CHAPTER 1

I NTRODUCTI ON
1.1. Types of Computi ng Mechani sms.Computi ng mechani sms
may be di vi ded i nto two di sti nct types: ari thmeti cal computi ng machi nes,
fami l i ar to the l ayman through thei r common use i n busi ness offi ces, and
conti nuousl y acti ng computi ng mechani sms and l i nkages that range i n
compl exi ty from si mpl e cams and l evers to enormousl y compl ex devi ces
for the di recti on of naval and anti ai rcraft gunfi re.
The ari thmeti cal computi ng machi nes accept i nputs i n numeri cal
form, usual l y on a keyboard, and wi th these numbers per for m the si mpl e
ari thmeti cal operati ons of addi ti on, subtracti on, mul ti pl i cati on, and
di vi si on-usual l y by the i terati on of addi ti on and subtracti on i n counti ng
devi ces. The resul ts are fi nal l y presented to the oper ator , agai n i n numer-
i cal form. I n thei r si mpl est forms these machi nes have the vi rtue of
appl i cabi l i ty i n a wi de vari ety of computati ons, i ncl udi ng those requi ri ng
ver y hi gh accuracy. By el aborati on of these devi ces, as by the i ntroduc-
ti on of punched-tape control , thei r possi bi l i ti es for automati c operati on
can be greatl y i ncreased. Characteri sti c of thei r operati on, however , i s
thei r producti on of numeri cal resul ts by cal cul ati ons i n di screte steps,
i nvol vi ng del ays whi ch are al ways appreci abl e and may be ver y l arge
i f the requi red cal cul ati on i s of compl ex form.
Conti nuousl y acti ng computi ng mechani sms are l ess fl exi bl e and have
l ess potenti al accuracy, but thei r appl i cabi l i ty to the i nstantaneous or to
the conti nuous sol uti on of speci fi c probl ems-even qui te compl ex ones
makes them of great practi cal i mportance. They may ser ve as mer e
i ndi cators of the sol uti ons of a probl em, and requi re further acti on by
human agency for the compl eti on of thei r functi on (speedometer, sl i de
rul e); or they may themsel ves pr oduce a mechani cal acti on functi onal l y
rel ated to other mechani cal acti ons (mechani cal gover nor s, automati c
gunsi ght).
Conti nuousl y acti ng computers fal l i nto two mai n cl asses: functi on
generators and di fferenti al -equati on sol vers. Functi on generators pr o-
duce mechani cal acti onsusual l y di spl acements or shaft rotati onsthat
are deti te functi ons of many i ndependent vari abl es, themsel ves i ntro-
duced i nto the mechani sm as mechani cal acti ons. Si mpl e exampl es of
such mechani sms are gear di fferenti al s, two- and three-di mensi onal cams,
sl i de mul ti pl i ers and di vi ders, l i nkage computers, and mechani zed nomo-
grams. Computers of the second cl ass gener ate sol uti ons of some defi ni te
1
1
2 I NTRODUCTI ON [SEC.1.2
di fferenti al or i ntegrodi fferenti al equati onoften an equati on that
i nvol ves functi ons conti nuousl y determi ned by vari abl e external ci r-
cumstances. El ementary devi ces of thi s type are the i ntegrators, com-
ponent sol vers, speedometers, andpl ani meters.
From these el ementary devi ces one can bui l d up compl i cated mecha-
ni sms that per for m el aborate cal cul ati ons. We may menti on thei r
appl i cati on i n gunsi ghts, bombsi ghts, automati c pi l ots (for ai rpl anes,
submari nes, shi ps, and tor pedoes), compensators for gyr oscopi c com-
passes, ti de predi ctors, and other robots of vari ed types.
The present vol ume wi l l deal onl y wi th the probl em of desi gni ng con-
ti nuousl y acti ng computi ng mechani sms.
1.2. Sur vey of the Probl em of Computer Desi gn.-There i s no set
rul e or l aw for the gui dance of a desi gner of compl ex mechani cal com-
puters. He must wei gh agai nst each other many di verse factors i n the
probl em: the accuracy requi red; the cost, wei ght, vol ume, and shape of the
computer; i ts i nerti a and del ay i n acti on; the for ces requi red to oper ate i t;
i ts resi stance to shock, wear, and changes i n weather condi ti ons. He must
consi der how l ong i t wi l l take to desi gn the computer, how easi l y i t can
be bui l t, how easi l y i t can be oper ated by a cr ew, whether sui tabl e sources
of power wi l l be avai l abl e, and so on. The compl exi ty of the theoreti cal
and practi cal probl ems i s so great that two desi gners worki ng on a gi ven
probl em wi l l never arri ve at preci sel y the same sol uti on.
For practi cal reasons, a desi gner shoul d be asked to fi nd a computer
that meets certai n speci fi ed tol erances, rather than the best possi bl e
computer for a gi ven use. He shoul d know what wi l l be the maxi mum
tol erated er r or of the computer, the maxi mum cost, wei ght, and vol ume
occupi ed, the maxi mum number of oper ator s i n the cr ew, the maxi mum
number of servomechani sms al l owed, and so on. Tol erances provi de a
conveni ent means for control l i ng the devel opment of the computer, and
i f establ i shed i n a practi cal waythey permi t some fr eedom of choi ce by
the desi gner.
Choi ce of Approach to the Desi gn Probl em.The type of computer to be
bui l t i s someti mes i ndi cated i n the speci fi cati ons. I f not, the fi rst task
of the desi gner i s to deci de whether the computer i s to be mechani cal ,
el ectri cal , opti cal , or a combi nati on of these. At the same ti me that thi s
i mportant deci si on i s made, the desi gner must wei gh i n hi s mi nd the path
that hi s thi nki ng wi l l fol l ow. Ther e are two pri nci pal methods for desi gn-
i ng a computer: the constructi ve method and the anal yti c.
The constructi ve method makes use of a smal l -scal e model of the real
system wi th whi ch the computer i s to deal .
For exampl e, a constructi ve
anti ai rcraft fi re-control computer mi ght determi ne the el ements of the
l ead tri angl e by mai ntai ni ng wi thi n i tsel f and measuri ng the el ements of a
smal l model of thi s tri angl e.
SEC.1.2] SURVEY OF THE PROBLEM OF COMPUTER DESI GN 3
I n usi ng the anal yti c method, the desi gner concentrates on the anal yti c
rel ati ons between the vari abl es i nvol ved. A rel ati on between vari abl es,
such as
(1)
can be gi ven mechani cal expressi on i n terms of di spl acements or shaft
rotati ons, wi thout r egar d to the nature of the quanti ty r epr esented by the
vari abl es x, y, and z. For exampl e, one may possess two devi ces that
gener ate output di spl acements xy and x/y, respecti vel y, gi ven i nput di s-
pl acements z and y. Combi ni ng these wi th a thi rd devi ce for addi ng thei r
output di spl acements, one can then pr oduce a computer that, gi ven i nput
di spl acements x and y, generates a fi nal output di spl acement z havi ng
conti nuousl y the val ue speci fi ed by Eq. (1). The computer i s then a
mechani zati on of Eq. (1), rather than a model of any speci al system
i nvol vi ng vari abl es x, y, and z thus rel ated.
Computers desi gned by anal yti c methods consi st of uni ts (cel l s)
that mechani ze fai rl y si mpl e rel ati ons, so connected as to provi de a
mechani zati on of a mor e compl ex equati on or system of equati ons. For
any gi ven probl em a great vari ety of desi gns i s possi bl e. Thi s vari ety
ari ses i n part from the possi bl e choi ce among mechani cal cel l s mechani z-
i ng a gi ven el ementary rel ati on, and i n part from the vari ety of ways i n
whi ch the rel ati on between a gi ven set of vari abl es can be gi ven anal yti c
expressi on. Thus, each of the equati ons
Z=; (yj+ l ),
( )
2=X y+~,
Y
Z7J= Z(yz + 1),
(2a)
(2b)
(2C)
[al l equi val ent to Eq. (l )] suggests a di fferent method of connecti ng
mechani cal cel l s i nto a compl ete computer. Thi s fl exi bi l i ty i n anal yti c
desi gn methods makes i t possi bl e to arri ve at desi gns that are i n general
mor e sati sfactory mechani cal l y than those obtai ned by constructi ve
methods.
I n the present vol ume we shal l be concer ned enti rel y wi th mechani cal
computers desi gned by the anal yti c method.
Bl ock Di agram of the CompuLer.To each formul ati on of the probl em
i n anal yti c terms ther e corresponds a bl ock di agram of the computer.
i n thk di agram each anal yti c rel ati on between vari abl es i s r epr esented
by a square or si mi l ar symbol , from whi ch emer ge l i nes representi ng the
vari abl es i nvol ved; a l i ne representi ng a vari abl e common to two rel ati ons
wi l l connect the correspondi ng squares i n the di agram. I n mechani cal
terms, each square then represents an el ementary computer that estab-
4 I NTRODUCTI ON [SEC. 1.2
l i shes a speci fi ed rel ati on between the vari abl es, and the connecti ng l i nes
r epr esent the necessary connecti ons between these el ementary com-
puters. By exami nati on of bl ock di agrams the desi gner wi l l be abl e to
see the pri nci pal vi rtues of each computi ng scheme: the compl exi ty of the
system, the wor khg range of vari abl es, the accuracy requi red of i ndi vi dual
components, and so on. On thi s basi s he can make at l east a tentati ve
sel ecti on of the bl ock di agram to be used.
Sel ecti on oj Components jor the Computer.-Knowi ng the accuracy and
mechani cal properti es requi red of each computi ng el ement, the desi gner
can sel ect the el ementary computers from whi ch the compl ete devi ce i s to
be bui l t.
As an exampl e of the di verse factors to be borne i n mi nd, l et us suppose
that i t i s requi red to provi de a mechani cal moti on proporti onal to the
product of two vari abl es, X1 and X2. A sl i de mul ti pl i er of average si ze
wi l l al l ow an er r or of from 0.1 per cent to 0.5 per cent of the whol e range
of the vari abl e; thi s er r or wi l l depend on the qual i ty of the constructi on
-on the backl ash and the el asti ci ty of the system. A l i nkage mul ti pl i er
wi l l have an er r or of some 0.3 per cent due to i ts structure, practi cal l y no
er r or from backl ash, and a sl i ght er r or due to el asti ci ty of the system i f the
uni t i s wel l desi gned; the space requi red by a l i nkage mul ti pl i er i s smal l ,
but i ts er r or cannot be r educed by i ncreasi ng i ts si ze. I f these devi ces do
not promi se suffi ci ent accuracy, the desi gner must use mul ti pl i ers based
on other pri nci pl es. I t i s possi bl e to per for m mul ti pl i cati on by use of
two of the preci si on squari ng devi ces i l l ustrated i n Fi g. 1.23, by connect-
i ng these i n the way suggested by the equati on
X1X2 = +(X1 + x2)* *(X1 X2)2. (3)
The er r or of such a mul ti pl i er may be as l ow as 0.01 per cent, but the
system has an appreci abl e i nerti a. About the same accuracy i s attai n-
abl e by a mul ti pl i er based on the di fferenti al formul a for mul ti pl i cati on,
d(xl xz) = xl dxz + x2 dxl ; (4)
thi s empl oys two i ntegrators, and i s commonl y used when two quanti ti es
are to be mul ti pl i ed i n a di fferenti al anal yzer. Thi s scheme i s useful onl y
when i t i s possi bl e to al l ow a sl ow change i n a constant added to the
product XI X~a change whi ch wi l l resul t from sl i ppage i n the i ntegra-
tors, negl i gi bl e for a si ngl e mul ti pl i cati on but accumul ati ng wi th repeti ti on
of the operati on.
From thi s di scussi on i t shoul d be evi dent that ther e i s no best
mul ti pl i er. Si mi l arl y, other components of a computer must be sel ected
wi th due r egar d for thei r speci al characteri sti cs and the demands to be
made upon them.
Mathemati cal Desi gn oj the fi @em.From the bl ock di agram one
shoul d pr oceed to the mechani cal desi gn of a system through an i nter-
SEC.1.3] ORGANI ZATI ON OF THE PRESE>T VOLUME 5
medi ate step-that of establ i shi ng the mathemati cal desi gn of the
system. The mathemati cal desi gn i gnores the di mensi ons not essenti al
to the nature of the computati on to be carri ed out-di ameters of shafts,
di mensi ons of bal l beari ngs, di mensi ons of the fram~but speci fi es the
di mensi ons of l evers measured between pi vots and joi nts, the si ze of fri c-
ti on wheel s, tentati ve gear di ameters and gear rati os. The properti es of
thi s desi gn shoul d be studi ed careful l y, because thi s usual l y l eads to a
change i n some detai l of the desi gn, and someti mes even to choi ce of a new
bl ock di agram.
Fi nal Steps i n the Desi gn.-From the mathemati cal desi gn of the
system one can pr oceed to the desi gn of a worki ng model .
The el ements
of thi s model shoul d be accessi bl e rather than massed together , i nexpen-
si ve, and qui ck to manufacture. I f the performance of the worki ng model
i s found to be sati sfactog, the fi rst model can be desi gned. Her e the
i ngenui ty of the desi gner must be used to the maxi mum.
The parts of
the mechani sm must be arranged compactl y to decr ease space requi re-
ments, wei ght, and the effects of el asti ci ty and thermal expansi on, but
they shoul d not be massed i n such a way that assembl y i s di ffi cul t, or
repai r or servi ci ng i mpossi bl e. Someti mes di vi si on of the whol e computer
i nto several i ndependent parts i s advi sabl e. Fi nal l y, the computer can be
bui l t and tested agai nst speci fi cati ons.
1.3. Organi zati on of the Present Vol ume.I t i s not possi bl e to di s-
cuss i n one vol ume al l el ements of the probl em of computer desi gn. Thi s
book wi l l deal pri nci pal l y wi th bar-l i nkage computersspeci fi cal l y, wi th
the mathemati cal desi gn of el ements for such computers. Bar l i nkages
are mechani cal l y ver y sati sfactory, and computers bui l t from them have
many i mportant vi rtues, but the mathemati cal desi gn of these systems
i s rel ati vel y di ffi cul t and i s not wi del y understood. Ther e are few stand-
ard bar-l i nkage el ements for computers; i t i s usual l y necessary to desi gn
the components of the computer, and not merel y to organi ze standard
el ements i nto a compl ex assembl y. I t i s hoped that the desi gn methods
to be descri bed her e wi l l l ead to thei r mor e general use.
Bar l i nkages can be used i n combi nati on wi th the standard computi ng
mechani sms. For thi s reason, and for the contrast wi th the bar l i nk-
ages whi ch are to be di scussed l ater, thi s vol ume begi ns wi th a bri ef survey
of some mor e or l ess standard el ements of mechani cal computers. Chap-
ter 2 i s devoted to a general di scussi on of bar l i nkages. Chapter 3
establ i shes termi nol ogy and descri bes graphi cal pr ocedur es of whi ch
extensi ve use wi l l be made. Chapters 4, 5, and 6 di scuss, i n or der of thei r
i ncreasi ng compl exi ty, bar l i nkages wi th one degr ee of fr eedomgener -
ators of functi ons of one i ndependent vari abl e, Chapter 7 i ndi cates some
mathemati cal methods of i mportance i n bar-l i nkage desi gn. Fi nal l y,
Chaps. 8, 9, and 10 devel op methods for the desi gn of bar-l i nkage gener -
6
I NTRODUCTI ON [SEC.14
ators of functi ons of two i ndependent vari abl esa fi el d i n whi ch bar
l i nkages have ver y stri ki ng advantages.
ELEMENTARY COMPUTI NG MECHANI SMS
The remai nder of thi s chapter wi l l gi ~e a bri ef survey of el ementary
computi ng mechani sms, or cel l s, of mor e or l ess standard type. Di s-
cussi on of bar-l i nkage cel l s wi l l be defer r ed to Chap. 2.
1.4. Addi ti ve Cel l s.- Addi ti ve or l i near cel l s establ i sh l i near
rel ati ons between mechani cal moti ons of the cel l , usual l y shaft rotati ons
or sl i de di spl acements. I f these are descri bed by parameters X1, X2, X3,
the cel l wi l l compute
X3= Q. X,+ Q. X,+C.
(5)
Her e Q, Q, and C are constants dependi ng on the desi gn of the cel l and the
choi ce of the zer o posi ti ons from whi ch Xl , XZ, and X3 are measured. By
FI G. 1 1.Bevel -gear cl i fferenti al .
pr oper choi ce of the zer o posi ti ons, C can al ways be made to vani sh; i n
what fol l ows i t wi l l be assumed that thi s has been done.
The bevel -gear deferenti al (Fi g. 1.1) i s a wel l -known l i near CCI 1f[Jr
whi ch al l thr ee parameters are rotati ons.
The parameter Xl i s the rot:~-
ti on of the shaft S1 from a predetermi ned zer o posi ti on, X1 = O; the posi -
ti ve di recti on of rotati on i s i ndi cated by symbol s representi ng the hewl
and tai l of an ar r ow wi th thi s di recti on. The parameter X2 i s the rotati on
of the shaft SJ from a si mi l ar zer o posi ti on; Xj i s the rotati on from i ts zer o
posi ti on of the cage C carryi ng the pl anetary bevel gears G. The zer o
posi ti ons are not i ndi cated i n the fi gure.
The equati on of the bevel -gear di fferenti al i s
X, = 0.5X, + 0.5XZ.
(6)
To der i ve thi s i t i s conveni ent to consi der the val ue of X2 correspondi ng to
gi ven val ues of X, and Xs. Let us consi der the di fferenti al to be ori gi nal l y
i n the posi ti on X1 = X2 = X3 = O. The parameters Xl and X3 can then
be gi ven thei r assi gned val ues i n two steps, the fi rst a rotati on of both the
SEC,14] ADDI TI VE CELLS 7
shaft S1 and the cage C through the angl e X3, and the second a rotati on of
the shaft S1through an addi ti onal angl e Xl Xi . I n the fi rst step the
di fferenti al moves as a uni t; the shaft Sz i s r otated through the angl e X3.
I n the second step, the cage i s stati onary and the movement of the shaft
S1 i s transmi tted to the shaft SZ wi th i ts sense of rotati on r ever sed; the
rotati on through angl e Xl X3 of the shaft SI causes rotati on through
Xt Xl of the shaft SZ. The total rotati on of the shaft SZ i s then
X2 = Xt + (Xt X,), from whi ch Eq. (2) fol l ows i mmedi atel y. I t i s,
of course, essenti al that al l rotati ons be taken as posi ti ve i n the same
sense.
I t i s remarkabl e that Eq. (6) i s i ndependent of the rati o of the bevel
geari ng of the cl i fferenti al ; the essenti al characteri sti c of thi s type of
X3
FI G.1.2.Cyl i ndri ca1-gear cl i fferenti al .
di fferenti al i s that the geari ng of the cage transmi ts the rel ati ve moti on of
the shaft S1 to the shaft SZi n the rati o 1 to 1, but wi th r ever sed sense. I t
i s not necessary to use bevel gears i n the cage to obtai n thi s resul t;
cyl i ndri cal gears can accompl i sh the same purpose. A cyl i ndri cal -gear
di fferenti al i s shown i n Fi g. 1.2. Thi s di fferenti al i s equi val ent to the
common bevel -gear di fferenti al , except i n i ts mechani cal features. I t i s
fl atter, and easi er to construct i n l arge numbers, but ther e i s one mor e
gear mesh than i n the common type; ther e may be mor e backl ash and
mor e fri cti on. I t shoul d be noted, however , that bevel gears are subject
to axi al as wel l as radi al for ces i n thei r beari ngs, and that these may al so
i ncrease fri cti on.
The spur-gear di fferenti al shown i n Fi g. 1.3 has onl y two gear meshes,
and i s qui te fl at. The pl anetary gears G i n thei r cage C do not i nvert the
moti on of the shaft SI when transmi tti ng i t to the shaft S3, but can be
made to transmi t i t at a rati o di fferent from 1. The eauati on of thi s
8 I NTRODUCTI ON [SEC.1.4
di fferenti al i s
X, = QX, + (1 Q)X,. (7)
To pr ove thi s rel ati on we can use the same method as before. Let us
begi n by consi deri ng the di fferenti al i n the zer o posi ti on,
X1=X,=X3=0.
We wi sh to fi nd the val ue of X3 correspondi ng to gi ven X, and X2. We
i ntroduce the angl es Xl and X2 i n two steps, fi rst turni ng both the shaft
FI G. 1.3.Spur-gear cl i fferenti al .
FI G.1.4.Di fferenti al wi th axi al l y di spl aced spi ral gear .
S, and the cage C through the angl e X2, and then the shaft S1 through an
addi ti onal Xl -
X,. I n the fi rst step the di fferenti al i s turned as a ri gi d
body; the shaft S, i s al so turned through the angl e X2. l n the second
step the shaft Ss i s turned through Q(xl X2); its total motion is
X, = X, + Q(X, X2), i n agreement wi th Eq. (7).
I f we make Q = Q = 0.5 by pr oper choi ce of the gear rati os, we can
obtai n a di fferenti al equi val ent to the bevel -gear di fferenti al .
The fact
that the fr ee choi ce of Q gi ves to thi s di fferenti al a l arger fi el d of appl i ca-
bi l i ty does not necessari l y mean that thi s di fferenti al shoul d be pr efer r ed
SEC.1.4] ADDI TI VE CELLS 9
tothosewi th Q = 0.5; i t i nconveni ent tousedi fferenti al s wi th Q =0.5
as prefabri cated standard el ements.
A deferenti al mth axi al l y di spl aced spi ral gear i s shown i n Fi g. 1.4.
The parameter,, whi ch measures the axi al di spl acement of the spi ral
gear andthepi n PZ, i nvari abl e onl y ~\-i thi nfi ni te l i mi ts. The mechani cal
structure of thi s di fferenti al i s, however , much si mpl er than that of the
di fferenti al s al ready menti oned, for whi ch al l parameters can change wi th-
out l i mi tati on. The equati on of thi s di fferenti al i s
X3=X1*9X2, (8)
wher e ni s the number of threads peri nch al ong the axi s of the spi ral gear
on the shaft Sz and m i s the number of teeth on the gear wi th whi chi t
FI G.1.5.Di fferenti al wor m gear i ng.
meshes. Thehel i cal mgl eofthe gears shoul d beat l east 450 for smooth
acti on and smal l backl ash.
Thedi ferenti al wor m geari ng shown i n Fi g. 1.5 i s used for the same
purpose asthe precedi ng di fferenti al , especi al l y i f the range of val ues of
X, corresponds to a l arge fracti on of a revol uti on of the shaft S, or even
to several revol uti ons of thi s shaft. The equati on of thi s d%-enti al i s
x,= ++ X1+ +X2 (radi ans) (9)
wher e t i s the number of teeth of the wor m gear, m i s the mul ti pl i ci ty of
the threads of the worm, and R i s the radi us of the wor m gear.
The si gn i n Eqs. (8) and (9) depends on the sense of the threads of the
spi ral or wor m gear.
The scr ew di fferenti al shown i n Fi g. 1.6 combi nes an axi al transl ati on
Xl of a scr ew wi th a transl ati on X2 of the nut N wi th r espect to the screw;
x3 = xl + x,. (l o)
I
10 I NTRODUCTI ON [SEC.1.4
To obtai n the fi rst transl ati on, the pi n P on whi ch the scr ew turns i s
di spl aced by X1. Therotati on of thescr ew comes fromthegear G,whi ch
meshes wi th a cyl i ndri cal r ack C and sl i des al ong i t. The real i nput
IWILI
x,
i
*
-+E=Eaii c-k
FI G.1.6.Screwdi fferenti al .
parameter of the di fferenti al i s not X,, but the angl e X, through whi ch the
r ack i s turned. The equati on of the di fferenti al i s then
X3 = x, * kx4. (11)
The si gn depends on the sense of the screw; k i s a constant determi ned by
the gear rati o, the number of threads per i nch on the screw, and thei r
.++
FI G.1.7.Bel t di fferenti al .
mul ti pl i ~i ty. Al l thr ee parameters of thi s
di fferenti al have constructi ve l i mi ts.
The bel t deferenti al (Fi g. 1.7) makes
use of the i nextensi bi l i ty of a bel ti ng on
several pul l eys. I n practi ce, chai ns,
stri ngs, and speci al cabl es are used as bel ts.
The equati on of the bel t di fferenti al i s
X3 = c 0.5XI 0.5X2,
(12)
wher e C i s a constant dependi ng on the
choi ce of zer o poi nts of the parameters.
The tensi on i n the bel t must not fal l
bel ow zer o at any ti me; i f i t does, the bel t
wi l l sag and the equati on of the di fferenti al
wi l l not hol d. To obtai n posi ti ve acti on i n the di recti on of i ncreasi ng X3,
i t i s necessary to prel oad the bel t by putti ng a l oad on the output pul l ey
i or i nstance, by a spri ng that can exer t a for ce l arge enough to pr oduce the
desi red acti on. The maxi mum dri vi ng for ce requi red for thk dMer-
enti al wi l l then be about twi ce the for ce necessary to oper ate i t wi thout
prel oadl ng.
The l oop-bel t di ~erenti al (Fi g. 1.8) has the bel ti ng i n the form of a l oop
wi th l ength i ndependent of the posi ti on of the pul l eys, The bel t can then
SEC-1.4]
MULTI PLI ERS
11
be prel oaded (turnbuckl e B) wi thout addi ng to the dri vi ng for ce of the
di fferenti al , except by the i ncreased fri cti on i n the beari ngs.
Bel t di fferenti al s are some-
ti mes used to add a l arge number
of parameters; they are easi l y
combi ned i n batteri es, as i ndi cated
schemati cal l y i n Fi g. 1.9. I n such -
an arrangement the parameter XT
may have so l arge a range that i t
i s i mpracti cal to use a sl i de as the
output termi nal . I t i s better
FI G.1S.-Loop-bel t cl i fferenti al .
practi ce to use a drum (dashed
l i ne i n Fi g. 1.9) on whi ch the bel t i s wound on. and at the same ti me
wound off. To pr event sl i ppage, the bel t shoul d make many turns on the
FI G. 1.9.Loop-bel t di fferenti al for
the eval uati on of
X1= CX12XZ+2X, 2X,
+ 2X, 2X6.
drum and be fastened to i t; a chai n
on chai n spr ockets may al so be used
as the bel t.
The above enumerati on does not
exhaust the possi bi l i ti es for l i near
mechani cal cel l s; ther e are many
vari ants the use of whi ch may be
di ctated by speci al ci rcumstances.
As a rul e, when a di fferenti al i s
used i n a computi ng mechani sm, two
of i ts members (the i nput termi nal s)
are moved by external forces; thi s
resul ts i n movement of a thi rd mem-
ber (the output termi nal ) whi ch i s i n
turn requi red to furni sh an appreci -
abl e for ce. I f di fferenti al s wer e fri c-
ti onl ess, any two of thei r thr ee
termi nal s coul d be used as i nput
termi nal s. I n real i ty, onl y a few of
the di fferenti al s descri bed her e have
compl ete i nterchangeabi l i ty of the
termi nal s. For i nstance, wi th the
scr ew cl i fferenti al [Fi g. 1.6) i t i s i m-
possi bl e to have Xd as the output
parameter i f the hel i cal angl e of the scr ew i s so l ow that sel f-l ocki ng of the
nut on the scr ew occurs; i t i s possi bl e to use Xl as an output parameter,
and, of course, al so X,. Wi th the chfferenti al wor m geari ng of Fi g. 15,
X, i s an i mpracti cabl e output parameter.
12 I NTRODUCTI ON [SEC.1.5
106. Mul ti pl i ers.-Mul ti pl i ers are computers that establ i sh between
thr ee parameters a rel ati on
RX3 = X, .X,, (13)
(
wher e R i s a constant that depends on the type of mul ti pl i er and on i ts
di mensi ons.
The acti on of the sl i de mul ti pl i er shown i n Fi g. 110 i s based on the
proporti onal i ty of the si des of two si mi l ar tri angl es. These are tri angl es
wi th hori zontal bases, and verti ces at the central pi n shown i n the fi gure:
I v;,:
.-
FI G.1.10.Sl i demul ti pl i er.
the fi rst has a base of l ength R and al ti tude Xl , the second a base of l ength
X2 and al ti tude X8. Thus
R x,
T, = X
(14a)
or
RX3 = X,X2.
(14b)
The fi gure gi ves a schemati c rather than a practi cal desi gn; the l engths of
the sl i di ng surfaces as shown are not great enough to pr event sel f-l ocki ng
i n al l possi bl e posi ti ons of the mechani sm. These l engths determi ne the
space requi rements for mul ti pl i ers of thi s type; they must be rel ati vel y
l arge i n two di recti ons. I t i s di ffi cul t to make thi s type of mul ti pl i er
preci se. The pi ns i n sl ots, as shown i n the fi gure, are mechani cal l y
i nadequate, and rol l er sl i des on rai l s must be used. One can not achi eve
the same end by i ncreasi ng the di mensi ons of the mul ti pl i er because the
SEC.1.5] MULTI PLI ERS 13
el asti ci ty of parts comes i nto pl ay, not onl y when the parts are operati ng
i n a computer, but al so when they are bei ng machi ned.
The sl i de mul ti pl i er shown i n Fi g. 1.11 saves space i n one di recti on.
Ther e are fewer sl i di ng contacts, and the sl i des are easi er to construct.
Fm. 1.11.Sl i demul ti pl i erwi thi nputsXI , XI - Xz.
.
I ?m. 1.12.I ntmwcti on nomogr am for mul ti pl i cati on z; = z j . z*.
Thi s devi ce cannot mul ti pl y Xl and X2 di rectl y to compute RX3 = X1X2;
the i nput termi nal s must be gi ven transl ati ons of Xl and X, X2. The
di fference i s easy to obtai n i f the parameters are gener ated as shaft
revol uti ons befor e enteri ng the mul ti pl i er; screws can then be used i nstead
14 I NTRODUCTI ON [SEC.15
of the sl i des shown i n the fi gure, and the requi red di fference can be formed
by a gear di fferenti al .
Nomographi c Mul ti pl i ers.A mul ti pl i er that i s structural l y rel ated to
a nomogram for mul ti pl i cati on wi l l be cal l ed a nomographi c mul ti pl i er.
Such mul ti pl i ers can be deri ved from i ntersecti on or al i gnment nomograms;
the exampl es to be gi ven her e are rel ated to i ntersecti on nomograms.
FI G.1.13.An i ntersecti on nomogr am for mul ti pl i cati on, obtai ned fr om the nomogr am i n
Fi g. 1.12 by a pr ojecti ve transformati on.
Fi gure 1.12 shows an i ntersecti on nomogram for mul ti pl i cati on i n an
unusual form, the ful l si gni fi cance of whi ch wi l l be made cl ear i n the l atter
part of thi s book. Thk represents the formul a
Xi = Xi Xk. (15)
I t consi sts of thr ee fami l i es of l i nes, of constant xi , xi , and z~, respecti vel y;
through each poi nt of the nomogram passes a l i ne of each fami l y, cor -
respondi ng to val ues of xi , ~i , and x~ whi ch sati sfy Eq. (15). (The l i nes
i n thk parti cul ar fi gure are drati for val ues of the zs that are power s of
1.25; thi s i s not of i mmedi ate i mportance f or our di scussi on.) The mul ti -
SEC.1.6] RESOLVERS 15
pl i er of Fi g. l .l Oi sstructural l y rel ated to thi s nomogram. .The rotati ng
sl i de can be brought to posi ti ons correspondi ng to the radi al l i nes i n the
nomogram; the hori zontal and verti cal sl ots cor r espond structural l y to
the hori zontal and verti cal l i nes on the nomogram, and the pi n that con-
nects al l sl i des mechani cal l y assures a tri pl e i ntersecti on of these l i nes.
The val ues of xi , xi , and xk correspondi ng to the posi ti ons of the thr ee
sl i des must then sati sfy Eq. (15); to compl ete the mul ti pl i er i t i s onl y
necessary to provi de scal es from whi ch these val ues can be read, or , as i s
done i n Fi g. 1.10, to provi de mechani cal connecti ons such that termi nal
di spl acements are proporti onal to these quanti ti es.
By a pr ojecti ve transformati on of the nomogram i n Fi g. 1.12 one can
obtai n the nomogram i n Fi g. 113, wher e l i nes of constant val ues of the
FI G.1.14.Nomographi c mul ti pl i er.
vari abl es Zi , ~j, and xk form thr ee fami l i es of radi al l i nes i ntersecti ng i n
thr ee centers. The obvi ous mechani cal anal ogue of thi s nomogram for
mul ti pl i cati on i s shown i n. Fi g. 1.14. I t consi sts of thr ee sl i des that r otate
about center s correspondi ng to the center s of the radi al l i nes i n Fi g. 1.13;
these sl i des are bound together by a pi n, whi ch establ i shes the tri pl e
i ntersecti ons found i n the nomogram, and the correspondi ng val ues of
Zi , Zi , and zk are read on ci rcul ar scal es.
I t wi l l be noted that the scal e
di vi si ons are not uni form. Such nonuni form scal es are of mor e general
use than one mi ght expect. Often one wi l l have to deal wi th vari abl es
gener ated wi th nonuni form scal es by some other computer; by pr oper
choi ce of the pr ojecti ve transformati on one can then hope to pr oduce a
mul ti pl i er of thi s type wi th si mi l arl y defor med scal es.
1.6. Resol vers.The r esol ver i s a speci al type of mul ti pl i er. I t
generates a parameter X3, and usual l y al so another parameter X4, as a
product of a parameter Xl and a tri gonometri c functi onthe si ne or
16 I NTRODUCTI ON [SEC. 1.6
cosi n~f a parameter X2. The equati ons are
X3 = Xl si n Xz, (I OU)
X4 = x, Cos X.2. (16b)
The name of thi s devi ce i s deri ved from i ts acti on as a r esol ver of a vector
di spl acement i nto i ts rectangul ar components.
A si mpl i fi ed desi gn of a r esol ver i s shown i n Fi g. 1015. I n the pl an
vi ew, Fi g. 115a, we see the materi al i zati on of a vector by a screw: the
axi s of the scr ew poi nts i n the di recti on of the vector , at an angl e X2 to a
zer o l i ne; the l ength Xl of the vector i s establ i shed as the di stance from
the pi vot O on whi ch the whol e scr ew i s r otated to a pi n T on the nut of
the screw.
To obtai n the components of the vector , sl i des are someti mes used,
as i n the case of the mul ti pl i er i n Fi g. 1.10. I n Fi g. 115 ther e i s sug-
gested a sol uti on that gi ves much better preci si on and saves space.
Perpendi cul ar shafts pass through the bl ock B that carri es the pi n P.
These shafts are carri ed by rol l ers on rai l s; thei r paral l el i sm to gi ven
l i nes i s wel l assured by gears that mesh wi th racks fastened to the frame.
For conveni ence of constructi on the axes of the shafts do not i ntersect
wi th each other and wi th the axi s of the pi n T. Thi s i ntroduces a con-
stant term e i nto the di spl acement of the shaftsthat i s, i t causes a di s-
pl acement e i n the effecti ve zer o posi ti ons of X3 and X4.
I t i s of i nterest to note how the parameter Xl i s control l ed from the
i nput shaft S, (Fi g. 1.15b.). Whi l e the scr ew i s r otated through the
angl e X2 on the shaft S2, i t i s necessary to control the val ue of Xl by a
gear G that rotates fr eel y on thi s shaft. I f such a gear i s turned through
an angl e proporti onal to Xl i s hel d fi xed when Xl i s constantthe
scr ew wi l l spi n on i ts axi s whenever X2 i s changed; the l ength of the vector
wi l l be affected by change i n XZ, and wi l l not r epr esent the desi red val ue of
Xl . I t i s thus necessary to keep the scr ew wi thout spi n wi th r espect to Sz
when onl y XZ i s changedto keep the gear G movi ng al ong wi th the shaft
S2 whenever Xl i s fi xed. Thi s i s accompl i shed by the so-cal l ed com-
pensati ng di fferenti al , D. As i s shown i n the fi gure, the pl anetary gear
of thi s bevel -gear di fferenti al i s gear ed to the shaft S2 i n the rati o 1 to 1;
the di fferenti al thus r ecei ves an i nput X2. When the i nput shaft S6 i s
r ot ated through Xc, the output shaft S6 i s r otated through an angl e
x, = X6 2x,. (17)
By geari ng the gear G to the shaft Sb i n the rati o 2 to 1, the angl e turned
by G can be made to be
x. = 0.5X6 = 0.5X6 + x2.
Then i f S6 i s stati onary, X. changes equal l y wi th X2, and the scr ew i s not
spun; X, remai ns constant. I f the shaft Se i s turned, the gear G turns
SEC.1.6] RESOLVERS 17
wi th r espect to the shaft 82 through an equal angl e. The change i n
Xl i s then proporti onal to the rotati on of the shaft S6:X6 = QXI , the
constant Q dependi ng on gear rati os and the threadi ng of the screw.
1 &+e I
1,
.
(a)
(b)
T&
,
,
I
X,+e
l @. 1.15.Raeol ver. (a) Pl an vi ew. (b) El evati on. The teeth of the racks are omi tted
fr om the fi gures.
The desi gn i n Fi g. 115 i s so oversi mpl i fi ed that the r esol ver i s sure to
be l acki ng i n preci si on. I n parti cul ar, the fl exi bi l i ty of the structure sup-
porti ng the scr ew i s excessi ve: shaft hs i s easi l y bent and easi l y twi sted.
18 I NTRODUCTI ON [SEC.1.6
Thi s can be remedi ed by pl aci ng the scr ew subassembl y on a ci rcul ar pl ate
wi th a l arge bal l beari ng on i ts ci rcumference, and usi ng a dri vi ng shaft of
reasonabl e di ameter.
A better constmcti on (but one that i s not al ways usabl e) i s presented
i n Fi g. 1.16. I n the pl an vi ew, Fi g. 10l &z, we obser ve the mai n cl i ffer ence
between the subassembl y of the scr ew i n Fi g. 1.15 and the present desi .zn.
-<
(a)
(b)
FI G.1.16.Al ternati ve r esol ver desi gn. (a) Pl an vi ew. (b) El evati on.
I n Fi g. 1~16a the pi n T i s carri ed on an arm of radi us R that rotates on a
pi vot P. Thi s pi vot i s pl aced at a di stance R from the center S of the
ci rcul ar pl ate H to whi ch i t i s fastened. By rotati ng the arm PT, the
vector ST can be changed i n l ength.
I ts di recti on woul d be changed at
the same ti me i f i t wer e not for a compensati ng rotati on of the pl ate H.
$i nce the tri angl e SPT i s i soscel es the angl e of rotati on of ST to be com-
SEC.1.7J
CAMS 19
pensated for i s exactl y hal f of the angl e of rotati on of the arm PT. To
i ntroduce thi s compensati on a di fferenti al i s used; to change the di recti on
of the vector ST i n a desi red manner, the tabl e H i s r otated through a
second di fferenti al . The two di fferenti al s, D, and Dz, are shown i n Fi g.
1.16b. Thei r functi on may be understood i n thi s way. To change the
di recti on of the vector ST we must r otate the whol e subassembl y of the
pl ate Has a uni t; we must turn the gears GI and Gz by the same amounts.
These gears are gear ed to the cages of the pl anetary gears of the di fferen-
ti al s D,, Dz., at the same rati o (1 to 1 i n the fi gure); these al so must be
turned equal l y. That i s accompl i shed by turni ng the shaft Sz and by
keepi ng the shaft S1 stati onary. To change the l ength of the vector ST
wi thout turni ng i t we have to turn the arm cl ockwi se, for exampl e, i n the
pl an vi ew, and the pl ate H countercl ockwi se by hal f the amount. Thi s i s
accompl i shed by turni ng the shaft S1. Thi s shaft i s gear ed to the i nput
of the di fferenti al D1 at the rati o 1 to 1 and to the i nput of the di fferenti al
Dz at the rati o 3 to 1; when the shaft S, rotates, the gear G, turns thr ee
ti mes faster than the gear GZ. To see that thi s gi ves a compensati ng
rotati on of the pl ate through an angl e X when the arm PT rotates
through 2X rel ati ve to the pl ate H, we obser ve that i f the gear Gl wer e
fi xed, a rotati on of Gz and the pl ate H through X woul d r otate the arm
wi th r espect to the pl ate al so by X. To bri ng i t to the cor r ect posi -
ti on, +2X, i t must then be r otated through an angl e of +3X wi th r espect
to the pl ate. To accompl i sh thi s the gear GI must be r otated through an
angl e 3X, si nce the di recti on of rotati on i s r ever sed i n the gear Gt.
Thus G, must turn i n the same di recti on as Gn, but thr ee ti mes as fast.
1.7. Cams.-A cam i s a mechani sm that establ i shes a functi onal rel a-
ti on between parameters Xl and X2:
x, = F(x,). (18)
I f X, i s the i nput parameter, XZ the output parameter, i t i s necessary i n
practi ce that F(XJ be a si ngl e-val ued, conti nuous functi on wi th deri va-
ti ves whi ch do not exceed certai n l i mi ts.
PZune cams exi st i n two pri nci pal vari ants, shown i n Fi gs. 1.17 and
1.18. I n the fi rst the cam has the form of a di sk shaped al ong a general
curve. Contact wi th thi s cam i s made by a rol l er on an arm; the contact
i s assured by tensi on of a spri ng. A cam of thi s type i s easy to bui l d
and has negl i gi bl e backl ash, but the for ce on. the arm i s rather smal l i n
one of the two senses of moti onnot l arger than the for ce of the spri ng.
I n the second vari ant ther e i s a sl ot mi l l ed i nto a fl at surface rotati ng
on a pi vot; contact i s made by a rol l er carri ed on a sl i de, as shown i n
Fi g. 1-18. The second form does not permi t use of as steep a spi ral ai i
does the fi rst, si nce sel f-l ocki ng i s mor e l i kel y to occur ,
!20
I NTRODUCTI ON
[SEC.1.7
The yl i ndti cal cam shown i n Fi g. 119 has a sl ot mi l l ed i nto the surface
of a cyl i nder; a smal l rol l er carri ed by a sl i de passes al ong the sl ot when
the cam i s turned on i ts axi s through the i nput angl e X1. The form of the
sl ot i s so chosen that the moti on of the sl i de, descri bed by the output
parameter X2, has the desi red
\.
character.
I WJ. l -17.Pl ane cam wi th sPri nK
Fm. 1.l S.Pl ane cam wi th gr oove
contact.
contact.
One vari ant of pi n geari ng, as shown i n Fi g. 1~, has a gear wi th a
speci al type of tooth meshi ng wi th a mi l l ed cur ved rack. (The mi l l i ng
tool has a cutti ng shape i denti cal wi th the shape of the teeth of the gear.)
Another form of pi n gear (Fi g. 121) has pi ns of speci al shape i nserted i n
a pl ate; these mesh wi th a speci al l y formed gear.
I n both vari ants the
gear i s keyed on a shaft, wi th fr eedom for l ateral moti on; thi s moti on
I 1
I - x l
FI G. 1,19.Cyl i ndri cal cam wi th gr oove
FI G. 1.20.Pi n geari ngwi th pi ns on gear.
contact.
of the gear i s assured by the acti on of the cur ved r ack on the pi ns on the
gear, or by a speci al cam constructed for thi s purpose.
The bel t cam shown i n Fi g. 122 i s a nonci rcul ar pul l ey or drum on
whi ch i s wound a bel t, or stri ng, or some other ki nd of bel ti ng. I f the
number of revol uti ons of such a cam i s to be gr eater than one, the stri ng i s
wound i n a spi ral ; the shape of thi s spi ral shoul d assure a smooth tan-
genti al wi ndi ng of the stri ng on the cams. Cams of thi s type can al l ow
SEC. 1.7] CAMS 21
ver y l arge travel s of the bel t and shaft, but they are mechani cal l y l ess
desi rabl e than pi n gears. They are not so safe i n operati on, and rather
\
Fl a. 1.21.Pi n-geari ng wi th pi ns on the di sk.
del i cate, especi al l y i n the compensated form i n whi ch
stri ng are si mul taneousl y wound off and wound on.
An exampl e of a compen.satd bel t cam i s the aquari ng cam shown i n
Fi g. 1.23. I n thi s, two stri ngs are wound partl y on a cyl i nder, partl y on a
cone. The wi ndi ng on the cone i s i n
the form of a spi ral wi th equal l y
spaced threads; the form of the wi nd-
i ng i s assured by a gr oove. One stri ng
begi ns on the l eft si de of the drum
and, after a number of turns, passes
on to the cone and conti nues i n the
gr oove to the tapered ri ght end of the
cone. The second stri ng begi ns on
the ri ght si de of the cyl i nder and after
several turns to the l eft passes al so
onto the cone, wher e i t conti nues
through the gr oove to the l eft, to end
at the l arger end of the cone. The
el ement of rotati on dXl of the cone
produces a moti on of the stri ng equal
to Rl dXl , wher e RI i s the average
radi us of the cone at the poi nts wher e the stri ng meets and l eaves the cone.
The correspondi ng rotati on of the drum i s ther efor e dXz = ( Rl dXJ/Rz,
equal l engths of
1+
Fm. 1.22.Bel t cam.
22 I NTRODUCTI ON [SEC.1.7
wher e RZ i s the radi us of the drum.
The radi us RI i s proporti onal to the
angl e X1 measured from a pr oper l y chosen zer o posi ti on of the shaft S1.
(Thi s zer o posi ti on i s, of course, not practi cal l y attai nabl e, si nce i t woul d
cor r espond to zer o radi us of the cone at the poi nt of contact. ) We have
then
-dX, = y, (19a)
X2 = &2x;, (19b)
i f the zer o poi nt for X2 i s pr oper l y chosen. Her e k i s the i ncrement of the
radi us RI per radi an rotati on of the shaft S~.
r
\
-1fqJJj1/ //
,,
X2\
,; 11,111
1
,, ;:,
i{~
1,
1;
1. I
t.
----
P
J
I
FI G.1.23.Compensated squari ng cam.
Thi s squari ng cam does not by i tsel f oper ate down to X, = O. I t can,
however , be used i n a range i ncl udi ng zer o i i i t i s combi ned wi th a di ffer-
enti al . Wi th
X,=x, +c, (20)
Eq. (19) becomes
X2 = KX: + 2KCXa + KC2. (21)
I ntroduci ng the new parameter
X, = X, 2KCX, + KC, (22)
we have
X4 = KX: ; (23)
thi s hol ds even i f X8 i s zer o or negati ve. The l arger the negati ve val ues
of X6 to be reached the l arger must be the posi ti ve constant C. The
SEC.18] I NTEGRATORS 23
preci si on of a cam of thi s type can be made ver y hi gh; the er r or maybe l ess
than 0.02 per cent of the total travel of the output shaft L%. The rel a-
ti vel y great i nerti a and bul k of the devi ce (especi al l y when i t i s combi ned
wi th a cl i fferenti al for squari ng negati ve numbers), l i mi ts i ts use to cases
wher e preci si on i s essenti al .
Threedi mensi onal cams or camoi ds, such as that shown i n Fi g.
1.24, are bodi es of general form wi th two degr ees of fr eedomfor i nstance,
a transl ati on of Xl and a rotati on Xzi n contact wi th another body wi th
one degr ee of freedom, for i nstance, a transl ati on X3. The parameter X3
wi l l then be a functi on of two i ndependent parameters, Xl and X2:
x3 = F(XI , X2). (24)
The body i n contact i s cal l ed the fol l ower; i t may be a bal l on a sl i de,
as shown i n the fi gure, or an arm
rotati ng on a pi n paral l el to the mai n
axi s of the camoi d and touchi ng the
surface of the cam. Camoi ds are
val uabl e i n that they can gener ate any
wel l -behaved functi on of two i nd~
pendent vari abl es. They are, how-
ever , expensi ve to bui l d wi th enough
preci si on, have consi derabl e fri cti on,
and take too much space. Bar l i nk-
ages are al ways to be pr efer r ed to
camoi ds when i t i s possi bl e to desi gn
such a l i nkage.
108. I ntegrators.-I ntegrators are
2
FI G.1.24.Three-di mensi onsl cam.
computers that have an output
parameter, Xi , and two i nput parameters, X, and X,, functi onal l y rel ated
by
/
x,
x3 X30 = F(x,)dx2. (25)
X,o
The si mpl est form of i ntegrator gi ves
/
x,
x3 x,, = Kxl dx2. (26)
x,,
The parameters Xl , X2, of an i ntegrator can be vari ed at wi l l ; they can, for
i nstance, be gi ven functi ons of ti me t. The val ue of the i ntegral , as a
functi on of t,wi l l depend on the form of these functi ons, and not merel y
on the i nstantaneous val ues of Xl and X2. Thus, unl i ke a functi on
gener ator , an i ntegrator does not establ i sh a fi xed rel ati on between the
i nstantaneous val ues of the parameters i nvol ved.
i
24 I NTRODUCTI ON [SEC. 1.8
The equati ons of i ntegrators are conveni entl y wri tten i n di fferenti al
form; Eq. (26) becomes then
dx, = Kx,dx,. (27)
Thi s i s parti cul arl y conveni ent i n schemati c di agrams of compl ete com-
puti ng systems.
A common type of i ntegrator i s the fri cti on-wheel i ntegrator shown i n
Fi g. 1.25. The output parameter X3 i s gener ated by a fri cti on wheel i n
contact wi th a pl ane di sk, the rotati on of whi ch i s descri bed by the
parameter X2. Si nce the moti on of the fri cti on wheel depends on fri cti on
between the di sk and the wheel , a normal for ce must act to mai ntai n the
FI G. 1.25.Fri cti on-wheel i ntegrator.
fri cti onal for ce at an adequate l evel ; for thi s reason the di sk i s pressed
agai nst the wheel by a spri ng. The fri cti on wheel i s transportabl e al ong
i ts axi s; the di stance from the axi s of the di sk to the poi nt of contact i s the
parameter Xl . I n preci si on i ntegrators the fri cti on wheel i s carri ed by a
fi xed shaft and the rotati ng di sk i s moved wi th r espect to the frame by the
amount Xl . The equati on of the i ntegrator i n the fi gure i s
dXs = : X,dXz, (28)
wher e r i s the radi us of the fri cti on wheel .
The doubl e-bal l i ntegrator of Fi g. 1-26 has the same equati on as the
fri cti on-wheel i ntegrator; the di fference between these two desi gns i s con-
structi ve onl y. The fri cti on wheel i s repl aced by two bal l s carri ed i n a
smal l cyl i ndri cal contai ner, as shown i n the fi gure, or i n a speci al con-
i
SW. 108]
I NTEGRATORS
u~ft$
Fm. 1.26.Doubl e-baU i ntegrator.
Fm. 1.27.Pl an vi ew of componentsol ver.
26 I NTRODUCTI ON [SEC.18
tai ner wi th rol l er gui des for the bal l s, to r educe fri cti on. These bal l s
transfer the moti on of the di sk (X1dXZ) to a drum of radi us r, whi ch
rotates through an angl e dXt gi ven by Eq. (28). The bal l s are easi l y
transportabl e, rol l i ng al ong the drum, wi th whi ch they are i n contact
under constant pressure. Thi s desi gn i s useful when one requi res an
effi ci ent compact computer but does not need the maxi mum accuracy
possi bl e wi th mechani cal i ntegrators. The mai n sour ce of er r or i s the
l ack of absol utel y sharp defi ni ti on of the di stance from the axi s of the
pl ate to the poi nt of contact of the pl ate wi th the bal l s. Any l ateral
fr eedom of the l ower bal l i mpai rs the preci si on of the resul ts.
The component sol ver shown i n Fi g. 1.27 i s a good exampl e of an i nte-
gr ator of the mor e general type. A l arge bal l of gl ass or steel i s hel d
between four rol l ers pl aced i n a square, wi th axes i n the same pl ane, and
two rol l ers wi th axes paral l el to that pl ane; the poi nts of contact are at the
cor ner s of a regul ar octahedron. (Fi gure 1.27 shows onl y fi ve of the si x
rol l ers.) The fi rst four rol l ers have fi xed axes, but the other two have
axes that are al ways paral l el , but may assume any di recti on i n the hori -
zontal pl ane. The rotati on of these l atter axes i n the hori zontal pl ane,
measured from a certai n zer o posi ti on, i s the i nput parameter Xl ; the
rotati on of these rol l ers on thei r shafts i s the second i nput parameter X2;
the rotati on of any one of the four rol l ers on fi xed axes maybe taken as an
output parameter. Si nce rol l ers on paral l el axes r otate through equal
angl es, ther e are two di fferent output parameters, X3 and Xl . I f al l
rol l ers have the same di ameter, the equati ons of the component sol ver are
dXs = COSX,dX,, (29a)
dX, = Si n X,dX*. (29b)
Thus the component sol ver i s descri bed by Eq. (25), but not by Eq. (26).
CHAPTER 2
BAR-LI NKAGE COMPUTERS
2.1. I ntroducti on.-A bar l i nkage i s, i n the cl assi cal sense of the wor d,
a system of ri gi d bars pi voted to each other and to a fi xed base. I n thk
vol ume the term bar l i nkage wi l l denote any mechani sm consi sti ng
of ri gi d bodi es movi ng i n a pl ane and pi voted to each other , to a fi xed base,
or to sl i des. Consi derati on wi l l be l i mi ted to essenti al l y pl ane mecha-
ni sms because these are mechani cal l y the easi est to construct. The
i ncl usi on i n bar l i nkages of ri gi d bodi es of arbi trary form i s not an essenti al
extensi on of the term, si nce any
ri gi d body can be repl aced by a
cor r espondhg system of ri gi d bars.
Si mi l arl y, the admi ssi on of sl i des i s
not a real extensi on, si nce bar l i nk-
agesi n the cl assi cal sensecan
be desi gned to appl y the same
constrai nts.
A l i nk i n a bar l i nkage i s a body
connected to two other bodi es by
pi vots. A l ever i s a body connected
,
L.
~
FI G. 2.1.Bar l i nkage: a noni deal bar.
moni ctransformer.
to thr ee other bodi es by pi vots. A crank i s a body pi voted to the fi xed
base, and to one or mor e other bodi es of the l i nkage. Fi gure 2.1 shows a
bar l i nkage that consi sts of a crank R, a l i nk L, and a sl i de S.
Bar l i nkages are ver y sati sfactory devi ces from a mechani cal poi nt of
vi ew. Pi vots and sl i des are easi l y constructed and have smal l backl ash,
smal l fri cti on, and good resi stance to wear.
As computi ng mechani sms, bar l i nkages can per for m al l the functi ons
of the el ementary functi on generators di scussed i n Chap. 1. They can-
not, however , be used to establ i sh rel ati ons between di fferenti al s; they
cannot per for m the functi ons of i ntegrators. As functi on generators i t i s
characteri sti c of bar l i nkages that they do not, general l y speaki ng, per -
form thei r i ntended operati ons wi th mathemati cal accuracy; on the other
hand, they can gener ate i n a si mpl e and di rect way, and wi th good approx-
i mati on, functi ons that can be gener ated onl y by compl i cated combi na-
ti ons of the cl assi cal computi ng el ements.
Ther e are few standard bar-l i nkage functi on generators; one must
usual l y desi gn a bar l i nkage for any gi ven purpose. Methods for desi gn-
i ng such l i nkages from the mathemati cal poi nt of vi ew are the mai n sub-
27
,
I
28 BARLI NKAGE COMPUTERS
[SEC.2.2
ject of thi s book. The probl em i s to fi nd a bar l i nkage that wi l l gener ate
a gi ven functi on. I t must be noted i mmedi atel y that i n general thi s can
be accompl i shed exactl y onl y by a l i nkage wi th an i nfi ni te number of
el ements; mechani sms wi th a fi ni te number of el ements cannot gener ate
the compl ete fi el d of functi ons. From a practi cal poi nt of vi ew, how-
ever , even the si mpl er bar l i nkages offer enough fl exi bi l i ty to permi t sol u-
ti on of the desi gn probl em wi th an acceptabl y smal l er r or . The approach
to the probl em must be syntheti c and approxi mati ve, not anal yti c and
exact.
The mechani cal desi gn of bar l i nkages cannot be di scussed i n thk
vol ume. I t i s of cour se possi bl e to treat anal yti cal l y the properti es of a
gi ven l i nkage: i ts moti on, the di stri buti on of vel oci ti es of i ts parts, accel er-
ati ons, i nerti a, forces. I n thi s r espect the theor y of l i nkages has been
wel l devel oped, even i n el ementary texts; the ki nemati cs of bar l i nkages
have been tr eated especi al l y thoroughl y. I t i s of cour se necessa~ that
the desi gner of l i nkages have knowl edge of the practi cal properti es of
these devi ces, even when he i s pri mari l y i nterested i n thei r mathemati cal
desi gn. I n the present vol ume ther e wi l l be some comment on the
mechani cal features of bar l i nkages, but onl y enough to gi ve the desi gner
the necessary base for reasoni ng when the desi gn pr ocedur e i s started.
2.2. Hi stori cal Notes.Engi neers and mathemati ci ans have i n the
past consi dered bar l i nkages pri mari l y as cur ve tracersthat i s, as devi ces
z
I
I
FI G. 2.2.Bar l i nkages i n a mi cr oscope pl ate
hol der .
servi ng to constrai n a poi nt of the
l i nkage to move al ong a gi ven
curve. The cl assi cal probl em i n
the fi el d has been that of fi ndi ng
a bar l i nkage that wi l l constrai n a
poi nt to move al ong a strai ght
l i ne. Thi s probl em was consi d-
er ed by Watt i n desi gni ng hi s
steam engi ne. Watt found a
suffi ci entl y accurate sol uti on of
the probl em, and i t was the cost
and space requi red that caused the
use of a sl i de i n hi s ori gi nal desi gn.
Bar l i nkages are now extensi vel y
used i n mechani cal desi gn because
of thei r smal l fri cti onal l osses and
hi gh effi ci ency i n transmi tti ng power-effi ci ency gr eater than that of any
gear or cam. The useful ness of bar l i nkages to the mechani cal engi neer
can be i l l ustrated by a l ocomoti ve: i ts transmi ssi on contai ns the famous
paral l el ogram l i nkage, and the val ve moti ons are control l ed by
l i nkages of some compl exi ty. A desi gner of l i nkage mul ti pl i ers
bar
wi l l
I
I
alla. 22] HI STORI CAL NOTES 29
r ecogni ze among these structures el ements that he i s accustomed to use
i n hi s own wor k.
Bar l i nkages are used i n heavy constructi on as counterwei ght l i nkages
and for the transmi ssi on of spri ng acti on.
They al so ser ve as el ements of
ti e i nstmments. The paral l el ogram l i nkage used to assure pure tran%
l ati onal moti on of a sl i de bei ng exami ned by a mi croscope i s i l l ustrated i n
Fi g. 22. Spri ngs are omi tted from the di agram. The fi el d of the mi cr~
scope i s i ndi cated at the center of the pl ate.
The probl em of produci ng an exact strai ght-l i ne moti on by a bar
l i nkage was fi rst sol ved by Peaucel l i er. 1 Thi s was accompl i shed by
appl i cati on of the Peaucel l i er i nversor to the conversi on of the ci rcul ar
moti on of a crank i nto a recti l i near moti on.
The Peaucel l i er i nversor i s
i l l ustrated i n Fi g. 2.3. I t consi sts
of a joi nted quadri l ateral wi th four
si des of equal l engths B, to the oppo-
si te verti ces of whi ch ther e are
joi nted two other bars of equal
l engths A; these l atter bars are
themsel ves joi ned at thei r other ends.
Thr ee joi nts of thi s structure neces-
sari l y l i e on the same strai ght l i ne,
and the di stances Xl and XZ between
Fm. 2.3.Si x-bar Peaucel l i er i nversor.
The sol i d l i nes i l l ustrate the case B < A,
the dashed l i nes the case B > A.
these joi nts vary i nversel y wi th each other . I t wi l l be noted that Xl i s
the sum of the l engths of the bases of two ri ght tri angl es of al ti tude T and
hypotenuses A and B respecti vel y, whereas X, i s the di fference of thess
base l engths. We have then
X, = ~A1 T + ~~, (l a)
x, = d~ d~. (l b)
I n these equati ons A, B, T, and the square r oots are necessari l y posi ti ve.
On mul ti pl yi ng together Eqs. (l a) and (l b) we obtai n
XI XZ = AZ Bz, (2a)
or
A2 _ B,
x,= xl . (2b)
Ther e are two vari ants of thi s i nversor, wi th A gr eater than B or wi th
B gr eater than A. I f B i s gr eater than A (dashed l i nes i n Fi g. 2.3), X2 i s
al ways negati ve; ther e i s no possi bi l i ty of havi ng X, equal X2. I f A i s
gr eater than B (sol i d l i nes i n Fi g. 2.3), i t i s possi bl e to have
X, = X2 = (A B2)~.
1A conci sesummaryof wor k i n thi s fi el d, by R. L. Hi ppi el ey,wi l l be found under
Li nkugm, i n the Encycl opedi a Bri tanni ca, 14th ed.
30
BARLI NKAGE COMPUTERS [SEC,2.2
At thi s poi nt the mechani sm exhi bi ts an undesi rabl e si ngul ari ty; the
joi nts P and Q of Fi g. 23 become coi nci dent, and sel f-l ocki ng of the
devi ce may occur .
These two forms of the Peaucel l i er i nversor al so
di ffer i n thei r useful ranges. These are
~<Xl <A+B,
i f A>B, (3a)
BA<X, <A+B, i f A<B. (3b)
The fr eedom from sel f-l ocki ng and the gr eater range make i t desi rabl e
to have B gr eater than A.
Fi gure 2.4 shows the Peaucel l i er i nversor i n a
form sui tabl e for use as a computer.
I 1
FI Q. 2.4.Three-bar PeauceUi er i nvemor . FI G.2.5.The Hart i nversor.
Another i nversor has been devi sed by Hart. I
The Hart i nversor (Fi g. 25) i s essenti al l y a bar-l i nkage paral l el ogram
wi th one pai r of bars r efl ected i n a l i ne through opposi te verti ces. Let any
l i ne OS be drawn paral l el to a l i ne UV through al ternate verti ces of the
quadri l ateral . I t can be shown that thi s wi l l i ntersect adjacent bars of
the l i nkage at poi nts O, P, Q, that remai n col l i near as the l i nkage i s
deformed; furthermore, the di stances Xl = ~ and X2 = ~ wi l l vary
i nversel y wi th each other .
Ther e have been descri bed l i nkages for the traci ng of coni c secti ons,
the Cassi ni an oval , the l emni scatej the l i magon of Pascal , the cardi oi d, and
the tri sectri x; i ndeed i t i s theoreti cal l y possi bl e to descri be any pl ane
cur ve of the nth degr ee i n Cartesi an coordi nates z and y by a bar l i nkage. z
Li nkages for the sol uti on of al gebrai c equati ons have al so been devi sed.8
1H. Hart, @ Certai n Conversi onsof Moti on, Messengero~Mathernatk.s, 4, 82
(1875).
A. Cayl ey, On the Mechani cal Descri pti on of a Cubi c Curve, Pmt. Math, SW.,
Lend., 4, 175 (1872). G. H. Dawson, The Mechani cal Descri pti onof Equi potenti al
Li nes, Proc. Math. Sot., Lo-d, 6, 115 (1874). H. Hart, On Certai n Conversi ons of
Moti on, Messengerof Mathemati cs,4, 82 and 116 (1875); On the Mechani cal
Descri pti onof the Li magonand the Paral l el Moti on Deduced Therefrom, Messenger
of Mathemati cs, 5, 35 (1876); On Some Casesof Paral l el Moti on, Proc. Math. SOC.,
Lend., 8,286 (1876-1877). A. B. Kempe, (On a Gener al Method of Descri bi ng Pl ane
Curvesof the rtthDegr eeby Li nkwork, Pr oc. Math. Sot., Lend., 7, 213 (1875); On
Some New Li nkages, Messenger of Mathematics, 4, 121 (1875). W. H. Laverty,
Extensi onof Peaucel l i ereTheorem, PTOC.Math. Sot., Lend., 6, 84 (1874).
3A. G. Greenhi l l , Mechani cal Sol uti on of a Cubi c by a Quadri l ateralLi nkage,
Messengerof Mathematics, 5, 162 (1876), A. B. Kempe, On the 8ol uti on of Equa-
ti ons by Mechani cal Means, Messengerof Mathemati cs, 2, 51 (1873).
!
Ii
SEC.2.3] DESI GN OF BAR-LI NKAGE COMPUTERS
31
Anal yti cal studi esl have been made of the three-bar moti on of a
poi nt C ri gi dl y attached to the central l i nk AB of a three-bar l i nkage
(Fi g. 26). Three-bar moti on i s ver y useful i n the desi gn of compl ex
computers, and wi l l be di scussed i n Sec. 10.4.
To comdete thi s survey of the bar-l i nkage
c
l i terature i n- Engl i sh, i t wi l l ~uffi ce to menti on t~e
papers of Emch and Hi ppi sl ey on cl osed l i nkages.z
A?
-2.3. The Probl em of Bar-l i nkage-com&ter
Desi gn.-I t i s onl y r ecentl y that much attenti on
fti
,
t
(
I
hae been pai d to the probl em of usi ng bar l i nkages
FI G. 243.Three-bar
i n computi ng mechani sms. The l i terature i n the $=~1$ ,t~~o~,~~ &
fi el d i s especi al l y restri cted. The author knows of centralbar.
onl y one publ i shed wor k that empl oys the syntheti c approach to bar-
l i nkage computer desi gn~and thi s i n a mor e restri cted fi el d than that of
the present vol ume.
The basi c i deas i n the syntheti c approach to bar-l i nkage desi gn are
si mpl e, but qui te di fferent from the i deas behi nd the cl assi cal types of
computers. Bar l i nkages can be characteri zed by a l arge number of
di mensi onal constants, and the fi el d of functi ons that they can gener ate
i s correspondi ngl y l arge-though not i ndefi ni tel y so. Gi ven a wel l -
behaved functi on of one i ndependent vari abl e, one shoul d be abl e to
sel ect from the fi el d of functi ons gener ated by bar l i nkages wi th one degr ee
of fr eedom at l east one functi on that di ffers from the gi ven functi on by a
rel ati vel y smal l amount. The characteri sti c probl em of bar-l i nkage
desi gn i s thus that of sel ecti ng from a fami l y of curves too numerous and
vari ed for effecti ve catal ogi ng one that agrees wi th a gi ven functi on
wi thi n speci fi ed tol erances.
The pr esence of a resi dual er r or sets bar l i nkages apart from other
computi ng mechani sms. The er r or of a computer of cl assi cal type ari ses
from i ts constructi on as an actual physi cal mechani sm, wi th unavoi dabl e
i mperfecti ons. I t i s possi bl e to r educe the er r or to wi thi n al most any
l i mi ts by suffi ci entl y careful desi gn-as, for i nstance, by enl argi ng the
I A. Cayl ey, On Three-bar Moti on, Pmt. Math. Sot., Lend., 7, 136 (1875). R.
L. Hi ppi el ey,A New Method of Descri bi ng a Threebar Curve, Pmt. Math. SoC.,
Lad., 16,136 (1918). W. W. Johnson,On Three-bar Moti on, Memengm of Mathe-
matics,6,50 (1876). S. Roberts, On Three-bar Moti on i n Pl ane Space, Proc. Math.
Sac.,Luzd.,7,14 (1875).
xA. Emch, I l l ustrati on of the El l i pti c I ntegral of the Fi rst Ki nd by a Certai n
Li nk-work, Annul s of Mathemati cs,Seri es 2, 1, 81 (1899-1900). R. L. Hi ppi sl ey,
Cl osedLi nkagea,Pmt. Maih. Sot., Lend., 11,29 (1912-1913); Cl osed Li nkagesand
Pori eti cPol ygons, Proc. Math. Sot., Lomi, 13, 199 (1914-1915).
8Z. Sh.Bl okh and E. B. Karpi n, Practi cal Methods of Desi gni ngFl at Four-si ded
Mechmi sms/ I zdatel stvo Akademi enauk SSSR, Moscow, Leni ngrad(1943). E. B.
Kar@m,Atl as of Nomograms, I zdatel etvo Akademi enauk SSSR, Moscow, Leni n-
grsd (1943).
32 BARLI NKAGE COMPUTERS [SEC.2.4
whol e computer. I n bar l i nkages ther e i s usual l y a resi dual er r or that
cannot be el i mi nated by any car e i n constructi on, an er r or that i s evi dent
i n the mathemati cal desi gn of the devi ce, as wel l as i n the fi ni shed product.
Thi s er r or wi l l be cal l ed structural er r or because i t depends onl y on the
structure of the computer, and not on i ts si ze or other mechani cal pr oper -
ti es. Reducti on of structural er r or requi res a change i n the structure of
the computer-usual l y the addi ti on of parts. The great number of
adjustabl e di mensi onal constants gi ves gr eater fl exi bi l i tyy and extends the
fi el d of functi ons that the l i nkage can generate; from thi s l arger fi el d of
functi ons one can then sel ect a better approxi mati on to the gi ven functi on.
The fact that bar l i nkages can be used to gener ate functi ons of a l arge
cl ass has been known for many years, and has been used (i nsti ncti vel y,
rather than wi th a ful l devel opment of the theor y) by desi gners of mecha-
ni sms. The fi el d of functi ons that can be gener ated by some si mpl e bar
l i nkages has been anal yti cal l y descri bed. Thi s, however , represents onl y
the easi er hal f of the probl em; what one needs i s to descri be the fi el d of
functi ons that can al most be gener ated by a gi ven type of l i nkage. The
fi rst attempts to sol ve thi s probl em for one i ndependent vari abl e have
been tabul ar or graphi cal . For ver y si mpl e structures i t i s possi bl e to
devi se graphs that al l ow one to determi ne whether a gi ven functi on can be
gener ated approxi matel y by such a structure, and what structural er r or
i s i nevi tabl e. These methods are practi cabl e i f the l i nkage can be
speci fi ed by means of onl y two di mensi onal parametersthat i s, i f the
fi el d of functi ons depends upon or dy two adjustabl e parameters. Such
graphi cal methods are di ffi cul t or are necessari l y i ncompl ete i f the fi el d of
functi ons depends upon thr ee adjustabl e parameters. Such a pr ocedur e
can hardl y be attempted when four or mor e di mensi onal parameters are
i nvol ved.
The desi gn methods presented i n thi s book are i n many cases based
on a graphi cal factori zati on of the gi ven functi on i nto functi ons sui tabl e
for mechani zati on by si mpl e l i nkages; the el ements of the mechani sm
desi gned i n thi s way can then be assembl ed i nto the desi red compl ete
l i nkage. By such methods i t i s possi bl e to desi gn l i nkages havi ng a
great many adjustabl e parameters, but the sol uti on obtai ned cannot be
cl ai med to be the best possi bl e. Usual l y i t i s easy to appl y these methods
to fi nd bar l i nkages that have er r or s ever ywher e wi thi n reasonabl e
tol erances. Thi s i s ordi nari l y suffi ci ent for practi cal purposes.
2.4. Characteri sti cs of Bar-l i nkage Computers.The speci al pr oper -
ti es of bar-l i nkage computers may be summari zed as fol l ows.
Advantages.
1. Bar l i nkages occupy l ess space than cl assi cal types of computers.
2. They have negl i gi bl e fri cti on.
~EC.2-41 CHARACTERI STI CS OF BAR-LI NKAGE COMPUTERS 33
3. They have smal l i nerti a.
4. They have great stabi l i ty i n performance.
5. Thei r compl exi ty does not necessari l y i ncrease wi th the compl exi ty
of the anal yti cal formul ati on of the probl em.
6. They are easy to combi ne i nto compl ex systems.
7. They are rel ati vel y cheap.
Di sadvantages.
1. Bar fi nkages usual l y possess a structural er r or .
2. The fi el d of mechani zabl e functi ons i s somewhat restri cted.
3. The compl exi ty of the l i nkage i ncreases wi th decreasi ng tol erances.
4. Li nkage computers are rel ati vel y di ffi cul t to desi gn. The di ffi cul ty
of the desi gn pr ocedur e i ncreases wi th i ncreasi ng compl exi ty and
decreasi ng tol erances.
5. The travel of the mechani sm i s usual l y l i mi ted to a few i nches.
Backl ash er r or and el asti ci ty er r or must be r educed by careful
constructi on: the use of bal l beari ngs i s essenti al , and ri gi di ty of the
structure perpendi cul ar to the pl ane of moti on must be assured.
The desi gn shoul d be such that mechani cal er r or s are l ess than the
awi gned tol erances for structural er r or .
Bar l i nkages can attai n extensi ve use as el ements of computers onl y as
effi ci ent methods of desi gn are establ i shed. The compl exi ty and di ffi cul ty
of the desi gn pr ocedur e depends l argel y on the nature of the gi ven func-
ti on. I t i s usual l y easy to desi gn a l i nkage wi th a structural er r or that
does not exceed 0.3 per cent of the whol e range of moti on of the computer.
I t becomes rel ati vel y l abori ous to r educe the structural er r or bel ow 0.1 per
cent. I f the tol erances are bel ow 0.1 per centas a typi cal val ue
al ternati ves to the use of a bar l i nkage shoul d be expl ored.
Bar Li nkages can advantageousl y be combi ned wi th cams when the
tol erated er r or i s smal l and a bar l i nkage al one woul d be excessi vel y com-
pl ex. For i nstance, i f a gi ven functi on of one i ndependent vari abl e wer e
to be mechani zed wi th an er r or of not mor e than 0.01 per cent, i t mi ght be
desi rabl e to mechani ze thi s functi on by a si mpl e bar l i nkage wi th an er r or
of, for exampl e, 1 per cent, and to use a cam to i ntroduce the requi red
cor r ecti on term. Si nce thi s cor r ecti ve term represents onl y 1 per cent of
the whol e moti on of the l i nkage, i t need not be gener ated wi th ver y hi gh
preci si on; for i nstance, i f the worki ng di spl acement of the cam i s to be
1 i n., i t can be fabri cated wi th a tol erance as rough as 0.01 i n.
I t i s a feature of bar-l i nkage computers that they can be used to
gener ate functi ons of two i ndependent vari abl es i n a ver y di rect and
mechani cal l y si mpl e way. Methods for the desi gn of l i nkages generati ng
functi ons of thr ee i ndependent vari abl es are not now avai l abl e when i t i s
34
BARLI NKAGE COMPUTERS [SEC.2.5
not possi bl e to r educe the probl em to the mechani zati on of functi ons of
one or two i ndependent vari abl es; ther e i s, however , some hope that
practi cal l y useful methods can be found.
Bar-l i nkage computers have great advantages when feedback i s to
beusedi n thedesi gn ofcompl ex computers. I ncomputers of thecl assi cal
type, feedback moti on must be a smal l fracti on of the total output moti on.
Li nkage computers can, however , oper ate ver y cl ose to the cri ti cal feed-
backthat i s, the degr ee of feedback at whi ch the posi ti on of the mecha-
ni sm becomes i ndetermi nate.
2.6. Bar Li nkages wi th One Degr ee of Freedom.Bar l i nkages wi th
one degr ee of fr eedom ser ve the same purpose as cams; they may be
cal l ed l i nkage cams. The paral l el ogram l i nkage of Fi g. 2.2 and the
l i nkage i nversors have moti ons expressed accuratel y by ver y si mpl e
formul as, but they are not general l y useful i n the mechani zati on of gi ven
functi ons. For thi s purposs, the fol l owi ng bar l i nkages are much mor e
i nteresti ng.
The harmoni c transjormer, shown i n Fi g. 2 1, establ i shes a rel ati on
between an angul ar parameter Xl and a transl ati onal parameter X$. I t i s
conveni ent to di sregard vari ati ons i n the form of thi s rel ati on due to
changes i n scal e of the mechani sm-to consi der as equi val ent two geo-
metri cal l y si mi l ar mechani sms.
The fi el d of functi ons
x, = F(X,) (4)
gener ated by the harmoni c transformer then depends upon two rati os of
di mensi ons: L/R and E/R, the rati os to the crank l ength of the l i nk
l ength and the di spl acement of the crank pi vot from the center l i ne of the
sl i de. As L i s i ncreased from i ts mi ni mum val ue, the pl ot of X, agai nst
XI changes (i n a typi cal case) from an i sol ated poi nt to a cl osed curve,
then to a si nusoi d, and fi nal l y, i n the l i mi t as L approaches i nfi ni ty, to a
pure si nusoi d. From a practi cal poi nt of vi ew, the pure si nusoi dal form i s
reached for l i nks short enough for practi cal use. I n the l i mi ti ng case,
L = w, the equati on of the harmoni c transformer i s
X,= Rsi n Xl +C.
(5)
Such a harmoni c transformer wi l l be cal l ed i deal .
Onl y rarel y i s the compl ete range of moti on of a harmoni c tra~former
used. When the range of the parameter X, i s l i mi ted to Xl ~ < X, < Xl ..
and the functi ons defi ned wi thi n these restri cted l i mi ts are taken as el e-
ments of a new functi onal fi el d, ther e i s obtai ned a four-di mensi onal
functi onal fi el d dependi ng on Xl ~ and Xl ,, as wel l as on L/R and E/R.
Methods for the desi gn of harmoni c transformers wi l l be di scussed i n
Chap. 4.
SEC.2.5] BAR LI NKAGES WI TH ONE DEGREE OF FREEDOM
35
The three-bar l i nkage shown i n Fi g. 27 consi sts of two cranks pi voted
to a frame and joi ned at thei r fr ee ends by a connecti ng l i nk. As a
computer, thi s serves to compute the parameter X2 as a functi on of
the parameter Xl . The l i nkage i tsel f i s descri bed by four l engths: A 1, B1,
AZ, Bt. The fi el d of functi ons gener ated by thi s type of l i nkage i s onl y
three-di mensi onal , because two geometri cal l y si mi l ar mechani sms estab-
FI G.2.7.Three-bar l i nkage.
FI G. 2.8.Three-barl i nkage modi fi ed by
eccentri c l i nkage.
fi sh the same rel ati on between Xl and X2. The fi el d of functi ons thus
depends on thr ee rati osfor exampl e, B1/A 1, Az/A,, and Bz/ A I . Usu-
al l y onl y a part of the possi bl e moti on of the mechani sm i s used. Li mi ts
of moti on can be assi gned for Xl or X2, though, of course, not i ndepend-
entl y for the two parameters; for i nstance, one may fi x Xl ~ < Xl < X 1~.
Thi s i ncreases the number of i ndependent parameters by two; the fi el d of
functi ons gener ated by a three-bar l i nkage operati ng wi thi n fi xed l i mi ts
/
1
FI G. 2.9.Harmoni c transformer modi fi ed by eccentri c l i nkage.
i s fi ve-di mensi onal . I n Chap. 5 we shal l see how to desi gn a three-bar
l i nkage for the approxi mate generati on of a gi ven functi on.
The euxntri c l i nkage i s not a bar l i nkage, but i s so conveni entl y used i n
connecti on wi th bar l i nkages that i t shoul d be menti oned her e. F@re
2.8 shows a three-bar l i nkage modi fi ed by the i nserti on of an eccentr i c
l i nkage. One crank of the three-bar l i nkage carri es a pl anetary gear that
meshes wi th a gear fi xed to the frame. The central l i nk i s then pi voted
eccentri cal l y to the pl anetary gear, rather than to the crank i tsel f. Li nk-
ages of thi s type wi l l be di scussed i n Sec. 7.9, wher e thei r i mportance wi l l
36 BAR-LI NKAGE COMPUTERS [SEC.2.5
FI G.Z.I O.Doubl e three-bar l i nkage gener ati ng the l ogari thmi c functi on.
be expl ai ned. Another i mportant appl i cati on of the eccentr i c l i nkage i s i n
the modi fi cati on of harmoni c transformers, as i l l ustrated i n Fi g. 29. I t i s
possi bl e to choose the constants of
the eccentr i c l i nkage i n such a way
that the l i nkage output i s an
al most per fect si n us o i d, even
though the l ength of the l i nk L i s
rel ati vel y smal l .
Combi nati ons of these l i nkages
to be di scussed i n thi s book are
the doubl e harmoni c transformer
FI G.2.1I .Bar l i nkage wi th two degr ees of
fr eedom.
(Sec. 4.9 and fol l owi ng), harmoni c
transformers i n seri es wi th thr ee-
bar l i nkages (Sec. 8.1 and foI l owi ng), and the doubl e three-bar l i nkage
(Sec. 88). Fi gure 210 shows a doubl e three-bar l i nkage that generates
the l ogari thmi c functi on through the range i ndi cated i n the fi gure.
8EC.2.6] BAR LI NKAGES WI TH TWO DEGREES OF FREEDOM 37
2.6. Bar Lfnkages wi th Two Degr ees of Freedom.Bar l i nkages wi th
two degr ees of fr eedom can be used i n the generati on of al most any
wel l -behaved functi on
X3 = F(xl , x2) (6)
of two i ndependent vari abl es. They provi de a mechani cal l y sati sfactory
substi tute for threedmensi onal cams, whi ch have many di sadvantages
and are to be avoi ded i f possi bl e.
Fi gure 211 shows a l i nkage wi th
two degr ees of freedom, whi ch
consi sts of thr ee cranks connected
by two l i nks and a l ever. The
l ever wi l l degener ate i nto a si mpl e
l i nk i f the pi vots A and B are
superposed; the resul ti ng struc-
tur e of thr ee l i nks joi nted at a
si ngl e pi vot wi l l be cal l ed a star
l i nkage. I ts properti es are di s-
cussed i n Chap. 9.
The bar-l i nkage adder shown
i n Fi g. 2.12 consi sts of essenti al l y
Fm. 2.12.Bsr-l i nkageadder.
the same parts as the l i nkage of
Fi g. 211, except that sl i des are used i nstead of cranks to constrai n the
l i nks. The di mensi ons obey the si mpl e rel ati on
A, 1?, C,
.
A, B, = r,
(7)
I t i s easy to show that when thi s proporti onal i ty hol ds, the thr ee pi vots
P,, P%, and Pa l i e on a strai ght l i ne. Thi s devi ce can, ther efor e, be used
to mechani ze any al i gnment nomogram that consi sts of thr ee paral l el
strai ght l i nes; i n parti cul ar, i t can be used to mechani ze the wel l -known
npmogram for addi ti on. I f Xl , X2, and X3 are thr ee parameters measured
al ong these l i nes i n the same di recti on from a common zer o l i ne, then
(A, + A2)X3 = A,XI + A,X,.
(8)
Thi s bar l i nkage i s fr ee from structural er r or .
I n contrast to the adders, baT-l i nkage mul ti pl i ers do not per for m the
operati on of mul ti pl i cati on exactl y, but wi th a smal l er r or ; the equati on of
such a mul ti pl i er i s
RX3 = X,X, + 6, (9)
wher e 8, the er r or of the mul ti pl i er, i s a functi on of the two i ndependent
parameters Xl and X2. The desi gn of mul ti pl i ers wi l l be di scussed i n
Chap. 9; a much si mpl i fi ed expl anati on of the pri nci pl e wi l l be gi ven her e.
38
BAR-LI NKAGE COMPUTERS
[SEC.2.6
Fi gure 2.13 shows the essenti al el ements of one type of mul ti pl i er.
Thr ee bars of equal l engths, El = RZ = Rs = 1, are pi voted together .
The fi rst i s pi voted al so to the frame at the poi nt O, the thi rd to a sl i de
wi th center l i ne passi ng through O.
I f the joi nts A 1 and A z are pl aced
at di stances Xl and XZ from the center l i ne of the sl i de, the di stance
OS = D wi l l be exactl y
D=~l X; @-(X,-X,) +~1X;.
(l o)
Expandi ng i n seri es the terms on the ri ght, one obtai ns
wher e X3 i s the di spl acement of the pi vot S from the posi ti on SOwhi ch i t
occupi es when Xl = X2 = O and
i
the thr ee l i nks are coi nci dent. I t
i s evi dent that X3 i s equal to the
product XI XZ to the approxi ma-
X2
---
ti on i n whi ch the terms of fourth
and hi gher degr ees can be neg-
l ected i n compari son wi th the
term of the second degr ee. For
FI G,
suffi ci entl y smal l val ues of Xl and
2.13.El ements of a bar-l i nkage
X2 thi s mechani sm i s thus a
mul ti pl i er. .
mul ti pl i er for these parameters.
Such a mul ti pl i er i s not practi cal , however , because of i ts smal l range of
moti on. I f the er r or i n the mul ti pl i cati on i s to be kept bel ow 1 per cent,
i t i s necessary to keep Xl , Xa s 0.2. [I f X, = X, = 0.2, then
x3 = (0.2) + +(0.2)4 + . . . ,
and the fracti onal er r or i s al most exactl y one per cent. ] Under these
condi ti ons, however , one has Xa = 0.04, an i mpracti cabl y smal l range of
moti on.
G
Ther e are i n pri nci pl e two ways to i mprove thi s mul ti pl i er. Wi th
ei ther method i t i s necessary to make the structure mor e compl i catedto
add new adjustabl e parameters. One possi bl e arrangement i s i ndi cated
i n Fi g. 214. Her e the parameter Xa i s a di spl acement of a sl i de (of
adjustabl e posi ti on) that control s the posi ti on of the joi nt A a through a
l i nk of adjustabl e l ength L,; X3 becomes an angul ar parameter, the angl e
turned by a crank wi th adjustabl e l ength and pi vot posi ti on.
Wi th the fi rst method, the output parameter X3 i s expressed i n
terms of X, and X,, i n the form of a seri es wi th coeffi ci ents whi ch depend
on the adjustabl e di mensi ons of the mechani sm.
These di mensi ons can
then be so chosen as to cause the terms of the fourth degr ee i n Xl and Xz
SEC.2.6] BAR LI NKAGES WI TH TWO DEGREES OF FREEDOM
39
to vani sh. I n thi s way, the mul ti pl i er can be made mor e accurate for
smal l val ues of Xl and X2, and the domai n of useful accuracy sub-
stanti al l y i ncreased. Toward the l i mi ts of thi s domai n, however , the
i naccurac~ of the mul ti pl i er wi l l
i ncrease ver y rapi dl y.
Ap
.. .
The second method for i m-
provi ng the mul ti pl i erthat
fol l owed i n thi s bookcan be i ndl - ~,
cated onl y ver y roughl y at thi s
.. . - -
poi nt. I t i nvol ves compari son of
the i deal product and the functi on
] [~
I
actual l y gener ated by the mul ti -
pl i er over the enti re range of
1[!
moti on, and adjustment of the
di mensi onal constants. of the sys-
FI G.2.14.Modi fi edbar-l i nkagemul ti pl i er.
ter n i n such a way that the er r or of the mechani sm i s brought wi thi n
speci fi ed tol erances ever ywher e wi thi n thi s domai n. To see i n pri nci pl e
how thi s can be done, l et us consi der the mechani sm of Fi g. 2.13. Let
X3 and Xl be gi ven a seri es of val ues that have the fi xed rati o
(12)
I f thi s l i nkage wer e an exact mul ti pl i er, the pi vot A, woul d i ndi cate al ways
the same val ue of X2; i t woul d move al ong a strai ght l i ne at constant
di stance X; from the l i ne of the sl i de. Actual l y, the pi vot A z wi l l descri be
a cur ve that i s tangent to thi s strai ght l i ne for smal l val ues of Xl and X3,
but wi l l di ver ge from i t as these parameters i ncrease. To each val ue of
X2 ther e wi l l cor r espond another curve; the curves of constant X2 form a
fami l y, each of whi ch can be l abel ed wi th the associ ated val ue of thi s
parameter. Now we can make thi s mul ti pl i er exact i f we can i ntroduce
a constrai nt whi ch, for any speci fi ed val ue of X2, wi l l hol d the pi vot A ~
on the correspondi ng cur ve of thi s fami l y. For exampl e, i f these curves
wer e al l ci rcl es wi th the same radi us Lz and center s l yi ng on a strai ght
l i ne, i t woul d be possi bl e to use the type of constrai nt i l l ustrated i n Fi g.
2.14. The X,-sl i de coul d then be used to bri ng the pi vot A a to the center
of the ci rcl e correspondi ng to an assi gned val ue of X2, and the pi vot
Az woul d stay on that ci rcl e, as requi red. Actual l y, the curves of con-
stant X2 wi l l not form such a fami l y of i denti cal ci rcl es. I t wi l l , however ,
be possi bl e to approxi mate them by such ci rcl es i n a way whi ch wi l l spl i t
the er r or and bri ng i t wi thi n tol erances hel d fai rl y uni forml y over the
whol e domai n of acti on. Unl i ke the mul ti pl i ers desi gned by the fi rst
method, a mul ti pl i er thus desi gned wi l l not have unnecessari l y smal l er r or s
i n one part of the domai n and excessi vel y l arge er r or s i n another part.

