawn by pivoted \
stylus. \
_____ _..,..c_ ______ - · · - .•.. -- �--'
246 K I N E M A T I C SYNT H E S I S OF LINKA G E S

( B
B,

\ � Coupler curve of C
as part of slider-crank
\ mechanism OA AB
I
�---�c nGl.RE 8-33 Principle of
the rectilinear instrument
recorder.

-----
--
ti,
12

i...-- _ 'P13

---------�----- . � 0,4 --- - - -

- --------
- 2 ➔i
:c
8 1:! �

I..
i - S1� I
___ _ ;
I_ _ _
£�

l
Oa(at oo)

FIGURE 8-34 Construction of slider-crank ffif'(1hanism for rectilinf'ar recorder,

three accuracy points.
 -,- -
G E O M E T R I C M E T H O D S-T H R E E A C C U R A C Y P O INTS 247

I-- - ---' +
.--
-
- --<
� "'C 1

-- -- ........ �
--:oA 2
/
C
• 3/
,,.,,
✓
,,.,

,,.,,,,.,, �
✓

/ ,,., ,,.,,

-!i �or
� /��O
� �
A,,.,���--
---
---
.-
:��-
-�-�
�--- 3-- x
A
�, �'S

FIGURE 8-35 Construction of revolute center Oc for rectilinear recorder,

three accuracy points.

0
Uniform Scale
0 1 2 3 4 5 6 7 8 9 10
Bi
• Actual positions of B
)( Accuracy points

(IO

_ ....

De
i,·rouRE 8-36 Rectilinear recorder, the link OcC now replacing the slider
B (three accuracy points, with Chebyshev spacing).
248 KINE M A T I C S Y N T H ESIS O F L I N KAGES

In this particular application, it is 1nost in1portant to li1nit the

"'eight of the 1noving parts to mini1nize the inertia of the instrument.
Choosing the coupler point C' along the line AB will allow the coupler

close to A., the center Oc of the circle passing through the three positions
ABC (Fig. 8-35) to be a slender bar. If, further, the point C is chosen

C1 , C 2, Ca will be near OA and the resulting four-bar OAACOc will have

poor force-transmission qualities. Should C be chosen close to B, the
center Oc will be remote and the space requiren1ents will be hard to meet.
The point C is therefore chosen about halfway between A. and B ; its
positions C 1, C2, and C3 are constructed (Fig. 8-35), and the center Oc is
found as the intersection of the 1nidnormals to C1 C2 and C 2C a. 1�he
resulting linkage OA ACOc, with its coupler point B, now complete, is
drawn once again in Fig. 8-36 and its accuracy checked in terms of 10
equal rotations of the crank.
A.s shown in Fig. 8-36, the error takes two aspects: there is a devia­
tion fro1n the straight line, approxin1ately equal to 0.5 percent of the
output travel, and a lack of linearity between crank rotations and coupler­
point travel, approximately equal to l percent of the output travel.
These errors follow a regular pattern and vanish at the accuracy points ;
they are n1ostly the result of the structural error and n1ay therefore be
reduced by the introduction of an additional accuracy point. Such a
synthesis, with four accuracy points, will be carried out in Sec. 9-5.

BIBLIOGRAPHY

Beyer, Rudolf: "Kinematische Getriebesynthese," chaps. I, II, Springer-Verlag

OHG, Berlin, 1953. English translation by H. Kuenzel, "The Kinematic
Synthesis of Mechanisms," McGraw-Hill Book Company, New York,
1963.
Hall, Allen S., Jr.: "Kinematics and Linkage Design." chap. 8, Prentice-Hall,
Inc., Englewood Cliffs, N .J., 1961.
Hirschhorn, Jeremy: "Kinematics and Dynarnics of Plane ::vrechanisms,"
:\IcGraw-Hill Book Company, New York, 1962.
 GEOMETRIC METHODS

OF SYNTHESIS WITH
FOUR ACCURACY POINTS

9-1 POLE TRIANGLES

Consider two successive rotations of a rnoving plane II with
respect to the fixed plane � (Fig. 9-1). A first rotation about
pole P 12 through angle 012 brings II from position II 1 to position
II 2 ; a second rotation about pole Pn through angle 023 brings
II from II 2 to II 3• The total displace1nent of II fro1n II 1 to Ila is
also (in general) a rotation about a pole Pia and through an angle
0 a. The triangle P12P2:1P1a is called the pole triangle associated
with the two successive rotations of pole P1 2 and P n and angles
1

012 and 02a.

The properties of pole triangles to be discussed in this section

will be used later in considering several positions of a plane in
planar motion. 1'hus, when two successive rotations 012 and 02a
about poles P 12 and P2a are known, the pole P13 and the angle 013
of a single rotation equivalent to the two may be found by con­
structing the corresponding pole triangle. The successive posi­
tions of a point of the moving plane n1ay also be found by simple
constructions involving the pole triangle.
250 KINEMATIC SYNTHESIS OF Lll".K.-\GES

l: (fixed plane)

'.
I \

l'· \
\

FIGURE 9-1 Pole triangle cor­

responding to three positions
of a moving plane.

THEOREM I I The angles of the pole triangle and the angles of rotation
satisfy the follow·ing equalities:

=-
823
P 12]� 13p23 =
813
2
in magnitude and direction, except for the possible addition of an angle
of 180°.

Proof Consider Fig. 9-2, in which link Ao2B2 in position 2 is coin­

cident with P12P23. In position 1, this link assun1es the position A 1B 1,
where Ao1 is coincident with P 12 and Ao2 . In position 3, this link assun1es
the position Ao3B 3, where Bo3 is coincident ,vith B2 and P 23 .
The pole P13 is the intersection of the midnorn1als to B1B3 and
A1Aa, ,vhence P 12P 1a bisects the angle B 1P 1 2B2 = 812 and PnP13 bisects
the angle A 2PnA a = O�a, or

and

l:Fixed plane

I As
·pl3
oso· + '13>
2

FIGURE 9-2 Proof of Theorem II.

 GEOMETRIC METHODS-FOUR ACCURACY POINTS 251

NOTE: A pole triangle con1pletely defines two rotations, since it

locates the two poles and gives the corresponding angles of rotation.

Cardinal Point
Given the pole triangle P12P23P13 corresponding to three positions II1, IJ 2,
and II 3 of a rnoving plane II and a point C of II which occupies position C 1
when II is in position II1, the further positions C2 and Ca corresponding
to positions II 2 and Ila n1ay be found by two rotations. First, C1 i�
rotated about Pe12 by the angle 812 = 2(PuP12P23) to obtain C2 ; then C::
is rotated about P 23 by the angle 92a = 2(P12P23Pu) to obtain Ca,
The points C2 and C3, ho,vever, may also be found by reflecting C1 about
side P12P13 of the pole triangle; this gives a point Cg called the cardinal
point (Fig. 9-3). Reflection of the cardinal point Cg about sides P 12P 23
and P13P 23 gives, respectively, C2 and C3. The validity of this construc­
tion 1nay be recognized fron1 consideration of the two angular relations
C1 P1 2C2 = 2(P1aP12Cg + Cg P12P2a) = 812
and C2 P23 C3 = 2(P1 2P2 aCu + Cg P2sP13) = 823
The cardinal point Cg was obtained in Fig. 9-3 by reflecting C1
about side P12 Pu . Note here that the subscript I of C1, indicating that
C is in position 1, is also the repeated subscript in side P12l"'1a. Sin1ilarly,
C2 is the in1age of C0 about P1 2P2a, subscript 2 repeated, and Ca is the
imag e of Cg about P13 P23, subscript 3 no,v repeated.
NOTE: A cardinal point is not a uniquely defined property of a
pole triangle (as an orthocenter, circumcenter, or centroid) but is associ­
ated with a given moving point. Thus, there is a cardinal point for
each point of the 1noving plane.

FIGURE 9-3 Use of cardinal point Co for

the construction of three positions C1, Ci,
Ca of a point.
252 KINEMATIC SYNTHESIS OF LINKAGES

Circle through Three Positions of a Point

THEOREM III The center of a circle through three positions of a moving
point and the corresponding cardinal point (Fig. 9-4) are related in tern1s
of the pole triangle by the following angular equalities:

Consider a pole triangle P 12P 23P 1 a and a point C of the n1oving

plane whose three successive positions are C i , C2, and c�. Given the
first position C i , the other t,vo positions C2 and C a may be found by
1neans of the cardinal point Cg (Fig. 9-4). The three points C 1, C 2,
and C 3 lie on a unique circle k<, whose center Co n1ay be found as the
intersection of the n1idnorn1als to C 1 C 2 and C2C 3• The 1nidnorn1al to
C 1 C 3 also passes through the center C o. The three midnorn1als pass
through the respective poles P 12, P 2a, and P ia, since P12 is equidistant
from Ci and C2, P2a equidistant from C2 and Co3 , and P i a equidistant
from Ci and C3.
To prove the equalities expressed in Theoren1 III, consider the
triangles P12CoC I and P1 2CoC 2 (Fig. 9-5). These triangles are equal,
since P 12C o is the n1idnormal to Ci C2, whence

C0Pi2C2 = CiPi2Co
or 2a' + -y = 2a + -y or a' = a

fron1 ,vhich the first equality results. The other equalities are obtained
in similar fashion.

FIOCRE 94- Circle through three positions

of a point, Theorem III.
 GEOMETH!C METHODS-FOUR ACCURACY POINTS 253

FIGURE 95
- Proof of Theo­
rem III.

Image Poles of a Pole Triangle

The pole triangle P1 2P23P13, corresponding to two successive rotations
of a moving plane II with respect to a fixed plane 2:, is a fixed triangle in
the plane :r. It will so1neti1nes be convenient to consider the inverse
situation in which II is fixed and� is n1oving. 1'he pole triangle P12P23P 13
then moves with 2:, and it is necessary to find the pole triangle fixed in n
that will describe the san1e relative rotations between II and 2:.
'fhe first rotation of IT with respect to 2: about P12 ,Yith angle 012
(Fig. 9-6b) yields, ,vhen considered with respect to 11 1 , a rotation also
abou t P 12 but of angle - 812 (Fig. 9 6 - c). This rotation ,Yill bring the pole
P2a fixed in 2: to a new· position Pn 1 (Fig. 9-6a). Note here that, because
of the angular relationships of a pole triangle expressed by Theorem II,
P23 is the image of pole P 23 about the opposite side P1 2P 13 • The point
1

P2a is thus the image pole of P2 3.

1
The second rotation of 11 ,vith respect
to }; about P23 with angle 82a (Fig. 9-6b) yields, when considered with
respect to Il 1 , a rotation about P2a 1 (new position of P2a) with angle - On.
The rotation that brings IT directly fro1n II1 to Ila, ,vhen :r is considered
fixed, has pole P13 and angle Ou (Fig. 9-6b). The corresponding rotation
of Z ,vith respect to 111 has the sa1ne pole Pia but an angle -813 (Fig.
9-6c). The pole triangle of the two successive rotations with respect
to Il1 is thus P12 P23 P13, shown in broken lines (Fig. 9-6a). It n1ay be
1

obs erved that the angles of this triangle satisfy Theorem II, namely,
812
--= angle of rotation from ! 1 to 2: 2
2 2
823 _ angle of rotation from Z2 to 1: 3
P12P2ao1Pu = -
2- 2

P1 2P1aP2� = - i
O u = angle- of rotation from 1: 1 to 2:-·3
1
--- -- 2
254 KINEMATIC SYNTHESIS OF LINKAGES

Once the pole triangle P12P23 1P1a describing the rotations of �

with respect to II 1 has been found, the corresponding pole triangles with
respect to II 2 and Ila are n1erely the new positions of P12P231P1a when IIe1
is moved successively to II2 and Ila. Thus, the pole triangle P12P23P1 a 2

,vith respect to II 2 is obtained by rotating P1 2P231 P13 about Pe12 by the

angle 812 (Fig. 9-7a). Because of the angular relationships in pole tri­
angles, P1 3 is the i1nage of Pia about P12P 23, and P13 is the image pole
2 2

of P13• Like,vise, the pole triangle P123P2aPu ,vith respect to Il3 is

obtained by rotating P1 2P23P132 about P23 by the angle 823 (Fig. 9-7b);
P123 is the image pole of P12 about P2aP1 a.
The discussion of i1nage :;:.-oles may be summarized as follows:
Let P 12P2aP13 be a pole triangle in the fixed plane i representing two
successive rotations of a 1noving plane Il with respect to � (fron1 Il1 to
11 2 to IIe3). The pole triangles representing the san1e relative rotations

(a)

With :E fixed: With IIfixed in positionfl 1:

pole triangle P1 2P 23P 13 :E takes positions l:" .E 2, :E3

I...
Il i �-::--:---112�

_(_
2
_,_
P _, 2, _>
___
8
�--
-

(P_2_3 _, _23__
8
II 3

) __,)
l
P23
- - -
--PJa

(b) (c)

FIGURE 9-6 Image-pole triangle P121'231P, 3 (in position 1).

 G E O M E T R I C M E T H O DS-FOUR ACCURACY P O I NTS 255
Pb
�--
\
"'
-
- - --
- P13

\ � n
\ A
b,
\
\
'
\

I \ I
\ I \ I
\ I
\ I \ I
\ I \ I
\ I \ I
\ I \ I
\ I \ I
�2
P13
(a) (b)

FIGURE 9-7 I_mage-pole triangles P12PnP132 (in position �) and P12 P2.J'13 (in
3

position 3).

of :Z ·with respect to IT in positions IT1, IT2, and Il3 are, respectively,

P12P2a 1 P1 :1, P12P23P 132, and P1 23P23P 13, in which the points P 23 , P13 , 1 2

and P 2 are the inlage poles of the given poles P23, P13, and P12•
1
3

9-2 FOUR POSITIONS OF A PLANE

The considerations of the last section will be extended here to
the case of three successive rotations, or four positions, of a n1oving
Plane IT with respect to a fixed plane l; (Fig. 9-8). The first rotation,
of angle 8 1 2 about P12, brings IT frorn position II1 to position Il 2 ; the second
rotation, of angle 823 about Pn, rnoves IT from Il2 to Ila; and the third
rotation, of angle 834 about Pa◄, deposits n at Il4.
In tern1s of these original rotations, three other rotations may be
derived. A rotation of angle 813 about P13 would bring II fron1 II1 to Ila;
a rotation of angle 82 about P2• ,vould bring II fro1n Il2 to II,; and a
4
rotation of angle 8 about P1 4 ·would bring II from II1 directly to Il4.
14
Thus, the consideration of four positions of a n1oving plane II with
respect to a fixed plane 2": involves a total of six poles,

P12 P23 l'a4 Pi:i P2, P1 4

These 1nay be grouped to fonn four different pole triangles,

as shown in Fig. 9-9.

 .I

256 KI�EMATIC SYNTHESIS OF LINKAGES

Note:
The constructions for only
P12 and P2a are shown

•Pu

F'IOURE !J-8 Construction of poles corresponding to four positions of a link.

When a n1oving plane II assumes three positions II 1, II2, and II 3

with respect to a fixed plane �, any point C of the moving plane takes
three positions G1, C2 , and G3 through which a circle (or a straight line)
may be drawn on 2:. If a straight line is considered as a circle of infinite
radius, it may be said that a circle n1ay always be drawn through three
positions of a moving point. The relationships between the center Co
of this circle and the corresponding pole triangle P12P2aP1 3 have been
discussed in the last section in terms of the cardinal point C0• With
four positions II1, Il2, Ila, and II., a point C of the moving plane will
assu1ne four positions C1, C2, Ca, and G4, ,vhich in general do not lie on
the same circle. Thus, we see (Fig. 9-10) that G 1 , C 2, and C 3 nlight lie
on a circle of center Co with radius R. = CoC1 = CoC2 = CoC3. If G4
is to lie on the same circle, however, the distance CoC4 1nust be equal to
the radius /?r and this ,vill occur only for certain points C of the rnoving
plane . Sin1ilarly, only a particular point of the fixed plane 2: may be
chosen as the center of a circle passing through four positions of a n1oving
point.
A point C of a n1oving plane II assuming four positions C1, c�,
C3 , C 4 lying on a circle ,vhen II assun1es four positions II 1• II:i, rr�, IT 4 is
G E O M E T R I C METHODS-FOUR ACCURACY P O I NTS 257

called a circle point. The center Co of such a circle in the fixed plane X
is called a center point.
Circle points and center points ,vill play an important role in
extending geon1etric synthesis to four positions. Circle- and center­
point detenninations, based on Theoren1 IV to follo,v, are conveniently
made in ter1ns of the so-called opposite-pole quadrilaterals.
Opposite poles are defined as t,vo poles carrying different subscripts :
there are three pairs of opposite poles,

connected in Fig. 9-11 by solid lines. An opposite-pole quadrilateral has

its diagonals connecting t,,,ro opposite poles; there are three opposite-pole

P24 (b)

FIGURE !J - H Pole triangles corresponding to four positions of a link.

258 K I N E M A 'f l C S Y N T H O
E � I S 0 1'' LIN K A G E S

FIGURE 9-10 Four positions of a

point do not always lie on a circle.

quadrilaterals,
(P2aP1aP1.P24)
The distinction betw·een sides and diagonals of a quadrilateral,
,vhich is the clue to the above definition, is son1ev. hat out of the ordinary
1

and merits particular attention. Consider the quadrilaterals of Fig.

9-12; the distinction between sides and diagonals depends only on the
order in ,vhich its vertices A, B, C, D are nan1ed and is independent of
the actual shape of the quadrilateral. In the case of the quadrilateral
A.BCD (Fig. 9-12a), the sides are AB, BC, CD, DA (in broken lines),
and the diagonals are .4 C and BD (in solid lines). Ho,vever, the quadri­
lateral ACBD (Fig. 9-12b), whose sa1ne four points are named in an
order different fro1n that of the first, differs fron1 A BCD in that its sides
are AC, CB, BD, DA (broken lines), ,vith diagonals A B and CD (solid
lines).
The three opposite-pole quadrilaterals are sho,vn separately in
Fig. 9-13; in each there are four sides (broken lines) and two diagonals
connecting the opposite poles (solid lines). Two sides are adjacent if
they have a vertex in comn1on and opposite if they do not.

F·1ounE 9 - 1 1 Opposite
poles (joined by solid lines)
have different subscripts.
GEOMETRIC METHODS-FOUR ACCURACY POINTS 259
C C

B \
I
\
\
I \
I \
I \
A D A D
(a) Quadrilateral ABCD (b) Quadrilateral ACBD

FIGtTRE 9-12 Diagonals (solid lines) and sides (broken lines) of

quadrilaterals.

P 13
(a) P2aP1aP14P24
Diagonals: P23 P14 and P13P24
Sides: P23 P24 opposite to P13 P 14

-
/
/
/ P13 P23 opposite to Pu P24
/ -
/,- ­
b
-
(b) P12P1aPa4P24
Diagonals: P12 P34 and P24 P13
I Sides: P12 P13 opposite to P24 P34
I
I P12 P24 opposite to P13 P34
I
I
I
I
I
I
I
I
-- - - -- - - -

Diagonals: P12 Pa4 and Pu P2a

Sides: P1 2 Pu opposite to P23 Pa4
P12 P23 opposite to P14 P34

FIGURE 9-13 Three opposite-pole quadrilaterals correspond­

ing to the poles of Fig. 9-11.
260 K I N E M A T I C S Y N T H E S I S O F LINKAGES

k'

l,..
·o

FIGURE 9-14 Construction for

.
h" renter potnts C'O an.
d C"
0 •

THEOREM IV 1 When a n1oving plane Il assun1es four positions ,vith

respect to a fixed plane, the opposite sides of each opposite-pole quadri­
lateral subtend angles that are equal or that differ by 180° at a center
point. Conversely, any point from which t,vo opposite sides of an oppo­
site-pole quadrilateral appear under equal angles or angles differing
by 180° is a center point.

The construction of a center point is then as follo,vs: Given the

opposite-pole quadrilateral P 1 3P1 4P24P23, as in :Fig. 9-14.
1. Construct the midnorn1als h' and h" of two opposite sides,
P 1aP1 4 and P23P24, for example.
2. Construct arbitrary angles
-�,,.,p 1 4P1 3
: 11 11 P24P23
.. = 1r� = a
3. Dra,v circle k' of center J.ll ' through PH and P ia, and k" of
center 111" through P24 and P 23•
4. The intersections of circles k' and k" are C� and C�'.
The central angles F'1 3ll/'P14 and P2a1.l,f''P24 are equal; the inscribed
angles P13C�PJ4 and PnC�P24, respectively equal to half their central
angles, are therefore equal to each other. Si1nilarly, the inscribed angles
P1aC�'Pu and P2 aC�'l'24 are also equal. 'fhe opposite sides P13P 1 4 and
P23P24 of the opposite- pole quadrilateral P1 aP 14P 2 4P 2a therefore subtend
equal angles from C� and C�, whence both C� and C� are center points
by the converse of 'fheore1n IV. The opposite sides P1 3P14 and P 23P24
of the opposite-pole quadrilateral P 13P 14P24P23 were chosen to carry
out the above construction. The san1e construction, however, could
1 Proof of this theorem is given in the next section.
GEOME T R I C METHO
ODS-FOUR A C C U RA C Y P O I NTS 26 1
have been made by using the other two opposite sides P 1 4P 24 and P 13P 23
or by using another opposite-pole quadrilateral.

Center-point Curve
For different values of the arbitrarily chosen angle a, different center
points C� and C� will be found. The locus of these points as a varies
fro1n - 90 to 90° constitutes the center-point curve Ce, This curve has
been drawn in Fig. 9-15, 1 the opposite sides P13P1, and P23P24 of the
1
The curve in the present case has a single branch, which extends to infinity.
In some situations there are two disconnected branches, one being closed and the
other extending to infinity.
1, 2, 3, 4
\\
\
Y,L
\
\
\
10
10 \
91
1 11

12 1
19 I
\ �
1 7 pl23
8 c, 16
Pu 813
2
5

18 19
1 c. P:u

\P;.

I
�"
I
I
ts
I
FIGURE H-15 Center-point curve c, (solid line) and circk--point
curve c; (broken line), with corresponding points identified.
262 K I N E M A T I C SYNTHESIS OF LINKA GES

opposite-pole quadrilateral P1 aP14P 2◄P 2J being used. Any point of this

curve is a center point, and it may be noted that the curve passes through
all six poles corresponding to the four positions under consideration.
In addition the center-point curve passes through the intersections
(not shown)
Q12 of P 1 aP 2a and P1◄P2•
Q1 a of P 1 2P2a and P 14Ps4
Q14 of P 1 2P24 and P1 aPa,.
Q2a of P 1 2P 1 a and P24Pa◄
Q24 of P 1 2P1 4 and P 2aPa•
Qa4 of P1aPu and P2aP24
The exact construction of a center-point curve is a lengthy pro­
cedure. Fortunately, a sketch of the curve may often be roughed out
on the basis of the six poles and the above intersections.

Construction of a Circle Point

1'he construction for a circle point, i.e., a point of the moving plane II
occupying four positions C1, C2, Ca, C4 lying on a circle for four positions
of the n1oving plane, is deduced fron1 that of a center point by simply
reversing the roles of the fixed and n1oving planes. If the motion between
IT and � is considered ,vith respect to Il1 , a center point Co fixed in I
occupies four positions C01, Coz, Coa, Co• equidistant fron1 the correspond­
ing circle point C1 . The circle point C 1 1nay therefore be constructed
as a center point, but by using the six poles representing the three succes­
sive rotations of � with respect to Il 1. These ne,v poles are found as
shown in the last section; thus P1 2, P1 3, Pu are unchanged, and P23, P24,

FIGURE 9-16 Construction for nircle

points.
 G E O M E T R I C M E T H O D S - F O U R A C C U RA C Y P O I N T S 263:
Pa• must be replaced by their image poles P2a1 , P241 , P·a41 • . Once these
image poles are found (see Fig. 9-7), an opposite-pole quadrilateral such
as PuP 1 4P24 1 P23 1nay be selected and the construction proceeds as in
1

the case of a center point to yield two circle points C� and c;' (Fig. 9-:- 16) ..

Circle-point Curve

As in the case of center points, the locus of the points c; and c;' ,vhen a
varies forn1s the circle-point curve Ci- This curve has been drawn in
Fig. 9-1.5 (broken line). Note that the circle-point curve in 111 passes
through the poles P12P1 :1 P1 4 and the i1nage poles P2: s 1, P2. 1, Pa, 1 as well
as through the intersections (not sho,vn)
Q1 2 1 of Pi:1P2aP
1
and P14P2◄ 1
Q1 3 of P1 2Pn 1 and Pul-'a4 1
1

Qu 1 of P1 2P24 1 and P1;1P:14 1

Q23 of P1 2P1 a and P24 1Pa4 1
1

Q2 4 of P1 2Pu and Pn 1P34

1 1

Q34 of P 13 [-''u and Pn 1 P24 1

1

To each circle point there corresponds a unique center point.

Additional inforn1ation n1ay therefore be displayed on the curves of
Fig. 9-15 if corresponding points are identified. The center point Co
corresponding to a given circle point C1 n1ay be found by using the pole
triangle P1 2P23P1 3 and the angular equalities of Theorem III. Examina­
tion of Fig. 9-4 sho·ws that
2 0
or 81 + 1 80
2

or 813 + 1 800
2

Consider, for exa1nple, the circle point C1, point 1 6 on the circle­
point curve (Fig. 9- 1 5) . Construct lines P12U and P1av such that

and

The lines P1 2u and P1 3v intersect on the center-point curve at the corre­

sponding center point Co identified as 1 6 on that curve. If the center­
point curve is already known, it may be possible to locate the correspond­
ing center point by means of a single line such as P12u, but ambiguity
may arise, since this line generally intersects the center-point curve at
several points: center- point location should be checked with a second line.
Problem 8 Design a four-bar linkage to transfer a plane IT
through four given positions II 1, II2, Ila, rr.
(Fig. 9-8).
264 K l N EM A T I C SY N TH ES I S O 1'' L I N K A GES

The plane II is taken as the coupler plane of the four-bar. The

revolute centers (A and B) of the coupler must be chosen as circle points
of Il, and the fixed revolute centers (OA and O B) must be the corresponding
center points.
Solution
1. Construct the six poles corresponding to the three successive
rotations. This may be done by first choosing two points D and E of
the moving plane. These points assume four positions D1E1, D2E2,
D3Es, D 4E4 with respect to the fixed plane �- The construction of two
of the poles P12 and P2a is shown in Fig. 9-8.
2. Select an opposite-pole quadrilateral, say P 1 4P1 3P23P24, and
construct two center points OA and OB• This construction n1ay be car­
ried out as shown in Fig. 9-14, 0A and OB being taken as C� and C�.
The points OA and OB n1ay, however, be chosen anywhere along the
center-point curve Ce (Fig. 9-15).
3. Since the circle-point and center-point curves have been co11-
structed with corresponding points identified by nun1bers, we shall n1ake
use of the available infonnation. Upon taking OA and Os at 5 and 10
on the center-point curve, the revolute centers A 1 and B1 of the coupler

FIGt:R� �-17 Probh!m 8. the• final linkage, using points 5 and 10 of the center­
and ,•irclE>-point <·urveR.
 GE O M E T R I C M E T H O DS-FOUR A C C U R A C Y POINTS 265

"11,

225°

37•--J
_..,._1,..-...,.._�--�----!- ,
1 ----L.-- ,J,P
--65° �,
FIGURE 9-18 Problem 9,dis­
placement specifications. - - - 1 17°-- -...i

plane 11 1 n1ust be taken at 5 and 10 on the circle-point curve, and the

desired linkage in position 1 is OAoA1B10s (Fig. 9-17).
The proble1n ad1nits oo 2 solutions because of the arbitrary choices
of OA and O8 on the center-point curve. It n1ay be verified that the
linkage of Fig. 9-17 does not satisfy the Grashof condition but goes
through dead points between specified positions. A few other trials,
,vith different center and circle points, failed to improve this unfortunate
state of affairs.
The further examples worked out in this chapter, however, show
that such difficulties may often be avoided. As a matter of fact, the
solution of many problems does not require the determination of the

FIGURE 9-19 Problem 9,

constru ction of relative
P<>les.
266 KINEMATIC S Y N T H E S I S O F L I N KA G E S

complete circle-point and center-point curves. It is necessary to locate

only one or two pairs of corresponding circle and center points. Note
that two center points may be found by a single application of the con­
struction shown in Fig. 9-14. Furthern1ore, the poles themselves and the
intersections Qi; are easily found center points and 1nay often be used as
frame revolute centers.
Problem 9 Design a four-bar linkage in which three successive
angular displacen1ents <1>12, <Pza , <Pa• of the crank produce, respectively,
three successive angular displacements "112, "1n, V'a• of the follower (Fig.
9-18). Franle OAOn is given.
Solution
1. Construct five relative poles of the follower with respect to
the crank (Fig. 9-19).
2. Choose an opposite-pole quadrilateral, such as R 1 4RuR 2aR 24 ,
and construct a center point A I in the crank. Take OA as the other center
point (Fig. 9-20).
3. Construct angles
'1>12 - ¥1 1 2
.. 4 1 R2U1
1

and A 1Ruv1 - '1> 1 a - ¥'13

J-'IGURE 9-20 · Problem . 9,

construction of revolute ct>n­
tt>rs.
 . F' O U R A C C U R A C Y P O I N T S
GEOMETRIC M E T H O D S- 267
At .,.

FIGURE 9-21 Problem 9,

the final linkage.

The intersection of R 12u 1 and R1:iV1 is B1. Theore111 I is invoked h�re_as

in Probs. 4 and 5 (Sec. 8-3). The desired linkage is OAA1B10s (Fig.e-9:21).

9-3 PROOF O F T H E O R E M IV
The theoren1 is repeated for easy reference : When a n1oving plane
IT assu1nes four positions with respect to a fixed plane, the opposite sides
ofeeach opposite-pole quadrilateral subtend angles that are equal or that
differ by 180° at the center point. Conversely, any point fro1n ·which
two opposite sides of an opposite-pole quadrilateral appear under equal
angles or angles differing by 180e° is a center point.
The first, or direct, part of the theorem n1ay be proved by referring
to Fig. 9-22, which shows four positions of a 1noving plane IT with the
corresponding four positions Ci, C 2, C3 C, of a point of this plane lying
1

on a circle centered at C0• Here Co is a center point, and the point C is

the corresponding circle point. The poles Pia, Pu, P2�, and P24 lie,
respectively, on the midnormals of the segn1ents C1Ca, C1C,, C2Cs, and
C2C4, that is, on the bisectors of the angles C1CoC2, C1CoC,, C2CoC3, and
C2CoC,. Let (CiCoC;) = a,; (for exa1nple, C1CoC4 = a1,); then

P 13 C]P 14 = au aiid p2 3C0p24 _ a2 4 a2� a:u

2-2
a34
-
-
a14
=
2 2 2 =
2
whence P13 C0P14 = P2aCoP24
. The segments P1 3P 14 and P23P24 are opposite sides of the opposite-
pole· quadrilateral P� 3P23P14 P2, and therefore subtend equal angles from
268 KI N E M A T I C SYNTHESIS OF L I X KA G E S

-- - --
--_ __-_______
,ol'l'.1

----
_____i_-- --
C2 _.,,. _ _.-
- to P1•

Co

-- ..._ ---- --
to p
2,

= =
FIGt:RE 9-22 Proof of direct part of Theorem IV. l "" a 13/2; 2 ""' au/2; 3 :::::
(au - a13)/2; 4 au/2; 5 = a2./2; 6 (a2• - a23)/2.

the center point Co. The same reasoning n1ay be repeated for the other
pair of · opposite sides of the quadrilateral as well as for the opposite sides
of other quadrilaterals. This completes the proof of the first part of the
theoren1.
To prove the second, or converse, part, the case of three positions
of a moving plane will be reconsidered, to establish the radius of the circle
passing through three positions of a n1oving point. 1'hree positions C1 ,
C2, and Ca of a 1noving point C are shown in Fig. 9-23, together with the

FIGURE 9-23 Det..-rmination of R. of

circle k•.
 GE OMETRIC M E THODS-FOUR A C C U R A C Y P O I N T S 269
pole triangle P 12P 23P 1 3 and the center Co of the circle kc passing through
Ci, C 2, and C3• Applying the sine law to the triangle P23C0C 2 yields
CoC2 CoPn
sin (C2P2aCo) - sin (C�C2P2 ;)
frorn ,vhich the radius Re of the circle kc rnay be expresia;ed as

Rc _- CoC2 -_ Cop2a ,/ sin

sin (C2P2aCo) !
(CoC2P2a)- I.
But, from Theorem III, C2P2�Co = 823/2, and, upon referring to Fig. 9-23,
83
CoC2P23 = 180° - ;,
..,
- b

However, still with reference to Fig. 9-23, the triangles P1 2C 0C1 and
P12CoC2 are equal, as already den1onstrated in the proof of Theorem III.
Similarly, P23 CoC 'i. is equal to PnCoCa and P13C0Ca equal to P 1 3C0C1,
whence
+ 2b + 2c = 360°
2a
or b = 180° - (a + c) = 180 - l>12CoP13
°

823
from which C0C2P23 = P12CoP13 -
9-
The radius of the circle kc is therefore found to be
sin (023 / 2) I
Re = CoP23 sin (P12C0P u - 82 a/2) I
It may be noted that the radius depends only on the distance C0P 23, the
angle 823/2 of the pole triangle at P2a, and the angle P12CoP 1a under ,vhich
the side P1 2P 1 3 appears from the center Co-
Consider now the set of six poles corresponding to four positions
of a moving plane (Fig. 9-8). These six poles n1ay be grouped to form
two pole triangles P1 2P23P13 and P2aP34P24 (Fig. 9-9), in which P23 is
con1mon to both triangles. The sides P12P1a and P2,Pa 4 opposite the
common pole P 23 are two opposite sides of the opposite-pole quadrilateral
P12P13P34P24 . Let Co be a point fron1 which these opposite sides appear
under the san1e angle. Three positions Ci, C2, Cs of a point C of the mov­
ing plane lying on a circle kc centered at Co 1nay be found as shown in
Fig. 9-24 : the cardinal point C0 is constructed according to Theore1n III
with respect to the triangle P1 2P2aP1a and then reflected about the sides
of the triangle. With this, the radius of the circle kc passing through
C1, C 2, Ca is

(9-1)
270 KINEMATIC SYNTHESIS O F LINKAGES

FIGURE 9-24 Proof of reverse part of Theorem IV.

When the sa,ne construction is repeated ,vith respect to the triangle

P24P23P34, it yields by reflection of the cardinal point C� about P24P�4 the
fourth position C4 of the 1noving point C and
sin (82a/2 180o)
CoC4 = CoP2 a . °"; (9-2)
Sin (82a/2 + 180 - P24C oPa4)
But since
a comparison of Eqs. (9-1) and (9-2) sho,vs
CoC4 = Re
and the circle kc also passes through C4 ,vhich con1pletes the proof of the
second part of the theorem.
1

9-4 EX A M P L E : S I N U S O I D A L - F U N C T I O N G E N E R A T O R
In the design of a sinusoidal-function generator with three accuracy
points (Sec. 8-5) it was noted that the inaccuracies of the resulting linkage
 G EO M E T R IC M E T H O O
D 8 - F O U .lt A C C U H A C Y P O I NTS 271
were due mostly to the structural error and that greater accuracy would
be obtained by increasing the number of accuracy points to four.
Such a new design will now be considered as an application of
Prob. 9. The function y = sin x ,vill be generated in the interval
0 < x < 90° on the basis of four accuracy points with equal spacing,
thus,
X1 = 0 X2 = 30 Xa = 60 X4 = 90
° ° °

Since the structural error is now expected to be small compared ,vith the
graphical error, no particular care is taken to n1inin1ize the structural
error by n1eans of a Chebyshev spacing of accuracy points. The cor­
responding values of y beco1ne
Yi = 0 Yz = 0.500 Ya = 0.866 Y• = 1 .000
By taking the ranges of variations of <f, and y; equal to those used in the
successful three-accuracy-point synthesis (Sec. 8-5), A<f, = 120° and
Ai/I = 60°, the rotations of the crank and follo·wer between accuracy points
Xi and xi 1nay be evaluated as

- 1 . 33(x,· -
120° -
X;
</>ii = ,Lj -
9Qo X;)

y;.; = Y; - Yi 600 = 60(y; - Yi)

1
whence
</>12 = 40° </>13 = 80° <f,14 = 120° <f,23 = 40° <f,24 = 80° <f,34= 40°
1/112 = 30° Yll3 = 52° V/ 14 = 60° V/23 = 22° y;2, = 30° V/34 = 8°
The relative poles R 12, R1a, Ru, R2a, R24, RH n1ay now be con­
structed on the basis of the above rotations. The construction for the

9-25 Function generator y = sin x, 0 < x < 90 , four accuracy points:

FIGURE °

construction for relative poles R 12, Rta, and R14.