40 BAR-LI NKAGE COMPUTERS [SEC.2.7

Thi s concept of mul ti pl i er desi gn must be ver y greatl y extended befor e
i t can l ead to the desi gn of sati sfactory computers. A powerful gui de i n
begi nni ng the wor k i s provi ded by the i dea of nomographi c mul ti pl i ers,
al ready di scussed i n Sec. 1.5. I t i s possi bl e to desi gn approxi mate i nter-
secti on nomograms for mul ti pl i cati on that have as thei r mechani cal
anal ogues bar l i nkages wi th two degr ees of freedom.
For i nstance, I Hg.
8 14 shows a nomogram for mul ti pl i cati on obtai ned by topol ogi cal trans-
formati on of the nomogram of Fi g. 1.12; i t consi sts of two fami l i es of
i denti cal ci rcl es and a thi rd fami l y of curves that can be ver y cl osel y
approxi mated by a fami l y of i denti cal ci rcl es. Thi s nomogram cor -
responds to the bar-l i nkage mul ti pl i er i l l ustrated i n Fi g. 8.15, whi ch, on
i mprovement of i ts mechani cal features, takes on the form shown i n Fi g.
8.16. The desi gn techni ques to be descri bed i n Chaps. 8 and 9 make i t
possi bl e to desi gn mul ti pl i ers wi th l arge domai n of acti on and good
uni formi ty of performance through thi s domai n.
Mul ti pl i ers can be used to per for m the i nverse operati on_of di vi si on;
that i s, they can be used to eval uate X2 = Xs/X1. I t i s, of course, not
possi bl e to di vi de by zer o; when a mul ti pl i er i s used i n thi s way Xl wi l l
never pass through zer o. I t i s ther efor e usel ess to attempt to r educe to
zer o the er r or of such a mul ti pl i er for val ues of X1 ver y near to zer o; i t
i s al so undesi rabl e to attempt to r educe the er r or s of the devi ce for nega-
ti ve val ues of Xl when onl y posi ti ve val ues can be i ntroduced. For thi s
reason thr ee types of mul ti pl i er may be di sti ngui shed.
1. Ful l -range mul ti pl i ers, for whi ch both i nput parameters can change
si gns.
2. Hal f-range mul ti pl i ers, for whi ch onl y one parameter can change
si gns.
3. Quarter-range mul ti pl i ers, for whi ch nei ther i nput parameter can
change si gns.
Di vi ders may be di vi ded i nto two types.
1. The pl us-mi nus type, for whi ch the numerator may change si gn.
2. The si ngl e-si gn type, for whi ch al l -parameters have fi xed si gns.
An exampl e of a practi cal ful l -range l i nkage mul ti pl i er i s shown i n Fi g.
8.16; a hal f-range mul ti pl i er i s shown i n Fi g. 915.
2.7. Compl ex Bar-l i nkage Computers.The el ementary l i nkage cel l s
al ready descri bed may be combi ned to form compl ex computers. Si nce
si mpl e l i nkages can add, mul ti pl y, and gener ate functi ons of one and two
i ndependent vari abl es, bar-l i nkage computers can sol ve any probl em that
can be expressed i n a system of equati ons i nvol vi ng onl y these operati ons.
The fi el d of appl i cati on of bar-l i nkage computers i s qui te l arge; they