 ;!

272 AGES
KINEM ATI C SYNTHESIS O F L l :-.i K O

first three relative poles is sho,vn in Fig. 9-25 and that for the other three
in Fig. 9-26. Referring to the solution of Prob. 9, the next step is the
determination of a cente.r point to he taken as the revolute center A 1 in
position 1. Since there are. an infinity of center points, all located on the
center-point curve, a choice 1nust be n1ade at this ti1ne. Ho,vever, this
choice is not con1pletely arbitrary, for, as in the three-accuracy-point
solution, the final linkage should be of the crank-and-rocker type, whence
the crank length O AA I should be less than the fra1ne length OAO B, Fur­
thermore, since the choice of point A 1 1nade in the three-accuracy-point
synthesis led to a solution having good force-transn1ission qualities, the

_ 1/13•
2

FIGURE 9-26 Function generator y = i:-in ;r, 0 < .r < 90°, four accuracy points ;
construction for relative poles R u, H2,, and R :,.
 GE OMET RIC METHODS-FOUR A C C URACY POINTS 273

A1
OA OB
wJM
��r
. -----....
m� Q,,
vA�
""' ""'

9-27 Function generator y = sin x, 0 < x < 90 , four accuraey points:

°
FIGURE
determination of revolute center B1, using pole triangle and cardinal point A".

new position of the point A 1 should be taken in the vicinity of the previous
one. It is to be recalled, now, that the relative poles R,j themselves are
center points. If one such point can be chosen as the revolute center A 1,
the construction will be sin1plified and 1nore accurate. The choice of the
relative pole R 34 (Fig. 9-26) as the revolute center A 1 see1ns to be a good
one, since it satisfies the above conditions.
The deterinination of the point. B1 that would co,nplete the solu­
tion of the problen1 n1ay be carried out by application of Theorem I to
the relative poles R12 and R1 4, for exa1nple. When this is tried, the point
B 1 is found as the intersection of two lines at a very sn1all angle and the
graphical error is large. A more accurate solution 1nay be obtained in
this case by an alternative n1ethod, that involving the cardinal point

!....- "''i
"' 0

A1
84

96
�
j �
108 OA Oa

\'2.0

PIGl:R•��)-28 Function generator Y = sin x, 0 < x < 90°, with four accu­
racy points.
274 KINEMATIC SYNTHESIS OF LINKAGES

associated with the center point A 1 (Sec. 9-1). A convenient pole tri­
angle may be chosen as R 1 2R HR u (Fig. 9-27), and the problem is to find
a point B whose positions B 1, B 2, B. are equidistant from A 1 • The cardi­
nal point Ao,, is first located by using 1'heore1n III, A 11R1 2R 1 4 = R24R1 2A 1
and Ao0R 2.R 12 = RuR 24A1, and B1 is found as the in1age of A 11 with respect
to R 12RH, thus completing the solution.
The resulting linkage OA ABOs is redrawn in position 1 in Fig.
9-28, and its accuracy is checked in terms of 10 equal rotations of the
crank. The results of this analysis are sun1marized in Table 9-1, from
which it is noted that the error, no,v reduced to approximately 1 per cent
of the range of variation of i/;, doei;; not vanish in the vicinity of the accu­
racy points. It is therefore attributed to graphical error. Further
reduction of this error could therefore be obtained only by 1nore accurate
drafting or recourse to analytical 1nethods.

Table 9-1 ERROR IN FUNCTION GENERATOR, y = sin x, 0 < x < 90o°,

FOUR ACCURACY POINTS

x, <I> - </>., "' - � ..

l>EG
Sin X !/mech Ymech - Sin X
DEG DEG

0 0 0 0 0 0
9 12 9.5 0 . 156 0 . 158 0 . 002
18 24 18.5 0 . 309 0 . 308 -0.001
Z1 36 27.0 0 . 454 0 . 450 -0.004
36 48 35.5 0 . 588 0 . 592 0 . 004
45 60 42.5 0 . 707 0 . 708 0 . 001
54 72 48.5 0 . 809 0 . 810 0 . 001
63 84 54.0 0 . 891 0 . 900 0 . 009
72 96 57 . 5 0 . 951 0 . 958 0 . 007
81 108 59. 5 0 . 988 0 . 993 0 . 003
90 120 60.0 1 . 000 1 . 000 0

9-5 E X A 1\-1 P L E : R E C 1' I L I X E A R R E C O R D E R M E C H A N I S M

The rectilinear recorder problen1 of Sec. 8-8 ,vill be reconsidered
here, this ti1ne on the basis of four accuracy points. The staten1ent of
the proble1n is the sa1ne as before : the range of rotation of the crank,
<t,1 - 4', = 90o° , and the total displace1nent of the pen, s, - s. = 4.5 in.
°
Four accuracy points are taken ,vith equal spacing, c/,1 2 = 30o°, q,13 = 60o,
q,14 = 90o, (/,23 = 30o, t/>2, = 60o, t/>a4 = 30o ; 812 = 1. 5 in., Sia = 3.0 in.,
° ° ° °

Su = 4. 5 , in., 823 = 1.5 in., 824 = 3.0 in., s34 = 1 .5 in.

GEOMET U I C MET H ODS- F O U R A C C U R A C Y POINTS 275
As before, the general 1nethod of approach is first to design a slider­
crank mechanism OAAB in which the displacements of the slider (as
point B) are proportional to the rotations of the crank OAA. A coupler
point C of this slider-crank n1echanisn1 following an approxin1ate circular
arc centered at Oc is then located, after ,vhich the pris1natic pair of the
slider-crank 1nechanisn1 is replaced by a link OcC.
The design of the slider-crank 1nechanis1n is carried out in Figs.
9-29 to 9-31. The construction of the relative poles R12, R1:1, R 1 4 is
shown in Fig. 9-29 and that of the relative poles R2a, R2,, R34 in Fig. 9-30.
A center point must next be chosen to serve as the revolute center A 1 •
Here, as in the last section, any point along the center-point curve would
yield a solution. In order to obtain a solution with good f orce-transmis­
sion qualities and sn1all inertia-as ,vas the case for the linkage previously
obtained in tenns of three accuracy points-the new point A 1 should be
chosen in the vicinity of the old one. Furthern1ore, for the sake of sin1-
plicity, one of the relative poles should be taken as A 1 if this is possible.
The point R 12 which best n1eets the two conditions is therefore chosen as
A 1 ; and B 1 is deduced by application of Theore1n I (as extended to the

<1> 12
2

_ q,13
2
l--+� - cj,H
2

S13
2

!0B (atco)

FIGURE 9-29 Rectilinear recorder, four accuracy points,

construction for relative poles R12, R,�, and R 14•
276 K I N EMATIC SYNTHESIS OF LINKAGES

FIGURE 9-30 Construction for relative poles R23, R2◄, RH for rectilinear recorder,
four aceuracy points.

o.,.,,----- --- r

0/J at co
Jo'IOURE 9-31 Rectilinear recorder, £our accuraey points, construction for
slider-crank n1echanism 0,..-lB.
 G E O M E T R I C l-!ETHODS-FO U R A C C U R A CY POINTS 277
B◄
r- - ,
Parallel to 0.4x

-- - -- -- -- A4

W/h��
-- X
0,4
}'IGURE 9-32 Layout of slider-crank mechanism for rectilinear recorder, four accuracy
points.

slider-crank n1echanisn1, Sec. 8-6). This construction is shown in Fig.

9-31. The resulting 1nechanis1n O,.AB is redrawn in Fig. 9-32, and the
positions of A and B corresponding to all four accuracy points are
constructed.
The second phase of the synthesis, i.e., the determination of the
coupler point C of the slider-crank 1nechanisn1 following an approximate
circular arc, is carried out in Figs. 9-33 to 9-35. The six poles Pi; of the
coupler AB for the four positions corresponding to the accuracy points
are constructed in Fig. 9-33. Note that P12 is at infinity and therefore

8
h2◄ -o-.....: 3:__
�=
- ---- -�-

b14 = b - -+---- - -
- - -----_.::,,,Pu P23

b13 -
B2 -- - - - ------
23

- - - - - - - --- - �
- - - ----= �

h 12 -+- ------------ P12, at 00

FIGURE 9-33 Rectilinear recorder, four ac�uracy points: construction for poles of
coupler with respect to frame.
 278 K I �EMATIC S Y :-JTHESJS 01<' LI :"li K A GES

:c.= c,
�
I
-•·e· I
I

6C1

t·1c;cnE!J-34 Rectilinear recorder, four accuracy points: determination of revolute

eenter C,, using pole triangle and cardinal point CQ .

cannot be used in further constructions. A center point Oc 1nust be

located in the vicinity of the revolute center Oc used in the three-accuracy­
point synthesis : the pole P24 is therefore chosen as Oc. 'fhe corresponding
circle point Ci , in position 1, is located as the in1age with respect to P 1 3P 1 4
of the cardinal point Cu (Fig. 9-34). The latter point is deduced fron1
Oc = P24 by application of Theorem III, a procedure already used in the
last section. The final linkage o... ACOc with its coupler point B is
redrawn in Fig. 9-35 and its accuracy checked in tern1s of 11 positions of
the crank at 9° intervals.

Uniform Scale
I
0 1 2 3B24 5 6 7 8 9 10
2 l.111;1,
0I B• l 3 o.,.
Bi <t
, ,,,
Sc-
,s <1/4,
6
Note: >
Differences between (A3)
the actual positions of B
and the desired positions
are too small to be shown here

FIGURE 9-35 Rectilinear recorder, the link OcC now replacing the slider B; four
accuracy points with equal spacing.
 GEOMETRIC METHODS-FOUR A-CCURACY POINTS 279

9-6 E X A M P L E : A C T U A T I O N O F P R E S S U- R E C Y L I N D E R
A n1echanism is to be designed to actuate a pressure cylinder as
shown in Fig. 9-36. As the piston descends from the starting position,
h, = 10 in. from the botton1, to the final position, h1 = 3 in. fro1n the
bottorn, the gas trapped in the cylinder is compressed, causing an increase
in the force F fron1 an initial value of 4 lb. The problen1 considered here
is to design the actuating mechanism OAAB such that the n1on1ent 11,f on
the crank OAA is constant throughout the n1otion, an ideal gas at constant
temperature being assumed.

Formulation of the Problem

With S the cross-section area of the cylinder, V the volun1e of gas, and
P its pressure,
V = Sh and F = SP
With an ideal gas and constant temperature,
P V = const
or hF = h,F, = h1F1

h, = lOin
F h

h t =3 in

FIGURE 9-36 Principle of actuation of pressure cylinder.

280 KINEMATIC SYNTH ESIS O F L I N KAGES

With the initial force F8 = 4 lb, the final force is

F1 = Fs �; = 4(1/-) = 13.3 lb

and, for any position h bet,veen h, and h1,

40
F = h
With friction neglected, the equilibriun1 bet,veen the force F and
the n1on1ent 111 applied to the crank requires that
M d<J> = F dh
from which d<J> =M!._ dh = 40
M h
dh

Since the n10111ent ill is assu1ned constant throughout the n1otion, this
equation rnay be integrated to yield
_ "'· 40 (" dh 40
<I> = .lf J "• h
= In �
1l1 h.
A value of the crank rotation <J,1 - </,8 corresponding to the total piston
displace1nent n1ust be chosen at this point; 60° is a likely value. Since it
is clock,vise, </,1 - <J,. = - ,r/3, and
40 3
- 1r-3 = - ln -
},1 10
The resulting n10111ent is
]f = 46.0 in.-lb
Substituting this value of J/ into the expression of q, - q,, yields
h
<I> - <I>, = 40 n
46 I 10
fro1n ,vhich h = 10 eH<<1>-<1>,>
Converting <I> - </>. to degrees,
h = l0c <<1>-<1>.> 1-19. s
as shown in Fig. 9-37.

Choice of Accuracy Points

The synthesis of the slider-crank n1echanisn1 OAAB ,vill be carried out

with four accuracy points. The departure of the actual n1on1ent M fron1
its hypothetical value is to be 111ini1nal. Since 111 depends on the velocity
ratio dh/dq, instead of displacen1ents, it is the derivative of the structural
 GEOMETRIC M ET H O DS-F O U R ACCURACY P O I NTS 281
error that must be n1inimized and the accuracy points chosen accordingly
(see Sec. 5-5).
It will be convenient at this point to make use of a dimensionle�s
variable x between the limits - 1 and + 1 instead of the variable q, - q,,,
which varies fron1 0 to - 60°,

X = -( <P ;0 q,8 + I)

Since h as a function of q, - <PB is to be generated with four accuracy

points, the derivative dh/d<t> will be exact at only three points and the
first step is to choose these three points. In Sec. 5-5 these three points
were chosen with Chebyshev spacing because a uniform distribution of the
derivative of the error was desired over the whole interval of generation.
Thus, the points for which dy/dx was to be exact were at x = - 0.866,
0, and +0.866. In the present case, however, the force F varies from 4
to 13.3 lb as q, - if,, varies fron1 0 to - 60°, that is, as x varies fro1n - 1 to
+ 1. A given error in the velocity ratio dh/dq, ,vill therefore result in a
greater error in the n101nent 1.lf when it occurs at x = + 1 than when it
occurs at x = - 1. The points for which dh/dq, is to be exact n1ust
therefore be shifted to the right fron1 the Chebyshev spacing above : they

h, in.-piston displacement

0 2 ' 4 6 8
! !
'l
i
. --+-
-10-� ----,f------1,,C-_O - -J-- �

I
-e
C
0
-.-:::
ro
.
<I>.- - </, -20 _l_ -
-O t---1--'4='-- - ..J--__j'..____ __j

�
C:
�
0
-30-+- --t---l-+-- --+- - -+- - --l
"'I
$
� -�-+---t-"1--+----l---+----l
Cl>
-c

;.. q,� - <I>,

I -50-t-- --iM-- -+- --t---+----l
�

<J>, - <I>, -60 -+-

I

"'· - ct,, -
-70_j____..__
__.__---'-___j___....J

FIGURE 9-37 Actuation of pressure cylinder, desired

relation between piston displacement and crank rotation.
282 KINEMATIC S Y NTHESIS OF LIN K A GES

are chosen here at x = -0.8, 0.1, 0.9. The derivative of the polynon1ial
P,.(x) is then
4
dJx = b(x + 0.8)(x - 0. I) (x - 0.9) - b(x3 - 0.2x2 - 0. 71x + 0.07)
Integration yields

1�4(X) = b (:
4
-
0 2 x3
3 -
0 -i 1 :r 2 + 0.07x + c)
and -3ince the coefficient of x 4 rnust be unity, b = 4 and
P4(X) = .r 4 - 0.27x3 - 1 .42x2 + 0.29.r + 4C
A plot of the variable portion of P4(x) is shown in Fig. 9-:3 8; the intersec­
tion of this curve with a "weighted" parallel to the x axis between the
n1axima and n1inima yields
a1 = - 1 .08 a2 = ·- o.35
The values of <J, - <J,. corresponding to the accuracy points are
therefore (.Fig. 9-37)
</,1 - "'· = 2.4° - "'· = - 19. 5
°

= - 46.0° -64.0°
4'2
q,3 - q,8 q,4 - "'· =

P4 (x) - 4C== x ◄ -0.27 x1 - 1 .42 x 2 + 0.29 x

�--�----- � o.5 ---· - - -1

-1.5 -1.0 0 1.0 1.5

-60.
01 ai 03 04 ' ' X

+ 15 ° I
I '
o· ' -75° q, - q, •
I' I !
I
I'
t" ..
'

l___ -·- · - ··-·- · .. -0.5 - - - -· - - ·-

- -- --·- ·-··- -0.7 - -- - . - ·-- - . . -·-

FIGURE 9-38 Actuation of pressure cylinder, determination of accuracy

points.
 G E OMETRIC M E T H O D S - F O U R A C C U R A C Y POINTS ·283
for which the corresponding values of h are computed as

h1 = 10. 49 in. h2 = 6. 77 in. ha = 3.93 in. h4 = 2.77 in.

Synthesis of the Linkage

The rotations of the crank bet,veen the first accuracy point and the other
three are
cf>12 = -21.9
°
cf> 1 a = - 48.4
°
cf>14 = -66.4°

with the corresponding translations of the slider,

812 = 3.72 in. 81a = 6. 56 in. S14 = 7.72 in.

Relative poles R12, R,a, and Ru are constructed as shown in Fig.

9-39, and one of them, R1 3 , is taken as the revolute center A 1• The cor­
responding point B1 is then obtained by application of Theoren1 I (inte r ­
section of R1 2u and R 14v), with

and

The resulting slider-crank mechanism in the position corresponding to

the first accuracy point is OAA,B,. The complete n1echanism is shown
in Fig. 9-40.

Determination of the Linkage by Analytical Geometry

A. 1nore accurate determination of the points A I and B1 n1ay be obtained
in the present case by application of analytic geo1netry. With respect to
the syste1n OAXY (Fig. 9-39), the coordinates of points R12, R 1 3, and Ru
may be evaluated as

= = .
- I . 86 Ill.
X12
- 812
2
S12 </J1 •
y12 =2 cot 22 = 9 .58 In.
813 .
X13 = XA, = - 2 = - 3 · 28 Jn.
Sia .
</>u = 7 . 30 Ill.
Yt3 = YA, = 2 cot 2
81 4
= -2 .
1:14 = - 3.86 111.
814 </) 14
y14 =2 cot 2 = 5.898 in.
284 KINEMATIC SYNTH E S I S OF LINKAGES

y
--
q,12

--
'1>13

\-�

_ 812
s o %
,.

! I · -· - _ -
s�4
-< -
t . · ·- -·-

08 at infinity
♦ ---.

FIGURE 9-39 Actuation of pressure cylinder, synthesis of slider-crank mechanism.

With 012 the angle between the line R1 2A 1 and the x axis,
- y.-1, - Y
tan
12 -
0 1 2 - ---- -
XA, - Xu
1 • 603 or 012 = 58.05e°
and with 014 the angle between the line R 1 4A 1 and the x axis,
YA, - Y14
tan 01 4 =- - -- = -2.41 9 or 01 4 = H7.54°
X.,1 1 - X14

Calling /31 2 the angle bet,veen the line R 1 2U and the x axis,
/312 = 012 + OAR120B = 47.10°
 GEOMETRIC METHODS-FOUR ACCURACY POINTS 285
and /31 4 the angle between the line R 14v and the x axis,
f3u = a14 + OAR120s = 34.33°
The equations of the lines R 1 2u and Ruv are then, respectively,
y - Y 12 and y -
----'y:..-1 �
-=--- = tan {3 14
= tan /312
X XH
X - X12

Since B1 is the intersection of these lines, its coordinates are obtained by

solving the above equations simultaneously,
= 1 . 08xs + 1 1.58
YB
ya = 0. 68xs + 8.53
from which x8 = - 7.74 in. YB = 3. 24 in.
A force analysis of this n1echanisn1 shows that the greatest devia­
tion of the moment }.If fron1 its desired value occurs when the piston is in
its upper position;e·it is 5.2 in.-lb, that is, 1 1.5 percent of the value of ill.

0.49in.

7.25in.

t--- 3.24 in.7

l ll1
0 B
-=-�- 1------ - �Y
-
�I '

.r <I>, I �
' \�

h,

FIGURE
j �¼
��¼�:t.WJll7J:%;
�;'
· ,-::;;,

9:40 Actuation of pressure cylinder, final linkage. ·

286 KINEMATIC S Y c'/ TH E S I S 0 1<' LIN KA G E S

9-7 EX A M P L E : A P P R O X I M A T E D W E L L L I N K A G E
A dwell mechanism is a device to convert a continuous rotation
into a reciprocating or oscillating motion having a period of rest. The
desired output n1otion VI as a function of the input rotation q, for such a
mechanisn1 is sho\\-·n in Fig. 9-41. The dwell period lies between q,, and
f/,1, and the an1plitude of the output n1otion is VIG• The design of a cam
1nechanisn1 to realize this 1notion conversion is well known ; the main
difficulty lies in avoiding discontinuities in the values of the acceleration
at points q,. and q,1. Not so widely known, however, is the design of a
four-bar linkage to acco1nplish a sin1ilar 1notion conversion. 1'he present
section will consider such a design on the basis of the following specifica­
tions: the input is to be rotation of constant angular velocity, and the
output an oscillation with a 40° drive for 135° of crank rotation, a dwell
for 90° of crank rotation, and a 40° return for the remaining 135° of
crank rotation.

General Approach

A si1nple four-bar linkage with q, as crank angle and VI as follower angle

could give the required oscillation but would not be able to furnish the
period of rest. A more sophisticated n1echanisn1 is required. The
general approach to the problen1 is to start ,,,ith a four-bar linkage and
n1ake use of a coupler curve having an approxin1ately circular arc for 90°
of crank rotation (Fig. 9-42a). If D1 is the center of the circular arc, t,vo
links CD and DOD are then connected to the linkage (Fig. 9-42b) to forn1
a six-bar linkage, the output link beco1ning DOD • Point D will vary little
from position D1 while C moves from C, to C1, and link 00D will be essen­
tially at rest, i.e., at dwell. The dwell becon1es 1nore co1nplete as the arc
C,C1 better approximates a circular arc. A true circular arc is impossible,

- - - - - -- �.,...------,......
40•

0 360°
135• 135°
</)I
90•

FIGURE 9-41 Output vs. input curve for dwell linkage.

 GEOMETRIC METHODS-FOUR A C C U R A C Y P O I NTS 287

C -- Ct
.........
"
///
c, \
)/
/

I
I ---- -./
/
/

__
l /

\ -- /
---;::::. /
Output -----

Perpendicular
D'5 (b)
bisector to D1 D's

FIGURE 9-42 Principle of dwell linkage usrng four-bar

coupler curve.

since the coupler curve is of the sixth order, but very good approximations
rnay be had.
1'he length of link O»D and the position of point O», however, are
not arbitrary. Let C5 be the position of the coupler point when the crank
angle is <f, 5, that is, rotated 135° fron1 t/,1, and Dr, the corresponding posi­
_
tion of D. For this position, link O»D must stop and return toward the
dwell position 00D 1 • Link CD for this position must therefore be normal
to the coupler curve, thus locating point Ds as either D; or D;'. Point O D
n1ust lie on the perpendicular bisector to D1Ds and the angle D10DDs
•nust equal the angle of oscillation of 40°.
The above considerations have shown that the design of a dwell
linkage depends on finding a four-bar linkage with a coupler curve having
an approximate circular arc. The diagrams in the Appendix of Chap. 6
288 K I N E M A T I C SYNTHESIS O F L I N KAGES

/
- - c,- ..........
"'
/ \
c,/ l
I
I --- T"
\
\
--- FIGURE 9-43 Transmission angle
r· /
-y and choice of a coupler curve for
dwell linkage.

show that a nu1nber of such curves P.Xist, and the designer is here presented
with a variety of solutions. The question to be decided is what type of
coupler curve ,vill be n1ost favorable from the viewpoint of force transn1is­
sion. This question n1ay be ans,vered by consideration of Fig. 9-43, in
which 'Y is the trans1nission angle bet,veen links CD and DOv . This angle
must never be zero- ,vhich ,vould 1nean a dead point- and its value should
ren1ain as close as possible to 90° . This requiren1ent excludes the choice
of D�' as Do in Fig. 9-42b and sho,vs that the center D 1 n1ust always be on
the concave side of the coupler curve. Furthern1ore, the angle T'Di 1'"
1nust be as small as possible. The radius of the "circular" arc for a
suitabJe coupler curve n1ust therefore be large co1npared ,vith the width
of th'e curve.

Determination of a Coupler Point

The four-bar linkage (Fig. 9-44) will be assu1ned given here, and the first
step is to obtain an approximate idea of the coupler curve to be used.
Use will be n1ade for this purpose of the curves of Fig. 6-33, whose link
proportions are roughly those of the given linkage. With the linkage in
the posi�ion shown at the beginning of the dwell period, a coupler point
between .4 ·and B and above the line A.B ,vould yield a satisfactory c(1rve.

( AC:-1------lin,�\o�; �
!Oa
!--- --- 2.75 in. ------
�

F'IOURE 0-44 Di1nensions of four-bar linkage to be used u1

design of dwell linkage. Position for first accuracy point.
 G E O M E T R I C M E TH O D S - I<' O U R A C C U R A C Y POINTS 289
An accurate determination of the coupler point wille·. now be
obtained by application of the n1ethods developed in this chapter. ·. Con­
sider four values of <J, in the interval </>1 - <J,. = 90e° with Chebyshev spac­
ing. Let <J,1 = 180e°, and assun1e clockwise rotations; thus, <J,, = 183.4e°,
t/,1 = 93.4°; and
<J,2 = 15,j ,6e° <j,3 = 121.2e° <j,4 = 96.8°
The corresponding poles of the coupler ,vith respect to the frarne are then
constructed (Fig. 9-4;'i ), and the problern is to detern1ine a circle point on
the coupler situated bet,veen A 1 and B1, This is done by locating the

b 12
\
.
Ba B•
'\\ \.B 2 ...Q-"--0-
, �
A4 B 1' , .
\

Note: The construction for only pole P 12

is shown. The other poles
are determined similarly

· r\
\
'

\
'

•'IGURE 9-45 Dwell link­

\
age: determination of op­
P<>site image poles.
290 K I X E M A 1 ' I C SYNTHE818 0 1'' LINKAGES

Portion of
circle-point curve

pl
23

C'1
Circle with
K· center at M'

M', 12.6 ·,n. trorn K'

.
... -
\.
\ '

J.' IGURE 9-46 Dwell linkage: construction of circle points c; and c;'.

image poles P23 1 and P3 4 and making use of the opposite-pole quadri­
1

lateral P12P23 Pa/P 1 4. A portion of the circle-point curve is then con­

1

structed in Fig. 9-46 by the 1nethod of Sec. 9-2. Xote that this curve
passes through points A I and B 1 ; any point of the curve between A and B
is then a suitable coupler point.

Determination of the Coupler Point by Analytic Geometry

As in the case of Sec. 9-6, a n1ore accurate detern1ination of the linkage
may be obtained by application of analytic geon1etry . The location of
the circle point C1 (in position 1) depends on the four poles P12, P2:i 1 . />a, 1 ,
J> u related to successive rotations of the frame v,ith respect to the coupler.
These points will therefore be most easily located in tern1s of a rectangu-
 GEO M E T R I C M E T H O D S - F O U R A C C UR A C Y POINTS 291

fJ•. =
-102.6°

0.4 A 1 = 1.00 in.

A I B1 = 3.00 in.
08 B1 = l.50 in.
OA Ou = 2. 75 in.
A 1 c,e..-...i..=�_....::::,i�wr,
, ----- -------�

FIGURE 9-4i Dwell linkage: determination of crank and fol lower rotations with
respect to coupler.

lar-coordinate systen1 A 1xy of origin A 1, with A 1X lying along .4 1B 1• The

rotations of the crank and follower with respect to the coupler for the four
positions under consideration are found graphically as in Fig. 9-47 or
may be calculated analytically. The coordinates of the poles are then
found (Fig. 9-48) by solving the equations
Yi; = tan a; + a, = tan /3; +
? 13,
- Yi;
X. .
•J 2 .,
X. . - 3 ..,

for x,; and y,;, the coordinates of the poles (or in1age poles) P1 2, P23 , P34 1 1

Pu. Thus, in inches,

,

X1 2 = 3.698 X23 = 3.823 :r34 = 1.824 . X14 = 3.267

Y12 - - 2.275 y 2:1 = -5.,572 y3, = - 1 1.920 Y14 = - 5.099
If, now, we consider A 1x as the real axis and .41Y as the in1aginary

Y lmag

.kPu - P;u
'
2

,,..
,/

--
__ ..
P14
i-­ \ K'
... --
pl
34 '\
(a) (b)

FIGURE 9-48 Dwell linkage: determination of point M', using complex numbers.
292 KINEMATIC SYl'\THESIS O F LIN K A G E S

A 1 H1 = x" = 1.55 in.

=
H1 C1 y"= 0.90 in.
C0C 1 = l.45O
in.

FlGt:RE 9-49 Four-bar linkage with coupler-point curve possessing an

approximate circular-arc segment.

axis in the plane of the coupler, the position of each pole is defined by a.
complex number,
P1 : = X12 + iy12
2
P12
Pn 1 : P2a = X2a + iy2a
Pu t : p34 = X34 + iy34
P14: Pi• =
X14 + iyu

A circle point C1 was found as the intersection of two circles centered at

M' and 1lf" (Fig. 9-46) ; the positions of ll1' and 1'1." are also defined in
terms of complex numbers as

M': m'= a' + ib'

M": m" = a" + ib"

Since M' is on the midnormal to P3, P 14, its position is defined in terms of
1

the complex number

m' _ Pu +
2
Pa4
+ i'k p14 -
2
Pa4
 G E O M E TRIC M E TH O D S - F O U R A C C U R A C Y POINTS 293
where k gives the position of M' on the perpendicular bisector,

(Fig. 9-48b). Separating real and i1naginary parts and taking k = 3.8 as
suggested by the graphical construction (Fig. 9-46) yields

aI = - 2.::>,-1 Ill.
• and b' = -4.16 in.

The radius of the circle is the distance Jf'P a4 (Fig. 9-48b), 1

R' = I p14 ; p3 4 + . 14 -
ik p
2
p34
I
= (1 + ik ) P14 ; p34
or R'2 = 42.04 in.2
The same operatio1�s for point Jl;/", with poles P2a 1P12 and the same value
of k, yield
a" = - 10. 42 in. and b" = -5.77 in.
R" 2 = 187.67 in. 2
The equations of the circles n1ay now be written as

(x - a')2 + (y - b')2 = R'2 and (x - a") 2 + (y - b") 2

= R" 2

c.
--oo-
c,
--.._
Ca
0- --...____
// "'
C2/ \
/ )
CI/ __.-/
c.
1\
--- ---
,,,,
/ ,;�. - v2 -= v a -= v◄
- v1 =
'-... _ _,,,,,,,,,- Co=

D .;

FIGURE 9-50 Dwell linkage: deterinination of revolute

center OD.
294 K I N E MATIC S Y N T H E S I S OF LINKAGES

Solving these equations simultaneously yields two intersections,

C� : X - :�.21 Hl.
y - - ""
I . 24 Ill.
c''.
1 . x" - 1.55 in.
y" - 0.90 in.
The point C�' is the desired circle point.
The re1nainder of the solution, i.e., the determination of points D
and OD , is shown in Figs. 9 - 49 and 9-50.

BIBLIOGRAPHY
Beyer, Rudolf: "Kinematische Getriebe:synthese," chaps. IVo-VI, Rpringer-Verlag
OHG, Berlin, 1953. English translation by H. Kuenzel, "The Kinematic
Synthesis of Mechanisms," McGraw-Hill Book Co1npany, New York,
1963.
F'reuden�tein, F., and G. N. Sandor: On the Burmester Points of a Plane,o1
ASJ1E J. Appl. Jfech., ser. E, vol. 28, no. 1, pp. 41--49, lvlarch, 1961.
1The scope of the geometric methods of synthesis presented in tLis text has
been limited to four accuracy points. Methods using five accuracy points, however,
have been devised on the basis of the so-called Burmester points. If a moving plane n
occupies five positions Il;, i = I, 2, 3, 4, 5, with respect to a (fixed) reference plane 2:,
only a few isolated points of Ilwill occupy positions lying on a circle in plane 2":. These
points, known as Burmester points after the German kinematician who first demon­
strated their existence, are the intersections of the circle-point curves corresponding
to two sets of four positions chosen from the five given positions of n. The determi­
nation of Burmester points by gcon1etric means is no easy task: this paper makes their
analytic deterinination possihle, thus bringin� together the generality of the geometric
1nethods with the ease and accuracy of analytic calculations using modern computers.
 ALGEBRAIC METHODS

OF SYNTHESIS USING
DISPLACEMENT EQUATIONS

10-1 DISPLACEMENT EQUATION OF THE

FOlTR-BAR LINKAGE

Consider a planar four-bar linkage OAABOs (Fig. 10-1). This

linkage is characterized by having four revolutes with parallel
axes, the distances between successive axes being the paran1eters
a1, a2, a:1, a4. The synthesis of four-bar linkages, or the deter­
mination of the four para1neters that will yield an approximation
to a desired function between the input (crank) and output (fol­
lower) angles, has been approached in the last chapters by geo­
rnetric methods. In this chapter, algebraic 1nethods for the
synthesis of four-bar linkages as ,vell as other planar n1echanis1ns
will be considered. Such n1ethods of synthesis are based on dis­
placen1ent equations, i.e., equations relating the input and output
variables of a n1echanis1n in tern1s of its fixed para1neters.

The displacement equation of the four-bar linkage n1ay be

obtained by considering a rectangular-coordinate system OAXY
(Fig. 10-1) with respect to ,vhich the coordinates of A and B
may be written as follows:
296 KINEMATIC SYNTHESIS OF LINKAGES

For A: X2 = 01 COS cf>

.1/2 = 01 sin cf>
For B: Xa = - 04 + Oa cos If
y3 = oa sin y;
Since the distance AB is fixed and equal to a2, application of Pythagoras'
theorem yields
(x2 - xa ) 2 + (y2 - y3) 2 = a2 2

or (a, cos cf> + a4 - aa cos f) - (a1 sin cf> - aa sin y;) = a2

2 2 2

After trigonon1etric si,nplifications this may be ,vritten

A sin f + B cos f = C (10-1)
where1
-4 + cos 'I'"- + aa 2 + a4 2
A = sin c/> =a =a
- cos "-'I' + 01
2
04 - 022
B C
a1 a 2a,ai:
Equation (10-1) n1ay be solved for a displace1nent analysis of the
four-bar linkage; that is, y; is found explicitly as a function of <I> and the
parameters a,, a2, aa, a4. Such a solution is obtained by expressing sin t/1
and cost/tin terms of tan (t/t/2),
. 2 tan (y; / 2 ) 1 - tan2 (y;/2)
sint/t = cos =
VI l + tan2 (t/t/2)
1 + tan2 (t/t/2)
and substituting those values in Eq. (10-1) to get

2A tan � + B ( 1 - tan2 t) = C ( 1 + tan2 t)

or (B + C) tan2t/t - 2A tan YI - B + C =0
2 2
f = A + VA 2 + B2 - C2
from ,vhich
tan2 - - - B+C
For each value of <I> the quantities A, B, C 1nay be obtained and
1 Do not confuse the quantities A, B, and C with the points A and B (Fig.
10-1).

Y3
Output

1 Input FIGURE 10-1 Planar four-bar

a� I
linkage; coordinates of .-t
:C3 :C2
08 0A and B.
 ALGEBRAIC METHODS-DISPLACEMENT EQUATIONS 297
Btor ',// +

FIGURE 10-2 Two solutions of the

equation of the four-bar linkage.

t,vo distinct values of i/; found as

.4 + VA 2 + B2 - C 2
if+ = 2 arctan B+C
= 2 arctan
A - vA2 + n2 - c2 (10- 2 )
i/;- B+C
These two values correspond to the t,vo ways in which a four-bar linkage
1nay be closed (Fig. 10- 2 ).

10-2 CRANK AND FOLLOWER SYNTHESIS:

THREE ACCVRACY POINTS
Consider the problen1 of designing a planar four-bar linkage such
that to three given positions of the crank, defined by angles </,1, q,2, and
c/,a, there correspond three prescribed positions of the follo,ver, i/;1 , f 2, and
'Ya. The solution consists in finding the proper values of a1 , a2, a3, and a4
for three related pairs (<J,1, i/;1) , (<J,2, i/;2), and (</>:,, fa). 1�he procedure is
based on the displace111ent equation 1
K1 cos </> - K2 cos i/; + Ka = cos ( <I> - y,) ( 10-3)
with

This equation "'as deduced fro1n Eq. (10-1) by rearranging the tern1s.
\Vhen ,vritten for three pairs of values, ( </,1 , i/;1) , ( </,2, V/2), ( <Pa, 'Ya), this equa­
tion yields a systen1 of three equations linear ,vith respect to K1, K2, K:.,
K1 cos ct,1 - K2 cos i/;1 + K3 = cos (ct,, - i/;1)
Ki cos cf,2 - K2 cos f2 + Ka = cos (cf,2 - t/12)
Ki cos <J,3 - K2 cos fa+ Ka = cos (</>a - i/;a)
1This is also known as the Freudenstein equation (see first reference u1
Bibliography at the end of this chapter).
298 KINEMATIC SYNTHESIS OF LINKAGES

Tedious third-order determinants n1ay be avoided by first subtracting the

second and third equations from the first, thus eliminating K:,,
K1 (cos</,1 - cos</,2) - K 2(cosi/; 1 - cosi/1 2)
= COS ( 'Pl - V11) - COS ( 'P2 - Y,,2)
K1(cos c/>1 - cos c/>a) - K2(cos i/;1 - cos i/;a)
= COS ( 'Pl - V11) - COS ( q,3 - Via)
and solving the resulting systen1 of two equations with two unknowns;
thus,
K- O = WtWij - W3W4
W2W4 - W1Wr,

in \vhich
W1 = COS cJ,1 - COS cJ,2
W2 = COS 1/;1 - COS i/;2
W3 = COS (,Pt - Y11) - COS (q,2 - V12)
W4 = COS q,l - COS q,3
Ws = COS Vil - COS Via
w6 = cos ( cJ, 1 - i/;1) - cos (q,3 - y;3)
Substituting values of K 1 and K2 into one of the three original equations
yields K 3 as
i = 1, 2, or 3
With the values of K1, K2, and K 3 known, the parameters of the
linkage may be found from the relations

The para1neter a4 n1ay be given a positive but arbitrary value, usually

taken as unity. This parameter merely determines the size of the linkage
and has no effect on the angular relationships.

10-3 EXAJ\1 PLES: FOUR-BAR FUNCTION GENERATORS

WITH THREE ACCURACY POINTS
The design of four-bar function generators, already carried out by
geo1netric 1nethods in Secs. 8-4 and 8-.5, is reconsidered here as an applica­
tion of the three-accuracy-point synthesis developed in the last section.
Example 1 The function y = log x is to be generated in the
interval 1 < x < 2 by means of a four-bar linkage 0.. 1.AB0 8 (Fig. 10-3).
The basic elements of the problen1 are here the same as in Sec. 8-4. The
variables x and y are represented, respectively, by the crank and foJlower
 ALGEBRAIC METHODS-DISPLACEMENT EQUATIONS 299
y-scale, uniform
•
y
Y, x -scale, uniform

at %,
'
a3

�1Pt
tr/tL\.,J- "',/
ui_
Oal
a4

F·1cuaE 10-:3 Principle of four-bar-linkage function generator.

angles <I> and y; through the relations

</> - </>a X - X,,
il<J, ilx
i/1 - t/1, - y - y.
flt/I . fly
The reader is referred to Sec. 8-4 for the details of the formulation of the
problen1 and the definitions of the syn1bols used. Three accuracy points
are taken in the interval 1 < x < 2 with Chebyshev spacing, whence the
corresponding values of the variables x and y are
;r1 = 1.067 y1 = 0.0282
X2 = 1.5 y2 = 0.1761
Xa == 1.933 Ya 0.2862
The ranges of variation of <I> and y; n1ust be selected. They are
chosen as !l</J = flt/I = 60°. The rotations of the crank and follower from
the position corresponding to the first accuracy point to the positions
corresponding to the other two are, with the con1putation carried to , *0
</> 2 - <J, 1 X2 - X1 fl<J, y2 - Y
= 26.0o 'f2 - Vil = I ill/; = 29.4°
YI - Y•
=

Xf - X8

<J,3 - </J 1 = Xa
- XI fl</> = 52.0° ./, ./, - 1/3 - Yi
-
A./, -1 • 40
-
\1'3 - \I'I
,'1; Y,
�\I' = .)
XJ - X, 1!r

With the present method, the angles <J,1 and V/1, crank and follower
Positions corresponding to the first accuracy point, n1ust also be selected
at the start. Choosing q, 1 = 0 and V11 = 0 yields

</> 2 = 26.0° i/12 = 29.4 °

<f>3 = 52.0° t/13 = 51.4°
300 KINEMATIC SYNTHESIS OF LINKAGES

from which W1 =
0.1011 = 0.1285
Wz W3 0.0017
=
W4 =
0.3539 W5 = 0.3766 w6 0 =
grv1ng K 1 = -0.0i'>777 K 2 = - 0.05900 K 3 = 0.99877
With the frame a4 = 1 unit of length, the other three paran1eters of the
linkage are found as
a1 = - 16.95 a.2 = 1.36 a3 = - 17.31
This linkage, ,vith t,vo long links (crank a1 = 16.95, follower
a3 = 17.:31) and two relatively short links (fra1ne a4 = 1.00, coupler
a 2 = 1.36), has poor force-transn1ission qualities and is not an acceptable
solution. In point of fact, application of the Grashof criterion sho,vs this
linkage to have change points; on second thought, this was inevitable,
since both crank and follower had starting angles of 0° . The unsatis­
factory solution was co1npounded fron1 unhappy choices of arbitrary
values-starting angles and ranges of n1otion.
If the idea of the spread associated with the 60° ranges seen1s
desirable and this feature is to be retained, only one alternative exists,
viz., different starting positions for <f, and if;.
A second attempt, in which <f,1 = 45° ( <1>2 = 71°, <f,:1 = 97° ) an<l
i/; 1 = 0° (i/; 2 = 29.5° , i/; 3 = 51.4°) ,vere assumed, with a4 = 1.0, yielded
a1 = - 1.031, a2 = 2.682, a 3 = - 2.310. These linkage proportions are
favorable to force transmission, and the design 1nay be considered as
acceptable, if it is recognized that it is a double rocker. The new linkage,
drawn in position 1, is shown in Fig. 10-4. 1�he negative signs for a1 and

I
\ �</,3 = \97•
51 4°
\t·f
I , •

/�tt:
'1'3- . \ . <!>2= 71•\
B0,1o::::-------a•;;...3 ________o-::Ba/,�: 1. t O
I l 1
=0 �, 'A
, / ·/ l/;-O I
.//1-.J, 2 = 29.5 ° 1
� ;• \ <I> 1 = 45•
a

./ I
, \
• I
// ----J I '
I \
Ai
\

- -- -- --
I/ __j

--
A 2 __ ....-bAa
// -- ---

-- -
/

/
/
/
/
/

- - -
B3 cf.,,--
E'IOURE10-4 Example 1, function generator y = log x, I < x < 2, with three
accuracy points, second attempt.
 ALGEBRAIC METHODS - D I S P LACEMENT EQUATIONS 301
aa are interpreted by considering OAA and OnB as vectors: the angles
ti> and VI define their direction; the paran1eters a1 anda3 define their magni­
tudes and the sense in ,vhich they are to be laid off. A graphical check of
this linkage for the three accuracy points sho,vs that no large error is
present. 1�0 detern1ine the structural error accurately, an analysis must
be carried out by using Eqs. (10-2) developed in Sec. 10-1. The results
of this analysis for values of t/, in the interval <f,, < <f, < <t,1 at 6° intervals
are sun1n1arized in Table 10-1.

Table 1 0- 1 ERROR IN LOG-FUNCTION GENERATOR, THREE ACCURACY POINTS

X (/,, DEG VI, DEG log x Yme<I, Ymeeh - log X

I .0 41.0 -6.1 0 -0.025 -0.025

1.1 47 . 0 2.8 0. 041 0 . 042 0 . 001
1 .2 53 . 0 10.4 0 . 079 0 . 080 0 . 001
1.3 59.0 17.3 0 . 114 0 . 115 0 . 001
1 .4 65.0 23.5 0. 146 0 . 146 0
1.5 71 . 0 29. 4 0. 176 0 . 176 0
1.6 77. 0 34 . 9 0 . 204 0 . 203 - 0 . 001
1 .7 83. 0 40 . 1 0. 230 0 . 229 - 0. 001
1 .8 89. 0 45. 1 0 . 255 0 . 254 -0. 001
1 .9 95.0 49.9 0 . 279 0 . 278 - 0 . 001
2.0 101 . 0 54 . 5 0. 301 0 . 302 . 0 . 001

By taking VI = y;+, the structural error, i.e., the difference betv,een

the values of Ymech given by the linkage and the corresponding values of
log x, is shown in the last colun1n. As expected, this structural error
vanishes at the accuracy points. The n1axi1nun1 structural error, occur­
ring at x = 1.0, is E = - 0.025, or 8.3 percent of the range of variation of y.
With algebraic n1ethods of synthesis and analysis, all quantities
n-tay be calculated to any desired degree of accuracy, which means the
elin1 ination of graphical error. Structural error ren1ains, as does the
mechanical error deriving fro1n 1nachining tolerances and deforn1ations
of links. 1'he evaluation of this 1nechanical error ,vill be considered in
a later section.
Example 2 A four - bar linkage is to generate the function y = 1/x
over the interval 1 < x < 2. The ranges of variation are to be /l<j, = 90°
and /li/; = 90° with three accuracy points having Chebyshev spacing.
'rhe accuracy points are then

X1 = 1.067 X2 = 1.5 X3 = 1.933

302 KINEMATIC SYNTHESIS O F LINKAGES

with associated functional values of

Y1 = 0.9371 Y2 = 0.6666 Ya = 0.5173

Taking 'P1 = 45° , v,,1 = 0 yields

'P2 = 'P l + :t:2 - Xi

:r..r - x.
�cp = 84.0o ¥12 = "1 1 + Y2 - Yi �1',, = - 48.7 0
YI - y.
q,3 = q, I + X3 - .i:1
- - -·
.r1 - · .
x.
A q,
i.J, = 1"'3
£, • oo Via = "11 + y 3 - Y i �if; = - 75.6 0
- YI - y•
fron1 ,vhich the linkage paran1eter.., are found as

a1 = 0.036 a2 = 0.970 aa = 0.056 ,vith a4 = 1.0

In this linkage the crank and follo,ver lengths are very sn1all co1npared
,vith the fran1e and coupler lengths, and another "try" is indicated. We
hopefully choose cp 1 = 4.5°, if;, = 90° . The crank angles ct, corresponding

Bi

A2

FIGURE 10-5 Example 2, function gencr:itor y = 1 /I, I < .r < :2, with three accuracy
points, second attempt.
 A L G E B R A I C M E 'l' H OI> S - D I S P L A CEMEX T E Q "l' .\TIO :\ � 303
B

a ,J,

.t/1
"'"'
FJ<a·aE 1 0-6 N'otation for velority '•.

and acceleration synthesis. o,.

to the accuracy points remain as above, but the follow·er angles becon1e
"1 1 = 90° "12 = 41.3
°
"1 � = 14.4°
With these new specifications the linkage para111eters are found as
a1 = - 0.547 a2 = l.035 a3 = 0.447
This ne,v linkage is shown in Fig. 10-5. A displace1nent analysis
si111ilar to that described in the last exan1ple shows that the 1naxi1nu1n
structural error is E = 0.015, that is, 3 percent of the range of variation
of y.

10-4 C R A N K A N D FOLLOWER S Y N T H E S I S :
A N G U LAR VELOCI T I E S A N D A C C ELE R A T I O N S

The design of a planar four-bar linkage having prescribed angular

velocities and angular accelerations of the crank and follo,ver for a desired
phase, i.e., for specified values of <J, and i/1, 111ay be carried out in a fashion
similar to the foregoing. If, in the four-bar linkage of Fig. 10-6, the crank
angle is q,, the angular velocity and angular acceleration of the crank
are defined by
d<J,
= and
W,t, dt
With a follov,er angle If, the angular velocity and angular acceleration of
the follower becon1e

Wy,
= dif
dt
and

A linkage n1ay be designed such that, when the crank has a specified posi­
tion, angular velocity, and acceleration, the follower will also have a
�Pecified position, angular velocity, and acceleration. In other words, it
1s necessary to detennine the paran1eters a1, a2, a3, a4 such that a given set
of values q,, wtf>, a"' ,vill give rise to desired values of t/;, wy,, at.
This problen1 n1ay be solved by taking the first and second time
304 K I N E M A T I C S Y N 'l' H B S I S O F L I N K A G E S

a , = 3.760

A
FIGURE 10-7 Example of velocity nnd acceleration synthesi�,
completed linkage.

derivatives of Eq. (10-3),

KI cos q, - K 2 cos tJ, + K 3 = cos ( q, - iJ,)
K,w"' sin q, - K 2w-1, sin tJ, = (w<1> - w"') sin (cf, - tJ,)
K,(a<1> sin cf, + w4, 2 cos q,) - K2(a-1, sin tJ, + w"' 2 cos tJ,)
= (a<1> - a"') sin (q, - tJ,) + (w<1> - w-1,)e2 cos (cf, - tJ,)
Solving the last t\vo equations of this systen1 for K1 and K2 and substitut­
ing the values obtained in the first equation to get K 3 yields

where W1 = w<t> sin q,

w2 = w"' sin tJ,
Wa = (w<t> - w-1,) sin (q, - tJ,)
W4 = a<I> sin q, + w,/ cos q,

·ws = a-1- sin 1/t + w-1-2 cos 1/t

w6 = (a<I> - a,J,) sin (<I> - tJ,) + (w<I> - w-1-)e2 cos (q, - 1/t)
and Ka = cos (q, - tJ,) - K1 cos <I> + K2 cos tJ,

As in the case of three-accuracy-point synthesis, the para1neters of the

linkage are given by

Example Design a four-bar linkage meeting the following

specifications:
<f, = - 144° w"' = - :3.0 rad/sec a"' = 0
iJ, = 6,:,o w-1, = 8.0 rad/sec af = 0
1'aking a. = 3.760, * the remaining parameters of the required linkage are
* This value of a, was chosen to a.llow a comparison with a solution gained
from complex numbers (Sec. 1 1 - 1 ) .
 ALGEBRAIC METHODS-DISPLACEMENT EQUATIONS 305
found as
a1 = 1.676 a2 = 2.640 aa = 0.606
The linkage is shown in Fig. 10-7.

10-5 G E � E RALIZATI O N O F T H E SYNTH ESIS J.\,1 E T H O D

B Y L I N EAR E Q UATIONS
The foregoing methods of synthesis of the foura-bar linkage in
terms of linear equations only "'ere made possible by the form of the dis­
placement equation [Eq. (10- 3)]. In this, the three coefficients K 1, K2,
and K 3 are (1) functions of three design parameters a1/a4, a2/a4, a3/a4
and (2) independent of the input and output variables, viz., the crank and
follo,ver angles cp and y;. It is therefore possible to generalize this method
to other linkages for ,vhich the displacement equation may be written in
a form having essent.ially the same properties. Consider a general linkage
in which the number of design parameters is n and for which cp and y; are
the input and output variables, analogous to the crank and follo,ver angles
in the case of the four - bar linkage. If the displacement equation of the
linkage is written in the form
n

i= 1
L K;( n design parameters)G; ( cj,, y;) = F( q,, "1)
then , ,vriting this equation for n pairs of values ( ¢;, "1,-) , j = 1 ,2 , . . . , n,
there ,vill be n equations linear in K 1, K2, . • • , K,.,

K1G1(c/> 1, "1 1) + K2G2(</>1, "1 1) + + K,,G,. ( <1>1, f1) = F( cJ, 1 , "11)

K I G1 ( 'P2, "12) + K2G2( cJ,2, "12) + + K,.Gn( </>2 , "12) = F(</>2 1 "12)
. . . . . . . . . . . . . . . . . . . , . . . . . . . . . . '

+ K,. G,. (cJ,,., y;,.) = F(cJ,,, , f,,)

which may be solved for K1 , K2, • • , K,,. Since these K's are functions
of n design parameters, this leads to a set of n equations from ,vhich thean
design parameters may be found.

10-6 SYNTHESIS O F T H E SLIDE R-CRANK l\.1 E C H Al\' ISl\-I

WITH T H R E E ACCU RACY POINTS
The general approach to synthesis outlined in the last section ,vill
be applied here to the slidera- crank mechanism ( Fig. 10-8). The first step
in the synthesis is to derive a displacement equation taking into account
all possible parameters of the linkage. 1'his may be achieved by ,vriting
- ·- -------... -· ·

306 K I N EMA TIC S YNTH E S J S O F L I N K A G E S

the coordinates of points A and B ,·vith respect to the set of axes O,1.xy as
For A : XA = a1 COS <j,
YA = a1 sin <J,
For B : XB =8
Yoe= a3
and then expressing the distance AB = a2 as
(AB) 2 = (xB - XA)2 + (ys - y,1)
2

or a22 = (s - a1 cos <J,)a2 + (a 3 - a 1 sin q,) 2

This, after manipulation and redu,�tion by means of trigono1netric identi­
ties, becomes
(10-4)
To carry out a three-point synthesis relating the crank position <J,
to the slider positionas, three coefficients K 1, K2, Ka must be defined in
terms of the three parameters a1 , a2, a3 of the linkage. On setting

Eq. (10-4) takes the form

K 1s cos <I> + K2 sin <I> - K3 - s2 (10-5)
which satisfies the conditions of the last section. With the notation of
the last section,
Gi = s cos <I> a� = -1 p=8 2

are recognized as functions of the input and output variables cf> and s but
independent of the design parameters a. The coefficients K, on the other
hand, are functions of the design parameters while independent of the

t--· . _ _ ,. · - · - "
s

10-8 Slider-crank mechanism 8howing pariuneters

FlOUftJo;
and variables used in synthesis.
 :\ L G E B R A I C M E T H O D S - D I S P L A C E M E X T J<;Q U A T I O X S 307
input and output variables. Writing Eq. (10- 5) for three pairs of values
(¢1 , 81), ( ¢2 , s2), and ( r/>a, s3) yields the system
K 1s1 cos <t>, + K� sin r/,,- K� = 8 1 2
K 182 cos </,2 + K2 sin </,2 - Ka = 822
K1 83 cos 'Pa a+ K2 sin q,3 - Ka = 8 3 2

W1 = St COS </, 1 - S2 COS </,2 w, = 81 cos </,1 - Sa cos <l> a

w 2 = sin </,1 - sin 4'2 Ws = sin 4'1 - sin 3 q,

W3 = S1
2
- 82 2
U16 = S1 - 8a 2 2

and K 3 is conveniently given by

Ka = -8, + K1 s; cos q,; + K2 sin <J,; i = 1 , 2 , or 3
'fhe parameters of the linkage now follow as

Example Design a slidera-crank mechanism in ,vhich the slider

(!/!_-
displacement is proportional to the square of the crank rotation, or
s - 8$
= q,8 )2
SJ - s. <Pt - 'Ps (10-6)

where /ls = 8t - s, is the total displacernent of the slider corresponding

to the crank rotation A</, = <J,1 - q,,. The starting and final values <f,,,
s, and <J,1, s1 are unspecified and n1ay be chosen by the designer. The
foilowing values are assun1ed for the present example:
<I>, = 45° s. = 8 in.
<Pt = 105° s1 = 12 in.

The range of variation of the crank is therefore A</, = <J,1 - q,, = 60°,
and the corresponding displace1nent of the slider is /ls = s1 - s, = 4 in.
Three accuracy points are chosen with Chebyshev spacing and yield

</,1 = 4 9° </,2 = 75° <f,3 = 101 °

The corresponding values of 8 are deduced from Eq. ( 10- 6) as

; ("'I - "'•) 2
"'' - "'·
.
St = 88 + ( St - S,) -'--- - = 8 ·02 lll.
s2 = 9.00 in.
sa = 11a.48 in.
308 K I NEMATIC SYXTHESIS O F LINKAGES

I
\
I
I
I
I
I
I
I
d �
A2 Aa
�-I<a·nE HJ-!J Example of synthesis of slider-crank mechanism,
completed linkage.

'fhe values of the coefficients w follo,v as

W1 = 2.93 W2 = - 0.211 W3 = - 1().8
W4 = 7.46 w, = - 0. 227 w6 = -67.8
from ,vhich K1 = - 1 1.7 K2 = - 84.0 Ka = - 189.1
The dimensions of the linkage are found to be
a1 = - 5.8,5 in. a2 = 16.55 in. a3 = 7.18 in.
This linkage is sho"\\'n in Fig. 10-9.

10-7 S Y N T H ES I S O F 'l' H E S L I D E R - C R AN K .M E C H AN I S l\{

WI'rH F O U R A C C u RA C Y P O I N T S
.
'fo extend the three-accuracy-point synthesis of the slider-crank
mechanism to a higher nu1nber of accuracy points, additional design
parameters must be taken into account. In the case of four accuracy
points, either the crank position <J,1 or the slider position s 1 corresponding
to the first accuracy point n1ust be included ,vith the link lengths a1 , a2,
and a3 as a fourth design parameter. The slider position s 1 ,vill be taken
in this section as a design parameter, leaving t/>1 to be chosen by the
designer to n1eet additional requirements. The slidei: position corre-
 A L G E B R A I C METHODS-D ISPLACEMENT ]<_;Q u A T I O N S 309
sponding to any accuracy point n1ay now be defined as 8; = 81 + 811,
where s1; is the slider displa·cen1ent from the position corresponding to
the first accuracy point to points I, 2 , 3, and 4 (811 being always zero).
In these tern1s, the displace1nent equation of the n1echanis1n becon1es

or
2a181 cos </,; + 2a1 8 1; cos </,; + 2a1aa sin <I>, - (a1
2
- al + aa 2 + 81 )
2

= 2S1811 + S1;2

An attempt to apply the general approach to synthesis by linear

equations to the present problen1 ,vould yield four linear equations ( cor­
responding to four accuracy points) with five unknowns,

K 1 cos <I>, + K 2s1; cos </,; + Ka sin </,; - K 4 = K s81; + 81;2

j = 1 , 2, 3, 4 ( 10- i)
in which

Ki = 2a1s1 K2 = 2a1
K3 = 2a1a3 K4 = a1 - 2
a2 2 + a3 2 + 81 2

Such a systen1 of four equations ,vith five unknowns is indetermi­

nate; however, exan1ination of the above definitions of the five K's indi­
cates that son1e of these unkno,vns are related to one another,

or (10-8)

This additional relation, called the compatibility equation, n1ust be added

to the syste1n of four linear equations ( 10-7) to yield a system of five
equations for the five unknowns Ki ,vith i = 1, 2, 3, 4, .5. ·
Since the con1patibility equation (I0- 8) is nonlinear, the above
system cannot be solved by the sin1ple rules involving five linear equa­
tions with five unknowns. I ts solution will be better understood if K"
is denoted as A. The compatibility equation then becon1es
2K1 - K2A = 0 (10- 9)
and the systen1 of equations (10- 7) is re,vritten as
K 1 cos </>; + K 281; cos </,; + Ka sin </,; - K 4 = A81; + 81; 2

j = 1, 2, 3, 4 (10- 10)
This system cannot be solved, since the value of A is unknown; however,
K ,, K , K 3, K 4 n1ay be expressed in tern1s of A. 1'o achieve this, consider
2
310 K INEMATIC SYNTH ESIS OF LINKAGES

Table 1 0-2 SYNTHESIS OF SLIDER-CRANK MECHANISM, FOUR

ACCURACY POINTS

Specifications:
Crank rotations c/> 12, c/>u, c/>u
Slider displacements 812, 813, Su
Crank position corresponding to first accuracy point, ,:/,1
Parameters:

Procedure:
1 . Compute
c/>2 = cf> I + c/>1 2 c/>a = c/>1 + cf>13

2. Solve the two systems of linear equations

l1 (cos c/>1) + [3(sin c/>1 ) - l4 = 0
l1 (cos c/>2) + l2(S12 cos c/>2) + la(sin c/>2) - l1 = s1 2
l,(cos c/>a) + l2(8 1 3 cos </>a) + la(sin cf>3) - /4 = 813
l1 (cos cf>4) + lls14 cos cf>4) + la(sin cf>4) - /4 = 8 1 4
and
rn,(cos c/> 1) + ma(sin c/>1) - m4 = 0
111,(cos c/>2) + m 2(81 2 cos c/>2) + ma(sin c/> 2) - m4 = 81 22
m 1 (cos c/>a) + m 2 (S13 cos c/>a) + ma(sin cp3) - 1114 = 8 1 32
m,(cos cp4) + m,2(8 1 4 cos cp4) + ma(sin <f,4) - m 4 = 8 1 42
NOTE: The coefficients of the l's and m's are the same in both systems; only
the second members differ, allowing the elimination of unknowns to be c•ar­
ried out in similar fashion for both systems.
3. Compute the discriminant A = (m2 - 2l1) 2 + 8m1l2 :
If A < 0, there is no solution.
A = O, "
If 1-1 , = 2l1 - m2 ; so1ut1on
. 1s. unique .
2l2
If A > O' ", = 2l1 - m2 + or " =
v� ,
2l1 - m2 - � .
' there are two
2l2 2l2
solutions, one for each X.
4. Compute for each X:
K
and a1 = -22
81 = 2X
Ka
Ka = Xla + ms aa = K2
a2
= V -
-_
_
a12+_
a_2_
a_+_
81 2 � K4
 ALGEBRAIC METHODS-DISPLACEMENT EQUATIONS 311
the t,vo syste1ns of linear equations
+ l2s1; cos </>; + la sin <f,; - z. =
l1 cos <f,; 81;
m1 cos <f,; + m2s1; cos <f,; + m� sin </>; - m4 = s1;
and 2 J = 1 , 2 , 3, 4

obtained by considering one ter1n of the second member at a tiine. These

systems n1ay be solved to yield l1, l2, la, [4 and m1, m 2, ma. m4 • The
unknowns K 1, K2, K a, K 4 of the systen1 of equations (10-10) n1ay now be
expressed by superposition as
K, = Xl; + m; i = 1, 2, 3, 4 (10- 1 1 )
These values substituted into the co1npatibility equation (10-9) yield
2(Xl1 + m1) - (Xl2 + m2)X = 0
or l2X 2 + (m2 - 2l1)X - 2m1 = 0 (10-12)
in which X is the only unkno,vn.
When the discrirninant of this equation is positive, two values of
X ,nay be found. For each of these values, K1 , K2, K a, K. may be evalu­
ated by n1eans of Eq. (10- 1 1 ) , after which the parameters a1 , a2, aa and
the slider position s 1 corresponding to the first accuracy point may be
found. Thus, if the discrin1inant of Eq. (10-12) is positive, the problen1
has two solutions·1 if the discri1ninant is zero, there is a unique solutionaI·
and if the discriminant is negative the problen1 has no solution. The
complete nun1erical procedure involved is summarized in Table 10- 2.

10-8 CRANK AND F O L LOWER SYNTHESIS:

FIVE A C C URACY POINTS
In the threea-accuracya-point synthesis considered in Sec. 10-2, the
crank and follower angles corresponding to all accuracy points were
specified. If the actual angular positions of the crank and follower are
left unspecified but if instead their rotations with respect to the position
corresponding to the first accuracy point are given, then t,vo additional
design parameters, the crank and follower angles </>1 and "11, may be con­
sidered in the synthesis. Since the number of design parameters is now
five, viz., the three ratios a1 /a4, a2/a., aa/a4 and the two angles q,1 and
¥1 1, a synthesis with five accuracy points 1nay be expected. Let the crank
and follo,ver angles for the five accuracy points be </>; and "1; ( j = 1 , 2, 3,
4, 5) ; then
</>; = </>1 + </>Ii ,vith </> = 0
11

and I/;; = ¥11 + ¥1 1; with "111 = 0

where <f,1,- and f1; are the rotations of the crank and follower relative to
the first accuracy point.
312 KINEMATIC SYNTHESIS O F LINKAGES

\Vhen these values are substituted into the displacement equation

of the four-bar linkage (10-3), they yield a system of equations
ll3 + ll 4
ll4 ll4
----=---+--
ll 12 - a22
--
2

+ </>1;) - - cos (Y11 + Y11;) +

2

-ll3- cos (<!> 1 a1 2a1aa

= cos (<1>1 - i/;1 + </>1, - i/;1;) j = 1, 2, 3, 4, 5 (10-13)
with five unknowns,

'Pt

As in the case of the synthesis of the slider- crank 1nechanis1n ,vith

four accuracy points (Sec. 10-7), tne solution of the above systen1 does
not reduce to that of linear equations ; and a con1patibility equation, of
third degree in this case, 1nust be considered. 1'he solution is lengthy,
but a digital-con1puter progran1 giving a co1nplete solution of the problem
has been written,1 and this progran1 n1ay be used in n1ost cases without
an understanding of the details of the solution perforrned.
Consider the problem of generating the function y = f(x) in the
interval X8 < x < x1 by n1eans of a four-bar linkage OAABOn (Fig. 10-10),
with five accuracy points such that the structural error is minimized. As
usual, the variables x and y are represented by the crank and follo,ver
rotations through the relations
</> - <f>, X - x.
and i/t - YI•_ = y - y,
<f>1 - <f>s X1 - Xs 'YI - YI, YI - y.
where y. = f(x.) and Y1 = J(x1). The ranges of variation fl<f, = <f,1 - <J,,
and t,.YI = Yli - if;, are chosen arbitrarily, and five accuracy points are
1
Freudenstein, second reference in the Bibliography at the end of the chapter.

a� . x,

10-10 Four-bar function generator with five accu­

�'IOU.RE
racy points, showing parameters and variables used in
synthesis.
 A L G F.�B RA IC METHODS-DISPLACEMENT EQUAT I O N S 313
Structural error E

FIGPRE 10-11 Structural-error r.urve.

selected along the curve of y = f(x) for x = x; (j = 1 , 2, 3 , 4, 5 ) in the

interval between x. and x1. The crank and follower rotations from the
position corresponding to the first accuracy point to the positions corre­
sponding to the other accuracy points are then
</>i . = X; - X1
A
</>
and
J Xt - Xa

with Y; = f(x;) and j = 1 , 2, 3, 4, 5.

A complete solution of the problem consists in solving the system
of equations (10-13) to find the parameters a1/a4, a2/a4, aa/a4, <1>1, and
¥11 ; analyzing the linkage to determine the structural error as a function
of x; respacing the accuracy points in order to reduce the structural errore;
and solving Eq. (10- 13) again. This process is repeated until the struc­
tural error is miniznized. Since there are five accuracy points in the inter­
val between x. and x1, the structural error will have the general appear­
ance shown in Fig. 10-1 1a. It will be zero at each accuracy point and will
reach a series of maxima and minima between the accuracy points as well
as at the beginning and end of the interval. For the curve shown in Fig.
10-11, the structural error e34, bet,veen the accuracy points Xs and x., has
the greatest n1agnitude. In respacing the accuracy points, points x;; and
X4 must therefore be brought closer together in order to reduce t34 at the

cost of other maxima or minima. The structural error is 1nini1nized when

the spacing is such that all maxin1a and minima are of equal n1agnitude.
The computer progran1 referred to above considers the complete
proble1n : the input data consist of the function y = f(x) ; the initial and
final values of x to be considered, x. and x1 ; the ranges of rotation of the
crank and follower, A</> and A1/;; and the five values X1, x2, xa, X4, X5 corre­
sponding to the five accuracy points to be used in the first cycle of co1n­
putations. With this progra1n, the four-bar linkages shown in Table
10-3e1 have been designed to generate elementary functions within speci-
1 This is table I from F. Freudenstein, Four-bar Function Generators, Trans.
Fifth Conj. 1lfechanisms, Purdue University, 1958; also ll1achine Design, Nov. 27, 1958.
Reprinted by courtesy of Penton Publishing Companv.
Tobie 10-3 OPTIMUM FOUR-BAR FUNCTION GENERATORS WITH FI VE ACCURACY POINTS

"'"e' Ian. •• z'·' ,, ,.. ,. ,,

IHT"tRVA\. (I t' � 0 5.i 50001<.·0 ;S z S 45usc O$r$ I O S .t $ 1

JIAS<l.t OF tp, Otl1 90 00 00 Jl()

..,,
t1.010.c n, .;,,, oe,: 90 90 90 90
ACCVR.ACY POIN't8:

..,.
1 .0IH678 2.7633673 1.62160-52 0.02781610 1.0130385 0.Ol 18359fl• J 0.03363927l 0.02262943! 0.0293:tMl22 - 0.9462
1.13$5580 21.988�:!b 12.8,2m o. 19327M2 1.1661093 0.lf-03"� l ).2◄917664, 0.21794876 0.w.120613 -0.5259
l.38f27'2 48.226392 'l7.476764 0.49740748 1.4991776 0.◄09'93302 0.54280174 0.64151659 0.l>S821332 0.0492
I.i0049S5 7UI06S 3S.54230'l 0. 70059339 1.7984197 0.7◄◄96432 0.$16361"3 0.818�46 0.8'128244 o G72R
., I.96Zi000 87..50520 H.269088 0 91582616 I. Q7S(i6n 0.9702'll\15 o.976113119 0.97795758 o.9Soi1s50 o 9761
(tt.\li'IC A/f(II,.� �• • 010 A�O
DICUUL8 5U28390 269.70917 33.80'400 - 5 . 1700711 -29. 3206"4 -88.314-667 -85.lnl!OS
N>LLOWlll AHG..,,.
I ,/,1, DlllO
AN"D DU'l'IAl.i:I 259.077'9 76.�709 lU.18!>611 120.21283 211.1'8878 233.S:St.30 44.49U3S 37.6367.l◄ - 53.tii
l,INl P1U)r(1Jlfl0Nl\;
ai/04 -3.23455 1.8348688 -2.11803189 - 3. '994859 -0.3$.fmSO 0.6/M816f6 2.52'335�5 -1.800908' -I.li056207 -0.6102
2.231!5439 7.430«40 0.87854083 1.030508 l.30'!Rij26 3.329592-� 0.9082980� 0.92.53iOO!l 0.6656
o,/o,
Gt/fh 0.84W0216
3.4!Ml87 -0.6�666 8.8866194 3.3�3H5 0.38463Sll-l -0.40061 )612 -0..56694832 1.2737203 1.1orn.11 0.3804
.u,ov1..ut u.uoR, 01:a Ar"IO
ueeuu.1.1 0.1900 0.008 0.0258 0.0161 0.140 0.0673 (HI:? 0.500-5 2 34
IIUTl'UT lftROfl, % O(ITPU1
u.,\·ia:1., 0.()002 0.21 0.042 0.0287 0.0119 0.162 0.467 0 5Gfi f.47
. .\LGEBRAICMETHODS-DISPLACEMENT EQUATIONS 315
y = x2

...-.---..--- -1.0 r-------,--- ---,,

FIGURE 10-12 Symmet­

rical function y = x2• -1 -0.5 0 +0.5 +I

fied intervals. The table shows the accuracy points corresponding to

optirnum spacing, i.e., for rninin1u1n structural error as ,vell as the mag­
nitude of this error. Note that the error varies widely fron1 function to
function and depends on the interval of generation as well. For exan1ple,
function y = :i:2 in the interval 0 < x < 1 is generated v,,ith a structural
error less than 0.075 percent of the output travel. When the interval is
extended to - 1 < x < + 1, the error becornes 4.47 percent, or 60 times
as large. The large structural error in this case is due to the syn1metry
of the function y = x2 in the interval - 1 < x < + 1 (Fig. 10-12). The
four-bar linkage is not suited to the generation of symmetrical functions.
}[echanisms better suited to this purpose will be considered in Chap. 12.

10-9 A N ALYS IS OF MECHAN I C AL ERRORS I N LIN K A GE S

Consider a linkage with n constant pararneters Qi, Q2, • • . , q ,,
transforn1ing a 1notion (as shaft rotation), defined by an input variable
q,, into another n1otion, defined by an output variable f. This linkage
has been designed to generate a given function in a given interval such
that, if the linkage ,vere built to perfection, the maxin1un1 deviation
between the desired function and the function generated by the linkage
would not exceed E,, the mathen1atical or structural error. This devia­
tion is present in all linkages designed by approxinlate synthesis. An
additional error� E , due to deflections of the links, play in the joints,
m

and rnanufacturing tolerances, will inevitably occur in addition to E, in

any actual linkage. This n1echanical error E will now be evaluated in
m

terms of din1ensional variations that n1ay be rnaintained on the values

of the parameters qi, q2, . . . , qn .
The displacen1ent equation of the linkage, relating the constant
paran1eters qi, q2, . . . , q,. to the input and output variables q, and YI,
may be written in general forn1 as
(10-14)
316 KINEMATIC SYNTHESIS OF LIN KAGES

The errors of the constant paran1eters qi, q2, . . . , q,. are assun1ed to be
llq1, llq2, • • • , llqn . For a given value of the input variable </,, the value
of the output variable will be y; + ily;, and we may write

F(q1 + ilq1, q2 + ilq2• . .a . , q,, + ilq,. , </>, t/1 + ily;) = 0 (10-1 5)

The mechanical error manifested at the output is

For small values of the errors llq 1, llq2, . . . , llq,, and fly;, the
differential of the function F n1ay be ,·ritten in terms of its partial deriva­
tives as

or Em =

1'he total n1echanical error E in the linkage is therefore the sum of the
m

individual errors due to each of the parameters considered separately.

10-10 M E C H A N I C A L E R R O R S I N F O "C R - B A R L I N K A G E S
For our present purpose, the displacen1ent equation (IO - I) of the
foura-bar linkage is wTitten as

D sin t + J,; cos t = F (10-1 6)

in ,vhich D = 2a1a3 sin </,
E = 2aaa4 + 2a1aa cos </,
F = 2a1a◄ cos q, + a1 2 - a2
2
+ 2
aa + a4 2

Errors !la in the link lengths a 1 , a 2, a3, a4 ,vill n1odify the coefficients D
and E and the term F by arnounts ilD, ilE, and ilF and will produce an
error ily; in the output. Each error ila in a given link will produce sepa­
rate errors in D, E, and F, whence the output error contributed by each
link must be considered separately. The total mechanical error E of the m

linkage ,vill be the sum of the separate errors.