SEC.2.7] COMPLEX BAR-LI NKAGE COMPUTERS

41
are especi al l y useful i f the computer must be l i ght, as when i t i s to be
carri ed i n ai rcraft or gui ded mi ssi l es.
An i mportant feature of bar-l i nkage computers i s the ease wi th whi ch
the cel l s can be assembl ed i nto a compact uni t. I t i s natural to spread
the parts of the computer out i n a pl ane, to pr oduce a rather fl at mecha-
ni sm wi th i ts parts easi l y accessi bl e.
The connecti ons between cel l s
areprovi ded by shafts or connecti ng bars.
Ther e i s a si mpl e tri ck that makes the connecti on of l i nkage cel l s even
easi er, and the structure of some cel l s l ess compl ex.
The si mpl i fi cati on
of l i nkage adders i s a characteri sti c
exampl e of thi s tri ck. The bar-l i nk- _ =
age adder shown i n Fi g. 2.12 has no
structural er r or . Any devi ati on from
the pri nci pl e of thi s desi gn i s l i kel y to
l ead to a structural er r or ; i t i s, how-
=
ever , possi bl e to change the pri nci pl e -
i n such a way that the structural er r or
x
i s negl i gbl y smal l . For i nstance, i f
FI G.2.15.Bw-l i nkageadder (approxi -
the l i nks B, and Bz are ver y l ong,
mate),
thei r l engths can be chosen at wi l l wi thout appreci abl y affecti ng the
accuracy of the addi ti on. Fi gure 2.15 shows such an approxi mate adder;
i ts equati on i s
(A, + A,)x, = AI X, + A,XZ. (13)
The l i nks L, and Lz must be so l ong that they l i e nearl y paral l el to the l i nes
of the sl i de, but they need not be exactl y paral l el to each other . The
acti on of thi s devi ce depends upon the essenti al constancy of the pr ojec-
ti on of the l engths of these bars al ong the l i ne of the sl i des. Let X4, x;,
and X: be defi ned as the di stances of the pi vots PI , Pz, and P~ from some
zer o l i ne perpendi cul ar to the l i ne of the sl i des. One then has, exactl y,
(A, + A,)X{ = A,X{ + A,X;. (14)
NOWl et 131be the angl e between the bar L1 and the l i ne of the sl i des.
Then
X,= X;+ LI COS(?, +C, (15a)
= X; L,(I COS6,) + (c + L,).
(15b)
Except for an addi ti ve constant (whi ch can be r educed to zer o by pr oper
choi ce of the zer o poi nt), X{ and Xl cl i ffer onl y by the vari abl e term
Ll (l cos 8,). As LI i s i ncreased, 01 decreases wi th I /L,, (1 cos 01)
decreases wi th l /L~, and L,(l cos 0,) decreases wi th l /Ll . Thus, by
maki ng L1 l arge and pr oper l y choosi ng the zer o poi nt, one can make Xl
and X; di ffer by a negl i gi bl y smal l term. I n the same way X2 can be
made negl i gi bl y di fferent from X4; Xa and X{ are i denti cal . Equati on
I
I
I
1
I
I
I
42 BAR-LI NKAGE COMPUTERS [SEC.2.7
(13) fol l ows as an approxi mati on to Eq. (14). I f 0, i s kept l ess that
0.035 radi ans (about 2) the di fference between Xi and X{ wi l l be about
0.0006 LI . Thus i f the bars devi ate from paral l el i sm wi th the sl i des by no
mor e than &2 duri ng operati on of the adder, the resul ti ng er r or i n the
output wi l l not exceed 0.06 per cent of the total l ength of the bars.
I f the l engths of the bars i n approxi mate adders are great enough, i t i s
even i mmateri al whether the sl i des move al ong strai ght l i nes; the essenti al
-7L-
*
~
FI G.2.16.Combi nati onof approxi mateadders.
1
thi ng i s that the parameters be measured as di stances from a zer o l i ne.
I t i s, ther efor e, possi bl e to connect addi ng cel l s through l ong connecti ng
bars, and to omi t some of the sl i des that woul d appear i n the standard
constructi on. Fi g. 2.16 shows a combi nati on of thr ee addhg cel l s that
wi l l sol ve (approxi matel y) the equati ons
(A, + A,)X, = A,X, + A,X,,
(D, + D,)X, = D,X, + D,X,,
1
(16)
(E, + E,)X, = E=, + E,X,.
CHAPTER 3
BASI C CONCEPTS AND TERMI NOLOGY
The present chapter wi l l defi ne the termi nol ogy to be empl oyed i n
di scussi ng bar-l i nkage desi gn and i ntroduce some co~cepts wi th wi de
appl i cati on i n the fi el d., Of parti cul ar i mportance are the concepts of
homogeneous parameters and homogeneous vari abl es, and a graphi -
cal cal cul us used i n di scussi ng the acti on of computi ng mechani sms i n
seri es.
3.1. Defi ni ti ons. I deal Functi onal Mechani sm.-Any mechani sm
can be used as a computer i f i t establ i shes defi ni te geometri cal rel ati ons
between i ts parts-that i s, i f i t i s suffi ci entl y ri gi d and fr ee from backl ash,
FI G. 3.1.Crank termi nal .
Zem
posi ti on
FI G.3.2.Sl i de termi nal ,
sl i ppage, or mechani cal pl ay. I n the fol l owi ng di scussi on we shal l be
concer ned onl y wi th such i deal functi onal mechani sms.
Termi nul s.-The termi nal s of a computi ng mechani sm are those el e-
ments that, by thei r moti ons, r epr esent the vari abl es i nvol ved i n the
computati on. The moti on of al l termi nal s i s usual l y speci fi ed wi th
r espect to some common frame of r efer ence. I f the posi ti on of a termi nal
i s control l ed i n or der to fi x the confi gurati on of the mechani sm, i t may be
cal l ed an i nput termi nal ; i f i ts posi ti on i s used i n control l i ng a second
mechani sm, or i s si mpl y observed, i t maybe cal l ed an output termi nal .
A termi nal maybe sui tabl e for use onl y as an i nput termi nal , or onl y as an
output termi nal , or i n ei ther way, accordi ng to the nature of the mechani sm.
43
44 BASI C CONCEPTS AND TERMI NOLOGY [SEC. 3.1
Termi nal s that are mechani cal l y practi cal are of two ki nds:
1. Crank or rotati ng-shaft termi nal s (Fi g. 31), whi ch r epr esent a
vari abl e by thei r angul ar moti on.
2. Sl i de termi nal s (Fi g. 3.2), whi ch r epr esent a vari abl e by a l i near
moti on.
Paranwters.-A parameter i s a geometri cal quanti ty that speci fi es the
posi ti on of a termi nal . Wi th a crank termi nal , i t i s usual l y the angul ar
posi ti on of the termi nal wi th r espect to some speci fi ed zer o posi ti on; wi th a
sl i de termi nal , i t i s usual l y the di stance of the sl i de from a zer o posi ti on.
Parameters may be dei i ned i n other ways for i nstance, as the di stance
of a sl i de termi nal from some movabl e el ement of the mechani sm-but such
parameters are l ess general l y useful than those just menti oned.
An i nput parameter descri bes the posi ti on of an i nput termi nal , an
output parameter that of an output termi nal .
Li nkage Computers.-A l i nkage computer establ i shes between i ts
parameters, Xl , X2, . . . X,, defi ni te rel ati ons of the form
I,(X1, x2, x.) = o, r=l ,2, ..., (1)
whi ch i nvol ve onl y these parameters and the di mensi onal constants of the
mechani sm. Wi th mor e general types of mechani sms these equati ons of
moti on may al so i nvol ve deri vati ves of the parameters. Such mecha-
ni sms are useful i n the sol uti on of di fferenti al equati ons, but they wi l l be
excl uded from our future consi derati ons; we shal l be concer ned onl y wi th
l i nkage computers, whi ch gener ate fi xed functi onal rel ati ons between the
parameters.
To descri be the confi gurati on of l i nkage computers wi th n degr ees of
freedom, one must i n general speci fy the val ues of n i nput parameters,
x,, x,, . . . Xn. The val ues of any number of output parameters can
then be expressed expl i ci tl y i n terms of these n parameters:
X.% = G, (Xl , X2. - x.),
r=l ,2, ?n. (2)
Dom.ui n.-The parameters of a computi ng mechani sm cannot, i n
general , assume al l val ues. The l i mi tati ons may ari se from the geometri -
cal nature of the mechani sm (a l i near di mensi on wi l l never change wi thout
l i mi t) or from the way i n whi ch i t i s empl oyed. To each possi bl e set of
val ues of the i nput parameters Xl , . . . Xm, ther e corresponds a poi nt
(x,, x2, . . . X,,) i n n-di mensi onal space; to al l sets of val ues that may
ari se duri ng a speci fi c appl i cati on of the mechani sm, ther e corresponds a
domai n i n n-di mensi onal space, whi ch wi l l be r efer r ed to as the domai n
of the parameters. I t must be emphasi zed that the domai n of the param-
eter s i s not necessari l y determi ned by the structure of the mechani sm,
but by the task set for i t.
SEC.3.1]
DEFI NI TI ONS 45
I n the most general case, the domai n of the i nput parameters maybe of
arbi trary form-except, of course, that i t must be si mpl y connected,
si nce al l parameters must change conti nuousl y. I n such cases the val ues
possi bl e for any one parameter may depend on the val ues assi gned to other
parameters. A mechani sm wi l l be sai d to be a regul ar mechani sm
when each i nput parameter can vary i ndependentl y of al l others, between
defi ni te upper and l ower l i mi ts,
(3)
whi ch defi ne the dai nai n of the parameter. Wi th angul ar parameters,
nei ther of these l i mi ts i s necessari l y fi ni te: i t i s possi bl e to have Xi ~ = cu,
or Xi ~
=+ Co.
The output parameters of a regul ar mechani sm wi l l vary between
defi ni te (though not necessari l y fi ni te) l i mi ts as the i nput parameters
take on al l possi bl e val ues. These l i mi ts ser ve to defi e a domai n for
each output parameter. Al though the i nput parameters vary i nde-
pendentl y through thei r r especti ve domai ns, thi s i s hot al ways tr ue of the
output parameters.
Travel .The range of moti on of a termi nal i s cal l ed i ts travel .
Thi s i s
AXi = Xi ~ _ Xi mj (4)
both for i nput and output termi nal s:
Vw-i ubfes.-The term vari abl e wi l l denote the vari abl es of the
probl em whi ch the computi ng mechani sm i s desi gned to sol ve. A vari a-
bl e wi l l be associ ated wi th each termi nal of a mechani sm, an i nput vari abl e
wi th an i nput termi nal , an output vari abl e wi th an output termi nal . To
each val ue of a vari abl e ther e wi l l cor r espond a defi ni te confi gurati on of
the termi nal ; each vari abl e, then, wi l l be functi onal l y rel ated to a param-
eter of the mechani sm:
Xi= @i(Xi).
~=l ,g,...o
(5)
I t i s i mportant to keep i n mi nd the di sti ncti on between parameters,
whi ch are geometri cal quanti ti es measured i n standard uni ts, and the
vari abl es of the probl em, whi ch are onl y functi onal l y rel ated to the param-
eters. I n thi s book, vari abl es wi l l be denoted by l ower-case l etters,
parameters by capi tal s.
Scul es.-The val ue of the vari abl e correspondi ng to a gi ven confi gura-
ti on of a termi nal can be read from a scal e associ ated wi th that termi nal .
The cal i brati on of thi s scal e i s determi ned by the form of the functi onal
rel ati on between z and Xi . I f xi i s a l i near functi on of Xi the scal e wi l l
be emnthat i s, evenl y spaced cal i brati ons wi l l cor r espond to evenl y
BpSCedval ues of xi . Such a scal e may al so be r efer r ed to aa l i near,
46 BASI C CONCEPTS AND TERMI NOLOGY [SEC.3.1
i n r efer ence to the form of the functi onal rel ati on r epr esented. (Thi s term
does not descri be the geometri cal form of the scal e, whi ch maybe ci rcul ar.)
A l i near termi nal i s a termi nal wi th whi ch ther e i s associ ated a l i near
scal e.
Range of a Vari abl e.As a parameter changes between i ts l i mi ts,
Xi n and Xi ~, the associ ated vari ~bl e wi l l al so change wi thi n fi xed, but not
necessari l y fi ni te, l i mi ts:
Xims Xis Xild. (6)
I n the case of a regul ar mechani sm, thi s may be r efer r ed to as the
domai n of the vari abl e; i ts range i s
Ax~ = XW Z& (7)
Mechani zati on of a Functi on.An i deal functi onal mechani sm estab-
l i shes defi ni te rel ati ons between i ts parameters:
F,(Xl , X2, . ) = O, r=l ,2, ... (8)
I t maybe sai d to provi de a mechani zati on of these functi onal rel ati ons
wi thi n the gi ven domai n of the i ndependent parameters.
Such a mechani sm, together wi th i ts associ ated scal es, si mi l arl y pr o-
vi des a mechani zati on of functi onal rel ati ons,
j,(zl , x*, ) = o, 7=1,2,, (9)
between the vari abl es xi , wi thi n a gi ven domai n of the i ndependent vari a-
bl es. The forms of these rel ati ons may be deri ved by el i mi nati ng the
val ues of the parameters Xi between Eq. (8), whi ch characteri zes the
mechani sm, and Eq. (5), whi ch characteri zes the scal es.
FI G.3.3.I nput scal e.
I f the output vari abl es are to be si ngl e-val ued functi ons of the i nput
vari abl es, the i nput parameters must be si ngl e-val ued functi ons of the
i nput vari abl es, and the output vari abl es must be si ngl e-val ued functi ons
of the output parameters; i t i s not, however , necessary that the i nverse
rel ati ons be si ngl e-val ued. Thus an i nput scal e may have the form shown
i n Fi g. 3.3, and an output scal e that shown i n Fi g. 3.4, but not the r ever se.
Li near Mechani zati on.A mechani zati on of a rel ati on between vari a-
bl es wi l l be ter med a l i near mechani zati on i f al l scal es are l i near.
SEC.32] HOMOGENEOUS PARAMETERS AND VARI ABLES
47
A nonl i near mechani zati on of a gi ven functi on may be useful when
i nput vari abl es are set by hand, and onl y a readi ng of the output vari abl es
i s requi red. When a computi ng mechani sm i s to be part of a mor e com-
pl ex devi ce, i t i s usual l y necessary that the termi nal s have mechani cal
moti on proporti onal to the change i n
the associ ated vari abl ethat i s, a l i near
mechani zati on of the functi on i s needed.
For i nstance, i f one has onl y to com-
pute the superel evati on angl e for an
anti ai rcraft gun i t may be qui te sati s-
factory to read thi s on an unevenl y
di vi ded scal e. I f, however , one wi shes
to use the computer to control di rectl y
the si ght on a gun, then a l i near mechw
ni zati on of the superel evati on functi on
wi l l be requi red.
I t i s a tri vi al matter to desi gn a
nonl i near mechani zati on of a functi on
of one i ndependent vari abl e. One
FI G.3.4.+utput scal e.
requi res onl y a si ngl e poi nter, servi ng
both as i nput and output termi nal , to i ndi cate correspondi ng val ues of
i nput and output vari abl es as paral l el scal es (Fi g. 3 5). For thi s reason
the term mechani zati on as appl i ed to functi ons of a si ngl e i ndependent
vari abl e wi l l al ways denote l i war mechani zati on; a di sti ncti on wi l l be
made between l i near and nonl i near mechani zati on onl y i n the case of
l i nkages of two or mor e degr ees of freedom.
Fm. 3.5.Nonl i near mechani zati on of a functi on of one i ndependent vari abl e.
3.2. Homogeneous Parameters snd Vari abl es.-Homogeneous vari -
abl es and parameters are ver y useful tool s i n the desi gn of i ndi vi dual
computi ng l i nkages, and al so i n the drawi ng up of schemati c di agrams for
compl ex computers. They are defi ned or dy for vari abl es and parameters
whi ch vary wi thi n fi ni te l i mi ts.
Associ ated wi th each vari abl e xi havi ng a fi ni te range Axi i s a homo-
geneous vari abl s defi ned by
hi =
Xi Z{m
Zi Af X*.
(10)
48
BASI C CONCEPTS AND TERMI NOLOGY [SEC.3.2
As w vari es from i ts l ower to i ts upper bound, h, vari es l i nearl y wi th i t,
from Oto 1. The i nverse form of Eq. (10) maybe wri tten
Xi= Xifn+ LAX<. (11)
Another homogeneous vari abl e, compl ementary to hl , i s defi ned by
(12)
or by
h<+~=l . (13)
I n the same way, ther e are associ ated wi th ,each parameter Xi , havi ng
a fi ni te travel AXi j two compl ementary homogeneous parameters,
~, = Xi Xi .
,
Xi 3f Xi m
(14)
~i =l _ Hi , (15)
whi ch change l i nearl y wi th Xi between bounds O and 1:
Xi = Xi . + HAXi = Xi . ~&Xi . (16)
I n a l i near mechani zati on, the homogeneous vari abl es and parameters
are ver y si mpl y rel ated. The quanti ti es Xi and x, are connected by a
l i near rel ati on,
Xi XjO) = ki (Z, Zj). (17)
I f k, i s posi ti ve, the mi ni mum val ues of X, and xi occur together , as do the
maxi mum val ues:
Xi ~ XjO = Jli(Zim X$)), (18a)
(ki > 1)
X,M X~~ = ki (Zi ~ ZjO). (18b)
I t fol l ows by i ntroducti on of these rel ati ons i nto Eqs. (10) and (14) that
H, ~ ~. (ki > 1). (19)
I f k<i s negati ve, the maxi mum val ue of X< occurs together wi th the mi ni -
mum val ue of ~i , and conversel y:
Xi m X!) = lCi(Zilf Z\O))j (20a)
(k< 1)
Xi .
jyj) = ~i (~i m ~.ol) ;
(20b)
then
Hi ~fi =l _hi . (ki < 1). (21)
Equati on (19) wi l l be r efer r ed to as the di rect i denti fi cati on of Hi
wi th hi . I t i mpl i es that Xi and Zi are l i nearl y dependent on each other ,
SEC.3.3] AN OPERATOR FORMALI SM 49
changi ng i n the same sense between mi ni mum and maxi mum val ues
whi ch they attai n si mul taneousl y; the scal e of xi i s even, and i ncreases i n
the di recti on of i ncreasi ng X;.
Equati on (21) wi l l be ter med the com-
pl ementary i denti fi cati on of Hi and h,; i t i mpl i es that the scal e of Zi i s
even, and i ncreases i n the di recti on of decreasi ng Xi .
I n terms of homogeneous vari abl es, the probl em of l i nearl y mechani z-
i ng a gi ven functi on takes on a parti cul arl y si mpl e form. For i nstance,
i f the gi ven functi on i nvol ves a si ngl e i ndependent vari abl e, i t may be
expressed i n terms of a homogeneous i nput vari abl e hi and a homogeneous
output vari abl e hz:
h, = j(hJ. @2)
A l i nkage wi th one degr ee of freedom, operati ng i n a speci fi ed domai n of
the i nput parameter,
X;. S Xi S Xi l ,f, (23)
wi l l gener ate a rel ati on between homogeneous i nput and output param-
eters, H.1and Hz, respecti vel y:
H, = F(H,). (24)
I t i s then requi red to fi nd a mechani sm and domai n of operati on such that
Eq. (24) can be transformed i nto the gi ven Eq. (22) by di rect or compl e-
mentary i denti fi cati on of HI wi th hl , wi th Hz wi th hz.
The useful ness of homogeneous parameters and vari abl es wi l l be
abundantl y i l l ustrated i n the chapters to fol l ow.
3.3. An Oper ator Formal i sm.-I t i s often necessary to combi ne
mechani sms i n seri es, i n such a way that the output parameter of the
!i .rst becomes the i nput parameter of the second, and so on. The fi rst
mechani sm determi nes an output parameter X2 as a functi on of the i nput
parameter Xl :
x, = +1(X1). (25a)
The second mechani sm determi nes an output parameter Xa i n terms of Xz,
X3 = +2(X2); (25b)
the thi rd determi nes an output parameter Xq i n terms of X3,
x, = @*(x,); (25c)
and so on. The fi nal output parameter, for exampl e, Xq, i s then deter -
mi ned as a functi on of Xl :
x4 = q$,{ @,[+,(x,)]]. (26)
The conventi onal notati on of Eqs. (25) and (26) i s ful l y expl i ci t, but some
ti mes cumbersome. For many purposes the author fi nds i t mor e con-
veni ent and mor e suggesti ve to use the f ol l owi ng oper ator notati on.
50 BASI C CONCEPTS AND TERMI NOLOGY [SEC.3.3
Equati on (25a) i mpl i es that the val ue of X2 can be obtai ned by carry-
i ng out an operati on (of character speci fi ed by the defi ni ti on of 01) on the
val ue of Xl . As an al ternati ve notati on we shal l wri te
x, = (X21X1) xl, (27a)
wher e (XZI XI ) denotes an oper ator converti ng the parameter Xl i nto the
parameter X2. Si mi l arl y, Eqs. (25b) and (25c) become
x, = (X,1 X,) . x,, (27b)
x, = (X,1X3) . x,. (27c)
I n thi s notati on Eq. (26) becomes
x, = (X,1X3) (X,[X2) (X,1 X,) xl. (28)
Thi s form shows cl earl y the successi ve operati ons carri ed out upon Xl
to pr oduce Xi . I t wi l l be noted, however , that the oper ator s are di s-
ti ngui shed from each other onl y by speci tl cati on of the parameters
i nvol ved; i t i s not possi bl e to change the argument of a gi ven functi on, as
i n the conventi onal functi onal notati on.
The over-al l effect of Eqs. (27) i s to defi ne X4 as a functi on of Xl :
x4 = (x,]Xl) oxl. (29)
On compari ng Eqs. (28) and (29) we obtai n the oper ator equati on
(X41X3) (X31X2) (X2[XI ) = (X4[XJ.
(30)
The form of thi s equati on cal l s our attenti on to a possi bl e mani pul ati on
of these functi onal operators. I n a meani ngful product of operati s,
each i nternal parameter wi l l occur twi ce i n nei ghbori ng posi ti ons i n
adjacent operators. One can, wi thout changi ng the si gni fi cance of the
oper ator , stri ke out such dupl i cated symbol s and condense the notati on
thus :
(X1]X,) s (X~l X2) (X,I X1) + (XA[X8) (XS[Xi ) ~ (X,[XI ), (31a)
or
(X,1 X,) . (X,1 X,) . (X,[xl) + (X41 X,) . (x,\X,). (31b)
Conversel y, one can descri be the structure of au oper ator i n mor e detai l ,
wi th consequent expansi on of the notati on:
(X,1 X,) + (X,1 X,) (X,1 X,) e (x41X3) (X,[X,) (X,1X,). (32)
The i nverse oper ator to (X,1 X,) wi l l be (X11X2). Thus
x, = (X,1 X,) ox,,
(33)
(X,1 X,) o (X,1 X,) = 1.
(34)
Both si des of an oper ator equati on can be mul ti pl i ed by the same
oper ator . Thi s must be done i n such a way that the resul ti ng oper ator s
have meani ng: the mul ti pl i ed operatom must have nei ghbori ng symbol s i n
SEC.3.4] GRAPHI CAL REPRESENTATI ON OF OPERATORS 51
common. Thus onecanmul ti pl y both si des of Eq. (30) from thel eftby
the oper ator (X21XJ, to obtai n
(X,lx,) .(x,lx, ).(x,lx, ).(x,[x,) = (X21 X,) (X,]XJ, (35)
whi ch may be condensed to
(xzlx,). (x,[x,) = (x21x4) (x41x,). (36)
Mul ti pl i cati on of Eq. (30) by (XZI XJ from the ri ght i s not defi ned,
butmul ti pl i cati on from theri ght by, for exampl e, (Xl l XJ i sdefi ned.
Thi s oper ator formal i sm can be appl i ed to vari abl es as wel l as to
parameters. Ani nput scal e, whi ch determi nes aparameter Xi asafunc-
ti on of a vari abl e ~ij can be r epr esented by an oper ator (Xi l zJ; an output
scal e woul d be r epr esented by an oper ator (Xklxk).
3.4. Graphi cal Representati on of Operators.The oper ator (X~l XJ,
fi ke the functi on +i(Xi), i s conveni entl y r epr esented by a pl ot of X~
agai nst Xi . Thi s representati on i s most uni form and most useful when
homogeneous parameters or vari abl es are used. A pl ot of Hk agai nst Hi
al ways l i es i n a uni t square (Fi g. 36); i t can be used i n the graphi cal
constmcti on of curves representi ng products of the oper ator (Hk lHi) wi th
other operators, and i n the sol uti on of other types of oper ator equati ons,
i n a way whi ch wi l l now be expl ai ned.
Gi ven the anal yti c form of the rel ati ons symbol i zed by
one can determi ne the form of the rel ati on
H. = (Hel Hi ) cHi (37C)
by el i mi nati ng the parameter Hk. I n the same way, one can determi ne
the graphi cal representati on of the product oper ator
(H,I H,) = (H,I H,) o (H~l H,)
(38)
by graphi cal el i mi nati on of the parameter Hk from pl ots of (H~l H~) and
(H~l H,). Fi gure 37 i l l ustrates the requi red constructi on. The opera-
tors (HdHi) and (H.l Hk) are r epr esented, i n the standard way, by pl ot-
ti ng the fi rst parameter verti cal l y agai nst the second hori zontal l y. I n t~e
representati on of (H~lHJ , H~ i s thus pl otted verti cal l y, but i n the r epr e-
sentati on of (H,I HJ i t i s pl otted hori zontal l y. The parameter Hi i s
pl otted hori zontal l y i n the fi rat case, and H, verti cal l y i n the second; i t i s
i n thi s way that they are to be pl otted i n the standard representati on of
the product oper ator (H,]Hi )) whi ch we must now construct. On the
mai n di agonal of the square, the l i ne (O, O) ~ (1, 1), we sel ect a poi nt A;
I
52 BASI C CONCEPTS AND TERMI NOLOGY
[SEC.3.4
thi s wi l l represent, by i ts equal hori zontal and verti cal coordi nates, a par-
ti cul ar val ue of the parameter H~. A hori zontal l i ne through A wi l l
i ntersect the cur ve (Hhl Hi ) at a poi nt B; the hori zontal coordi nate of
B i s a val ue of Hi correspondi ng to the chosen Hk. A verti cal l i ne through
A wi l l i ntersect the cur ve (H.~Hk) at a poi nt C; the verti cal coordi nate of
c i s the val ue of H. correspondi ng to the chosen Hk. The poi nt D, con-
structed by compl eti ng the rectangl e Al l DC, then has the hori zontal
coordi nate Hi and the verti cal coordi nate H, correspondi ng to the same
F1~. 3.6.Graphi cal representati on of
typi Cd oper ator (~kl~i).
a Fm. 3.7.Construction of a pr oduct of
operatora.
val ue of Hk; i t i s a poi nt on the cur ve of the product oper ator (Hs\ Hi). I t
wi l l be noted that the hori zontal l i ne through A i ntersects the cur ve
(Hk[Hi) at a second poi nt, B, to whi ch corresponds a second val ue of
Hi compati bl e wi th the same val ues of Hk and Ha. The poi nt D deter -
mi ned by constructi ng the rectangl e ABDtC i s thus a second poi nt on the
cur ve (H,[Hi). By carryi ng out thi s constructi on for a suffi ci ent number
of poi nts A, one can determi ne enough poi nts D, D, on the cur ve (H#l Hi )
to permi t i ts constructi on wi th any desi red accuracy.
The sl opes of the factor and product curves are si mpl y rel ated. The
anal yti c rel ati on
dHa dH, dH~
. -
dH, dHk dHi
(39)
becomes, i n the notati on of Fi g. 37,
[Sl ope of (Hml H,) at D] = [Sl ope of (H.I HI J at Cl
x (Sl ope of (H~[Hi) at B]. (40)
I f the factor curves i ntersect at a poi nt -4 on the mai n di agonal , the rectan-
gl e ABDC reduces to a si ngl e poi nt; the product cur ve passes through thi s
same poi nt, wi th a sl ope equal to the sl opes of the factor curves. h
i mportant speci al case i s that i n whi ch both factor functi ons are con-
, ;/-.
SZc.3.4] GRAPHI CAL REPRESENTATI ON OF OPERATORS 53
ti nuous and monotoni cal l y i ncreasi ng i n the range of defi ni ti on. The
factor curves then i ntersect at the poi nts (0,0) and (l , l ), atthe endsof
the mai n di agonal : the termi nal sl opes of the product cur ve are equal to
the products of the correspondi ng termi nal sl opes of the factor curves.
I t i s someti mes desi rabl e to construct the product (H.I HJ . (H~l HJ,
usi ng, i nstead of a pl ot of (H, I HJ, a pl ot of i ts i nverse, (H~\HJ.
The
requi red constructi on i s shown i n Fi g. 38. A hori zontal l i ne through a
poi nt A, correspondi ng to an arbi trari l y chosen val ue of Hk, wi l l i ntersect
the cur ve (Hkl H,) at a poi nt C wi th hori zontal coordi nate H., and the
cur ve (Hkl Hi ) at a poi nt B wi th hori zontal coordi nate Hi . A verti cal
l i ne through c wi l l i ntersect the mai n di agonal at a poi nt D wi th verti cal
coordi nate H,. Fi nal l y, by compl eti ng the rectangl e CDEB, one can
FI G. 3.S.Constructi on of the FI G. 3.9.Graphi cal sol uti on of
product (H.I HL) . (HhlH;), usi ng Z (H,lH/ J = (I f,lHk).
pl ot of (H~lI f,).
determi ne the poi nt E, wi th verti cal and hori zontal coordi nates H, and
H,, respecti vel y; thi s poi nt, then, l i es on the requi red cur ve (H.I HJ .
Thi s constructi on i s essenti al l y a sol uti on of the oper ator equati on
(H,I HJ . (H31HJ = (H, I H,),
(41)
the fi rst and thi rd of these oper ator s bei ng known. Otherwi se stated, i t i s
a graphi cal sol uti on of the oper ator equati on
(Hkl ~e) . Y = (Hkl Hi )
(42)
for the unknown oper ator Y, whi ch i s obvi ousl y the desi red (H.]H,). I t
wi l l be noted that the constructi on of Fi g. 3.8 i s that requi red for the
mul ti pl i cati on of (H~l HJ and (H, IHi ) to pr oduce (H~l Hi ), accordi ng to
the method fi rst expl ai ned.
Another oper ator equati on often encounter ed i s
Z . (H,I H,) = (H,l Hk). (43)
11
54 BASI C CONCEPTS AND TERMI NOLOGY [SEC.3.5
The constructi on for Z i s sketched i n Fi g. 39 i n the case of monotone
oper ator s (HilH/ c) and (H,I HJ.
3.6. The Square and Square-root Operators.I t i s someti mes desi r.
abl e to connect i n seri es two i denti cal l i nkages wi th equal i nput and OUG
put travel s. The fi rst l i nkage carri es out the transformati on
H~ = (H~l Hi ) , H;, (44a)
the second l i nkage, the transformati on
H, = (H,\H,) ~H~, (44b)
wher e the oper ator s (Hk\H;) and (H,l Hk) are i denti cal i n form, though
not, of course, i n the arguments. Then
(H,I H,) = (H.I H,) . (H,I H,) (45)
I
FI G. 3.10.Squari ng an oper ator .
J
FI G. 3.11.Squari ng an oper ator W
r epr esented by a cur ve whi ch crosses the
mai n di agonal .
i s essenti al l y the square of the oper ator
W ==(H,I HJ - (HklHJ ;
(46)
Eq. (45) may be wri tten as
(H,I H,) = W . W = W. (47)
The constructi on for the oper ator W i s i l l ustrated i n Fi gs. 3.10 and 3.11.
I n pri nci pl e, i t i s the same as the constructi on of Fi g. 3.7; cl i ffer ences i n
appearance ari se from the fact that, si nce the functi ons are i denti cal , the
poi nts B and C l i e on the same curve, i nstead of on two di fferent ones.
The cur ve representi ng W2 l i es beyond the W-curve, away from the
mai n di agonal . Where the W-curve crosses the mai n di agonal , the
W-curve al so crosses i t, wi th a sl ope equal to the square of the sl ope of the
W-curve; termi nal sl opes are rel ated i n the same way when the termi nal
poi nts are (O, O) or (1, 1). Thus the vari ati ons i n sl ope of the W2-curve,
and i ts curvature, are gr eater than those of the W-curve.
Sl tc.3.5] THE SQUARE AND SQUARE-ROOT OPERATORS 55
The ~cul ty i n desi gni ng a l i nkage to gener ate a gi ven functi on tends
to i ncrease wi th the curvature of the functi on.
I t i s often i mpossi bl e to
w a l i nkage of gi ven type to mechani ze a gi ven functi onal oper ator
(H,l Hi ) = W wi th l arge curvature, but qui te feasi bl e to mechani ze the
l ess strongl y cur ved square-root oper ator W. I f i t i s possi bl e to sol ve
Eq. (47) for the oper ator W, and to mechani ze thi s by a l i nkage wi th equal
i nput and output travel s, i t i s then possi bl e to mechani ze the gi ven f unc-
ti on by two such l i nkages i n seri es.
Thi s techni que wi l l be di scussed i n
Chap. 6; we shal l her e consi der onl y the
graphi cal method for sol vi ng for the
square-root oper ator W, when W2 i ncreases
monotoni cal l y.
The general nature of the probl em of
sol vi ng for W can be understood by i n-
specti on of Fi g. 310. One needs to fi l l
out the regi on between the mai n di agonal
and the W2-curve by a system of rectangl es
wi th hori zontal and verti cal si des, such
that one comer of each rectangl e l i es on the
mai n di agonal , the opposi te cor ner l i es on
the W2-curve, and the other two cor ner s
fal l on a conti nuous curve, the W-curve.
FI G. 3.12.C0nstructi 0n of a
square-root oper ator .
Thi s can al ways be done, and i n an i nfi ni te number of ways; the square-
r oot oper ator i s not uni que, but has the mul ti pl i ci ty of the curves that
can be drawn between two gi ven poi nts.
A square-root oper ator can be constructed i n the fol l owi ng way.
Between the mai n di agonal and the W2-curve, l et a poi nt C be chosen,
qui te arbi trari l y (Fi g. 3.12). Begi nni ng at the poi nt (7, construct the
hori zontal l i ne c@, the verti cal l i ne P7, the hori zontal l i ne Y13,and so on;
these form a step structure wi th ver texes al ternatel y on the mai n di agonal
and the W2-curve, extendi ng through the regi on between these l i nes. A
second step structure passi ng through c i s formed by the verti cal l i ne
~a, the hori zontal l i ne CI D,the verti cal l i ne p+, and so on. These two
step structures defi ne a seri es of rectangl es wi th opposi te ver texes on
the mai n di agonal and the W2-curve. The other ver texes defi ne a
sequence of poi nts, . . . , A, B, C, D, . . . , such that a W-curve whi ch
PMSWthough any poi nt of the sequence, say C, must pass al so through
i dl the others. Thi s sequence of poi nts wi l l have a poi nt of condensati on
wher e the W2-curve crosses the mai n di agonal , and cannot be extended
through such a poi nt. I n I ?i g. 3.12 the poi nts of condensati on are the
termi nal poi nts (O, O) and (1, 1); i n a case l i ke that of Fi g. 3 11, i nde-
pendent sequences must be defi ned i n regi ons separated by poi nts of
condensati on.
i
56 BASI C CONCEPTS AND TERMI NOLOGY [SEC.3.6
Let us choose to construct a square-root oper ator , W, whi ch passes
through the sequence of poi nts, . . . , A, B, C, D, . . . , i ndi cated by
sol i d ci rcl es i n Fi g. 312. We can al so requi re that i t pass through any
other si mi l arl y constructed seri es of poi nts, . . . , A, B, C, D, . . . ,
such as that i ndi cated i n Fi g. 3.12 by smal l ci rcl es. We can, i n fact, com-
pl etel y defi ne W by requi ri ng that i t pass between poi nts B and C i n an
arbi trari l y chosen conti nuous curve. Correspondi ng to the poi nts of thi s
curve, the above constructi on wi l l defi ne sequences of poi nts that con-
nect A to B, C and D, and so on; these poi nts defi ne a conti nuous W-curve
extendi ng from one condensati on poi nt to the next. The r eader wi l l
fi nd i t easy to pr ove that i f W i s to be si ngl e-val ued ever ywher e, i t must
i ncrease monotoni cal l y between B and C.
FI G. 3.13.C0nstructi on of a sauare,aot FI G. 3.14.Sauare-root-onerator cur ve
oper ator near a poi nt af condensati on. havi ng a deri vati ve at a poi ;t af canden~
sati on.
The square-root oper ator s thus defi ned do not, i n general , have
deri vati ves at the l i mi ti ng poi nts of condensati on. I n Fi g. 3.12 i t i s
zwi dent that the W-curve osci l l ates mor e and mor e rapi dl y as the ori gi n i s
approached, and i t i s hardl y to be expected that a deri vati ve wi l l exi st at
that poi nt. Fi gure 3.13 represents the part of Fi g. 312 ver y near the
ori gi n, i n a nei ghborhood i n whi ch the W2-curve can be repl aced by a
strai ght l i ne wi th fi ni te sl ope S # 1. The poi nts a, b, c, d, e, fal l i n the
same sequence as the poi nts A, B, C, D, of Fi g. 3.12. No attempt i s made
to r epr esent the forms of the i nterveni ng cur ve segments, whi ch are
repl aced by strai ght l i nes. The step structure shown dashed i s the con-
ti nuati on of the structure aDY& . . . of Fi g. 3.12; i t wi l l be unchanged i f
the poi nt (7 i s shi fted hori zontal l y, say to C*. The other step structure i s
the conti nuati on of aP-Y& . . . , and i t wi l l be changed by a hori zontal
shi ft of C. I t i s easy to show that the segments ab, cd, ef, . . . , are
paral l el , as are the segments bc, de, fg, . . . . The segments ab and cd
are i n general not paral l el to each other ; the average sl opes i n successi ve
sac. 3.5] THE SQUARE AND SQUARE-ROOT OPERA TORS 57
segments of the W-curve remai n constant and di fferent as the ori gi n i s
approached, and no deri vati ve exi sts at the ori gi n.
As we have al ready noted, a shi ft of the poi nt C of Fi g. 3.12 to the l eft
wdl modi fy one of the step structures, defi ni ng a new sequence of poi nts
a*, b*, c*, . . . , correspondi ng to the new poi nt C*. By pr oper choi ce
of C* the new sequence of poi nts can be brought to l i e on a strai ght l i ne
through the ori gi n, as shown i n Fi g. 314. Onl y through thi s parti cul ar
sequence of poi nts can one pass a W-curve havi ng a deri vati ve at the
ori gi n; the l i mi ti ng sl ope of that cur ve must be the sl ope of the l i ne
a*b*c* . . . , wh ch i s easi l y shown to be v%. Thi s geometr i c argu-
ment thus l eads to the al ready stated concl usi on that the sl ope of the
W-curve at a poi nt of condensati on must (i f i t exi sts) be equal to the
square r oot of the sl ope of the W2-curve.
The argument of the precedi ng paragraph al so l eads to the concl usi on
that on any gi ven hori zontal l i ne ther e i s one and onl y one poi nt (7 that
fi es on a W-curve wi th deri vati ve at the ori gi n. I t i s evi dent, then, that
the condi ti on that the W-curve shal l have a deri vati ve at the ori gi n (or
any other poi nt of condensati on wher e the W*-curve i ntersects the mai n
di agonal wi th a fi ni te di fference of sl ope) i s suffi ci ent to determi ne uni quel y
the form of the W-curve as far as the next adj scent poi nt of condensati on.
Si i ce an i ndependentl y determi nabl e secti on of the W-curve usual l y l i es
between two such poi nts of condensati on, the condi ti on that i t have a
deri vati ve ever ywher e pl aces on i t two condi ti ons, whi ch may or may not
be consi stent. Thus for any gi ven monotoni c W2-curve ther e can exi st,
at most, one W-curve wi th a deri vati ve ever ywher e; ther e may exi st none
at al l .
I f the W-curve i s to be mechani zed exactl y, i t i s obvi ousl y necessary
that i t have a deri vati ve ever ywher e.
For an approxi mate mechani zati on
i t i s onl y necessary that the W-curve osci l l ate wi th suffi ci entl y smal l
ampl i tude about a mechani zabl e cur ve wi th a deri vati ve ever ywher e. I n
ei ther case, the anal ysi s just outl i ned forms a practi cal basi s for the
determi nati on of W-curve. Tryi ng i n turn several poi nts (7, one can
qui ckl y fi nd a poi nt C* such that the sl opes of the segments a*D*, B*Y*,
. . . approach equal i ty as one of the two l i mi ti ng poi nts of condensati on
i s approached. The correspondi ng sl opes may then osci l l ate near the
other poi nt of condensati on, at whi ch thi s. W-curve wi l l have no deri vati ve.
I t i s, however , usual l y possi bl e to choose C* so that the osci l l ati ons of the
W-curve are negl i gi bl y smal l near both poi nts of condensati on. By i nter-
pol ati on one can then determi ne a smooth approxi mate W-curve sui tabl e
for mechani zati on.
CHAPTER 4
HARMONI C TRANSFORMER LI NKAGES
We turn now to the probl em of desi gni ng a bar l i nkage for the mecha-
ni zati on of a gi ven functi onal rel ati on between two vari abl es. The devi ces
used wi l l be di scussed i n the or der of thei r i ncreasi ng fl exi bi l i ty and the
i ncreasi ng compl exi ty of the desi gn pr ocedur e requi red: i n Chap. 4,
harmoni c transformers and doubl e harmoni c transformers; i n Chap. 5,
three-bar l i nkages; i n Chap. 6, three-bar l i nkages i n combi nati on wi th
harmoni c transformers or other three-bar l i nkages. Ful l exampl es of the
desi gn techni ques wi l l be provi ded by detai l ed di scussi ons of the probl em
of mechani zi ng the tangent and l ogari thmi c functi ons.
THE HARMONI C TRANSFORMER
4.1. Deti i ti on and Geometr y of the Hzrmoni c Transformer.-An
i deal harmoni c transformer i s a mechani cal cel l for whi ch i nput and output
parameters Xi and X, are rel ated by
X~ = R si n X,,
(1)
R bei ng an arbi trary constant. Such a rel ati onshi p can be obtai ned wi th
si mpl e mechani sms model i ng a ri ght tri angl e, such as are sketched i n Fi g.
t+
x
FI Q. 4.1.I deal harmoni c transformers.
4.1. These harmoni c transformers are cal l ed i deal because they
gener ate the si ne or arc-si ne functi ons accuratel y; unfortunatel y, they are
somewhat unsati sfactory mechani cal l y, and are ther efor e used onl y
excepti onal l y i n practi cal wor k. I t i s usual l y preferabl e to empl oy
noni deal harmoni c transformers, such as those shown i n Fi gs. 4.2 and 4.3,
58
sm. 4.1] DEFI NI TI ON OF THE HARMONI C TRANSFORMER
59
whi ch gi ve or dy an approxi matel y si nusoi dal rel ati on between i nput and
output parameters.
The mechani sm shown i n Fi g. 42 i s an ordi nary crank-l i nk system wi th
~ymmetri cal l y pl aced sl i de. The devi ati on of the output parameter
from i ts i deal val ue depends upon the angl e c between the l i nk L and
the l i ne of the sl i de. Representi ng the output parameter by X;, one has
x: = R si n X, L(I cos ,).
(2)
Thi s may be wri tten as
Xi = X~ + l $xk, (3)
f
w
y
.-.
w
.. -
S
[
r. I
i
.....
I
I
L Cusf
I
1
~
Fm. 4.2.Crank-l i nk system as a noni deal harmoni c transformer.
wher e bxk i s the structural er r or of the mechani sm as compared wi th the
i deal harmoni c transformer:
LTxk = L(l Cos c). (4)
I n the mechani sm of Fi g. 4.2, c i s vari abl e, bei ng gi ven by
Lsi n~=Rcos X~ W. (5)
I rI Fi g. 42 the sl i de di spl acement W has been so chosen as to keep q
and hence bxk, smal l as the crank turns through i ts l i mi ted operati ng
angl e. As wi l l be di scussed,i n detai l l ater, i t may be desi rabl e to make a
di fYerentchoi ce of W i n or der to obtai n a desi red nonvani shi ng form for
8XL.
60 HARMONI C TRANSFORMER LI NKAGES [SEC.4-1
Fi gure 43 represents a harmoni c transformer connected to another
l i nkage such that the pi vot P may be found anywhere wi thi n the shaded
area. Equati ons (2), (3), and (4) hol d i n thi s case, but c and the struc-
tural -error functi on ~xk now depend not onl y on xi, but al so on the
posi ti on of the pi vot P wi thi n the possi bl e boundary.
An i deal harmoni c transformer generates a secti on of si ne or arc-si ne
curve, the form of whi ch can be fi xed by speci fi cati on of the angul ar l i mi ts
of the rotati on of the crank, X~mand Xi ~. The noni deal harmoni c trans-
for mer requi res four parameters for i ts speci fi cati on-f or i nstance, Xi m,
Xi ~, L/R, and W/R. The pr esence of these addi ti onal parameters per -
\
,
\
=-wh.- --
1
FI G. 4.3.A noni deal transformer wi thout fi xed el i de.
I
mi ts a consi derabl e extensi on of the fi el d of mechani zabl e functi ons-an
extensi on whi ch becomes stri ki ng i f c i s permi tted to assume l arge val ues.
I n most practi cal wor k e and 6X~ are kept fai rl y smal l ; ~xkthen appears
ei ther as an er r or ari si ng from the use of a noni deal desi gn, or as a smal l
cor r ecti on to the si nusoi dal form, by whi ch one makes the mechani zed
functi on cor r espond mor e cl osel y to a gi ven, not exactl y si nusoi dal ,
functi on.
I n worki ng out the mathemati cal desi gn of a system that i ncl udes a
noni deal harmoni c transformer, i t i s usual l y desi rabl e to car r y through the
fi rst cal cul ati on as though the transformer wer e i deal . The er r or ari si ng
from use of the noni deal desi gn can then be cor r ected i n the fi nal stages
of the wor k (i f thi s i s requi red by ver y ri gi d tol erances), or so chosen as to
mi ni mi ze the over-al l er r or of the system.
SEC.4.2] USE OF THE HARMONI C TRANSFORMER
61
4.2. Mechani zati on of a Functi on by a Harmoni c Transformer.
I n the harmoni c transformer one parameter i s a rotati on, the other a
transl ati on. Ei ther of these may be taken as the i nput parameter. I f
the crank R i s the i nput termi nal , the l i mi ts of the i nput parameter Xi may
be chosen at wi l l ; the crank can descri be any angl e or make any number
of revol uti ons. The mechani zed functi on wi l l al ways be a si nusoi d or a
part of a si nusoi d between chosen l i mi ts ,
u
(Fi g. 4.4). I f the sl i de i s the i nputter- i
mi nal , the range of the i nput parameter 1
xk must be l i mi ted to keep the mecha-
1[
L_
Maxi mum
AXk
sl ope
J
AXi
Xw
x~
t /
J-xim
/ ~ x~
xkm
Fm. 4.4.Si nusoi d gener ated by an i deal FI m 4.5.Arcsi nusoi d gener ated by an
harmoni c transformer. i deal harmoni c transformer.
ni sm far enough from the sel f-l ocki ng posi ti ons.
The mechani zed
functi on i s then a porti on of an arcsi nusoi d (Fi g. 4.5) wi thi n whi ch the
sl ope does not exceed some maxi mum val ue determi ned by mechani cal
consi derati ons.
The si mpl est probl em i n i deal -harmoni c-transformer desi gn i s that of
mechani zi ng a harmoni c rel ati on, anal yti cal l y expressed, between vari a-
bl es xi and Xk:
X.k Sk, = r si n (~i ~i o), r>o, (6a)
or
( )
Zk Xk,
Xi $i , = si n 1
T
(6b)
gi ven speci i i ed l i mi ts for the i nput vari abl es. To determi ne the constant
R of the harmoni c transformer and the requi red rel ati on of the vari abl es
zi , ~k,to the parameters Xi , X~, one need onl y compare Eqs. (1) and (6):
Zk Xjco Xk
.,
R
Zi Xi , = Xi .
r
(7)
The wdue of R, chosen at wi l l , determi nes the scal e factor Kk of the param-
eter xk:
K~ =
zk xk, _Zl c-Z~O_r
Xhxh xk ~
(8)
ii
62 HARMONI C TRANSFORMER LI NKAGES [SEC.4.3
[X,, = O, by Eq. (7)]. The scal e factor for Xi i s uni ty. These constants
bei ng fi xed, the harmoni c transformer i s determi ned. The range of
parameter val ues for whi ch i t must oper ate i s determi ned by the l i mi ted
range of the i nput and oUtpUt vari abl es, xi ~ ~ Xi S zi ~, 2% ~ Xk~ XkM:
Xtm= Z,m z,,, xiM = $iM XioJ
(9)
Xkm Xko
Xk. = ~k ,
zkM Zko
Xkkf = ~k . (l o)
A l ess tri vi al probl em i s that of mechani zi ng a functi on that has a
general l y si nusoi dal character, but i s gi ven onl y i n tabul ated form. One
possi bl e method i n such a case i s to fi t the gi ven functi on as wel l as possi -
bl e (for exampl e, usi ng the method of l east squares) by the anal yti c
expressi ons of Eq. (6), and then to pr oceed as just expl ai ned. A qui cker
way, maki ng use of homogeneous vari abl es and parameters, wi l l now be
presented.
4.3. The I desl Hsrmoni c Transformer i n Homogeneous Parameters.
Befor e expressi ng the equati on of an i deal harmoni c transformer i n homo-
geneous parameters, we must defi ne the parameters mor e preci sel y.
The posi ti on of the crank R (Fi g. 4.2) i s descri bed by the parameter
Xi , the rotati on of the crank cl ockwi se from a zer o posi ti on perpendi cul ar
to the center l i ne C of the sl i de.
The other parameter, xk, i s defi ned as
the normal pr ojecti on of the arm R onto the center l i ne of the sl i de. The
crank R i n the zer o posi ti on i s pi ctured as di r ected upwards, and xk i s
taken as posi ti ve toward the ri ght from the poi nt S.
The homogeneous parameters t%,Hk, are rel ated to the parameters
X,, xk, by
~, = Xi Xi m ~k = xk xkm.
,
AX, Axk
(11)
(The symbol Oi i s chosen to r epr esent one homogeneous parameter,
i nstead of Hi, to emphasi ze the fact that i n thi s case one i s concer ned
wi th a rotati on. ) From these defi ni ti ons i t fol l ows that both homm
geneous parameters i ncrease i n the same sense as the ori gi nal parameters:
Oi i ncreases al ways cl ockwi se, Hk i ncreases to the ri ght.
The connecti on between ordi nary and homogeneous parameters i n a
harmonic transformer i s i l l ustrated i n Fi g. 4.6. The arc of the angl e of
travel AXi , scal ed evenl y cl ockwi se from Oto 1, permi ts di rect readi ng of
Oi . The pr ojecti on of that arc on a strai ght l i ne perpendi cul ar to the
zer o l i ne SO, scal ed evenl y from Oto 1, from l eft to ri ght, permi ts di rect
readi ng of H~. Any l i ne paral l el to OS passes through correspondi ng
val ues of & and H~. The correl ati on of val ues of Hk to those of Oi i s
uni que so l ong as AXi < 360; the conver se correl ati on may be
val ued i n some cases, as i s i l l ustrated by Fi gs. 4.6(b) and 4.6(c).
doubl e-
,
SEC.4.4] TABLES OF HARMONI C-TRANSFORMER FUNCTI ONS
63
From the defi ni ti on of homogeneous parameters and from Eq. (11) i t
i s evi dent that, al ways,
H, =
si n (Xi ~ + O~AX,) (si n X,)~ti
(si n X,)~.= (si n Xi)min
(12)
o
o
,i
x~
xhM
(b)
o
(c)
(a)
FUJ.4.6.I deal harmoni c transformer wi th homogeneous parameters.
(a) (si n Xi )= = si n Xu,r. @) (si n Xi ) ~= = 1. (c) (si n X,)~ti = si n X,.af,
Speci al forms of thi s rel ati on, appl i cabl e i n cases of the types i l l ustrated
i n Fi gs. 4.6(a), 4.6(b), 4.6(c), respecti vel y, are as fol l ows:
H~=
si n (X{~ + Oi AXi ) si n Xi ~
si n Xt~ si n Xi ~
(13a)
H, =
si n (Xi ~ + Oi AXi ) si n Xi ~
1 si n Xi ~
(13b)
Hk =
si n (X~~ + O~XJ si n X;~
1 si n XW
(13C)
444. Tabl es of Harmoni c-transformer Functi ons.-The use of har-
moni c transformers as parts of compl ex l i nkages i s so extensi ve and the
desi gn probl em i s so greatl y si mpl i fi ed by the use of homogeneous param-
eter s that i t i s ver y conveni ent to have avai l abl e a fai rl y compl ete tabl e
of the functi ons appeari ng i n Eq. (13). Tabl e Ao1 gi ves H~ for @i = 0.0,
0.1, 0.2, . 0.9, 1.0, and for AX~ = 40, 50, . . . 140. Smal l er
val ues of AX~ are of Li ttl e i nterest, si nce wi th smal l angul ar travel the
er r or s due to mechani cal pl ay become rel ati vel y i mportant, and other
devi ces can ser ve as wel l for mechani zati on of the correspondi ng nearl y
l i near functi ons H~(tl ~). Two faci ng pages are requi red for each val ue of

64
HARMONI C TRANSFORMER LI NKAGEL9 [SEC. 4.4
AX<. Col umns of val ues of Hk are gr ouped i n pai rs i n a way i ntended
to faci l i tate the cal cul ati on of structural er r or functi ons, as di scussed i n
Sec. 45. The fi rst of these col umns has the correspondi ng val ues of
Xi ~ and XW i ndi cated at the top, and i s tabul ated wi th& (i ndi cated to the
l eft) i ncreasi ng downward.
The second col umn has the val ues of X,~ and
Xi ~ i ndi cated at the bottom, and i s tabul ated wi th tl (i ndi cated at the
ri ght) i ncreasi ng upward. The associ ated col umns cor r espond to har-
moni c transformers wi th Xi ~ S Xi S Xt~ and wi th
(90 Xi M) S Xi S (90 Xi m),
respecti vel y; the si gni fi cance of thi s and of other features of the tabl e
whi ch are not of i mportance at thi s poi nt wi l l be expl ai ned i n Sec. 45.
Tabl e A.2 gi ves Oi for Hk = 0.0, 0.1, 0.2, . 0.9, 1.0, and for the
same AX, as Tabl e A. 1. The arrangement i s si mpl e and shoul d requi re
no expl anati on her e. Onl y si ngl e-val ued rel ati onshi ps between Hk and
Oi are tabul ated, si nce the tabl e i s i ntended for use when Hk i s the i nput
vari abl e; al l regi ons that i ncl ude a poi nt wi th i nfi ni te d6i /dHk may be
excl uded.
I n usi ng Tabl es Al and A*2 to mechani ze a tabul ated functi on wi th a
pronounced si nusoi dal character, the functi on shoul d fi rst be expressed i n
homogeneous vari abl es. We shal l caU the homogeneous i nput vari abl e
h,, the homogeneous output vari abl e h., i n or der to avoi d any commi tment
as to whi ch i s to be the angul ar parameter i n the mechani zati on.
Next, ther e shoul d be tabul ated i n a col umn the val ues of the output
vari abl e h., for h, = 0.0, 0.1, 0.2, s
1.0, maki ng such i nterpol ati ons as
may be necessary.
I t remai ns onl y to compare thi s col umn of numbers wi th those i m
Tabl es A. 1 and A% One can easily fi nd whi ch of these col umns gi ves
the best fi t to the gi ven set of numbers; each col umn, i t i s i mportant to
note, may be read ei ther up or down.
Thi s determi nes the best val ues of
x,m and xi ~ for the harmoni c transformer, to wi thi n 10; by i nterpol ati on
one may fi x these val ues even mor e preci sel y. The remai nder of the
desi gn process i s then tri vi al .
I f the best fi t i s found i n Tabl e A 1, the output vari abl e h, i s bei ng
i denti fi ed wi th H~; the output termi nal of the mechani zati on wi l l be the
sl i de, the i nput termi nal the crank.
I f the best fi t i s found i n Tabl e A.2,
the r ever se i s true.
Suppose that the best fi t i s found i n Tabl e A-1, and that, i n readi ng the
correspondi ng col umns, hr and Oi i ncrease together . Then one has
h, = 0,, h. = Hh. Knowi ng xi na and xi ~, one can construct scal es of Oi
and Hk as descri bed i n the precedi ng secti on; these are the requi red scal ee
of h, and h,, whi ch one can recal i brate i n terms of the ori gi nal vari abl es, i f
thi s shoul d be desi red.
SW 44] TABLES OF HARMONI C-TRANSFORMER FUNCTI ONS 65
I f the best fi t i s found i n Tabl e A 1, but cor r espondence of the col umns
requi res that they be read i n such di recti ons that h, decreases as i %
i ncreases, then 1 h, = &, hs = Hk. The h,-scal e thus di ffers from the
O,-scal eonl y i n that h, i ncreases to the l eft i nstead of the ri ght; the rest
of the constructi on i s as before.
I f the best fi t i s found i n Tabl e A2, one hash. = tt, h, = H,, i f h, and
8i i ncrease together , and otherwi se 1 h, = &, h, = Hk.
I n the operati onal l anguage i ntroduced i n Chap. 3 thi s process maybe
descri bed as fol l ows: A functi onal oper ator (h.l b,) i s gi ven, and ther e i s
sought a functi onal oper ator (H~10i )or (4 IHJ of a harmoni c transformer
whi ch transforms i nto the gi ven oper ator (h.\ h,) when the pai r of vari abl es
(h,, h,) i s transformed i nto the pai r of parameters (o,, H,) or (H,t, 0,)
through a di rect or compl ementary i denti fi cati on.
When the tabl es are empl oyed i t i s useful to make graphs of oper ator s
and sketches of mechani sms i n or der to pr event mi stakes. I t i s r ecom-
mended that the H~-scal e run al ways from l eft to ri ght, that the zer o l i ne
for Xi be dh-ected upward, and that the scal e for 0, i ncrease cl ockwi se, as
i n Fi g. 46.
Exampl e: Use an i deal harmoni c transformer t o mechani ze the rel ati on
X2 = tan xl (14)
wi th the range of the i nput vari abl e $1 f rom 0 to 50. The homogeneous
vari abl es are
=S
h, = +.
tan 50
Tabl e 41 gi ves the rel ati on of h. to h, i n tabul ar form.
Tmm 4.1.z* = TANz,, O S z, S 50, I N HOMOGENEOUS
h, h,
0.0 0.0000
0.1 0.0734
0.2 0,1480
0.3 0,2248
0.4 0.3054
0.5 0.3913
0.6 0.4844
0,7 0.5875
0.8 0.7041
0.9 0,8391
1.0 1,0000
(15)
VARL4BLRS
I n seeki ng a correspondi ng col umn i n the tabl es, we need exami ne onl y
those whi ch show no maxi mum.
I n such cases the fi rst and l ast val ues
are al ways O and 1; ever y such col umn matches the gi ven col umn at the
two ends.
k,
56 HARMONI C TRANSFORMER LI NKAGES [SEC. 4.4
Consi der fi rst Tabl e A 1. Fi xi ng on a val ue of AX,, we seek a col umn
that gi ves a match at the mi ddl e as wel l as at the ends; for exampl e, wi th
AX~ = 70 the best match i s obtai ned for Xi ~ = 70, Xi ~ = OO.
However , thi s col umn contai ns val ues that are too smal l at smal l Oa,too
l arge at l arge 0,. Repeati ng thi s process for smal l er AXi , one obtai ns a
better over-al l fi t, but the i mprovement i s sl i ght; one must ei ther use
ver y smal l val ues of AX~ or tol erate er r or s of over 2 per cent of the total
range.
Next we exami ne Tabl e A-2. Agai n the best match i s obtai ned for
rel ati vel y smal l AX,a consequence of the nearl y l i near character of the
tangent functi on i n the gi ven range. Here, however , a much better
match i s possi bl e. Compari ng wi th the gi ven h, the val ues of O,shown i n
Tabl e A.2 for Xi n = 30, Xi x = 70 and for Xi n = 35, Xi ~ = 75,
one fi nds the di fferences shown i n Tabl e 42.
TABLE 4.2.VALUES OF h.-tl
H,
0.0
0,1
0.2
0.2
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Xi. = 30
X,X = 70
0.0000
0.0004
0.0023
-0.0050
0.0076
0.0097
0.0106
0.0095
0.oow
0.0009
O.0000
X;m = 35
X,M = 75
0.0000
0.0036
0.0056
0.0065
0.0072
0.00s0
0.0095
0.0120
0.0152
0.0160
0.0000
Xi m = 31.5
X%kf= 71.5
0.0000
0.0008
0.0001
-0.0015
-0.0032
-0.0044
0.0046
0.0030
0.0003
0.0042
0.ooOo
Li near i nterpol ati on between these col umns shows that wi th Xi - = 31.5.
Xi ~ = 71.5: the di fference between h, and Oi remai ns l ess than 0.005; ari
i deal harmoni c transformer wi th these constants woul d have a structural
er r or ever ywher e l ess than 0.5 per cent of the travel .
Fi gure 47 shows the harmoni c transformer thus desi gned, wi th func-
ti onal scal es for ha = Hk and h = Oi . The travel , Axk, can be gi ven any
desi red val ue by pr oper choi ce of R:
AX, = R(si n 71.5 si n 31.5) = 0.4258R. (16)
I t i s i nteresti ng to note that i n thi s exampl e the angul ar vari abl e Z1 of
Eq. (14) has been mechani zed as a sl i de di spl acement, the l i near vari abl e
ZZ as an angul ar di spl acement, whereas i n a constructi ve computer the
r ever se woul d be the case.
SEC. 4.5] TOTAL STRUCTURAL ERROR 67
I n thi s desi gn pr ocedur e we have tr eated the harmoni c transformer as
i deal . To construct i t as noni deal woul d i ntroduce an addi ti onal struc-
tural er r or , 6X~, descri bed by Eq. (4)an er r or that can be made suf-
fi ci entl y smal l by maki ng the l i nk L ver y l ong and by so pl aci ng the center
l i ne of the sl i de as to r educe the maxi mum val ue of the angl e e as much as
possi bl e. I n general i t i s better to make posi ti ve use of the term I sX~,so
choosi ng the desi gn constants that ~xktends to cancel out the structural
er r or ~hkof the i deal -harmoni c-transformer component of the mechani sm.
I n the present case thi s may seem hardl y wor th the troubl e, as the fi t
obtai ned wi th the i deal
l
transformer i s ver y good. However , i t i s to be
0.0 0.5 1.0
Hk
+X*
\
/
\ .o
T
fi a. 4.7.Harmoni c transformer mechani zi ng X, = tan X,, O < Xl <50. A htter
desi gn i u shown i n Fi g. 4.12.
noted that thi s desi gn i s unsati sfactory i n that the angul ar travel AXi i s
rather smal l . I n practi ce, i t woul d be better to empl oy a noni deal trans-
for mer wi th l arge angul ar travel , keepi ng the total structural er r or smal l
by judi ci ous choi ce of L and the sl i de posi ti on (Fi g. 4-12). The requi red
desi gn techni que i s di scussed i n the next secti ons.
4.6. Total Structural Er r or of a Norddeal Harmoni c Tranaformer.
I n fi ndi ng a harmoni c transformer to mechani ze a gi ven rel ati on,
hk = (h~l hi ) h{, (17)
one begi ns, as al ready descri bed, by fi ndi ng an i deal harmoni c transformer
tbt gi ves an approxi mate fi t. Then i f Oii s i denti fi ed wi th hi, Hk can al so
be i denti fi ed wi th I w, except for the smal l structural er r or ti hk:
Hk = h~ + ti h~. (18)
68 HARMONI C TRANSFORMER LI NKAGES [SEC. 4.6
I f the transformer to be used i s noni deal , i ts output parameter wi l l be not
H~ but Hi . Representi ng by 6H~ the change i n output ari si ng from the
noni deal character of the transformer, we wri te
Hi = Hk + 6Hk. (19)
T,~.ecompl ete mechani sm then has a structural -error functi on
M; = Hi h~ = 6Hh + ~hk; (20)
i t i s thi s er r or that shoul d be r educed to tol erabl e l i mi ts over the whol e
range of operati on.
A noni deal harmoni c transformer has been sketched i n Fi g. 4.2. Of
the four desi gn constants, Xim and Xi ~ characteri ze the i deal -harmoni c-
transformer component and determi ne the form of ti hk;L/ R and W/R
affect onl y the form of 8Hk. I t i s of cour se i mpossi bl e i n general to make
dhj vani sh i denti cal l y by any choi ce of these parameters. I deal l y,
one woul d mani pul ate al l four parameters i n or der to make bh~ever y-
wher e sati sfactori l y smal l , wi thout r egar d to the resul ti ng magni tude of
8Hk and M,. An easi er techni que i s to make bhk as smal l as possi bl e by
choi ce of Xi ~ and XW, and then to choose L/ R and W/R so as to mi ni mi ze
bhj; however , one can often arri ve at mor e sati sfactory desi gns, and even
appreci abl y r educe the over-al l er r or , by some other choi ce of X~~ and
X,~.
4.6. Cal cul ati on of the Structural -error Functi on ti Hk of a Noni deal
Harmoni c Transformer.-I n desi gni ng harmoni c transformers i t i s i mpor-
tant to have a qui ck, effi ci ent way to compute the structural -error func-
ti on 8Hk. Use of Eq. (4) i s nei ther qui ck nor wel l adapted for wor k wi th
homogeneous parameters; better methods to be descri bed her e and i n
Sec. 4.7 depend upon r efer ence to Tabl e Ao1. The di scussi on wi l l be
i l l ustrated by Fi g. 48, whi ch shows a harmoni c transformer wi th al terna-
ti ve posi ti ons for the l i nk L, extendi ng from the crank toward the l eft or
toward the ri ght. Here, and throughout the di scussi on that fol l ows,
the uni t of l ength, i n whi ch al l di mensi ons are stated, i s taken to be the
l ength of the Hk-SCtde; thus, flk = 1. As before, we consi der the
harmoni c transformer i n i ts basi c posi ti on, wi th Oi i ncreasi ng cl ockwi se,
the zer o for Xi verti cal l y upward, and scal es H~ i ncreasi ng from l eft to
ri ght.
The change from the i deal harmoni c transformer (scal e H~) to the
noni deal one (scal e Hj) wi l l be traced through two steps.
Fi rst, the Hk-sCal e maybe shi fted bodi l y to the ri ght or l eft by a di s-
tance L. On thi s scal e, shown i n Fi g. 4.8 above the sl i de, the readi ng
opposi te the poi nter wi l l be Hk, modi fi ed by an er r or
DH~ = +- I LI (l COSc). (21)
fihw.4.6] CALCULATI ON OF THE STRUCTURAL ERROR
69
The si gn of thi s er r or depends onl y on whether the crank extends to the
fi ght or to the l eft. Taki ng L as posi ti ve when the l i nk extends toward
the l eft, negati ve when i t extends toward the ri ght, one has al ways
DH~ = L(I COSe). (22)
As H~changes from zer o to one, Hk + DHk changes between l i mi ts whi ch
are i n general not zer o and one:
(Hk + DHk)ti 5 H. + DHk ~ (Hk + DH&mx; (23)
thus H~ + DHk i s not i n general a homogeneous P~rameter.
As the second step, the H~ + DHk-scal e i s repl aced by the homo-
geni zed Hi scal e, shown i n Fi g. 48 bel ow the sl i de:
H, + DHk (Hk + DHk)mi n
; =(Hk + DH,)max (Hk + DH&im
P
L L
_ AX,
Hkfi!m.. ~_ ~=,
(24)
I
i
0.0 0.5
--- ..
. ---
\
I
K
-- .
.
.,jr,,,~ w . ~ ~..,,::,,,,, ~
H; 0.0 0.5 1.0 0.0 1.0H;
FI G. 4,S.Notati on used i n harmoni c transformer
When DH, i s reasonabl y smal l the maxi mum val ue
occur at essenti al l y the same @i as the maxi mum of
H~ = 1. One can then wri te
(Hk + DH,)ma, = 1 + (DHk) ,,
and si mi l arl y
(H, + DHk)min = (DH,)o,
desi gn.
Of H, G DHk Wfl ~
H,that i s, when
(25)
(26)
(DHb)O and (DHk)l bei ng the val ues of DHk for H, equal to O and 1
rmpecti vel y. As an approxi mati on good enough for al l prel i mi nary
cal cul ati ons one has then
Hk + DHk (DH,), .
Hi =
1 + (DH,) 1 (DHk)o
(27)
To compute DHk, we obser ve that i f Yki s the di stance above the sl i de
center l i ne of the pi vot between L and R, then
Yk
si n~=
L
(28)
70 HARMONI C TRANSFORMER LI NKAGES [SEC.4.6
and
H = L[ - ( - W1
( 29
When c i s smal l
DHk z g.
(30)
The quanti ty yk i s conveni entl y found as a functi on of 0; by use of Tabl e
A-l , by taki ng advantage of the speci al rel ati onshi p of associ ated col umns
of that tabl e. The rel ati onshi p of the correspondi ng harmoni c trans-
0.0 i 0.5 1.0
O* Ofi o &
Hh
, , )
J
- E
/
H; / /
x~
x;
/
0.0,
1.0
0.0
/
e:=
/
ei -o
90-xiM
/ ;/
0.5 AX;
0.5
xf,/
H: 0,0
90-x,m
o*
/
(a)
/
/
(b) - Lo
/
/
/
/
FI G. 4.9.Harmoni c transformers associ ated i n Tabl e A.1.
formers i s i l l ustrated i n Fi g. 4.9. The fi rst transformer (parameters
Xi , X~; Oi ,H/$) operates through the range
Xi m S Xi S Xi .l f
(X,~ = 15, X~M = 75 i n Fi g. 4.9a). The second transformer (param-
eter s X~, X:; o:, H:) operates through the range
X,2 = 90 Xi ~ ~ X,* S X,; = 90 X~~
(XL = 15, X3 = 105 i n Fi g. 4.9 b). Now i t wi l l be obser ved that i f
Fi g. 4.9b i s r efl ected i n the l i ne X: = 45 and superi mposed on Fi g. 4.9a,
the angul ar scal es wi l l then coi nci de, but wi th 19,*= 1 & The
H~-scal e, however , becomes a verti cal scal e, as compared wi th the hori -
zontal H~-scal e. The entri es i n a secti on of Tabl e A. 1 may then be
i nterpreted as fol l ows: For a harmoni c transformer wi th l i mi ts on Xi as
gi ven at the top, the four entri es i n each r ow, from l eft to ri ght, cor r espond
to (1) di for some i ndex poi nt P on the angul ar scal e, (2) Hk, whi ch meas-
ures the hori zontal di spl acement of that poi nt to the ri ght of a verti cal
sm. 4.7] STUDY OF THE STRUCTURAL-ERROR FUNCTI ON 71
r efer ence l i ne, (3) H?, whi ch measurea the di spl acement of the same poi nt
upward from a hori zontal r efer ence l i ne (though i n di fferent uni ts, si nce
the H:-scal e i s not i n general of uni t l ength), and (4) O: = 1 19i . I f
the l i mi ts on Xi are those gi ven at the bottom of the secti on, the entri es i n
each r ow have the same meani ng i f they are taken i n or der from nght to
kjt.
The quanti ty Hk( = xk) measures the actual di stauce of the poi nt P
from the verti cal r efer ence l i ne, si nce the Hk-SCtdei s one uni t l ong by
defi ni ti on. The l ength of the verti cal scal e i s
(COSXi)- (COSX,)*;
g = (si n Xi )- (si n X,)ti
(31)
hence the actual di stance of the poi nt P from the hori zontal r efer ence l i ne
i s
X: = gH~. (32)
Val ues of g are gi ven i n Tabl e A. 1, i n the same l i ne wi th those of Xi and
Xi M and i n the same col umn wi th the val ues of H:.
Returni ng now to the computati on of Yk,we defi ne E? as the val ue on
the H:-scal e at the poi ut wher e the sl i de center l i ne i ntersects i t. I n Fi g.
48, E: l i es on the cal i brated part of the scal e. Thi s i s not necessari l y so;
~ i s a parameter i n the desi gn whi ch may be assi gned negati ve val ues, or
val ues gr eater than one. I n any case
y, = g(Hf E:). (33)
I t i s conveni ent to speci fy a noni deal harmoni c transformer by gi vi ng
X{-, X,~, E:, and L. Cal cul ati on of i ts structural -error functi on for a
seri es of val ues of Oi or Hfi then requi res readi ng from Tabl e A 1 the cor -
respondi ng val ues of H~, fol l owed by computati on of Yj by Eq. (33),
DHk by Eq. (29) or (30) (accordi ng to the accuracy requi red), Hi by I Zq.
(27), and fi nal l y 6H~ by Eq. (19). An i l l ustrati ve cal cul ati on wi l l be
found i n Tabl e 45. Thi s pr ocedur e i s qui ck and easy i f Et and L are
known, but when i t i s desi red to determi ne the approxi mate form of 8Hk
for a consi derabl e seri es of val ues of El and L, or to fi nd requi red val ues of
Et and L, the method to be descri bed i n the next secti on i s to be pr efer r ed.
4.7. A Study of the Structural -error Functi on MI ~.For a general
i nvesti gati on of the structurs&error functi on 6Hfi or for a prel i mi nary (and
usual l y fi nal ) choi ce of ~~ and L i n the process of desi gni ng a noni deal
harmoni c transformer, i t i s suffi ci entl y accurate to use Eq. (30) i n com-
puti ng DHk, and to assume that
I DH, (DHJ oI <<1.
(34)
To thi s approxi mati on 6Hk has a si mpl e dependence on I I ? and L whi ch
faci l i tates i ts computati on for a seri es of val ues of these parameters, or ,
72 HARMONI C TRANSFORMER LI NKAGES
[SEC. 4.7
conversel y, the fi ndi ng of val ues of E; and L whi ch gi ve 6Hk a desi red
form and magni tude.
Expandi ng Hj, as gi ven by Eq. (27), i n power s of the smal l quanti ty
(DHk) 1 (DH~)o, and negl ecti ng terms of the second or der of smal l ness,
one fi nds
Hi = H, + [DHk (DHJ o] + Hk[(DH,)o (DHk) ,]
(35)
and
JH, = [DHk (DHk),] + H,[(DH& (DHk) d.
(36)
Thi s approxi mati on to 6Hk, l i ke the functi on i tsel f, vani shes when Hk = O
or 1.
By Eqs. (30) and (33),
2 H* Ey)z,
DH, =
~L( k
(37)
When thi s i s i ntroduced i nto Eq. (36) the quadrati c terms i n E; cancel ,
and one fi nds
8H~ =
~ Ul (di ) + .i Y~(oi )l .
(38)
wher e jl ((?J and jj (19Jdepend onl y on the parameters Xi ~ and Xi ~ of the
harmoni c transformer.
Wi th (H~), and (H:), the val ues of H: when Hk
has the val ues O and 1, respecti vel y,
Knowi ng the form of f,(tl ) and j, (@i ), one can easi l y compute 6Hk for a
l arge seri es of val ues of L and E;.
To thi s approxi mati on the magni tude of the structural -error functi on
vari es i nversel y wi th L, but i ts form i s determi ned enti rel y by E:. The
possi bl e range i n forms i s easi l y i nvesti gated by computi ng 8Hk for some
val ue of E&for exampl e, for ET = O, i n whi ch case one has si mpl y the
fi rst term of Eq. (38)and then addi ng to thi s the functi on jZ(OJ i n di f-
fer ent proporti ons.
Al thougn f2(@i ) i s easi l y computed by Eq. (40), i t i s wor th whi l e to
take note of i ts si mpl e anal yti c form. AS functi ons of Xi , one has
si n Xi (si n Xi )fi
k = (si n XJm= (si n X,)ti
(41)
H: =
Cos xi (Cos Xi)mti
(Cos X,)w (Cos Xi)ti
(42)
Let X. and Xh be the val ues of Xi for whi ch si n Xi has i ts mi ni mum and
i ts maxi mum val ues, respecti vel y. (These are not necessari l y Xi ~ and
Xi ~, nor are they al ways the angl es at whi ch cos Xi has i ts mi ni mum or
SEC.4q STUDY OF THE STRUCTURAL-ERROR FUNCTI ON 73
maxi mum val ues. ) Then on combi ni ng Eqs. (40), (41), (42), one fi nd%
after some tri gonometri c mani pul ati on, that
,,= _ , co: ; f ~: ) )
*m
[ ( ) - cosr i xa) l 43)
X COS Xi-xojxb
jti s thus symmetri c about the val ue of 0; correspondi ng to X, = ~ b,
mi dway between the val ues of 9~for whi ch Hk = Oand Hk = 1; i t i s of the
form of a si nusoi d mi nus a constant, and vani shes for Hk = O and
0
-OT+s-
0.5 - E:=O.25
0.- -
(r-l
1,0)
E:=-O.5
Fm. 4,10.Structural -error functi ons for noni deal harmoni c transformers. The
functi onsshown are (a) j!(%), and (b) to (h) (2L/a2)8Hk for a seri es of val ues of Ek*, when
Xi - = -15, XW = 75.
H&= 1. I ts general form i s thus easi l y sketched wi thout r efer ence to
Tabl e Al . When H, i ncreases monotoni cal l y wi th .9L,f,(di ) i s sym-
metri cal about Oi = ~~a fact whi ch makes computati on even si mpl er.
To i l l ustrate the change i n form of 6Hk wi th changi ng E: l et us consi der
the speci al case of a harmoni c transformer for whi ch 15< X; <75.
The vari ati on of H~ wi th $i for thi s transformer i s shown by the mi ddl e
cur ve of Fi g.411. Fi gure 4.10 shows the form of j~(Oi ), and of
for a seri es of val ues of El. When E: i s l ess than 0.5 or gr eater than
1.5, JH~haa nearl y the same form as fZ(OJ, whi ch i s symmetri cal about
74 HARMONI C TRANSFORMER LI NKAGES [SEC.4.7
Oi = 0.5. To pr oduce a desi red form of H; that di ffers from the gi ven
(H;l @i)
(a) E:=O.O
(Hhle,)
L=-2
/
/
(b) ~= 0.5
(c) G=l .o
Fm. 4.1l.~k(&) for noni desd har-
moni c transformers. Xi fn = -15,
Xi H = 76, L k 2, Ek as i ndi cated.
H~ by a symmetri cal cor r ecti on 6Hk,
one woul d thus choose Ek < 0.5 or
E~ > 1.5; to rai se the Hk curves i n the
center one woul d use a posi ti ve L
(l i nk to the l eft) i n the fi rst case and
a negati ve L i n the second, whereas
to depress the cur ve i n the center
these ori entati ons of the l i nk woul d
be r ever sed. To l i ft or depress the
Hk-curves for smal l 0,, wi th l i ttl e
change for Oi near 1, E? = O i s an
appropri ate choi ce; to make a change
near Oi = 1 but not near f?~= O, one
shoul d take Et = 0.75. Wi th E:
rangi ng from 0.25 to 0.5 i t i s possi bl e
to depress one si de of the cur ve whi l e
rai si ng the other , and so on. These
observati ons of cour se appl y onl y to
the parti cul ar harmoni c transformer
her e con si de red; si mi l ar sketches
woul d need to be made as the basi s for
a di scussi on of other cases.
The magni tude of 6Hk i s di rectl y
control l ed by the choi ce of L. I t wi l l be
noted however , that when (2L/gz) 6Hk
i s smal l , as for E; = 0.5, a parti cul arl y
smal l val ue of L may be requi red i n
or der to gi ve ~Hk a desi red magni tude.
I n general , i t i s rel ati vel y di ffi cul t to
depress one si de of the cur ve Hk(oi )
whi l e rai si ng the other , and one may
fi nd that an i mpracti cal l y smal l val ue
of L i s requi red to pr oduce a desi red
effect. On the other hand, i f one de-
si res merel y to r educe bHk bel ow some
establ i shed tol erance one can wi th
advantage make E: = 0.5, si nce con-
veni entl y smal l val ues of L are then
accept abl e.
The magni tude of C?Hki n typi cal
cases i s i l l ustrated by Fi g. 4.11, i n
whi ch ~Hk i s gi ven for thr ee val ues of
SEC. 4.8] DESI GN OF NONI DEAL HARMONI C TRANSFORMERS 75
Ep(O.0, 0.5, and 1.0) wi th L = f 2. The di fference between the exact
cal cul ati ons on whi ch these graphs are based and approxi mate cal cul a-
ti ons usi ng the resul ts of Fi g. 4 10 woul d not be evi dent to the eye. I t
i s to be emphasi zed, however , that fi nal cal cul ati ons shoul d be made usi ng
the exact formul as i n al l cases i n whi ch c approaches 45 (a val ue whi ch, for
mechani cal reasons, ought never to be much exceeded).
4.8. A Method for the Desi gn of Noni deal Harmoni c Transformers.
The exper i enced desi gner of noni deal harmoni c transformers wi l l fi nd i t
possi bl e to guess sati sfactori l y the requi red val ues of Et and L, gui ded
onl y by vi sual compari son of the Hk-curves wi th the desi red form of H;,
and perhaps a few expl or ator y computati ons. On the other hand, a
si mpl e and strai ghtforward desi gn pr ocedur e can be based on the resul ts of
the precedi ng secti on. To i l l ustrate thi s, we return to the probl em (Sec.
4.4) of usi ng a harmoni c transformer to mechani ze the rel ati on X2 = tan z1,
for OO< z, <50. Here, however , we shal l add the requi rement that the
angul ar travel of the transformer shal l be twi ce as great as that previ ousl y
Used: AXi = 80.
Despi te the i mposi ti on of thi s addi ti onal condi ti on, i t remai ns tr ue
that i t i s best to mechani ze xl as a l i near di spl acement, w as an angul ar
di spl acement: the best fi t for Tabl e 41 i s to be found i n Tabl e A2, rather
than i n Tabl e A 1. Si nce Tabl e A. 1 i s to be used i n the determi nati on of
Es and L i t i s conveni ent to retabul ate the rel ati on of the homogeneous
vari abl es h, and h, for equal l y spaced val ues of the vari abl e h,, whi ch i s to
be i denti fi ed wi th Oi . The resul t i s shown i n the fi rst two col umns of
Tabl e 43. The best fi t for the rel ati on thus expressed i s to be found i n
Tabl e A-l , for Xi ~ = 5, XW = 750the same val ues, of course, for
whi ch one fi nds i n Tabl e A2 the best fi t to Tabl e 4 1. The fi t coul d be
i mproved somewhat by i nterpol ati on i n the tabl es, the best val ue of Xa~
l yi ng between 10 and 5. We shal l not bother wi th thi s i nterpol a-
TARLE 4.3.COMPUTATI ONSI N DESI GNI NG A HARMONI C TRANSFORMER
h, = e, h, Hk h, Hk f,
fz
(W,).,vwx. (~),mox. (f)e-t
0.0 0,00000.0000 0.0000 0.000o 0.0000 0.000o 0.000o 0.0000
0.1 0.13590.1325 0.0034 0.1378 0.2701 0.0048 0.0014 0.0018
0.2 0.26830.2640 0.0043 0.2225 O .4858 0.0051 0.0008 0.0010
0.3 0.3934 0.3919 0.0015 0.2565 O .6433 0.0015 0.0000 0,0000
0.4 0.5097 0.5139 O .0042 0.2465 o .7387 o .0049 0.0007 0.0012
0.5 0.6158 0.6274 0.0116 0.2027 O.7708 0.0127 0.0011 0.0018
0.6 0.7115 0.7304 0.0189 0.1396 O.7387 0.0197 O.0008 0.0018
0.7 0.79670.8207 O.0240 0.0724 0.6433 O.0240 0.0000 0.0010
0.8 0.87260.8967 0.0241 0.0173 O.4858 O.0233 0.0008 0.0002
0.9 0.9401~0.95690:0168 0.0110 0.2701 0.0158 0,0010 0.0002
1.0 1.00001.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.000o
76 HARMONI C TRANSFORMER LI NKAGI X [SEC.4.8
ti on, but shal l choose Xi - = 5, Xi ~ = 75, and thr ow the enti re bur-
den of cor r ecti ng our desi gn on the choi ce of E~ and L. The val ues of H,
read from Tabl e A 1 are shown i n Col umn 3 of Tabl e 4.3. The desi red
val ue of 6H~i s then h, Hk, shown i n Col umn 4 of thi s tabl e.
As the next step, j,(ei ) and j,(13J are computed (COI U~S 5 and 6).
By Eq. (38) we can express ~Hk i n terms of these functi ons:
c$Hk= aj,(&) + tJ.f2(&), (45)
wher e
~=ti
2L
(46)
and
b g?.
(47)
Our probl em i s then to make a l i i ear combi nati on of Col umns 5 and 6 that
wi l l approxi mate Col umn 4 as wel l as possi bl e. I t i s a si mpl e matter to
fi nd the best fi t i n the sense of l east rms er r or , but an even si mpl er method
wi l l suffi ce: we shal l fi t 6Hk to h, H~ exactl y at two chosen poi nts. I n
appl yi ng such a method some di screti on i s necessary as a poor choi ce of
these poi nts may l ead to a bad over-al l fi t. We choose to make the fi t
exact at @i = 0.3 and at fl = 0.7, assuri ng a pr oper hei ght for the pri nci pal
maxi mum i n 6Hk and a change i n si gn near the cor r ect val ue of Oi . The
er r or i n the mechani zati on wi l l then vani sh for four nearl y equal l y spaced
val ues of Oi : 0.0, 0.3, 0.7, 1.0. We requi re then
0.2565cJ 0.6433b = 0.0015,
0.0724a 0.6433b = 0.0240.
Hence
a = 0.1385, b = 0.0529.
By Eqs. (46) and (47),
E? = ; = 0.382,
}
(48)
(49)
(50)
The correspondi ng val ues of 6Hk (as computed by thi s approxi mate
method) appear i n Col umn 7 of Tabl e 4.3, and val ues of
the resi dual er r or i n the mechani zati on, i n Col umn 8. The maxi mum
er r or i n the mathemati cal desi gn thus appears to be about 0.1 per cent of
the total travel . The maxi mum val ue of si n c for thk desi gn i s
(si n f)mu =
g(l 0.382) _ 0243
1.788
(51)
SEC.4.9] TWO I DEAL HARMONI C TRANSFORMERS I N SERI ES 77
a suffi ci entl y smal l val ue to assure good accuracy of the approxi mate
formul as empl oyed. Exact cal cul ati on of the total desi gn er r or i n the
mechani zati on (l ast col umn of Tabl e 43) shows that i t nowher e exceeds
0.2 per cent, a hi ghl y sati sfactory resul t. The devi ce i tsel f i s sketched i n
Fi g. 4.12.
I f excessi vel y l arge val ues of c occur i n a desi gn thus determi ned, the
exact val ues of 6H~ will not be i n sati sfactory agreement wi th h, Hk.
A further cor r ecti on i n 6H.t is then necessary. Thi s maybe added to the
oti gi nal val ues of h, H~, and the process of determi ni ng E: and L car,
ri ed through as before. The quanti ti es 6Hk, computed wi th the resul ti ng
constants by the exact formul a, shoul d now show better agreement wi th
the desi red val ues (the ori gi nal h, Hk). Repeti ti on of thi s process wi l l
usual l y l ead to a sati sfactory desi gn, except when excessi vel y l arge val ues
H/ii
+~ --H--+-
_+L
Fm. 4.12.Hmmoni c transformer mechani zi ng Z, = tan z,, 0 < z, <50.
of 6are cal l ed for. I n such cases another choi ce of X;~ and X~~ may hel p,
or another type of l i nkage may be requi red.
HARMONI C TRANSFORMERS I N SERI ES
49. TWO I deal Harmoni c Transformers i n Seri es.-Wi th a si ngl e
harmoni c transformer one can mechani ze onl y a rel ati vel y narrow fi el d of
functi ons. These devi ces have al so a mechani cal di sadvantage i n that
one termi nal rotates or i s r otated by a shaft, whi l e the other pushes or i s
pushed by a sl i de; usual l y one desi res that al l cel l s i n a computer have
termi nal moti ons of the same type.
As a fust step i n the extensi on of the fi el d of mechani cal functi ons we
consi der the combi nati on of two i deal harmoni c transformers i nto an
i deal doubl e harmoni c transformer, as shown i n Fi g. 413. Thi s
mechani cal cel l has sati sfactory mechani cal properti es, wi th both ter -
mi nal s movi ng i n strai ght l i nes. The fi el d of functi ons that i t can gener -
ate can be descri bed by thr ee i ndependent parametersfor i nstance, by
78 HARMONI C TRANSFORMER LI NKAGES [SEC.49
AXi , Xi ~, Xj~, wher e AXi i s the range of angul ar moti on common to both
arms of the rotati ng member, and Xi - and Xi ~ are the mi ni mum val ues
for the angul ar parameters Xi and Xj, whi ch descri be the ori entati on of
the two arms. Al though a consi derabl e vari ety i n form of the gener ated
functi on i s obtai nabl e by pr oper choi ce of these parameters, the i deal
doubl e harmoni c transformer i s best sui ted to the mechani zati on of
monotoni c functi ons wi th a mi l d change i n curvature (as i n Fi g. 4.14)
and functi ons of roughl y si nusoi dal character (as i n Fi g. 4.16).
FI Q. 4.13.I deal doubl e harmoni c transformer.
Mechani cal l y, the acti on of the doubl e harmoni c transformer may be
thus descri bed: The i nput parameter Xl i s transformed i nto a r otar y out-
put parameter X3 by the fi rst harmoni c transformer; thi s rotati on i s
i mparted to a second harmoni c transformer, for whi ch i t serves as a
r otar y i nput parameter, Xl ; Xi i s transformed by the second harmoni c
transformer i nto the fi nal output parameter X2.
of the correspondi ng homogeneous vari abl es,
OS= (@@J . H,,
04 = (e4p3) . 1%= e3,
Hz = (H2104) 04,
or , combi ni ng these rel ati ons,
Symbol i cal l y, i n terms
(52)
(53)
(54)
H, = (~,lt94) (L%IL%) (631H1) HI = (H2]83) (031H1) H,. (55)
SEC.4.9] TWO I DEAL HARMONI C TRANSFORMERS I N SERI ES
(H2IHJ
(f331HI )
79
Fm. 4.14.-Graphi cal constructi on of the functi on gener ated by a doubl e harmoni c trans-
for mer (Fi g. 4.13),
0.0
0.5
HI i
~:,>>~ ~
,4
7
.
\
\
Hz
o 0.5 1.0
Fm. 4.15.I deaf doubl e harmoni c transformer.
L
80 HARMONI C TRANSFORMER LI NKAGES [SEC.4.10
From thi s symbol i c equati on i t i s evi dent that one can fi nd the oper ator
for a doubl e harmoni c transformer,
(56)
by the graphi cal mul ti pl i cati on of oper ator s for the component harmoni c
transformers, as expl ai ned i n Chap. 3. The oper ator (031H1) may be
obtai ned from Tabl e A.2, the oper ator (Hz{OJ from Tabl e A 1; they must
of cour se cor r espond to the same val ue of AX2.
As an exampl e we take a doubl e harmoni c transformer (Fi g. 4.13) for
whi ch 75 s X3 s 15; 25 ~ X4 ~ 65; AX, = AXe = 90.
We fi nd i n Tabl es Al and A.2 the fol l owi ng rel ati ons:
H,
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
e,
O.0000
0.1942
0.3207
0.4249
0.5175
0.6033
0,6649
0.7641
0.8422
0.9204
1.0000
e,
0.0
0.1
0.2
0.3
0.4
0.5
0,6
0,7
0.8
0.9
1.0
M,
0.0000
0.1106
0.2263
0.3443
0.4616
0.5754
0.6828
0.7813
0.8684
0.9419
1.0000
Fi gure 4.14 shows graphs of these two operators, and the geometr i c con-
structi on requi red for thei r mul ti pl i cati on as requi red by Eq. (56). The
graphi cal representati on of the product (HZI HJ i s an al most ci rcul ar arc,
qui te di fferent from the functi ons mechani zabl e by a si ngl e harmoni c
transformer.
Another typi cal exampl e of two harmoni c transformers i n seri es i s
shown i n Fi g. 4.15. The travel s are AXS = AX4 = 90, wi th
The oper ator (oSI H,) i s the one used i n the precedi ng exampl e, and the
oper ator (H~\ @J wi l l be found i n Tabl e A 1. These oper ator s are pl otted
and thei r graphi cal mul ti pl i cati on i ndi cated i n Fi g. 4.16. The resul ti ng
oper ator i s r epr esented by a defor med si nusoi d wi th i ts maxi mum di s-
pl aced to the l eft.
4.10. Mechani zati on of a Gi ven Functi on by an I deal Doubl e Har-
moni c Tranaformer.-As the fi rst step i n mechani zi ng a functi onal rel a-
ti on by an i deal doubl e harmoni c transformer, i t shoul d, as usual , be
expressed i n homogeneous vari abl es:
h, = (h2[h1) h,, (57)
wi th hl the i nput vari abl e, hz the output vari abl e. One then desi res to
fi nd i deal -harmoni c-transformer oper ator s (Hzl d,) and (L9,[H,) whi ch
SEC.4.10]
MECHANI ZATI ON OF A GI VEN FUNCTI ON
81
cor r espond to the same val ue of AX1 and whi ch make
(H,I HJ = (H,]e,) . (e3[H,)
(56)
approxi mate as wel l as possi bl e to the gi ven oper ator (h21hJ. I t i s
necessary for mechani cal reasons, whi ch appl y whenever the sl i de
termi nal of a harmoni c transformer i s used as the i nput, that (@,~HJ not
,
Fm. 4.16.+rapbi cal constructi on of the functi on gener ated hy a doubl e harmoni c tran%
for mer (Fi g. 4.15).
i nvol ve an i ti ni ty i n ~; we need consi der onl y those cases for whi ch
dH I
Tabl e AZ i s constructed, wi th 90< X~_, X,~ <90.
Sol uti on of thi s probl em fal l s i nto two steps:
1. A prel i mi nary sol uti on of the probl em, by whi ch an appropri ate
val ue of AXi i s fi xed upon and a prel i mi nary choi ce of X3. and
X4- i s made.
2. I mprovement of the choi ce of Xon and X,n by a process of succes-
si ve approxi mati ons.
I L,
82 HARMONI C TRANSFORMER LI NKAGES [SEC.4.1~
To gai n a prel i mi nary esti mate of an appropri ate val ue of AXi one
may fi t the gi ven cur ve ver y roughl y by a secti on of si nusoi d (by r efer ence
to Tabl es A. 1 and A2, or even by a vi sual esti mate); the angul ar range of
thi s secti on of si nusoi d wi l l be approxi matel y the desi red val ue of AXC
The roughness of the approxi mati on wi l l be evi dent from i nspecti cm of
Fi gs. 4.14 and 416, i n both of whi ch the curves cor r espond to AXi = 90.
However , the nature of the cal cul ati ons requi red i n computi ng doubl e
harmoni c-transformer functi ons i s such that i t i s desi rabl e to begi n an
attempt to fi t a gi ven functi on by fi xi ng on a val ue AXi , even when the
choi ce must be made qui te arbi trari l y. By adjusti ng the parameters
Xa~ and X4- one can then, i n pri nci pl e, obtai n the best fi t of the mechan-
i zed functi on to the gi ven functi on consi stent wi th the chosen AXi ; by
repeati ng thi s for a seri es of val ues of AXi one coul d at l ength determi ne
the best val ue of thi s parameter and the best possi bl e fi t to the gi ven
functi on. I n practi ce, i t i s not necessary to fi nd the best fi t careful l y for
each AXi . I n the prel i mi nary cal cul ati ons i t i s suffi ci ent to use a si mpl e
and easi l y appl i ed method of fi t i n choosi ng X3- and Xtm, to establ i sh an
equal l y si mpl e cri teri on for the accuracy of the over-al l fi t thus obtai ned,
and to choose the best AXi i n the sense of thi s cri teri on. When a val ue
of AXi has been establ i shed i n thi s way, i t then becomes wor th whi l e to
use mor e careful methods, descri bed i n Sec. 4 13, i n the further adjust-
ment of Xa~ and Xd~.
We shal l consi der separatel y the qui te di fferent methods of getti ng a
prel i mi nary fi t to monotoni c functi ons (Sec. 4.11) and to functi ons wi th
maxi ma and mi ni ma (Sec. 4.12).
4.11. Prel i i ary Fi t to a Monotoni c Functi on.-A monotoni c func-
ti on wi l l i n general be fi tted by a monotoni c functi on; the range of Xt wi l l
not i ncl ude ei ther +90 or 90. I n thi s case one has automati cal l y a
fi t of the gener ated functi on to the gi ven functi on at both ends of the range
of vari abl es. I n addi ti on, for any gi ven AX{ the val ues of X3- and X,n
can be so chosen that the gener ated functi on wi l l (1) agr ee wi th the gi ven
functi on at any chosen pai r of i nteri or poi nts, or (2) have the same sl opes
as the gi ven functi on at the two ends of the range of the i nput vari abl e,
or (3) have the same rati os between the sl opes at any thr ee poi nts i n the
range of the i nput vari abl es. The fi rst of these methods of fi tti ng woul d
i n many cases be the most sati sfactory; however , i t i s the most di ffi cul t to
appl y and wi l l not be consi dered further. The second method has som~
what wi der uti l i ty than the thi rd and wi l l be made the basi s of our further
di scussi on.
When X* and X4- are so chosen that the gener ated functi on not onl y
fi ts the gi ven functi on at the end poi nts but has the same sl ope as wel l , a
sati sfactory fi t i s assured throughout a mor e or l ess broad regi on near
both ends of the range of vari abl es. The fi t wi l l then be good ever ywher e
Sm. 4.11] PRELI MI NARY FI T TO A MONOTONI C FUNCTI ON
83
if the gi ven functi on i s wel l adapted to mechani zati on by an i deal har-
moni c transformer wi th the chosen val ue of AXi . I f the chosen val ue of
ni i s not appropri ate, the central porti on of the gener ated functi on,
havi ng been subject to no control duri ng thi s si mpl i fi ed fi tti ng process,
may show marked di fferences from the gi ven functi on.
As an i ndi cati on
of the over-al l accuracy of fi t attai ned i n thk process, and of the appro-
pri ateness of the chosen val ue of AXi , i t i s natural to take the di fference
between the gener ated and the gi ven functi ons at the mi dpoi nt of the
curve, HI = i; AX; shoul d then be so chosen as to mi ni mi ze thi s di fference.
The fol l owi ng steps can thus be used i n obtai ni ng a prel i mi nary fi t to a
monotoni c functi on:
1.
2.
3.
4.
5.
6.
I t
Chooses val ue of AXi J arbi trari l y i f ther e i s no gui de.
Choose XS?I Iand X~ (by a method to be descri bed bel ow) such that
the sl ope of the gener ated functi on has the pr oper val ues for
H~=Oand Hl =l .
Wi th these val ues of the parameters, fi nd the val ue of Hz when
H, = i. (fh can be read from Tabl e A2, si nce AXs and Xs~ are
known; usi ng thi s val ue of &to enter the col umn of Tabl e Al that
corresponds to the known val ues of AXa and X4., i nterpol ate to
fi nd the requi red val ue of Ht.)
The di fference d between thi s and the desi red val ue of HZ i s taken
as a measure of the over-al l er r or i n the fi t.
Repeat the precedi ng steps for several other val ues of AX;, unti l
the trend of d as a functi on of AXi i s establ i shed.
Choose as the val ue of AX{ to be used i n further cal cul ati ons the
one whi ch mi ni mi zes I dl .
remai ns to descri be a qui ck and easy methmd for fi ndi ng those
val ues of X3~ and XA~ for whi ch the gener ated functi on has speci fi ed
termi naJ sl opes:
For mechani cal reasons the i nput transformer must be such that
08=04=0
(58)
(59)
84 HARMONI C TRANSFORMER LI NKAGES [SEC.4-11
bl.i+tf
+0.2K44+
_z -04
!!@
f
-0.6
t
-0.8
--95
- . .
--

-.
-1.01
I I I I I I I I I
1
406060708090 100 110 120 130 140
Fm. 4.17.Logari thm of the i ni ti al sl ope,
()
dH
J of i deal -harmoni c-tranaformw
z *-O
functi ons, pl otted agai nst angul ar travel for a seri es of val ues of Xi ~. Dashed l i nes i n~-
cate that the i ni ti al dope i s negati ve.
Sm. 4.11] PRELI MI NARY FI T TO A MONOTONI C FUNCTI ON
85
+0.4
+0.2
40 50 60 70 so 90 100 110 J
1 1 1 I
1
120 130 MO
( ).
dH
FI G. 4.18.Logari thm of the fi nal sl ope, ~ ~ ,J of i deal -harmoni c-transfor mer
functi ons, pl otted agai nst angul ar travel for a seri es of values of X{~, Dashed l i nes i ndi -
cate that the fi nal dope i s negati ve.
l
86
when HI = O,
and
HARMONI C TRANSFORMER LI NKAGES [SEC.411
and I %= (l , = 1 when H, = 1. Thus
(-)
dHZ
()
deg g,~
i %= $Ho= dH1
,=
(-)
dth e-o
(60)
(61)
I n other words, each termi nal sl ope of the graph of the doubl eharmoni c-
transformer oper ator (Hz IHI ) i s equal to the correspondi ng termi nal
sl ope for the output oper ator (Hz 10s)di vi ded by that for the i nput oper ator
(HI ]190). Our probl em i s thus, i n effect, to pi ck out of the part of Tabl e
A-l that corresponds to a gi ven val ue of AXi two col umns such that the
rati o of thei r i ni ti al sl opes i s SOand the rati o of thei r fi nal sl opes i s S1.
Consi der now Fi g. 417, whi ch shows the vati ati on wi th AXi of the
quanti ty
( )-
10!310 $~o
(
COSXim
)
= l ogo (si n X,)- (si n XJ&
(62)
for a seri es of val ues of Xi n. On thi s chart the di stance al ong the verti cal
l i ne AX = AXi from the cur ve Xi . = X3. to the cur ve Xi . = X,.
(counted as posi ti ve upward, negati ve downward) i s
( )
dHZ
()
dH,
l ogl o so = 10!410 ~ ~j=o 1%10
dds .9,-o
(63)
( )
dH8
the l ogari thm of the i ni ti al sl ope
dHl a,-o
for an i deal doubl e harmoni c
transformer characteri zed by the parameters AXi , Xi ~, X4*. Con-
versel y, i f we draw a l i ne of l ength l ogl o So on a stri p of paper and move
thi s, al ways i n a verti cal posi ti on, over Fi g. 4.17, i ts ends wi l l conti nual l y
i ndi cate the parameters AXi , X3~, and Xd~ for an i deal doubl e harmoni c
()
dHZ
transformer wi th i ni ti al sl ope
dHl H,==il
equal to the chosen val ue of SO.
Fi g. 4.18 presents i n a si mi l ar manner val ues of
10glot:)o.l=lOglO[(sinx):::Linx)til 64)
I t i s obvi ous that i f we draw a l i ne of l ength l ogl o S, on a stri p of paper
and move i t, al ways i n a verti cal posi ti on, over Fi g. 4.18, i ts ends wi l l
Sm. 4.11] PRELZMI NAR Y FI T TO A J UONOTONZC FUNCTI ON
87
conti nual l y i ndi cate the parameters AXij X~, and X4-for an i deal doubl e
( )
dH,
harmoni c transformer wi th termi nal sl ope -
dHl xl -,
equal to the chosen
val ue of S1.
I n or der to determi ne the parameters of an i deal doubl e harmoni c
transformer for whi ch the i ni ti al and termi nal sl opes have val ues SOand
S, respecti vel y, one may pr oceed as fol l ows. (Attenti on wi l l be restri cted
to cases i n whi ch SOand S1 are both posi ti ve; a case i n whi ch both sl opes
are negati ve can be r educed to tti s case by repl aci ng X1~ by Xk~ + 1800.)
At the edge of a stri p of paper draw an ar r ow of l ength I l og,o SOI (usi ng
the scal e at the l eft of Fi g. 4-17) and pl ace i t on Fi g. 4.17, di recti ng i t
upward i f I ogl o SOi s posi ti ve and downward i f thi s i s negati ve. Si mi l arl y
construct an ar r ow of l ength l l ogl a Sll and pl ace i t on Fi g. 418, di recti ng
i t upward or downward accordi ng as l ogl o S 1i s posi ti ve or negati ve. I f
these arrows are pl aced on verti cal l i nes correspondi ng to the same
AX = AX,, wi th the heads of both arrows on curves comespondi ng to the
same X~ = Xk~ and the tai l s on curves correspondi ng to the same
X. = Xs~, then these val ues of AXi j Xa~, and X4. gi ve si mul taneousl y
the desi red i ni ti al and fi nal sl opes.
Such posi ti ons for the arrows can be
found qui ckl y, for any speci fi ed AXi , by pl aci ng the tai l s of the arrows
successi vel y at several val ues of XS~, unti l a val ue i s found for whi ch the
heads of the arrows al so l i e at the same X4-.
Exampl e: As our pri nci pal exampl e of doubl e-harmoni c-transformer
desi gn we shal l take the probl em of mechani zi ng the rel ati on
zz = tan xl , (65)
previ ousl y consi dered, over the l arger range 0 s z, s 70,
0 S X, s 2.7475.
On i ntroducti on of homogeneous vari abl es
hl = &,
2 = 2.7275
(66)
thi s rel ati on becomes
2.7475hZ = tan (h, o70) (67)
Thi s i s tabul ated for uni forml y spaced val ues of h, i n Tabl e 4.4. The
sl ope of the cur ve i n homogeneous vari abl es i s
dh,
= 0.4447 secz xl ,
dh,
(68)
and the termi nal sl opes are 0.445 and 3.802.
For a prel i mi nary fi t we tr y AXt = 90. We pl ace on the cor r espond-
i ng l i ne i n Fi g. 4.17 an ar r ow of l ength I l ogl o 0.4451, and on that l i ne i n
88 HARMONI C TRANSFORMER LI NKAGES [SEC.4.11
TABLE 4.4,z, = TANz,, O = z, s 70, I N HOMOGENEOUS VAEUABLES
h, h,
0.0 0.0000
0.1 0.0447
0.2 0.0907
0.3 0.1397
0.4 0.1935
0.5 0.2549
0.6 0.3277
0.7 0.4187
0.8 0.5396
0.9 0.7143
1.0 1.0000
Fi g. 4.18 an ar r ow of l ength l og,, 3.802. We note that i f Xa~ = 10, cor -
r ect i ni ti al sl ope requi res X4~ = 58 (Fi g. 4.17), and cor r ect fi nal sl ope
requi res X4~ = 52 (Fi g. 4.18); i f Xi - = 15, cor r ect i ni ti al sl ope
requi res X4m = 62, cor r ect fi nal sl ope requi res Xd~ = 75. I nter-
pol ati ng to zer o di fference of the val ues of X4~, we have a set of constants
assuri ng cor r ect termi nal sl opes:
AXi = 90, X,m = 12, x4m = 59.
Assumi ng these constants, we now compute Hz for H, = +. Fi rst
we center attenti on on the i nput harmoni c transformer and determi ne
tJ8= 19q:i n Tabl e A.2, AXi = 90, we i nterpol ate between col umns for
Xi n = 15 and Xi . = l OO; for H = +, Xi. = 12 we fi nd
6, = 0.385 z L93= 04.
Turni ng attenti on to the second transformer, we can now determi ne Hz:
i nterpol ati ng between col umns of Tabl e A 1 for AX~ = 90, Xi ~ = 60
and Xi . = 55, we fi nd that H = 0.325 when Xi . = 59 and
Oi = 0.385. The desi red val ue of H,, read from Tabl e 4.4, i s 0.255; the
cur ve thus fi tted l i es too hi gh i n the center by d = 0.070.
Next we tr y AX~ = 70. Movi ng the arrows to the correspondi ng
l i nes of Fi gs. 4.17 and 4.18, we fi nd that cor r ect termi nal sl opes are
obtai ned by usi ng
AX~ = 70, X3~ = 8, Xi . = 56.
TWth these constants, i f H, = +, then es = 0.371, Hz = 0.308, d = 0.053.
Tri al of sti l l smal l er val ues of AXi shows that d can be decreased onl y
sl i ghtl y bel ow thi s val ue; an exact fi t of termi nal sl opes wi l l al ways l ead to
a gener ated cur ve too hi gh i n the mi ddl e. The best val ue of AXi , i n
thi s sense, i s a l i ttl e smal l er than i s mechani cal l y desi rabl e, and not much
can be gai ned by adopti ng preci sel y thi s val ue i nstead of a l arger and mor e
conveni ent one. I n the further di scussi on of thi s probl em we shal l
ther efor e fi x AXi = 90.
1
I
I
SEC.4,12] PRELI MI NARY Y FI T TO A NONMONOTONI C FUNCTI ON 89
4.12. Prel hi nary Fi t to a Nonmonotoni c Functi on.-Nonmonotoni c
functi ons that can be gener ated by an i deal doubl e harmoni c transformer
possess onl y a si ngl e maxi mum or mi ni mum. Expressed i n homogeneous
vari abl es, they fal l i nto four types i l l ustrated i n Fi g. 4 19:
(a) Hz = O when H, = O.
(b) H, = O when HI = 1.
(c) H, = 1 when HI = O.
(d) H, = 1 when Hl = 1.
As wi th monotoni c functi ons, i t i s possi bl e to fi nd, for any gi ven AXi ,
val ues of Xam and X4. that make the termi nal sl opes of the gener ated
functi on equal to those of a gi ven nonmonotoni c functi on. However , a
fi t of the val ue of the gener ated functi on to that of the gi ven functi on i s
assured at onl y one end of the range of H 1: for HI = Owi th types (a) and
mm
2 U L(LJ
HI Hl
FI O.4.19.Types of functi ons mechani zabl e by an i deal doubl e harmoni c transformer.
(c), and for HI = 1 wi th types (b) and (d). Agr eement of the sl opes at
the other end of the range of H, thus does not assure tangency of the
gi ven and the gener ated functi ons, and the fi t may be ver y unsati sfactory.
For thi s reason i t i s not advi sabl e to make a prel i mi nary fi t to a gi ven
nonmonotoni c functi on by the method of Sec. 4 11. I t i s usual l y best to
choose a val ue of Xf~ such that a fi t i n the val ue of the functi on i s secured
at the end wher e thi s i s not otherwi se assured, and then make the maxi -
mum or mi ni mum i n H* occur for the pr oper val ue of H 1; as an i ndi cati on
of the accuracy of the over-al l fi t one can take the di fference between the
gi ven and gener ated functi ons at a chosen poi nt between the maxi mum
or mi ni mum and the mor e r emote end of the range of H1.
The pr ocedur e for securi ng a prel i mi nary fi t to a nonmonotoni c func-
ti on i s then as fol l ows:
1. Choose a val ue of AXi , arbi trari l y i f necessary.
90
2.
3.
4.
5.
6.
7.
8.
HARMONI C TRANSFORMER LI NKAGES
[SEC. 4.12
Referri ng to Tabl e A-l , choose X,~ such that Hz has the d-i red
val ue when i l ~= HI = 1, for types (a) or (c), or when % = HI = O}
for types (b) or (d).
From the same col umn of Tabl e A-l read the val ue of& for whi ch
HZ = 1 [types (a) or (b)] or the val ue of OSfor whi ch HZ = O[types
(c) or (d)].
From the gi ven functi on, determi ne the val ue of H, for whi ch
Ha = 1 [types (a) or (b)] or the val ue of H1 for whi ch Hz = O[types
(c) or (d)]
By r efer ence to Tabl e A-l or A-2, for the same AXi l fi nd the val ue of
Xa~ for whi ch the val ue of L%determi ned i n Step (3) corresponds to
the val ue of H, determi ned i n Step (4).
For these val uea of AXi , Xa~, and X4-, determi ne the di fference d
between the gener ated functi on and the gi ven functi on at the
chosen test val ue of HI .
Repeat the precedi ng steps for several other val ues of AX{, unti l the
trend of d as a functi on of AXi i s establ i shed.
Choose as the val ue of AXi for use i n further cal cul ati ons that whi ch
mi ni mi zes I dl .
Exampl e: As an exampl e, we take the probl em of maki ng a prel i mi nary
fi t to the cur ve (Hz]H1) of Fi g. 4 10-a case i n whi ch we happen to know
that an exact fi t can be obtai ned.
The cur ve i s of a borderl i ne type,
bel ongi ng to types (a) and (b). For the purposes of the prel i mi nary fi tti ng
we desi re
H%=O when HI = O,
H2=0 when HI = 1,
H,=l when HI = 0.38.
For test purposes, we shal l compare the gener ated functi on wi th the gi ven
functi on when Hl = 0.70 (desi red val ue, Hz = 0.710).
Fi rst, choose AX~ = 70. I n Tabl e Ac1 we fi nd that Hz = Ofor both
0s = O and 03 = 1 i f X,~ = 55; from the same col umn we see that
f+t = 1 for 133= 0.5. The desi red X% must then make 63 = 0.5 cor -
respond to HI = 0.38. From Tabl e A-l i t i s evi dent that
75 < Xh < 70;
i nterpol ati ng, we obtai n Xs~ = 72.
To test the over-al l fi t gi ven by AX, = 0.70,X% = 72, X% = 55,
we compute Hz for HI = 0.70. I nterpol ati ng i n Tabl e A*2 (si nce HI has a
val ue appeari ng ther e) between col umns correspondi ng to X* = 70
and X3n = 75, we fi nd 93 = 0.771. Returni ng to Tabl e A-l ,
X,m = 55,
1 I
SEC.4.13] A METHOD OF SUCCESSI VE APPROXI MATI ONS 91
we obtai n by l i near i nterpol ati on Ht = 0.692 for OS= 0.771. Li near
i nterpol ati on, however , i s her e obvi ousl y i nadequate; quadrati c i nterpol a-
ti onyi el ds Hz = 0.700, d = 0.010.
Repeati ng the process wi th AX~ = 90 we fi d that l i ttl e i nterpol ati on
i s necessary. To make HZ = O for 83 = O and for 19S= 1 requi res
Xk = 45; the maxi mum comes for 03 = 0.5, HI = 0.38; hence
X3. = 75.
Computi ng HZ for Hl = 0.7, we obtai n essenti al l y the graphi cal l y deter -
mi ned val ue, 0.710, and d = O.
Al though AX{ = 90 i s the best val ue, i t i s evi dent that the fi t i s not
ver y sensi ti ve to the choi ce of AXi .
4.13. I mprovement of the Fi t by a Method of Successi ve Approxi ma-
ti orM.-A sati sfactory fi t of the gener ated to the gi ven functi on i s not
assured by the si mpl e and rather arbi trary methods just descri bed; these
shoul d be depended upon onl y i n choosi ng a val ue of AXi . The fi nal
adjustment of Xs~ and X1~, to obtai n the best over-al l fi t possi bl e wi th
the chosen AX;, i s most sati sfactori l y accompl i shed by a graphi cal method
of successi ve approxi mati ons whi ch gi ves a compl ete vi ew of the fi t at
each stage of the process. Conver gence of the successi ve approxi mati ons
on the fi nal resul t can be speeded up by exer ci se of the superi or judgment
of an exper i enced desi gner, but a sati sfactory resul t i s assured even for a
begi nner.
The probl em to be sol ved i s that of fi ndi ng i deal harmoni c trans-
for mer oper ator s (Hz Ith) and (@tl Hi ), both correspondi ng to the chosen
AXi j whi ch make the approxi mate rel ati on
(H,[8,) . (e,] H,) = (H,I H,) = (hzl h,) (69)
as nearl y exact as possi bl e over the enti re range of vari abl es. Thi s wi l l
be done by al ternatel y i mprovi ng the choi ce of the two harmoni c-trans-
for mer operatorsthat i s, the choi ce of the parameters X1~ and Xs~,
respecti vel y.
Let the harmoni c-transformer oper ator s chosen after S stages i n the
approxi mati on be (Hzl 63)8 and (OSI H1)s.
Then
(H21L%)s (@31HJs = (h,lh,).
(70)
Let i t be desi red to repl ace (Hzl 03).9by an oper ator (HZl &).+1, whi ch
wi l l make the approxi mati on of Eq. (70) mor e exact. Let the oper ator
Z, be defi ned by
(h,l h,) (H,[fh)s = 2s. (71)
Then
Z.s = (H2]@3)s, (72)
92 HARMONI C TRANSFORMER LI NKAGES [SEC.4.13
as may be shown by mul ti pl yi ng Eq. (70) from the ti ght by the oper ator
(H, I0,)s. I f thi s approxi mati on wer e exact, Eq. (70) woul d necessari l y
be exact; i f tl us approxi mati on i s i mproved, that of Eq. (70) wi l l be
i mproved. Now Z, can be computed wi th su5ci ent accuracy by graphi -
cal methods. I f i t i s possi bl e to fi nd an i deal -harmoni c-transformer
oper ator (Hzl (?a)~+1whi ch gi ves a better fi t to Z~ than does (H2103)s, then
thi s i s the desi red i mproved oper ator ; the approxi mati on i n the rel ati on
(H,i o,)s+, (O,pl)s = (1+,) (73)
i s better than that i n Eq. (70).
Next one wi l l wi sh to repl ace (osI HI )S by an i mproved oper ator
(0,117,)~+1. Let the oper ator Y~+l be defi ned by
(h,lh,) . Ys+, = (H2183)s+,. (74)
By Eq. (73)
(H,\&)s = Y.+,. (75)
An i mproved oper ator (HI I63)8+1woul d make thi s approxi mati on mor e
exact; one can ther efor e determi ne i t by computi ng Y~+l by graphi cal
means and fi ndi ng the i deal harmoni c transformer functi on that best
fi ts thi s functi on. The approxi mati on i n wri ti ng
(H2103)S+I (03] HJ S+I = (h,ph) (76)
i s then even better than that i n Eq. (73).
I t i s now possi bl e to make a further i mprovement i n (H2183), comput-
ing ,z~+l by Eq. (71) and fi tti ng (Hzl e3)~+Zto thi s as exactl y as possi bl e.
The oper ator (OSI H1)can then be i mproved agai n, and the process r epeated
unti l the i mprovement obtai ned does not repay the effor t expended. I t i s
of cour se possi bl e that a sati sfactory fi t can not be gi ven by any i deal
doubl e harmoni c transformer; i t wi l l then be necessary to make use of
methods to be descri bed l ater i n thi s chapter.
Exampl e: We return to the Exampl e of Sec. 411, the mechani zati on
of the tangent functi on from 0 to 70.
We ther e fi xed on the val ue
AXi = 90 and found approxi mate val ues of Xa~ and Xl ~. Roundi ng
off these val ues to those appeari ng i n Tabl e A 1, we mi ght take
(H,l O,) ~ X,. = 60 (77)
(O,I H,) x X,~ = 5. (78)
These val ues, especi al l y the second, are good. I n or der to provi de a
better i l l ustrati on of the method of successi ve approxi mati ons we shal l
del i beratel y sel ect a poor er val ue, Xs~ = 15, wi th whi ch to start the
computati ons.
Fi gure 4.20 shows a graph of the gi ven functi on (h21hJ, poi nts on
the graph of (Hl l Q 1= 15 s Xa S 75, as read from Tabl e Al , and
I
SEC.4.13] A METHOD OF SUCCESSI VE APPROXI MATI ONS
93
the constructi on needed to determi ne correspondi ng poi nts on the graph
of Zl , whi ch has been drawn i n as a conti nuous curve. A good over-al l
I
fi t to Z, can not be found i n Tabl e Al (Axi = 90 !), but a reasonabl e
fi t at the l ower end i s obtai ned by taki ng Xd~ = 70, as shown i n
the same fi gure. Ther efor e, as the basi s for the next step i n the com-
putati on we make (H210s)2 cor r espond to Xd~ = 70.
I I I
I I I
11A I I I I I
I
M
Fm. 4.20.Mechani zati on of z! = tan m. Fi rst step i n the method of successi ve
approxi mati ons: constructi on of the oper ator Z, and approxi mate fi tti ng of thi s by
(H,)d,) , * AX, = 90, Xi m = 70.
Next, Fi g. 4.21 shows the constructi on used i n determi ni ng Y,. (I n
G
practi ce thi s woul d be carri ed out on the same graph as the constructi on
for Z,, but for the sake of cl ari ty a new fi gure i s used here.) The
oper ator (HI I03)2 can be made to fi t thi s fai rl y wel l by taki ng x% = 5.
Repeti ti on of thi s process wi l l l ead to l i ttl e further i mprovement.
I t can be seen i n Fi g. 423 that Z~ i s perhaps best fi tted by the val ue of
X,-, 70, arri ved at i n Fi g. 420.
I t woul d be of l i ttl e val ue to r educe
the er r or further by i nterpol ati on i n the tabl es; the sol uti on woul d i n
any case appl y onl y to an i deal doubl e harmoni c transformer, whi ch
coul d be real i zed onl y by usi ng undesi rabl y compl ex mechani sms or
. -..
1
I 94 HARMONI C TRANSFORMER LI NKAGES [SEC.413
(
(
/
Y2
[
/
/
( )
()
f
Fm. 4.21.Mechani zati on of 22 = tan z,, Second step i n method of successi ve
~pprO~mati ons: constructi on of the oper ator Yz and approxi mate fi tti ng of thi s by
(H,I OJZw AX, = 90, X,~ = 5.
X4M=20
x3m=-9
/
x,
1
LI
I
/,-
/
t
I
Ii
H2
I
I , t 1
0.0 0.5 1.0
X2
FI G. 4.2!2.-I deaf doubl e harmoni c transformer approxi matel y mechanki ng Z* - tanz,.
SEC.4.14] NONI DEAL DOUBLE HARMONI C TRANSFORMERS 95
nomdeal doubl e harmoni c transformers wi th ver y l ong l i nks. I t i s
better to desi gn the transformer as noni deal , further reduci ng the er r or
by adjustment of l i nk l engths and sl i de posi ti ons, as expl ai ned i n Sec. 413.
Fi gure 4.22 shows the i deal doubl e harmoni c transformer cor r espond-
i ng to the present stage of sol uti on of the probl em. The cel l has been
normal i zed by maki ng
R,[(si n X,)- (si n X,)ti .] = I i l J(si n X~)~.. (si n X&,]. (79)
4.14. Noni deal Doubl e Harmoni c Transformers.-The fi el d of
mechani zabl e functi ons i s ver y substanti al l y extended i f noni deal har-
moni c transformers are coupl ed i nstead of i deal ones. (A typi cal non-
i deal doubl e harmoni c transformer i s shown i n Fi g. 4.26.) I nstead of
thr ee i ndependent parameters, ther e are seven to be adjusted: AXS = AX4,
Xa~, X,~, LI , LZ E?. and E?.
Here, as before, the l engths L1 and Lz
of the l i nks are to be measured i n terms of the hori zontal travel s AX1
and AX2, respecti vel y. E? i s the readi ng on the H~-scal e wher e i t i s
i ntercepted by the center l i ne of the X{ sl i de, and E: i s the readi ng on
the H#scal e wher e thi s i s i ntercepted by the center l i ne of the Xi sl i de.
The Peaucel l i er i nversor shown i n Fi g. 2.4 i s a speci al case of the non-
i deal doubl e harmoni c transformer, wi th X3~ = X4*, L = L, and
E? = E? = O; i t i s thus evi dent that such devi ces can ser ve for the
mechani zati on of functi ons that are not even roughl y of si nusoi dal form.
To determi ne the functi on gener ated by a gi ven noni deal doubl e
harmoni c transformer one can appl y the method descri bed i n Sec. 4.8,
obtai ni ng the oper ator (HZI H1) as the product of oper ator s (Hzl oS) and
(t9*]H~),whi ch descri be the component noni deal harmoni c transformers.
I n the conver se probl em of mechani zi ng a gi ven functi on by a non-
i deal doubl e harmoni c transformer i t i s not feasi bl e to vary al l seven of
I
) the avai l abl e parameters si mul taneousl y. One shoul d begi n as though
(
the doubl e harmoni c transformer wer e to be i deal , carryi ng out an
approxi mate fi t (Sees. 4 11 and 412) to determi ne a val ue of AX{, whi ch
I
i s hel d constant thereafter, and then i mprovi ng the choi ces of Xn~ and
X4. (Sec. 4.13) unti l thi s ceases to be profi tabl e. At thi s poi nt i t becomes
necessary to begi n the adjustment of LI , Lz, E?, E;. Si nce the devi ce
t
may be r egar ded as two- noni deal harmoni c transformers i n seri es, the
probl em to be sol ved i s sti l l formal l y the same as that consi dered i n
Sec. 413-that of maki ng the approxi mati on i n
(H,\ O,) (0,111,) = (H,I H,) = (h,l h,) (69)
as nearl y exact ae possi bl eand the method of sol uti on by successi ve
approxi mati ons i s the same. Here, however , each of the component
)
transformed i s characteri zed not by one, but by thr ee constants, (X8.,
96 HARMONI C TRANSFORMER LI NKAGES [SEC.4.14
~~, LJ or (X,~, E~, L2), whi ch must be chosen at each stage of the process
for i nstance, by the methods of Sees. 44 and 4.8.
Exampl e: We conti nue the exampl e of Sec. 4.13, that of mechani zi ng
the tangent functi on from 0 to 70. Fi gure 4.23 shows the constructi on
of Zn, to fi t whi ch we shal l now adjust the thr ee constants characteri zi ng
(Hnl th): X4~, E:, and Lg. As al ready noted, the best fi t obtai nabl e wi th
1
)/ 1 1 1 1
(H2I e3)3(-70+20; E*=O.5, L =-2)
( (H2I e3)2(-70- 20)
I I 1
FI G. 4.23.Mechani zati on of zz = tan z,. Thi rd step i n method of successi ve appr oxi -
mati ons: constructi on of the oper ator Z~ and approxi mate fi tti ng of thi s by
(H*[6%),-J AXi = 90, X;m = 70, ( . . . ),
and by (HzI OJZ* AXi = 90, X;- = 7o, E* = 0.5, L = !2(crosses).
an i deal -harmoni c-transformer oper ator i s gi ven by X~ = 70,
X4- = 20; the resi dual er r or then changes si gn twi ce, tendi ng to be
l arge near the ends of the range of vari abl es. Now the l i mi ts of X, her e
are roughl y the same as those of the exampl e of Sec. 4.8 ( 15, 75)
except for a change i n si gn, and the geometri cal si tuati ons di ffer onl y as
mi rror i mages i f one repl aces a l i nk to the l eft by a l i nk to the ri ght, and
vi ce versa. Correspondi ngl y, one easi l y sees, the structural er r or func-
ti ons 6Hk appl i cabl e her e di ffer from those of Fi g. 4.10 i n repl acement of
SEC.4.14] NONI DEAL DOUBLE HARMONI C TRANSFORMERA9
97
ei by 1 6,, that i s, refl ecti on of the curves i n a verti cal l i ne. Thus i t
i s evi dent that i t i s not possi bl e, by any choi ce of EZ and Lz, to obtai n a
structural -error functi on that changes si gn twi ce, such as i s needed to
gi ve a good fi t to 22 over the whol e range of vari abl es. We shal l ther e-
for e concentr ate our attenti on on i mprovi ng the fi t for l ow val ues of 6,
rai si ng the cur ve i n thi s regi on, and attempti ng onl y to keep the change
smal l el sewhere. I nspecti on of Fi g. 4.10 then shows (the di fferences of
V i
/ -7,5+i2.5) \
1}
1 I
{
1b
/
(HI 103)3(-7.5+81.5;
E*=O.2L=+ 2)
()
/
(H*I e3)3(-70s+ 20;
E~0.5, L=-2)
/ .
Fm. 4.24.Mechani zati on of %x = tan m. Fourth step i n the method of successi ve
apprOfi mati Ons: cOnstructi Onof the Omrati r Yt, and approxi mate fi tti ng of thi s by an
i deal -harmoni c-transformer functi on wi th AX; = 90, X;n = 7.5 (ci rcl es) and by
(H@)~ - AXi = 90, Xi rn = 7.5, E* = 0.2. L = 2 (crosses).
the present from the for mer case bei ng borne i n mi nd) that Et = ~ i s
an appropri ate val ue, and that L shoul d be negati ve, the l i nk to the ri ght.
Rough consi derati on of the magni tudes i nvol ved l eads to choi ce of
L = 2, The resul ti ng fi t, as shown i n Fi g. 4.23, i s qui te sati sfactory
for l ow val ues of O.
The process of successi ve approxi mati ons i s conti nued i n Fi g. 4.24
wi th the graphi cal constructi on of Ys. We have now to fi t (Hl l OJ3 to
thi s by choosi ng X~., E?, L,. I nspecti on of Tabl e A. 1 shows that wi th
98 HARMONI C TRANSFORMER LI NKAGES [SEC,414
Xj~ = 5 one has a bad fi t at the l eft, and wi th X3~ = 10 the cur ve
i s much too l ow i n the mi ddl e. I n such a case i t i s desi rabl e to i nter-
pol ate. We choose Xs~ = 7.5. The val ues of the correspondi ng
functi on are found wi th suffi ci ent accuracy for our graphi cal method by a
l i near i nterpol ati on i n Tabl e A. 1; the resul ti ng val ues are pl otted i n
Fi g. 4.24 as a seri es of poi nts i n smal l ci rcl es. I n further i mprovi ng the
fi t one wi l l wi sh to rai se the central part of the curve, to depress the
(
(b
(3
(>
(H, 163)~(-7,5--82.5, E*=0.2,L =+2)
f
()
/
\ ,
/ r
(i-f*I e3)4(-~50-15;
E*=0,42,L =-0.8)
/
\ I
.
1 1 1 1
FI G. 4.25.Mechani zati on of Z* = tan z,. Fi fth step i n the method of successi ve
wwofi mati ons: constructi on of the oper ator Za, and approxi mate fi tti ng of thi s by (Hzl fh)c
AXi = 90, X;m = 75, E* = 0.42,L = 0.S.
extr eme upper end, and to l eave the l ower and unchanged. The val ues
of XS~ and Xi ~ are suffi ci entl y l i ke those of Fi g. 4.10 for thi s to be used
as a gui de; i t i s agai n evi dent that no choi ce of constants can accompl i sh
everythi ng that i s desi red. We choose ther efor e to al l ow a consi derabl e
er r or at the l ower end of the curve, l eavi ng thi s to be cor r ected (as befor e)
by our choi ce of E; and Lz; we concentr ate on i mprovi ng the fi t at the
upper end, and, secondari l y, that i n the central regi on. I nspecti on of
Fi g. 4.10 l eads to choi ce of EY = 0.2 and L = +2. Computati on of the
structural -error functi on then l eads to the cor r ected poi nts of Fi g. 4.24,
SEC. 4.14] NONI DEAL DOUBLE HARMONI C TRANSFORMERS
99
i ndi cated by crosses. I t i s evi dent that a val ue of E? near er to zer o
woul d have been preferabl e, as gi vi ng a depressi on of the cur ve over a
l ess extended regi on. However , i t i s hardl y wor th whi l e at thi s stage
of the computati on to make a mor e careful choi ce of constants, and we
accept the resul ti ng functi on as (HI I0s)3.
The next stage of the cal cul ati on, the determi nati on of (H21oS),, i s
shown i n Fi g. 4.25. When 23 i s constructed i t i s found that a better
fi t can be obtai ned at the upper end by taki ng Xs~ = 75 than by
taki ng Xtm = 70. The er r or functi ons shown i n Fi g. 4.10 then
appl y exactl y, wi th the substi tuti on Oi ~ 1 Oi . Si nce prel i mi nary
fi ts have been made i n al l parts of the range of Oi , i t i s now wor th whi l e
to make a careful adjustment of the constants E; and L2, as by the
methods of Sec. 48. Wi th Et = 0.42, L2 = 0.8, one fi nds exactl y
computed poi nts that gi ve an excel l ent fi t except at the extr eme upper
end of the curve. Thi s i s ver y sati sfactory, as i t i s i n thi s l ast regi on
that the fi t i s bei ng control l ed by choi ce of E? and L1.
The fi nal graphi cal stage of the sol uti on, the determi nati on of (HI I Q,,
i s not i l l ustrated by a fi gure. I t l eads to the choi ce of X% = -7.5,
E? = 0.2, L1 = +1.8, wi th an excel l ent over-al l fi t. Thi s i s as far as
the fi t can be carri ed by these graphi cal methods; further refi nements are
best obtai ned by the methods di scussed i n Chap. 7.
We have thus arri ved at the fol l owi ng choi ce of constants:
x% = 7.5, Xs~ = 82.5, E? = 0.2, L, = 1,8, (80a)
X4. = 75, X,af = 15, E: = 0.42, L, = 0.8. (80b)
Cal cul ati on of the resul tant total structural er r or i s i l l ustrated i n
Tabl e 4.5, whi ch consi sts of thr ee secti ons. The fi rst shows the cal cul a-
ti on, by the method descri bed i n Sec. 4.6, of val ues of the homogeneous
i nput parameter H{ for a seri es of val ues of 08. The second shows the
cal cul ati on of val ues of the homogeneous output parameter H: for the
same seri es of val ues of 83. I n the thi rd secti on ther e are shown cor -
respondi ng val ues of
xl = Hi 70, (81)
the gener ated tangent functi on
zt~ = Hi tan 70,
(82)
the i deal tangent functi on X2 = tan xl , and the i deal gener ated homd
geneous vari abl e h~. The er r or i n the gener ated functi on, ~hz = H; h2,
i s found to be l ess than 0.8 per cent of the total vari ati on of the output
vari abl e.
The l i nkage correspondi ng to these constants i s drawn i n Fi g. 4.26.
Reducti on of the l i nkage to the normal i zed form shown her e requi res a
100 HARMONI C TRANSFORMER LI NKAGES f8EC.4.14
TABLE4.5.COMPUTATI OXOFmm TOTALSTRUCTURALERROR(T_JNRAGE OFFI G.4.26)
si n 61
H:
DH,
93 HI
E;
fi x 1 Cosc, =L, x
= L,
(1 Cos61
(H: E?)