A L G E B R A I C M E'I' H O D S- ·D lSPLACEMENT EQUATIONS 317

link-error Equation
In the presence of link-length errors the displacernent equation (10-16)
n1ay be written as

(D + llD) sin (iJ., + lliJ.,) + (E + llE) cos (iJ., + lli/t) = F + llF (10-17)

After expansion of this equation (by use of trigonon1etric identities, small­

angle approxin1ations, and neglect of higher-order tern1s), the subtraction
of Eq. (10-16), and ordering of terms, we find

ilY1(D cos i/1 - ll sin i/1) = - llD sin 1/; - fl.E cos "1 + !l.F
.
from ·which fl.y; = - llD sin "1 + llb' cos iJ., - AF = (tm)link (10-18)
D COS VI - E S111
. "'

This is the link-error equati<m.

Error Due Only to fl.a1

An error Aa1 in link dimension a1 produces deviations fl.D, llE, and llF in
D, E, and F to yield a mechanical error Em 1 = Ai/11. The deviation AD is
found as follows :

AD = (D + AD) - D = [2(a1 + lla1)aa sin cl>] - 2a 1a3 sin cl>

= (2aa sin cl>) fl.a1

In like n1anner
llE = (2a3 cos q,) /la1
llF = (2a4 cos q, + 2a1) Aa1

Substitution of these values into Eq. ( 1 0 -18) yields

=2 cos ct, + a1 -. aa cos -=-<! 't2 Aa

fl.y; =
a4
E"' 1 1
1
D cos YI - E sin YI

Error Due Only to lla2

The deviations are

and
318 KINEMATIC SYNTHESIS OF LINKAGES

Error Due Only to Aaa

Here
AD = (2a1 sin q,) Aas AE = (2a4 + 2a1 cos q,) Aa3

"·'· = = _ (
AFe= 2as Aaa
a cos cJ, + aa - a1 cos "') ,/\
and 2 4 '-las
.,. - E"
U'f'J Em3
D COS ¥' , ¥'
" Sln .,.

Error Due Only to Aa,4

Here
ADe= 0 AF = (2a1 cos t/, + 2a,) Aa,
and _
".,. - Em4 _ 2 a1 cos t/, + a4 - aa cos if; ua4
- ,\
u.,,4
D cos .,, . .t.
.t. - E sin .,,

Numerical Example Consider the four-bar linkage designed by

the five-point synthesis method to generate the function log x for values
of x bet,veen 1 and 2. The din1ensions of this linkage, given in Table
10-3, are approximately
a1 = - 3. 23 in. a2 = 0.84 in. aa = 3. 48 in. a, = 1.0 in.
with the first accuracy point at
'Pl = 53° Vil = 259°
and a maxin1u1n structural error E. = 0.0037°. The n1echanical error will
now be evaluated at the first accuracy point in ter1ns of errors Aa1 , Aa�,
Aa3, Aa 4 that may be assun1ed for the link lengths a1 , a2, aa, a,. At the
first accuracy point,
=
D = - 18.7 E = -7.04
cos VI = -0.19 sin VI = F 19.64
-0.98
frorn which D cos if; - E sin i/t = - 3.:�8
Em, = - 0.24 Aa1 rad = - 14 Aa1 deg
Em, = 0.50 Aa2 rad = 29 Aa2 deg
Em, = - 0.38 Aa3 rad = - 23 Aaa deg
E,,., = 0.20 Aa4 rad = 12 Aa4 deg

Assuming IAa1I = IAa2I = IAaa l = IAa4 I = 0.001 in.,

I Em I max = IEm, I + IEm, I + IEm, I + JE.,,. J = 0.001 (78) = 0.078°
or IEm lrm, = � + Em, 2 + E ,2 + Em. 2 = 0.001 (42) = 0.042°
m

1'he maximum mechanical error with the 0.001-in. tolerance is thus n1ore
than 20 times the structural error (0.0780/0.0038 '.::::: 20) ; the ratio of the
r,ns error is more than 1 1 tin1es the structural (0.0420/0.0037 � 11).
ALGEBRAIC METHODS-DISPLACEMENT EQUATIONS 319

10-11 GEOMETRIC IN'l' E R P R E TATION OF THE

ERROR D E NOM INATOR
Intuition and experience indicate that the errors in a four-bar link­
age are closely related to the value assu1ned by the angle 'Y (Fig. 10-13),
already identified in Sec. 2-10 as the transmission angle. The role played
by the angle 'Y in the mechanical errors and force transmission of the
mechanism indicates that there should be a relation between this angle
and the denominator, G = D cos "1 - E sin 1/t, in the expressions of the
mechanical errors.
The angle 'Y may be expressed in tern1s of the linkage parameters
and the input variable q, by application of the cosine law to triangles
OA ORA and ABOn (Fig. 10-13),

or

To express G in terms of the sa1ne quantities, the angle y; n1ust be elin1i­

nated by using the displacernent equation of the linkage [Eq. (10-1)].
Solving sirnultaneously the equations

D sin t/t + E cos 1/t =F

- E sin 1/t + D cos "1 =G
F E l D Fl
. t/t =
sin
G D
cos t/t = I - E G l1
yields 1 D E -:-- D-----'-
E
1 -E n -Ee D I

/ Direction of vBA
I ..-----

I
/
,,---------- .--- Direction of v8
+ -
B I-- -

� ·· � A

FIGURE 10-13 Determina­

tion of transmission angle
'}'.
320 KI N E M A T I C SYNTHESIS O F L I N K A G E S

,vhich by substitution into the trigono1netric identity

sin2 y., + cos2 y; = 1
F E 2 E z
D F 2 D
gives
G D
+ - J.J G -E D
Expanding the determinants, collecting tenns, and dividing through by
D 2 + E2 yields
02 = D2 + E2 - F2 (10-20)
Note that, according to this relation, G is the square root which appears
in the expressions of y.,+ and y.,- dP.rived in Sec. 10-1. When G = 0,
i/1+ = y;- and the linkage is in dead-center position, whence no torque can
be transmitted from the crank to the follower.
Introducing the expressions for D, E, and fl' (Sec. 10-10) into Eq.
(10-20) yields
G2 = 4a1 2a32 sin 2 q, + (2aaa4 + 2a1aa cos q,) 2
- (2a1a4 cos q, + a1 2
- a2 2 + a3 + a4 2)
2 2

By combination of ter1ns and algebraic sin1plifications, this reduces to

G2 _
- - 4a2 as 2 2a1a4
cos q, + a1
2
- a2
2
- al + a42 + 4a
2 a3 2
2 2

2a 2aa
or G2 = 4a22a32 ( - cos2 'Y + 1)
or G = + 2azaa sin -y (10-21)
This equation shows that the quantity G and the angle 'Y are indeed two
equivalent expressions of the same property of a four-bar linkage. ·e·eIt
confirn1s the fact that a four-bar linkage having poor force transmission
is also subject to large n1echanical errors.

BIBLIOGRAPHY

Freudenstein, F.: Approximate Synthesis of Four-bar Linkages, Trans. ASJfE,

Yol. 77, pp. 853-861, August, 1955.
- - -: Structural Error Analysis in Plane Kinematic Synthesis, ASllfE J. Eng.
Ind., ser. E, vol. 81, no. 1, pp. 15-22, February, 1959.

� ,i : :
 ALGEBRAIC METHODS
OF SYNTHESIS USING

COMPLEX NUMBERS

11-1 VELOCITY AND ACCELERATION SYNTHESIS

BY COl\.lPLEX NUMBERS 1
The problen1s of synthesis considered in the previous chapter have
been solved by using the displacen1ent equation relating the input
and output variables in terms of design parameters. Values of
the design parameters for which this input-output relation satis­
fied given conditions of motion were found analytically without
further reference to the geometry of the problem. The n1ethod
to be considered here follows the same general pattern, but the
displacement equation to be used will be written in tern1s of com­
plex nun1bers. The use of con1plex nun1bers makes it possible to
consider not only angles and distances, as rotations of cranks or
translations of sliders, but also vectors, to express analytically the
arbitrary 1notions of points in a plane.

In the four-bar linkage OAABOB (Fig. 11-1) the fran1e (link 4)

is stationary, but the other three links (1, 2, and 3) possess
angular velocities w1, w2, and w3 and angular accelerations a1,
1 Much of the material of this section appeared in Machine Design,
Mar. 20, 1958, and is reprinted by courtesy of the Penton Publishing Com­
pany, Cleveland.
322 KINEMATIC SYNTHESIS 01<' LINKAGES

lmag B

-84=11'
"'
I
,··
--· -9" \
.I ' \ ___
__.\..__j...._ .,
08 Real
4
FIGURE 11-1 Four-bar linkage and vector polygon.

a 2, and a3• The problem considered in this section is to find not only the
link lengths a1 , a2, a3, and a4 but also the relative positions of the links
satisfying angular velocity and acceleration specifications.
The four-bar linkage OA�4BOn may be considered to be defined by
four vectors, since four points are involved. In fact, the linkage will now
be taken to consist of a closed vector polygon (Fig. 11-1), for which we
n1ay write
04 + 01 + 02 = 03 (11-1)
or 01 + 02 - 0:1 + 04 =0
Here the relation of one revolute connection to another is given by the
directed distances 0 (vectors).
Since a vector such as 0 n1ay be written 0 = aei8, in which a is a
distance and 8 a counterclockwise angle n1easured fro1n the real axis, the
vector equation of the polygon n1ay be written in con1plex-number form,

The last term may be simplified, for ei.- = -1, ,vhence 0, = -a., and
(11-2)
This equation represents the space relation of the points O..t, A, B, and
0 s, the points of connection between links.
On differentiating with respect to tin1e, setting d8/dt = w, and
ordering terms,
(11-3)
The tern1s are recognized as defining the linear velocities of the points.
This is then the velocity relation and represents the velocity-vector
diagram.
ALGEBRAIC METHODS-COMPLEX NUMBERS 323

A second differentiation yields, after setting dO/dt = w, dw/dt = a

and ordering,
i(a 1 a1 )e'81 + i2(a1w1 2)ei81 + i(a2a2)ei82 + i 2(a2w2 2)e' 82 - i(aaaa)ei8•
- i2(aawa2)e•8• - (a4)Q = 0 (11-4)
Here the factors i(aa) and i2(aw 2) are recognized as associated with the
linear acceleration components of the points. Equation (11-4) is thus
an acceleration relation involving the points; it represents the accelera­
tion-vector diagram.
A more convenient fornt for future manipulation results on dividing
by i and rearranging,
(a1 + iw12)a1ei81 + (a2 + iwl)a�"
2

- (aa + iwa 2)aaei8a - (a4)0 = 0 (11-5)

Assembling Eqs. (11-2), (11-3), and (11-5) as a group and replacing
each aei6 by its vector a yields
101 + 102 - lo 3 + 104 = 0
W101 + W 202 - W303 + 004 = 0 (11-6)
(a1 + iw12)01 + (a2 + iw 22)02 - (aa + iwa2)aa + Oo 4 = 0
This is a system of three hornogeneous equations in four unknowns,
the vectors (or complex numbers) a1, a2, aa, o., in which some of the
coefficients are complex numbers which involve the w and a values of the
links. Since this system, consisting of only three equations, involves four
unknowns, one of the unknowns, a 4 for example, may be chosen arbi­
trarily and the systen1 rewritten as
101 + 10 2 - la 3 = -a•
W101 + W202 - W3 03 = Q (11-7)
(a1 + iw1 )01 + (a 2 + iw2 )a2 - (aa + iwa )aa = 0
2 2 2

The solution may then be carried out by determinants. With

1 -1
-wa (11-8)
-(a3 + iwa2)
the determinant of the system, the unknowns 01, 02, o 3 are expressed in
complex-number form as
-o,[ -w2(a3 + iwa2) + ws(a2 + iw 22)]
D
-04(-wa(a1 + iw12) + w1(aa + iwa2 )]
(11-9)
D
-a4[w1(a2 + iw22) - w2(a1 + iw12)]
oa =
D
324 KINEMATIC SYNTHESIS OF LINKAGES

'fhe difficulty with this solution lies in the con1plexity of the third­
order determinant D. This difficulty, ho,vever, may be overcome if the
arbitrary 04 is taken proportional to the determinant D itself. With

(11-10)

the above values of 0 , 02, 0a becon1e independent of D and are expressed

1

in simple form as shown in Table 11-1. The value of 04, ,vhich still
remains to be found, n1ay then be obtained from (11-1),

(11-11)

In Table 11-1, the links are defined as complex numbers of the

form 0 = c + id, whose real and imaginary parts are themselves defined
by the angular velocities and accelerations specified for the links.
When the vectors represented by the con1plex nu111bers are
assembled in order (see exan1ples), they ,vill define the proportions of a
mechanism having the specified velocity and acceleration values. In
general, the fixed link (in this case 04) will not "corne out horizontal,"
even though the basic sketch (Fig. 11-1) placed it that way; 04 will
generally have an i con1ponent, ,vhich n1eans a position rotation. How­
ever, all links will be rotated by the same amount. 'fhis is a consequence
of having set 04 = - D in the solution of the system of equations (11-6)
instead of setting 04 real as shown in Fig. 11-1.
If the inputs to the definitions of Table 11-1 are incompatible with
physical reality, then no mechanis1n will result in the sense that a vector
of zero length ,vill be given. 1'he definitions cannot be called upon with-

Table 11-1 FOUR-BAR LINKAGE

VECTOR REAL COMPLEX

i COMPONENT LENGTH
FORM
+ d2
OR LINK COMPONENT
d a= yc2
a C a=c+id

a1 f1 = waa2 - w2a, d, = W2Wa(w2 - wa) a1 = c1 + id1 a1 = Vc12 + d12

a2 C2 = W1a3 - W3a1 d2 = waw1(wa - w1) a2 = ,2 + id: a2 = Vc2 2 + d22

as ca = w1a2 - w2a1 ds = w1wlw2 - w1) aa = C3 + -id, 03 = vca2 + a/1

a. C4 = C3 - C1 - C d4 = da - d, - d2 a. = c. + id a4 = Vc42 + d42
ALGEBRAIC METHODS-COMPLEX NUMBERS 325
out some regard for sensible and compatible magnitudes, but poor esti-
1uates are quickly found and revised.
In making entries into the definitions of the components of Table
11-1, it must be re1nen1bered that the positive directions of angular
velocity and acceleration are the san1e as the po$itive directions for the
angles from which they derive; positive is counterclockwise.
Example l To determine the links of a four-bar mechanism that
will in one of its po5itions satisfy the following specifications:
w1 = 8 rad/sec a1 =0
w2 = 1 rad/sec a2 = 20 rad/sec 2
w� = -3 rad/sec Q'.3 =0
Substituting values into the definitions of Table 11-1,
01 = -3(20) - 1(0) + i(l)(-3)(1 + 3)
= -60 - i(12)
a1 = 61.19 units
02 = 8(0) - (-3)(0) + i(-3)(8)(-:3 - 8)
= 0 + i(264)
a2 = 264.00 units
03 = 8(20) - 1 (0) + i(8)(1)(1 - 8)
= 160 - i(56)
a3 = 169.52 units
a = 160 - (-60) - 0 + i(-56 + 12 - 264)
4

= 220 - i(308)
a4 = 378.46 units
The vectors represented by the con1plex nurnbers are shown in
Fig. 11-2a. The 1nechanisn1 is formed by assembling the vectors in
sequence, starting with o 4 (Fig. 11-2b). The proportions of a n1echanis1n
responding to the specified motion characteristics are now on a relative
basis. The mechanism has appeared with a rotation, a consequence of a
1nathen1atical manipulation which is of no importance to our physical
proble1n. The relative lengths of the bars and their terminal points have
been established as functions of the specified w and a values. Needless
to say, the bar a 4 must ahvays be the fixed link. This example may be
compared with the result obtained by a different method in Sec. 10-4.

Dead Points
A dead point occurs ,vhen the follow·er is n1on1entarily at rest just prior to
reversing its direction of rotation, that is, when w3 = 0. With continu­
ously rotating crank, the crank and coupler are either (1) extended in a
straight line or (2) folded over each other into a straight line. Condition
326 KINEMATIC SYNTHESIS OF LINKAGES

i
300
�

100

260

-100

-200

-300

-360
-j
(a) (b)

FIGURE 11-2 Example.

1 calls for 81 = 82, ,vhence d1/d2 = c1/c2 or d1 = d�1/c2. Condition 2

requires that 81 + 180 = 82, for ,vhich d1 = d2C1/c 2 also applies. If

the link values reduce to

Ct = -w2aa + 0
02 = W1a3 + 0
Ca = w1a2 + iw1w2(w2 - w1)
04 = w1a2 + w2a3 - w1aa + iw1w2(w2 - w1)
Example 2 Condition 1, crank and coupler extended in a straight
line. Here w2 will be negative, with a2 and a3 both positive. Consider a
mechanisn1 in which w1 = 3, w2 = -2, w3 = 0, a1 = 0, a2 = ¾, and
aa = 8 (radian-second units). The links are then a1 = 16, a2 = 24,
ca = 8 + i30, and 04 = -32 + i30 units. This mechanism is shown
in Fig. 11-3.
Example 3 Condition 2, crank and coupler folded over each other
into a straight line. Here w2 and a2 will be positive, with aa negative.
ALGEBRAIC METHODS-COMPLEX NUMBERS 327

30 i

20

-20 -10

FIGURE 1 1-3 Dead point : crank and coupler extended

in straight line.

Consider a mechanisn1 in ,vhich w1 = 3, w2 = 1.5, wa = 0, 0:1 = 0,

o:2 = 1.7, and o:3 = -3.7 (radian-second units). The links are then
0 = 5.55, 02 = - 11 . 10, aa = 5.10 - i6.75, and 04 = 10.65 - i6.75
1

units. The n1echanisn1 is sho"'n in Fig. 11-4.

11-2 C O lJ P LE R - C U RVE S Y N T H E S I S :
F I V E A C C U R A C Y PO I N T S

1'he proble1n to be considered in this section is the synthesis of a

four-bar linkage (Fig. 1 1-5a) that is to generate a coupler curve prescribed
by means of the coordinates of specified accuracy points (Fig. 1 1-5b).
As the coupler point passes through these accuracy points, the crank n1ust
rotate through prescribed angles c/>2, c/>a, . . . measured fron1 position 1,
which corresponds to the first accuracy point (Fig. ll-.5c). The design
paran1eters to be used in this synthesis are the link lengths a1, a2, aa, a.;
the coordinates x and y of the point OA with respect to a coordinate system
Oxy; the angle 8 between the line OAOn and the axis Ox; the distance b and
the angle u defining the coupler point to be used; and finally the initial
crank angle <J,1 • A total of 10 design parameters is thus at hand. Since
each accuracy point is given by two coordinates, a n1aximu1n of five accu­
racy points 1nay be specified on matching 10 coordinates ,vith 10 design
parameters.

-10 - i

FIGURE 11-4 Dead point: crank and coupler folded over each
other.
328 K I � E MATIC S Y NTHESIS O F L I N KAGES

y
5

(b) Coupler-curve passing through

(a) Four-bar linkage OA AB08 with coupler-point C five specified points
to be used in synthesis, shown in position l

(c) Crank rotations correspond to

coupler-point positions of (b)

FIGt:RE 11-5 Notation for coupler-curve synthesis.

A solution of this five-accuracy-point coupler-curve problern has

been developed by Freudenstein and Sandor. This solution, making use
of complex numbers, contains such lengthy nun1erical calculations that it
is feasible only ,vhen acco1nplished with a digital co1nputer. Available
on an IBl\f-6,50 progra1n, the solution n1ay be applied to ::i. proble1n ,vithout
a con1plete understanding of the convolutions of the 1nethod. However,
the utility of any n1ethod does gain fron1 an understanding, for, among
other things, it may lead to extensions and different applications. This
chapter will provide understanding by treating the procedures and speak­
ing of the details of the solution of a rather con1plex problen1 of kine1natic
synthesis. As a matter of convenience, one part of the solution-the
reduction of the first pair of co1npatibility equations requiring con1plicated
algebraic 1nanipulation- is treated separately (Sec. 11-3) and may be
on1itted in a first reading.
ALGEBRAIC M E T H O DS-COMPLEX NUMBERS 329

The radius vectors defining the accuracy points are denoted by the
con1plex nun1bers
. - 1 •) -
J - (11-12)
J -, • • • ' :)

referred to an arbitrary coordinate systen1 Oxy in the plane of the fixed

link. The configuration of the linkage in position 1 , its location relative
to the coordinate system Oxy, and the location of the coupler point are
defined by the con1plex nun1bers
k = 1, . . . , 7 (11-13)
as sho,vn in Fig. 11-6.
. ;\rotation of the crank fro1n position 1 to positionj, already defined
as 'Pii ,nay be expressed by the con1plex number

Thus, the product X;z1 denotes the crank in position j. Sin1ilarly, rota­
tions 'Yi and Vii of the coupler and follower from position 1 to position j
may be expressed by the cornplex numbers
and
The radius vector of the coupler point in position 1 (Fig. 11-6) may now
be expressed as a su1n of vectors in two different ways ·as

r1 = Z1 + Z.; + Z1 + Z2 = Z7 + Z4 + Z3

In position j ( Fig. 11-7) the radius vector may si1nilarly be expressed as

c,

-- .... --

FIGURE 1 1 -6 Vectors to be used in coupler-curve synthesis.

330 KINEMATIC SYNTH ESIS OF LINKAGES

0 X

FIGURE 11-7 Displacement o f linkage from position l to position j.

If o; denotes the displacen1ent of the coupler point fron1 position 1 to

position j, the vectors (or co111plex nun1bers) o;, with j = 2, 3, 4, and 5,
are kno,vn. and
o; = r; - r1
or o; = Zt (X; - 1) + z2 (v; - 1) = Z4(µ; - 1) + za(v; - 1)
With five accuracy points, these last equations 1nust hold for j = 2, 3, 4, 5.
Since the vectors o; and the coefficients X; are known, this condition yields
two systen1s of equations. From this point on, the vector designation of
the r, z, and o vectors will be dropped, since it is understood that such
vectors are expressed in terrns of complex numbers. The two systems of
equations are then

(11-14)
(vs - l)z2 + (Xs - l)z1 = Os
and (µ2 - l)z, + (112 - l)za = 02
(11-15)

Each system involves two unknowns, z1, z2 and za, Z4, present in four equa­
tions. Since each system has only two unknowns, solutions will be possi­
ble only if the matrix of each systen1 is of rank 2 and their third-order
characteristic determinants are zero.s1 On the assumption that
V2 - 1 1'2 - 1 ,pf 0
and
"�' -. 1 1'3 - 1
1 See Sec. A-4 for definition of characteristic determinant.
A L G E B R A I C M E T H O D S - C O M P L E X :!li' U MB E R S 331

these determinants may be taken as principal determinants fore_�ch

system. Two characteristic third-order determinants may then be
formed for each system as shown in Sec. A-4. The first system (11-14)
will be con1patible if
V2 - 1 X2 - 1 lh 112 -1 X2 - 1 02
Vs - 1 A3 - 1 03 = 0 and va - 1 >--a - 1 oa = 0 (I 1-16)
v, -1 >--. - I 04 vs - 1 Xs - 1 0&
For the second systen1 (11-15) the conditions of con1patibility are
JJ2 -1 V2 - l V2 - 1
µ3 - 1 J/3 - 1 and J/3 - 1 (1 1-17)
µ, - 1 J/4 - 1 J/5 - 1
The first set, Eqs. (11-16), which guarantee the compatibility of
the system of Eqs. (11-14), are called the first pair of compatibility equa­
f,ions. The solution of these equations yields values of 112, • • • , vs for
\Vhich the syste1n of Eqs. (11-14) may be solved. The last set, Eqs.
(1 1-17), which guarantee the con1patibility of the system of Eqs. (1 1-15),
constitute the second pair of compatibility equations.
Since the known quantities are the >-.'s and o's, the first pair of
compatibility equations (11-16) must be solved first. Note that these
equations involve complex numbers : when their real and imaginary parts
are equated to zero, they yield four equations ,vhich may be solved for the
four unknowns v2, • • • , vr,, which are con1plex numbers of unit magni­
tude. With these values of the v's, the system of Eqs. (11-14) is compati­
ble. After a solution of the first pair of compatibility equations has been
obtained, the corresponding values of 112, • • • , vs 1nay be substituted in
the second pair (1 1-17), which then yield values of JJ.2, • • • , µ5 for which
the system of Eqs. (11-15) may be solved. The desired linkage is then
obtained by solving the systems of Eqs. (11-14) and (11-15).
In order to solve the first pair of compatibility equations, the vari­
ables vs, v4, and v3 n1ust be successively eliminated to leave v2 as the only
unknown. The resulting equation is then reduced to algebraic forn1 in
terms of the unkno,vn

-r = tane-
'Y2
2
This process of elimination and reduction is carried out in Sec. 11-3 and
yields a fourth-degree algebraic equation
(11-18)
whose coefficients bm (m = 0, 1, 2, 3) are real and n1ay be evaluated in
tern1s of the coordinates of the accuracy points and the corresponding;
332 K l .N E M A T I C S Y � T H :l!: S I S O F LIN K A G E S

crank rotation. The solution of this quartic, which must be carried out
by iteration, yields zero, two, or four real roots, from which zero, two, or
four values of 112 may be obtained ; these values are denoted as 1121: with
k = 1, 2 or k = 1, 2, 3, 4. The corresponding values of 113, 114, v,, may then
be obtained fro1n equations derived in the course of the elitnination (see
Sec. 1 1-3).
When one set of solutions v2k, • • • , vr,1c is substituted for 112, • • • ,
11 5 in the det.enninants of the second pair of compatibility equations
(11-17), the proble1n appears identical to that of the first pair. The
coinplex nun1bers X2, . . . , A5 n1ust no,v be replaced by 1121:, • • • , 1161c,
and a sin1ilar process of elin1ination 1:1.nd reduction would obviously yield
a fourth-degree equation similar to (1 1-18), so that four roots may be
expected at n1ost. These roots, however, rnay be found more directly.
Let v2h, • • • , vs,, be a set of solutions of the first pair of compati­
bility equations different from 1121c, • • • , llok• Substituting these values
for µ2, • • • , µ5 in the detern1inants of (11-17) yields

V2Ji - 1 112k - 1 02 112h - 1 112k - 1 02

1131, - 1 1131: - 1 03 and Jl3h - 1 ll3A: - 1 03 (11-19)
J/41, - 1 J/4k - 1 04 115!. - 1 Jlf,k - 1 Or,

We shall show that, under general conditions, the above determinants are
zero. With both sets v2h, • • • , 11,,1,, and 1121<, • • • , 11s1: satisfying the first
pair of con1patibility equations, one set n1ay then be used as solution of
the first pair (11-16) and the other as solution of the second pair (11-17).
In order to establish that the first detern1inant (11-19) is zero, it
will be convenient to consider the con1patibility of a system of four
equations
02U + 03V = 04
(X2 - l)u +
(X3 - l)v = A4 - 1
(11-20)
(112A: - l)u + (113k - l)v = V4k - 1

(112h - l)u +
(1131, - l)v = J14h - 1

v.1ith t,vo unkno,vns u and v. On the assun1ption that

ch �0
Xa - 1

this determinant n1ay be chosen as principal determinant of the systen1 ,

and the characteristic determinants

and
vn - 1 vak - 1 vu - 1 V2h - 1 Jl3h - 1 Jl4h - I
ALGEBRA IC METHODS-COMPLEX NUMBERS 333

are zero since the sets v2k, • • • , vsk and v2h, • . • , vsh satisfy the first
pair of con1patibility equations (1 1-16). The syste1n (1 1-20) is therefore
compatible. Ho,,'ever, the second-order determinant

v2 - 1 1 V2k - 1 Vzh - 1 Vah - 1

V3 - 1 1 V3k - 1 V2k - 1 V3k - 1

is also different fron1 zero. Upon choosing D2 as principal determinant of

(11-20) the characteristic determinants

V2k - 1 Vah - 1 V4h - 1 112k - 1 V3h, -1 Jl4h, - 1

V2k - 1 V3k - 1 V4k - 1 and V3k - 1 Jl4k - 1
02 03 04 A3 - 1 >-. - 1

1nust be zero, since the systern has already been shown to be compatible.
But the first detern1inant above is identical to the first determinant
(1 1-19), which n1ust therefore also be zero. In similar manner, one could
show that the second detern1inant ( 11-19) is also zero.
If the set v2h, • • • , vs,. were identical to v2k, • • • , vi;i., the deter-
1ninants ( 1 1 -19) would also be zero because of identical columns, in which
case, ho,vever, the system of Eqs. (1 1-15) would be impossible. There­
fore, every set of solutions of the first pair of co1npatibility equations is
also a set of solutions for the second pair. The sets used as solution of
the first and second pairs of equations 1nust not, however, be identical.
In su1nn1ary, three cases n1ay be considered, depending on the
nu1nber of real roots in Eq. (1 1-18) :
1 . No real root-the synthesis problen1 has no solution.
2. 1lwo real roots-there are t,,,o sets of solutions to the first pair
of co1npatibility equations, v2k, • • . , vsk, ,vith k = 1, 2. The synthesis
proble1n has t,vo solutions, denoted as (1, 2) and (2, 1), where the first
nu1nber between parentheses denotes the value of k defining the set of
solutions used for the first pair of co1npatibility equations and the second
nu1nber the value of h defining the set used for the second pair of con1-
patibility equations.
3. Four real roots-there are four sets of solutions to the first pair
of con1patibility equations, v2k, • • • , vak, ,vith k = 1, 2, 3, 4. The
synthesis problem has 12 solutions; with the above notation, these solu­
tions n1ay be denoted as (1, 2), (1, 3), (1, 4), (2, 1), (2, 3), (2, 4), (3 , 1),
(3 , 2), (3, 4), (4, 1), (4, 2), (4, 3).
Example' Deterinine the dirnensions of a four-bar linkage to
generate a curve passing through the five points sho,vn in Fig. 1 1-8, ,vith
the successive crank rotations indicated. Note that points C2, C3, C 4,
1
This example is ta.ken from Freudenstein and Sandor.
334 KINEMATIC SYNTHESIS O F LINKAGES

_,,,.,,.,
.,,,,,,✓

,
- - -­ - -- ,�
/ -_,,,.,,,- _,,, ;;,,/
--- -- _
- _M ...,. C•

-
/
/

..... -­ -- - --
.,,,,,,
_- - rI C1
0 .........- - - - -
......... - r1

·,, .........
'
'--"•
5

..................
',

l 1-8 Example of coupler-curve synthesis, specification of crank rotations,

FJGl'.RE
and coupler-point positions.

and Cs lie on a circle centered at the origin of the coordinate systen1 ; the
desired coupler curve should approxi1nate the circle as closely as possible
between these points. With such specifications, the radius vectors of the
five points are defined by their 1nagnitudes and angles r; and 8; (j = 1, 2,
3, 4, 5), and the crank rotations are defined by the angles c/>; (j = 2, 3 ,
4, 5) (see Table 1 1 -2).

Table 1 1 -2 SPECIFICATION OF FIVE ACCURACY

POINTS FOR COUPLER- CURVE
SYNTHESIS

POSITION c/>i, Dt;Q r; = Jr;!, 1N. ()i, DEG

1 ..... 1.0 0
2 117.0 1 . 740 -29.50
3 150.0 1 . 740 - 1 0 . 70
4 191 .0 l . 740 1 0 . 30
5 228 . 0 1 . 740 25. 90
A I, G EBRAIC METHODS-COMPLEX NUMBERS 335

Table 1 1 -3 DIMENSIONS OF LINKAGES (1, 2) AND (2, 1) *

LINKAGE (1, 2) LINKAGE (2, 1)

X COMPONENT y COMPONENT X COMPONENT y COMPONEN•r

Z1 -0.e48357 0 . 21870 -0.e82939 - 0 . e53627

Z2 1 . 15443 0 . 1 2010 0 .96500 - 0 . 04671
Z3 1 . 23473 -0.e61734 0 . 02689 0 . 61689
Z4 0 . 02689 0 . 61689 1 . 23473 -0.61734
Z5 0.e59077 - 0 . 33925 1 . 12601 0. 58253
Z6 - 0 .08030 0. 73745 0 . 93810 - 0 . e66360
Z1 - 0 . 26163 0 . 00045 - 0 . 26163 0 . 00045

• Linkage (l, 2) corresponds to the choice T1 = tan (-rd2), T2 =

tan (,/,2/2); linkage (2, I ) corresponds to the choice -r2 = tan (-y2/2),
-r1 = tan (,/,2/2).

Such a four-bar could, for exan1ple, be designed in view of applica­

tion as part of a dwell linkage as in Sec. 9-7. The crank rotation cor­
responding to the dwell would be 111° in the present case.
The first step in the solution is the computation of the coefficients
bo, b1, b2, ba of Eq. (11-18). This is done in tern1s of the equations devel­
oped in Sec. 11-3. Solution of the equation follows and 1nust be carried
out by iteration, since the equation is of the fourth degree. The four
roots all turn out to be real,

T1 = - 0.046634 T3 = 0.091166
T2 = - 1.1799 T4 = 1.2553

With four real roots, the proble1n has 12 solutions: 2 of these are considered
in 1'able 11-3.
1. Linkage (1, 2), in which r 1 is used as solution of the first pair of
compatibility equations and r2 as solution of the second pair of compati­
bility equations. This linkage is sho,vn in position 1 in Fig. 11-9.
2. Linkage (2, 1), in ,vhich r 1 and r2 have exchanged places; the
resulting linkage is sho"·n in Fig. 1 1-10.
As noted earlier, the con1putations involved in this problen1 are
beyond the practical li1nits of a desk calculator, and an auton1atic digital
con1puter must be used. 1
1 A program for use with an IBl\iI-650 has been written by Freudenstein and
Sandor.
336 K I X E M A T l C S Y :-. T H E S I S 0 1'' L l :X K A G E S

B1

z4

FIGURE 11-9 Example of <;oupler-<'urve synthesis, linkage (I, 2).

Bi

FIGURE 1 1-10 Example of coupler-curve synthesis, linkage

(2, I).
ALGEBRAIC METH O D S-CO MPLEX NUMBERS 337

-- linkage (1, 2)
- - - linkage ( 2 , 1)
- • - third cognate linkage

r1oeRE 1 1- 1 1 Application of Roberts-Chebyshev theorem to

linkages ( 1 , 2) and (2, 1 ).

The relation between the linkages (1, 2) and (2, 1) is ,vorth noting :
the construction (Fig. 11-11) sho,vs them to be cognate (Roberts-Cheby­
shev theoren1). The third cognate, constructed from the rules, traces the
san1e coupler curve as the first t,vo but does not appear fron1 the computa­
tion synthesis, since its crank is not related to the cranks of (1, 2) and
(2, 1); that is, its crank rotation is not related to the spacing of the accu­
racy points. It n1ay be shown that the ren1aining 1 0 solutions also occur
in pairs of cognate linkages, for example, (1, 3) cognate of (3, 1), etc.

11-3 R E D l:' C T I O N O F T H E F I R S T P A I R
O F C O M P A T I B I LI T Y E Q U A T I O N S

Consider the detenninant of the first co1npatibility equation

(11-16); on expansion it yields

"· "·
Xa - 1 l,3 X2 - 1 l>2
(112 - 1)
x. - 1 - (113 - 1)
x. - 1
1
+ ( 114 - 1) X2 -
X3 - 1
l>2
= 0
l>a

or A1 + A2112 + Aa113 + A4114 = 0 (11-21)

with
l>
Xa - 1
= -
1 X2 - 1
A2 = 03
Aa
X2 -
2 A4 = l>2
A4 - 1 04 A4 - 1 l,4 \ >,.3 - 1 03
(11-22)
A1 - - A2 - A 3 - A•
338 KINEMATIC SYNTHESIS O F LINKAGES

When the determinant of the second compatibility equation (11-16) is

expanded in similar fashion, it may be ,vritten
A1
I
+ AI2 112 + AI3113 + d4115 = 0 ( 1 1-23)
>-a - 1 03 >-2 - 1
= .1 3 =
I I 02
with A2 (11-24)
Ar. - 1 0� >-s - 1 05
I I
A� - - A2 - Aa - A4

The coefficients of Eqs. (11-21) and (1 1-23) are cornplex numbers

which may be evaluated fron1 the given data, i.e., fro1n the >./s and the
o;'s. 1,he unknowns are the 11/s, ,vhich depend on four real angles, 1'2 ,
1'a, 1'4, 1's- Since two co1nplex equations are equivalent to four real equa­
tions, and since there are a total of four real unknowns, a si1nultaneous
solution of these equations is possible and will be carried out here by
elin1inating successively the unkno,vns 114, 115, 113•
To eliminate v4, Eq. (1 1-21) is written as

and each nte1nber is 1nultiplied by its o,vn cornplex conjugate. Since

]114l 2 = v4ii4 = 1, this yields
(A1 + A2112 + Aava) (A1 + A2ii2 + Aaiia) = A4A4
or (1 1-25)
in which
Ct = AJ(A1 + A2ii2) and
,vith (1 1-26)
and we note that d, is real. The eli1nination of 11e5 fro1u (l 1-2a) by si1uilar
operations yields
(1 1-27)

and
with (1 1-28)
and, again, d2 is real.
Solving Eqs. ( 11-2,5) and (1 1-27) as a syste1n of t,vo linear equa­
tions shows the uuknowns 11:i and ii3 to be
Id di
l d:
l1 Ct

c2 - d2
J/ 3 = and JI ;1
C2
( 1 1-29)
c1 c1 !
I C1 Ct

i C2 C2 C2 C2 I
A L G EBRAIC METHODS-C OMPLEX NUMBERS 339

However, since lvale2 = vaiia = 1, the coefficients c1 , c2, di, d2 n1ust satisfy
the condition
di Ct I = Ct e1
d2 C2 J C2 e2
or NN + D 2 = o (11-30)
in which (11-31)

and D = C1
(11-32)

When c1, c 2, d1 , d2 are replaced by their definitions (1 1-26) and ( 1 1-28),

the detern1inants N and D become
N = - Aa(A1 + A2ii2)(n' + 3.;A�v2 + A�3.�v2)
+ A�(A� + .3.;ii2)(n + A1A2v2 + A1A2ii2) ( 1 1 -33)
D = Ai(A1 + A2ii2)A;(A� + A;v2) - Aa(A1 + A2v2)A�(.3.� + A;ii2) (11-34)
Expanding and ordering terrns produces
1V = av2 + b + cii2 + dii,/
D = kv2 + 2g11 + kii2
in which a = A1.3.� (AaA; - A�A2)
b = Aa.3.1n' - A�.3.;n + AaA2A�A; - A�A;A1 A 2
c = A3A 2n - A 3A2n + AaA1A1A2 - A 3A1A 1A2
- I I-I - f I f-f

(11-35)
d = AaA2A�.3.; - A�A;A1A2
k = �aA1A�A� - AsA2A;A�
gy = in1aginary part of [AaA�(�1A� + A�A2)]
Substitution of the above values of N and D into (11-30), with expansion
and ordering of terms, produces
aa + bb + cc + dd - 4g112 - 2kk + (ab + be + cd + 4ikg11)v 2
+ (ab + be + cd - 4ikg11)ii2 + (ac + bd + kk)v22
+ (ac + bd + kk)iil + adv23 + adv23 = o (1 1-36)
It may be noted that this last expression is the sum of a real term and of
complex terms which occur in complex conjugate pairs; thus ab + be +
cd + 4ikg11 is the con1plex conjugate of ah + be + ed - 4ikg11, etc.
When a con1plex nu1nber z is added to its con1plex conjugate, their sum
is a real nu1nber equal to t,vice the real part of z, that is,
z + z = 2(real part of z)
Applying this property of co1nplex nu1nbers to the above equation gives
(11-37)
340 K I X E M .-\ T I C S Y X T H E S T S O F L l � K A G E S

where p = ad
q = ac + bd + kk
s = ab + be + cd + 4ikgu (11-38)
t= ½(aa + bb + cc + dd - 4gy2 - 2kk)
Kow, since v2 = e;-r,, vi = e2;-r,, v2 3 = e 3i-r,, Eq. (11-37) 1nay be \\ ritten 1

Pz cos :l-y2 - Pu sin 3-y 2 + q,, cos 2-y2 - qu sin 2-y2

+ Sx cos "Y2 - 811 sin "Y2 +t=0 (I 1-39)
in which P:z , q,,, Sz and p11, q,,, s11 are the real and in1aginary parts of p, q, s.
A solution of the trigono1netric equation (11-39) would yield values of -y 2
for which the system of Eqs. (11-14) has a solution. To solve (11-38),
the trigonometric functions are expressed in tern1s of the tangent of the
half angle, r = tan (-Y2/2). Thus, substitution of the identities
1 - r2 . ?
COS i'2 = J + T2 Sill "Y2 = -T
l + T2
2 + r4
cos 2-y2 = ---- 2 2 ­
1 - 6r . ,
sin 2-y2 =
4r(l - r2)
(1 + r ) (l + r2) 2
(1 - r )(1 - 14T2 + r )
2 4
_ 2r(3 - 10r2 + 3r4)
COS 3-y2 = . 3j'2 -
Sill
(l + r2) 3 (l + r�)a
into (11-39) yields (note that Pz + q,, + Sz + t = 0 since 'Y2 = 0 is a
�olution)
( - p,, + Qx - S:r. + t)r • +
( - 6py + 4qu - 2su)r4
+ (1.5p,, - 5qz - s., + 3t)r 3 + (20p11 - 4s11)r 2
+ (1.5p,, - ,5qz + Sx + :�l)r - 6py - 4qy - 2s71 = 0
or r� + a4r + a3r + a2r2 + a1r + ao =
4 3
0 (11-40)

- 11 +--- 11 - --
4q 2s11
with a41= - -6p
- Pz + +l
qx - Sx
15p:,; - 5q,, - s,, + 3t
aa =
-p + qz; - S + t
:,; :r

=
20pu - 4s11 (11-41)
a2 -- - - - - ­
- p,, + q,, - Sr + t
a1 = ----'-----'C..
l 5pz - .5q"'- +-s,,- +- 3t
- P:r + q,, - s,, + t

-6p - 4q - 2s
-
a0 -
- pz; + qz - Sx + t
y y u

Since this equation is of the fifth degree, it will adn1it five solutions, real
or complex. Observation of the determinants in (11-16), frorn which
(11-40) ,vas derived, shows that a possible solution is v; = 'A; (j = 2, 3,
4, 5), and therefore ro = tan (q,2/2) is a solution of (11-40). It n1ay be
A L GEBRA I C METHO DS-COMPLEX NUMB E R S 34 1

noted, how4iver, that such values of the v2's would make (11-14) impos­
sible, and the solution must be eliminated by dividing (11-40) by T - To
to obtain the 'luartic equation (11e-18) of Sec. 11-2,
T4 + baT3 + b2T2 + b1T + bo = 0 (11-18)
Synthetic division yields the coefficients of this quartic as
ba = a4 + To
b2 = a:1 + a To + To
4
2

(1 1-42)
b1 = 0,2 + a:1To + U4To + r0 2 3

bo = a1 + a2To + aaTo2 + a4To3 + To 4

In sun1n1ary, the solution of the first pair of con1patibility equa­

tions, therefore, proceeds as follo,vse:
1. Con1pute the f ollo,ving quantities:
a. The >.'s and o's fro1n the specifications of the five accuracy
points (see Sec. 11-2).
b. A t, A 2, A 3, A 4, A 1 , A2, A 3, using Eqs. ( 1 1-22) and ( 11-24 ) .
I I I •

c. a, b, c, d, k, Y.v, using (11-35).

d. p, q, s, t, using (11-38).
e. ao, a1, a2, aa, a4, using (11-41).
f. bo, bi, b2, ba, using ( 11-42).
2. Solve the quartic [Eq. (1 1-18)). An iteration process is required
and, when real roots exist, yields either four or t,vo roots Tk ,vith
k = 1, 2, 3, 4 or 1, 2.
3. For each root, i.e., for each value of the subscript k, evaluate
the corresponding i'2 by n1eans of the equation
i'2k = 2 arctan Tk
as well as v2 in tenns of the equation

4. For each root, evaluate the corresponding v3, using (11-26),

(11-28), and ( 11-29).
5. For each root, evaluate the corresponding v4 and v5, using
(11-21) and (1 1-23).
When the quartic has four solutions, this numerical procedure
yields four sets of solutions

k = 1, 2, 3, 4
for the first pair of compatibility equations. When there are but two
roots, then k = 1 or 2 and there are only two sets of v's. When there are
no (real) roots, the synthesis is impossible.
342 KIN E M A T J C S y N T H E S I S O F L I N K A G E S

BIBLIOGRAPHY

Beyer, Rudolf: "Kinematische Getriebesynthese," Springer-Verlag OHG, Berlin,

1953. English translation by H. Kuenzel, "The Kinematic Synthesis of
1fechanisms," McGraw-Hill Book Company, New York, 1963.
Bloch, S. Sch.: On the S�·nthesis of Four-link :\Iechanisms, Acad. Sci. U .S.S.R.,
Bull. Dept. Tech. Sci., no. 1 , 1 940 (in Russian).
Freudenstein, F., and G. Sandor: Synthesis of Path-generating :\-Iechanisms by
Nleans of a Programmed Digital Computer, ASJfE J. Eng. Ind., ser. B,
vol. 81. no. 2, �fay, I 959. Program available from IB:Vl-650 Program
Library, File 9.5.003.
Hartenberg, R. S.: Complex Numbers and Four-bar Linkages, .l[achine Design,
vol. 30, 1958.
Rosenauer, N.: Complex Variable :\fethod for Synthesis of Four-bar Linkages,
Australian J. Appl. Sci., vol. 5, no. -l, 195-l.
 SYNTHESIS OF

SPATIAL LINKAGES

12-1 IXTRODUCTION TO THE STUDY OF

SPATIAL LINKAGES

The problems of kinernatic synthesis considered so far have all

been concerned with planar mechanisn1s, i.e., n1echanisms in
which the moving links, with respect to the fixed link, had planar
motions all parallel to the san1e plane. The positions, displace­
n1ents, paths, velocities, etc., could then be truly projected into
a single plane in \\1 hich the 1notions of the n1echanisms could be
visualized and studied. Thus, a planar 1nechanis1n may be
described v,ith a sin1ple dra\\'ing, and its analysis or synthesis
1nay be carried out with the help of planar geon1etry.

In the study of spatial 1nechanisn1s this ease of representation

and visualization no longer exists. A description of spatial mech­
anisn1s free of an1biguity requires a careful recognition of all
paran1eters and variables involved, and a perspective drawing is
often insufficient and n1isleading. The sy1nbolic notation intro­
duced in Chap. 2 will be en1ployed in the description of spatial
linkages by syn1bolic equations to define the roles played by
the fixed parameters and variables of the linkage. The analy­
sis or synthesis of spatial 1nechanis1ns by geo1netric methods
is greatly complicated by the fact that at least two projections
344 KINEMATIC SYNTHESIS OF LINKAGES

�Zt

Follower

FIOURE 12-1 Two-revolute two-spheric-pair mechanism.

are required to define a geometric quantity such as a point or a line.

Even descriptive geometry yields a satisfactory solution only for the
simpler problems. The synthesis of spatial linkages will therefore be
approached in terms of analytical methods similar to those of Chap. 10.
Such analytical n1ethods are based on displacement equations, which in
the case of planar mechanisms could be obtained by straightforward appli­
cation of analytic geometry. Some spatial linkages may be handled in
similar fashion, and an example of this "direct" method is given in the
next section. In general, ho\\'ever, the co1nplexity of the direct method
is such as to 111ake it in1practical1 and another n1ethod, such as one based
on n1atrix algebra, must be used. A. con1plete presentation of the syn1-
bolic notation and of the ,natrix 1nethod of analysis is beyond the scope
of this book. Only the course of these develop1nents is outlined in Sec.
12-3, and the reader is referred to the bibliography for specific sources
on the n1atrix method.

12-2 DISPLACEMENT AN A LYSIS OF THE 1'WO-REVOL UTE

TWO-SPHERIC-PAIR MECHAN ISi\-1
1'his section is concerned with a spatial 1nechanisn1 with two revo­
lute (or turning) joints and t,vo spheric pairs (or ball-and-socket joints).
This 1nechanisn1a1 shov,n in Fig. 12-1, will be referred to as a 2R, 2G
n1echanis1n. 'fhe following axes are convenient to define the geo1netry:
z1 axis of rotation of the crank, oriented arbitrarily
z. axis of rotation of the follower, oriented arbitrarily
X1 con1n1on perpendicular to z, and z., oriented fro1n its inter�ec­
tion with Z4 toward its intersection with z1
SYNTHESIS OF SPATIAL LINKAGES 345

x2 perpendicular to z1 through the center of G2, oriented arbitrarily

x3 perpendicular to Z4 through the center of Ga, oriented arbitrarily

In terms of these axes, the n1echanisn1 parameters and variables are:

a4 fra1ne length, 1neasured along X1 fron1 Z4 to z1; this distance is

always positive
a4 fran1e angle, 1neasured frorn Z4 to z1 counterclockwise when
looking from the end of X1; this angle may have any value
fro1n 0 to 360°
81 crank axial position, n1easured along z1 f ron1 X1 to x2; this dis­
tance may be positive or negative, depending on the posi­
tion of the crank with respect to the corn,non perpendicular
Xi
84 follower axial position, 1neasured along Z4 f ro1n xa to xi; like
81, this distance may be positive or negative

a1 crank length, measured along x2; this distance n1ay be positive

or negative depending on the orientation of x2
a2 coupler length, always positive
a3 follower length, measured along x3 ; like a1 , this distance may
be positive or negative
ff, crank angle, measured fro1n X1 to X2 counterclock,vise when
looking fron1 the end of z 1 ; this angle 1nay have any value
from O to 360°
YI follower angle, 1neasured fro1n X1 to x3 counterclockwise when
looking from the end of 24; this angle 1nay have any value
from O to 360°

A displacement equation relating ff, and YI may be found in much

the san1e way as for the four-bar linkage (Sec. 10-1). The coordinates
of points G2 and G3, centers of the spheric pairs, will be written in terms
of the fixed coordinate system 0 1x1 y 1z 1 (Fig. 12-2a). The coordinates of
G2 are then

Xia, = a1 cos </,

Yia, = a1 sin <f,
Zia, = 81

The coordinates to G3 will first be expressed with respect to the systen1 of

axes R4�11z,, where axis R 4 � is parallel to x1 and axis R 4'17 is perpendicular
to both R4� and z,, ,vhence

Eo, = a3 cos 1"

110, = aa sin YI
zw, = 0
346 KINEMATIC SYNTHESIS 01" LINKAGES

Xz

/Ga

'11
(a) Coordinates of points G 1 and G 2

Y1
(b) Projection in planeO1 y 1 z 1

FIGURE 12-2 Displacement analysis of the 2R, W

mechanism.

In Fig. 12-2b, the coordinates of Ga with respect to 0 1X1Y1Z1 n1ay now be

written

X ia, = aa cos VI - a4
Yio, = Projoy, (01R4) + Projo111 (R4G�) - -s• sin a4 + aa sin VI cos a4
z10, = Projo,, (01 R4 ) + Projo,, (R.G�) - -S4 cos a4 - aa sin VI sin a4

The distance G2Ga, which must be equal to a2, may be expressed in tenns
SYNTHESIS OF SPATIAL LINKAGES 347
of the coordinates of G2 and G3 as
(x10, - X10,)2 + (y102 - Y10,) 2 + (z10, - z1c,)2 = (G1G2) 2
(a1 cos <I> - a3 cos i/t + a4) 2 + (a1 sin <I> + S4 sin a4 - aa sin VI cos a4)2
+ (s1 + S4 cos a4 + aa sin i/t sin a 4) = 2
a2 2

By application of trigono1netric identities and ordering of tern1s this

becomes
A sin VI + B cos i/t = C (12-1)
1 sin a4
with A = .
- COS a4 Sill <p + 8-- -­
a1
a
B= - -4 - cos <I>
aI
a1 2 - a2 + a3 + a4 + s12 + S42 + 2s1S4 cos a4
2 2 2
(12-2)
C= 2a1aa
a4 84 sin a4 . ,1,.
- - COS ,1,.
'I' - ---- Sill -,,
as aa

We note that this equation is of the sa1ne form as that obtained for the
four-bar linkage in Sec. 10-1 and that it reduces to that equation if the
paran1eters s1, S4, and a4 are zero.

12-3 AN OUTLINE OF THE l\lATRIX :METHOD

OF ANALYSIS
In brief, the matrix method of analysis is based upon the use of four
paran1eters needed for the geometric description of a connection between
links. Since a 1nechanism 1nay be viewed as being a sequence of con­
nections (Chap. 2), a mechanism n1ay be described by means of a sym­
bolic equation whose terms contain the paran1eters defining the successive
connections. On replacing each term by a matrix also formed from the
paran1eters, an algebraic manipulative procedure suited to analysis
becon1es possible. Because of the close relation of sorne of the definitions
of Chap. 2 to the 1natrix method, the more pertinent ideas of that chapter
will be reviewed here.
'fhe 1notions of the parts (links) of a mechanis1n are enforced
(constrained) because of the geometry of the connections between adja­
cent parts. It is convenient to visualize each connection between two
parts as consisting of two contacting elements ,vhose relative n1otion may
be described n1athen1atically. Taken together, the t,vo ele1nents form
the pa-ir of the connection, the connection itself being characterized by
a pair variable able to describe relative motion between the two elements
of the pair. A.. connection per1nitting only rotation of two parts may be
constructed in n1any ways, but it is convenient to picture this as a journal
348 KINEMATIC SYNTHESIS OF LINKAGES

bearing, for then one element is the surface of the shaft (journal), while
the other is the surface of the bearing; the rotation is uniquely described
by an angle, the pair variable. Two or n1ore elen1ents properly placed on
a n1achine part constitute a link, and an assen1blage of links constitutes
a kinematic chain. Such a chain is closed ,vhen all elen1ents are connected;
it is otherwise open. If all links are connected to two and only two other
links, the chain is simple-closed; if some links are connected to n1ore than
tv.o other links, the chain is compound-closed.
The relative motion between pair elen1ents-and hence the relative
motion between the links bearing the elen1ents-is described by pair vari­
ables such as O for rotation and s for translation. Such variables, singly
or in combination, are used to define the so-called six lower pairs-the
revolute, prisn1atic, screw, cylindric, spheric, and planar. Both the revo­
lute and pris1natic pairs are litniting cases of the screw pair; the first cor­
responds to a screw having zero lead, the second to a screw of infinite lead.
Each is located in space by defining the positions of its axis.

Symbolic Equations
The syn1bol for a scre,v pair is SL, where Lis the lead; the syn1bols for the
revolute and prisn1atic pairs are thus So and S00 but it simplifies the nota­
,

tion to designate those pairs by Rand P, respectively. The two elements

of a pair are distinguished by + and -: thus SL+ n1eans the element giving
the surface of the screw, or full body, and SL- 1neans the elen1ent that is
the nut, or hollow body.
A description of a sin1ple chain involves a description of the rela­
tive positions of the successive pair axes, and this 1nay be done by the use
of the unique co1nn1011 perpendicular between successive pair axes. The
manner of deseribing the relative positions of successive pair axes is given
by Fig. 12-3, which depicts two links of a spatial chain. The two links
are connected by a revolute pair R1, and the revolute elements Ro+ and
R2- show how connections are 1nade to the rest of the chain. The axes
zo, z1, and z2 of arbitrary orientation define the positions of the revolutes
chosen in the illustration, but they would equally ,vell define the positions
of screw or prisn1 pairs. The axis x1 is chosen as an extension of the con1-
mon perpendicular of length a0 bet·ween z0 and z1 oriented fron1 Ho to 01.
1'he axis Y1 is chosen to give a right-handed rectangular syste1n. Simi­
larly, X2 is the extension of the conunon perpendicular of length a1 between
z1 and z2 oriented fro1n H I to 02, ,vith Y2 coinpleting a right-handed
systen1. Then,

81 = (01X1, H1X2) a1 = (01z1, 0222)

s1 = 0,H1 a1 = H102
SYNTHESIS O F SPATIAL LIXKA GES 349

1-'IGURE12-3 Two links of a spatial mechanism, with definition of axes and param­
eters used in the symbolic notation.

A rectangular cartesian systen1 of coordinates is thus defined in

each link, and the relative positions of successive links are expressed in
terms of the paran1eters a, a, 0, andes defining the relative positions of the
successive systen1s of coordinates. "f he set of parameters needed to define
a simple-closed chain is assernbled by means of a general symbolic equation
of screw pairs,
an
. . an
SL. I (12-3)
On
81r s,.
350 K I N E M A T I C S Y N T H E S IS O F L I N K A G E S

'fhe symbol I is used to indicate that the kinematic chain is closed; later,
in the matrix method of analysis to be outlined, it plays the same role
mathematically.
For a scre,v SL. (Lk � 0 or oo ) both parameters fh and s,. vary,
being related by the lead as

A.Ok =
-
Ask
21r L,.

Either 81i: or Sk n1ay be designated as the pair variable.

For a revolute pair (Lk = 0) only the parameter 81r. varies; for a
prismatic pair (L" = oo ) only the parameter s1c varies. 'fhe symbolic
equations of specific mechanisms co1nposed of only revolute and prismatic
pairs are not so fonnidable as Eq. (12-3).
Examples Three linkages of four revolutes are known to have
1notion with one degree of freedon1, F = 1 . They are the planar four­
revolute mechanism (Fig. 12-4), where the four revolute axes are parallel;
the spherical four-revolute mechanism (Fig. 12-5), where the four revolute
axes intersect at one point (the Hooke joint is a special case of this n1echa­
nism)a; and the Bennett mechanism (Fig. 12-6), where the axes are skewed
with prescribed geometric relations. These mechanis1ns will serve as
examples to illustrate the use of symbolic equations.
In the planar four-revolute mechanis1n (Fig. 12-4), the z axes, or
revolute axes, nun1bered as shown, are all oriented in the sarne sense. The
successive con1n1on perpendiculars define xi, x2, :r3, X4, and four y axes not
shown would complete four systems of rectangular axes. Kote that

IB--,---a ,-----Ea,...;��-
R•�;��=====-=r-====�2
1

FIGl:RE 12-4 Four-revolute planar mechanism (four-bar

linkage) showing axes used in symbolic notation.
SYNTHESIS OF SPATIAL LINKAGES 35 1

01x1y1z1 is fixed in link 1, 0 2X2Y 2Z2 is fixed in link 2, etc. The symbolic
equation is
a1
I a2 i
a-1
•
I
I
I a40 I1
0 1 0 0
R1 R2 ;
. 8 2 Ra 8a
. R4 = I (I 2-4)
81 84 i
0 0 0o1 oj
Comparing it ,vith the general fonn of Eq. (12-3), the four pairs are revo­
lutes R1 , R 2, Ra, R4. The paran1eters a1, a2, aa, 0.4, or link lengths, are
the distances bet,veen z axes along the con1n1on perpendiculars. The
angles a are all zero, since the z axes are parallel. 1'he angles 81, 02, 83 ,
84 between successive x axes are the pair variables of the revolutes. Fi­
nally, all distances s are zero, because successive x axes have been chosen
to intersect.
The sy1nbolic equation of the spherical four-revolute 1nechanis1n
(Fig. 12-5) is
!I o 0 0 1o0
R2 Ra 0:' 3 R4 I
a1 a2 I a4
R1 (12-5)
I 81 82 83 0.
lo 0 0 0
In this all z and x axes intersect, whence all parameters a and s
are zero, as shown in the syn1bolic equation. The angles a define the link

FIGURE 12-5 Four-revolute spherical n1echa.nism.

352 KINEMATIC SY NTHESIS O F L I N K A GES

%�
I

--

-- - - -
--- -
FIGURE 12-6 Bennett mechanis1n (see also Fig. 3-9).

dimensions, and the 8's are as usual the pair variables of the revolutes.
The Hooke joint (universal joint) is a special case of this type of n1echa­
nism, with a2 = a 3 = a4 = 90° ; its symbolic equation is
0 0 0 0
(12-6)

The Bennett n1echanisn1 (Fig. 12-(j) is characterized by opposite

links of equal length and twist (paran1eters a and b, a and /3) and succes­
sive common perpendiculars (x axes) which intersect, so that the s paran1-
eters are all zero. The 8's are still the pair variables of the revolutes, and
the symbolic equation is
a b a b
a {3 a (3 I (12-7)
R1 R2 Ra R4
81 82 83 84

1 0 , 0 0 0 1
with the condition that
a (12-8)
b
+
sin a sin (3

Matrix Method of Analysis

The principle on ,vhich the matrix 1nethod is based may now be indicated.
Once a linkage has been described by means of a sy1nbolic equation, a
S Y XTHESIS OF S P AT I A L LINKAGES 353
rectangular-coordinate system is defined on each link. A change of coor­
dinates or linear transformation between adjacent systems, as from
02x2y2z2 to O 1X 1Y1Z 1 (Fig. 12-3), may be represented by the 4 X 4 matrix
:1 1 0 0 0
i
a 1 COS 81
l cos 81 - cos a1 sin 81 sin a 1 sin 81
A 1 == 1 a sin 8, (12-9)
1 sin 8 1 cos a 1 cos 81 - sin a1 cos 8 1
Il St 0 sin a 1 COS a 1

involving the paran1eters a, a, 8, and s of the syn1bolic notation. In the

case of each of the mechanisms considered above, there are four coordinate
syste1ns as sho·wn in the diagram (Fig. 12-7), and each change of coordi­
nates bet,veen adjacent systems may be represented by a n1atrix A involv­
ing the proper paran1eters a, a, 8, and s. When changes of coordinates
are taken in succession, the corresponding matrices are multiplied ; thus,
the change of coordinates fron1 O3x3y 3z3 to O 1 x 1y1z1 is represented by the
matrix A 1 X .4 2- When all four changes of coordinates are taken in suc­
cession, starting fro1n 01.t1y1z1 , going around the chain, and returning to
O 1x1 y1z1 , the 1natrix of the resulting change of coordinates is ,4o1 X A 2 X
A a X A4. But this change is an identity transforn1ation because of
return to the original system, whence
(12-10)

Coordinate system
02 % 2 Y2 Z2 ,
fixed on link 2
Matrix A 1 , Matrix A 2,
function of function of
o 1 , a 1 , 81 , s 1 a2, a2, 82 , s2

Coordinate system Coordinate system

01%1Y1Zl' 0 3 % 3 Y3Z 3 ,
fixed on link 1 fixed on link 3

Matrix A 4 , Matrix A3 ,
function of function of
a 4 , a4 , 8, , s 4 Q3,a3,83 , S3
Coordinate system
04 %4 Y4 Z 4 ,
fixed on link 4

12-7 Closed loop of coordinate changes among the links of a

FIGl:'RE
simple-closed mechanism.
354 K I N EMATIC S Y .N T H � S I S OF LINKAGES

in which I is the unit matrix

1 0 0 0
0 1 0 0
l= (12-11)
0 0 1 0
0 0 0 1

The symbol I, here associated with a sequence of transforn1ations around

a closed chain, is the I of Eq. (12-3) that indicated chain closure.
All displacement equations relating the pair variables 8 n1ay be
extracted from the above n1atrix eql:.�tion without any further geon1etric
considerations: the kinen1atic problem has been forn1ulated in purely
algebraic form. As an example, the displacement equations of the Hooke
joint will be derived from the matrix equation (12-10). With the values
of the para1neters given in the symbolic equation (12-6) of the Hooke
joint the rnatrices A 1, A 2, .4 a, A • are

1 0 0 0
0 cos 81 - cos a1 sin 81 sin a1 sin 81
Ai =
0 sin 81 cos a1 cos 81 - sin a1 cos 81
0 0 Sln a1 cos a1
1 0 0 0
0 cos 82 0 sin 82
A2e=
0 sin 82 0 - cos 82
0 0 1 0
i 1 0 0 0
A a e=
l o cos 83 0 sin 8a
lo
sin 83 0 - cos 83
1 0 0 1 0
1 0 0 0
0 cos 8, 0 sin 84
A4e=
0 sin 8• 0 - cos 84
0 0 1 0

To reduce the nun1ber of n1atrix products involved, both sides of

Eq. (12-10) are n1ultiplied by the inverse 111atrix A 1- 1, thus,

1
A1- X Ai X A2 X Aa X A4 = A1- 1 X I
or A 2 X A a X A4 = A1- 1 (12 -1 2 l
S Y N T H E S I S O F SPATIAL LINKAGES 355

since A 1-1 X A 1 = I. The inverse matrix A 1- 1, which in this example

may be obtained by simply interchanging rows and columns in the matrix
A1, is
I 1 0 0
- cos 81 sin 81
A 1- 1 -
0 0
0 - cos a1 sin 81 cos Cl'.1 cos 8 1 Sin a1
0 sin a1 sin 81 - sin a1 cos 81 cos Cl'.1

When the 1natrix products in the left 1nember of (12-12) are carried
out, the matrix equation of the Hooke joint becomes

12I 13
'
1 l ll 0
· -----· 0 0 -
II
14
-
0 cos 82 cos 8:1 cos 84 cos 82 sin 83 - cos 82 cos 8a sin 84
21 1 + sin 8 sin 84 22 - sin 82 cos 8• 24 1
1
23
··--. 2
0 sin 82 cos 8:i cos 84 sin 82 sin 8a - sin 82 cos 83 sin 84
31 - cos 82 sin 84 32 33 + cos 82 cos 84 34

0 41 sin 83 cos 84 42 - cos 83 43 sin 83 sin 8, 44

11 11 0 12 0 13 0 14

0 21 cos 8 1 22 sin 81 23 0 24
(12-13)
0 31 - cos a1 sin 81 32 cos a1 cos 81 33 sin a1 34

10 41 sin a1 sin 81 42' - sin a1 cos 81 43 cos a1 44

Corresponding elements of the rnatrices in both 1nen1bers of this equation

1nust be equal. If 81 is the input (known) variable, we seek relations
giving 8 2, 83, and (}4 in terms of 81. Equating the ratios of elements 33
and 23 in both matrices yields 8 2 in terms of (Ji as

(12-14)

Equating elements 43 in both 1natrices gives 83 in terms of 81 as

cos 8a = sin a1 cos 81 (12-15)

and the ratios of elements 44 and 42 gives e. in terms of 81 as

tan 8� = tan a11sm 81 (12-16)

It may be verified that, with values of 82, 8a, and 84 satisfying Eqs. (12-14)
to (12-16), all corresponding elements of the matrices in Eq. (12-13) are
equal.
356 KINEMATIC S Y N T H E S I S OF' LIN K A G E S

12-4 SYNTHESIS O F F O VR-R EVOLUTE

S P HERICAL l\fE C H A N IS.MS
Because of the difficulty of the geo1netric representation of three­
dimensional situations noted earlier, the synthesis of spatial mechanisms
,vill be restricted to analytical methods 1naking use of displacement equa­
tions; the ren1aining sections of this chapter are thus an extension of
Chap. 10.
The first exan1ple to be considered is the four-revolute spherica]
mechanis1n shown in Fig. 12-5. Upon choosing (Ji as the input variable
and 82 as the output variable, a displacement analysis leads to the equa­
tion (see Denavit, 1958)
. cos aa
. a 1 cot a4 cos 82 + sin
sin a1 cot a2 cos 8 1 + sin •
a2 sin a4
- cos a1 cot a2 cot a4 = sin 81 sin 82 - cos a1 cos 8 1 cos 82 (12-17)
frorn ,vhich the synthesis will be n1ade. The angle a1 will be specified,
and the design paran1eters are the three angles a2, a3, a4. The above
equation may be put in a form convenient for synthesis, as shown in
Sec. 10-5. Upon setting
K2 = sin a1 cot a 4
K3 . cos
= sin �3
a2 sin a4
- cos a1 cot a2 cot a4
G1 = cos 81 G2 = cos 82 U:i = 1
F = sin 81 sin 82 - cos a1 cos 81 cos 81
the equation beco,nes
K1 cos 8 1 + K2 cos 82 + K a = sin 81 sin 82 - cos a 1 cos 81 cos 8 2 (12-18)

y = log10 x

3
Y3 =l

y2 = 0.477

FIGURE 12-8 Logarithmic function to be generated by a 4R

spherical n1echanism with three accuracy points.
SYNTHESIS O F SPATIAL LIN KAGES 357

Table 1 2-1 SYNTHESIS OF FOUR-REVOLUTE SPHERICAL

MECHANISMS WITH THREE ACCURACY POINTS

Positions of input link:

81i
i = l, 2, 3
Positions of output link:
82i i = 1, 2, 3
a1 specified (different from zero)
W1 = COS 01 1 - COS 81 2 W2 = COS 82 - COS
1
822
W3 = COS 8 1
1
- COS fJ1
3
W4 = COS
821 - COS 82 3

Fi = sin 81i sin 82• - cos a1 cos 81i cos 8i

W5 = F1 - F2 WG = F1 - F3

K _ W4Wr, - W2W& _ W1Ws - WaWr,

1 - K2 -
W1W, - W2W3 W1W4 - W2W3
Ka = F• - k1 cos 81' - Ki cos 82i with i = 1, 2, or 3
K1 K2
a2 = arccot . a4 = arccot - .--
sin a1 sin a 1
aa = arccos (Ks sin a2 sin a4 + cos a1 cos a2 cos a4)

To detennine the design para111eters a2, a 3, a4, such that to three values
81 , 8 1 , 8 1 3 of the input variable 81 there correspond three prescribed
1 2

values 821 , 8 22, 8l of the output variable 82, the above equation is written
for each pair of values (8/, 8i),
K cos 8 1i + K2 cos 8i + K 3
1 = sin 8 1 i
sin 82i - cos a1 cos 81i cos 8ii (12-19)
Subtracting the equations corresponding to i = 2 and 3 fro1n the equation
corresponding to i = 1 gives two equations which will yield K 1 and K 2 ;
fron1 these Ka n1ay be obtained by means of Eq. (12-19), written for any
value of i (1, 2, 3). The angle a 2 is deduced fron1 Ki, a4 is deduced fron1
K 2, and a3 fro1n K3• rrhe co111putation procedure is surnlnarized in
Table 12-1.
Ex.ample Design a four-revolute spherical n1echanisn1 to gen­
erate the function y = log x in the interval 1 < x < 10 bet1,veen t,vo per­
pendicular shafts (Fig. 12-8). Further specifications are as followse: ranges
of variation of 81 and 82 to be 6.81 = 60e° and 6.82 = 90e° ; accuracy points
at x = 1, 3, 10; initial value of 8 1 (corresponding to x = 1 ) equal to 45e° ;
initial value of 82 equal to --15e°
. 'fhen
81 1= 45° 81 = .18.33e°
2
81 3 - 105e°
821
= - 45° 8l = -2.1e° 8�3 _ -l.jo
1'able 12-1 ,Yith a 1 = 90e° yields w 1 = 0.182, w2 = - 0.292, w 3 = 0.966,
YOGmU<l2--0 Ca,,H,,ardamoddoh!Nao<>hea· ,1 ..,..-.,,- ., .. -,,.,,.,r ·•..,;
, - Josa, ina""'a;""''•lal ,c; ,a e, 10aw;l.hal.h,..a""'"'ae'·apoiuto, s;..,aw i-,
�f by Si;,,.

lz..5 M Y 'I T II Y. S l S O F T W O • lt U O J.l11'�

T W O - S PII E R ! C - P ., I R >! E C H <. 'llS!tfS
The �R, W .,,oct>.rii,,., aualy...J ;., S....::, l�-2 will bo rooonridered
1,e..,armmatbcp0intofavi<o..-ofait.,a,yuth,,oi,. Thi,a,,...,.hanisma;,,a01,o,..a
in Hg, 12-1. Th,ainputaanglea♦ andatl><aoutputangl,a</, ar<arebt.te.Jaby
tl,eequatiO!la(-See.a!2-2)

a,' - a,• + u,' · · a,' + •, ' + ••' +az,,,,aro,a, + 2'1,«, ooo ♦

+ 2a,,, Bin a,oin ♦ - rooa,j,(24o'h + 2a,a, WII ♦)
- •ini(2a1a,amo ,,.oin ♦ - 2,,a,.ina.) • 0 ( 1 2 -20)
SYNTHESIS OF S P A T I A L LINKAGES 359
where q, is measured counterclockwise around z1 from x1 to the axis of
the crank and ,/I is measured counterclockwise around z4 from x1 to the
axis of the follower.
In the synthesis to follow, the values of a. and a., which determine
the relative position of the input and output shafts, are specified ; the
remaining five paran1eters are used in the synthesis of the linkage. In
addition, the angle c/>o corresponding to the initial position of the crank,
which ,vas chosen arbitrarily in the preceding exan1ple, rnay now be taken
as a design parameter, and c/> = <l>o + p.
An atten1pt was made also to use if; o, the angle corresponding to
the initial position of the follower, as a design para1neter. This failed
to give an equation of the type described in Sec. 10-5 that could be used
in a synthesis with linear equations. A. total of eight coefficients K were
required, with only seven design paran1eters, so that a nonlinear co,npati­
bility equation would have had to be solved as in the slider-crank problen1
considered in Sec. 10-7.
With the value of i/;o specified, the values of VI corresponding to each
accuracy point are known, and the displacement equation n1ay be put in
the form prescribed in Sec. 10-5,
KI cos p + K2 sin p + Ka cos VI + K 4 sin VI
+ K s(sin p cos i/; - cos a4 cos p sin i/;) + K 6
= cos p cos y; + cos a4 sin p sin i/t
in which

a3

K2
_ 84 sin a4 - a4 tan c/>o
-
a3
- a4
Ka =
-cos-
a1 </>o
.
_ 8l Sill Cl'.4
K4 -
a1 cos c/>o
Ks = tan <J,o
a12 - a2 + aa 2 + a4 + 81 + S4 + 28184 cos � 4
K6 =
2 2 2 2

2a1aa cos </>o

I'his allows a synthesis for six positions of the input and output shafts ;
the con1putation procedure is sumn1arized in Table 12-2. A program for
use with an IB�l-650 has been prepared to perfonn this synthesis. With
this progra1n the co1nputations indicated in 1'ahle 12-2 are perforn1ed in
35 sec. Another program perforrns an analysis of the 1nechanism, and
thus not only checks the results of the synthesis but also evaluates the
errors between accuracy points.
360 K I N EMATIC S Y ?I. T H ESIS O F L l � K A G E S

Table 1 2-2 SYNTHESIS OF TWO-REVOLUTE TWO-SPHERIC-PAIR MECHANISMS

WITH SIX ACCURACY POINTS

Positions of crank :
q, = <l>o + p; i = 0, 1, 2, 3, 4, 5
Positions of follower:
V' = V'o + q; i = 0, 1, 2, 3, 4, 5
a 4, a4, and v,,o specified (a4 .,,_ 0 and a4 .,,_ 0)
w;i = cos p; W;2 = sin p; W;3 = cos (v,,o + q;) W;4 = sin (v,,o + q;)
w,-5 = sin p; cos (v,,o + q;) - cos a, cos p; sin (v,,o + q;) w,-6 = 1
v,- = ('OS p,: cos (y,0 + q.-) + cos a4 sin p; sin (v,,o + q;)
Solve the system of six linear equations,

l W;;K; =
r;

V; i = 0, 1, 2, 3, 4, 5
j=l

to find K 1 , K2, K3, K4, K6, Kg.