(H2]e,)4 0.0 0.0000 0.7887 0.3396 0.0594 0.1069

0.1 0.1398 0.7986 0.3438 0.0610 0,1098
0.2 0,2791 0,7801 0.3358 0.0581 0.1046
0.3 0,4143 0,7337 0.3159 0.0512 0.002
0,4 0.5421 0.6605 0.2844 0,0413 0,0743
0.5 0.6586 0.5624 0.2421 0.0297 0.0535
0.6 0.7633 0.4418 0.1902 0.0183 0.0329
91 =0.7749 0.7 0.8513 0.3016 0.1298
0.0085 0.0153
0.8 0.9211 0.1453 0.0626 0.0020 0.0036
~ =0.4305
0.9 0.9711 O .0233 0.01043 0.0000 0.0000
L, 1.0 1.0000 0.2000 0.0861 0.0037 0.0067
e, H,
H:
E:
sin S* 1 Cosq
-llll-1
(H,]8J , 0.0 0.0000 0.4200 0.3178
0.1 0.0428 O .2204 O. 1667
0.2 0.1038 0.0344 0.0260
0.3 0.1819 0.1336 0.1011
9, =0.6052 0.4 0.2748 0.2793 0.2113
0.5 0.3804 0.3992 0.3020
0.6 0.4961 0.4904 0.3710
g,
0.7 0.6189 0.5505 0.4164
z,
= 0.7564 0.8 0.7459 0.5782 0.4374
0.9 0.8740 0.5726 0.4331
1.0 1.0000 0.5340 0.4040
0.0518
0.0140
0.0003
0.0051
0.0226
0.0467
0.0714
0.0908
0.1007
0.0986
0.0852
es
ZI , degr ees
0.0 0.00
0.1 11.10
0.2 21,53
0.3 31.09
0.4 39.63
0.5 47.08
0,6 53.63
0.7 59.10
0,8 63.62
0.9 67.24
1.0 70.00
x2g
O.0000
0.2061
0.4094
0.6187
0.8416
1.0653
1.3562
1.6592
1.9952
2.3615
2.7475
tan21
0,0000
0.1962
0.3945
0.6030
0,8282
1.0754
1.3579
1.6709
2.0163
2.3836
2.7475
DH,
0.0414
0.0112
0.0002
0.0041
0.0181
0.0374
0.0571
0.0726
0.0806
0.0789
0.0682
H,
+ DH, H,
(DHJo
O.0000 0.0000
0.1427 0.1586
0,2768 0.3076
0,3996 0.4441
0.5095 0.5662
0.6052 0.6726
0.6893 0.7661
0.7597 0,8443
0.8178 0.9089
0.8642 0.9805
0.8998 1.0000
I
I
H,
+ DH, H:
(DH,)o
O.0000 0,0000
0.0730 0.0750
0.1450 0.1490
0.2192 0.2252
0.2981 0,3063
0.3844 0.3950
0.4804 0.4936
0.5877 0.6039
0.7067 0.7262
0,8365 0.8595
0.9732 1.0000

h,
O.0000
0.0714
0.1436
0.2195
0.3014
0.3914
0.4942
0.6081
0.7339
0.8675
1.0000
ah~
O.0000
0.0036
0.0054
0.0057
0.0049
0.0036
0.0006
0.0042
0.0077
0.0080
O.0000
SEC.4.15] ALTERNATI VE METHOD OF TRANSFORMER DESI GN
101
sl i ghtl y di fferent cal cul ati on from that i n the case of i deal doubl e har-
moni c transformers (Eq. 79). Perhaps the si mpl est method i s to choose
arbi trary val ues of RI and Rz, and to compute the correspondi ng travel s
AX{ and AX; from the geometr y of the l i nkage. Si nce these travel s
are proporti onal to the Rs, and are to be equal i n the normal i zed cel l ,
one has as the rati o of the normal i zed arm l engths
1
AX;

--r%--L
/
.~
-
(83)
I I L !--/
-L. . . =
Fra. 4.26.Noni deal doubl e harmoni c transformer gener ati ng, approxi matel y, Z2 = tan ZI ,
0< z, <70. Val ues of the constants are gi ven i n Eq. (SO).
4.15. Al ternati ve Method for Doubl e-harmoni c-transformer Desi gn
The graphi cal method descri bed i n Sec. 4-14 has the advantage that i t
permi ts readjustment of the constants X3* and XA~ at al l stages of the
desi gn process. The al ternati ve method to be descri bed i n the present
secti on i s useful when val ues of AX3, Xs_, and Xl ~ can be consi dered as
fi xed; i t i s essenti al l y an extensi on of the method of Sec. 4.8, whi ch per -
mi ts si mul taneous adjustment of the constants Ll, L2, ET, E;, of the
two harmoni c transformers.
Let us assume that a gi ven rel ati on
h* = (h,ph) h
(84)
has been mechani zed approxi matel y by an i deal doubl e harmoni c trans-
for mer that generates the rel ati on
H2 = (H21HI ) . H1 (85)
102 HARMONI C TRANSFORMER LI NKAGES [SEC.4.15
between i ts i nput and output parameters. Thi s rel ati on we cw express
parametri cal l y i n terms of the angl e 63:
H, = H,(6,),
H, = H,(8,).
}
(86)
l Wthout changi ng the constants AX3, Xi -, X4* of thi s l i nkage, l et the
i deal harmoni c transformers be repl aced by noni deal ones. The i nput
and output parameters wi l l then be H; and H;, di fferi ng from HI and
H, by the structural -error functi ons 6H, and 6%:
H~(19S)= H,(oa) + ~H1(@,),
H;(6,) = H,(03) + 6H2(L93).
}
(87)
The resul ti ng noni deal doubl e harmoni c transformer wi l l then gener ate
a rel ati on
H; = (Hj]H;) H;. (88)
Our probl em i s to assi gn to the constants L,, Lz, E?, l??, val ues such that
Eq. (88) wi l l approxi mate as cl osel y as possi bl e to the gi ven rel ati on,
Eq. (84), when H; takes over the r ol e of H,, H; that of H,.
I t was shown i n Sec. 48 that when 13H1and 8He are smal l one can
wri te
6H1 = @1(@3) + f#2(03),
(89a)
6H, = cj3(03) + dj4(193), (89b)
wher e
(90)
The functi ons f,(e,) andj,(e,) can be computed usi ng Eqs. (39) and (40),
wi th Hk repl aced by HI (%); fi (th) and jl (OS) are al so computed by Eqs.
(39) and (40), respecti vel y, wi th H, repl aced by HZ(19J.
Let i t be desi red to choose the constants a, b, C, d, i n such a way that
the l i nkage generates the desi red rel ati on exactl y for some fi xed val ue of
03:
h,(es) = H,(@,) + af,(@ + b.fz(o,), (91a)
I
h,(03) = Hz(@3) + c~3(L93) + dj4(193).
(91b)
Equati on (84) speci fi es whi ch val ues of hl and hz, must cor r espond to
each other , but ther e i s nothi ng to prescri be whi ch pai r of val ues (h,, h2)
must cor r espond to any gi ven val ue of 0,. We coul d, for i nstance, pi ck
thi s pai r arbi trari l y and sti l l sati sfy Eqs. (91) by pr oper l y choosi ng the
di sposabl e constants. However , we do know that i f Eqs. (91) are to
be accurate ~H, and ~Hz must be smal l ; h,(OJ must be nearl y equal to
SEC.4.15] ALTERNATI VE METHOD OF TRANSFORMER DESI GN
103
Hl (th), h,(OS) nearl y equal to Hz(03).
We ther efor e pl ace on our choi ce
of the pai r of val ues (hi , hz) onl y the condi ti on that
h,(O,) = H,(h) + Ah,, (92)
wher e Ahl i s smal l . The correspondi ng val ue of hz(Os)i s easi l y computed.
Let
h, = hj)(e,)
when hl = H1(oJ. (93)
Then
(94)
to terms of the fi rst or der i n the smal l quanti ty Ahl . Combi ni ng Eqs.
(91,) (92), and (94), we fi nd that the condi ti ons to be sati sfi ed are
afl (h) + bj.d%) = Ahl ,
(95a)
cj3(83) + d~4(L93) = h~)(fl a) Hz(6a) + ~ . Ahl . (95b)
El i mi nati ng Ah, from these equati ons, we have
= h\J(Q I I ,(6J. (96)
T~s i s the onl y condi ti on that must be sati sfi ed by the constants a, b,
c, d, so l ong as no attempt i s made to control the val ue of the smal l
quanti ty Ah,.
Si nce one can sati sfy si mul taneousl y four condi ti ons such as Eq. (96),
i t i s possi bl e to make the l i nkage gener ate the desi red rel ati on exactl y
at four chosen val ues of 0s. One has to sol ve four si mul taneous l i near
equati ons i n the four unknowns a, b, c, d:
( )
dh,
()
dh, ,,,, j,(f3$1)) b ~ a
dhl ,,,,, f(ef)) + cf3(19$) + dj,(ef)
,
= hjo)(e$)) H2(o$1),
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
(-) ()
(97)
dh,
j1(@$4)) ~ d!@
a dhl o,(o
dhl ,,,4$(0$4)) + cf3(~$4)) + df,(ef)
.
h\ 0)(O~4)) H,(o$4) .
The constants of the l i nkage can then be computed by Eqs. (9o). Thi s
shoul d be fol l owed by exact cal cul ati on of the functi on gener ated by the
l i nkage, as i n the exampl e of Sec. 4.14.
Exampl e.To i l l ustrate thi s method we shal l treat agai n the exampl e
consi dered i n Sec. 414. Here, however , we shal l accept as fi xed the
104 HARMONI C TRANSFORMER LI NKAGES
[SEC.414
constants arri ved at i n Sec. 4.13,
AX3 = 90, X3. = 5, X4. = 70,
(98)
(cf. Fi g. 422) and shal l adjust onl y the constants L,, L,, E?, E?.
The functi on Hi (HI ) gener ated by the l i nkage of Fi g. 4.22 can be
wri tten down i n parametri c form by r efer ence to Tabl e A 1, the val ues
of HI bei ng found i n the col umn AXi = 90, Xi ~ = 5; the val ues of
H,, i n the col umn AXt = 90, Xi = 70. These are shown i n Tabl e
4.6, together wi th the correspondi ng val ues of h~) [computed by Eqs.
(65) and (66) wi th hl set equal to H,] and the over-al l structural er r or .
The structural -error functi on exceeds 3 per cent of the total travel ; by
choi ce of the four di sposabl e constants we shal l now attempt to r educe
thi s er r or to zer o for I & = 0.2, 0.4, 0.6, and 0.8.
TABLE4.6.FUNCTI ON GENERATED BYLI NKAGEOF FI G.4.22
08 H, Hz
h(:)
h(:) H,
0.0 0.0000 0.0000 0.0000 0.0000
0.1 0.1448 0.0508 0.0650 0.0142
0.2 0.2881 0.1183 0.1337 0.0154
0.3 0.4262 0.2010 0,2087 0.0077
0.4 0.5559 0,2969 0.2938 0.0031
0.5 0.6738 0,4034 0.3926 -0.0108
0.6 0.7771 0.5181 0.5083 0.0098
0.7 0.8632 0.6381 0.6413 0,0032
0.8 0.9301 0.7604 0.7843 0.0239
0.9 0,9761 0.8820 0,9159 0.0339
1,0 1,0000 1.0CN30 1.0000 0.0000
We have fi rst to gi ve expl i ci t numeri cal form to Eqs. (97), whi ch
determi ne the constants a, b, c, d. Val ues of H; and H; are read from
Tabl e A 1, for the chosen val ues of L93;the js are then computed as
expl ai ned bel ow Eq. (90). Val ues of ~ can be computed by Eq. (68)
TARLE47.-CONSTANTSREQUI REDI N DESI GNPROCEDURE
T
Oa H?
0.0 0.9958
02 0.9719
0.4 0.8435
0.6 0.6232
0.8 0.3326
1.0 0.0000
f,(h) f2(M
o: 25s7 O,5260
0.2836 O.8025
0.1736 O.8025
0.0433 O 5260
Hz
0.0000
0.4159
0.7402
0.9411
0,9991
0.9083
f3(eJ
I
f4(e,)
I
dh,
m,
... . . . . .
0.0754 0.6169 0.5046
0 3030 0.9410 0.7344
0.4582 0.9410 1.3121
0,3709 0.6169 2.5094
. . . ,. ..,.
SEC.4.15] ALTERNATI VE METHOD OF TRANSFORMER DESI GN 105
wi th z, = 70 x HI . Al l these quanti ti es appear i n Tabl e 4.7. By
usi ng al so the l ast col umn of Tabl e 46 we can easi l y determi ne al l the
constants of Eqs. (97):
0.1305a + 0.2654b + 0.0754c 0.6169d = 0.0154
0.2083a + 0.5894b + 0.3030c 0.9410d = 0.0031
0.2278a + 1.0530b + 0.4582c 0.9410d = 0.0098
I
. (99)
0.1087a + 1.3199b + 0.3709c 0.6169d = 0.0239
Sol uti on of these equati ons yi el ds
a = 0.1966, b = 0.0566, c = 0.1874, d = 0.0651. (1OO)
Hence, by Eqs. (90),
L, = 1.806, L, = 0.703,
E? = 0.288,
E; = 0.347. (101)
The constants speci fi ed by Eqs. (98) and (101) are not ver y cl i ffer ent
from those found i n Sec. 4.14, and the l i nkage woul d cl osel y resembl e
that of Fi g. 426. The exact total structural er r or of the new l i nkage i s
gi ven i n Tabl e 4.8: i t i s about a thi rd of that of the fi rst desi gn. At fi rst
si ght i t may appear surpri si ng that the er r or does not vani sh for OS= 0.2,
0.4, 0.6, 0.8, si nce thi s was the condi ti on appl i ed i n determi ni ng the
constants of the l i nkage. I t i s to be remembered that the equati ons on
whi ch thi s method i s based are approxi mati ons obtai ned by treati ng c
TABLE4.8.ToTALSTRUCTURAL EmoE I NSECONDMECHAN1ZATI ON OFm = tan zI ,
L
Oa, H: = h,
0,0 0.0000
0,1 0.1589
0.2 0.3074
0.3 0.4429
0.4 0.5643
0.5 I 0.6709
0.6 0.7628
0.7 0.8409
0.8
i
0.9061
0.9 0.9589
1.0 1.0000
0<2, <70
H:
0.0000
0.0734
0.1460
0.2212
0.3020
0.3908
0,4905
0.6024
0.7268
0.8609
1.0000
h,
0.0000
0.0715
0.1435
0.2187
0,3001
0.3899
0.4901
0.6026
0,7280
0.8624
1.0000
ah~
0.0000
0.0019
0.0025
0.0025
0.0019
0,0009
0.0004
-0.0002
-0.0012
0.0015
0,0000
as a smal l angl e. The er r or computed by these formul as does vani sh
at the speci fi ed val ues of oS,but ther e are present other and l arger er r or s
due to the use of the smal l angl e approxi mati ons, whi ch are, essenti al l y,
those di scl osed by the exact cal cul ati ons on whi ch Tabl e 4.8 i s based.
We coul d make al l owance for these er r or s, approxi matel y, by repeati ng
the cal cul ati on, taki ng as the constants on the ri ght-hand si de of Eqs.
106 HARMONI C TRANSFORMER LI NKAGES [SEC.4.15
(99) sums of correspondi ng entri es i n the l ast col umns of Tabl e 46 and
4.8. Such a refi nement woul d be wor th whi l e onl y i f the mechani sm
wer e to be constructed wi th excepti onal care.
I n Sec. 6.6 we shal l meet a probl em i n whi ch the strai ghtforward
appl i cati on of thi s method l eads to a l ess sati sfactory resul t; the requi red
mcdfi cati on of the method wi l l be descri bed ther e.
CHAPTER 5
THE THREE-BAR LI NKAGE
6s1. Fundamental Equati ons for the Three-bar Li nkage.-A three-bar
l i nkage (Fi g. 5.1) consi sts of two cranks, l ?,, AZ, pi voted on a frame and
connected through a l i nk B,. The symbol s AI , BI , AZ, Bz wi l l be used
to r epr esent di stances between the pi votal poi nts wi thi n the cor r espond-
i ng mechani cal parts: B1 and A z are the l engths of arms of the cranks,
Bz i s the l ength of the connecti ng l i nk, and A I k the di stance between
pi vots i n the frame.
The three-bar l i nkage serves as a mechani cal cel l havi ng one of the
cranks as the i nput termi nal , the other as the output termi nal .
The
FI G.5.1.Symbol s used i n the di scussi on of three-bar l i nkages
i nput and output parameters, Xl , X2, are rotati ons of those cranks
measured cl ockwi se from a zer o l i ne passi ng through the pi votal poi nts,
SI and SZ, of the cranks; the zer o posi ti on for each crank i s that i n whi ch
~t poi nts toward the l eft.
; The functi onal rel ati onshi p of the parameters X,, X2 fol l ows from
the geometr y of the quadri l ateral S 1?1TzXZ.
To fi nd X2 graphi cal l y for a gi ven XI and di mensi ons ~ 1, B], Az, Bz,
one woul d fi rst construct the zer o l i ne SS2 and the line of the i nput
crank, SI T1. The end Tz of the output crank must l i e on a ci rcl e of radi us
Bz about Tl, and on a ci rcl e of radi us AZ about S,. I f these two ci rcl es
i ntersect, a sol uti on for X2 exi sts; i n general they wi l l i ntersect i n two
poi nts TZ+, Tz-, whi ch are verti ces of two congruent tri angl es, TSZTZ+
and Tl SzT2_. Let VI be the pri nci pal val ue of the angl e S&Tl , l yi ng
107
108
THE THREE-BAR LI NKAGE [SEC.5.2
between 180 and +180, and w the pri nci pal val ue of the angl e
T,S,T,+, l yi ng between 0 and +180. Ther e are then two possi bl e
val ues of X2:
X2+ = ~1 + ?72,
x2_ = 111 72. (1)
I n terms of the probl em speci hed her e, X2 i s thus a doubl -val ued func-
ti on of Xl ; the functi onal rel ati on XL. = (X21X1) cXl has two branches,
r epr esented by the oper ator s (X2+1X1) and (X21X1). I f ql i s not
restri cted to i ts pri nci pal val ue, X2 i s of cour se a Kl ghl y mul ti pl e-val ued
functi on of X,. Cases i n whi ch thi s mul ti pl e-val uedness i s of si gni fi cance
i n actual mechani cal cel l s wi l l appear l ater.
For numeri cal cal cul ati on of X, the fol l owi ng pr ocedur e i s probabl y
the best:
1, Compute the di agonal D of the quadri l ateral usi ng the cosi ne l aw:
(2)
D2 = A! + B? + 2A@l COSX,.
2. Compute q, and T, by further appl i cati ons of thi s l aw:
D2 + A! Bf,
Cos ql =
2A I D
wi th si n VI si n X1 >0, (3)
D2+A~ B:
Cos ?72=
2A,D
wi th O <72 < 180. (4)
3. Fi nd X?+, X2-by Eq. (l ).
5.2. Cl assi fi cati on of Three-bar Li i ,ages.-Three-bar l i nkages are
conveni entl y cl assi fi ed accordi ng to the i nherent l i mi tati ons on the range
of the i nput parameter Xl .
To fi nd these l i mi ts, wi thi n whi ch the func-
ti on X2(X,) i s defi ned, we obser ve fi rst that the di agonal D i s a si de of
the tri angl e SSZT1 and as such i s l i mi ted:
Si mi l arl y, si nce D i s a si de of the tri angl e Td3zTz+,
I -42
B21SDSA2+B2. (6)
These are the onl y l i mi tati ons on D, and they i mpl y the l i mi tati ons on Xl
wi th whi ch we are concer ned.
Si nce Eqs. (5) and (6) appl y si mul taneousl y, they must be consi stent;
unl ess ther e i s overl appi ng of the i nterval s set by them for D, i t wi l l not
be possi bl e to construct a cel l wi th the gi ven di mensi ons. When the
two i nterval s overl ap, D can take on any val ue wi thi n the range common
to them. As i s i l l ustrated i n Fi g. 5.2, the i nterval s can overl ap i n four
di fferent ways, whi ch form the fi rst basi s for our cl assi fi cati on of these
l i nkages:
SEC.5.2] CLASSI FI CATI ON OF THREE-BAR LI NKAGES
109
Cl ass u: I Az B2{ < 1A, B,\ < A, + B, < A2 + ~z,
(7)
Cl ass b: I A1 B,l < I Az B21<AI +B1<A2+B2, (8)
Cl ass c: I A2 -BZI <I A1 B11<A2+B2<A1 +B,, (9)
cl ass d: I A, Bll<lA2 Bzl<A~+B2<A1 +B,. (l o)
I n each case, D can take on al l val ues between the two i ntermedi ate
quanti ti es of the correspondi ng l i ne.
I A1-B,I A1+B,
The l i nkages of Cl ass a have an cl a~~. ~
unl i mi ted i nput, si nce Eq. (5) i mpl i es
4
I A~-Bzl
AZ+B2
no l i mi tati on on Xl , and Eq. (6) i s
automati cal l y sati sfi ed. Li nkages of
1A,-Bll AI +BI
the other thr ee cl asses have a l i mi ted ~l a~~~ ~
i nput range. Wi th l i nkages of
Cl ass b, passage through the val ue
~+Bz
I A2-B21
XI = 180 i s i mpossi bl e, si nce D can-
not assume the correspondi ng val ue
I AI - Bl l A,+B1
I AI Bl l . Wi th l i nkages of Cl ass c, cl assc , ,
passage through Xl = 0 i s excl uded,
lA2-B~
A2+B2
si nce D cannot assume the cor -
respondi ng val ue Al + B1. Fi nal l y,
1A*-Bl l A,+B1
wi th l i nkages of Cl ass d, passages
Cl assd
through Xl = O and Xl = 180 are
I
{
I A2-B21 A2+B~
both i mpossi bl e; D cannot attai n FI G. 5.2.Cl assi fi cati on of three-bar
val ues correspondi ng to ei ther of
l i nkages.
these poi nts. The r ange of the output vari abl es can be di scussed
si rnjl arl y.
From what has been sai d i t i s obvi ous that the four l i nkages wi th
1. Al =p, Bl =q, Az=r, B2= s
2. Al =q, Bl =p, A2=r, B~= S
3. Al =p, Bl =q, AZ=s, B2=r
4. Al =q, Bl =p, Az=s, Bz= r
bel oug to the same cl ass. Now the rel ati ve magni tudes of AI , A z, B,,
Bz form the basi s of a further subcl assi fi cati on of three-bar l i nkages, the
subcl asses bei ng gi ven the numeri cal desi gnati on above i f one takes
al ways p > q, r > s. That i s, Cl ass a l i nkages are di vi ded i nto four
subcl asses :
al : Al > B1, Az > Bz, (11)
u2: Al < I ll, Az > Bz, (12)
u3: Al > Bl, Az < Bz, (13)
a4: Al < Bl, Az < Bz, (14)
and the other cl asses are si mi l arl y di vi ded i nto four subcl asses.
1
1
0
T
H
E
T
H
R
E
E
-
B
A
R
L
I
N
K
A
G
E
[
S
E
C
.5
.
2

4
,
.
.
S
E
C
.
5
.
2
]
C
L
A
S
S
I
F
I
C
A
T
I
O
N
O
F
T
H
R
E
E
-
B
A
R
L
I
N
K
A
G
E
S
0

u
8
-
~
(
1
C
u

,
%

)
/

,
,
T
1
\
-
-
-
.
,
.
.
$
q
.
.
.
.

.
k

,
,
0
T
.
.
-
.
!
,
e -
u
112 THE THREE-BAR LI NKAGE [SEC.53
Fi nal l y, i n each subcl ass X2 i s a functi on wi th two branches, X2.+ and
X*, whi ch we pl ace i n separate sub-subcl asses of the subcl ass. Thr ee-
bar l i nkages are thus di vi ded i nto 4 X 4 X 2 = 32 sub-subcl asses i n al l .
A sub-subcl ass wi l l be i ndi cated by a symbol such as c3+, whi ch appl i es
to the posi ti ve branch of a l i nkage for whi ch
[A, B,I <I AI B,I <A, +B, <A,+ B,,
Al > Bl, Az < B,.
The general forms of the functi ons gener ated by al l these types of
three-bar l i nkage are i l l ustrated i n Fi g. 5.3. I n each case the X2 has
been pl otted as a functi on of X,, for a three-bar l i nkage wi th di mensi ons
i l l ustrated i n the adjoi ni ng sketch. A mechani cal confi gurati on and the
gener ated cur ve are both shown for the posi ti ve branch by conti nuous
l i nes, for the negati ve branch by dotted l i nes. The val ue of X2 shown i s
not necessari l y the pri nci pal val ue. I n some cases the posi ti ve and
negati ve branches joi n conti nuousl y, but al ways at a poi nt of i nfi ni te
sl ope near whi ch the l i nkage i s not operabl e. The r eader shoul d study
thi s fi gure careful l y, si nce one shoul d not attempt to mechani ze by thi s
means functi ons that obvi ousl y are not i ncl uded i n the cl ass of functi ons
of the three-bar l i nkage.
5.3. Si ngul ar Cases of Three-bar Li nkages. -Certai n speci al thr ee-
bar l i nkages that bel ong to mor e than one of the cl asses defi ned above,
n +.,
FI G. 5,4.Three-bar l i nkage wi th .4, + B, = AZ + B2.
as l i mi ti ng cases, have speci al properti es that enti tl e them to separate
menti on.
Case A: Al + I ?l = AZ + B,.
G
(15)
A l i nkage of thi s type (Fi g. 5.4) has a si ngul ar poi nt for Xl = O. So
l ong as the i nput vari abl e i s restri cted to a range not i ncl udi ng the poi nt
X, = O, the confi gurati on of the mechani sm and the val ue of the output
vari abl e are uni quel y determi ned. When Xl = O the val ue of X2 i s
sti l l uni quel y determi ned, but the mechani sm has at thi s poi nt an i nde-
termi nate moti on, ther e bei ng two possi bl e fi ni te val ues for dX~/dXl .
SEC.5.3] SI NGULAR CASES OF THREE-BAR LI NKAGES 113
Thus, when the i nput parameter i s al l owed to pass through the val ue
Xl = O, X, mayor may not pass from the posi ti ve to the negati ve branch
of the functi on, or conversel y; the val ue of X2 i s no l onger uni quel y
determi ned by the val ue of X 1, but may have ei ther of two val ues, unl ess
appropri ate stops are i ntroduced.
Case B: 1A, B,l = 1A, B,l. (16)
I n thk case (Fi g. 5.5) a si mi l ar si ngul ari ty exi sts for Xl = 180.
x~
\___
o
x,
\
\-.~
- . _
FI G.5.5.Three-bar l i nkage wi th A, BI = A2 l?,.
Case C: AI + BI =A*+B*
1A, B,l = I A, B,\
I
si mul taneousl y. (17)
I n thi s case ther e are, of course, si ngul ar poi nts for both X1 = O and
Xl = 180, as wel l as some other i mportant features that shoul d be
menti oned.
The condi ti ons i n Eq. (17) can be sati sfi ed i n two ways:
cl: Al1?, = (AZB,); BI = A2; A, = B,. (18)
c,: AI B1=A2 B2; Al = Az; BI = B,. (19)
The Case cl , the paral l el ogram l i nkage (Fi g. 5.6), i s ver y wel l known.
I ts posi ti ve branch (for O < Xl < 180) i s used to transmi t rotati on
from one shaft to another at the rati o 1 to 1, wi thi n l i mi ts set far enough
from the poi nts of si ngul ari ty, at whi ch backl ash may become i mportant.
(A good range i n practi ce i s 30< Xl <150, but l arger ranges can be
attai ned by i ncreased car e i n manufacture.) The correspondi ng nega-
ti ve branch of the l i nkage functi on, shown dotted i n Fi g. 5.6, i s rarel y
used; i ts curvature decreases as the l ength of the l i nk Bz i s i ncreased.
I t wi l l be noted that the vari ous posi ti ve and negati ve branches, di fferi ng
by changes i n Xl and X2 whi ch are mul ti pl es of %r, form a connected
networ k through the whol e of the X l X2-pl ane. I f no stops are i ntro-
duced the gener ated XE may or may not pass from a posi ti ve branch to a
114 THE THREE-BAR LI NKAGE [SEC.5.3
negati ve branch, or vi ce versa, ever y ti me Xl passes through a val ue
that i samul ti pl eofr. Theval ue of Xzi sthus notuni quel y determi ned
by the val ue of Xl ; i t i s not even restri cted to one of two val ues, asi n
Cases A and B; itmay take on an i nfi ni te number of val ues, whi ch fal l , of
course, i nto two sequences wi th spaci ng 27r,correspondi ng to the posi ti ve
and negati ve branches.
n ~:;~,
[
.+..
[x,
- YT
, / 0.,
n
// ?/
/ \ /
/
\
/
\ / /
/
A
FXQ. 5,6.Three-bar l i nkage wi th BI = A,, AI = B,.
Li nkages of Cl ass CZ (Fi g. 5.7) are of speci al i nterest i n that X2
remai ns zer o on part of the posi ti ve and negati ve branches, whatever the
val ue of X1; how thi s can happen wi l l be evi dent from the geometr y of
the sketch. [The cl assi fi cati on of branches as posi ti ve and negati ve i s
her e qui te formal ; physi cal l y i t woul d be mor e appropri ate to thi nk of the
branches as (1) the strai ght l i ne X2 = O, and (2) the osci l l atory cur ve
wi th conti nuous deri vati ve. ] I f the gener ated X2 i s fol l owi ng the posi ti ve
x,
----------- .-----
-P
;
---
--;\\
x,
o ~\
,.
\ \ _< .s
.;
FI Q. 5.7.Three-bar l i nkage wi th A ~ = A~, B, = B,.
branch between Xl = O and X, = r, and Xl passes through the for mer
poi nt, then X2 may conti nue to change at a uni form rate by passi ng over
to the negati ve branch, or i t may fol l ow the posi ti ve branch and remai n
zer o thereafter; thi s l atter behavi or can be assured by the i ntroducti on
of stops. Thi s type of l i nkage i s ther efor e of val ue i n mechani zi ng func-
ti ons wi th a di sconti nui ty y i n the deri vati ve. Unfortunatel y, these cel l s
cannot suppl y any appreci abl e effor t near the poi nt of si ngul ari ty;
SEC.5.3] SI NGULAR CASES OF THREE-BAR LI NKAGES 115
torques must be appl i ed to both cranks i n the di recti ons of the desi red
moti ons.
I n practi cal appl i cati ons the author uses a sti l l mor e speci al l i nkage,
wi th Al = B1 = AZ = Bz (Fi g. 5.8). Thi s i s al so a speci al case of the
other si ngul ar cl asses, A, B, and C1; i t i s i nteresti ng to obser ve how the
di verse curves of Fi gs. 5.4 to 57 can pass over i nto the curves of Fi g. 5.8
as a common l i mi ti ng case. Wi th thi s l i nkage thr ee types of confi gura-
n/- /
X2
(a)
,4
I /
1,~
r
(a)
A
(b) (b)
(b)
x1=(2n+l )ff
X2f
-IT
(c) (c) R
---- -------
/
xl
/
I
L
I
(a) . (b) ~ ,,,
(c) *
:(b)
I
1
X2=2n7r / /
I / /
;(ay
/ t - -
X
FK~. 5.8.Three-bar l i nkage wi th AI = B, =Aa = B2.
ti on are possi bl e, r epr esented by thr ee sets of l i nes on the graph i n Fi g.
5.8:
a.
b.
c.
The paral l el ogram l i nkage confi gurati on, r epr esented by the curves
X2 = xl + Zmn.
Confi gurati ons i n whi ch the i nput termi nal i s l ocked i n a defi ni te
posi ti on, Xl = (2n + l )m, whi l e the output termi nal can assume
any posi ti on.
Confi gurati ons i n whi ch the output termi nal i s l ocked i n a defi ni te
posi ti on, X2 = &n, whi l e the i nput termi nal can assume any
posi ti on.
Of parti cul ar i nterest are the transi ti ons between confi gurati ons of types
(b) and (c), whi ch can be assured by the use of stops. We shal l now see
how these can be used i n generati ng a functi on wi th a di sconti nuous
deri vati ve.
Fi gure 5.9 shows a mechani cal cel l for whi ch
X~ = aXi when Xi > 0,
Xk =. bxi when Xi <0,
mth
cz>l >b.
116 THE THREE-BAR LI NKAGE
[SEC. 5.3
I t consi sts of the l i nkage of Fi g. 5.8, wi th added i nput and output termi -
nal s whi ch are push-rods pi voted to the central l i nk Bz.
The i nput
and output parameters, Xi and X~, are di spl acements of these rods
perpendi cul ar to the l i ne of the pi vots S, and Sz of the three-bar l i nkage.
When X, = X~ = O, the l i nkage
i s i n i ts cri ti cal posi ti on, wi th
X, = 180, X, = OO; the two
cranks then just touch stops Cl ,
C2, whi ch l i mi t thei r moti ons to
X, a 180, X2 > 0. I f X, i s now
i ncreased by a push exer ted on the
Xi termi nal , the crank A z wi l l be
hel d fi rml y by the stop Cz, whi l e
the crank BI and the l i nk Bz wi l l
r otate together about thei r col -
l i near pi votal axes S, and Tz i nto,
FI G. 5.9.Mechani ca1 cel l gener ati ng a func-
ti on wi th di sconti nuous deri vati ve.
for exampl e, the confi gurati on
i l l ustrated i n Fi g. 5.9. The param-
eter X~ wi l l then i ncrease mor e rapi dl y than Xi , i n the rati o of the
di stances of the correspondi ng push-rod pi vots from the axi s of rotati on:
B, d,.
a=Bzdz
(21)
I f the di recti on of moti on i s r ever sed by a push exer ted on the X~ termi nal ,
AZ wi l l be hel d agai nst the stop Cz both by the l i nkage constrai nt and by
the torque due to any resi sti ng for ce at the Xi termi nal , unti l the crank
B, touches the stop Cl . At thi s poi nt the si tuati on changes abruptl y:
the crank B1 can no l onger rotate; the crank A z i s no l onger l ocked i n
posi ti on by the l i nkage constrai nt; a further push on the xk termi nal
wi l l cause the crank A z and the l i nk BZ to r otate together about thei r
now col l i near pi votal axes Se and T1. Then Xi and xk both become
negati ve, the rati o of thei r val ues bei ng
(22)
The change i n dXz/ dX, as the l i nkage i s pushed through i ts cri ti caf posi -
ti on, i n ei ther di recti on, i s qui te abrupt; i t i s associ ated wi th a si mi l arl y
sudden i ncrease i n the dri vi ng for ce necessary to over come a resi sti ng
for ce at the other termi nal , when the mechani cal advantage i s r educed
by the change i n ful crum.
The desi red di sconti nui ty i n the deri vati ve i s not so per fectl y real i zed
i f the l i nk i s pul l ed rather than pushed through i ts cri ti cal posi ti on.
the confi gurati on i s that i l l ustrated i n Fi g. 5.9, a pul l exer ted
When
on the
SEC. 5.4] PROBLEM OF DESI GNI NG THREE-BAR LI NKAGES 117
X,-termi nal and a resi sti ng pul l on the Xk-termi nal wi l l pr oduce a torque
tendi ng to move the arm AZ away from i ts stop. Thi s arm, however ,
i s l ocked i n posi ti on by the l i nkage constrai nt, and the l ocki ng wi l l be
effecti ve unti l the cri ti cal posi ti on i s approached, and mechani cal pl ay
i n the l i nkage becomes i mportant.
Thi s wi l l al l ow crank .4P to begi n to
move away from stop Cz befor e crank B1 qui te reaches stop Cl ; the
resul t i s some roundi ng off of the otherwi se abrupt transi ti on from one
sl ope to another, but ther e i s no tendency for the mechani sm to jam.
5.4. The Probl em of Desi gni ng Three-bar Li nkages. -We have now
I
to consi der the probl em of determi ni ng the el ements of a three-bar l i nkage
that wi l l mechani ze a gi ven functi on
X2 = ( 4ZJ . Z1. (23)
I f thi s functi on i s to fal l wi thi n the cl ass of l i nkage functi ons, i t must be
requi red to gener ate i t onl y for a fi ni te range Ax1 of the i nput vari abl e
x 1, or , i f the range of x 1 i s i nfi ni te, X2 must be a peri odi c functi on of xl
wi th peri od Azl . I n ei ther case, restri cti ng attenti on to the range Axl
of the i nput vari abl e, one can wri te the rel ati on i n homogeneous form:
h, = (h,lh,) . h,. (24)
To mechani ze thi s rel ati on we have to desi gn a three-bar l i nkage descri bed
by
x, = (X2JX,) . x,, X1. s xl ~
such that when homogeneous parameters I I I , Hz
correspondi ng rel ati on
Hz = (H4H,) . H,
Xl w, (25)
are i ntroduced, the
(26)
becomes i denti cal wi th Eq. (24) on di rect or compl ementary i denti fi ca-
ti on of the pai r of vari abl es (hi , hz) wi th the pai r (Hl Hz). I f the func-
ti on to be gener ated i s peri odi c, i t i s necessary, i n addi ti on, that
AXt = X,.w X,. = 360;
the i nfi ni te range of x, then corresponds to the i nfi ni te range of Xl , when
passage to the next peri od of the gener ated functi on i s permi tted.
A three-bar l i nkage may be descri bed by the constants A ~Bl, A ~,
Bz, Xlm, X,~, AX1, X2~, Xz,r, AXZ; of these onl y fi ve are i ndependent.
The form of the functi on (X2 I X,) i s determi ned by thr ee i ndependent
rati os of the si des of the quadri l ateral ; the angl es Xl and X2 do not
depend on the over-al l scal e of the mechani sm. We shal l choose the
thr ee si de-rati os, B,/ A 1, Bt/ A z, A 1/ A z, as the i ndependent constants
that determi ne the form of (X,1X,). Now, the fi el d of functi ons (X,I XI )
118 THE THREE-BAR LI NKAGE [SEC. 5.5
of the three-bar l i nkage i s three-di mensi onal , but each functi on (X21X 1)
can gener ate a whol e fi el d of functi ons (H2 I HJ that depend on the choi ce
of addi ti onal constants: two constants (for exampl e, Xl - and X 1~) i n
the case of a nonperi odi c functi on, and one (for exampl e, X2~) i n the
case of peri odi c functi ons. The fi el d of al l functi ons (HZI H,) of a thr ee-
bar l i nkage i s ther efor e fi ve-di mensi onal wher e nonperi odi c functi ons
are concer ned, and four-di mensi onal wi th r espect to peri odi c functi ons.
We shal l henceforth concentr ate our attenti on on the mor e di ffi cul t case
of nonperi odi c functi ons.
I n practi cal terms, the probl em i s that of approxi mati ng a gi ven func-
ti on (hzl h,) as wel l as possi bl e by a three-bar-l i nkage functi on (H,I H1)
characteri zed by fi ve i ndependent constants.
I t i s ver y di ffi cul t to
fi nd the best fi t by varyi ng al l fi ve constants i ndependentl y; one must
begi n by assi gni ng fi xed val ues to at l east two of them, even when choi ce
of these val ues must be made rather arbi trari l y. Fortunatel y, i n prac-
ti ce one has usual l y some i ndi cati on of an appropri ate val ue for one or
mor e of these constants.
The way i n whi ch a l i nkage i s used i n the computer as a rul e suggests
an appropri ate val ue for AX1 and AXZ..
I n parti cul ar, i n generati ng a
monotoni c functi on one can hardl y have AX1 > 180; on the other hand,
AX1 must not be chosen too smal l l est the l i nkage degener ate i nto what
i s essenti al l y a harmoni c transformer. I t i s thus evi dent that i t wi l l be
useful to have a method for fi ndi ng the best fi t to the gi ven functi on
consi stent wi th speci fi ed val ues of AX1 and AX2; the si de-rati os (or thei r
equi val ent) wi l l then be the adjustabl e parameters.
The nomographi c
method, to be descri bed i mmedi atel y, i s sui ted for thi s type of cur ve
fi tti ng. I t shoul d be used for al l monotoni c functi ons and i s useful i n
many other cases.
When the gi ven functi on i s not monotoni c, i t i s someti mes di ffi cul t to
choose AX,. The geometr i c method, to be descri bed l ater i n thi s chapter,
i s then useful . I n appl yi ng thi s method, AX2 and B,/A, are fi xed and
the fi t to the gi ven functi on i s obtai ned by adjustment of AX,, B,/ A,,
and Al /AZ.
THE NOMOGRAPHI C METHOD
The nomographi c method her e di scussed i s a method of cur ve
fi tti ng by three-bar l i nkages wi th gi ven AX I and AX2. I t takes i ts
name from the use made of an i ntersecti on nomogram, whi ch appears as
an i nsert i n the back of thi s book. Thi s nomogram, Fi g. B. 1, i s al so
useful i n many other types of cal cul ati ons on three-bar l i nkages.
5.5. Anal yti c Basi s of the Nomographi c Method.For anal yti c pur-
poses i t i s conveni ent to speci fy the si de-rati os of the quadri l ateral
through the thr ee i ndependent constants
WC. 5.5J ANALYTI C BASI S OF THE NOMOGRAPHI C METHOD 119
( )
B,
bl =l n ~,, (27)
()
b~=l n ~, (28)
()
AI
a=ln z
(29)
Correspondi ngl y, we may speci fy the confi gurati on of the quadri l ateral
i n terms of the di agonal -to-si de rati o, through one or the other of the
new vari abl e parameters
()
D
~=ln x,
(30)
()
D
= ln 12
=Pl+a, (31)
whi ch wi l l repl ace the i nput parameter XI i n our di scussi on.
I n terms of these new symbol s the equati ons of Sec. 5.1 take on a l ess
fami l i ar but ver y useful form. Si nce
D B, Bz Al
z = p 11 = b z = b X2 =
(32)
one can r ewr i te Eq. (2) as
ezpl = 1 + ezb,+ 2eb, cos Xl ,
or
(
e%+ e%
e% = 2eb,
)
+ COSXI .
2
Hence the rel ati on between the vari abl e parameters X,
cos X1=*e
271b, Cosh bl,
or
(33)
(34)
and p 1i s gi ven by
(35)
PI = + h (2 cos XI + 2 cosh h) + + bG
(36)
By si mi l ar mani pul ati ons Eqs. (3) and (4) become, respecti vel y,
cos ql = cosh pl i ezb,epl , (37)
cos 7J2= cosh PZ i e2bZeP2

cosh (pl + a) ~ezb,e(p+j. (38)