ExampleDesign a 2R, 2G n1echanism to generate the function

!I = cos x in the interval 0 < x < 180° between perpendicular shafts
(a4 = 90°) a distance 1 in. apart (a 4 = 1). Further specifications are as
follows: range of variation of q, is 1.50°, range of variation of v,, is 100°,
and initial value of v,, is v,, 0 = 0°. Six accuracy points with Chebyshev
spacing are detern1ined as shown in Fig. 12-10 according to the method
presented in Sec. 5-5. They correspond to the following values of x and Y :

Xo = 3.069° X1 = 26.361 ° X2 = 66.708° X3 = 1 1 3.292°

X4 = 153.639° X5 = 176.931
°

Yo = 0.9986 Y1 = 0.8961 Y2 = 0.3953 Yae= - 0.3953

.4 =
Y -0.8961 Ys = -0.9986

Upon taking the initial crank and follower angle </,o and t/lo to correspond
to the first accuracy point (x0, y 0) , the successive crank and follower
SYNTHESIS OF S P A T IA L L I N K A GE S 361

rotations are given by

= tll(x, - Xo)
p; i = 0, 1, 2, 3, 4, 5
or Poe= 0 p1 = 19.41e° p2 = 53.03e°
p3 = 91.85e P• = 125.47e° p5 = 144.88e°
1 00
°

and q, = -,- (y; - Y o) i. = 0, 1, 2, 3, 4, 5

or Qo = 0 qi = - 5.125e °
q2 = -30.165e°
q3 = - 69.695e° -q4 = - 94.735e °
q5 = - 99.86°
1'he pararneters of the resulting mechanis1n, found by application
of the procedure described in Table 12-2 (or from the IBl\1-650 computer
program), are
a1 = - 0.4949 a2 = 2.7460 a3 = - 1.3158 a4 = 1.0
= - 0.8388 S4 = -2.0749 Q'.4 = 90e
°
St

with the initial value of </,o = 40.291 ° . The maximum structural error,
i.e., the maximum difference between cos x and the value given by the

y = cos .x

1.0 - - - - - - - - - - --------·1
I
!
I
I
%3 X4 X5!
0 .xo .X1 .x2
1180 .x (deg)
I

-
-••� - - - --· -· · - ·- - - - - -- - --=�

7
I /

;;
;

'

FIGURE 12-10 l)etermination of accuracy points for example

(Chebyshev spacing). x0 = 3.069° ; X i = 26.361 °; x2 = 66.708° ;
x� = 113.292° ; x. = 153.639 ° ; x� = 176.931 °.
, a ... ,1�i.1ao1a. 2� _.. .a.. .,, ..�....a..,.a,..... · - - ·
· ,,._ . ,....a!a0 <; , ,; 1il(l"aw·u, ·•a-•••.,, ·au. .....,.. ,.,a_..a,....
-•a•ith jin .a..... in•a""'a·a "" ol - S b1 i ;,,.

link>@.aioa, - 0. 00".!8, whlcl, oor.,.,.>OOO• I<> 0.14 per oont of tho ran�c
of ,·ariation of f . .\ modelaofth;,,an,..:han,.maio•ho•n in FiK , 12-1\
It m&}" be noted th.&t. by repeatin.1< th<a,ynth,,.i<, fo• aa""']uen.,. ol
,·ah- ofj,. it i,ap<>Mible toaobla.io aa-,u,nc,,aol"°\uhon,.(,onoamonf'
wl, id, \l,o m'"''· �..i,,.bO, from ti,, >t&ndpoin\ ,� (o<ee ,,,.n.,-.,i,w,inn '"
•J>&Ot!a""l''i"""'""la< may l.. c�n. Hyaoh., .. nwat<•l)<'\iti,·eaproe.- it i•
ab.,r<:-il,l,.ainall<"<"'l,aloaobta.inaa ,·ah�aoff,for-,,·hichaaJ(i,·oofuoclion
-,,·ill bo ll<IN'l'&tod -,,·ith ..,,,�n ..,.,U...,}' poinlL

12-6 Pl·x•·•·w� (a � X 1' k A Y O H a • o K

S l " ! \I E T U I C A L r , - � c T J O X S
l\a-,,·..,,a.,ol<e<faina,;.,c.a\0-8a\h.&ta1l,eaplana,a/ou,...baralinUjOea1i,·,.
poora,...011,a-,,·h.·,.aapplied lo theagenerahon ofahrnc\io, .. ..,.hicha,. •ym·
Hw\nt in thoaint<:rvalolaF"'""'"'"• ""a'°'a"'""'Plo,tl,.a/urn;\ionar - x'
"' Ibo inte,...a1 - I a.$ L .$ la. Thia W du" ro tbo fact th•I to g;,•e ,ym·
mctrya&b.,,.,ttboa,·olue La- 0 tl.. r,Hl<-1,.,.,Mldahaveatoal><aoompk,t,ly
"flat'" fo• th< pooition "'"""'P'>'"!l"� to � o, ¥ - 0. l�u.., of \],e
urnatraru,miocionaafll(i.a.,a,.,cha&ap<>o;,io., tt.isoohditionaio in,p,.,.tical
SYNTHESIS 0 1'' SPATIAL L I N KAGES 363

Symmetric functions, however, ntay be generated n1ore accurately

by mechanisms which then1selves have symmetric positions. This is the
case with the four-revolute spherical mechanism and the 2R, 2G mecha­
nism when some of the parameters are properly chosen.

Symmetric Four-revolute Spherical Mechanisms

The displacement equation of the four-revolute spherical mechanism,
used before in Sec. 12-4, is

sin a1 cot a2 cos 91 + sin a1 cot a4 cos 8 2 + sin

. cos aa
a2 sin a4
- cos a1 cot a2 cot a4 = sin 01 sin 82 - cos a1 cos 81 cos 02
If the angles a1 and a2 are chosen equal to 90e° , the equation beco1nes
cos aa
cot a4 cos 02 =
.
+ sin
.
a4
Sill 81 SIU 82 (12-21)

This equation relates the output variable 0 2 to the input variable 01 , and
it is noted that, if 81 is taken as 180e
°
- 81, the corresponding value of
02 is unchanged. Thus the associated mechanisn1 will have symmetric
positions with respect to the position for which 8, = 90e° (Fig. 12-12).
By using the paran1eters aa and a◄, a synthesis with two accuracy
points (points 1 and 2, Fig. 12-12) n1ay be carried out, and a third point
(point 3) then auton1atically falls on the prescribed curve because of its
symmetry. The synthesis thus an1ounts to a three-accuracy-point
synthesis.

'
821 -
- 83
2

--t-----� ----
I
I
13
I
I
I I
I
I
--r - _e _ ,.,,,..-==-
12 :
I

0 8� = 90• 83I 180°

(= 180° -:81}

FIGURE 12-12 Symmetric function to be generated by a

spherical 4R mechanism with three accuracy points.
,,.,..•• ,:.,-13o ),fo,lol <>f • •pbm,,.l '1' ._1ww,,.o1o�a1'>1ho fun<l;,u , • •'
·• 1h, ,.,�-.1 - 1 o,; _ , • .,• .,,...., .,...,...y 'n1o. /lio,of t..o� .•Jhy "f i

hampl• Dffi�n•ofoo,....,rnlutcoph<,rioalomochaniomtoo1t<no"'"
1heofuru,tion r • x' in \he intcrvo.l -1 ::i , ::i 1 Ranp ofovariatiol'
we to M .l.f, • l80'a00.l.f, • foO",with..,.,un.oypoint.at.< • - l , 0. 1
ltinin,um ,·o.lueof ,,. oorrespondin� to y - 0 (o,rond "'""'""'Y point). ;.

�=: -
toobo!o#,1 • \;t)"_ Tt.:,n
,., - 0 ,, , • 00- ,,, • 180"
1, 1 - 1'.!0" ,,, - roO" ,,, - 1w•
WritinKo!,,q.o(12-21)/oropooitioo, I and 2 yield,

-Of,oola, + 0

0 ..5 oot ,,.o+ � • 0.!!66

andoool,·i,,g /.., - ,,. •nd (coa .. ,}/(,in �,)olli,..,.

rot a, • o
0.8Gf, �; - o . ..s:i
a, • 49" a, 71 °
,.-ith a,o• 00-, a,o• 00'. A mo,l,,I of the on,.,...p,mdin� mtthani,,m 1'
.b.,... n in Vi&. 12-13, ,\odi,,plaoemen1 analy,i,,•ho,.-, that intheop....,."1
SYNTHESIS OF SPATIAL LINKAGES 365

case the maxin1um structural error is E = 0.033, or 3.3 percent of the

range of variation of y.
It may further be noted that the present n1echanisn1 ·will lend itself
to a more sophisticated synthesis by using the minimum value of 02 as a
design parameter. If this is done, a total of five accuracy points (count­
ing symn1etries) is obtained and the structural error is consequently
reduced. We may note that the structural error of the spherical four­
revolute linkage with three accuracy points is less than that of a planar
four-bar linkage with five accuracy points; this is seen on examination of
Table 10-3.

Symmetric 2R, 2G Mechanisms

The displacen1ent equation of the 2R, 2G n1echanis1n, fan1iliar from Sec.
12-5, is
a12 - a22 + aa2 + a42 + s1 2 + S42 + 2s1S4 cos a4 + 2a1a4 cos q,
+ 2a,s, sin a4 sin cf, - cos y;(2aaa4 + 2a1aa cos cf,)
- sin y;(2a1aa cos a4 sin </, - 2s1a 3 sin a4) = 0
With a4 = 90° and s1 = 0, this equation beco111es
a12 - a2 + aa2 + a4 2 + sl + 2a1a4COS q, + 2a1S4Sin <J,
2

- cos i/;(2aaa• + 2a1a4 cos </,) = 0 (12-22)

In this equation ,ve shall take the angle i/; as input and the angle cf, as
output; i.e., we interchange the crank and follo,ver fro1n what they are
in Fig. 12-1. It n1ay then he noted that changing i/; into -y; in the above
equation does not change cf,. 1'hus, the 1nechanisrn ,vill have sy1nmetrical

------ -- "'•- 4's___ _

</>3 = 4'1
--- -

-180" V,8 1/11 VIG I/ls 1/1, 1/12 V/3 w. 180

°

¥10 =0
f'IOURE 12-14
Syn1metric function to be generated by a
2R, 2G mechanism with nine accuracy points.
366 KINEMATIC SYNTHESIS OF LINKAGES

positions with respect to the position corresponding to 1/,- = 0, as shown in

Fig. 12-14. The mechanisn1 itself, in the position corresponding to the
center of syn1n1etry (I/I = 0), is shown in Fig. 12-15.
By using the paran1eters a1 , a2, aa, a4, S4 and the n1inimum value
q,0 of q,, a synthesis with five accuracy points (0, 1, 2, 3, 4 ; Fig. 12-14)
1nay be perforn1ed; and because of synunetry, four additional points (5,
6, 7, and 8) also fall on the curve, so that a nine-accuracy-point synthesis
is achieved in this way.
Example Generate the function y = x 2 in the interval - 1 <
x < 1 by n1eans of a sy1n1netrical 2R, 2G n1echanis1u. The accuracy
points are taken at x = 0, 0.25, 0.50, 0.75, 1.0 and also at the symmetric
points x = -0.25, - 0.50, -0.7,5, - 1.0. The range of q, is taken as
100°, while 1/,- will vary between -100 and 100° . With these assump­
tions the input and output angles corresponding to the five accuracy
points are
I/lo = 0 1/11 = 2,3° 1/12 = 50° 'f'3
.I. = 'i)
- �o
lf4 = 100°
poe= 0 Pl = 6.25° p2 = 25.0° Pa = 56.25° p4 = lOO(J
,vith t/>i = t/>o + p; i = 0, 1, 2, 3, 4
The para1neters of the resulting n1echanisms are
a1 = 0.5171 a2 = ;3.5660 = - 2.4990
aa a. = 1.0
S1 = S4 = a4 = 90
°
0 4.5792
with q,0 =
-62.17 °. Note that s1 = 0 and a4 = 90°
are required to
obtain a mechanism which is symmetrical with respect to the plane X1Z4
(Fig. 12-15). A 1nodel of this mechanisn1 is shown in Fig. 12-16.

---- ,. - - -� •;.,ll
I

FIGURE 12-15 Sy1n1netrical 2R, W mechanisn1 shown in its position cor­

responding to the center of symmetry.
 A diapW,O.....,nt. analy•io. of. the m..hanism. ,bo.ro. that. the. ,,....i•
mum.otructu,..,luror.io. , 0.0013,. ,.'h,ch. wr,..pondo.i o 0 . 1 3 .puttto\.of
the �.d.v•mtionof ,

lie)"'• ltuoo.11' Dao .llotriMnk&lk�l .t. lfilfomit!<l ,ur L'nu,......,huuctf.umlieh,,

G,,1.o<c<t�. F<i•--•;t, vol. �I. ru:,. �. ,�;
-:."T«hni><bo.HaumklJ>tm,,Lik,".Sprin.,..\·eriao;.OllG.. lle,!in,. 11163
U,..,.vit, L. � -nt. .\n&l,'W. ol )!.,,h,,o_. ll_t.,.. (2 >< 2) )1atntfl
olDuolXuml..... l'/JI 11,,.,,-ol.29.. 1 11.'\S
- on,J.II.S..llo,........ , .,.Kh,•.,._.,;.,:,...,.,.,.r,.,.1.<:,,,,,..pair..ll«honiom,
11...,.1 on �r.,-, A.s.11,r J. A.pp/. .11«• .. -- i;, vol. n. JuM, ,�
- &ad Approximate. Si·nt..... ol. Spati&I. J�nk- ,IS.\fF:. /
,l �• .lfri.,.1tt. E,. vol. 27,...,... l,. ,\h,.\,. l060.
368 KINEMATIC SYNTHESIS O F L I :-J K A G E S

Goldstein, H.: "Classical }1echanics," chap. 4, Addison-Wesley Publishing

Company, Inc., Reading, Mass., 1953.
Harten berg, R.. S.: Die Darstellung und Handhabung der niederen Elemen­
tenpaare in einer auf �Iatrizt:>nre<"hnung gegruendeten Zeichensprache,
VD! Ber., vol. 12, 1956.
l(eler, :.vr. L.: Analyse und Synthese der Raumkurbelgetriebe n1ittels Raumlinien­
geometrie und dua.ler Groessen, Forsch. lngenieurw., vol. 25, nos. 1 , 2, 1959
 APPENDIX:
MATHEMATICAL TRACTS

Often the reader of a mathematically oriented paper does not

have the mathematical background of the writer of the paper.
Such a reader would like to understand the n1aterial (or he would
not be trying to read the paper) but "gets lost" because of the
mathen1atics he has forgotten or never learned. For a writer to
attempt to tie up the number of loose ends that n1ay occur by
reference to "standard ,vorks" 1nakes life difficult for the reader,
who is uncertain about what he needs for clarification. The
matter of reference is not 1nade easier by the fact that many
different notations are used.

These tracts attempt to 1neet the above proble1n by presenting

the n1athematical tools in an orderly fashion, ,vith a degree of
con1pleteness adequate for their application in this book. Obvi­
ously an exhaustive treat1nent of each topic is quite i1npossible;
if the reader is sufficiently curious to want to know how much is
missing, then he is referred to the standard ,vorks.

In preparing this volun1e it was realized that it would be desirable

to hold all the n1aterial to a unifor1n notation, to achieve maximum
carry-over between sections. It was in1possible to do this, how­
ever, for there are not enough symbols available to allow each to
have a universally unique 1neaning, and consequently the sa1ne
sy1nbol may have a variety of n1eanings. These different quali­
ties will be indicated as they occur. After that, association and
context will establish ,vhich rneaning prevails.

Whenever feasible, illustrations of the 1naterial treated ,vill be in

ter1ns of pertinent kine1natic proble1ns.
370 KINEMATIC SYNTHESIS OF LINKAGES

The material presented is intended to serve as a guide or refresher

to those whose acquaintance with these topics is slight. The reader will
have to take on faith the fact that rigorous proofs are available for state-
1nents n1ade and conclusions drawn. Derivations judged useful in dis­
playing the operations involved are included.

A-1 COMPLEX :'-ll:MBERS

About Symbolism

:.\Iathe1natics is a sy1nbolic language. It n1akes use of sy1nbols-letters,

characters, abbreviations, or signs-that represent an idea or quality, as
an operation or relation. :.\1agnitudes and operations are usually indi­
cated by single letters, operators by particular signs, although an operator
,nay be a letter such as i. These syn1bols con1prise a shorthand used
either to 1nake staternents or to give instructions in a cornpact 1nanner.
1\. n1athe1natical systen1, as algebra, concerns itself ,vith a systen1
of 1nagnitudes and operations built up of certain chosen elen1ents. It
sets up a nu1nber of "rules of play" that pern1it certain combinations to
be considered as equivalent. ,vhen expressed in the sy,nbolic language,
the equivalence is declared in the sy1nbolic equation. The staten1ent of
equivalence (in ,vhich the various necessary operations are included) could
be given in ordinary language, but it is n1uch 1nore concise when written
in the syn1bolisn1 pertinent to the nature of the state1nent.

About Numbers

The concept of the size, or magnitude, of a thing, isolated f ron1 all other
qualities of the thing, is the concept of a natural nun1ber, as 1, 2, 3, ... ,
tha.t is, positive integers. For the establish1nent of a calculating proce­
dure, the positive integers n1ust be aug1nented by rational fractions,
irrational nu,nbers, negative numbers, and zero. The positive, zero, and
negative nu111bers are called real nu1nbersa1 and rnay be co,nposed of
several natural nun1bers.
Any real nun1ber, positive or negative, has the property of desig­
nating a point by acting as a coordinate. A group of such nurnbers, each
represented by a single coordinate, fonns o n e d- i1nensional space (,ve shall
find it convenient, in the n1anner of the ancient Greeks, to give our algebra
a geo1netric interpretation). When these nu1nbers are displayed as points
along a straight line, each point is the coordinate of its nun1ber. A one­
dimensional nun1ber c, co1nposed of the sum of t,vo others, a and b, each
1 Real numbers are the numbers that correspond to the distance-positive,
negative, or zero-measured from a fixed point O on a straight line.
APPENDIX: MATHEMATICAL TRACTS 371

c = 8 units
-►◄
i
ol--1--2-- - 3 4- 5 6 1 s
a = 5 units b =�
...._,:;._ 3 units
:..:.: ,.J
T • =--...i
(a) c =a+ b, where a= +5, b = +3

b=
!
c =

!+2 units"j• -3 units

0 1 2 3 4 5 6

•I
7

l a= 5 units
(b)c=a+b,wherea= +5,b=-3

b=+3units ,. t
t �
c=
, 1 -2 units
_
'
-5 -4 -3 -2 -1 0

L
a= -5 units
l
(c)c =a+ b,where a= -5,b = +3
FIGt:REA-l The one-dimensional number c = a + b. The ele­
ments a and b must be real numbers, and are laid off collinearly.

of which is itself a real number, is shown in Fig. A-1. In symbolic form,

c = a + b. The directions are to lay off a and then perforn1 the opera­
tion indicated by the + sign. This rneans the laying off of b collinearly
and to the right of a; the operation is that of addition.
When an ordered pair of nun1bers, written (a, b), defines a third
number c = (a, b), then c is called a co1nplex number. To discuss co1n­
plex numbers, it will be necessary first to develop the concept of the
"imaginary" unit, V -1.

The Meaning of the Negative Sign

Suppose· that we have a line with marks on it (Fig. A-2), the distance
between the n1arks being in units of, say, a. One n1ark is designated as
the origin 0, the direction to the right being agreed on as positive, to the
left as negative. The distance OA is then 3a, and the distance OA' is

---- -... .......

/
''
.....

/
/
I �
I \
I \
I \
I A't... t ..,\A I +
-Sa -4a -3a -2a -a
- I I
2a 3a 4a
:=J �
O a 5a
FIGt;RE A-2 The negative sign-a reverser of direction.
372 KINEMATIC SYNTHESIS OF LINKAGES

- 3a; the distinction between the t,vo is that they are 180e° "apart," in
the sense that, if we imagined 0A to be pivoted at 0, it could be turned
counterclockwise until it coincided with 0A'. That the counterclock"'ise
direction of turn is taken as positive is also a n1atter of agreement. Pivot­
ing OA' about O in a counterclockwise turn of 180e° "'ould also bring it into
coincidence with 0A. Both these operations of the 180e° turn have
involved reversal of direction, an association "'ith the negative sign, as
( - l)0A = 011'.

v
The square root of a negative nu1nber, as v-x2 , occurs frequently.
This can be written as (-1 )x2 . = v<=TI x. This y=-i is then an
operation to be perfor1ned on x. If the notation i 2 = -1 is introduced,
then it follows that (i) (i) = i2 = -1; two operations with i are the san1e
as n1ultiplying by -1. Now -1 has been seen to rnean reversal of direc­
tion, but we n1ay associate i2 with - 1 ; i2 = - 1 is thus an operator signi­

v
fying a counterclock,vise turn of 180e°. A single i indicates a turn of 90e°
in the counterclockwise direction. If we write -1 x = ix, the geo-
111etrical interpretation is that the nu1nber x shall he tu1·ned through 90e°
counterclockwise. =
Nlathen1aticians used the sign v-1 for about 200 years hefore
knowing what it n1eant. �rhe symbol i originated with Euler (1777), who
introduced so n1uch of our present notation, including J(x), e, and �
(sun1111ation).

Imaginary Quontiti�

,.:\.n expression consisting of an ordered pair of nun1bers c = (a, b) is n1ore

generally ,vritten c = a + £b; it is not the sa1ne as b + ia. This nu1nber
c = a + ib is called a co1nplex 1 number and is two-din1ensional, since it
involves both a and ib. Both a and b are real numhers; the number a is
called the real part and b the imaginary part. A complex number of
the form a + iO is said to be a real nu1nber, whereas one of the form
0 + ib is called a pure in1aginary. The validity of most of the theorems
of arithmetic and algebra is unaffected by their transference to complex
quantities, a very useful fact, since every algebraic equation in con1plex
algebra has at least one root (Gauss).

The Complex Number in Geometrical Representation

Suppose that we have a distance r laid off fron1 0 as shown in Fig. A-3.
Geometrically this r is a radius vector whose length is r times the length
1Rather, c = a + ib should be called the ordinary, or usual, complex number.
There are others, for example, c = a + wb, where it is agreed that w 2 = O. These
are not our concern at this time.
APPENDIX: MATHEMATICAL TRACTS 373
I
I
i �P, a point
ib.--------- -2 =
-......
a+ ih
I '-.

w\
I
I t / I "'-
I // I
/ I \
I \

- --���--
----<0>-- - - - - -
0 8
--l ----l------'-'----'-+
�------+----1 --�-----i
-a a b

1''1GtJRE A-3 The complex number c = a + ib shown by its two dimensions

a and ib.

of a vector of unit length. We can describe the position of the end point
P of the radius vector in several ways by the use of pairs of nun1bers; e.g.,
the polar fonn would require the specification of rand 0. We could also
use the real positive nun1bers a and b, saying, "Lay off a to the right, turn
90° to the left, and lay off b as shown." This n1ethod of locating the point
P uses a lot of words. We could use the syrnbol i to indicate the 90 °
turn to the left (counterclock,vise) by writing ib; this suggests an axis
perpendicular to the original axis. On the other side of O we could define
a number -a (represented by a point), since i2a = ( - l)a. The other
end of the ib axis is 180° beyond the first; that is, i2ib = ( - 1 )ib = -ib.
A.ssembling all this, we have Fig. A-4. The expression c = a + ib looks
like an algebraic equation even though it defines the con1plex nu1nber c,
consisting of a real part a and an i1naginary portion ib. Geo1netrically
we can interpret it as a vector equation, since ,ve now understand the
implication of the operator i, namely, that ib 1neans laying off b from the
end of a after a counterclock,vise 90° turn. It is nevertheless a picture
of the number c, shown in what is called the Gauss-A.rgand, or complex,
plane. Any point in this plane is a co1nplex number. The absolute value
of the con1plex number c is lrl, also called the modulus; 0 is called the
argument, or angle, and is always n1easured counterclockwise, as
illustrated.
Certain relations among con1plex numbers are shown in Fig. A-5,
a consideration of which shows that:
1. Two con1plex numbers can be equal only if their real and in1agi­
nary parts are separately equal.
2. Co1nplex nurnbers add vectorially; their sun1 is found by adding
the real parts to give the real part of their sun1 and adding the imaginary
parts to give the i1naginary part of their sum.
3. The difference of t,vo con1plex numbers is found by taking the
difference of their real parts to give the real part of their difference, etc.
374 K I NEMATIC SYNTHE S I S O F L I N KA G E S

lmag
2ib

ib c = a + ib
- - - - --- ;-,o- - ....__ '-
/
/ "
/ \
1 /
/ \
\
- - - - -------¥----' +
,,,, ,,,- {8
-
-a a
b
Real
also (i2)a = (-l)a also (i )(i 2)a == (- 1)(- l)a
2

-ib = (i2)ib = (-l)ib

FIGURE A-4 The complex number c = a + -ib on display. Geometrically,

r2 = a + bZ2

8 = tan-• -
b

a = r cos 8 ib = r sin 8
Forms of the complex number c:
c = a + ib orthogonal
c = r(cos 8 + i sin 8)
c = ·re•9
C = r/8 polar or circular

:'vlultiplication and division follow the rules of ordinary algebra

with the additional relation i2 = - 1 ,
(a + ib)(c + id) = ac + (ib)c + a(id) + (ib)(id)
= ac + i(bc + ad) + i2 (bd)
= ac + i(bc + ad) + ( - l)(bd)
= (ac - bd) + i (bc + ad)
a + ib - (a + ib) (c - id) - ac + (ib)c - a(id) - (ib)(id)
c + id (c + id)(c - id) c2 - (id)2
ac + i(bc - ad) - ( - l)bd
c2 - i2d2
ac + i(bc - ad) + bd
c2 + a2
ac + + bd . be - ad
= c2 + a2 i c2 + d2
Multiplying and dividing in the above manner is cu1nberson1e. The use
of exponentials is less tedious, but involves the use of Euler's theorem,
A P P E N DIX: MATHEMATICAL TRACTS 375

lmag
iba

FIGURE A-5 Addition of complex numbers c1 and C2 to yield c3.

which de1nonstrates that

(cos fJ + i sin fJ) = ei8 (A-1)
Since
c - r(cos (J + i sin fJ) = rei8
we may write
(A-2)
From this,
C:i = C1C2 = (r1 e'81)(r2e'61) = r1r�i<9,+9,>
= r1r2[cos (81 + 02) + i sin (81 + 82)] (A-3)
and - = -- - T1 ei(9,-8,)
r2

(A-4)

'fhe Gauss-Argand representation of the complex nun1ber makes

use of the cartesian-.coordinate plane xy and calls the co1nplex number z,
that is, z = x + iy. The number is thus displayed as the point whose
lmag
%2 = - x + iy z 1 = x + iy
=
r(-cos 6 + isin6) iy = i rsin6 =r(cos 6 + i sin 6)
- i=i =i 5 9

Real
i' x = r cos8
'

z� = -x -iy -iy = - i r sin8 z 4 =x -,y

= - r = (cos 8 + isin6) = r(cos 8 -isin6)
FIGURE A-6 The coniplex plane (Gauss-Argand plane). Complex numbers that
are mirror in1ages about the real axis are called conjugate; this means that -iy
replaces iy; thus z, and z4, and z2 and za are conjugate.
376 KINEMATIC SYNTHESIS O F LINKAGES

coordinates are (x, y). Note that z is the sy1nbol for the co1nplex nun1ber;
it has no relation to the usual third cartesian coordinate. The length of
the radius vector is designated by r. Figure A-6 shows in sun1mary many
properties of the complex number in all quadrants of the complex, or
z, plane.

The Angle of Turn

rrhe angle of turn of a unit vector is defined by the operator (cos 8 + i sin 8).
If we perform this operation again, turning through another angle 8, we
shall have
(cos 8 + i sin 8)(cos O + i sin 0) = (cos O + i sin 0) 2
However, the total angle of turn is 20, ,vhence we n1ay write the identity
(cos (J + i sin 0) 2 = cos 20 + i sin 20
On extending this toen operations we have
(cos o + i sin o)» = cos n.O + i sin no
'fhis last is known as De l\-loivre's theoren1.

Transformation of Vectors
Consider t,"·o vectors OP, and OP2 represented by their con1plex nun1bers.
The length of OP1 is r(l) = r, and the length of OP2 is kr (Fig. A-7).
To transform OP1 into OP 2, two operations are necessary on OP1 : (1) OP 1
n1 ust be turned counterclock,vise through the angle <f,, giving the point
P;, and (2) OP� n1ust be stretched into a length OP2, accomplished by
n1ultiplying the length r by the factor k. Now OP1 = r(cos 8 + i sin 8),

lmag

\
\

\
\
\
\
\
I
I
Real
-- -- r(l) = r---- J
- - - - - k(r) = kr ----

FIGUBE A-7 Transformation of vector OP1 into OPt-

A P P EN D IX : MA1' H E M A T I C A L TRACTS 377

lmag . lmag
2iy

iy . - - - - - z
' -z
'
,z
l.Y · · - · -- . - -
r
Real Real

(a) (b)

FJGL"RE A-8 Conjugate complex numbers z = x + iy and z = x - iy.

and turning OP1 through <J, yields
OP� = OP 1 (cos <J, + i sin <J,)
= r(cos 8 + i sin 8) (cos <J, + i sin <J,)
= r[cos (8 + <J,) + i sin (8 + <t>)}
Stretching OP� into OP2 1neans 1nultiplying by k, or
OP2 = kr[cos (8 + <J,) + i sin (8 + <J,)]

Coniugate Complex Numbers

Two complex nun1bers are said to be co1nplex conjugates of one another

when they have the sa1ne real part and in1aginary parts equal in rnagni­
tude but opposite in sign. For exan1ple, the con1plex conjugate of x + iy
is x - iy. In general, ,ve shall use a bar over the sy1nbol of a co1nplex
number to denote its conjugate ; thus, if z = x + iy, then z = x - iy.
Points of the con1plex plane representing two conjugate cornplex
numbers are synunetric with respect to the real axis (Fig. A-8a). Note
that the n1agnitudes of two conjugate con1plex numbers are equal and
that their angles are opposite in sense. Further, if z is real, then z = z ;
if z is pure imaginary, then z = - z.
Conjugate con1plex nu1nbers are useful in a nun1ber of situations,
such as in calculating the n1agnitude of a con1plex nurnber,
lzl2 = r2 = x2 + y2 = (x + iy) (x - iy) = zz
The square of the magnitude of a con1plex nun1ber is therefore its product
with its own conjugate. The real and imaginary parts of a complex nun1-
ber rnay also be expressed in ter1ns of the conjugate (Fig. A-8b),

Real part of z = x = ½[(x + iy) + (x - iy)] = z 1�

- iy) ] = z ;z
2\ [(x + iy)
Imaginary part of z = y = - (x
378 K I N E MA'l' I C S Y X T H E S I S O F L I N K A G ES

Even when a complex nun1ber is expressed in terms of sums, prod­

ucts, ratios, and exponents of other numbers, its conjugate may be found
by simply taking the conjugate of each of the numbers appearing in the
expression. Consider, for example, the complex nun1ber

= c1l•1 + C2e •s i

z d1ei"1, + d2e;"' 2

where c1, c2, di, d2 are then1selves complex and the angles </, 1 , <J,2, i/;1, i/;2
are real. The conjugate of z is
- c1e-i•,
= ·- - -- +---
c�-i•,
z
die- if, + a�- N,,
Example Differentiation of re;8 with respect to time. .\n expres­
sion of the forn1 re•8 really defines a point A in the con1plex plane, although
it is often convenient to think of rei8 as a position vector r tern1inating at
the point A . According to this last interpretation, we think of the 1notion
of A as being dependent on both the rotationa() of the position vector and
its change in 1nagnitude. If this expression is differentiated with respect
to time, the velocity of A results. 'faking the time derivative of rei8,

d (reiB) = i(riJ)eiB + fe iB (A-5)

dt
where 9 = dO/dt is the angular velocity of the line represented by the vec­
tor rei8 and t = dr/dt is the rate of change of n1agnitude of the position
vector r. The first term i(riJ)ei8 is interpreted as follows: A distance rfJ
is laid off in the direction r and then rotated through a counterclockwise
angle of 90° (indicated by the i). As seen from Fig. A-9, this first term
may be interpreted as the transverse velocity v8 of point A (Sec. 4-4).
The second term fei8 represents a vector of magnitude f in the direction
of the position vector r. As seen from Fig. A-9, this second tern1 n1ay
be interpreted as v', radial velocity of point .4 .

lmag lmag lmag

.,,..---- ........90•'-
. + .r)e•'
\

" \
\
v = (ir8
\
i(J
A ;::?
�r(J
/
✓
\

A
Real Real Real
0 0 0
(a) Position of A (b) Transverse velocity (c) Total velocity given as sum of
is given by re i8 v 8 given by ir8e ii transverse and radial velocities
FIGURE A-9 Kinematic interpretation of first derivative of rei8•
APPENDIX : MATHEMATICAL TRACTS 379
Plane 1
lmag lmag

•�2 = irBeil
8cor= 2if(Je i 6
&A/2= reil

,,,,,,,, aA 2 == -,iJ2e i6
,/
,/
Real ,,,,, Real
0
(b) Acceleration of A is given by the sum
of four components. Note that
(a) Position of A is gi1,1en by re i6
8A
2
+ a!2 == 8A 2

FIGURE A - 10 Kinematic interpretation of the serond <lE'rivative of re•11•

Taking the second time derivative of rei11,

d2 d
dt2
(reill) = dt
[i(rB)eiB + feill j
= fe;8 - r82ei11 + ir6ei 8 + 2if8ei8 (A-6)
The rnost natural interpretation of this equation is made in tern1s of the
concepts of relative acceleration and Coriolis acceleration presented in
Sec. 4-9. Consider .. Ato be a point of a plane 3 rnoving along the line
0-u of a plane 2, itself rotating about the origin O (Fig. A-lOa). 'l'he
reference plane 1 is the con1plex plane carrying the real and in1aginary
axes. The path sho\vn as KA in the figure is that traced by A on the
reference plane. We know that the absolute acceleration aA of A (with
respect to plane 1 ) is the sun1 of three con1ponents,

(A-7)

. ,vith respect to plane 2 ; OA2 is the accelJ

where oA,2 is the acceleration of A
eration of the point A 2 of plane 2, coincident ,vith A at the instant con­
sidered ; and ac0r is the Coriolis acceleration as described in Sec. 4-9.
With this in 111ind, we 1nay now lay out the four terrns of Eq. (A-6) as
shown in Fig. A-lOb and co1npare the1n with the tern1s of Eq. (A-7).
The first tenn fe;8 is the acceleration of A relative to plane 2 : its
magnitude is given by f and its direction by e' 8• The next t\VO tern1s rep­
resent the nonnal and taugential accelerations of the coincident point A2;
these are the only tern1s which do not vanish when r is n1ade a constant
(so that f = f = 0), thus fixing the point 1\ in plane 2. Finally, the ter1n
2if8e•8, of magnitude 2f8 and direction rotated 90e° from the direction of
e•8 in the sense of 8, is the Coriolis acceleration.
380 KINEMATIC SYNTHESIS O F LINKAGES

A-2 L I X E A R S Y S T E .'.\f S O F A L G E B R AI C
E Q l" A T I O � S - I : D E F I N l 'f I O N S

A linear algebraic equation is one in which the unkno,vns, ho"'e,·er

n1any, appear to only the first power (degree).
1'he general for1n for m equations in n unkno,vns is

a1 1X1+ a12X2 + . . ' + a1 :l

0
+ ' ' . +a X = 1n ,, b1
a21X1 + a22X + • . . + + ' . ' + a2nX =
j j

2 a2jXj 11 b2

(A-8)
a;1X1 + a,:2X2 + · · · + u;;X; + · · · + a;,,x,. = b,
. . . . . . . . .