As before, si n ql and si n Xl must have thk same si gn, whi l e
o s q, s 180.
Then the output parameter i s gi ven by
X2+ = ql + ?l Z, (39)
120 THE THREE-BAR LI NKAGE [SEC. 5.6
or by
X2- = q, qz. (40)
Equati ons (36) to (40) descri be al l three-bar-l i nkage functi ons (X,1X,).
The i mportant feature of thi s formul ati on i s the expressi on of q, and q2
i n terms of the same functi on of two i ndependent vari abl es,
G(p, b) = COS- (cosh p ~ezbn); (41)
one has
ql = G(P1, h)
(42)
and
q2 = G(P2, ~2) = G(PI + a, bz).
(43)
Thi s makes i t possi bl e to compute ~~ and qz by the same i ntersecti on
nomogram, wi th other advantages that wi l l become cl ear as the di scussi on
pr oceeds.
6.6. The Nomographi c Chart. -I n three-bar-l i nkage cal cul ati ons one
repeatedl y encounters the rel ati ons
q = G(p, b) = COS-l (cosh p ~e~p) (44)
and
p = F(X,b) = ~ I n (2 cos X + 2 cosh b) + i b, (45)
wher e p stands for pl or pz = pl + a, b for bl or bz, X for Xl , and q for
71 or ~2.
I t may be requi red to sol ve these equati ons si ngl y or si mul -
taneousl y, wi th vari ous choi ces of the unknown. For rapi d cal cul ati ons
of thi s type the use of an i ntersecti on or gri d nomogram i s ver y conveni ent.
The i ntersecti on nomogram representi ng a gi ven functi onal rel ati on
i s not uni quel y determi ned, but may be gi ven an i nfi ni te vari ety of
forms. I n the present case i t i s desi rabl e to take l i nes of constant p as
verti cal l i nes, l i nes of constant ~ as hori zontal l i nes, and to pl ot on the
(P, q)-plane curved lines of constant b and constant X (Fig. 510). I t i s
at once evi dent that choi ce of consistent val ues of any two of the vari abl es
wi l l determi ne a defi ni te poi nt on the chartthe i ntersecti on of the l i nes
correspondi ng to the gi ven val ues of these vari abl es; correspondi ng val ues
of the two other vari abl es, as determi ned by Eqs. (44) and (45), can then
be read off at the same poi nt. Befor e i l l ustrati ng thi s process, however ,
we must consi der i n mor e detai l the structure and properti es of the chart.
As shown i n Fi g. 5.10, the hori zontal scal e i s uni form i n p wi th the
verti cal l i nes spaced at i nterval s of 0.1 i n 10; they are l abel ed i n terms
of the vari abl e
( )
D
pp = l ogl o )
A
(46)
for whi ch the i nterval s are 0.1. The verti cal scal e i s uni form i n q, wi th
l i nes of constant q shown i n Fi g. 5.10 at i nterval s of 30, from 180
to +180.
SEC.5.6] THE NOMOGRAPHI C CHART
121
On the gri d thus establ i shed ther e have been pl otted l i nes of constant b
at i nterval s of 0.1 i n 10; they are l abel ed i n terms of the vari abl e
/Lb = l og,, ;, (47)
for whi ch the i nteNal s are 0.1. (The factor p = I /i n 10 i s i ntroduced
i n thi s way to faci l i tate computati on wi th deci mal l ogari thms. ) The
cur ve b = O i s open, wi th the hori zontal asymptotes ~ = + 90. Curves
I
+4++tH
-6
A I 1/ r -
/1 fi r
KDu4=H+@
$zEEli
+140
+120
+100
\
-
~,
I f
*
-180~
I I I I I I I I I I I I I \
0.7 -0.6 -0.5 -0.4 -0.3-0.2-0.1 0,0+0.1 +0.2 +0,3 +0.4
111/11 I+HEF +
PP
FI G. 5.10.I ntersecti on nomogr am sol vi ng Eqs. (44) and (45).
of constant b <0 are cl osed. Curves of constant b >0 are open and
peri odi c i n q wi th peri od AV = 360; they have a pronounced si nusoi dal
character, bei ng symmetri c to refl ecti on i n the l i nes
~= . . .
180, 0, 180, . . .
and central l y symmetri c about the poi nts wher e they cross the l i nes
~= . . .
90, +90, . . . .
For a mor e detai l ed di scussi on the r eader i s r efer r ed to Appendi x B.
I
122 THE THREE-BAR LI NKAGE [SEC. 5.7
Li nes of constant Xhave been pl otted at i nterval s of 10. Al l are
open curves, and each has the same shape as apart of the cur ve b = O.
I ndeed, the curves X = XO and X = XO 180, whi ch joi n smoothl y
at p = ~ = O if O < XO < 180, together form a cur ve congruent wi th
the cur ve b = O. Al l have asymptotes paral l el to the p-axi s, but run
to i nfi ni ty toward the ri ght (p = + m), i nstead of toward the l eft
(p = m) as does the cur ve b = O. Agai n the r eader i s r efer r ed to
Appendi x B for a mor e compl ete di scussi on.
Si nce the parameters p and q have no l i mi ts, the nomogram extends
i n pri nci pl e over the whol e pl ane. I t i s peri odi c i n ~ wi th peri od 360;
the part shown h Fi g. 5.10 coul d be suppl emented by the addi ti on of
si mi l ar fi gures above and bel ow, extendi ng i ndefi ni tel y to posi ti ve and
I
negati ve q. The chart coul d al so be extended to l arger and smal l er
p, but the added porti ons woul d be of l ess practi cal i mportance si nce
ver y l arge or ver y smal l val ues of b are not much used.
I n actual wor k one does not need the whol e fi el d cover ed by Fi g. 510
but onl y i ts upper hal f, si nce the l ower hal f i s a mi rror i mage. By sup-
pressi ng the l ower hal f, l onger scal es can be used i n a gi ven avai l abl e
space. Thi s has been done i n the preparati on of Fi g. B*1, whi ch pr e-
sents thi s nomogram on the l argest scal e possi bl e i n thi s book. Thi s
fi gure i s qui te adequate for a study of the method; i n actual desi gn wor k
i t, i s desi rabl e to have i t drawn on a scal e twi ce as l arge and wi th a gr eater
nubber of curves. Tabl e B. 1 of Appendi x B presents the i nformati on
needed for redrawi ng the nomogram-the coordi nates (pp, V) of the poi nts
of i ntersecti on of the curves yb = O, +0.01, ~ 0.02, 0 o , +0.5, wi th
the curves X = O, + 5, ~ 10, s , ~ 180.
6.7. Cal cul ati on of the Functi on Gener ated by a Gi ven Three-bar
Li nkage.-The i ntersecti on nomogram permi ts sol uti on of Eqs. (36) to
(4o), whi ch compl etel y descri be any three-bar l i nkage; i t ther efor e suffi ces
for the graphi cal constructi on of any three-bar-l i nkage functi on (X,1X,).
The pr ocedur e wi l l be descri bed i n connecti on wi th i ts appl i cati on to
the speci al l i nkage sketched i n Fi g. 511, for whi ch pb, = O. 1, pbz = 0.3,
~a = 0.3. For thi s l i nkage ~ = 0.795, ~z = 1.995, ~ = 1.995; wi th
B, taken as uni ty, the l i nks have l engths BI = 0.795, Al = 1.000,
Az = 0.501.
To determi ne the val ue of the output parameter X2 correspondi ng to
a gi ven val ue of Xl i n the exampl e, 140, as i l l ustrated i n Fi g. 5-1 l
we pr oceed as fol l ows:
(1) Knowi ng Xl and b,, one can determi ne p, and q, by Eqs. (36) ~
and (37). I nstead, on the nomogram, Fi g. 5 12, we fol l ow the cur ve
X = Xl (= 140) unti l i t i ntersects wi th the cur ve pb = pbl ( = 0.1) at
SEC.5.7] FUNCTI ON GENERA TED BY THREE-BAR LI NKAGE 123
the poi nt P(o). At thi s poi nt we can read off the correspondi ng val ues
of ppl ( =
0.191) and q, (= 52.50).
(2) Knowi ng pp, and ~a, one can compute Y(p, + a)i n the exampl e
0.109. I nstead, on the nomogram we l ocate a poi nt pa uni ts to the ri ght
of P(o) (by the scal e at the bottom of the fi gure) and through thi s con-
struct the verti cal l i ne Pp = P(P1 + a).
(3) Knowi ng ~(pl + a) and pbz, one can compute m by Eq. (38).
I nstead, on the nomogram we fol l ow the verti cal l i ne ~p = p(pl + a)
unti l i t i ntersects the cur ve pb = ~bz ( = 0.3), as i t does at the two poi nts,
FI G. 5.11.Three-bar l i nkage used i n i l l ustrati ve cal cul ati ons.
Q? and Q!, wi thi n the fi el d of Fi g. 5.12, and at an i nfi ni te sequence of
poi nts outsi de thi s fi el d. I t i s at the poi nt Q(!) that q l i es between O
and r, and i t i s, ther efor e, at thi s poi nt that we can read off the val ue
of qz (= 1210).
(4) Computati on of X2 i s now si mpl e:
X,+ = q, + q, = 52.5 + 121 = 173.5,
and
X,_ = q, ~, = 52.5 121 = 68.5.
These val ues can be checked on Fi g. 511.
I t wi l l be noted that i n Fi g. 5.12 the val ue of n, i s r epr esented by a
verti cal l i ne from the l i ne v = O to the poi nt P(o), and the val ue of m i s
r epr esented by a si mi l ar l i ne to the poi nt Q$). Graphi cal methods for
addi ng these l engths can be used to construct the val ue of X2+. I I I the
same way, the verti cal l i ne from ~ = O to the poi nt Q!!) represents the
(negati ve) quanti ty whi ch must be added to TJ~to get X2-; we cal l thi s
qz-, to di sti ngui sh i t from the pri nci pal val ue, ?Z+, and wri te
X2* = 71 + m*.
(48)
124 THE THREE-BAR LI NKAGE [SEC. 5.7
We shal l often use thi s rel ati on i nstead of Eqs. (39) and (40). The
poi nt Q!) wi l l then be r egar ded as correspondi ng to the posi ti ve branch
of the sol uti on for X2, and Q! as correspondi ng to the negati ve branch.
By use of the nomogram we can get a graphi c presentati on of the
enti re cour se of the functi on gener ated by a gi ven l i nkage.
To pi cture
I I l\ I I I \l 1-
0
I I I I I I i
.
x
-60
A I f I
1
/1
I
-
-+
#P
FI G. 5.12.Cal cul ati on of the functi on gener ated by a gi ven three-bar l i nkage.
the functi on (X21X1) we may wi sh to compute Xz for a spectrum of
val ues of X1:
X\O), Xp, X\*),. . .
Xy) .
(I n Fi g. 512, Xy) = 140 T.1O, r = O, 1, ~. ~4.) Correspondi ng
to thi s sequence of val ues, ther e i s a sequence of poi nts P(o), P(]), . . ,
on the our ve pb = pbl, at whi ch we can read off the spectra of val ues of
~pl and of ql :
PP\ O))#Pil)) . . . PPi) ;
~io), w! . . . 7i).
SEC,58] COMPLETE REPRESENTATI ON BY THE NOMOGRAM 125
Verti cal l i nes from q = O to these poi nts r epr esent the spectrum of
val ues of pp 1by thei r hori zontal spaci ng, the spectrum of val ues of ~1by
thei r l engths. On shi fti ng each of these l i nes to the ri ght by an amount
pa we next obtai n l i nes representi ng, by thei r hori zontal spaci ng, the
spectrum of the vari abl e parameter ppz = ~(pl + a), whi ch can assume
the val ues
dPiO) + a), K(P!l) + ~), . P(PP + a).
Si nce thi s shi ft does not di sturb the di stri buti on of the l i nes, one can
speak of the spectrum of val ues of pp~ as congr uent to the spectrum of
val ues of pp,. The spectral l i nes for pp,, by thei r i ntersecti ons wi th the
cur ve pb = pbz, defi ne two sequences of poi nts:
Q~), Q:), . . . Q!!),
and
Q!!, Qg), . . Q!?,
from whi ch one may read off the spectral val ues of 72:
By termi nati ng the spectral l i nes of ~pz at the poi nts Q$, we can make
them r epr esent the spectral val ues of q:< by thei r upward and downward
extensi ons, just as the spectral l i nes for ~pl r epr esent the val ues of T1.
Ther e resul ts a ver y cl ear pi cture of the way i n whi ch v, and qz~ change
together wi th X,. Fi nal l y, the spectrum of val ues of the output vari abl e,
Xg, X$2, . . Xpi,
can be obtai ned by addi ng correspondi ng spectral val ues of ~1and qzf.
5.8. Compl ete Representati on of Three-bar-l i nkage Functi ons by
the Nomogram.I t wi l l soon become evi dent to the r eader who attempts
to use the nomogram that i t i s not possi bl e to car r y through for al l X,,
and for gi ven pbl, pbz, and pa, the cal cul ati on outl i ned i n Sec. 57. Thi s
l i mi tati on corresponds to restri cti ons on Xl i nherent i n the geometr y of
the l i nkage consi dered, and i s not a shortcomi ng of the nomogram; that
the nomogram gi ves a compl ete representati on of the whol e cl ass of
three-bar-l i nkage functi ons wi l l be evi dent from the fol l owi ng di scussi on.
I n the cal cul ati ons di scussed i n Sec. 5.7 i t i s conveni ent, but not neces-
sary, to sel ect val ues of Xl correspondi ng to l i nes appeari ng on the nomo-
gram. We shal l now consi der a conti nuous spectrum XT), whi ch i ncl udes
al l val ues i n the range CCI< Xl < + ~. We shal l cal l such a con-
ti nuous and i nfi ni te spectrum of Xf) the compl ete spectrum X~.
Correspondi ng to a conti nuous spectrum X~) ther e wi l l be a conti nuous
spectrum X~J. The val ues of Xf), however , as computed by Eqs. (36)
to (40), wi l l not be real for the compl ete spectrum X~) but onl y for
certai n bands of that spectrum. Real confi gurati ons of the l i nkage
126 THE THREE-BAR LI NKAGE [SEC, 58
cor r esp end, of course, onl y to real val ues of XY); thus, by obser vi ng the
l i mi ti ng val ues of X~) and X!) i n the bands i n whi ch the sol uti on i s
real , one mi ght deter mi ne the l i mi ti ng confi gurati ons of the l i nkage.
We now use the nomogr am i n studyi ng the condi ti ons for the exi stence
of a real sol uti on X;) cor r espondi ng to a gi ven val ue of X~). We note
fi rst that, as X~) goes thr ough al l val ues, ppl can go onl y thr ough a
l i mi ted r ange of val ues deter mi ned by the fi xed val ue of Pbl . Thi s cor -
responds to the l i mi tati on on the magni tude of the di agonal D, whi ch has
180
150
120
V 90
60
30
0
PPf)
FI Q. 5.13.Range of oper ati on of a three-bar l i nkage.
been expressed anal yti cal l y i n Eq. (5) and i n our present notati on can
be r ewr i tten as
l ogl o I I I obl l s pp, s I ogl o (1 + l o@l ). (49)
The range of ppl k fi ni te i f pb, # O, and extends to co when pbl = O;
we shal l speak of the val ues i n the range of ~pl as maki ng up the com-
pl ete spectrum pp~). I n Fi g. 5.13, whi ch appl i es to a l i nkage wi th
pbl = 0.2, pbt = 0.3, pa = 0.5,
i t i s cl ear that ypl can not be l ess than 0.432 (for Xl = 1800), nor
gr eater than 0.215 (for Xl = 0). We have, then, for the compl ete
spectrum, 0.432 S pp~) s 0.215. By shi fti ng the compl ete spectrum
SEC. 5.9] RESTATEMENT OF THE DESI GN PROBLEM
127
~Py) to the ri ght by a di stance pa, we obtai n the compl ete spectrum
g(p~) + a). I n our exampl e
+0.068 s p(py) + a) s 0.715.
I n the same way i t wi l l be obser ved on the nomogram that ther e i s a
l i mi ted range of val ues of ypz consi stent wi th the fi xed val ue of ~bz.
Thi s corresponds to the restri cti on on D expressed anal yti cal l y by Eq. (6),
whi ch i n our present notati on i s
l og,, \1 l o@,l s pp, s l og,, (1 + I obj).
(50)
The val ues i n thi s range make up the compl ete spectrum pp~). I n the
case i l l ustrated i n Fi g. 513, 0.002 S gpcrj s 0.476.
I n the nomographi c computati on of Xa one has to i denti fy ~p, and
/dP1 + ~).
Thi s wi l l be possi bl e onl y for val ues of PP9 whi ch l i e i n the
compl ete spectrum Ppt) and al so i n the compl ete spectrum I J (P!) + a);
such val ues make up the l i mi ted spectrum pp~). By shi fti ng the
l i mi ted spectrum ALpi) to the l eft by an amount pa, we obtai n fi nal l y the
l i mi ted spectrum pp~). The nomographi c computati on can be carri ed
through for al l val ues of ppl that l i e wi tbi n thi s l i mi ted spectrum; for
the correspondi ng val ues of Xl , the l i mi ted spectrum X~), one can com-
pute real val ues of X2. The range wi thi n whi ch thk cal cul ati on i s pos-
si bl e corresponds exactl y to the range wi thi n whi ch both Eqs. (5) and
(6) are sati sfi ed, as i l l ustrated i n Fi g. 5.2. Thus al l physi cal l y possi bl e
confi gurati ons of the l i nkage, al l real three-bar-l i nkage functi ons (X91X1),
are cover ed by the nomogram.
The r eader wi l l fi nd i t i nstructi ve to appl y the nomogram to the di s-
cussi on of the paral l el ogram l i nkage.
6.9. Restatement of the Desi gn Probl em for the Nomographi c
Method.-The nomogram i s conveni entl y used i n three-bar-l i nkage
desi gn onl y when i t i s possi bl e to preassi gn val ues for two of the desi gn
constants, AX1 and AX2. Ther e remai n thr ee desi gn constantsbl , tv,
and a, or thei r equi val entsto be adjusted i n the process of fi tti ng the
gener ated to the gi ven functi on.
When the angul ar ranges of the i nput and output vari abl es are thus
speci fi ed, i t becomes possi bl e to express the gi ven functi on i n terms of
angul ar vari abl es PI and q9, i nstead of the homogeneous vari abl es hl
and h~:
91 = Axl hl ,
I
(51)
w = AX9hz.
I n compari ng the gi ven functi on wi th the gener ated functi on, one wi l l
correspondi ngl y express the l atter i n terms of the angul ar parameters
X, and X,:
1
128 THE THREE-BAR LI NKAGE
Xl X1- = AXI Hl ,
X2 Xzm ==AX2H2.
I
The desi gn probl em can then be stated as fol l ows.
three-bar l i nkage generati ng a functi on
x, = (X,1 X,) . xl
whi ch can be i denti fi ed wi th the gi ven functi on
P2 = (92)91) P1
on di rect or compl ementary i denti fi cati on of H]
[SEC. 5.10
(52)
I t i s desi red to fi nd a
(25)
(53)
wi th hl, Hz wi th hz (cf.
Sec. 5.4). Di r ect i denti fi cati on i n the two cases i mpl i es
PI = xl Xl m,
$D2= x2 x2m;
I
(54)
compl ementary i denti fi cati on i mpl i es
AXI PI = Xl XI .,
I
(55)
Axz I pJ= x2 x2nu
The desi gn probl em i s essenti al l y the same i f the i denti fi cati on i s di rect i n
both cases or compl ementary i n both cases; i f the i denti fi cati on i s di rect
i n one case and compl ementary i n the other i t does not matter i n whi ch
case i t i s di rect. I t wi l l be conveni ent to assume that i t i s al ways di rect
i n the case of the output vari abl e. The rel ati ons to be sati sfi ed by the
angul ar parameters are then
(56)
wi th the upper si gn correspondi ng to di rect i denti fi cati on.
I t i s i mportant to note that the pr ocedur e to be descri bed does not
necessari l y l ead to a uni que sol uti on of the probl em. Ther e usual l y exi st
two qui te di fferent approxi mate sol uti ons, wi th a posi ti ve and a negati ve
val ue for b,, respecti vel y. Duri ng the desi gn process the constants
of both of these sol uti ons shoul d be determi ned suffi ci entl y accuratel y to
permi t a rati onal choi ce between them. Thi s poi nt wi l l be ful l y i l l us-
trated i n l ater secti ons.
6.10. Sur vey of the Nomographi c Method.Fi tti ng the gener ated to
the gi ven functi on by si mul taneous and i ndependent vari ati ons of the
thr ee remai ni ng desi gn constants i s hardl y practi cabl e. We ther efor e
(1) make a defi ni te choi ce of b,, and then fi nd the best fi t obtai nabl e bv
., .
i ndependent vari ati on of the other two desi gn constants; (2) fi nd a better
val ue of bl, aa descri bed i n Sec. 5.13; (3) fi nd the best fi t obtai nabl e by
vari ati on of the other desi gn constants, usi ng thi s i mproved val ue of bl;
I -- _____ ._ _ _
SEC.5.10] SURVEY OF THE NOMOGRAPHI C METHOD 129
(4) fi nd a oetter val ue of b,; and so on, approachi ng the opti mum choi ce
of al l thr ee constants by successi ve approxi mati ons.
I t woul d be qui te natural to choose bz and a as the desi gn constants
to be adjusted i n the fi rst step of thi s pr ocedur e. However , to deal wi th
1
these constants di rectl y i nvol ves, i n effect, the fi tti ng of the gi ven cur ve
to a member of a two-parameter fami l y of three-bar-l i nkage curves. I t
i s preferabl e to choose X 1~and Xz~ as the addi ti onal constants on whi ch
attenti on i s concentrated, si nce i t i s then possi bl e to wor k i nstead wi th
two one-parameter fami l i es of curves, one easi l y constructed from the
gi ven functi on, the other appeari ng on the i ntersecti on nomogram. To
make i t cl ear how thi s can be done we shal l consi der thr ee i ncreasi ng y
di ffi cul t probl ems, The di scussi on wi l l be i l l ustrated by Fi g. 5.14.
180
K \l !C II I
\ 180
1 v I
. .
30
.%
-iin-\ l I \-u 160
150
v 90
60
0
.
,, ,!,
-0.7 -0.6 -0!5 -0:4 -0.3-0,2 -0,1 0!0 +0:1+0.2 +0:3-+0~
PP
FI Q. 5.14.Curve fi tti ng i n Probl ems 1 and 2, Sec. 5.10.
Problem 1: Determine whether or not a given functian (pzlp,) is generated
by a linkage of speci $ed constants AX1, AXZ., 51, X1~, Xz~.Si nce AX~,
Xl ~, and X2* are known, i t i s possi bl e to transform the gi ven functi onal
rel ati on (PZ. IPI ) i nto a functi onal rel ati on of the angl es Xl and Xa,
(
(X21X,), by use of Eqs. (56), whether the i denti fi cati on i s di rect or com-
pl ementary. The probl em i s then: I s thi s functi on (X21X,) real l y
I
gener ated by a l i nkage characteri zed by the gi ven constants?
Let X~] be any val ue of Xl i n the speci fi ed range. On the nomogram
we l ocate at the i ntersecti on of the l i nes X = X~) and pb = pbl the poi nt
P(); at thi s poi nt we can read off the val ues of pp~) and q~) gener ated by
the l i nkage. We do not yet know the gener ated val ues of q,f and X,, but
we do know the correspondi ng gi ven val ue of X2, X~j, and can compute
the gi ven val ue of T2+usi ng Eq. (48):
7$; = x:; nf).
(57)
130 THE THREE-BAR LI NKAGE [SEC.5.10
On the verti cal l i ne through the poi nt P(J we er ect a spectrum l i ne of
hei ght vg~, endi ng at poi nt R().
By movi ng thi s spectrum l i ne to the
ri ght by an amount pa (sti l l unknown) i t can be brought i nto the posi ti on
of the spectrum l i ne pp$). I f the gi ven val ue q~~i s the one actual l y
generated, thi s spectral l i ne wi l l then extend exactl y to the cur ve pb = pbz,
(also unknown at the moment); the amount by whi ch i t fal l s short of that
cur ve i s the amount by whi ch the gener ated val ue of Xl ) exceeds the
gi ven X~~. The compl ete spectrum of such l i nes, l i mi ted by the cur ve
R(0) . .. R(.)...
R(), can be outl i ned qui ckl y.
Let the cur ve R(o) . . . R() . . . R(n) be drawn on a transparent
overl ay. Suppose now that, by movi ng the overl ay to the ri ght by a
di stance ~a, thk cur ve can be made to coi nci de wi th some cur ve pb = pbz.
I t wi l l then fol l ow that the gi ven functi on i s i ndeed gener ated by a l i nkage
wi th the speci fi ed constants, AX1, AX2, bI , Xl-, XZ~, and, furthermore,
that thi s l i nkage i s al so characteri zed by the constants bz and a thus
determi ned. For, i n vi ew of the methods of computati on outl i ned i n
Sec. 57, i t i s cl ear that a l i nkage wi th the above val ues of b,, b,, and a wi l l
gener ate a spectrum of val ues of m+ whi ch i s just the spectrum of the
gi ven
val ues, ~;~, and hence a spectrum of val ues of X2 whi ch i s al so
the gi ven spectrum Xfj, for X,. S X S Xl . + AX1. The spectrum of
Xl i s determi ned by the gi ven constants Xl and AXI ; the gener ated
spectrum of X2, whi ch r epr oduces the gi ven spectrum of X2, must
cor r espond to the constants Xz~ and AX2. Si nce the l i nkage that
generates the gi ven functi on i s characteri zed by the fi ve speci fi ed
constants and by the deri ved bz and a, the truth of the statement
fol l ows.
As an exampl e, shown i n Fi g. 514, we take a case i n whi ch the
speci fi ed constants are AX1 = 60, AXZ = 105, pbl = 0.2, XI - = 90,
x2m = 45. r he gi ven functi on (Pzl W) has been assumed to i ncrease
monotoni cal l y from 92 = 0 when w = 0 to PZ = 105 when w = 60;
t J Eq. (56) (wi th the upper si gn) X2 then i ncreases monotoni cal l y from
X, = 45 when X, = 90 to X2 = 150 when Xl = 150. Taki ng
(o>= 33; hence qj~i s 45 33 = 12,
X!) = 45, we l ocate P(o) and read q,
correspondi ng to poi nt R(). Wi th XT) = 150, we l ocate P() and read
W) =
36; hence & = 150 36 = 114, correspondi ng to poi nt
R(). We shal l assume that si mi l ar computati ons for i ntermedi ate Xl
ser ve to determi ne the cur ve R(o) . . . R() . . . R() as shown. I f now
thi s cur ve i s moved to the ri ght by an amount pa = 0.4, the end poi nts
can be brought to l i e at poi nts A and C on the same contour of constant
pb: pbz = 0.3. The i ntermedi ate porti ons of the cur ve do not then l i e on
that contour, and i t i s evi dent that no other contour can gi ve a fi t. I t
fol l ows that the gi ven functi on can not be gener ated by a l i nkage wi th the
speci fi ed constants. The di fference between the gi ven and the gener ated
SEC. 5.10] SURVEY OF THE NOMOGRAPHI C METHOD 131
functi ons i s i mmedi atel y evi dent. A- l i nkage wi th g.a = 0.4, pbz = 0.3
does gi ve a fi t at the ver y ends of the range of X, (poi nts A and C), and
thus at the ends of the range of X2. Thi s l i nkage has ther efor e the
speci fi ed val ues of AX1, AX2, XI ~, and X2~, as wel l as that of bl; i t i s the
speci fi ed l i nkage. I t generates val ues of q,t gi ven by the cur ve AABC,
i nstead of the gi ven val ues of the cur ve ABC; the verti cal separati on
between these curves i s then the di fference between the gi ven and the
gener ated val ues of X2.
Problem 2: Determine whether or not a given function (pz[pJ can be
gener ated by a linkage of spec-ijied constants AX1, AX2, bl, Xl ~.I t i s now
possi bl e to consi der al l val ues of XZ~ i n seeki ng a fi t of the gener ated to
the gi ven curve, i nstead of onl y one val ue. Let the cur ve R(o) . . . R(n)
be constructed as before, for an arbi trari l y chosen val ue of Xz~for
exampl e, X~~. I f a fi t can not be found for thi s among the curves of
constant ~b on the nomogram, one wi l l desi re to make a si mi l ar tri al for
another val ue of X2* for exampl e,
x;. = X;m + A. (58)
By Eq. (56), thi s i ncrease i n Xz~ wi l l pr oduce a uni form i ncrease, by
the same amount, i n the gi ven val ues X2a and qzo; the new cur ve
R(o) . . .
R@j wi l l be the ol d one rai sed by an amount A, and the fi t wi l l
be sought as before. Of course, i nstead of redrawi ng the curve, one can
si mpl y shi ft upward by A the overl ay on whi ch the fi rst cur ve was drawn.
Thus by al l owi ng al l verti cal shi fts of the overl ay i n seeki ng a fi t one
treats X2~ as a di sposabl e parameter.
The stated probl em can then be sol ved as fol l ows: On a transparent
overl ay draw a cur ve R(o) . . . R() . . . R(n), assumi ng X2- = Xi ~.
I f, by transl ati ng the overl ay to the ri ght by an amount pa and upward
by an amount A (as read on the scal e of n), thi s cur ve can be made to
coi nci de wi th some porti on of the cur ve pb = pbz, then the gi ven functi on
(wl qJ can be gener ated by a l i nkage wi th the speci fi ed constants AX I ,
AX2, b,, and Xl ~. Furthermore, that l i nkage wi l l al so be characteri zed
bY the constants bz, a, and Xz~ = X~~ + A, determi ned in this fi tti ng
process.
I n the exampl e of Fi g. 5.14 a fi t can be obtai ned by movi ng the over -
l ay, prepared as previ ousl y descri bed, upward by an amount A= 57,
and to the ri ght by pa = 0.255; cur ve R(o) . . . R() then l i es on the
cur ve pb = pbz = 0.3, extendi ng from A to C. Thus the gi ven cur ve i s
actual l y gener ated by a l i nkage wi th the constants
AX1 = 60, AX2 = 105, ~bl = 0.2, Xl m = 90,
and, as now determi ned, pbz = 0.3, pa = 0.255, X~~ = 45 + 57 = 102.
I f the fi t wer e not exact the di l Terence between the gener ated and the
132 THE THREE-BAR LI NKAGE [SEC. 511
gi ven functi ons coul d be read asthe verti cal separati on of the cur ve
R(o) . . . R~nJ andthecontourpb = pbs.
Problem 3: Determine whether or not a given funci%on (WZ[WJ can be
generated bya linkage of spemfied constants AXl, AXa, bl .Both Xl ~and
X2~ ar enowto r etr eated as dkposabl e constants; the probl em i s then
essenti al l y the same as that encounter ed i n Step (1) of the desi gn pr o-
cedur e descri bed at the begi nni ng of thi s secti on.
We have al ready seen how a ti t can be sought for a gi ven val ue of
X,~for exampl e, X~~by a process that begi ns wi th the constructi on
of a correspondi ng cur ve-for exampl e, RjO) . . . R~). To make the
same test for another val ue of Xl ~for exampl e X~~one woul d si mi l arl y
construct another curve, R$O) . . . R~). Unfortunatel y, thi s i s not
of the same form as the fi rst curve; the actual constructi on of thi s cur ve
i s not to be avoi ded, though i t can be made rel ati vel y easy by methods
to be descri bed.
The probl em i s then to be sol ved as fol l ows. On a transparent over -
l ay, construct a fami l y of curves R(o) . . . R@j for suffi ci entl y cl osel y
spaced val ues of X 1~, and for X2~ = O; l abel each cur ve wi th the cor r e-
spondi ng val ue of Xl ~. Now suppose that, by transl ati ng the overl ay
to the ri ght by an amount pa and upward by an amount X2. (as read
on the scal e of q), the cur ve of thi s fami l y l abel ed Xl ~ can be brought i nto
coi nci dence wi th a part of the cur ve pb = ybz on the nomogram. Then
the gi ven functi on can be gener ated by a l i nkage wi th the gi ven con-
stants AX1, AX2, and bl; thi s l i nkage woul d be characteri zed al so by the
constants Xl ~, XZ~, pa, @Z thus determi ned.
The essenti al features of the nomographi c method shoul d now be
evi dent to the reader. To fi nd the three-bar l i nkage wi th gi ven AX1,
AXZ, b1, whi ch most accuratel y generates a gi ven functi on (wI PI ), one
constructs on an overl ay a fami l y of curves correspondi ng to Xz = Oand
to vari ous val ues of Xl ~. Movi ng the overl ay over the nomogratn, one
seeks the best possi bl e fi t of a cur ve of thi s fami l y to a cur ve of the ybz
fami l y on the nomogram. The di spl acement of the overl ay necessary
to pr oduce thk fi t determi nes X2~ and pa for the l i nkage; the choi ce of
curves for thi s fi t determi nes Xl - and pbz. The er r or i n the resul tant
mechani zati on i s di rectl y evi dent i n the fai l ure to obtai n an exact fi t
between the overl ay and nomogram curves, and i s measured by thei r
verti cal separati on. The steps i nvol ved i n thi s process wi l l be di scussed
i n detai l i n Sees. 5.11 and 5.12. After the method of i mprovi ng the
choi ce of bl has been descri bed i n Sec. 5 13, the whol e pr ocedur e wi l l be
ful l y i l l ustrated i n Sec. 5.14.
5.11. Adjustment of b2 and a, for Fi xed AX1, AX2, bl .We shal l now
descri be i n ful l detai l a practi cal pr ocedur e for the constructi on of the
- ---- -
I
I
SEC. 5.111
ADJ USTMENT OF B%AND A 133
overl ay menti oned i n Sec. 5 10 and i ts use i n determi ni ng the best val ues
of b~and a.
Constru.ct~on oj the Overlay.
(1) Choose a spectrum of val ues of XI ,
whi ch fi l l s the enti re range from Oto 360 at i nterval s 3 smal l compared to
AX1. One shoul d choose ~ as the di fference i n X between consecuti ve
curves on the nomogram, or a mul ti pl e of thi s, so that ther e wi l l be on the
chart a cur ve correspondi ng to each val ue X!). Usual l y ~ = 10 i s
suffi ci entl y smal l ; J = 5 i s possi bl e wi th the chart pl otted from Tabl e
B.1.
(2) As the spectrum of val ues of pl , take
# = m
((33)
wi th r= 0,1, . ..n. Si nce these val ues shoul d fi l l the range AX 1, one
must have nh = AXI .
(3) Compute the correspondi ng spectrum of ~z:
# = ($4$0,) d. (61)
Usi ng the same scal e as the ?-scal e of the nomogram, construct thi s
spectrum as a seri es of ti ny hol es al ong a strai ght l i ne on a stri p of paper
(see Fi g. 5.15). On thi s stri p mark the i ndex r for each of the l i nes of the
spectrum; i ndi cate by an ar r ow the di recti on of i ncreasi ng pt.
(4) Fasten over the nomogram the materi al on whi ch the overl ay i s
to be constructedfor i nstance, a pi ece of traci ng paper. Copy onto
the overl ay al l the poi nts P@J at whi ch the contour pb = @1 i s i nter-
sected by the l i nes X = X~. (Fi gure 515 shows the compl ete contour ,)
Mark the poi nts P(J on the over l ay wi th the subscri pt s. Al SO COPY
onto the over l ay the l i nes q = O, P = O. Thi s posi ti on of the over l ay
wi l l be cal l ed i ts starti ng posi ti on.
(5) Dr aw on the over l ay the verti cal l i nes pp = ~p(j through al l
poi nts P(). These are the spectral l i nes for the vari abl e pp. The Over -
l ay can now be separated fr om the nomogr am.
(6) Pl ace the stri p of paper carryi ng the spectrum q;) on the over l ay,
al ong each l i ne of the spectrum Pp(a), maki ng the ar r ow poi nt downward
and the fi rst poi nt p~o)of thi s spectrum coi nci de wi th the poi nt P~). For
each such posi ti on of the stri p, mark on the l i ne pp = @sJ of the overl ay
the posi ti ons of the poi nts & on the stri p, l abel i ng each wi th the cor r e-
spondi ng val ue of r. These poi nts we shal l i ndi cate as
Py)Py . . . P$@ . . . P#.
(7) Starti ng at each poi nt P(n) on the overl ay, pass a cur ve succes-
si vel y through the poi nts P(), P~+l), P~+2),. . . P:+), . . . P$+).
134 THE THREE-BAR LI NKAGE [SEC. 5.11
Thi s fami l y of curves weshal l cal l thepl usfarni l y. I ti s unnecessary to
use a French cur ve i n thi s constructi on; i t i s suffi ci ent to connect the
poi nts by hand wi th strai ght l i nes, i n or der to make cl ear the way i n
Fm. 5.15.Scal e andoverl ayfor6rst appl i cati on of the nomogr aphi c method,
whi ch they are associ ated. The pl us fami l y of curves appears as con-
ti nuous l i nes i n Fi g. 5.15.
(8) Agai n, starti ng wi th each poi nt P@J on the overl ay, pass a cur ve
successi ve y through the poi nts P(a), P~ll, Pg2)
. . p$a-r) . . p;n).
Thi s fami l y of curves (dashed l i nes i n Fi g. 5.15} we shal l c~l l the mi nus
fami l y.
Thi s compl etes the constructi on of the overl ay. As wi l l appear from
our l ater exampl e, constructi on of the compl ete overl ay i s not al ways
necessary.
We must now exami ne the si gni fi cance of the curves thus drawn.
(1) The successi ve poi nts P@, P(+), . . . P(a+n), represent, by thei r
di stances from the l i nes p = O and q = O, the val ues of pp and q, for a
sequence of val ues of XI : X~), X~+l I , . . . By Eq. (54) these poi nts al so
cor r espond to the sequence of val ues of PI : O, Pjl ) . . . P$), for the case
SEC.5-11] ADJ USTMENT OF B, AND A
135
i n whi ch XI ~ = Xr) and the i denti fi cati on of hl and HI i s di rect. I n
parti cul ar, P(+) represents the val ues of pp() and q!) when Xl ~ = X~).
(2) The separati on of the poi nts on the stri p correspondi ng to qj and
w represents, on the same scal e, the val ue of # ~!) = &). BY
~q. (54) thi s i s al so the val ue of X~j i f X,- = O, and the i denti fi cati on of
h, and H, i s di rect.
(3) The poi nt ~~+) thus corresponds to pp = pp() and has as i ts v
coordi nate [by Eq. (57)]
(r)_ x;; =
VI N1
for the case i n whi ch X1~ = Xy), x~~ = O, and the i denti fi cati on of h,
and HI i s di rect.
(4) We thus see that the cur ve of the pl us fami l y passed through the
poi nts P(), P~+l), . . . 5+), . . .
P~~J on the overl ay corresponds to
the sequences of val ues
xl : Xy), Xy+l ), . . . Xy+), . . . Xy+),
PP: / J P(0),/ J P(l), . . . PP(), o . . PP(-),
7: 0) M, . . . d% . . . d% 72L7?
for the case descri bed under (3).
(5) The si gn of q can be r ever sed by rotati ng the fi gure through 180
about the axi s q = O. Thus if the overlay as constructed ~s turned jace
down by rotating d through 180 about the axis T = O, then the cur ve of the
pl us fami l y l abel ed wi th ans gi ves the rel ati on between m. and ~p for the
case i n whi ch Xl ~ = XY), Xzm = O, and the i denti fi cati on of hl and HI i s
di rect.
(6) I n the same way the r eader wi l l be abl e to show that, with the over-
lay turned jace down as above, the cur ve of the mi nus fami l y l abel ed wi th
ans gi ves the rel ati on between m~and pp f or the case i n whi ch X l U = X~),
X2. = O, and the i denti fi cati on of h, and I l l i s compl ementary.
Cse oj the Overl ay.Wi th the overl ay face down on the nomogram
and the l i ne T = O hori zontal , a fi t i s sought between any cur ve on the
1
overl ay and a l i ne pb = pbz of the nomogram. I f a fi t i s found, the
constants of the l i nkage are determi ned as fol l ows:
(1) AX,, AX,, Kbl have been previ ousl y chosen.
(2) pb, i s read from the cur ve of the nomogram wi th whi ch the fi t i s
found.
(3) pa i s the shi ft of the overl ay to the ri ght needed to establ i sh the
fi t. I t may be read at the i ntersecti on of the l i ne P = O of the overl ay
wi th the pp-scal e on the nomogram.
136 THE THEEE-BAR LI NKAGES
[Sm. 5.12
I
I
(4) Xa~ i s the shi ft of the overl ay upward needed to establ i sh the fi t.
I t maybe read at the i ntersecti on of the l i ne q = Oof the overl ay wi th the
q-scal e of the nomogram.
(5) I f the fi t i s establ i shed i n the pl us fami l y of curves, one has ,
X,. = s~, s bei ng read from the overl ay cur ve wi th whi ch the fi t i s made.
The angl es w and Xl wi l l i ncrease together ; thi s, of course, i s o.Zwagstr ue
of pz and X2.
(6) I f the fi t i s establ i shed i n the rui nus fami l y of curves, one has
Xl ~ = s6, Xl ~ = S6 AXI . The angl e WI wi l l decr ease as Xl i ncreases.
The l i nkage and the associ ated scal es are then compl etel y determi ned.
The actual l y gener ated val ues of TZwi l l not be q~~,si nce these wer e
computed on the assumpti on that X2. = O; i nstead, they wi l l be
whi ch can be read di rectl y on the nomogram scal e. I f the fi t i s estab-
l i shed on the upper hal f of the nomogram, qz i s gr eater than zer o and the
gener ated functi on bel ongs to the posi ti ve branch; i f the fi t i s establ i shed
on the l ower hal f of the nomogram one has to do wi th a negati ve branch.
Usual l y the fi t obtai ned between the nomogram and overl ay curves
wi l l be onl y approxi mate. The constants determi ned al one wi l l then not
al l be mutual l y consi stent, unl ess the approxi mate fi t i s so made that the
er r or vani shes at the extr eme val ues of X 1 and X2. Thi s i s easi l y done
when monotoni c functi ons are bei ng deal t wi th; i n other cases one shoul d
remember that when fi ve of the desi gn constants have been fi xed the
others must be determi ned by appropri ate cal cul ati ons rather than read
as above.
5.12. Al ternati ve Methods for Overl ay Constructi on.-Modi fi cati ons
of the above pr ocedur e are necessary when use i s made of a nomogram
l i ke Fi g. B.1, whi ch i ncl udes onl y the range from q = 0 to q = 180.
The mi ssi ng porti ons of thi s chart can be constructed as mi rror i mages
of the part shown; or , mor e conveni entl y, the same effect can be obtai ned
by appropri atel y turni ng the overl ay.
For exampl e: To construct the poi nts P() on the overl ay, copy the
poi nts P(o) to F(16J(assumi ng 6 = 10) from the nomogram, and draw the
r efer ence l i nes. Then turn the overl ay face down by rotati ng i t about
the l i ne q = O, and copy the same poi nts onto the overl ay. The poi nts
thus constructed are i n fact P(o), P(lJ , Ptz), . . . P(is), and shoul d be
so l abel ed.
I n the curve-fi tti ng process descri bed above, wi th the overl ay face
down, the l ower part of the nomogram wi l l be mi ssi ng, and di rect fi tti ng
to functi ons of the negati ve branch wi l l not be possi bl e. One can, how-
ever , turn the overl ay agai n (so that .i t i s now face up) and seek a fi t on
I
I
SEC. 5.13] CHOI CE OF BEST VALUE OF b, 137
the upper part of the nomogram. I t must, of course, be remembered that
readi ngs made on the q-scal e (for i nstance, readi ngs of XZ~) must then be
taken wi th a mi nus si gn.
When pb, >0, an overl ay constmcted as descri bed above becomes
excessi vel y l arge; another modi fi cati on i n the overl ay constructi on then
becomes conveni ent. I t wi l l be noted that i f the l ower hal f of an overl ay
i s turned ab~ut the l i ne q = O, the poi nt P$a) will be brought i nto
coi nci dence wi th the poi nt Pf, and the poi nt P$) wi l l l i e as far above
g)l i es bel ow i t. We shal l speak of these poi nts i n thei r new posi - P~) as P,
ti on as the transferred poi nts; they extend through. the transferred
regi on of the overl ay. These transferred poi nts can be constructed
di rectl y by the method descri bed above wi th the one change that, i n
l ocati ng the transferred poi nts ~~s), ~j), . . . &, one pl aces the
poi nt pjo) of the spectrum stri p on the poi nt P$) wi th the ar r ow di r ected
upward befor e copyi ng off the successi on of poi nts Pjo), p$l ), . . . q$].
I n worki ng wi th thi s transferred regi on one must remember that i t i s
equi val ent to a normal regi on turned face down. When fi tti ng cur ves i n
a normal regi on, one turns the overl ay face down and reads val ues
di rectl y from the q-scal e of the nomogram; when fi tti ng cur ves i n the
transferred regi on, one uses the overl ay face up. The pl us and mi nus
fami l i es of curves i n the transferred regi on are most readi l y i denti fi ed by
turni ng over the overl ay.
5.13. Choi ce of Best Val ue of bl for Gi ven AX,, AXz.I n the pr eced-
i ng secti ons we have seen how to fi nd the el ements of a three-bar l i nkage
whi ch gi ves an approxi mate mechani zati on of a gi ven rel ati on,
P2 = (WIP1) p], (53)
when AX 1, AX2, and fibI are speci fi ed i n advance. We have now to con-
si der the probl em of fi ndi ng an appropri ate val ue for bI when onl y AX I
and AX2 are speci fi ed.
A method of tri al and er r or i s obvi ousl y appl i cabl e, One can car r y
through the above process for an arbi trari l y chosen ~b,; i f an acceptabl e
fi t i s not found another val ue can be chosen for ~bl and the process
r epeated, unti l a good fi t i s found or the useful r ange of pbI has been
cover ed. Fortunatel y i t i s necessary to tr y onl y a rel ati vel y smal l
number of val ues, such as @l = 0.5, 0.2, 0.0, 0.2, 0.5, to determi ne
roughl y the val ue of pb I or to establ i sh that the pr oposed type of mecha-
ni zati on i s not appropri ate.
Such a process of r epeated tri al s can be abandoned as soon as even a
poor approxi mate fi t i s found between the overl ay and nomogram curves.
Usual l y one fi nds at l east a ver y rough fi t wi th the fi rst chosen val ue of
pbl, and can begi n to appl y a second methodone of successi ve approxi -
138 THE THREE-BAR LI NKAGE
[SEC. 5.13
mati ons. Let the l i nkage that gi ves the 6rat r ough fi t be characteri zed
by the constants
Axl , AX2, Pb!l ), @\l ), P@, (62)
of whi ch the l ast two have been found by the process al ready descri bed.
Now l et us consi der the probl em of si mi l arl y mechani zi ng the i nverted
functi on
w = (PI I P2) pa, (63)
wi th PZpl ayi ng the r ol e of the i nput parameter, I P1the r ol e of the ouput
parameter. The parameters P1and P2wi l l then be i nterchanged through-
out the previ ous di scussi on, P1 varyi ng wi th the angl e Xz, w wi th the
angl e Xl . The l i nkage that mechani zes thi s rel ati on wi l l be the same as
that whi ch mechani zes the ori gi nal rel ati on, Eq. (54), except that i nput
and output are i nterchanged. I f Fi g. 5.1 represents the l i nkage for Eq.
(54), the l i nkage for Eq. (63) can be obtai ned from thi s by mi rrori ng i t i n
a verti cal l i ne, al ong wi th the associ ated scal es for pl and 92. Thi s new
l i nkage di ffers from the ol d i n that BI and A z are i nterchanged, as are
AX I and AX2; X, i s repl aced by 180 XW, XZ~ by 180 Xl ~. As
for the constants ~bl, pbz, pa, we note that i nterchange of B1 and At
carries
Al
w = 10gO X2
i nto
l og,o .* = l ogl o : = l d)l ,
and conversel y, whi l e
(64)
(65)
(66)
becomes
10f 4%) l O+ww) lb2-a-pblo67)
Di sti ngui shi ng the constants of the i nverted l i nkage by a ti l de, we may
wri te
I .LG= pbl, (68a)
pbl = pa, (68b)
L& = pbs pa pbl, (68C)
~, = AX,, (68d)
and so on.
I n attempti ng to mechani ze Eq. (63) one mi ght appl y the nomographi c
SEC.5.14] AN EXAMPLE OF THE NOMOGRAPHI C METHOD 139
method as before, choosi ng arbi trari l y a val ue of & and fi ndi ng cor r e-
spondi ng val ues of y;z and @. However , ~afl j i s a known fi rst approxi -
mati on to the desi red val ue of p%>,and an appropri ate choi ce for the fi xed
val ue of thi s quanti ty. We ther efor e take
-(2) = _ ~a(l),
pb~ (69)
and by the nomographi c method determi ne the correspondi ng constants
-(z) i n the mechani zati on of the i nverted probl em. Thi s mecha-
pb~z)pa
ni zati on of the rel ati on between w and w must be at l east as good as
that descri bed by the constants i n Eq. (62), si nce i t i s chosen as the best
of a f arni l y of l i nkages whi ch i ncl udes the mi rror i mage of that fi rst
l i nkage; usual l y i t i s much better. From these constants one can then
obtai n second approxi mati ons to the constants requi red for the di rect
probl em:
pa
(2) = _ #&:) = #a(l )
>
pb;z) = @c2), (70)
~bjz) = pb\2) pa(z) ~b;z).
The val ues of pbl and ~bz have been i mproved; the val ue of pa was fr ozen
i n passi ng to the i nverted probl em, and i s hence unchanged.
We can now return to a consi derati on of the probl em as fi rst formu-
l ated. I t i s obvi ousl y desi rabl e to take @(2) = pb~z) as the chosen
val ue of pbl; repeti ti on of the curve-fi tti ng process l eads to a sti l l better
mechani zati on of the rel ati on between PI and P2, characteri zed by the
constants
pb\ 3)(= @t2)), pb!f), pa(3J.
Thus by al ternatel y consi deri ng the probl em as formul ated i n Eqs.
(53) and (63), and appl yi ng the methods of Sees. 5.11 and 5.12, one
obtai ns successi vel y better approxi mate sol uti ons, whi ch usual l y con-
ver ge rapi dl y to a l i mi t. The method i s l ess l abori ous than mi ght at fi rst
be supposed, si nce the constants to be expected i n al l sol uti ons but the
fi rst are known approxi matel y, and the compl ete overl ay need not be
constructed.
I t wi l l be found that i f one obtai ns a fai r fi t wi th a gi ven pbl, one wi l l
obtai n al so a reasonabl y good fi t wi th ~bl. On appl i cati on of the
method of successi ve approxi mati ons, these two approxi mate sol uti ons
usual l y l ead to two di fferent sol uti ons of the probl em, whi ch are equi val ent
nei ther wi th r espect to the resi dual er r or , nor wi th r espect to mechani cal
qual i ti es. These two possi bi l i ti es shoul d r ecei ve separate consi derati on.
6.14. An Exampl e of the Nomographi c Method.-As a fi rst exampl e
of the nomographi c method, we shal l appl y i t i n attempti ng to mechani ze
the functi on presented i n Tabl e 5 1 i n both di rect and i nverted forms.
140 THE THREE-BAR LI NKAGE [Szc. 5.14
TABLE 5.1.GI vEN FUNCTI ONSFORTEE EXAMPLZ
(P21W1)
w,
degr ees
0.0
22,3
34.1
43.6
52.1
60.0
68.3
75.9
83.1
90.0
n,
degr ees
o
10
20
30
40
50
60
70
60
90
(wl l fP.)
n,
degr ees
0.0
3,4
9.0
16,9
26.6
37.7
50.0
62,6
76.1
90.0
W3,
degr ees
o
10
20
30
40
50
60
70
80
90
Thi s tabul ated functi on i s i n fact the one gener ated by athree-bti
l i nkage wi th the fol l owi ng constants:
X,m = 170, AXI = 90, Xzm = 16&, AXZ = 90,
pbl = O, pa = 0.286, ~bz = 0.0367,
B,
I
(71)
=1, = 1.932,
B,
x T &
= 2.102.
I n appl yi ng the nomographi c method we shal l assume the i deal val ues for
the angul ar travel s,
AX1 = AX2 = 90,
(72)
but shal l begi n by choosi ng a val ue of pb, whi ch i s not the best:
pbjlJ = 0.1.
(73)
I n thi s way we can make parti cul arl y evi dent the conver gence toward the
best constants that i s usual l y afforded by the method of successi ve
approxi mati ons. A second and qui te di fferent exampl e, wi thout thi s
ad hoc character, wi l l be found i n Sec. 6.4.
Fol l owi ng the steps outl i ned i n Sec. 5.11, we pr oceed thus:
(l ), (2) I n tabul ati ng the gi ven functi on, ~ = 10 has been chosen as
suffi ci entl y smal l compared to the ranges of Xl and X2; thi s wi l l permi t
use of Fi g. B-1 (fol di ng i nsert i n back of book) i n appl yi ng the method.
I n mechani zi ng the functi on i n the di rect form the spectrum of val ues of
pl i s 0, 10, . . . 90. We have her e n = 9.
(3) The val ues of q:) appear i n the fi rst col umn of Tabl e 5.1. Usi ng
the q-scal e of the nomogram, we transfer thi s spectrum of val ues to a stri p
SEC. 5.14] AN EXAMPLE OF THE NOMOGRAPHI C METHOD 141
of paper, as i l l ustrated i n Fi g. 5 15. The di recti on of i ncreasi ng ~ i s
shown by an arrow.
(4) Traci ng paper i s used i n maki ng an overl ay. Thi s i s taped to the
nomographi c chart, whi ch shoul d be made on cardboard, and 36 poi nts,
from Pc-laj to P(+lJ , are constructed and marked wi th the pr oper val ue
of s (Fi g. 5.15). The r efer ence l i nes are traced onto the overl ay.
(5) The verti cal l i nes of the spectrum Pp(s) are omi tted from Fi g. 5.15
for the sake of cl ari ty.
(6) Pl aci ng the zer o poi nt of the stri p successi vel y on each of the
poi nts P(), wi th the ar r ow di r ected downwards, the 36 poi nts P~) are
l ocated and l abel ed wi th thei r r-val ues. (I n fi rst approxi mati ons one can
someti mes ski p hal f the val ues of r and hal f the val ues of s.)
(7), (8) The pl us fami l y of curves i s now sketched (ful l l i nes i n Fi g.
5.15) through poi nts wi th r-val ues successi vel y i ncreasi ng by 1 as s
i ncreases; curves of the mi nus fami l y (dashed i n Fi g. 5.15) pass through
poi nts wi th r-val ues successi vel y i ncreasi ng by 1 as s decreases. The
compl ete fami l y of curves i s shown i n the fi gure. Thi s i s real l y unneces-
sary, si nce one can tel l at a gl ance that some of them cannot l ead to a fi t.
I n parti cul ar, si nce q~) i s a si ngl e-val ued functi on of s, one coul d her e
omi t the numerous curves that have i nfi ni ti es i n thei r sl opes.
The overl ay i s now turned about the hori zontal r efer ence l i ne and
transl ated over the nomogram unti l a fi t i s founda qui te sati sfactory
fi t, as i t happens, between the overl ay cur ve s = 16 of the pl us fami l Y
and the cur ve pbz = 0.075 on the nomogram. Fi gure 5.16 shows, on the
nomogram gri d, the constructi on of the parti cul ar overl ay cur ve for whi ch
the fi t was obtai ned, and the posi ti on of fi t on the chart (dotted cur ve at
l ower l eft). The fi t has been made exact at the ends. The overl ay cur ve
then devi ates downward from the nomogram curve; on a l arge chart i t
can be seen that the maxi mum er r or i n q i s a l i ttl e mor e than one degr ee.
The r efer ence l i nes on the overl ay are al so shown i n the posi ti on of fi t.
The el ements of the l i nkage are thus establ i shed:
pb~l )= 0.1, as assumed.
~~i u = 0.075, read from the nomogram cur ve on whi ch the fi t was
made.
~a(o =
0.265, read at the i ntersecti on of the verti cal r efer ence l i ne
wi th the pp-scal e.
x~~ =
202.5 or + 157.5, read at the i ntersecti on of the hori zontal
r efer ence l i ne wi th the q-scale. (When thi s r efer ence l i ne fal l s
off the nomogram, as i t woul d her e, an auxi l i ary r efer ence l i ne
on the overl ay can be used.)
X~~ = 160 = s~, si nce the cur ve that gi ves the fi t i s of the pl us
fami l y.
142 THE THREE-BAR LI NKAGE [SEC.5.14
By Eq. (56) we have (usi ng the upper si gns i n the fi rst equati on, si nce the
fi t was obtai ned wi th a cur ve of the pl us fami l y)
w = Xl + 160,
pa = x2 157.5. (74)
These l ast equati ons r epr esent the gi ven functi on wi th er r or s vi si bl e as the
verti cal separati on of the fi tti ng curves i n Fi g. 5 16. Si nce the fi t i s exact
+30
0
-30
-60
~
-90
-120
-130
-1s0
-210
w
FI G.5.16.Constructi on of over l ay l i ne and posi ti on of fi t i n fi st appl i cati on of the
nomogr aphi c method. I n the posi ti on of fi t the over l ay cur ve l i es sl i ghtl y bel ow the
contour @J = 0.075.
at the ends of the range of Xl , the travel s of both i nput and output of the
l i nkage as desi gned wi l l have the requi red val ue, 90. (I t i s not necessary
to make the fi t exact at the ends, except perhaps i n the l ast stage of the
desi gn process. I n earl i er stages one can often accel erate conver gence
on the i deal constants by seeki ng a good fi t on the average rather than an
exact fi t at any gi ven poi nts i n the range.)
As a check i t i s useful to make a drawi ng of the l i nkage, showi ng the
cranks i n thei r extr eme posi ti ons (Fi g. 5.17). The di stance A fl ) between
the crank pi vots may be taken as the uni t of l ength; the rel ati ve crank
l engths are drawn i n as
(75)
SEC. 5.14] AN EXAMPLE OF THE NOMOGRAPHI C METHOD 143
The constancy of the requi red l ength of the connecti ng l i nk,
B~ ~~1) ~~1)
Ajl ) = ~~
= l og(~z(l )-ufl ))= l oO.a10= 2188,
(76)
provi des a check on the quanti ti es determi ned i n the fi tti ng process.
Of the constants thus determi ned for the l i nkage, b, has been hel d at a
preassi gned val ue, but the others have taken on val ues that are good
approxi mati ons to those known to gi ve an exact fi t. Such behavi or
i s of cour se essenti al i f the method of successi ve approxi mati ons (Sec.
5.13) i s to be effecti ve. We now appl y thi s method to the i mprovement
of the l i nkage desi gn.
Fm.
\
\
\
\
\
\
\
\
\
\ i
NJ
5.17.Fi rst approxi mate l i nkage for mechani zati on of the gi ven functi on, Tabl e 5.1.
The rol es of w and qs are to be i nterchanged throughout our next
treatment of the probl em. The i nverted functi on has al ready been
gi ven i n Col umns 3 and 4 of Tabl e 5.1. Now ~ i s to be associ ated wi th
the parameter Xl of a new l i nkage; to remi nd us that thi s i s a parameter
of the i nverted probl em we shal l di sti ngui sh i t by a ti l de: XI. Si mi l arl y
pl i s to be associ ated wi th the parameter ~z i n the new l i nkage.
Accordi ng to Eq. (69) we shoul d begi n the process of mechani zi ng the
i nverted functi on by choosi ng ph~z)= 0.265. To faci l i tate constructi on
of the overl ay we shal l use an approxi mati on to thi s:
Yi )~2)= 0.25. (77)
Such roundi ng off of val ues i s general l y useful i n practi cal desi gn wor k;
we have her e del i beratel y done i t i n such a way as to retard rather than
accel erate the conver gence of the method.
We know that the l i nkage to be desi gned wi l l not be ver y di fferent, i n
i ts di mensi ons and i n the arrangement of the scal es, from that of Fi g.
144 THE THREE-BAR LI NKAGE [SEC.5.14
5.17. I t must, however , cl i ffer from that l i nkage by refl ecti on i n a
verti cal l i ne, si nce the pi vots are to be i nterchanged; and i t may cl i ffer
al so by refl ecti on i n a hori zontal l i ne. One can determi ne whether or not
thi s addi ti onal refl ecti on i s i nvol ved by exami ni ng the ql -scal e, whi ch, by
the conventi on i ntroduced i n the di scussi on l eadi ng up to Eq. (56), must
i ncrease i n the di recti on of i ncreasi ng ~z. I n the present case i t i s
\
I
\
~fib=Q.25
1
\ .
\
o
\
~
u
2
.3
.4
.5
.6
.7
.0
.9
evi dent that the two l i nkages must
di ffer al so by a refl ecti on i n a hori zontal
l i ne; the appearance of the new l i nkage
wi l l then be that of Fi g. 5.17 turned
upsi de ~own. One must accordi ngl y
expect X 1~ = 20, and on the overl ay
wi l l need to construct onl y a few curves
of the pl us fami l y wi th s = 2.
Fi gure 5.18 shows the nomogram
cur ve pb = 0.25 used i n constructi on of
the overl ay, and the few l i nes of the
pl us fami l y that need to be drawn. I t
i s obvi ous that a good fi t can not be
obtai ned wi th curves for whi ch q i s not
a si ngl e-val ued functi on of p. The
cur ve s = 2, for whi ch we expect thi s
fi t, i s, however , essenti al l y si ngl e-val ued.
The r etr ogr ade porti on of thi s cur ve
,.
cl osel y overl aps the rest of i t and i s no
FI Q. 5.18.ScaJe and over l ay fol
second appl i cati on of the nomogr aphi c
bar to an accurate fi t; i ts ~r esence i ndi -
method.
. .
cates or dy that q2 may r ever se i ts
di recti on of change as the l i nkage operates.
When the over l ay i s tur ned about a hori zontal l i ne and moved over
the nomogr am a ver y accurate fi t can be found between the over l ay l i ne
s = 2 and the l i ne pbz = 0.275 of the nomogram, i n the posi ti on i ndi -
cated i n Fi g. 5.19. Al so shown are the usual hori zontal r efer ence l i ne
and the auxi l i ary r efer ence l i ne Pp = 0.3, whi ch appears al so i n Fi g. 5 18.
From the posi ti on of fi t we read the fol l owi ng val ues of the constants:
(z) = 0,25,
~bl as assumed,
(4 = 1).275,
~bz
J @) = 0.011,
t
(78)
jf~~ = 50
x!: = _@.
Si nce the fi t was obtai ned wi th a cur ve of the pl us fami l y, we have
@=xl +20
~1=x2_ 5.
}
(79]
I
I
SEC. 5.14] AN EXAMPLE OF THE NOMOGRAPHI C METHOD 145
To make mor e evi dent the change i n constants due to thi s second
cal cul ati on, we can r ewr i te the above resul ts i n terms of the constants of
the uni nvested l i nkage. Rememberi ng that the t\vo l i nkages di ffer, i n
thi s case, by refl ecti ons i n both hori zontal and verti cal l i nes, we have
x, =x, 180, 2, = X, 180,
X,m = X,m + 180, X,m = ~,m + 180.
90
60
v
30
0
-30
0:2 0:3 0;4 0:5 0:6
PP
FI G. 5.19.Posi ti on of fi t i n second appl i cati on of the nomogr aphi c
By use of these rel ati ons and of Eqs. (68), we have
jtby) = 0.011,
JLby = 0.014,
pa<21= 0.25,
x~~ = _ 1750,
I
(81)
x~~ = l~oo,
B~
= 0.9975,
A;2)
= 1.778,
~
A\ 2)
~ A~2)
= 1.837.
1
These quanti ti es r epr esent di sti nct i mprovements over the fi rst
approxi mate val ues, except for ya (2J(whi ch was rounded off i n the wrong
di recti on and not al l owed to i mprove duri ng the second fi tti ng) and the
rati os A .J A I and Bz/A 1 (whi ch depend upon paj). I n parti cul ar, the
val ue of ~bl devi ates from the i deal by onl y one-tenth as much as the val ue
i ni ti al l y assumed. I t i s evi dent that a second appl i cati on of the nomo-
graphi c method to the mechani zati on of the gi ven functi on i n the di rect
form, wi th pbl = 0.011, woul d l ead to val ues of the constants ver y near
to the i deal .
THE GEOMETRI C METHOD FOR THREE-BAR LI NKAGE DESI GN
We shal l now di scuss a geometr i c method for the desi gn of three-bar
I
l i nkages for whi ch the i nput travel AX1 i s not fi xed but may be tr eated as
L________
.. .
(80a)
(80b)
method.
. ....=
146 THE THREE-BAR LI NKAGE [SEC. 5.15
a vari abl e parameter. Thi s i sal ess common probl em than that sol ved
by the nomographi c method, i n whi ch both i nput and output travel s are
tr eated as fi xed ; neverthel ess, themethod i sanecessary and frequentl y
useful compl ement to the nomographi c method.
The basi c probl em tr eated by the geometr i c method i s that of fi ndi ng
the three-bar l i nkage wi th gi ven val ues of AX2 and BZ/A 9 whi ch most
accuratel y generates a gi ven functi on. I n essence, the method i s one
by whi ch a rapi d compari son can be made between the desi red and the
actual posi ti ons of the i nput crank, for a seri es of posi ti ons of the output
crank, for any gi ven l i nkage of a l arge fami l y. Thi s compari son i s made
so easy that i t becomes a rel ati vel y si mpl e matter to fi nd that l i nkage
of the gi ven fami l y whi ch gi ves the best fi t. Thi s sol uti on can be
i mproved, i f desi red, by a method of successi ve approxi mati ons l i ke that
empl oyed wi th the nomographi c method: the val ues of Bz/B1 and AX1
determi ned by the fi rst appl i cati on of the pr ocedur e are tr eated as fi xed,
and the i ni ti al l y chosen val ues of Bz/A 2 and AX2 i mproved by a second
appl i cati on of thi s pr ocedur e to the i nverted functi on; then BZ/Bi and
AXI are readjusted, and so on. When thi s method i s empl oyed, no
constant of the l i nkage i s hel d at an arbi trari l y fr ozen val ue.
5.16. Statement of the Probl em for the Geometr i c Method.-The
probl em to be sol ved by the geometr i c method i s that of mechani zi ng a
gi ven functi onal rel ati on,
(82)
as accuratel y as possi bl e by a three-bar l i nkage wi th gi ven output travel
AX2 and gi ven crank-l i nk rati o B2/A,.
A l i nkage wi l l gener ate a rel ati on
x, = (X,1 X,) . x, (83)
between i ts i nput and output parameters; i t wi l l consti tute a mechani za-
ti on of the gi ven functi on i f ther e exi sts a l i near rel ati on between the
parameters Xl , X2 and the vari abl es m, X2:
xl Xp = kl (m Z\J)), (84a)
X2 Xf) = IC4Z2 Xy). (84b)
Her e Xi o) and Zi are correspondi ng val ues of Xl and xl , X$ and Z!
correspondi ng val ues of X2 and X2; Zjo)and z~o)do not stand i n any neces-
sary rel ati on to the l i mi ts of the i nterval of defi ni ti on i n Eq. (82).
I n the probl em at hand one knows both
AXZ = XZ~ Xz., (85)
and
AX2 = x2df X2. (86)
SEC.5.16] SOLUTI ON OF A SI MPLI FI ED PROBLEM
The magni tude of .k, i s thus determi ned:
147
(87)
Al so, i t wi l l be noted that a posi ti ve si gn of kt i mpl i es di rect i denti fi cati on
of the homogeneous parameters hz and Hz correspondi ng to X2 and Xt; a
negati ve si gn i mpl i es compl ementary i denti fi cati on. As i n Sec. 5.9 we
can, wi thout l oss of general i ty i n the desi gn process, assume di rect
i denti fi cati on of hz and Hz, whi l e admi tti ng ei ther di rect or compl e-
mentary i denti fi cati on of hl and HI . Thus kz may be consi dered as
compl etel y known,
kz = ~, (88)
but ki i s unknown both as to magni tude and si gn. The fi xed parameters
of the probl em are thus Bz/ A a and kz; attenti on wi l l be focused, i n the
actual desi gn process, on the adjustment of Al/ A z, B1/ A Z, and k 1.
5.16. Sol uti on of a Si mpl i fi ed Probl em.As i n the case of the nomo-
graphi c method, we fi rst consi der a si mpl i fi ed probl em i n whi ch ther e are
onl y two adjustabl e parameters. Her e we shal l treat Bz/AtJ k,, and k,
as fi xed, and seek the best possi bl e fi t of the gener ated to the gi ven func-
ti on by adjusti ng A l/ A* and BJ A z. We r eser ve for Sec. 5.17 an expl ana-
ti on of the method for varyi ng kl.
To sol ve thi s probl em we choose a spectrum of val ues of the vari abl e
xl :
2$), l +), . . . ,Zy), ...,
&),
extendi ng through the i nterval of defi ni ti on of Eq. (82). Equati on (82)
then defi nes a correspondi ng spectrum of val ues of X.Z:
Z\O), Xp, . . . , x:), ...,
,@ .
Si nce both kl and k2 are known, Eq. (84) woul d defi ne correspondi ng
spectra of Xl and X2, i f ther e wer e not present the unknown addi ti ve
constants X~O)and X~O). Gi ven val ues of these constants, one coul d
compute X~) and X:), and make sketches showi ng, for each T, the cor r e-
spondi ng posi ti ons of the i nput and output cranks, each i n i ts cor r ect
rel ati on to i ts own zer o posi ti on. Now, even though X~O)and X~O)are
unknown, one can sti l l compute such quanti ti es as
One can thus make a sketch showi ng the rel ati ve posi ti ons that the i nput
crank must have for a sequence of val ues of T, and a sketch showi ng the
148 THE THREE-BAR LI NKAGE [SEC. 5.16
correspondi ng relat~we posi ti ons that the output crank must have, i f the
gi ven functi on i s to be generated. Fi gure 5.20 shows such a set of rel ati ve
posi ti ons for the i nput crank, r epr esented by the radi al l i nes B~J from
thepi votpoi nt S1. Theori entati on ofthi sfi gure wi th r espect tothe zer o
posi ti on of Xl -or, to put i t another way, the di recti on on thi s fi gure of
the l i ne ~ between the crank pi vots of the l i nkage-i s unknown, si nce
i t depends on X\o). Si mi l arl y, Fi g. 521 represents, by the radi al l i nes
A:) from the pi vot poi nt S,, the correspondi ng rel ati ve posi ti ons of the
F-7x-
FI Q. 5.20.The radi al l i nes r epr esent a seri es of rel ati ve posi ti ons of the i nput crank of a
three-bar l i nkage.
output crank; i n thi s fi gure, too, the di recti on of the l i ne ~ between
the pi vots of the l i nkage cannot be speci fi ed, si nce i t depends on Xj).
Fi gures 520 and 521 can be combi ned i nto a si ngl e fi gure representi ng
a sequence of correspondi ng crank posi ti ons i n the desi red l i nkage, by
pl aci ng them i n pr oper rel ati ve posi ti ons. What the requi red rel ati on.
shi p of these fi gures shoul d be we do not yet know, but we do know
enough about i ts characteri sti cs to hel p us i n fi ndi ng i t. For i t i s evi dent
that (1) i f the crank l engths A z and I l l are l ai d out on the same scale, and
(2) i f the rel ati ve posi ti ons of the two fi gures are cor r ect, and (3) i f the
gi ven functi on can actual l y be gener ated by a l i nkage wi th the gi ven
BJ A ~, kl, and kg, then the di stances between the ends of the cranks, i n
al l correspondi ng posi ti ons, must be constant, and i ndeed equal to B2
SEC.5.16]
SOLUTI ON OF A SI MPLI FI ED PROBLEM
149
on the chosen scal e. By appl yi ng thi s i dea one can determi ne the rel a-
ti ve posi ti ons of F@. 5.20 and 5.21 whi ch cor r espond to that three-bar
l i nkage (wi th the gi ven constants) whi ch most nearl y generates the gi ven
functi on; from the combi ned fi gure one can then read off the constants
of thi s l i nkage. To understand how thi s can be done we consi der Fi gs.
5.20 and 521 i n mor e detai l .
The l ength Az of the output crank has been taken as the uni t ~,
l ength i n both Fi gs. 5.20 and 5.21. I n Fi g. 5.21, the poi nts P(o), . . . ~,
FI O. 5.21.The radi al l i nes r epr esent a seri es of rel ati ve posi ti ons of the output crank
of a three-bar l i nkage: the ci rcl es r epr esent correspOndi ns pOssi bl epOsi ti OnsfOr the r ~Ote
end of the connecti ng l i nk.
p(r), . . . , P(-I , rePre~ent a sequence of posi ti ons of the pi vot T2 between
the output crank and the connecti ng l i nk. I n Fi g. 520, one cannot con-
struct correspondi ng defi ni te posi ti ons for the pi vot T1 si nce the crank
l ength B1 i s unknown; i nstead, ther e i s shown a sequence of ci rcl es of
di fferent radi i , each of whi ch defi nes, by i ts i ntersecti ons wi th the radi al
l i nes, correspondi ng posi ti ons 62(0), . . . , Q(r), . . . , Q(n), of thi s pi vot
when the i nput crank has the appropri ate l ength.
I n Fi g. 5.21 ther e has been constructed about each poi nt P() a ci rcl e
c(,) hati ng as i ts radi us the known length 1?, of the l i nk; the r emote end
of the l i nk, the pi vot Tl , must l i e somewher e on thi s ci rcl e.
I f i t i s
possi bl e to gener ate the gi ven functi on by a three-bar l i nkage wi th the
150 THE THREE-BAR LI NKAGES [SEC. 5.16
gi ven constants, i t must now bepossi bl e to pl ace Fi g. 521 on Fi g. 5%3i n
such a way that poi nt Q(o) l i es on the ci rcl e C(o), poi nt Q(l ) on ci rcl e C(l ),
and so on, as shown i n Fi g. 5.22. The val ue of A I in the requi red l i nkage
wi l l then be the l ength of SW!%on the common scal e of the fi gures; the
val ue of B1 wi l l be the radi us of the ci rcl e on whi ch the poi nts Q(o), . . . ,
Q(), . . , Q(n) l i e; and successi ve confi gurati ons of the l i nkage wi l l be
defi ned by the poi nts S1, S2, Z(o), Q(o); S1, S2, P(l ), Q(); etc.

F1~. 5.22.Rel ati ve posi ti ons of Fi gs. 5.20 and 5.21 cor r espondi ng to a thre+bar l i nkage
gener ati ng the gi ven functi on.
I n practi cal terms, the geometri cal method of sol vi ng our restri cted
pr obl em may be summari zed thus:
1.
2.
3.
4.
5.
6.
Choose a spectrum of zI .
Compute the spectral val ues of G, Xl Xi o), X2 XjO).
Construct Fi g. 5.21 as a chart, on a suffi ci entl y l arge scal e.
Construct Fi g. 5.20 as a transparent overl ay, on the same scal e.
Move thi s overl ay freel y, usi ng both transl ati on and rotati on, over
the chart, seeki ng a posi ti on such that the ci rcl es (Y), C(), . . . ,
C(n) pass through the poi nts Q(J), Q(l ), . . . , Q(m), on some ci rcl e
of the overl ay. (I n maki ng thi s fi t i t may be necessary to consi der
each overl ay ci rcl e i n turn.)
I f a fi t i s found, the unknown constants of the l i nk can be read off,
A l/ Az as the di stance S,S2, I ll/ A, as the di stance SI Q(0),. and
X1*, X,~ as the correspondi ng angl es i n the combi ned fi gure.
SEC.6.17J SOLUTI ON OF TEE BASI C PROBLEM 151
7. I f onl y an approxi mate fi t i s found, the er r or i n the i nput angl e, for
any gi ven val ue of the gener ated output angl e, can be read as the
angl e subtended at & by the arc from the correspondi ng Q poi nt
(for exampl e, Q()) to the i ntersecti on of the correspondi ng C ci rcl e
(C(T))&th the ~r c -.
Thus one shoti d seek a posi ti on of the
overl ay whi ch makes these er r or s as smaU as possi bl e, and deter -
mi ne the constants of the l i nkage as above.
I t wi l l be evi dent to the r eader that a change i n si gn of k~ wi l l l eave
Fi g. 5.21 unchanged, but wi l l pr oduce the same effect on Fi g. 5.20 as
turni ng the overl ay face down. A ti ngl e overl ay, used face up or face
down, thus suffi ces for a gi ven I kl j.
6-17. Sol uti on of the Basi c Probl em.We now turn to the basi c
probl em of the geometr i c method, that of obtai ni ng the best fi t of the
gener ated to the given functi on by si mul taneous variations of thr ee
parameters of the linkage, keepi ng fi xed the val ues of BJ A * and k,. l%is
can be accompl i shed wi thout any essenti al compl i cati on of the pr ocedur e
descri bed i n Sec. 5.16, by maki ng a speci al choi ce of the spectrum of XI .
Thi s has al so the advantage that the overl ay correspondi ng to Fi g. 5.20
then has the same form for al l probl ems and can be used agai n and agai n.
Let the spectrum of val ues Xr) be chosen as
Xf =
gl *
Zi o) +=. , T=o, l ,-. .,n, (90)
where 8 and 9 are constants such that al l val ues XT)l i e wi thi n the range of
defi ni ti on of Eq. (82). Equati on (89a) then becomes
(91)
The separati ons of consecuti ve spectral val ues X~J, the angl es between
successi ve posi ti ons of the i nput crank, wi l l then change i n geometri cal
progressi on. I ?i ,gure 5.20 has, i n fact, been drawn for such a case.
So l ong as k, i s unknown, one cannot construct an overl ay l i ke Fi g.
5.20. To over come thi s di ffi cul ty we construct an overl ay, Fi g. 5.23, on
whi ch appear radi al l i nes L(t) wi th separati ons
y(t+l) _ y(f) = ~gt,
t=o, 1,2, .. (92)
(I n pri nci pl e, the sequence of ts mi ght start wi th other val ues than O;
such cases can be r educed to the above by changi ng the choi ce of a and
1,
renumberi ng the l i nes.) Let us consi der the n + 1 l i nes of thi s system
I
l abel ed t = s,s+l , .-, ,s+ r,... ,s+n, wi th separati ons
y(sir+l) - yb+r) = ~ga . @-
(93)
I
I
I
I
152 THE THREE-BAR LI NKAGE {$EC. 5.17
These wi l l have thesame separati ons as, andcanbe i denti fi ed wi th, the
l i nes B~m,11~), . . . , B~), . . . , EYI J,provi ded
k,a = aga (94)
or
(95)
Thus by i denti fyi ng vari ous l i nes L(u) of Fi g. 5.23 as the l i ne B~O),one can
i n effect assi gn to kl any val ue gi ven by Eq. (95) for an i ntegral s. The
FI G. 5.23.Overl ay for the geometr i c method.
overl ay i s compl eted by the system of concentri c ci rcl es whi ch appea
al so i n Fi g. 5.20; i t i s used i n the same way as that fi gure.
The pr ocedur e i s then as fol l ows:
ws
(1) Choose a spectrum of val ues z~)) as gi ven by Eq. (90). I t i s
usual l y sati sfactory to take g = 1.1; 6 may be posi ti ve or negati ve and
shoul d be so chosen that n, defi ned by
(96)
I
,
SEC. 5.17] SOLUTI ON OF THE BASI C PROBLEM 153
l i es i n the range between 8 and 12. I t i s advantageous to choose the
si gn of 8, and the correspondi ng val ue of x~, so as to make the spectrum
of val ues x!) as evenl y spaced as possi bl e. Thus i n the case i l l ustrated
t
i n Fi g. 5.24a, i n whi ch dxZ/dxl decreases as xl i ncreases, i t i s desi rabl e to
choose Zjo) at the l ower end of the range of z I and to make 6 posi ti ve;
when dxz/dzl i ncreases as xl i ncreases, as i n Fi g. 5.24b, z(o) shoul d l i e
at the upper end of the range of xl and 6 shoul d be negati ve.
(2) Compute the correspondi ng spectral val ues of X2 and X2 X~Ol ,
usi ng Eqs. (82) and (84b).
I
&>g
6<0
(a) (b)
FI G. 5.24.Choi ce of z,(o) and 6 to make the spectral val ues z, @J as evenl y spaced as
possi bl e.
(3) Construct a transparent overl ay si mi l ar to Fi g. 5.23, wi th succes-
si ve radi al l i nes at angl es
(97)
measured cl ockwi se from the zer o l i ne. The val ue of g must be the same
as that chosen i n Step (1); a may be chosen arbi trari l y but shoul d be
smal l . Fi gure 5.23 has been drawn wi th g = 1.1, a = 1. Label each
radi al l i ne wi th the correspondi ng val ue of t.
(4) Usi ng the spectral val ues of X2 XL), construct a chart cor r e-
spondi ng to Fi g. 5.21. The l ength of the crank, A z, shoul d be one uni t
on the scal e used i n constructi ng the overl ay. Lay down the successi ve
crank posi ti ons, A ~), and about the end poi nts P() construct ci rcl es C(r)
wi th the known radi us Be.
(5) Pl ace the overl ay on thi s chart, face up, and seek a posi ti on for i t
such that the (n + 1) ci rcl es, Cc), C(l ), C@J, on the chart pass through
(n + 1) poi nts Q(a), Q(S+), . . . Q@+) on the overl ay i n whi ch a ci rcl e
of the concentri c fami l y, l abel ed B1, i ntersects n + 1 consecuti ve radi al
l i nes, L(), L(+l ), . . . L(a+n).
154 THE THREE-BAR LI NKAGE [SEC.5-18
I n seeki ng thi s fi t one has to consi der:
a. Al l possi bl e posi ti ons on the chart of the poi nt S1 of the overl ay.
b. Al l ori entati ons of the overl ayi .e., al l val ues of s.
c. Al l ci rcl es of the concentri c fami l y.
The probl em i s not as di ffi cul t as i t mi ght seem. Let the poi nt S1 of the
overl ay be pl aced i n a fi xed posi ti on on the chart. Each ci rcl e of the
overl ay wi l f be i ntersected by the C ci rcl es i n a sequence of poi nts whi ch
wi l l be unchanged by rotati on of the overl ay. Unl ess successi ve i nterval s
between these poi nts change i n geometr i c progressi on, by a factor g, ther e
i s no possi bi l i tyy of obtai ni ng a fi t by turni ng the overl ay. Thus, for each
posi ti on of the overl ay center S,, a qui ck i nspecti on of the spaci ngs of the
i ntersecti ons of the two fami l i es of ci rcl es wi l l suffi ce to determi ne whether
ther e i s any chance of a fi t on any ci rcl e of the overl ay. By a systemati c
survey of thi s type one can r eject l arge areas of the chart as possi bl e
posi ti ons for S1.
When a sequence of i ntersecti ons has been found i n whi ch the i nterval s
change i n about the ri ght way, i t becomes wor th whi l e to turn the overl ay
unti l the radi al l i nes i n the regi on of i ntersecti on have si mi l ar spaci ngs
for exampl e, unti l ans i s found such that ci rcl es CCjand C@j pass through
the poi nts Q(s) and Q@+), respecti vel y. Thi s confi gurati on wi l l cor r e-
spond to a l i nkage i n whi ch the er r or s i n the gener ated functi on woul d
vani sh at the ends of the range of xl ; the er r or s i n the gener ated functi on
i n the i ntermedi ate range are evi dent, bei ng measured by the angul ar
di stances on the overl ay between the poi nts Q(S*) and the i ntersecti ons
of the ci rcl es C() wi th the B 1 overl ay ci rcl e. Wi th practi ce one rapi dl y
devel ops a techni que for i mprovi ng thi s fi t by smal l er adjustments i n the
posi ti on of S,, wi th correspondi ng rotati ons of the overl ay.
(6) I f an acceptabl e fi t i s not found wi th the overl ay face up, turn
the overl ay face down, and r epeat the process.
(7) When a fi t has been found, the el ements of the l i nkage can be
read di rectl y on the overl ay scal e; B1/A j i s the val ue of B for the overl ay
ci rcl e on whi ch the fi t i s obtai ned, and AI / -4 z i s the val ue of B for the
overl ay ci rcl e that passes through the poi nt LSZon the chart. Li mi ti ng
confi gurati ons of the l i nkage are evi dent from the arrangement, and
val ues of Xl ~, Xz~, and Xl ~ can be read.
Fi gure 522 actual l y represents an appl i cati on of thi s method, si nce
Fi g. 5-21 i s, i n fact, the porti on of Fi g. 5.23 i n whi ch s changes from 22
to 30. A ful l exampl e of the method i s presented i n Sec. 5.19.
5.18. I mprovement of the Sol uti on by Successi ve Approxi mati ons.
A fi rst sol uti on of the probl em of mechani zi ng a gi ven functi on can be
i mproved by successi ve appl i cati ons of the geometr i c method, i n essen-
ti al l y the same way as wi th the nomographi c method.
SEC. 5.18]
SOLUTI ON BY SUCCESSI VE APPROXI MATI ONS
155
The fi rst approxi mate sol uti on wi l l have been found wi th fi xed val ues
of the constants AX2 and B2/A 2.
The fi rst of these constants may be
determi ned by other factors i n the probl em, but the choi ce of the second
wi l l have been to some degr ee an arbi trary one. I f the choi ce of Bs/A Z
was ver y unfortunate, the fi t obtai ned may be so bad that the process
must be r epeated wi th another val ue of thi s constant. I n most cases one
wi l l fi nd a reasonabl y good mechani zati on of the functi onone whi ch i s
at l east suffi ci entl y good to ser ve as a gui de i n fi ndi ng a better one. I n
parti cul ar, note shoul d be taken of the val ues found for the constants
B2/B1 and AX1 of thi s l i nkage.
Now l et us consi der the i nverse of the functi on of Eq. (82),
*1 = (X11X2) X2,
(98)
wi th X2 tr eated as the i nput vari abl e.
I nterchangi ng the rol es of z, and
Xz i n Sees. 5.16 and 5.17, one can appl y the geometr i c method to the
mechani zati on of thi s rel ati on and thus obtai n a second mechani zati on
of the ori gi nal rel ati on. The i nverted probl em di ffers from the ori gi nal
i n the i nterchange of B1 and Az, Xl and 180 X2 (cf. Sec. 513). Thus
i t i s evi dent that appropri ate choi ces for the fi xed constants of the new
probl ems are
(99)
I f the condi ti ons of the probl em di ctate a speci al choi ce of AX2, one
shoul d treat A~j2) = AX!) al so as a constant; the probl em i s then that
di scussed i n Sec. 5.16. [I t can, of course, be tr eated by the method of
Sec. 5.17, wi th s restri cted to a constant val u~ determi ned by Eq. (94)
or Eq. (95)]. I n other cases one wi l l treat AX1 as a vari abl e parameter
i n the i nverted probl em. I n any case the i nverted functi on wi l l be
approxi mated by a l i nkage sel ected from a fami l y whi ch i ncl udes the
mi rror i mage of the ori gi nal l i nkage; the fi t, i f pr oper l y made, must be at
l east as good as that found as a fi rst approxi mati on, and wi l l usual l y be
appreci abl y better.
A thkd approxi mati on can then be found by returni ng to the consi dera-
ti on of the ~ni nverted functi on and appl yi ng the geom~tri c method wi th
the fi xed constants
(100)
\
As a rul e thi s process conver ges toward a certai n opti mum sol uti on of
the probl em. I t i s to be noted, however , that ther e maybe several such
. .-
156 THE THREE-BAR L.I NKAGE [SEC.5.19
approxi mate sol uti ons wi thi n the cl ass of three-bar l i nkages; whi ch of
these i s found wi l l depend upon the i ni ti al choi ce of Bt/Az and AXZ.
When one fi nds a mechani cal l y unsati sfactory sol uti on of the probl em, i t
i s usual l y profi tabl e to start the process agai n wi th a di fferent val ue of
B,/A*.
I n appl yi ng the geometr i c method i t wi l l be found that the val ues of
AX1 and AXZ conver ge mor e rapi dl y to a l i mi t than do the rati os of the
si des of the quadri l ateral . I t i s ther efor e suggested that thi s method be
abandoned as soon as the val ues of AX1 and AXZ are suffi ci entl y wel l
determi ned, the cal cul ati on bei ng compl eted by the nomographi c method.
6.19. An Appl i cati on of the Geometi c Method: Mechani zati on of the
Logari thmi c Functi on.-We shal l now appl y the geometr i c method to the
mechani zati on of the l ogari thmi c functi on
X2 = l ogl o Z1 (101)
i n the r ange
1< ZI <1O, O<z, <l . (102)
I n terms of the homogeneous vari abl es
xl l
h, = ~, (103)
h~ = x2, (104)
the rel ati on to be mechani zed becomes
h, = l og,, (9h, + 1). (105)
Si nce the l ogari thmi c functi on i s of the type i l l ustrated i n Fi g. 5.24a,
we shal l choose a posi ti ve ~. The spectrum of val ues of the homogeneous
vari abl e hl can then be wri tten as
hf) = (),
. . . . . . . . . . .
hy).=.~
~1 k
. . . . . . . . . . .
h~)=gn l.&
g1
(106)
We shal l choose g = 1.1, n = 10. Sol uti on of the l ast of Eqs. (106),
wi th hf) = 1, gi ves
8 = 0.0627. (107)
The val ues of h~)can then be computed by Eq. (106), and the cor r espond-
i ng val ues of h:) by Eq. (105). The resul ti ng val ues arc shown i n
Tabl e 52,
I
SEC. 5.19] MECHANI ZATI ON OF THE LOGARI THMI C FUNCTI ON
157
TAZLE 5,2.SPECTRAL VALUES FOR TEE LO~AEI TEMI CRELATI ON
r
o
1
2
3
4
5
6
7
8
9
10
h~)
O.0000
0.0627
0,1318
0.2077
0.2912
0.3830
0.4841
0,5952
0,7175
0.8520
1.0000
h?)
0,0000
0.1943
0.3397
0.4578
0.5588
0.6481
0.7289
0.8032
0.8726
0.9379
1.0000
I f we express thi s rel ati on i n the i nverted form, treati ng ZZor hz as the
i nput vari abl e, the functi on i s of the type shown i n Fi g. 524b. I n
mechani zi ng thi s by the geometr i c method the spectral val ues of hz
shoul d be chosen wi th 6 negati ve.
Di sti ngui shi ng by a ti l de the spectral
val ues requi red i n thi s i nverse mechani zati on, we have
(108)
. . . . . . . . . . . . . .
@)=l+c_
6=0.
g1
)
Wi th g = 1.1, n = 10, as above, one fi nds, on sol vi ng the l ast of these
equati ons, the same magni tude as befor e for 8:
6 = 0.0627. (109)
Thus
~y) = 1 _ ~y);
(110)
the correspondi ng val ues of @, computed by Eq. 105), are shown i n
Tabl e 5.3.
We begi n mechani zati on of the rel ati on i n the di rect form by choosi ng
arbi trari l y
B~l )
~ = 1.25; AX~l ) = 100. (111)
The overl ay requi red for the wor k i s determi ned as soon as g and an
arbi trary smal l angl e a are chosen; wi th a = 1 i t has the form shown i n
Fi g. 523. The chart to be constructed depends, however , on the
1
5
8
I

H
E
T
H
R
E
E
-
B
A
R
L
I
N
K
A
G
E
[
S
E
C
,5
.
1
9
SEC.5.19] MECHANI ZATI ON OF THE LOGARI THMI C FUNCTI ON 159
TABLE5.3.SPECTRAL VALUESFOETHELOGARI THMI C RELATI ONI N I NVERSEFORM
9
0
1
2
3
4
5
6
7
8
9
10
T(7)
I
i [r]
\ 55 x L:),
,,:
I .0000
0.9373
0.8682
0.7923
0.7088
0.6170
0.5159
0.4048
0.2825
0.1430
0.000o
parti cul ar probl em her e consi dered.
,,
degr ees
1.0000 55.0
0.8506 46.8
0,7092 39.0
0.5777 31.8
0.4572 25.1
0.3489 19.2
0.2533 13.9
0.1711 9.4
0.1018 5.6
0.0451 2.5
0.0000 0.0
On thi s chart (cf. Fi g. 5.25) the
l i nes A~) radi ate from the poi nt S2, maki ng angl es h~JAXZ = h~)l OOO
wi th the zer o l i ne. The poi nts P{) lie on these l i nes at uni t di stance from
S2. About each of these i s drawn a ci rcl e C@J wi th radi us
B,
= Bz = 1.25.
A,
Thi s compl etes preparati on of the equi pment. The overl ay i s now
pl aced face up on the chart, and i t i s found (as shown i n Fi g. 5.25) to be
possi bl e to make the ci rcl es C() pass, approxi matel y, through the poi nts
Q(o) . . .
Q(0) at whi ch the i nterpol ated ci rcl e B = 0.95 on the overl ay
(dashed ci rcl e i n Fi g. 525) i ntersects the radi al l i nes L() to D). The
fi t, however , i s rather poor at the poi nts Q() Q@J, Q(). I n addi ti on, the
I i nkage woul d be mechani cal l y unsati sfactory because of the smal l angl es
between the output crank and the l i nk at smal l r, and between the i nput
crank and the l i nk at l arger. (The extr eme confi gurati ons are i ndi cated
by dashed and heavy sol i d l i nes i n Fi g. 525.) The fi t coul d be i mproved
by the method of Sec. 518, but the approxi mate sol uti on thus found
woul d probabl y have the same unsati sfactory mechani cal characteri sti cs.
No sati sfactory fi t can be obtai ned by turni ng the overl ay face down.
We ther efor e r epeat the process wi th another choi ce of B,/Az.
We now tr y
B&l )
~)
= 1.8, AX~l ) = 100. (112)
The overl ay i s unchanged, and the chart i s changed onl y i n that the
ci rcl es C() have the l arger radi us Bz = 1.8. The same chart can thus be
used agai n, wi th the new ci rcl es drawn i n i nk of another col or. A mor e
sati sfactory fi t can now be obtai ned, thi s ti me wi th the overl ay face
1
6
0
T
H
E
T
H
R
E
E
-
B
A
R
L
I
N
K
A
G
E
[
S
E
C
.5
.
1
9
8Ec. 519] MECHANI ZATI ON OF THE LOGARI THMI C FUNCTI ON
161
down (Fi g. 526); the ci rcl es C() to C(O) pass ver y nearl y through the
poi nts Q() to Q(L), at whi ch the overl ay ci rcl e B = 1.2 i ntersects the
radi al l i nes Lts~ to L(). From thi s fi gure one reads the constants of
the l i nkage:
gJn
B\ ] )
= 1.8; ~ = 1.2;
Ajl )
Ajl )
= 1.79.
A(J)
(113)
2
Hence
(114)
The angl e AX1 can be measured on the overl ay, but i s even mor e
easi l y obtai ned as the di fference of tabul ated val ues of 1{~1:
AxI = Y@*) Y(a). (115)
These val ues are gi ven, for the overl ay Fi g. 5.23, i n Tabl e 5.4. I n the
present case
AX, = Y() Y() = 89.54 34.52 = 55.02. (116)
TABLE5.4.Y(~),FORg = 1.1, a = 1
t
Y(t),
t
Y(t),
t
Y(t),
t
Y(t),
degr ees degr ees degr ees degr ees
o 10.00 10 25.94 20 67.27 30 174,49
1 11.00 11 28.53 21 74.00 31 191.94
2 12.10 12 31.38 22 81.40 32 211.14
3 13.31 13 34.52 23 89.54 33 232.25
4 14.64 14 37.97 24 98.50 34 255.48
5 16.11 15 41.77 25 108.35 35 281.02
6 17.72 16 45.96 26 119,18 36 309.13
7 19.49 17 50.54 27 131.10 37
340.04
8 21.44 18 55.60 28 144.21 38 374.04
9 23.58 19 61,16 29 158,63
Al though the l i nkage thus obtai ned i s not mechani cal l y sati sfactory
when r i s smal l (xl and X2 near thei r l ower l i mi ts), we attempt to i mpr ove
i t by appl i cati on of the geometr i c method to the i nver ted functi on, wi th
~~) Bjl)
_
= 1.5
~ B~l )
(l 17a)
and
Aj@ = AX~l J = 550.
(l 17b)
A compl etel y new chart must be constructed, wi th radi al l i nes A~) maki ng
angl es @A~2 = %~).55 wi th the zer o l i ne; the requi red val ues wi l l be
found i n Tabl e 5.3. (I t must be remembered that i n thi s i nverse probl em
~, and ~z vary together , as do ht and l %,. I n the pr ocedur e, %~) now
1
6
2
T
H
E
T
H
R
E
E
-
B
A
R
L
I
N
K
A
G
E
[
S
E
C
.5
.
1
9
SEC.5.19] MECHANI ZATI ON OF THE LOGARI THMI C FUNCTI ON 163
takes the pl ace of hf); the val ues of ~~~ have been so chosen that the
di fferences i ncrease i n geometri cal progressi on and cor r espond to succes-
si ve l i nes on the overl ay. ) The poi nts P() are constructed at radi us
~~a = 1, and about these are drawn ci rcl es C() wi th radi us ~~) = 1.5
(Fi g. 5.27). When the overl ay i s pl aced on thi s chart, face up, the
ci rcl es (W can be made to pass ver y nearl y through the poi nts Q(r) at
whi ch the ci rcl e B = 0.75 i ntersects the radi al l i nes L( I s) to L (Ze); on a
l arger scal e i t can be seen that the fi t i s perhaps a l i ttl e better than that
obtai ned i n the precedi ng step, but the accuracy obtai ned i n both cases
i s about the best that can be expected of
reads from the fi gure
the geometr i c method. One
1
hence
@
= 2.0.
jj\2)
(l 18b)
The i nput angul ar range i s
Aj@ = Y(28) Y(18) = 8861.
(119)
I n terms of the constants of the uni nvested probl em the above resul ts
become
and
AX\2) = 88.61.
(121)
The val ues of Bf)/ A\ 2J and AX~2) are not ver y di fferent from those of
Eq. (112), wi th whi ch we started; i t i s evi dent that the sol uti on i s not far
from the best one-or , at l east, the best one wi th approxi matel y these
constants. I t i s ther efor e reasonabl e to fi x on defi ni te travel s,
AX1 = 55, AXZ = 90, (122)
as suffi ci entl y cl ose to the best val ues, and to determi ne a fi nal desi gn
usi ng the nomographi c method.
The r eader wi l l fi nd i t a useful exer ci se to car r y through thi s step,
usi ng the pr ocedur e of Sec. 5.11.
We have, by Eq. (51),
pl = hl .55, (123)
p~ = h~.90.
164 THE THREE-BAR LI NKAGE [SEC.5.19
TABLE5.5.SPECTEAL VALUESOFTEE PAEAMETEES
T
o
1
2
3
4
5
6
# ,
degr ees
h?)
0 0.0000
10 0.1818
20 0.3636
30 0.5454
40 0.7272
50 0.9090
60 1,0908
h;)
O.0000
0,4209
0.6306
0.7715
0.8777
0.9627
1.0341
+&,
degr ees
$
0.0
37.9
56.8
69.4
79.0
86.7
93.1
To make i t possi bl e to use Fi g. B.1, we choose ~ = 10, though, i n vi ew
of the smal l -val ue of AXl , i t w-oul dbe better to use ~ = ~. T-he spectral
10
/
U
\ 0 -90
\. o
O&
o
-116.5-
0.5
=Xz=I ogl oxl
\
Fm. 5.2S.Approxi mate mechani zati on of z~ = l og~oz~.
val ues P!) and #, computed wi th the ai d of Eq. 105, appear i n Tabl e
5.5. The choi ce of pb, suggested by the l ast appl i cati on of the geometr i c
method [Eq. (120)] i s
( )
B: , )
J & = l ogl o ~, = l og,, (1.39) = 0.143 = 0.15. (124)
1
~EC.5.19] MECHANI ZATI ON OF THE LOGARI THMI C FUNCTI ON 165
Only a few l i nes need be drawn on the overl ay. Pi cturi ng the mi rrored
form of Fi g. 527 wi th S1 to the l eft of S2, one sees that the scal es of ~
and X2 i ncrease together , whereas PI and Xl i ncrease i n opposi te senses;
the fi t i s to be expected wi th an overl ay cur ve of the mi nus fami l y,
probabl y wi th s = 15, si nce Xl ~ = 150.
Choosi ng
pbl = 0.15,
we fi nd
~bz = 0.283,
pa = 0.260,
Xuf = St = 150,
X,m = 116.5.
Hence
(125)
~= 0707 B, =1919 Al
A, x x
= 1.8X),
and fi nal l y
!
(126)
Q= 1055 ~= 0550
A, A,
The l i nkage i s sketched i n Fi g. 5.28. I t wi l l be di scussed further i n a
l ater exampl e (Sec. 7.8).
CHAPTER 6
LI NKAGE COMBI NATI ONS WI TH ONE DEGREE OF FREEDOM
I t i s onl y rarel y that one can mechani ze a gi ven functi on wi th hi gh
accuracy by a harmoni c transformer or a three-bar l i nkage. Usual l y a
mor e compl ex l i nkage must be empl oyed i n or der to gai n the fl exi bi l i ty
requi red i n fi tti ng the gi ven functi on wi th suffi ci ent accuracy. I nstead
of devi si ng enti rel y new structures i t i s better to combi ne the el ementary
l i nkages; the doubl e harmoni c transformer di scussed i n Chap. 4i s such a
combi nati on. Other useful combi nati ons are the doubl e three-bar
l i nkageanal ogous i n structure to the doubl e harmoni c transformer-and
combi nati ons of si ngl e or doubl e three-bar l i nkages wi th one or two
harmoni c transformers. Choi ce of the pr oper combi nati on shoul d of
cour se be determi ned by the type of functi on presented for mechani zati on.
Techni ques for the desi gn of such l i nkages wi l l be i ndi cated i n the present
chapter.
COMBI NATI ONOF TWO HARMONI C TRANSFORMERSWfTH A THREE-BAR
LI NKAGE
6.1. Statement of the Probl em.The combi nati on of two harmoni c
transformers wi th a three-bar l i nkage, as sketched i n Fi gs. 61 and 62, i s
parti cul arl y useful when i t i s desi rabl e to use sl i de termi nal s at both i nput
and output. (I n these fi gures both harmoni c transformers are i ndi cated
as i deal ; i n practi ce both wi l l usual l y be constructed as noni deal .) The
i nput l i nk and the crank R I S I consti tute a harmoni c transformer that
transforms the homogeneous i nput parameter HI i nto the homogeneous
angul ar parameter O1. The angul ar parameter correspondi ng to 0, wi l l
be cal l ed Xl (Fi g. 6.2); the constants of the harmoni c transformer are
then X,n, AX,. (I t i s i mportant to remember that 01, not H,, i s the
homogeneous parameter correspondi ng to Xl .) The crank T,S,, ri gi dl y
l i nked to R 1S1, i s descri bed by an angul ar parameter Xt and a homo-
geneous angul ar parameter 63, whi ch wi l l be i denti cal l y equal to 01. The
i nput harmoni c transformer thus carri es out the transformati on:
03 = (03[HI ) HI . (1)
The cranks TI S, and TJ 2, wi th the l i nk T, TZ, form a three-bar l i nkage
(constants X,n, AX, = AXI , X,-, AXA, etc.) that transforms the param-
eter Oti nto another homogeneous angul ar parameter,
e, = (0,[03) ol %, (2)
166
8EC.6,1] STATEMENT Or THE PROBLEM 167
associ ated wi th the angul ar parameter Xi . The crank R&z, rigidly
l i nked to 2,S,, i s descri bed by the angul ar parameter X,, or by the
homogeneous angul ar parameter 82, i denti cal l y equal to 04. Fi nal l y, the
crank RSZ and the output l i nk form a harmoni c transformer (constants
X,., AX, = AXi ), whi ch trans-
forms 84 = th i nto the homogeneous
output parameter
Hz = (HZ104) .04.
3)=
I t wi l l be noted that the angl es
Xl and X, descri bi ng the harmoni c
transformers cannot i n general be
FI G.6.1.Three-bar l i nkagecombi nedwi th
twoharmoni ctransformers.
measured from the same zer o l i nes
as the angl es X3 and Xq descri bi ng the three-bar-l i nkage confi gurati on, i f
the conventi ons of the precedi ng chapters are to be mai ntai ned. I n the
parti cul ar cases i l l ustrated i n Fi gs. 6.1 and 62, i n whi ch the i nput and
output l i nks of the transformers are paral l el to the l i ne of pi vots of the
three-bar l i nkage, the zer o l i nes for Xl and X, are perpendi cul ar to those
for Xa and Xi .
t t
7fl l 17i l fl J(fJl JJ~l ~JJJ
=.
*.
& -._-+
FI Q.6.2.Combi nati on of three-bar l i nkage wi th two harmoni c transformers, sketched i n
i ts extr eme posi ti ons.
The l i nkage as a whol e carri es out the transformati on
H, = (H21H,) . H,, (4)
wher e
(H21HI ) = (H,lt?,) . (0,103) . (031HJ. (5)
Gi ven a functi onal rel ati on i n homogeneous form,
h, = Ot,lhl) hl,
(6)
one wi l l wi sh to fi nd harmoni c-transformer functi ons, (H21191) and (&l H1),
and a three-bar-l i nkage functi on, (o4[0s), such that the product oper ator
(H,]HJ will approxi mate as cl osel y as possi bl e to (hzl hl ), on di rect or
168 LI NKAGE COMBI NATI ONS [SEC.6.2
compl ementary i denti fi cati on of the parameters (HI , Hz) wi th the
vari abl es (hi , ;L2).
I t woul d be ver y di ffi ctdt t: fi nd the best approxi mati on to (h,l h,)
wi thi n the twel ve-parameter fami l y of avai l abl e functi ons. The tech-
ni que to be descri bed i s i ntended onl y as a practi cal l y useful method for
obtai ni ng a good resuh i n a reasonabl y short ti me. Thi s i nvol ves a
prel i mi nary resol uti on of the desi red oper ator (HZI H1) i nto thr ee factors:
two harmoni c transformer oper ator s (usual l y i deal ), and a thi rd oper ator
to be mechani zed by the three-bar l i nkage. When the three-bar l i nkage
has been desi gned, by the methods of Chap. 5, the harmoni c transformers
are redesi gned, al most i nvari abl y as noni deal , i n or der to get a better fi t
to the gi ven functi on. Fi nal l y, the over-al l er r or i s further r educed by
smal l si mul taneous vari ati ons of al l constants of the l i nkage, by methods
to be di scussed i n Chap. 7.
6.2. Factori zati on of the Gi ven Functi on.-A rapi d method for fi ndi ng
a sati sfactory prel i mi nary factori zati on of (H21HJ i s essenti al to the
fi uccess of thi s pr ocedur e. Let Eq. (5) be mul ti pl i ed from the l eft by
(0,1H2), from the ri ght by (H,l %). One obtai ns
(d,l e,) = (e41H,) (2721HI ) (H,[03). (7)
Of the quanti ti es on the ri ght, (H,I H,) has a prescri bed f orm i n the gi ven
probl em, and the oper ator s (841H2)and (Hl l 0s), though unknown, are of a
rel ati vel y l i mi ted cl assparti cul arl y when attenti on i s restri cted to the
i deal -harmoni c-transformer oper ator s of Tabl es A 1 and A2 i n carryi ng
out the prel i mi nary factori zati on.
Mor e or l ess reasonabl e choi ces of the
oper ator s (194 IHz) and (Hl l fk) can be based on consi derati on of the form
of the gi ven functi on. One can then qui ckl y determi ne, by the graphi cal
mul ti pl i cati on correspondi ng to Eq. (7), the requi red form of (Oi (tk).
I nspecti on of thi s functi on wi l l suffi ce to i ndi cate whether i t can be
approxi mated by a three-bar-l i nkage functi on. I f so, the constants of
that l i nkage can be found by the methods of Chap. 5; i f not, the probl em
must be r econsi der ed and another choi ce of harmoni c-transformer func-
ti ons tri ed. Thi s process of tri al and er r or i s not excessi vel y burdensome
si nce each tri al i nvol ves onl y r efer ence to Tabl es .4.1 and A.2 and a
graphi cal constructi on. The speed wi th whi ch i t can be carri ed out
depends, of course, on the judgment and exper i ence of the desi gner, both
i n sel ecti ng the harmoni c-transformer functi ons and i n assessi ng the
possi bi l i tyy of mechani zi ng the deri ved (84163) by a three-bar l i nkage.
Some suggesti ons on the fi rst of these matters are contai ned i n the fol l ow-
i ng paragraphs.
I t i s possi bl e, though not usual l y desi rabl e, to mechani ze a gi ven
functi on approxi matel y by a doubl e harmoni c transformer and to use an
i nterposed three-bar l i nkage to make a smal l cor r ecti on; i t wi l l rarel y be
I
SEC.6.2] FACTORI ZATI ON OF A FUNCTI ON
169
sati sfactory to mechani ze the gi ven functi on approxi matel y by a three-bar
l i nkage and then attempt to conver t to sl i de i nput and output by harmoni c
transformers that make onl y smal l changes i n the form of the gener ated
functi on. I nstead, i n mechani zi ng monotoni c functi ons i t i s better to
make al l thr ee components of the l i nkage combi nati on contri bute about
equal l y to the curvature of the gener ated functi on. Perhaps the si mpl est
way of accompl i shi ng thi s i s to focus attenti on on the termi nal sl opes
of the factor functi ons, whi ch become mor e and mor e di fferent from 1
as the curvature i ncreases. When al l factor functi ons are monotoni c
one has
the termi nal sl opes of the gener ated functi on are products of the cor r e-
spondi ng termi nal sl opes of the factor functi ons. For a fi rst ori entati on,
to make sure that no factor functi on need have excessi ve curvature, one
can requi re that al l factor functi ons have the same termi nal sl opes:
Speci fi cati on of both termi nal sl opes i s suffi ci ent to fi x the i deal -harmoni c-
transformer functi ons compl etel y; they may be i denti fi ed by r efer ence to
Fi gs. 4.17 and 4.18. By use of Eq. (7) one can then determi ne the
correspondi ng requi red form of (OJ163), for exami nati on as to the, possi -
bi l i ty of mechani zi ng i t by a three-bar l i nkage. I t i s to be remembered
that thi s l i nkage must be one of speci fi ed angul ar travel s AX3 and AXJ,
these bei ng fi xed as the angul ar travel s of the i nput and output harmoni c
transformers, respecti vel y.
I f ther e exi sts no i deal -harmoni c-transformer functi on wi th the speci -
fi ed termi nal sl opes, or i f the angul ar travel s AX, = AX8 and AX, = AX4
are unsati sfactory, one can l i ghten the restri cti ons on the termi nal sl opes
by requi ri ng onl y that
I n addi ti on, any conveni ent angul ar travel s AX1 and AX, can be speci fi ed
and the constants of the two i deal harmoni c transformers then deter -
170 LI NKAGE COMBI NATI ONS [SEC.6.2
mi ned by use of Fi gs. 4.17 and 4.18. (An exampl e i s provi ded i n Sec.
63.) The requi red three-bar-l i nkage functi on, found as before, wi l l
agai n have termi nal sl opes gi ven by Eqs. (9).
When the gi ven functi on has one maxi mum or mi ni mum, at l east one
of the thr ee factor functi ons must al so have a maxi mum or mi ni mum.
Onl y one of Eqs. (8a) and (8b) can then be val i d, and a di fferent pr o-
cedur e must be empl oyed. I t i s usual l y best to choose the output-
harmoni c-transformer functi on as nonmonotoni cthat i s, to attempt to
mechani ze the functi on by a l i nkage of the sort i l l ustrated i n Fi g. 6.2.
The constants of thi s transformer shoul d be such that the functi on to be
gener ated by the other two el ements of the combi nati on,
(e41HJ = (%1%) (%I HJ = (041H2) (H21HJ,
(11)
i s monotoni c and as smoothl y cur ved as possi bl e.
The functi on (o4I HJ wi l l be monotoni c onl y i f the harmoni c-trans-
for mer functi on (H21~,) has the same val ues as the gi ven functi on (H,I H,)
.
FI ci . 6,3.Resol uti on of a gi ven functi on
(HzIHI) i nto an output-harmoni c-trans-
for mer functi on (Hzl 8J and a monotoni c
functi on (O,I HJ.
at the ends of the range of vari -
abl es (Fi g. 6.3). Thi s requi re-
ment fi xes the form of (Hzl oJ, and
hence (@,\H,), for any gi ven AXZ;
i t remai ns to choose the val ue of
thi s constaut. Thi s shoul d be
done wi th some attenti on to
mechani cal sui tabi l i ty but pri -
mari l y so as to assure that (oI I HJ
i s a smoothl y cur ved functi on, as
i n Fi g. 6.3; i t i s mor e i mportant
to avoi d poi nts of i nfl ecti on i n
(o,\ZZJ than to make i ts curvature
smal l . When (Hzl OJ has been
determi ned and the correspondi ng
functi on (19,1HJ has been obtai ned
by graphi cal constructi on (Sec.
3.4), i t remai ns to r esol ve thi s
l atter functi on i nto the product on one harmoni c-transformer functi on
and a three-bar-l i nkage functi on, as expressed i n the fi rst part of Eq. (1 1).
As when r esol vi ng a gi ven functi on i nto thr ee factors, i t may her e be
desi rabl e to choose the harmoni c-transformer factor by r efer ence to i ts
termi nal sl opes, fi xi ng
(%)..O= [ ( %) . l - O1
( $O. l - l = [ ( %) HJ
(12a)
(12Z))
Sm. 6.3] EXAMPLE: FACTORI NG THE GI VEN FUNCTI ON
171
I l k wi l l determi ne the constants of the second harmoni c transformer,
and i t wi fl remai n onl y to wor k out the requi red form of the three-bar-
I i nkage functi on by a second graphi cal constructi on. I n some cases i t wi l l
not be possi bl e to sati sfy both of these condi ti ons; one can then, for
i nstance, sati sfy one or the other , and i n addi ti on fi x AX1.
I t i s to be emphasi zed that the precedi ng paragraphs do not contai n
a prescri pti on that assures i mmedi ate success, and are i ntended onl y to be
suggesti ve. A sati sfactory resol uti on of the functi on may be found onl y
after several tri al s, i n whi ch the desi gner must be gui ded by hi s i magi na-
ti on and exper i ence.
Secti ons 6.3 and 64 wi l l car r y an exampl e through the stages of
factori zati on of the gi ven functi on and mechani zati on of the three-bar-
I i nkage factor, to the poi nt wher e ther e i s obtai ned a fi rst approxi mate
mechani zati on of the gi ven functi on by a combi nati on of a three-bar
l i nkage and two i deal harmoni c transformers. I n Sec. 6.5 we shal l then
return to a general di scussi on of the next stage of the desi gn pr ocedur e
i mprovement of the fi t by i ntroducti on of noni deal harmoni c transformers.
6.3. Exampl e: Factori ng the Gi ven Functi on.-To i l l ustrate the
detai l s of the method we shal l consi der agai n the probl em of mechani zi ng
the tangent functi on, but through a wi der range of vari abl es than was
attempted i n Chap. 4:
X2 = tan xl , 0< X, <80. (13)
As usual , we i ntroduce homogeneous vari abl es,
111= *,
h2=
5.;;13
TABLE6.1.z* = TANz,, O S z, S 80, I N HOMOGENEOUS VAKUABLES
h,
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.95
1.0
h,
O.0000
0,0248
0,0506
0.0785
0.1102
0.1480
0.1958
0.2614
0.3615
0,5427
0.7072
1.0000
(14a)
(14b)
172 LI NKAGE COMBI NATI ONS [SEC.6.3
Equati on (13) then becomes
h, = 0.17632 tan (h, . 80) = (h,l h,) oh,. (15)
Thi s functi on i s tabul ated i n Tabl e 6.1 and pl otted, as the desi red
(H21H,), i n l l g. 6.4. The termi nal sl opes, obtai ned by di fferenti ati ng
Eq. (15), are
( )
dh,
= 0.246,
()
dh,
Z2 hs-O %2 by-l
= 8.165. (16)
We fi rst consi der the possi bi l i ty of appl yi ng Eqs. (9) i n factori ng thi s
monotoni c functi on. Thi s woul d requi re
~+) @4=0=( *) HI =O= 0246) 06n)
(17a)
@),,=, =(&)~,=,= (8.165)~ =2.013.
(17b)
I nspecti on of Fi gs. 4.17 and 4.18 shows that ther e exi st no i deal -harmoni c-
transformer functi ons wi th the requi red termi nal sl opes; i t i s necessary
to use the l i ghter condi ti ons of Eqs. (10), whi ch become
( )
dH2
m =0 = (0.627) = 0.393,
(-)
dH,
(18a)
df3a e-o
(-)
dH,
de, eel
( )
= (2.013)2 = 4.057.
dH,
(18b)
z 9.1
I n addi ti on, val ues can be assi gned to AX1 and AX2. I f one requi res that
AX1 = AX2, the probl em becomes i denti cal wi th that di scussed i n Sec.
4.11. Appl yi ng the method of sol uti on descri bed ther e, one fi nds, for
exampl e, the f ofi owi ng sets of constants that sati sfy Eqs. (18):
1. AX1 = AXZ = 90, X1m = 7.5, X2r n= 67.5.
2. AX, = AX, = 100, X,~ = 17.5, X2m= 70.
Other sets of constants wi th AX1 = AX, are easi l y found by the same
method; a sl i ght and obvi ous modi fi cati on of the method i s requi red
i f one wi shes to have AX1 # AX2. For a fi rst tri al we shal l choose
AX1 = AX2 = 100, these val ues bei ng both mechani cal l y sati sfactory
and especi al l y conveni ent for the computati ons to be made. Then the
harmoni c-transformer functi ons to be used are
SEC.6.3] EXAMPLE: FACTORI NG THE GZVEN FUNCTI ON 173
(Hllo, ): 17.5 S X, s 82.5,
(11,/11, ): ~70 SX,~300.
These functi ons are pl otted i n Fi g. 64, the fi rst as a set of enci rcl ed
poi nts, the second as a conti nuous curve.
The desi red form of the three-bar-l i nkage functi on can now be com-
puted by appl i cati on of Eq. (7), or some equi val ent equati on. As i s
0.9
0.8
0,7
0
0.6
0.5
0.4
bf2i~1J (H1I q
v
o Q.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 M)
FI G. 6.4.I &oI uti on of the tangent functi on, (Ht\HJ, i nto a pr oduct of hmmmni c-
transformer functi ons, (H,l 0,) and (r9,1HJ ,and a functi on (0.1E%)that i s to be mechani sed
by a three-bar l i nkage.
usual l y desi rabl e, i n Fi g. 6.4 we have pl otted al l functi ons wi th the
O-scal e hori zontal . (Systemati c use of thi s conventi on hel ps to pr event
mi stakes and nakes easi er the change to noni deal transformers.) To
make use of these pl ots i n a graphi cal constructi on for (Oi l 0s) we must, i n
effect, sol ve an equati on that i nvol ves harmoni c transformer functi ons
onl y i n the for m (HI 0). Thi s we can obtai n by mul ti pl yi ng Eq. (7) fr om
the l eft by (l YJ OJ:
(HJL) (0,16,) = (H,p7J (H,\eJ. (19]
174 LI NKAGE COMBI NATI ONS
[SEC.6.4
The product on the ri ght can be formed by graphi cal mul ti pl i cati on of
known operators, as i n Fi g. 64, wher e constructi on of the rectangl e
0 + A * B * C l eads to l ocati on of the poi nt C on the (unpl otted) cur ve
representi ng thi s product. A correspondi ng poi nt F on the cur ve of
(0410J i s then found by graphi cal sol uti on of Eq. (19), thr ough construc-
ti on of the rectangl e C + D * E + F. I t i s, of course, unnecessary to
l ocate poi nt C. The compl ete constructi on for the poi nt F of the functi on
(oAI o,), correspondi ng to the poi at O of the functi on (HI I os), then i nvol ves
(]) constructi on of a verti cal l i ne through the poi nt O, and (2) l ocati on
of the poi nt F by constructi on of the l i nes O * A * B - D ~ E + F.
The compl ete cur ve (0,10,) shown i n Fi g. 64 i s qui ckl y determi ned by
r epeated appl i cati on of thi s constructi on. I t appears to be a functi on
that can be mechani zed by a three-bar l i nkage. We ther efor e tentati vel y
accept the resol uti on of the gi ven functi on as sati sfactory, and turn, i n
Sec. 6.4, to the probl em of desi gni ng the correspondi ng three-bar-l i nkage
component.
6.4. Exampl e: Desi gn of the Three-bar-l i nkage Component.-We
have now to consi der the probl em of mechani zi ng a functi on, gi ven
graphi cal l y i n Fi g. 6.4, by a three-bar l i nkage wi th fi xed angul ar travel s,
AX, = AX4 = 100.
(20)
Si nce both angul ar travel s are fi xed, the nomographi c method must be
used i n determi ni ng the other constants of the l i nkage. Al though thi s
method has been i l l ustrated i n Sec. 5.14, i t may be desi rabl e to show al l
stages of the pr ocedur e al so i n the present case, whi ch di ffers from the
earl i er exampl e i n that the method of successi ve approxi mati ons descri bed
i n Sec. 5.13 conver ges ver y sl osvl y. Thi s exampl e wi l l al so sei ne to
i l l ustrate the fact that a gi ven functi on can often be mechani zed by
several qui te cl i ffer ent l i nkages, among whi ch one must make a choi ce
on the basi s of mechani cal sui tabi l i tyy.
The functi on to be gener ated by the three-bar l i nkage, as read from a
careful l y constructed chart, i s gi ven i n Tabl e 62 i n both the di rect and
the i nvemed form. The vari abl es used are not the homogeneous vari abl es
8, and 0,, but the angul ar vari abl es [cf. Eq. (5.51)],
PI = AX393,
(21a)
P2 = AX413.i . (21ZI )
(I t i s to be remembered that i n thi s exampl e X8 wi l l repl ace the Xl of
Sees. 57 to 513, and Xi wi l l repl ace X,.)
We begi n by taki ng
@\) * -0.2, (22)
the usual fi rst choi ce of the author,
SEC.6.4] EXAMPLE: DESI GN OF A THREE-BAR LI NKAGE 175
TABLE6,2.GI vENFUNCTI ONFOETHEEXAMPLE
(Ad
V2,
degr ees
0.0
6.7
12.6
18.5
24.7
32.1
40.7
51.3
65,5
82.5
100.0
w,
degr ees
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
90.0
100.0
(91144
PI ,
degr ees
0.0
15.5
32.4
47.3
59.2
68.9
76.4
82.8
88.6
94.3
100.0
V2,
degr ees
0.0
10.0
20,0
30.0
40.0
50.0
60.0
70.0
80.0
90.0
100.0
Choosi ng b = 10, n = 100/10 = 10, we fi nd the val ues of q~) and
9:) i n C!ol umns 1 and 2 of Tabl e 6.2. Constructi on of the overl ay
fol l ows preci sel y the steps descri bed i n Sec. 514 and need not be expl ai ned
her e. On turni ng the overl ay face down on the nomographi c chart, a
sati sfactory fi t i s found between the cur ve s = 8 of the mi nus fami l y
on the overl ay, and the cur ve pbz = 0.075 suppl i ed by i nterpol ati on
on the nomogram. Fi gure 6.5 shows on the nomogram gri d the con-
structi on of the parti cul ar overl ay cur ve for whi ch the fi t was obtai ned,
and the posi ti on of fi t on the chart. The fi t i s exact at the ends, ver y
good for the l arger val ues of pp and the smal l er val ues of X3, and some-
what l ess sati sfactory for the l arger val ues of X3. The r efer ence l i nes
of the turned overl ay are al so shown i n the posi ti on of fi t.
The el ements of the l i nkage are thus establ i shed:
/J)\l )=
-0.2, as assumed.
~b~l) = 0.075.
~a(l ) = 0.~7, read at the i n~mcti on of the vefl i cal r efer ence l i ne
wi th the pp-scal e.
X~~ = 7.2, read at the i ntersecti on of the hori zontal r efer ence l i ne
wi th the q-scal e.
x\~ = 80. (s6 = X,., not Xti , because the cur ve that gi ves a
fi t i s of the mi nus fami l y.)
176 LI NKAGE COMBI NATI ONS [SEC.64
By Eq. (5.56) we have (usi ng the l ower si gns i n the fi rst equati on because
the fi t was obtai ned wi th a cur ve of the mi nus fami l y)
PI = X3 80
1
(23a)
P2 = X4 + 7.2.
(23b)
30
I \ I I I I 1/ I hv x
01
I /1 I I ! I I I I\l I k 1/
-0:5 - oh -0:3 -02 -91 +0:1 +02 +Oi +0.4 +0.3
:
FI G. 6.5.Mechani zati on of the tangent functi on: Fi rst appl i cati on of the nomogr aphi c
method. The dashed l i ne i s the i nterpol ated contour Mb=0.075.
Equati ons (23) are onl y as exact as the fi t obtai ned. Si nce the fi t i s
exact at the ends of the range of X3, the i nput and output travel s of the
l i nkage as desi gned wi l l have the requi red val ue, 100. The pl - and
Os-scal es wi l l i ncrease wi th decreasi ng Xa, and p~, @i , and Xt wi l l i ncrease
together .
As a check, the l i nkage i s drawn as i n Fi g. 6.6, whi ch shows the cranks
i n thei r extr eme posi ti ons. The di stance A ~1)between the crank pi vots i s
taken as the uni t of l ength; the rel ati ve crank l engths are drawn i n as
B~l )
AT
= 1041 = 10-0.2 = o.&o,
(24a)
Ai l )
~
= l o~ = 1OI J.ZOT = 0.6~.
(24b)
I
(
SEC.64] EXAMPLE: DESI GN OF A THREE-BAR LI NKAGE
177
The constancy of the requi red l ength of the connecti ng l i nk,
%=(%)(%)
= l o,(b~d c 100.132= 0,TS6,
(25)
provi ded a check on the quanti ti es determi n~d i n the fi tti ng process.
The fi t obtai ned wi th thi s fi rst val ue of I .A, i s so good that one cannot
expect i t to be changed greatl y by further cal cul ati ons.
Neverthel ess, we
now attempt to i mprove i t by the process of successi ve approxi mati ons,
i nterchangi ng the rol es of 91 and q.w
The i nverted functi on (Sec. 5.13)
has al ready been gi ven i n Col umns 3 and 4 of Tabl e 6.2. We begi n the
process of mechani zi ng thi s by taki ng
the approxi mati on bei ng cl ose enough for the purpose at hand.
Al -l
t
I
I
t
94
X3M=-80
g
0
$
FI G. 6.6.Mechani zati on of the tangent functi on: Fi rst three-bar-l i nkage
desi gn.
We know that the l i nkage to be desi gned wi l l not be ver y di fferent
from that of Fi g. 66 r efl ected i n a verti cal l i ne. For the i nverted prob-
l em we must then have XW = 170, s = 17. (The fi t wi l l agai n be
found i n the mi nus fami l y of curves.) Thus onl y a few l i nes need be
drawn on the overl ay.
Fi gure 6.7 shows the constructi on of the overl ay l i ne for whi ch the
best fi t i s obtai ned (s = 17 i n the mi nus fami l y, as predi cted), and the
posi ti on of fi t wi th the turned overl ay. Thi s overl ay l i me i s pecul i ar i n
that i t abruptl y r ever ses i ts trend at the poi nt P~-l, wi th the resul t that
P#l n and P~lg) fal l together ; thi s, however , i s not an i ndi cati on of any

1
7
8
L
I
N
K
A
G
E
C
O
M
B
I
N
A
T
I
O
N
S
[
%
C
.
6
.
4
.
I I
~Ec. 6.4] EXAMPLE: DESI GN OF A THREE-BAR LI NKAGE 179
pecul i ari ty i n the l i nkage. From the posi ti on of fi t we read the fol l owi ng
val ues for the constants:
m = 0.2, as assumed,
@l
~b~z)= 0.025,
@2) = (),200,
~,~ = 93.8,
X3,1 = 170,
f,m = 270,
i
@
3!2)
= 10-0.2 = (),63()
4(21
$,
= 100.2 = (3,630,
fi j2)
~
= 100.176= 0.668.
1
(27)
Si nce w and P2 have i nterchanged i n thi s probl em, Eq. (5.56) becomes
P2 = X3 170
q = X4 + 93.8. (28)
When the fi tti ng process i s carri ed out on a l arge scal e i t can be seen that
these constants gi ve a fi t good to wi thi n 1.
To make mor e evi dent the change i n constants due to thi s second
cal cul ati on, we r ewr i te the above resul ts, usi ng Eqs. (5.68) and the
obvi ous rel ati ons
(30)
To thr ow Eq. (28) i nto a form comparabl e to Eq. (23), one must al so
remember that X8 i n the i nverted probl em corresponds to Y 180 X4 i n
the di rect probl em, and X4 corresponds to + 180 X8. We have then
w = Xa 86.2
>
(31a)
@2 = x4 + 10.OO. (31b)
1so LI NKAGE COMBI NATI ONS [SEC64
Fi gure 6.8 shows the mi rror i mage of the l i nkage descri bed by Eqs.
(27) and (28)that i s, the l i nkage descri bed by Eqs. (30) and (31).
Di r ect compari son can then be made wi th the l i nkage of Fi g. 6.6, wi th
r espect to whi ch thi s i s supposed to be an i mprovement. The scal es of
X3 and ~s i n these fi gures are mi rror i mages, as are those of X, and ~,,
but the scal es of th and e,, whi ch al one are of real i nterest, are al most the
same.
I t wi l l be obser ved that our consi derati on of the i nverted probl em has
l ed us back to the i ni ti al l y assumed val ue of pb,; i ndeed, al l constants of
the l i nkage are essenti al l y the same i n the second approxi mati on as they
Fm. 6.8.Mechani z8ti on of the tangent functi on: Second three-bar-l i nkage
desi gn.
wer e i n the fi rst. Thi s, together wi th the good fi t obtai ned, mi ght l ead
one to suppose that the i ni ti al choi ce of pbl was unusual l y fortunate, that
another choi ce woul d have been deci dedl y wor se, and that the method of
successi ve approxi mati ons woul d have l ed to conver gence on the val ue
~bl = 0.2. Thi s, however , woul d be i ncorrect: we have her e a case i n
whi ch a good fi t does not depend upon a parti cul ar choi ce of pbl, and the
method of successi ve approxi mati ons conver ges ver y sl owl y, i f at al l .
For exampl e, i f we had chosen Pbf) = 0.3 we woul d have found a good
fi t for pa() = 0.314. Passi ng to the i nverse probl em, we woul d have
assumed ~~~~= Pl J = 0.314 and then found pb~2) = PCV2) = 0.3,
ver y cl osel y i ndeed.
Conver gence to a defi ni te val ue of pbl i s her e so sl ow as to be undetecta-
bl e i n a graphi cal method, and i s not of any practi cal i mportance for
obtai ni ng a good fi t. I t wi l l , however , be obser ved that ther e i s one
constant whi ch i s the same i n al l these l i nkages:
I
Sec. 6.4] EXAMPLE: DESI GN OF A THREE-BAlt LI NKAGE
181
or
I.& + pa = 0.01, (32a)
()
B,
l og,o --& = 0.01. (32b)
I t i s evi dent that to obtai n a good fi t to our gi ven functi on one must have
the crank l engths ver y nearl y equal . Any adjustment of parameters
r, .
1s01i
, , ! 1 1 [ 1 II I 1!/ 1! I
J Inrv r
+. l/- I x L.--Y.
II I I
L{
*
I l \ [ I I {/ I AV A A ~1 \_&i
-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 +0.1 +0.2 +0.3 +0.4 +0.5
w
Fm. 6,9.Mechani zati on of the tangent functi on: Thi rd approxi mati on of the nomogr aphi c
method.
whi ch tends to di sturb thi s rel ati onfor i nstance, change of a when bl
i s fi xed-wi l l l ead to ver y l i ttl e i mprovement i n the over-al l fi t, even
though i t mi ght resul t i n a marked i mprovement i n some other constant
of the l i nkage. The i deal method for adjusti ng constants i n thi s probl em
woul d be one i n whi ch bl + a coul d be tr eated as fi xed and the other
constants vari ed. Thi s i s, however , a matter of rather academi c i nterest
as the fi t al ready obtai ned i s qui te sati sfactory.
Thk same gi ven functi on can be mechani zed by other radi cal l y
di fferent l i nkages. As noted at the end of Sec. 5.13, befor e accepti ng any
desi gn one shoul d seek a sol uti on of the probl em wi th ~bl of the opposi te
si gni n thi s case posi ti ve. Tri al of pbfJ = 0.2 l eads to so poor a fi t
182 LI NKAGE COMBI NATI ONS
that i t i s uncertai n what val ue of pa i s real l y best.
desi rabl e to tr y another val ue of pb~lJ . We take pb~)
[SEC.6.4 ~
I n such a case i t i s
= 0.4. The curves
i
of the pl us fami l y l abel ed s = 9 and .s = 10 then gi ve the best fi t, but an
i ntermedi ate overl ay curve, s = 9.5, i s appreci abl y better. Fi gure 6.9
FI G. 6.113.-Mechani zati on of the tangent l i nkage: Thi rd three-bar-l i nkage desi gn.
shows the constructi on of thi s cur ve and the fi t obtai ned on the contour
gb = 0.1. The constants of the l i nkage are
Pbf) = 0.4,
#b\) = 0.1,
~afl ) = 0.686,
X,~ = 326.4 or 33.6,
X,m = 95,
By
Ajl )
= 2,54,
A ~1)
AT = 4-857
B$
AT = 610
(33)
13EC. 6.4] EXAMPLE: DESI GN OF A THREE-BAR LI NKAGE 183
Her e Xti = sJ, si nce the fi t i s found i n the pl us fami l y. Thi s l i nkage i s
sketched i n Fi g. 6.10. By Eq. (5.56), usi ng the upper si gns, we i i nd
pl = x3 95, (34a)
92 = X4 + 33.6. (34b)
The di recti ons of i ncreasi ng 03 and 194are then those i ndi cated i n the
sketch.
I n appl yi ng the method of successi ve approxi mati ons one woul d
normal l y take ~;~z) = paf I ) = 0.686. I nsteadz i n or der to keep wi thi n
the range of the nomogram, we shal l take @~21 = 0.6. The l i nkage
generati ng the i nverted functi on must di ffer from that of Fi g. 6.10 by
refl ecti on i n a verti cal l i ne, and al so i n hori zontal l i ne, i n or der that the
scal e of the output quanti ty may i ncrease cl ockwi se (cf. Sec. 5.14). The
fi t i s to be expected agai n i n the pl us fami l y of overl ay curves, for
X3. = 150,
or p = 15. I t i s i n fact wi th thi s cur ve that the best fi t i s found, on the
contour pb = 0.3. Thi s fi t i s shown i n Fi g. 6.11, whi ch makes use of the
extensi on of the nomogram i nto the range q > 180. The constants of
the l i nkage as thus determi ned are
#i i
Z,m
23.
.
.
.
.
.