This may also be ,vritten

LI a
n

;_;:rj = b. i = 1 , 2, . . . , m (A-9)
j•

When dual indices are used, the first ,vill ahvays denote the row, the second
the cohunn. The general coefficient a;i stands in the ith row and jth
column.
The syste1n is said to be:
1 . Nonhon1ogeneous if at least one b, � 0
2. Ho1nogeneous if all bi = 0
3. Consistent or con1patible if there is at least one solution common
to all equations
The syste1n of equations 1nay be interpreted as a linear transfor­
n1ation; i.e., the syste1n assigns a unique set of quantities b i, b2, • • •
b,, . . . , b,,. to every set of quantities .t,, .i:2, • • • , x.-, . . . , x,, . Such
an assignn1ent is called a linear transforn1ation of the set xi , . . . , Xn
into the set b i , . . . , bm.

A-3 D ET E R M I � A � T S

1'he application of detenninants gives a \\'ay of solving a systen1

of nonhon1ogeneous consistent linear equations ,vhen there are as rnany
equations as unknowns. It is possible to solve these equations by the
1nethod of eli1nination using judicious 1nultiplication, addition, and sub­
traction. Determinants furnish rules for speeding so1ne phases of the
111anipulation and n1ake it possible to forn1ulate solutions in a concise
manner.
A P P E N D IX : MATHEM A T I C A L TRACTS 381

Meaning of Determinant

Determinants developed with the study of linear equations. The sin1-

plest forn1 of an algebraic equation of the first degree in one unknown is
ax = b, in which a and b are supposed given an<l x is to be found. The
answer appears to be x = (1/a)b, but ,vhether or not this is really so needs
to be determined, for, in an algebraic problen1, literal symbols such as a
and b are used to discuss relations regarding all nu1nbers. But all num­
bers includes zero, and particular situations exist if either a or b or both
happen to be zero.
Our inunediate concern. is ,vith 1/ a, a nu1nber ,vhich ,vhen multi­
plied by a yields 1 ,vhen a -:;c 0. \Vith a = 0, 1/a does not exist, since
the product of any nu1uber and O is always 0. It follows that the pos­
sible existence of a solution depends on having a -:;c O ; the situation is
determined (at least in part) by the value of a, ,vhence a is called the
determinant of the equation. This necessary exa111ination of a leads to
the well-known theoren1e: If a -:;c 0, the equation ax = b has but one solu-
.
tion, x - - biI a.
The value of b is also part of the total picture. With b = 0 and
a -:;c 0, x has the value of 0. When both a and IJ are zero, any value of x
satisfies the equation and the solution is indetern1inate. With a = O and
b -:;c 0, there is no answer to the problen1. The nu1nber b is son1etin1es
called the secondary detenninant of the equation.
The theore1n derived fron1 the exan1ple based on a single linear
equation ,vith one unknown applies n ti1nes to a set of n linear equations
inen unkno,vns. Nun1bers corresponding to the foregoing a and b result
fro1n the sun1s of products of the constants defining the equations: these
nun1bers are the determinants.
Linear equations began to be studied in Ne,vton's time. Seki
(1642-1708), '"'hon1 the .Japanese consider their n1ost i1nportant rnathe­
n1atician, discovered detenninants prior to 1683. Leibnitz expounded on
the subject in a letter to De L'Hospital in 1693, a letter not made public
until 18ti0. Deter1ninants \Vere again discovered by Cramer (1704-1752),
popularly ren1en1bered for his ''rule" (1750). i\-Iany n1ade further contri­
butions, of who1n we can n1ention only a fe-w. Vanderinonde (17 53-1796)
gave the first logical and connected exposition of the theory. Cauchy
(1789-1857), the great 1naster of deter1ninants, rounded the theory into
son1ething like its present fonn and introduced the name determinant, in
1812. Sarrus (1798-1861), also associated '"'ith a "rule," is better known
to kinen1aticians for having devised the first true straight-line motion
in the form of a spatial n1echauis1n. I(inematicians often misspell his
name as Sarrut. Cayley (1821-1895), a prolific worker, gave us the
familiar upright lines, or determinant bracket.
382 K I N E M A T I C S Y N T H E S I S O F L I N KA G E S

Second-order Determinants
The operations pertaining to the solution of linear equations ,vith two
unknowns are readily given and allo,v a perception of the process culmi­
nating in the shorthand of detenninants. The second-order deterrninant
to be developed is in addition always useful.
Consider a systen1 of two linear nonhon1ogeneous equations in t"·o
unknowns x1 and x2 ,
a11x1 + a12X2 = b1
(A-10)
0-21 X1 + a22:'l;2 = f> 2

To solve this syste1n, the first equation is 1nultiplied by a22, the second
by - a12, and the two equations added. This gives

(A-11)

Similarly, 1nultiplication of the first equation by - a21 , of the second by

au , with subsequent addition of the two, produces

We note that x1 and x2 have identical coefficients. On solving for

the unkno,vns we have

:i· 1 = a---- b2a12

b1a22
and :l:2 = - b2a11
- -- b1a21
- ­
- (A-13)
11
a22 - a21a12 a11a22 - a2 1 a12

in which each x is expressed in the fonn already established for an equa­

tion in only one unkno,vn, na1nely, x = b/a. Each numerator, and the
common denon1inator of x1 and x2, although expressed as the sun1 (or
difference) of several products, will reduce to single nun1bers equivalent
to b or a, thus becon1ing the determinants of their respective equations.
In particular, there is the quantity a11a22 - a-21a12, con1mon to both
x1 and x2. It is represented by the syn1bol D,

a square array of the coefficients of the unknow·ns, ,vith the four ele1nents
arranged in two row·s and two coh1n1nse: this is a deterrninant of the second
order.
We n1ay view· the array as a practical device for producing the
single number D. The evaluation (or expansion) of D follows from the
sum of the signed products of-the elements of the t,vo diagonals according
APPE!IID IX: MATHEMATICAL TRACTS 383
to the following rule:
'\ /,
au

a21
X ; a12 I

a22 1
= ( +) (a11a22) + ( - ) (ai2a21) = a11a22 - a12a21

( - ),: �( + ) (A-14)
2 5
- +[(2) ( - 6)) + ( -)[(5)(-3)) = - 12 + 15 = +3
-3 -6

The polyno1nials on the right sides of (A-11) and (A-12) may also
be written in determinant forn1, giving the sy1nbolic relations

au a12 b1 au a12 au b1
= a12
and X2 = (A-15)
a21 a21 b2
X1
a22 b2 a22 a21 a22
b1 ai2
b2 a22 Di
whence Xi = =
au a12 75

a21 a22 (A-16)

au bi (Cramer's rule)
a21 b2 - D2
and X2 = au a1 2 I 75
a21 a22
yielding x1 and x2 if D � 0.
It should be noted that the "nu,nerator" detenninant D 1 of Xi is
like D except that the coefficients of x 1 are replaced by the constant terms
b 1 and b2; in D 2 the coefficients of x2 are replaced by the constants b 1 and
b 2. Consequently the "numerators" may be evaluated in a 1nanner
similar to that for D.

Third-order Determinants

If three linear nonho111ogeneous equations in three unkno,vns as

a11x1 + a1 2X2 + a1aXa = b1

a21X1 + a22X2 + anxa = b� (A-17)
a31X1 + a32:r2 + aa3X:1 = bi
are n1anipulated toward a solution, x 1 , x�, and xa ,vill have an identical
polynon1ial coefficient,

(a11a22aaa - a11aa2a2a + a21aa2a1a - a:n a12aaa + aa1a12an - a31a22a13)

(A-18)
384 KINEMATIC SYNTHESIS OF LINKAGES

This coefficient is again the detern1inant of the system and is written

l au a12 a1a
D - a21 022 an (A-19)
a�1 Oa2 Q3;1

The evaluation is conveniently carried out as follov,s: Write the detern1i­

nant as usual, and repeat the first two colun1ns. The tenns ,vill follow
from the signed products of the elen1e11ts according to Sarrns' rule,

(A.-20)

CAUTION : Sarrus' rule is not applicable to detenninants of higher

than third order.
Further manipulation of (A-19) would give
l
b1 012 a1 3
b2 a22 a23
a32 a33 i D
X1 - ba
= -
D D
a1 I b1 a1 3
a21 b2 a2�
aaa
(A-21)
a31 ba _____,_ D2
-
X2 = -
D D
au a12 b 1
a21 022 b 2
aa1 aa2 ba Da
--
X3 =
D D
yielding x1, x2, and x3 if D � 0.
This solution is si1nilar in fonn to that of the second-order equa­
tions. Note again that in the "nun1erator" detern1inants the b's stand
in the place of the coefficients of the x being sought.

Determinants of Higher Order

A set of n linear nonho1nogeneous equations in n unknowns will have

detern1inants of the sa1ne character as those of the second and t.hird
orders, viz., if of nth order, then there will be n! n-fold tenns containinp;
one, and only one, ele1nent from each row and each column in each deter-
A P P E N D IX : M A T H E M A T I C A L T R A C T S 385
minant. The evaluation of determinant� of even the fourth order becomes
tedious (24 fourfold terms), and unn1anageable if n > 4 (with n = 8,
there are 40,320 eightfold tern1s in each of the nine necessary determi­
nants!). In addition, there is no direct 1nethod of evaluation such as
Sarrus' rule, which gives not only the terms but also provides the proper
signs. The simplest sche1ne for the expansion of higher-order determi­
nants involves the use of the minor1 and the cofactor, also called the
signed minor.

Minor
Consider a determinant of the nth order,

au a12 a1; ll1n

ao·
.
a21 a22 a2n
. . .
-1
. . .
D - l atJ--1I - l
a,1 ll;i a;; a.,.
:
I
I
. . .
l anl an 2 a,.; a,in
- 1, 2, 3, . . . ' n (A-22) i, J

where the first index i is for the rows and the second index .i denotes the
columns.
The first n1inor of any elen1ent aii is the deterrninant A;h forn1ed
on suppressing the ele1nents of row i and column j. Thus, fron1
au a12 au
D - a21 a22 a23 (A-23)
aa1 aa2 a33

nine detern1inants Ai; can he formed, as

lviinor of a 1 1 :

.-l u - a2!? au
21 a22 a2a 1
Uit aaa
31 U32 a33 !
N!inor of a21 :

a1 a
A 21 - a12

' -
a:1•.• aaa
�Iinor of a�2:
au
.- l �2 - ais
a2a
a21

1 Minor is the short form of "minor determinant."

386 K I N E M A T I C S Y N T H E S I S O F L I N KA G E S

and so on, for the rest. Each of the new deter1ninants A i; is of one order
less than the determinant from which it ,vas formed.

Cofactor
A signed minor is called a cofactor. The sign of the cofactor is positive
if the sum of the nun1bers of the ro,v 1: and colu1nn j of the elen1ent ai; is
even and negative if the sun1 is odd. This signed 1ninor is written C,;
and may be defined as C,; = ( - l)i+iA i;- The rule of sign 111ay also be
re1nembered in tenns of the cha.rt

+ +
+ +
+ +
+ +
By ,,,ay of exarnple,

Cofactor of a1 1 :
a2 2 a2a a23
Cu - ( - 1) 1+1 - + a22
a a2 aaa a32 a u
Cofactor of a21:
a12 au a12 au
- ( - 1) 2+1 !
I

C21 -
! a 32 a3a a32 au
Cofactor of a�2:
au a1 3 an a 1a I
Ca2 - ( - })3+2 -
a21 an a21 a.23 I

Rules for the Evaluation of the Determinant of Any Order

1. Fonn the product of each elen1ent in a ro,v or column by its
cofactor; i.e., form a.;C;;.
2. Repeat this process on C,; until it can be easily expanded, as on
reaching the third- or second-order detenninant form.
3. The algebraic sun1 of the terms so found is the value of the
higher-order determinant.
Example 1 Given the necessary equations to form

D=
A P P E N D I X : MATHEMATICAL TRACTS 387

1. Suppose the first column to be chosen:

a22 a2a a24 a12 a13 a14 I a12 a12 au
D - + au a32 aa3 aa4 a21 aa2 aaa aa4 + a11 a22 an a2"
! a4� a4� au 1 a42 a"a a44 a42 a4a a44
I
I a12 a1a a14
- a n a22 a2 3 a2,
aa2 aa3 aa,
The third-order determinants could no,v be expanded by Barrus' rule, or
they could be further reduced. Taking the second a,1C,:; as an example:,
,ve have:
...
')
I
a1 2 a1a au
a21C21 - - a21 aa2 aaa aa4
a42 a4a a44
and, choosing the first ro,v,

- - a21[a,2(aa3aH - a34a43) - au(aaza44 - a34a42)

+ a14(aa2a43 - a42aaa)]
On con1pleting the 1nultiplication of this a21C 21 there will be 6 fourfold
products, ,vhich, together vvith the products of a11Cll, aa1Cs1, and aHCo,
,vill give a total of 2-1 fourfold products.
3. The algebraic sun1 of the 24 fourfold products is the value of
the detern1inant D.
Example 2 Consider the detern1inant
01
3 5 0 4
D - · 2o
I

2 2 -1
6 0 0
0 1 -1 3,
1. The first or third colu1nn or the third ro,v rnight be chosen
because of the zeros. Suppose that vve take the third ro,v :
5 0 4 3 0 4
D - +2 2 ...,
') -1 - 6 0 2 -1
1 -1 3 0 -1 3
3 5 4 3 fi 0
+o 0 2 · -1 - 0 0 2 2
0 1 31 0 1 -1
388 KINEMATIC SYNTHESIS OF LINKAGES

2.

-1 ! 2 -1 !
- 2 (+5 2 +4 2 -i , )
I-
D -1
o 31
3 1 1

( - -1 ! _ 0 1 o 4
)
I

7;+0 °
I

- ij 3
- 1•)
I 3j I -1 ,1 I i2 -1
3.
D - 2[5(6 - 1) + 0 + 4 ( - 2 - 2)] - 6(3(6 - l)]
- 2[5(5) + 4 ( - 4)] - 6(3(,5 )] = 18 - 90 = -72

Cramer's Rule
The solution of a systen1 of n nonhon1ogeneous linear equations in n
unkno\vns n1ay be obtained by the application of the above excerpts fron1
the theory of determinants by the application of Cra1ner's rule. The
procedure follows:
Given the systen1 of equations

a11x1 + a1 2:i·2 + · · · + a1i:i·J + · · · + a1,,x,, = b,

aux, + a�2X2 + . . . + a2;Xj + . . . + a2nX,i = b2
. . . . . . . . . . . . . . . . . . . . . . . . . ' . . .

a n 1:1:1 + Q II 2:r� + . + a,,jl:j + · · + a,,,,x,. = bn

. au a12 a 1_; a, n !
: a21 ·-
a .,. a 2i
.
a2,. ;
.
If D - laiJl - ;;e 0
' a;1 a;2 a;i a;,,
i . . . . • • •
I

I a,, 1 a,,2 a,,; ann :

then by Cramer's rule the unique solutions arc

X1 = D1 X2 = D2 .lj =D
1)
D
:r" = n.
D D D
1n which D; is the detern1inant forn1ed by replacing the elenients a1;,
a2i, . . . , a;)) , a,,i of the jth colurnn by b 1, b 2, . . . , b;, . . . , b,, ,
respectively.

Properties of Determinants

The most useful of the properties of detern1inants (but not all) 1nay be
noted:
APPENDIX: MA1'HEMAT ICAL TRACTS 389

1. A detern1inant vanishes if:

a. T,vo rows (or colu1nns) are identical.
b. A row (or colun1n) has only zero elen1ents.
2. A detern1inant is unchanged if:
a. l�o,vs and colun1ns are interchanged, e.g.,
lg 3 1' 8 4 7
14 5 2 = 3 5 9
11 9 6 1 2 6
b. To each element of a ro,v (or colun1n) is added the corre­
sponding element of another ro,v (or colun1n) n1ultiplied by
any constant, e.g.,
8 ;; I j + (1)2 3 1 '! ! 10 :; i I
!8
,.
2 : = 4 (2)2 2 ' - 'i 8 ,J 2
I;
4 + 5 I
1
' ;.)

7 + (6)2 9 6 1 1 19 9 6
17 9 o! J

3. A detern1inant is multiplied by a constant if the elements of a

ro,v (or colun1n) are 1nultiplied by that constant, e.g.,
: 8 3 1 8 (3)3 1 3Is 1
i (3)4 (3)5 (3)1 = 4 (3)5 1 = 3 4 ;) 1
7 9 6 I7 (3)9 6 ', 7 9 6
4. A. detenninant changes its sign if t,Yo of its rows (or columns)
are interchanged. e.g.,

Is � 1 '3 s 11
,4 .-:> 2 = - 5 4 2
:7 9 6 9 7 f)A,

The foregoing has shown only the briefest ,vorking outline of the
theory of detern1inants as needed for the present purpose. Refinen1ents,
"tricks," and short cuts, helpful to the dedicated user ,\,ho has the time
and inclination to learn the1n through repeated application: have been
on1itted. Nothing has been said about addition and multiplication.

A-4 L I N E A R SYSTEMS O F A LGEBRAIC

EQUATIONS- I I : RANK A N D S O L U TION
A determinant has been described as a symbol that defines poly­
non1ials of a type frequently encountered, as in the solution of nonhon10-
geneous linear equations. \Vith n equations in n unknowns, the deter­
minant is a square array of quantities co1nprised of the coefficients of the
unknowns or the constants of the equations. In the course of a solution,
n + 1 polynomials, each containing n ! n-fold tern1s, have to be established.
390 KINEMAT I C S Y N T HESIS O F LINKAGES

The polynomials have a specific numerical value, and the determinant

provides not only a shorthand for indicating it symbolically but also a
bookkeeping scheme for finding the corresponding number.
In the exan1ples given, unique solutions of linear equations did not
exist ,vhen D = O; also, the nun1ber of equations and unknovrns had to
he the san1e. A ne,v concept, the matrix, is eminently useful in studying
those cases ,vhere D = 0 or ,vhere the nun1ber of equations differs fro111
t.he number of unkno,vns.

Matrix
l\ matrix is defined as an ordered array of elen1ents in row, colu1nn, rec­
tangular, or square fonn, the elen1ents being the coefficients of a systen1
of linear equations (transformations). A n1atrix does not represent so1ne
polynomial, as does a determinant : it is a syn1bolic representation of an
ordered array of nurnbers. Because of the array, there is a superficial
resemblance bet,veen a 1natrix and a determinant, and although a relation
does exist between the two, each is based on a quite different concept.
An m X n rnatrix is a syste1n of mn quantities arranged in m ro,vs and
n columns. On croi2sing out appropriate rovvs and columns ,ve get square
arrays of nu1nbers which rnay be considered determinants.
Consider a set of m linear nonhon1ogeneous equations in n
unknowns,
a11X1 + a12X2 + · · · + a1;X; +

.. .
(A-24)
. . . . . . . . . . . . . . .
+ am,,X1, = bm

The matrix A. of the syste111 1s fonned fron1 the coefficients of the

unknowns,

a·,,
,_ (A-25)

I a,,. } am2 a,..; a,,u,

Double bars are used to distinguish n1atrices from detern1inants.

Determinants of o Matrix
It has been remarked that by crossing out certain rows and columns the
resulting square arrays of nun1bers n1ight be considered as determinants.
Any such determinant is a determinant of the matrix.
A P P E N D I X : M A T H E M A T I C A L 'f H A C T 8 391

Example 1 Given
·1 3 7 l'.,1[
A = 2
;) ��

This 1natrix contains six first-order and three second-order determinants.

Example 2 Given
1,
4 I
�
-I• ')
f· ;)

.4 - 5 2 7 I

8 1 10 ,!
This matrix contains the follo,ving determinants: nine first-order, nme
second-order, and one third-order
.

Ronk of a Matrix
The rank of a matrix is defined in ter111s of its detern1inants: the rank of
a matrix is r if at least one nonzero detern1inant of order rexists, ,vith all
deter1ninants of order higher than r equal to zero.
Example Consider the n1a.trix
1 0 1 3
A - 2 1 0 -2
-1 -1 1
The four third-order detern1inants are
I 0 1 1 0 3
2 I 0 = 0 2 1 -2 = 0
-1 -1 I -1 -1
I I 3 0 1 3
2 0 -2 = 0 1 0 -2 = 0
i -1 I 5 -1 1 5
However, since
I O! � 0
2 1t
there is at least one second-order detern1inant different froru zero and
hence the rank of this rnatrix is r = 2.

Solution of Linear Systems of Equations

The solution of systen1s of linear equations is usually carried forward in
English-language texts by a 1nethod based on the use of the augn1_ented
matrix of the coefficients. However, the concepts of principal and char­
acteristic determinantsa1 represent useful additions to the general n1ethod.
1
Introduced by E. Rouche.
392 K I N E M AT I C S Y N T H E S I S O F L I N KA G E S

They are presented in the follo,ving, for they are ,vell suited to our
problems.
Returning to the syste1n of Eqs. (A-24), let r be the rank of the
n1atrix A of its coefficients (A-2,3). Since at least one detern1inant of
order r of the n1atrix A 1nust be different frorn zero, \\"e n1ay assun1e that
the equations and unknowns are taken in such order that the deterrninant
D of the first r ro,vs and columns of the n1atrix A is different from zero.
This determinant is called the principal determinant. The first r equa­
tions and the first r unknowns (whose coefficients appear in the principal
determinant) are the principal equations and the principal unknowns.
The detenninants

' .
D
p =r+ 1, . . . , m (A-26)

1------- - -----' br
ap, bp

formed by adding a ro,v and a colun1n to the principal detern1inant as

shown are called characteristi"c determ.1·nants. Note that the last row of
DP is formed of coefficients of one of the nonprincipal equations. Since
there are m - p such equations, a total of m - p characteristic deter­
minants may be fonned. If r = m, the rank of the 1natrix is equal to the
nun1ber of equations and there are no characteristic detern1inants.
It 1nay be shown that a systen1 of m. equations with n unkno,vns
such as (A-24) is con1patible, i.e., has solutions, if all its characteristic
detern1inants are zero. If a syste1n has no characteristic deter1nina.ntf­
(r = m), then it ahvays has solutions.
The number of solutions depends on the number of unknowns and
rank. With n = r, the systen1 has a unique solution. With n > r, the
syste1n has an infinity of solutions, and there are n - r nonprincipal
unknowns which can be specified arbitrarily.
Example 1 Consider
X + + 3t =Z 1
2x +y = - 2t 2
-X - y + Z + 5t = 3
The ,natrix of its coefficients is
1 0 1 3
A - 2 1 0 -2
-1 -1 1 5
A P P E N D I X : MATHEMATICAL TRACTS 393

We have seen that the rank of this matrix is r = 2 and that the second­
order detern1inant
·1 0
D = 1 l
)2
forn1ed by the elernents of the first two ro,vs and colu1nns is different
from zero. We n1ay take this detern1inant as our principal deterrr.1inant,_
so that x and y are the principal unknowns, With only one nonprincipal
equation (the third) ,ve have one characteristic determinant,
1 0 1
Da - ' 2
'
1 2 - 4 ;,,e 0
i - 1 -1 3.
Since this detenninaut is different fron1 zero, the systen1 is incon1patible,
No set of four nu1nbers such as x, y, z1 and t will satisfy all three equations.
Example 2 Same system as above, except that the second n1en1-
ber of the first equation is changed to 5. The matrix coefficient is
unchanged, but the characteristic detern1inant is no\\'
1 0 5
D� = 2 1 2 =0
-1 -1 3
Since it is zero, the systen1 is compatible and the two nonprincipal
unknowns z and t may be specified arbitrarily ; take, for exan1ple, z = 1
and t = 4. The principal equations (the first t,vo) become
X = 5 - 1 - 12 = -8
2x + y = 2 + 8 = 10
Solving them for x and y yields x = -8, y = 26. The reader may
observe that the values x = - 8, y = 26, z = I, t = 4 satisfy the third
equation.
Example 3 Consider
X + y =I
X - y =2
X + 2y =�
The 1natrix of the coefficients is
1 1
A = 1 -1
1 2
The determinant
1
D = I
1 -1
- -2

is different from zero and may be taken as principal determinant. The

394 KINEMAT I C S Y NTHESIS OF L I N KAGES

rank of the matrix is r = 2. The first t,vo equations are principal equa­
tions, and the.re is one characteristic detern1inant,

D:: =
l1
11
1
-1
i
2
I= -5
11 2 3
Since this determinant is not zero, the system is incompatible.

Solution ol Homogeneous Equations

A special case of linear systems of equations occurs when all second n1em­
bers b 1, . .., b,, . . . , bm are zero so that each equation is homogeneous;
i.e., all terrns are of the sa1ne degree in the unknowns x1, ... , x;, . . . ,
x,. . Because such syste1ns of equations are of particular interest, we shall
see how the general rules apply to them.
Consider a system of m hon1ogeneous equations in n unkno,vnsJ
a11X1 + a12X2 + · · · + a1jXj + · · · + a1nXn =0
(A-27)

Since all l>/s are zero, all characteristic determinants have a column of
zeros and are therefore zero. We conclude that systern (A.-27) is ahvays
co1npatible. As a n1atter of fact, a look at the equations reveals that we
can always count on the trivial solution
Xi = · · · = X; = · · · = X,, = 0
Still applying the general rules, we note that, if the rank r of the n1atrix
of the coefficients is equal to the nun1ber n of unknowns, this trivial solu­
tion is also unique. If, however, r < n, there will be other (nonzero)
solutions and n - r nonprincipal unkno,vns n1ay be specified arbitrarily.
In particular, if there are n1ore unknowns than equations (n > m), the
rank is Jess than the number of unknowns and there ,vill be solutions other
than the trivial zero solution.
Example 1 Consider

2x1 - X2 + 3xa = 0
X1 + 2x2 - X:1 = 0
3x1 + 4x2 + X3 =0
2 -1 3
Here D - 1 2 -1 = 8
,3 4 1
 - - - - ... - - · �'-'---

APPEN D I X : MATHEMATICAL T H A C T S 395

Sinc.e the third-order detenninant of the rnatrix is D = 8 � 0, the

rank ris 3, ·which n1eans, since n = :3, that x1 = x2 = x3 = 0 is the only
solution.
Example 2 Consider
2x1 - x2 + 3x3 =0
3x1 + 2x2 + xa =0
X1 + 3x2 · - 2xa =0
;"5x1 + X2 + 4xa =0
The matrix of the coefficients
2 -1 3
3 2 1
A
1 3 -2
5 1 4

contains three third-order detenninants, all of which are found to be zero

on investigation. Inspection shows that r = 2. Since r < n (2 < 3),
there are solutions other than X 1 = x2 = X a = 0. Since we n1ay specify
n - r, that is, 3 - 2 = I , unknowns arbitrarily, set 1 x3 = k, creating a
set of nonhomogeneous equations
2x1 - X2 = -3k
3x1 ·+ 2x2 = -k
X1 + 3x2 = 2k
.5:t1 + :t2 = -4k
and solve any two e(Juations for x1 and :t2. Choosing the first two,
2x1 -· X2 = -3k
3x1 + 2x2 = -k
2 -1
D - 3 =7
2

-3k -1
-k 2 - -6k - k
- ---=-----'
whence X1 - -k
7 7
2 - 3k
3 -k _ - 2k + 9k +k
and X2 = '---7- ---' 7
=
(The solution was carried out by detern1inants for the exercise.)
1 The symbol k is used for the arbitrary value of :r3, which could have been
chosen as unity or any other number, positive or negative. The use of the symbol k
prevents the value assigned to x3 from being lost sight of.
396 K I N E M AT I C SYNTHESIS OF LINKAGES

Example 3 Consider the two ho1nogeneous equations in three

unknowns,
2x1 - 4x2 + xa = 0
3x1 + X2 - 2xa = 0
The matrix of the coefficients
-4
...,1 -
-
1
i I
-2 !
has the rank r = 2. Since r < n (2 < :3), ,ve may specify n - r, or one
unknown.
Set X:i = - k (for variety ,ve use the 1ninus sign). creating a set of
nonho1nogeneous equations
2x1 - 4x2 = +k
3X1 X2 = -2/...:
+
and solve for x1 and x2,

D = 2
3 -1 = 2 + 12 - 14

- - 1I-
k
-2k k - 8k
,vhence x, =- -
14 14
- -½k
k
-- · - -2k
--
- -4k - 3/, - -½k
and
14 14

BlBLIOGRAPHY

Frazer, R. A., ,v. J. Duncan, and A. R. Collar: "Elementary �'latrices," Cam­

bridge lfniversity Press, New York, 1938. Reprinte<l in paperback
edition, 1960.
Haag, J.: "Mathematiques speciales," vol. 1 , chap. 23, Gauthier-Villars, Paris,
1940.
Rouche, E.: Note sur Jes equations lineaires, J. ecole polytech. (Pari.9 ). vol. 29.
no. 48, 1880.
Sokolnikoff, I. S., and R. �f. Redheffer: "::\'!athematics of Physics and �Iodern
Engineering," 1"1cGraw-Hill Book Company, �ew York, 1958.
 PROBLEMS

Chapter 2
2-1 Show the transmission and deviation angles for the two mechanisms, taking
the input to be
a. The link to the left.
b. The link to the ri�ht

D
,,_��-
6 Oo
A 3 1

o,,.
FIGURE p 2-1
398 KINEMATIC SYNTHESIS OF LINKAGES

2-2 Show transmission and deviation angles, assuming

a. Input at link 6, output at link 2.
b. Input at link 2, output at link 6.
2-3 Find the transmission and deviation angles of
a. The direct-contact mechanisn1.
b. Its equivalent linkage.

D
A

1 Os

1''IGURE p 22
- FIGURE p 2-3

2-4 to 2-8
a. Show equivalent linkages.
b. Describe the linkages in the symbolic notation.
c. ,Yhirh fan1iliar chain::- do you recognize?

�
1

w})m.
(b) 1

FIGURE p :l-4
PROBLEMS 399

1
W/////////4

:VUHU//.m.
(a) Radius r

.(b)

FIGURE p 2-5

(a)

l Cz (c)
$2 (b)

FIGURE p 2-6
400 KINEMATIC SYNTHESIS OF' LINKAGES

�
1

FIGURE p 2-7

I 3
¥
1

FIGURE p 2-8
 - - . . ...-- -~
�-o- - -

PROBLEMS 401

2-9
a. l)oes the wedge arrangement represent the mechanism equivalent to
the linkage?
b. Give your reasons.

.,,,....,,..,. . w/#.im#.i.i#u#m
1
(b)

1
(a)

FIGURE p 2-9

2-10 The cylinder C is trapped between the two forks. Considering the motion
of C to be the output, describe the mechanism in terms of the syn1holic
notation.

FIGURE p 2 10
- 1 1
402 KINEMATIC SYNTHESIS OF LINKAGES

Chapter 4
4-1 Determine the centrodes of the rectangular ellipsograph.

o+
_____ 2
B
1

FIGURE p 4-1

NoTB (to be read after completing the problem): The two circular cen­
trodes are known as the Cardan circles; the smaller is half the diameter of
the larger and rolls inside it.

4-2 The sketch shows a graphical construction for determining the magnitude
of the Coriolis acceleration component.a1 Show that it is correct. Note
that this construction adds itself conveniently to the graphical determina­
tion of a R�".

'
_,/

/
/
/
/
W2
fi/
1 FIGURE p 4-2

-1-3 to 4-7 Find all instantaneous centers for the mechanisms shown in Figs.
P 2-3, 2-4, 2-6, 2-8, and 2-9.
1H. Parkus, "lvlechanik der fest.en Koerper," p. 2:3, Springer-Verlag OHG,
Vienna, 1960.
 -- . - . ··- ·-·· - --��-- -
-------- --·· .

PROBLEMS 403

4-8 Determine the direction of the common tangent to the ceiitrodes of links 1
and 3 in the mechanisms shown. Draw the centrodes for a short dis­
tance to either side of the given position. The dimensions are·
a. O.._A = 2.25
AB = oBB = o..oB = 4.5
°
b. O.tA = 1.2 at 40
OAB = 1.5
c. OAA = 1.0 at 40°
AB= 2.8
B

(a)

A
A (b)

----------..:;ti]--
1 ���'-��
FIGURE p 4-8 (c)

4-9 Point Pa of plane 3 is rnoving so that it traces a curve JIKon the moviniz
plane 2; rectangular axes 0?;X2 and 02y2 are fixed to it. The fixed plane 1
is the complex plane with origin at O and axes as shown. The position of
plane 2 is defined by x, y, and -./t; and the position of point P3 with respect to

(Imaginary)

Plane 3

y ------

Fixed plane 1

FIGURE p 4-9 %
· (Real)
404 KINEMA'l'IC SYNTHESIS OF LINKAGES

plane 2 is defined by p and q,. The position of point Pa with respect to

the fixed plane (complex plane) is defined by the vector r, which may be
written in complex form as
(1)
a. Taking the first derivative of Eq. (I) with respect to time, write in
complex form the velocity of Pa with respect to the reference plane.
.
b. This velocity is the sum of two parts: (1) the velocity of P3 with respect
to the moving plane; and (2) the velocity of the coincident point P2
of the moving plane 2. \Yrite the expression for each of these velocities,
and show that their sum equ�!s the total velocity of Pa-
c. Taking the second derivative of Eq. (1) with respect to time, write in
complex form the acceleration of Pa with respect to the fixed plane.
d. This acceleration is the sum of three parts: (1) the acceleration of Pa
with respect to the moving plane; (2) the acceleration of the coincident
point P2 of the moving plane 2; and (3) the Coriolis acceleration.
Write the expression for each of these accelerations, and show that
their sum equals the total acceleration of Pa.
e. Show that the Coriolis acceleration is normal to the trajectory traced
by point Ps on the moving plane 2.

Chapter 5
5-1 Using the Grubler criterion, determine the movability (number of degrees of
freedom) of the mechanisms shown.

(b)

(a)

(d)

(C)

FICilUBE p 5-1
 ·-···-··

PROBLEMS 405

(h)

(i)
(k)

FIGURE P 5-1 (continued)

406 KINEMATIC SYNTHESIS OF LINKAGES

5-2 Determine the movability of

a. The schematic of locomotin• \·al\'e gear.
b. The schemati<' of �Iusgrave $team engine.

(a) •
FIGURE p 5-2

5-3
a. Check the chain for movability.
b. Identify the links as binary, ternary, and so on.
1

Jjl ,c ; -Jr-��-�

,;

FIGURE p 5-3

5-4 Derive all planar revo)ute kinematic chains that have constrained motion
with eight links. Do not consider redundant chains.
 . ----�- --··- ·-- -

PROBLEMS 407
5-5 Shown is the linkage of a Sterling-cycle engine developed in Holland. 1
Demonstrate that the chain is movable. (5, transfer piston; 4, power
piston; 10, compressor; 2, crank.) What are your thoughts about the
ease of balancing this mechanism?
1 Product Eng., Aug. 24, 1959.

l'IGURE p 5-5
408 KINEMATIC SYNTHESIS OF LII'.KAGES

5-6 The essential parts of a folding top (as for a buggy or a conYertible) are
shown in two positions:
a. Folded.
b. "Up."
Investigate each position by Grtbler's criterion.

Fabric
(folded)

(a)
-- C

-------
_
.A fab� c____

I
I
I
I
I
(b) . c., I
.::I
-Q
�I
I
I
I
I
I
I
I
�

FIGURE p 5-6

5-7
a. .-\ppl�· the philo::::ophy that led to the planar Grlibler criterion to
:,;patial ('hains ronnE>rted b�· lower pair:-; of f = 1, and develop the
spatiul-ehuin GrliblPr rriterion.
b. \\'hat i,; the minimum number of link:,; and pair::- for a ron::;trained
sin1ple-clos<>d ehain "?
c. ""hat is the next combination of link:,; and pair:::: for a ('Onstrained chain·?
P R O B L E :'.\-I S 409
5-8 A campstool is pictured in the partially folded state. Apply the Gri.ibler
criterion to three conditions:
a. The stool as shown.
b. The stool with the canvas seat taut.
c. The stool extended a.nd on the ground. The ground "connections"
must be pictured as appropriate kinematic pairs.

FIGURE p 5-8

5-9
a. Determine three accuracy points a 1, a 2, a3 in the interval - 1 < x < 1
for which the derivative of the error is minirnized according to the first
approximation given by the method of Chebyshev polynomials.
b. Same question, but with five accuracy points a 1, a2, a3, a4, a 5.
5-1 O Approximate the polynomial y(x) = 3x3 + l lx2 + 20x + 1 i in the inter­
val O < x < 2 by means of a second-degree polynomial y,. (x) = A.x2 +
Bx + C. Find the coefficients A, B, C giving the smallest possible value
of the maximum error E = y - Ya in the interval specified. ,Yhat is the
n1aximum value of this error, and where does it occur?
5-11 :\pproxiinate the polynomial y(x) = 12x• + 3x4 + 4x3 + 2x2 + 8x + 3 in
the neighborhood of x = 0 (i.e., for values of x close to zero) by means of a
fourth-degree polynomial y,.(x) = Ax4 + Bx3 + Cx2 + Dx + E.
,'>-12 •.\ pproximate the parabola of equation y = x2 in the interval 0.5 < x < 1
by means of a straight-line segment defined by the equation Ya = Ax + B;
find the coefficients A and B giving the smallest possible value of the
maximum error in the interval specified. \Vhat is the maximum error
under those conditions, and where does it occur?
5-13
a. Construct graphically three, four, five, six, and seven points with
Chebyshev spacing on a 4-in. segment.
b. Develop a formula for spacing of the form ak = f(h, n, k), where k = 1,
. . . , n, h is the half interval, and n is the number of accuracy points.
c. Check answers to (a) by means of (b).

Chapter 6
6-1
a. Trace the locus of the coupler point .11 of a ,Yatt-type mechanism.
assuming the proportions of the four-bar linkage to be OAOs = 6.5 in.,
410 KINEMATIC SYNTHE818 OF L I N KAGES

o..A = OsB = 3 in., Ailf = J.1B = 0.75 in. The curve is of the
sixth order.
b. Change the proportions to = AB = v'2
o..oB A = v'2 OsB,
o..
maintaining A2\f = MB, and trace the curve of M. The linkage ha.5
change points: if it runs as a parallelogram linkage, the coupler curve
is a circle, a curve of second degree; if it runs as an anti parallelogram
(contraparallelogram), the curve is that known as Bernoulli's lemnis­
cate, a curve of the fourth order.
6-2 Trace the locus of the coupler point M of the Roberts type of approximate
0
straight-line mechanism shown for three cases: 8 = 45, 60a° , and 75a°.
Note the symmetry, O..A = A21f = MB = BOs and AB = ½O..Os.
Construct also the circles of foci, zind discuss the intersections of the
circles with the coupler curves.

A B

FIGURE p �2

6-3 Trace the locus of point 1lf for the mechanism shown, in which the two
links O.tA and BM constitute what is known as a Chebyshev dyad,
narnely, O.tA = AB = A1lf. Other dimensions are OsB = 0.4(OAA),
and OAOs = 2(0sB). Construct the circle of foci and the cognate
linkages.

FIGURE p 6-3

6-4
a. Recast the four-bar coupler-curve equation [Eq. (6-1)] by (1) sub­
stituting expressions containing k for sin a and sin B and (2) writing
the result in the form ij'J. + V2 = W2 •
 - --- - - ----- ···--·- ·- - - ··- . . ... . .

PROBLEMS 411
b. Consider the case in which the coupler point M lies on the line AB
between A and B. Deterinine appropriate values of 'Y and c, and
evaluate U, V, and W.
c. A special case, reminiscent of the ,,,att linkage, comes with setting
r = s and placing 1'1 at the midpoint of AB. Evaluate U, V, and W.
d. Consider the case in which the coupler point .1'1 lies on the line AB,
but beyond either A or B. Evaluate U, V, and W.
e. Determine the conditions for a coupler curve symrr1etrical with respect
to the x axis.
6-5 Derive the equation of the lemniscate, and compare with the usual forms of
the equation (x2 + y2) 2 = a2(x2 - y2) or p2 = a2 cos 28, where the origin
is taken at the node (see page 80). Determine the parameter a in terms
of linkage dimensions.
6-6 Determine the four-bar cognates of the linkages shown.

2.6"
(b)

FIGURE p 6-0

6-7 Determine the six-bar cognates for the linkages of Fig. P 6-6.
6-8
a. Find two other linkages having coupler curves identical with the path
of M.
b. If in the given linkage the member OBB is driven at a constant angular
velocity, which frame link of each cognate may be driven at the same
velocity to have each cognate's M describe the curve at the same rate
as the given M?

2.85"

FIGURE p 6-8
412 KINEMATIC SYNTHESIS O F LINKAGES

6-9 Show that the locus of point ill of the Free,nantle mechanism is an exact
straight line. The dimensions are OA.A = A. B = A. .ilf. A modified ver­
sion of this mechanism came to be known as the Evans ''straight-line"
,notion.

FIGURE p 6-9

6-10 Give a direc-t µroof that the Peaucellier mechani:,;m gives an exact straight
line .
6-11 The cubic of stationary curvature (see See. 7-2) i::; a C'urve defined in polar
coordinates by Eq. (7-10).
a. l)etennine its equation in reetangular coordinates and verify that it i,-;
of the third degree.
b. l)eter,nine the direction and location of it.-:- asymptote.

Chapter 7
7-1 Determine the inflection circle for link 3 of the n1echanisms in Fig. P 4-8.
7-2 Determine the inflection circle for the planetary gear train, the ratio of
diameter::, being 1 : 2.

Jo"IGURE p 7-2

X OTB (to be read after finishing the problem): A point on the pitch
circle of the srnall gear traces a hypocycloid ; the arches in this case are
two, and they are straight, having reached the limit by becoming diameters
of the large gear. In other words, a point on the pitch line of the small
gear traces a straight line. This circu1nstance seem� to have been first
exploited for a true straight-line mechanism by Jame� \Vhite, an English
engineer, who received a medal for this inv0ntion from Napoleon in 1801
on the occasion of an industrial exposition in Paris.
PROBLEMS 413

7-3 Consider the circular centrodes of diameter ratio I : 2. Develop a graphical

construction for the inflection circle.

+C 11'2

FIGURE p 7-3 +o

7-4 A Bobillier construction for the inflection circle of a four-bar linkage is as

follows:
a. Locate / from the intersection of O....A and OBB.
b. Locate l' from the intersection of OAO» and AB.
c. Draw II'; this is called the collineation axis.
d. Through I, draw a line parallel to l'OAO».
e. Extend AB to intersect the line of step d in M.
f. Draw a line through M parallel to II'.
g. The intersection of the line of step f with AOA (extended if necessary)
marks JA. Js is found similarly.
h. Locate the center of the inflection circle from the three point!--/ . .r.�,
and J»-lying on the circle.

FIGURE p 7-4

Show that this construction satisfies the Euler-Savary relations.

414 KINEMATIC SYNTHESIS OF LINKAGES

7-5 A construction for the inflection circle of a slider crank is as follows:

a. Locate /, the instantaneous center of links 1 and 3.
b. Dra, w two arbitrary but parallel lines u and w. through o,.._ and /.
c. Draw any ronYenient line through .4 . <lefinin11; points G and H.
d. Draw IH.
e. Through G, draw a line parallel to JH
f. Extend O ,.._A. to intersect the line of step e in J,.._.
g. Locate the center of the inflection circle from the three points-I, B,
and J,.._-lying on the circle.
Show that this construction satisfies the Euler-Savary equation.

(Note that B, undergoing rectilinear translation, describes a curve having

an infinitely great radius, whence B is a point on the inflection circle.)

u B
�

.t-'IGURE P 7-5

7-6 Locate the inflection circle of the four-bar linkage shown, an<l inve;;tigate
the coupler curves of
a. Point C.
b. A point D lying at the intersection of A C and the inflection circle.
c. Point K of the inflection circle.
Use a transparent overlay for plotting the curves.

FIOURE p 7-6
PROBLEMS 415

7-7 Consider a plane 2 moving with respect to a fixed plane 1. The fixed and
moving centrodes are 1r1 and 1r2. The angular velocity at the instant
considered is w, and the instantaneous center is at I. A curve C2 of plane
2 as it moves with respect to plane 1 generates an envelope C1, and the
point of contact (at the instant considered) is llf. Note that point M
moves with respect to both planes 1 and 2. Let
vM11= velocity of JI with respect to plane 1
= velocity of !lf with respect to plane 2
v,v12

Then V.u11 = v.,112 + wl with JJf = l (1)

Note also that, since C1 is the envelope of C2, their common tangent is
perpendicular to the line I11[. Let t be the unit tangent vector to C1
and C2 ; let � be the unit tangent vector to 1r1 and 1r�. The angle i/; and
the distance l define the point M.

)'IOURE p 7-7

The Euler-Savary equation for envelopes relate1S the curvature of C1 to

that of the generating curve C2 (the radii of curvature are Jf0 1 = R1
and 1'102 =·R 2) . The equation is to be derived from Eq. (1) by obtaining
two independent expressions for v_,,1 1 and VM/2·
a. Show that
v1 sin 1/t
VMfl = l + l/R1 (2)

HINT : Using complex-number notation for vectol'8, take the deriva­

tive with respect to time and substitute.
b. Show that

VM/2 = 1u,+sinl/R1/t2 (3)

c. Making use of Eqs. (1) to (3), derive the Euler-Savary equation

for envelopes,
1 1 w 1
102 - ro. = v, sin "' (4)
416 K I NEMA T I C SYNTHES I S OF L I N KAGES

7-8 Consider a rnoving plane 2, moving with respect to a fixed plane l . The
fixed and moving centrodes are 1r1 and 1r2, their centers of curvature being
located at Q1 and Q2, respectively. ..\t the instant considered the angular
velocity is w, the instantaneous center is at /, and the IC velocity is
v1, as shown in the figure.
a. Express the acceleration of the instantaneous center in terms of v, and
w. ,Yhat is the direction of the acceleration?
b. Using Coriolis' theorem applied to the acceleration of the point of
contact I as it 1noves along both centrodes, derive the equation
1 1 w
/Q I =
(1)
/U2 - v,
c. Compare this equation with the Euler-Savary equation found 111
Prob. i-7, and show that the latter may be written in the forrn

,�-� - 1� 1 = (1t-2 - 1�·1 ) si� i/t (2)

I
I
+n2
I
I
I
I
I
I
I

1,
I
I
I
I
I

I
I
I
+n, FIGl'RE p 7-8
I

7-9 A plane 2 1noves with respect to a fixed plane 1. The fixed and n1oving
centrode,,; are 1r 1 and 71"2. At the instant considered, the instantaneous
center is l, the angular velocity is w, the angular acceleration is a, and the
inflection circle (of diarneter /K) is as shown. A point A of thC' ltlOYing
plane has tangential and normal accelerations a.t1 and a.�". it::- po�itiou
being defined in polar coordinates by r = IA and 1" = TIA . \Ye know
from Sec. 7-1 that. the inflection circle is the locus
. in plane
2 of. point::,
having zero normal acceleration.
a. By similar reasoning, using Eq. (7-3) in particular, show that the locu�
of points having zero tangential acceleration is also a circle with
<lia1neter IK' = w2/K/a. This circle is called the Bresse circle.
J> H O D L E M S 417
K
A

----
C B
I
K' !
I
T I
I
/
...
i.- �'-------'-�,:.i·
aA'

FlGl.RE p 7-9

b. The intersection C of the circles of inflection and of Bresse is an instan­

taneous center of accelerations (zero normal and tangential acceleration
at this point). L, I also a center of acceleration?
c. A geometric property of the motion between planes 2 and l is one which
i::; independent of the time law (independent of the angular velocity
and arceleration) but depends only on the centrodes. A kinematic
property, on the other hand, depends on the centrodes as well as the
tin1e law. State which of the following properties-instantaneous
center of velocity, instantaneous center of acceleration, inflection
circle, BressP. circle-are geometric properties and which are kinematic
properties.
d. Again using Eq. (7-3), show that the acceleration of any point A may
be written in the form
aA = (ia - w2)CA
so that, if the acceleration of one point A. is known, the acceleration of
any other point B may be found by the construction shown. Kote
the similarity and differences of this construction in comparison with
that of velocity using instantaneous centers.
7-10 .\ four-bar-linkage vertical posthole drill rig is to be developed. The
figure �hows the sche1natic. The idea is that the drill is to be :-:uspended

L.----- Path

n
Kr
- - ��---- --- - �l
9'

/
A/ I
5'
t
I OA
I �

r- "
2.2·: �1.7'
t- - - - - --��
FIGURE p 7-10 I 4.5' . OB
418 K I N E M A 'l' I C S Y N T H E S I S O J<' L I N K A G E S

from a coupler point having an approxin1ate straight-line motion in the

vertical. It will be advantageous to choose the coupler point on the
normal through J, that is, at the inflection pole K. The links O..tA and
0BB are to be determined and the "vertical" section of the coupler curve
traced.
7-11 Design a wall-mounted level-luffing crane similar to that of the figure.
Bear in mind the discussion of Prob. 7-10.

r --- - - - - - -�
10.25'
I .,.- / l

11 • A, / ,,,, / /
'I18.25'
14 25 ' / ,.,. / I
I /
/
I
I
I
1
- ---'"n -- Path
K

7-12 Determine the centers of eurvature of the coupler curve shown at points C1,
c�,and C3. 1·,,<' either numerical calculations or graphical constructions.

�
OB },'JGt.iRE P 7-12

7-13 Determine the cubic of stationary curvature for the coupler of the linkage
of Fig. 7-7.
PR OBLEMS 419

7-14 Consider the cycloid generated by the point A on the rim of a wheel of
radius r rolling on a straight path. The parametric equations of the
coordinates of A are

;i; = r( (J - sin 0) and y = r( l - cos 0)

a. DeriYe the values of the radii of curvature uf the cvcloid when O has
values of 0, 90, and 180°. ,\lso, locate each center of curvature.
b. Compare your answers with values derived from the Euler-Savary
equation.

l'IGURE p 7-14 I

Chapter 8
8-1 Determine fixed pivot$ OA and Os of a four-bar linkage OAAB0 8 to guide
the door shown from closed to open positions.

Door in
closed
position

,..---- - - - --- --,

'-:--,----
,O, ----TT
... ... ,o,
._,,

p 8-1
A2 B2
J.l'IGURE Open position

8-2 Design a four-bar linkage to convert a clockwise rotation of 60° into

a. A. clockwise rotation of 90° .
b. A. counterclockwise rotation of 90°.
420 K I N EMATIC SYNTHESIS OF L I N KA G E S

Draw the mechanisn1:,; in their extreme positions, and check your answer
by means of a simple model.
8-3 Design a four-bar linkage to produce a follower displacement of i/t 2 = 60° 1

clockwise for a crank displacement of <1>1� = 100° counterclockwise.

8-4 Detern1ine the length of the coupler and follower (i.e., find the coupler
pivot B1) of the four-bar linkage OAABOs for which two successive rota­
tions q, 12 = 15° and <Pn = 45° of the crank OAA produce, respectively,
two ::-uccessive rotations i/t 2 = 43° and Y12a = 47 of the follower OsB .
°
1

.\11 rotations are clockwise.

FJGl'RE p 8-4

8-5 Design � central ,.;li<ler-crank mechanism O.�AB so that a 90° clockwise

rotation of link o.◄.4. n>,.;uJt,- in a rectilinear tran,.;lation of point B frorr1 left
to right along the line ,;hown. Di,.;plaee1nE>nt,- of B along the line n1ust
be approximately proportional to the eorre,;pon<ling rotations of link O.◄A.
within thE> 90° inter\'al. t·_,f' threE> aceuracy points with ChebysheY
spacing. To cheek your result, draw a dbplacement eurve showing the
translation as a function of the rotation for seven equidi:;tant points.
Con1pare the graphical and :;tructural errors. \Yhat would be your sug­
gestions to in1prove this synthesis'?

•Oc
-o-

I
;I.l.
1 0,◄
- -I. -- -- -·· - �"'''°"''"°"'°"'°"''-'�°"'�""�""°"'�'-'-�'1,
Path ofB
�
I:=1.3· - -►.'
3· ____..,....____ 3"

Fl(ll'RK p S-5

8-6 Referring to the n1echanisn1 de::.igned in Prob. 8-5, find a point C of the
roupler who,_l, three po,-itions C" C2, C3 eorre:-::ponding to the three accuracy
points lie on a cirde centE>red at Oc. If a rigid link OcC is now introduced
and the ::.lider ren10\·ed, 0 .4,4 COc forms a four-bar linkage whose coupler
point B traces an approxi,nately straight. segment. Check your design
by construction of the locus of point B.
8-7 If the crank OA A of the slider-crank mC'ehanism OAAB shown in solid lines
is rotated with constant angular velocity, the time Atd required for the
direct stroke (i.e., left to right) will be the same as the time At, required for
the return stroke (i.e., right to left). In a quick-return mechanism it is
PROBLEMS 421
desired to have a ratio ll.td!ll.t, greater than unity, and this may be achieved
by driving the crank OAA at a variable angular velocity. In the present
problen1, the unknown four-bar linkage OcCA.O,.., shown in broken lines,
is to be used for this purpose. Find the dimensions OcC and CA for which
ll.td/ll.t,. = 2 when the crank OcC i$ rotated at constant angular velocity.

A
-- --- -
c
q--
'
-- --­
I
I
'I
I
" -- ·-I B--+·-
�w.."""""""""�
09-6·
FIGURE p 8-7

8-8 The figure shows the principle of a lift truck in which the fork is guided by a
linkage. The fork must move up and down in nearly rectilinear transla­
tion. Determine the dimensions of the required linkage. Note tha.t
OAABOs is a four-bar linkage, and that point C should have an approxi­
mate straight-line path. Link DE maintains the fork in a motion of
translation: to detern1ine this link, use three positions of accuracy for
which the fork is to be exactly horizontal at ground level, halfway up. and
all the way up. The points O,.. and On may be chosen in any convenient
location within the 4- by 8-ft rectangle. Keep in mind that the truck
must have maneuverability in close quarters where the headroom is low.

,- tI
I I
6"
I
I

I 4'
I +
I

k-- - - -
- s· - - -
t'lGlJRE p l:> - 8
422 KI NEMATIC S Y S T H E S I S OF LINKAGES

8-9 The figure shows the principle of the linkage guidance of a mixer motor.
The motor must move up and down in nearly rectilinear translation.
Determine the dimensions of the required linkage for three positions of the
motor, lowest., midway, and highest. Keep space requirements in mind
when rhoo�ing your links.

rs·

HGvRE p 8-9
PROBLEMS 423
8-10 An overhead garage door is to be supported by a four-bar linkage. The
sketch shows the two positions of the 7-ft-high door.
a. Lay out the linkage, and show what brackets are necessary, and where.
Brackets may be fastened directly t.o the 2- hy 4-in. door frame and the
2- by 6-in. doorpost.
b. Provide for a helper spring to aid in opening the door and the mainte­
nance of the open position.

%¥Hi @
t. f%�& A
f; i �
��;t. t.tt��r1£�1.���' 1
Ht. &::���!:!��;:t�1;t1lM}t.�@MS:
f.::r.@tjf�t�llf.Brn :
I
I
I

I
I
I

I
I
I
I
I
I
I
I
I
I
I
I
I
I
I

I
I

Post, 2 X 6

I
I

I
I

I
I

'
•
L-

FIGURE p 8 - 10
424 K I N E M A 'f l (' S Y NTHESIS O F LINKAGES

8-11 The four-bar linkage O,.A BOB assumes three positions as shown . Find
a point C of the coupler AB that will assume three positions Ci, C2, Ca
equidistant from Oc when the linkage takes its three positions. Locate
the positions C1, C2, Ca of C. and check your answer.

1 3.30"

I 1.87" 1.05·
23°
\.1
.--,
,o..v -- - - � - �
� OA Oc 08
t'JOtTRE p 8-11

8-12 The level-luffing crane shown is a European type, u:::ed for shipbuilding
and cargo handling . Note that point. C is a coupler point of the four-bar
linkage O,. ABOa. Locate the pivot point A so that the coupler point C
will pa1,s Pxac-tly through posit.ions C1, C2, and Ca,

Water

FlCICllE p 8-12
PROBLEMS 425

Chapter 9
9-1 \\'hich of the quadrilaterals shown are opposite-pole quadrilaterals? For
those which are, indicate the two pairs of opposite side:-i and the two
diaionals.

Ras
Ras

FIGl:RE P \} - ]

9-2 Six pole:;; are given, corresponding to four positions of a moving plane.
Select an opposite-pole quadrilateral, and construct the center-point curve.

'
2.55H I 2.63#
82" i
1.7 1H
I

.:.

j
i l IE 1.89" - -··
iI H

'
� 2.25
2.72"
I..:--- H -
- 3.25"
FIGURE P 9-2 1"4-- ----- 4,75
426 KINEMATIC S Y N T HESIS O F LINKAGES

9-3 Construct the center-point curve for the relative poles of the sinusoidal­
function generator (Sec. 9-4).
9-4 The figure shows a four-bar linkage 0,1ABOs and four positions of its crank,
O.tA 1, 0,1A 2, 0,1A3, 0,1A.. For these four positions,
a. Construct five of the poles of the coupler with respect to the fra,me.
b. Select an opposite-pole quadrilateral, and draw the center-point curve
in the vicinity of point OB-
e. Using a center point D l in. from O B, and abo,·e O B, determine a coupler
point C tracing an approximate circular arc as the crank rotates from
position O,1A I to position O.tA •.

OB FHHlRE p 9-4

9-5 Determine the equation of the center-point curve defined by the opposite­
pole quadrilateral P 1zP, 8P34P2•• Take a rectangular system with x axis
along the side P12P13 and origin 1nidway between P12 and P 13.
9-6 Find the following algebraic properties of the cente r -point curve considered
in Prob. 9-5:
a. Degree.
b. Real and imaginary points of infinity.
c. Real and imaginary asymptote:,;.
d. Real intersections of real and imaginary asymptote:,;.
9-7 Fro1n the results of Prob. 9-6, show that the real asymptote of the center­
point curve may be obtained by joining the midpoints 1n' and m l' of the
diagonals P12P34 and P1 �2• of the opposite-pole quadrilateral.
PROBLEMS 427

Chapter 1 0
10-1 Design a four-bar linkage to generate the function y = z 1 . 0 for an interva.l
in x from 1 to 4. The input link is to start from 30° and is to have a range
of 90° ; the output link is to start from 90° and is to have a range of 90° .
Use three-point Chebyshev spacing.
10-2 Design a four-bar linkage to generate the function y = xt . 8 for an interval
in x from 1 to 5. The input link is to start from 315° and is to have a range
of 90° ; the output link is to start from 20° and is to have a range of 70° .
Use three-point Chebyshev spacing.
10-3
a. Using Eq. (10-4), develop formulas for the mechanical errors .::1s in s
due to errors Aa1, Aa2, Aaa in the parameters a,, a2, aa of the slider-crank
mechanism.
b. \Vhat is the value of the error As for <I> = 45° in the mechanism defined
by the parameters
a, = 5 + 0.004 in. a2 = 10 + 0.004 in. aa = +0.004 in.
10-4 Design a slider-crank mechanism (Fig. 10-8) in which the translation of B
parallel to Ox from s; = - 1 in. to s1 = 3 in. is approximately proportional
to the rotation of OAA around OA from <J,; = 60° to <J,1 = 120°. Use three­
point Chebyshev spacing.
10-5 Design a slide r -crank mechanism OAAB (Fig. 10-8) to convert a counter­
clockwise 90° rotation into a 4-in. rectilinear translation from left to right.
The velocity ratio between the translation and the rotation must be kept
approximately constant. Take the accuracy points at 0, 1, 3, and 4 in.
The initial crank angle q, 1 is to be 70° . Check your result by means of a
graphical construction.
10-6 Design a four-bar linkage to meet the following specifications of position,
velocity, and acceleration:
<J, = 900 t/1 = 90
°

w4, = 3 rad/sec w.,, = 1.20 rad/sec

Ot4> = 0 at = 1.62 rad/sec2

Chapter 1 1
1 1-1 Deter1nine the proportions of a four-bar linkage that will 1n one of its
positions satisfy the specification;;
w, = 8 rad/sec a, = 0
w2 = 1 rad/sec a2 = 20 rad/tieC2
wi = 3 rad/sec aa = 0
428 K I N EMATIC SYNTHESIS OF LINKAGES

11-2 Determine the proportions of a four-bar linkage that will in one of its
positions have parallel crank and follower and satisfy the specifications
w1 = 3 rad/sec a1 = 0
ws = 1 rad/sec a2 = 1 rad/sec2
aa = 0
1 1-3 Devise a method, using complex numbers, to determine the dimensions of
a slider-crank rnechanism and the positions of its links, for which the
following are specified:
w, and a 1 angular velocity and acceleration of the crank
w2 and a2 angular velocity and ac\,,eleration of the coupler
v and a linear velocity and acceleration of the slider

Use the notation shown in the figure, and summarize your findings in
tabular fashion.

lmag axis

I
•1 1&
3

e I
��� - ---- --'------
- 8
Real axis 1-'IGURE p 11-3

Chapter 1 2
12-1 Evaluate the meehanieal error of the 2R, 2G function generator y = co,- x
of Se<'. 12-5 at the first accuracy point. Assume deviations as follows:
Aa2 = 0.001 in., As, = 0.001 in., Aa4 = 6 min. Values of pararneterR
given in Sec. 12-5 are in inches.
1 2-2 Extend the method of synthesis of the 4R spherical mechanism (Sec. 12-4)
to the case of four accuracy points. Take a, = 90°, and use 82 1, crank
position corresponding to the first accuracy point, as a fourth design
pararneter.
12-3 Design a 4R spherical mechanism to generate the function y = cos x in the
interval -60 < x < 60°. Ranges of variation are to be A81 = 180° and
A82 = 50°. with accuracy points at :i: = -45, 0, 45°. The maximum
value of 82 (eorresponding to the second accuracy point) is taken a.s
8t2 = 50° .
 - -- -- -

PROBLEMS 429
12-4 Determine the displacement equation relating s and () in the R, 2G, P
mechanism shown. ,vhat choice of parameters will produce a mechanism
capable of generating symmetrical functions·.>

G2 rotates in xy plane

FIGURE p 12-4

12-5 Develop a method of synthesis of the n1echanism con::::idered in Prob. 12-4

to generate symmetrical functions with a total of five accuracy points.
Use s1 (value of s corresponding to the first accuracy point), r, and a as
design parameters. .-\pply the method to the function y = x2 in the
interval - 1 < x < 1, with zo = 0, a = 30° . Ranges of ,·ariation a,re to
be f:l(} = 180° and As = 3 in.
12-6 Develop a method of ,-ynthesis for the 4R ::;pherical mechanism with four
accuracy points, using a i , a2, aa, a4 as design parameters. Does this
synthesis procedure require the use of con1patibility equations?
!

Acceleration, 117
Coriolis, 120, 379, 402, 403
difference, 120, 125
image, 20, 128
normal, 119
relative, 120
tangential, 119
Alt, H., 47
Ampere, A. M., 14, 22
Analysis, kinematic, 88, 130
Archimedes, 3, 70
.\.rea contact, 34
.-\ronhold, 19, 24, 106
Asymptotes, 154, 156
four-bar coupler rurve, 159
Autofocus mechanism, 184
Bennett mechanism, 52, 85, 350
Bernoulli, Johann, 15
Bernoulli, lemniscate of, 80, 411
Betancourt, 12
Block, S. S., 170
Bobillier construction, 413
Bock, A., 48
Body, resistant, 30
rigid, 8, 30
Borgnis, 12, 14
Bottema, 0., 198n.
Bresse, 204
Bresse circle, 416
Bricard mechanism, 85
Burmester, L., 19, 20, 22
INDEX
Cardan circles, 402, 412
Cardan universal joint, 50
Carnot, L. N. M., 11, 23
Cayley, A., 170, 172, 174
Centrode, 113
Chain, constrained, 133, 134
kinematic, 53, 348
movable, 134
Chasles, lVI., 15, 23
Chebyshev, P., 22, 23, 176
Chebyshev, straight-line motion, 146
(See also Roberts-Cheb,vshev theorem)
Chebyshev polynomials. 22, 142, 144
Chebyshev spacing, 142, 147
Cinematique, 14
Circle, Bresse, 416
inflection, 194, 199
Classification, 16, 18, 29
Cofactor, 386
Collineation axis, 413
Compatibility equation, 309, 331, 337
Complex numbers, 321. 370
conjugates, 377
Cornponents, mechanical, 2
Connections, 31
cylindric, 34
planar, 34
prismatic, 34, 60
revolute, 34
screw, 34
spheric, 34
Connectors, 31, 32
higher-pair, 32
431
432 KINEMATIC SYNTHESIS 01'' LI:"llKAGES

Connectors, lower-pair, 32 Directions, normal, radial, tangential.

wrapping, 32, 37 transverse, 93
Constrained chain, 133, 134 Di�placement, 92
Coriolis, De, 12, 14
Coriolis acceleration component, 120,
379, 402, 403 Eccentric, 59
Coupler, 39, 45 Element, I 7, 35, 347
curve;· 40, 46, 74 full, 35
point, 40 hollow, 35
Cramer's rule, 383, 388 interrupted, 37
Crank, 2, 39, 40, 45, 77 Elliptic trammel, 56, 402
Crosshead, 34 Engine, 28
Crunode, 152, 162, 164 Newcomen atmospheric, 14
Cubic of stationary curvature, 195, Watt rotative, 15, 41
207 Equations, compatibility, 309, 331. 337
Curvature, center of, 194 homogeneous, 394
centrode, 416 linear algebraic, 380, 389
envelope, 203, 415 symbolic, 343, 348
path, 203 Error, derivative of, 142, 281
radius of, 194 graphical, 299
stationary, 205 mechanical, 299, 315
Curves, center-point, 263 structural, 141, 230
circle-point, 263 veloeitY• · ' 144
circling-point, 195 Euler, L., 7, 8, 23, 204
coupler-point, 15, 16, 40 Euler-Savary equation, 94, 200, 201
asymptotes of. 1.59 Euler's theorem, 374
atlas of, 186
equation of, 151
symmetry of, 166 Farish. \V., 70
triple generation of, 161 Ferguson, E. S., 22
second order, 154 Follower, 39, 45
asymptotes of. 156 Four-bar linkage, 19, 38
three-bar, 150 cognates of, 168
Cusp, 152, 162 four-bar, 169, 337
six-bar, 179
displacement equation, 296, 297
D' Alembert, 8 Frame, 38, 45, 92
Davis steering gear. 66 reference, 31
l)e Coriolis, 12, l 4 Freedom, degrees of, 133
De Jonge, A. E. R .. 21, 22 Freudenstein, F., 22
De la Hire, 204 Freudenstein equation, 297n.
De Moivre theorem, 376 Function generator, 40, 138, 225
Description, symbolic, 65
Determinants, 380
characteristic, 391 Garnier, R., 204
principal, 391 Gauss-Argand plane, 375
Deviation angle, 48 Goldberg mechanism, 85
INDEX 433

Goodman, 20 Leupold, J., 7, 24

Grashof criterion, 77, 134 Line contact, 38
(}ri.ibler, M , 21, 24, 134 Link, 52, 348
Link-error equation, 317
Linkage, 39
Hachette, 12 cognate, 168, 176, 179
Hall, A. S., Jr., 195, 210 deltoid, 78
Hammer, trip, 6 dwell, 81, 286
Hart, H., 170, 186 equivalent, 63, 168
Hero of ..\lexandria, 2, 4, 24 spatial, 343
Hire, De la, 204
Hodograph, 119
Hoist, mine, 8 :VIachine, 1, 2, 28
Hooke joint, 16, 50, ;354 elements of, 11
simple, 2
1\1atrix, 390
Image, acceleration, 20, 128 method, 347, 352
velocity, 20, 101 rank of, 391
Imaginary quantities, 372 Mechanical components, 2
Inflection circle, 194, 199 Mechanical powers, 2
fnstantaneous center, 15, 103 Mechanism, 1, 18, 28, 54
absolute, 106 Bennett, 52, 85, 350
acceleration, 417 Bricard, 85
relative, 106 crank-and-rocker, 55, 77
Inversion, geometric, 180 direct- and indirect-contact, 108
kinematic, 18, 55 double-crank, 55, 77
lnversors, 181 double-rocker, 55, 77
Isometric perspective, 71 drag-link, 55
Galloway, 78n.
Goldberg, 85
Joint, ball-and-socket, 34
planar, 50
.Jonge, de, A. E. R., 21, 22
planar four-revolute, 350
quadric-crank, 40
Kant, 8 slider-crank, 44
Kardan universal joint ·(see Cardan spatial, 50, 343
universal joint) spherical, 50
Kennedy, A. B. \V., 17, 19, 24, 35, 106 spherical four-revolute, 350. 356
Kinematic analysis, 88, 130 straight-line, 179
Kinematic chain, 53, 348 sy1nmetric-, 362
Kinematic model, 68 two-revolute, two-�pheric pair, 344,
Kinen1atic :synthesis, 18, 21, 88, 130 358
Kinematics, 8, 14, 15 i\Iehn1ke, 20
Kinetics; 8 I\1ill, grain, 5
l\1inor, 385
Model, kinernatic, 68
Lagrange, 7 Mohr, 20
Lanz, 12 Moivre, De, theorem, 376
434 KINE.MATIC SY:-iTHESIS OF LINKAGES

Monge, G., 11, 12, 24, 29 Pole, 110, 218

l\Jotion, absolute, 31, 92 image, 253
coupler-point, 7, 19 inflection, 199
planar, 49 opposite, 257
relative, 31, 92 relative, 221
spatial, 50 slider-crank mechanism, 236
three-bar, 8, 19 triangle, 249
variable, 32 Polygon, vector, 101, 128
Movability, criteria of, 21, 134 Polynomials, Chebyshev, 22, 142, 144
Movable chain, 134 Powers, mechanical, 2
Pressure angle, 48
Prony, 16, 150, 175
Nerge, G., 48
Notation, 91
symbolie, 19, 64, 343 Quadric-crank mechanism, 40
Quadrilateral. opposite-pole, 257
Obelisk, Luxor, 18
Vatican, 12
Odometer, 4 Rankine, ,v.
: J. lVI., 17, 25
Oldham coupling, 55 Rapson's slide, 66
Restraint, 135
Reuleaux, F., 2, 17-19, 25, 29, 34, 35,
Pair, 17, 35, 347 38, 64, 75
higher, 17, 38, 138 Revolute, 34
lower, 17, 35 Roberts, Richard, 16, 25
variable, 35, 347 Roberts, Samuel, 8, 19, 22, 25, 150
wrapping, 36 Roberts-Chebyshev theorem, 154, 168,
(See also Connectors) 337
Parallel-motion, 41, 74 extensions of, 176
Peaucellier, 16 Hocker, 77
Poinsot, L., 15, 23 Rotary engine, 57
Points, accuracy, 139, 140
spacing of, 140
cardinal, 251 Sarrus' rule, 384
center, 257, 260 Savary, 204
change, 78 Scotrh yoke, 55
circle, 257, 262 Singular foci, 160, 161
circular at infinity, 156n. circle of, 152
coincident, 89 Slider-crank mechanism, 44
cyclic, 156 central, 45
dead., 325 cognate, 176
double, 162 displacement equation of, 306
imaginary, 154 eccentric, 45
infinity, 155 offset, 45
multiple, 152 poles of, 236
precision, 139 Smith, R. H., 19, 20, 26
INDEX 435

Speed, 93 Universal joint, Hooke, 16, 50, 354

Straight-line mechanism, 179
approximate, Chebyshev, 180
Veldkamp, G. R .. 198n.
Evans, 43, 180
Velocity, absolute, 97
Watt, 41, 180
angular, 93
true, Hart, 185
difference, 100
Peaucellier, 16, 181
displacement of IC, 196n.
Structure, 53, 132
IC, 115, 196
Symbolic description, 65
image, 101
Symbolic notation, 19, 64, 343
linear, 93
Synthesis, algebraic methods, 295
pole, 196n.
approximate, 138
polygon, 101
dimensional, 19, 21, 132, 138
relative, 97, 100
exact, 138
Vinci, Leonardo da' 5
geometric methods, 215
Vitruvius, P. �1., 2-5, 26
kinematic, 18, 21, 88, 130
number, 19, 21, 132
type, 19, 21, 132 Watenvorks, Lon.don Bridge, 10
Marly, 11
Watt, J., i, 8, 26
Three-center theorem, 106 Whitworth quick-return 56
'fransmission angle, 47, 319 Willis, R., 8n., 16, 26, 29, 70