.
.
0.6,
0.3,
0.435,
279.8,
150,
3.98,
2.72,
5.43.
(35)
The form of thi s l i nkage, after refl ecti on i n verti cal and hori zontal l i nes,
i s shown i n Fi g. 6.12, for compari son wi th Fi g. 6.10. By Eq. (5.56),
qz = X3 150 (36a)
PI = X4 279.;0. (36b)
The scal es for Osand 04, as associ ated wi th the r efl ected l i nkage, then have
the senses i ndi cated i n the sketch. The l i nkage of Fi g. 6.12 provi des an
excel l ent fi t to the gi ven functi on, and ther e i s no reason to pr oceed to a
thi rd approxi mati on, wi th ~b~3) = ~dc2J = 0.435.
We have thus mechani zed the gi ven functi on by three-bar l i nkages
184 LI NKAGE COMBI NATI ONS [SEC.6.4
1
x )) J Iu f 1 11 1 I I
+
KnlL/ P#4LLLY
. 1.
a IF
\ I
-- .- -- --

l\
g
.
wi th @2 <0 and pbz > 0, respecti vel y. I deal fy, ei ther of these l i nk-
ages mi ght be used. Mechani cal l y, the second l i nkage i s much l ess sati s-
factory than the fi rst, both as regards space requi red and the magni tude
of backl ash er r or (acute angl es between the cranks and the connecti ng
l i nks wi l l tend to magni fy backl ash). I n our further di scussi on of thi s
exampl e we shal l ther efor e use
gi ven by Eqs. (30) and (31).
the l i nkage of Fi g. 6.8, wi th constants
Di r ect cal cul ati on, by the methods of
SEC.6.4] EXAMPLE: DESI GN OF A THREE-BAR LI NKAGE 185
\
FI G. 6.12.Mechani zati on of the tsngent l i nkage: Fourth three-bar-l i nkage desi gn.
Sec. 5.1, shows that thi s l i nkage generates the rel ati on between es and
84gi ven i n Tabl e 6.3 (cf. Tabl e 6.2).
TABLE6.3.(0,]f?,) AS GENERATED BY THREE-BAR LI NKAGE
&)
0.0
0.1
0.2
0.3
0,4
0,5
0.6
0.7
0.8
0.9
1.0
d ,
0
10
20
30
40
50
60
70
80
90
100
x$),
degr ees
- 86.2
96.2
-106.2
-116.2
-126.2
136.2
146,2
156.2
-166.2
176.2
186.2
Xo ,
degr ees
10.10
2.80
3.08
8.71
14.66
21.48
29.88
40.80
55.17
7256
88.90
d)
degr ees
o. 10
7.20
13.08
18.71
24,66
31.48
39.88
50.80
65,17
82.56
98.90
0:) gen.
0.0000
0.0737
0.1331
0.1900
0.2501
0.3190
0.4038
0.5141
0.6592
0.8349
1.0000
0:) gi ven
O.000
0.067
0.126
0.185
0.247
0,321
0.407
0.513
0.655
0.825
1,0000
Wi th the constants as gi ven the fi t i s not exact at one end of the curve, and
AX, does not have exactl y the desi red val ue of 100. Thk di screpancy
coul d be r emoved by readjustment of the constants but wi l l be cor r ected
186 LI NKAGE COMBI NATI ONS [8EC.6.5
l ater i n an easi er way. The gener ated or ) i n Col umn 6 of Tabl e 6.3 i s the
homogeneous vari abl e correspondi ng to the gener ated P!), rather than
that gi ven by Eq. (21b). The er r or i n thi s quanti ty nowher e exceeds 1
per cent of the total travel .
We have thus arri ved at a fi rst approxi mate mechani zati on of the
tangent functi on, Eq. (13), by a combi nati on of two i deal harmoni c
transformers and a three-bar l i nkage, wi th the constants
X,m = 17.5 AX, = AX~ = 100,
X,m = 70, AXZ = AX, = 99,
X,m = 186.2, X,m = 10.1O,
B,
z,
= 0.630, 2
z
= 0.630,
z
= 0.668.
(37)
The er r or i n thi s mechani zati on i s most easi l y determi ned by extendi ng
Tabl e 6.3 to ei ther si de. Usi ng the harmoni c-transformer constants of
Eq. (37), one can compute the val ues of H, and Ha associ ated wi th the
tabul ated val ues of % and 01; these can be compared wi th val ues of hl and
h, computed by Eq. (15). The resul ti ng val ues appear i n Tabl e 6.4.
TABLE6.4.(H,I H,) AS GENERATED BY FI RSTAPPROXI MATE LI NKAGE
h, = H,
0.0000
0.1317
0.2665
0.4002
0.5288
0.6485
0.7555
0.8467
0.9191
0.9708
1.0000
e, 94
0.0 0.0000
0.1 0.0737
0.2 0.1331
0.3 0.1900
0.4 0.2501
0.5 0.3190
0.6 0,4038
0.7 0.5141
0.8 0.6592
0.9 0.8349
1.0 1.0000
H,
O.0000
0.0358
0,0721
0.1127
0.1612
0.2234
0.3085
0,4300
0.6017
0.8136
1.0000
~y
0,0000
0.0328
0,0688
0.1103
0.1604
0.2247
0.3108
0,4306
0.5965
0.8064
1.0000
0.0000
0.0030
o .0033
0.0024
0.0008
+0.0013
+0.0023
+0. 0006
0.0052
0.0072
0.000o
The over-al l er r or thus remai ns l ess than 1 per cent of the total travel .
6.5. Redesi gn of the Termi nal Harmoni c Trsnsformers.-The
methods descri bed i n Sec. 6.2 wi l l l ead one to a prel i mi nary mechani zati on
of the gi ven functi on by a combi nati on of a three-bar l i nkage and two
i deal harmoni c transformers. Accepti ng the three-bar-l i nkage constants
as fi xed, one can then i mprove the accuracy of the devi ce, and at the
same ti me bri ng i t i nto a mor e sati sfactory mechani cal form, by redesi gn-
i ng the termi nal harmoni c transformers as noni deal . The probl em of
desi gni ng the two termi nal harmoni c transformers di ffers but l i ttl e from
that of desi gni ng a doubl e harmoni c transformer and can be sol ved by the
same methods. (Cf. Sees. 413 to 415.)
SEC.66] REDESI GN OF THE TRANSFORMERS 187
Graph&al Melhod Of Successive Approx~matz2w.-The probl em i s to
choose oper ator s (031HJ and (Hal 04), each characteri zed by thr ee di spos~
bl e constants X~, L, E*, whi ch gi ve the product oper ator
(H21HI ) = (H21e4) (84108) (e,l Hl ) (38)
as nearl y as possi bl e a speci fi ed form. We fi rst tr y to make (HZI HI )
i denti cal i n form wi th (hl l hz) by changi ng onl y one of the transformer
oper ator s-for exampl e, (HZ104)and assi gni ng to (%I HI ) i ts fi rst approxi -
mate form, (o@,),. The requi red form of (H,I O,) can be determi ned by
sol uti on of
G
(h,pl ,) = (27210,) (04103) (031H1)1,
(39)
by the graphi cal constructi on i l l ustrated i n Fi g. 6.13 (whi ch appl i es to
the exampl e di scussed i n Sees. 63 and 6c4). A judi ci ousl y chosen
approxi mati on to thi s wi l l be (HZ184)2.
The form of (&l H,) requi red, i n
conjuncti on wi th thi s form of (H2IBl ), to make the mechani zati on exact,
can then be determi ned by graphi cal sol uti on of
(h,l h,) = (H,164)2 (thl %) (o,\HI ) ; (40)
(031HI )Z i s determi ned as a sui tabl e approxi mati on to thi s. Next,
(H,I 04) i s readjusted, and so on unti l
the fi t can no l onger be i mproved or
unti l the l i mi ts of appl i cabi l i ty of the
graphi cal method are reached.
Numencal Method.-The numeri -
cal method for the desi gn of noni deal
doubl e harmoni c transformers (Sec.
4.15) can be appl i ed to the present
probl em wi thout essenti al change.
I n parti cul ar, Eqs. (4.89) to (4.97)
are val i d her e al so, provi ded onl y
that I fz (0S) i s taken to mean the
val ue of 112 correspondi ng to the
speci fi ed val ue of OS; al ternati vel y,
we may consi der H2(th) to be an
abbrevi ati on for ~2[04(&)], wher e
F1o. 6.13.Graphi cal constructi on of the
r equi r ed for m of (Hz!0,).
H2(t?J i s defi ned by Eq. (4 12) and the functi onal rel ati on 0,(0,) i s deter -
mi ned by the three-bar l i nkage under consi derati on, The method wi l l be
ful l y i l l ustrated i n the next secti on.
6.6. Exampl e: Redesi gn of the Termi nal Harmoni c Transformers.
I n conti nui ng the exampl e of Sees. 63 and 6.4 we appl y numeri cal meth-
ods to the redesi gn of the termi nal harmoni c transformers. Thi s exampl e
i s of speci al i nterest i n showi ng that strai ghtforward appl i cati on of the
188 LI NKAGE COMBI NATI ONS [SEC.6.6
method of Sec. 4.15 does not al ways l ead to a sati sfactory resul t; the
modi fi cati on requi red i n the present case wi l l be descri bed.
We shal l keep fi xed al l constants of the l i nkage speci fi ed i n Eq. (37)
and shal l adjust onl y the constants L1, L2, ET, ET, whi ch characteri ze the
i nput and output l i nks.
We shal l fi rst of al l attempt to make the er r or i n the mechani zati on
vani sh for 03 = 0.2, 0.4, 0.6, 0.8. I n Tabl e 6.5 wi l l be found the quanti -
ti es needed to gi ve expl i ci t numeri cal form to Eqs. [4.97). Val ues of OS
and 64wi l l be found i n Tabl e 6.4. Val ues of H ~can be found from Tabl e
A*1 by an i nterpol ati on between correspondi ng efl tri es i n the col umns
AXi = 100, X;~ = 20, and AXi = 100, Xi ~ = 15. TO obtai n
val ues of H; woul d requi re i nterpol ati on i n both AX; and 81; i t i s advi sabl e
to make a di rect cal cul ati on by Eq. (4.42). The val ues of H~ and H; i n
Tabl e 6.5 have been thus obtai ned. The fs have been computed fr om the
Hs and H* s, and the val ues of dh,/ dh, have been comput~d as
( )
dh~ = 47r
~h~ h,-~, 9 tan 80
. see (80 . HI ).
TABLE6.5,CONSTANTS REQUI REDIN DESI GNPROCEDURE
I
l
0.0 0.9468
0.2 0.9989
0.4 0.9124
0,6 0.6978
0,8 0.3809
0.9 0,1957
1,0 0.0000
j,(h)
0.0000
0.3403
0,4101
0.2678
0.0726
0.0121
0.000o
j2(f3J
O.0000
0.6087
O .9325
O .9325
O .6087
0.3361
0.0000
H:
0.0000
0.3120
0.5505
0.7962
0.9848
0.9630
0,8094
.f,(o,)
0,0000
0,0501
0.1974
0.4318
0.5757
0.3944
0.000o
j,(h)
0.000o
o .5073
O,8400
1.0930
-0.9956
0.6090
0.0000
(41)
dh,
dh,
o 0000
0.2837
0.4501
1.0115
3.0622
5,399
Equati ons (4.97) become, for OS= 0.2, 0.4, 0.6, 0.8, 0.9, respecti vel y,
0.0965a + 0.1727b + 0.0501c 0.5073d = 0.0033, (42a)
0.1846a + 0.4197b + 0.1974c 0.8400d = 0.0008, (42b)
0.2709a + 0.9432b + 0.4318c 1.0930d = 0.0023, (42c)
0.2223a + 1.8640b + 0.5757c 0.9956d = 0.0052, (42d)
0.0653a + 1.8146b + 0.3944c 0.6090d = 0.0072. (42e)
Attempti ng to r educe the er r or to zer o at the fi rst four val ues of L9S, we
sol ve si mul taneousl y the fi rst four of these equati ons, and obtai n
a = 0.0246, b = 0.0206, C= 0.0702, d = 0.0017. (43)
Equati on (4.31) gi ves
g, = 0.6751, g, = 0.4627.
(44)
SEC.6.6] REDESI GN OF THE TRANSFORMERS 189
By Eqs. (4.46) and (4.47) we have then
0.2279
L, = ~,
0.1070
L,=?
b
E:=z
E;=: (45)
or
L, = 9.264, L, = 1.525,
E? =
0.837, E!, = 0.024. (46)
Cal cul ati on, by the methods of Tabl e 45, of val ues of H{, H~, and of
h~)(the val ue of hz correspondi ng to h, = H;), yi el ds the resul ts shown i n
Tabl e 6.6. The over-al l er r or , h~) Hi , of thi s mechani zati on i s
actual l y l arger than that wi th whi ch we started, rather than zer o at the
chosen val ues of oS. Thi s i s evi dentl y due to excessi vel y l arge er r or s i n
the approxi mate l i near equati ons used i n the desi gn pr ocedur e. I n or der
TAZLE6.6.PERFORMANCE OF TEELI NXA~E[I Zb. (45)]
0.2
0.4
0.6
0,8
0.9

H,
0.2665
0.5288
0.7555
0.9191
0.9708
H;
0.2889
0.5602
0.7831
0,9344
0.9785
0.0721
0.1612
0.3085
0.6017
0.8136
H;
0.0746
0.1732
0.3362
0.6398
0.8630
)&)
0.0688
0,1604
0,3108
0.5965
0.8064
0.0753
0.1752
0.3409
0.6469
0.8499
h$QD H;
0.0007
0.0020
0.0047
0,0071
0.0100
to make a smal l cor r ecti on i n the over-al l gener ated functi on we have been
for ced, by i ts pecul i ar form, to use harmoni c transformers that devi ate
strongl y from the i deal form; h~) h~o) and Hz H: are l arge. Accor d-
i ng to our approxi mate equati ons, these l arge cor r ecti ons shoul d nearl y
cancel , l eavi ng an over-al l cor r ecti on of much smal l er magni tude and of
the desi red form. Unfortunatel y, i n computi ng thew l arge cor r ecti ons
wi th the l i near equati ons we have made er r or s that do not tend to cancel
out-er r or s that, i n thei r aggregate, are even l arger than the di fference
that i t was desi red to compute. Accordi ngl y, the expected accuracy i n
the cor r ecti on has not been real i zed.
Di ffi cul ti es of thi s type can someti mes be avoi ded by ver y sl i ght
modi fi cati ons i n the condi ti ons i mposed. I n the present case, for i nstance,
one need admi t onl y a ver y smal l er r or at OS= 0.4 i n or der to use harmoni c
transformers that are mor e accuratel y descri bed by the l i near equati ons;
the l i nkage thus desi gned has a performance much cl oser to ones expecta-
ti ons, and correspondi ngl y mor e sati sfactory.
We desi re to make a posi ti ve cor r ecti on at t%= 0.2, a negati ve one at
OS= 0.6. I t i s evi dent, then, that the cor r ecti on made wi l l tend to be
smal l at Os = 0.4, whether or not speci al car e i s taken wi th thi s poi nt.
We shal l ther efor e rel eaae thi s poi nt from di rect control , and shal l sol ve
the fi rst, thi rd, and fourth of Eqs. (42) for the constants a, c, d, i n terms of
190 LI NKAGE COMBI NATI ONS [SEC,6.6
the constant b. (The choi ce of the constant b for thi s speci al treatment i s
qui te arbi trary.) One fi nds
a = 0.226 847 11.400571 b, (47a)
C = 0.015 059 3.980738 b, (47b)
d = 0.048170 + 2.115946 b. (47C)
The er r or to be expected at O, = 0.4 i s 0.0008 pl us the quanti ty on the
l eft-hand si de of Eq. (42b):
~,., = 0.00076 0.03895 b. (48)
As one woul d expect, 80.4i s qui te i nsensi ti ve to the choi ce of b; we can
choose thi s constant wi th the i dea of getti ng a good mechani cal desi gn,
wel l descri bed by the l i near equati ons. We desi re, then, that -Ll and Lz
shal l not be ei ther ver y l arge or ver y smal l , and that E? and E; shal l l i e
between zer o and one. I t fol l ows that a and c shoul d be of the or der of
magni tude of 0.1, that b shoul d have the same si gn as a, and that d shoul d
have the same si gn as c. We can gi ve LI and Lz roughl y equal magni t-
udes, and obtai n the desi red si gn rel ati ons, by setti ng b = 0.0135:
a=
0.07294, b = 0.0135, C = 0.03868, d = 0.01960, (49)
L, = 3.1245, L, = 2.7676, ET = 0.185, E; = 0.507.
The expected val ue of 60., i s then 0.0002. The actual performance of
the l i nkage i s shown i n Tabl e 6.7.
TABLE6.7.PERFOE~ANCE OF THELI NKAGE[EQ. (49)]
83
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1,0
H:
O.0000
0.1217
0.2504
0.3819
0.5120
0,6361
0.7489
0.8457
0.9219
0.9743
1.0000
O.0000
0.0302
0.0640
0.1036
0.1523
0.2159
0.3037
0.4293
0.6045
0.8170
1.0000
0.0000
0.0001
0.0003
0.0005
0.0008
0.0010
0.0006
0.0003
0.0005
0,0086
0.0000
The er r or s due to use of the approxi mate equati ons are smal l , and the
performance of the l i nkage i s sati sfactory at the poi nts control l ed.
Unfortunatel y, the er r or i ncreases rapi dl y for OS>0.8, and the desi gn can
not be consi dered acceptabl e. I t i s evi dent that i n the desi gn process
mor e attenti on must be pai d to the er r or for 03 = 0.9.
I
SEC.6.6] REDESI GN OF THE TRANSFORMERS 191
An attempt to control the er r or at 0, = 0.9 i nstead of 0, = 0.8, by
usi ng Eq. (42e) i nstead of Eq. (42d), l eads to si mi l ar resul ts: one can
actual l y make the er r or at 63 = 0.9 ver y smal l , but the er r or at Os = 0.8
takes on a l arge negati ve val ue.
An attempt to make the er r or vani sh for
both 193= 0.8 and Os = 0.9, by sol vi ng si mul taneousl y Eqs. (42a),
(42c), (42d), and (42e), l eads to cal cul ati on of the constants
LI = 1.030, L~ = 0.771, E? = 0.171, Et = 0.250. (50)
These val ues are such that the l i near equati ons can not be expected to be
accurate; the l i nkage wi l l not gi ve the expected good performance even
at i % = 0.8 and Oa= 0.9.
The case her e encounter ed i s i n fact one i n whi ch adjustment of the
constants LI , Lz, E?, E; can not bri ng the over-al l er r or wi thi n ver y stri ct
tol erances, such as f 0.001. Readjustment of Xl ~ and XZ~, or even a
new resol uti on of the gi ven functi on and redesi gn of the three-bar-
l i nkage component, woul d be requi red i f such accuracy wer e demanded.
On the other hand, a tol erance of f 0.0025 can be met wi thout such
redesi gn, by a somewhat di fferent approach.
Our probl em i s to cor r ect the er r or appeari ng i n the l ast col umn of
Tabl e 64 by maki ng the pr oper l i near combi nati on of four cor r ecti on
uncti ons (~),8f1(8a)1 (%
d~l ,,~2(e3), fs(%), and f,(h). We have
been deal i ng wi th speci al val ues of these functi ons as coeffi ci ents i n
Eqs. (42). These are r epr oduced, wi th the er r or to be cor r ected, i n
Tabl e 6.8. What i s requi red i s that we make l i near combi nati ons of
TABLE 6.8. EFLaoR CORRECTION FUNCTI ONS
dh,
e, *f, G. fl f s
fd
h~j H, FI F,
0.2 0.0965 0.1727 0,0501 o .5073 0.0033 0,0469 0, 1503
0.4 0.1846 0,4197 0.1974 0.8400 0.0008 0.0641 o. 1344
0.6 0.2709 0.9432 0.4318 1.0930 0.0023 0.0000 0.0000
0.8 O.2223 1.8640 0.5757 O.9956 0.0052 0.3130 0.1824
0.9 0.0653 1.8146 0.3944 0.6090 0.0072 0.4559 0.1538
entri es i n Col umns 2 to 5, i ncl usi ve, wi th coe5ci ents a, b, c, d, such that
the sums approxi mate as wel l as possi bl e to the correspondi ng entri es
i n Col umn 6. Exami nati on of Tabl e 68 wi l l make i t cl ear that our di ffi -
cul ti es have arisen from the attempt to make a posi ti ve cor r ecti on i n the
center of the range and a negati ve cor r ecti on at both ends, whereas not one
of the er r or -cor r ecti on functi ons changes si gn. The er r or at 03 = 0.6 can
be tol erated; l et us ther efor e make no cor r ecti on at thi s poi nt and attempt
onl y to r educe the er r or s at the ends of the range. We shal l i n fact desi gn
192 LI NKAGE COMBI NATI ONS [SEC.66
the i nput and output transformers so that nei ther changes the gener ated
functi on at % = 0.6, taki ng
b
.
= E? = 0.2872,
a
(51)
d
= E; = O.395O.
z
The form of the cor r ecti on made i n each termi nal transformer (F I and
FZ i n Tabl e 68) i s thus fi xed; i t remai ns to determi ne the constants g and c
wi th whi ch these shoul d be added.
Next, l et the (approxi mate) er r or at es = 0.2 be requi red to vani sh:
0.0469a 0.1503c = 0.0033. (52)
The er r or s at th = 0.8 and 83 = 0.9 wi l l then be
&g = 0.0052 0.3130a 0.1824c = 0.0092 0.2561a, (53a)
6,.9 = 0.0072 0.4559a 0.1538c = 0.0106 0.4079a. (53b)
I t i s evi dent that for best resul ts one must use a negati ve a, and al l ow a
negati ve er r or at 193= 0.8, a posi ti ve one at 93 = 0.9. An appropri ate
choi ce i s
a = 0.0306, c = 0.0315. (54)
We thus fi nd as constants for the l i nkage
L, = 7.438, L, = 3.398, E? = 0.2872, E; = 0.3950. (55)
The performance of thi s l i nkage i s shown i n Tabl e 6.9; i t i s about the best
that can be attai ned by adjustment of these four constants.
TABLE6.9.PERFORMANCE OF TEELI NKA.E [Et. (55)]
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0,0000
0.1285
0.2615
0.3948
0.5244
0.6461
0.7555
0.8487
0.9222
0.9734
1.0000
O.0000
0.0324
0.0674
0.1077
0.1570
0,2209
0,.3085
0.4334
0.6075
0.8185
1.0000
h~)r _ Ht
2
0.0000
0.0004
O.0000
0.0007
0.0015
0.0023
0,0023
0.0008
0.0015
0.0021
0.0000
!
I
SEC.6.7]
ASSEMBLY OF THE LI NKAGE 193
6.7. Exampl e: Assembl y of the Li i age Combi nati on.--The fi nal
step i n the mathemati cal desi gn of a l i nkage combi nati on i s to coordi nate
pr oper l y i ts component parts. Careful attenti on must be pai d to si gn
conventi ons and to the varyi ng zer o l i nes from whi ch angl es are measured
i n the several types of components.
I t i s safest to begi n by sketchi ng the component l i nkages i n thei r basi c
posi ti ons. Wi th each component ther e shoul d be i ndi cated the scal es
for the output parameters. I n our exampl e the l i nkages and scal es are
ful l y characteri zed as fol l ows:
I nput harmoni c transformer:
X,m = 17.5,
AX1 = 100,
82 i ncreases
Three-bar l i nkage:
X,~ = 186.2,
AX* = 100,
X,m= 10.1O,
L, 7.438,
E? = 0.287,
wi th X,.
~ = 0,630
A,
Az
x
= 0.630,
B2
z
= 0.668,
Ax, = 99,
100. t+ = 86.2 X3,
99 e, = 10.1 + x,.
Output harmoni c transformer:
X2* = 70, Lz = 3.398,
AX, = 99, E; = 0.395,
6, i ncreases wi th X2.
These l i nkages are sketched i n Fi g. 6.14; scal es of HI , H,, I %, and O, are
shown. Ther e i s an adjustabl e scal e constant i n the desi gn of each com-
ponent. The scal e constants of the harmoni c transformers (a and c,
respecti vel y) can be adjusted to control the travel s at the i nput and out-
put termi nal s; choi ce of the scal e constant of the three-bar l i nkage (b) i s
subject onl y to consi derati ons of mechani cal conveni ence.
I n the l i nkage combi nati on the readi ngs on the Ot-scal es of the i nput
harmoni c transformer and the three-bar l i nkage must al ways be the same.
We have desi gned the two t%scal es to cover the same angul ar range, but
the si gn conventi ons have for ced us to al l ow the &scal e of the three-bar
l i nkage to i ncrease countercl ockwi se, whereas that of the transformer
i ncreases cl ockwi se. These components mi ght, for i nstance, be con-
nected by the geari ng i ndi cated i n Fi g. 6.14. On the other hand, the
@l -scal es of the three-bar l i nkage and the output transformer i ncrease
. ..
194
LI NKAGE COMBI NATI ONS
[SEC.6.7
f33=o.o
gla
~%
0.287g1a)
HI =0.0 0.5 1.0
L:
1-
.
-..J=oc-
-
Fm. 6.14.COI I
/
+
s
bi
\
ei i,. ,/
o.395g2c
.-
Lponentaof the tangent l i nkage.
/ -17.5
Fm. 6.15.4ompl eted l i nkage mechani zi ng zz = tan ZI , O < z, <80.
SEC.6%] THE DOUBLE THREE-BAR LI NKAGE 195
i n the same sense; one possi bl e method of connecti ng these components
I i s i ndi cated i n Fi g. 6.14. Thi s compl etes the prel i mi nary representati on
of the l i nkage combi nati on.
Fi nal l y, one must conver t the prel i mi nary representati on i nto a prac-
1
ti cal desi gn wi thout changi ng the essenti al rel ati ons of the components.
One possi bl e arrangement of thi s tangent l i nkage i s shown i n Fi g. 6.15.
The three-bar-l i nkage component i s r otated through 90 from i ts posi ti on
i n Fi g. 6.14, l argel y to gai n cl ari ty i n the representati on. The output
crank of the three-bar l i nkage and the crank of the output transformer
are made to r otate together as arms of the same bel l crank. I n or der to
I
use the same type of connecti on between the i nput transformer and the
three-bar l i nkage, we must r ever se the sense of rotati on of one or the other
I
of these cranks. Thi s can be done by refl ecti ng the i nput transformer
and i ts associ ated scal es i n a verti cal l i ne. The two cranks can then be
I
joi ned i nto a bel l crank, and the l i nkage appears as i n Fi g. 6.15.
THREE-BAR LI NKAGES I N SERI ES
I t i s not desi rabl e to use harmoni c transformers i n a computer i n whi ch
al l vari abl es are r epr esented by shaft rotati ons si nce the l i near moti on of
the i nput or output sl i des must then be transformed i nto a r otar y moti on
by a r ack and pi ni on; i t i s much better to take the r otar y moti on di rectl y
from a rotati ng termi nal . Thi s remai ns tr ue even when the angul ar
travel must l ater be i ncreased si nce thi s can be accompl i shed by gears that
permi t a mor e compact desi gn than the r ack and pi ni on.
For such computers a si ngl e three-bar l i nkage i s i deal , except that i t
does not permi t generati on of a suffi ci entl y l arge cl ass of functi ons to cover
al l practi cal cases. Systems of two or mor e three-bar l i nkages provi de
gr eater fl exi bi l i ty, together wi th the same sati sfactory mechani cal
characteri sti cs.
6.8. The Doubl e Three-bar Li nkage.-I n a doubl e thr~bar l i nkage,
such as that sketched i n Fi g. 6.16, the homogeneous i nput parameter 01
i s transformed i nto an i ntermedi ate parameter os by the fi rst three-bar
l i nkage; thi s serves as the i nput to the second three-bar l i nkage, whi ch
generates the output parameter oZ. I n the oper ator symbol i sm,
,
(0,10,) = (e,le3) . (e,le,). (56)
The three-bar-l i nkage oper ator s are each characteri zed by fi ve constants,
(AX,, AX3, bl ,, bz,, a,) and (AX,, AX,, b,,, b,,, a,), respecti vel y. Si nce
the l i nkages must have a common val ue of the constant AXS, the number
of di sposabl e constants i n the combi nati on i s ni ne, The desi gn probl em
i s to choose oper ator s (&103) and (Oal Ol )such that thei r product (OZI OI )
I
approxi mates as cl osel y as possi bl e to the gi ven functi on
I
h, = (h,ph) oh, (57)
196 LI NKAGE COMBI NATI ONS [SEC. 6+3
on di rect or compl ementary i denti fi cati on of the vari abl es 01, 02 wi th the
vari abl es h,, hz.
Formal l y thi s probl em resembl es cl osel y that of desi gni ng a doubl e
harmoni c transformer, and the general
approach to i t i s the same.
For i n-
stance, one can appl y the method of
successi ve approxi mati ons descri bed
e,
i n Sec. 4.13. I n each stage of the pr o-
cedur e one must then fi t a three-bar-
l i nkage functi on of speci fi ed AX, to a
known functi on by an appl i cati on of
F1~.6.16.Doubl ethree-barl i nkage.
the nomographi c or geometr i c method.
Asi de from thk i ncrease i n mani pu-
l ati ve di ffi cul ti es, the pri nci pal di fference between thi s probl em and
the earl i er one l i es i n the fi rst step, i n whi ch one must make an i ni ti al
F1rJ. 6.17.Constani s of a doubl e three-bar l i nkage mechani zi ng the l ogari thmi c rel ati on
fOr l <z <50.
choi ce of one of the factor operators.
I t i s to be noted that thi s choi ce
fi xes a val ue of AX~ whi ch wi l l be used throughout the desi gn pr ocedur e.
SEC. 6.8] THE DOUBLE THREE-BAR LI NKAGE 197
One can begi n by usi ng an oper ator (0, l &), for exampl e, whi ch by
i tsel f gi ves a rough fi t to the gi ven oper ator ; the second factor wi l l then
ser ve to make rel ati vel y smal l correcti ons. Thi s mocedur e l eads to the
desi gn of combi nati ons i f qui te di fferent l i nkages, ~uch as that i l l ustrated
i n Fi g. 6.16.
A general l y sounder pr ocedur e i s to tr y to fi nd a combi nati on of mor e
or l ess si mi l ar l i nkages whi ch wi l l make roughl y equal contri buti ons to the
curvature of the gener ated functi on (cf. Fi g. 6.17). An appropri ate begi n-
Fm. 6.l S.A possi bl e physi cal for m for the l ogari thmi c l i nkage.
ni ng i s then made by factori ng the gi ven oper ator i nto the product of two
i denti cal oper ator s W:
(h,ph) = w w = w. (58)
The oper ator W, cal l ed the square-root oper ator , has been di scussed
i n Chap. 3, wher e i t has been shown that i t i s not uni quel y determi ned.
I f any one of the square-root oper ator s can be mechani zed by a three-bar
l i nkage m-th equal input and output travels, then two of these l i nkages i n
seri es wi l l gener ate the gi ven functi on. I f onl y an approxi mate mecha-
ni zati on of W can be found, the correspondi ng oper ator can at l east ser ve
M a fi rst approxi mati on to (831d,), wi th whi ch to begi n appl i cati on of the
method of successi ve approxi mati ons.
198 LI NKAGE COMBI NATI ONS [SEC.6.8
An exampl e of a l i nkage obtai ned by use of the square-root oper ator
W i s provi ded by a patented l i nkagel mechani zi ng the rel ati on
X2 = l ogl o Z1, l SZI S 50, (59)
wi th an er r or ever ywher e l ess than 0.003 of the output travel . I n desi gn-
i ng thi s, a three-bar l i nkage was used to mechani ze, wi th good approxi ma-
ti on, that one of the square-root oper ator s W whi ch has deri vati ves at the
ends of the domai n. Two such l i nkages i n seri es gave a mechani zati on
of the gi ven functi on whi ch was suffi ci entl y good to permi t i mmedi ate
appl i cati on of the methods of Chap. 7 i n a fi nal adjustment of the con-
stants of the l i nkage combi nati on. The fi nal l i nkage i s shown sche-
mati cal l y i n Fi g. 6 17; the component l i nkages have si mi l ar, but not
i denti cal , constants. The angl e B of the combi nati on can be chosen at
wi l l . Fi gure 210 shows the l i nkage obtai ned on setti ng B = 14234.
A mechani cal l y preferabl e form i s that shown i n Fi g. 618, i n whi ch the
two l i nkages sha;e the i ntermedi ate crank:
@=()
and 2.37259a =
1A. Svoboda, U.S. Patent 2340350, Feb, 1, 1944.
1.80124b.
fJI 14PTER 7
FI NAL ADJUSTI VI EMT OF LI NKAGE CONSTANTS
7.1. Rol es of Graphi cal and Numeri cal Methods k Li nkage Des@.
The pr ecedi ng chapters have been concer ned wi th methods for l i nkage
desi gn that are l argel y graphi cal , rather than numeri cal . @aphi cal
methods are easi l y appl i ed, and have the i mportari t vi rtue of maki ng
evi dent the character of the over -al l fi t to the gi ven functi on, not mer el y
the fi t at a sel ected set of poi nts. Thei r accuracy, however , i s l i mi ted;
when hi gh accuracy i s r equi r ed, the ti al adjustment of l i nkage constants
must be carri ed out by numeri cal methods because these al one permi t
suffi ci entl y careful adjustment of the constants and suffi ci entl y accurate
eval uati on of the performance of the l i nkage.
Numeri cal methods, on the other hand, tend to be excessi vel y compl ex,
except when the y rel ate to changes i n l i nkage constants so smal l that one
can assume that the er r or functi on depends l i nearl y on each of these
changes. Graphi cal methods are thus ver y i mportant i n maki ng i t possi -
bl e to fi nd, qui ckl y and easi l y, a l i nkage wi th constants whi ch need to be
changed onl y a l i ttl e to bri ng i ts structural errors wi thi n the speci fi ed
tol erances of the probl em; i t i s onl y at thi s poi nt that numeri cal methods
become effecti ve and conveni ent.
I n general , then, graphi cal methods are desi rabl e for the fi rst stages
of l i nkage desi gn, whi ch must yi el d a l i nkage wi th smal l er r or over the
whol e range of travel . The er r or can then be further r educed by numeri -
cal methods; often i t can be made to vani sh at several sel ected poi nts.
Thi s was, for i nstance, the method empl oyed i n Sees. 4.7 and 4.15.
The present chapter wi l l provi de a general di scussi on, for l i nkages wi th
one degr ee of freedom, of the probl em of maki ng fi nal adjustmen~s of al l
di sposabl e constants of a l i nkage. I t wi l l be a basi c assumpti on that the
structural er r or at any poi nt i s nearl y a l i near functi on of each of the
vari ati ons of constants to be consi dered. Thus the di scussi on wi l l i n
most, but not al l , cases appl y onl y to smal l changes of the constants.
Someti mes these methods are conveni ent even when an i mproved basi c
outl i ne of the system i s to be obtai ned by a substanti al change i n some
constant. Such may be the case when the graphi cal method has been so
appl i ed that i t does not establ i sh a near opti mum desi gn wi tbi n a whol e
cl ass of l i nkagesfor i nstance, when a combi nati on of a three-bar l i nkage
and harmoni c tranaf ormers has been desi gned wi th fr ozen angul ar travel s,
199
200
FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS [SEC.7.2
and one must consi der the possi bi l i ty of maki ng fai rl y l arge changes i n
these travel s.
The chapter wi l l concl ude wi th a di scussi on of a qui te di fferent method
of reduci ng structural er r or s, whi ch i s parti cul arl y useful after the usual
numeri cal methods have been appl i ed: the i ntroducti on of smal l cor -
recti ons by the eccentr i c l i nkage.
7,2. Gaugi ng Parameters.-Let us consi der the probl em of checki ng
the performance of a l i nkage desi gned to mechani ze a gi ven rel ati on
between vari abl es xl and x2:
X2 = (Z21ZJ Z1. (1)
The l i nkage wi l l gener ate a rel ati on between an i nput parameter Xl and an
output parameter X2:
X2 = (X2] X,) oxl . (2)
The form -of the oper ator (X21X1) wi l l depend upon di mensi onal constants
of the l i nkage, the preci se nature of whi ch we need not speci fy. We
denote these by go, gl , gz, . . . g~d.
At the i nput termi nal ther e wi l l be a
l i near scal e whi ch rel ates the val ues of the i nput vari abl e and the i nput
parameter:
xl = X!) + l rl (z, Zf)), (3)
X~O)and z~j bei ng correspondi ng val ues of these quanti ti es. At the out-
put termi nal ther e wi l l be a si mi l ar scal e rel ati ng the output parameter to
the actual l y gener ated (not the i deal ) val ues of the output vari abl e.
Denoti ng by xz~these actual output val ues of the mechani sm, we have
X2 = Xp + lr2(z2a zig), (4)
X~OJand X~~Jagai n bei ng correspondi ng val ues. The l i nkage and scal es
together gener ate a rel ati on between z, and x2., whi ch depends on the
constants of the l i nkage and on the four addi ti onal constants, k 1, X!o), RZ,
and Xi o), whi ch characteri ze the termi nal scal es.
These l atter constants
we denote al so by g~a, g~g, g~1, and g..
We have then
x*C = F(z], go, gl, . .
9n),
(5) \
a functi on of the i nput vari abl e and n + 1 constants of the mechani sm.
Perhaps the most obvi ous way to study the structural er r or of the
mechani sm i s to compare the desi red and the actual l y gener ated val ues of
the output vari abl e for a spectrum of val ues of the i nput vari abl e,
2y), Z\), . . .
~y) .
The correspondi ng spectrum of val ues of X2 i s determi ned by Eq. (1):
& = (X2121). Zy); S=o, l ,. ..r (6)
I
I
SEC.7.3] USE OF THE GAUGI NG PARAMETERS 201
Si mi l arl y, Eqs. (2), (3), and (4) determi ne spectra of val ues of X,, X,, and
Xti . I n parti cul ar,
x$: = F(xy), go, gl ,
- 9n);
S=o, l ,. ..r.
(7)
The structural er r or , ~z2, of the mechani sm has the spectrum
&@) = -@: &;
S=o, l ,. ..r.
(8)
The cor r ecti ons whi ch one woul d l i ke to make i n the output of the mecha-
ni sm are the negati ve of these quanti ti es.
I n such a test a compari son of the i deal and the actual l y gener ated
val ues of X2i s used as a gauge of the preci si on of the l i nkage; we shal l say
that X2i s used as the gaugi ng parameter.
I t i s not at al l necessary to use X2 i n the gaugi ng process. I n most
cases thi s i s not even desi rabl e; i t i s better to use as a gaugi ng parameter
one of the di mensi onal constants of the l i nkage, go, g1 . . . g,. Let us
sol ve Eq. (7) for thi s gaugi ng parameter, for i nstance go:
90 = G(W, x% gl , gz, c c . g).
(9)
I f we substi tute on the ri ght any correspondi ng val ues of X1 and xZ., we
shal l compute al ways the same val ue of gothe actual val ue of thi s con-
stant i n the l i nkage consi dered. I f, however , we use i deal val ues of x2,
~~), i n5tead of the actual l Y gener ated val ues, x~~, gowi l lnot i n general
have a constant val ue, but i nstead a spectrum of val ues,
gy) = G(x~), z$J, gl , g.z, o . . gfi ); 8 =0,1,2,...?-. (l o)
The di fference between the actual val ue go of the constant and the val ue
g:) whi ch i t woul d need to have to make the l i nkage exact at the poi nt s
we shal l cal l the gaugi ng er r or ,
ti gf) = go g$~,
S=o, l , . . . ?- . (11)
Such quanti ti es are useful as gauges of the preci si on of l i nkages, al though
they do not gi ve di rectl y the er r or at the output. A wi sel y chosen
gaugi ng parameter i s usual l y si mpl er to cal cul ate and easi er to i nterpret
(at l east as regards desi rabl e changes i n the constants) than i s the er r or at
the output; i n parti cul ar, i f bg~)i s i ndependent of s i t i s onl y necessary
to r educe goby thi s amount to make the l i nkage exact. That the pr oof of
per fect performance of the l i nkage i s r educed to demonstrati on of the
constancy of the resul ts of a seri es of computati ons i s al so of val ue for the
avoi dance of computati onal er r or s.
7.3. Use of the Gaugi ng Parameter i n Adjusti ng Lti age Constants.
I n the precedi ng chapters we have seen how to desi gn l i nkages wi th smal l
gaugi ng er r or s. I t may sti l l be desi rabl e to i mprove these l i nkages by
maki ng smal l vari ati ons i n the di mensi onal constants go, gl , gz, . . . g..
202 FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS [SEC.73
A per fect l i nkage wi l l be obtai ned i f val ues of gl , . . . g., can be found
such that g%)as computed by Eq. (10) i s the same for al l possi bl e sets of
val ues (2$), 2$)). I n general , one can at best hope to make go constant
at a preassi gned set of poi nts equal i n number to the i ndependent con-
stants of the l i nkage and thus to obtai n a l i nkage whi ch generates the
gi ven functi on exactl y at these poi nts.
I f the di mensi onal constants are changed by amounts A9i , becomi ng
g: = g; + Agi,
i=(), l .. .~,
(12)
the gaugi ng parameter wi l l have the spectrum of val ues
9:() =
G(zy), x$), g;, gj, . g:), (13)
and the gaugi ng er r or wi l l become
ag;(1 = g: g;(), S=o, l ,. ..p. (14)
Expandi ng Eq. (13) i n a Tayl ors seri es, we may wri te
i n
d2gf) . A9i Aqi + . ,
+jj
zz
agi dgi
(15)
~=lj=l
the parti al deri vati ves bei ng eval uated at (z~), Z$), gl , 92, . . . g.). The
gaugi ng er r or can thus be wri tten as
~gf) = go + Ago g$)
I t i s desi red to r educe thi s to zer o at a chosen set of p + 1 preci si on poi nts:
s=o, 1,2, . ..p.
The general sol uti on of thi s probl em i s prohi bi ti vel y di ffi cul t, and i t i s
necessary to make an approxi mati on whi ch wi l l be val i d onl y i f the requi red
changes i n the constants are suffi ci entl y smal l . Terms i n Eq. (16) of
hi gher than the fi rst or der i n the smal l quanti ti es Ag; wi l l be negl ected. Let
Then, by use of Eq. (11) one can r ewr i te Eqs. (16) thus:
.
z
G~)Agi = ~g~) ;
S=o, l ,. ..p.
;..0
(17)
(18)

SEC. 7.3] USE OF THE GAUGI NG PARAMETERS 203

A set of Agi s whi ch sol ve these equati ons wi l l ser ve as cor r ecti ons to the
ori gi nal l y chosen gi , as i ndi cated i n Eq. (12), under restri cti ons whi ch
must be di scussed.
One can sol ve thi s system of l i near equati ons for the Agi i f the ranks of
the matri x of coeffi ci ents [G:)] and the augmented matri x ([G~)] wi th the
added col umn ~g~)) are equal . I n l ess preci se but mor e di rect terms,
the equati ons wi l l usual l y be sol ubl e i f and onl y i f the number of i nde-
pendent constants characteri zi ng the gener ated functi on i s equal to or
gr eater than the number of equati ons, p + 1. I t woul d be natural to
i nfer from thi s statement that the l i nkage can be made to gener ate a gi ven
functi on exactl y at m arbi trari l y chosen poi nts whenever the gener ated
functi on i s characteri zed by m mathemati cal l y i ndependent constants,
(m s n + 1). I n practi ce i t wi l l be found that thi s i s not the case; the
number of preci si on poi nts whi ch can be obtai ned depends upon the
nature of the l i nkage and the gi ven functi on, and on the way i n whi ch
the preci si on poi nts are chosen.
Even when the l i nkage under consi dera-
ti on i s wel l adapted to generati on of the gi ven functi on one must often
be content to fi x fewer than m preci si on poi nts, or to use other methods
of reduci ng the er r or .
Thk di fference between the mathemati cal probl em of sol vi ng Eqs. (18)
and the practi cal probl em of fi ndi ng a l i nkage wi th p + 1 preci si on poi nts
ari ses from the fact that Eqs. (18) are mathemati cal approxi mati ons val i d
onl y for suffi ci entl y smal l Agi .
For practi cal purposes one must not
onl y solve Eqs. (18), but must sol ve them with Agi which are so small that the
quadratic terms in i?qs. (16) are negligible. We have seen i n Sec. 66
(see Tabl e 6.6) how di fferent may be the expected and the actual per -
formance of a l i nkage desi gned by usi ng approxi mate l i near equati ons ver y
si mi l ar to Eqs. (18), when the Agi are so l arge that negl ect cd terms are
i mportant.
Di ffi cul ti es are most l i kel y to ari se i n the ~trai ghtforward appl i cati on
of Eqs. (18) when the restri cti on to smal l Agi has, for practi cal putposes,
the effect of establ i shi ng a rel ati on between mathemati cal l y i ndependent
parameters.
To si mpl i fy the di scussi on of thi s poi nt we shal l assume that the
parameters gi (z = O, 1, 0 n) whi ch occur i n thi s equati on are al l i nde-
pendent of each other . One can then attempt to fi x n + 1 preci si on
poi nts, determi ni ng the Ag~by sol vi ng n + 1 of Eqs. (18). Because of the
i ndependence of the parameters, the determi nant of the coeffi ci ents Gjg)
wi l l not vani sh; the sol uti on for the Agi wi l l be uni quel y expressi bl e as a
fracti on i n whi ch the numerator i s the determi nant I G$)I wi th one col umn
repl aced by the col umn of coeffi ci ents ~g~), and the denomi nator i s the
determi nant l G~) I i tsel f. The smal l er the gaugi ng errors ~g$ the smal l er
wi l l be the Agi . However , even when the ~g$) are ver y smal l i t may be
204 FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS [SEC. 7.3
found that the Ag~are l arge and that the l i nkage wi th constants gi ven by
Eq. (18) does not have the desi red preci si on poi nts, or even an i mproved
performance. Thi s happens most frequentl y when the determi nant
l Gja)1, al though not exactl y zer o (as i t woul d be i f ther e wer e an exact
rel ati on between the parameters), i s ver y smal l . I n such cases one can
make l arge and pr oper l y rel ated changes i n the parameters whi ch pr oduce
onl y a smal l net change i n the gener ated functi on. For i nstance, as
i l l ustrated bel ow, i t may be possi bl e to make l arge changes i n two param-
eters, gi and gj, whi ch wi l l change the gener ated functi on ver y l i ttl e i f
gi /gi i s hel d constant. When one restri cts
attenti on to smal l changes i n the parameters
the gener ated functi on then depends, i n
effect, on a smal l er number of parameters;
i n our exampl e i t woul d depend, not on gi
and gj i ndi vi dual l y, but onl y on gi /gj.
Thus the number of effecti vel y i ndependent
parameters may be decreased by restri cti ng
consi derati ons to smal l A~i , and wi th i t the
number of preci si on poi nts whi ch one can
Bz
hope to establ i sh.
n
X2
0=
An al most tri vi al exampl e i s i l l ustrated
by Fi g. 7.1. A si ngl e pi voted arm (Fi g.
B1
A2
7-l a) can be used i n the generati on of l i near
functi ons. The mechani sm i tsel f i nvol ves
Al
no adjustabl e constant. The i nput scal e i s
(b)
characteri zed by the two parameters kl and
FI G.7.1.(a) Mechani zati on
X~OJ ,the output scal e by the two parameters
of a l i near functi on. (b) Mechan-
kz and X~OJ. The gener ated l i near functi on
i zati on of an al most l i near
functi on.
i s characteri zed by onl y two i ndependent
constants; equal changes i n X~O)and X~O)or
proporti onal changes of al l four vari abl es do not pr oduce any change i n
the gener ated functi on. I n usi ng such a devi ce as a mechani zati on of an
al most l i near functi on one cannot i n gener al r educe the er r or to zer o at
mor e than two preassi gned poi nts. Now consi der the three-bar l i nkage
i n Fi g. 7. l b. I t i s al most a paral l el ogram l i nkage, and generates an al most
l i near rel ati on between xl and xaone whi ch i s characteri zed by seven
mathemati cal l y i ndependent parameters. The determi nant I G$J1, wi th
seven rows and col umns, wi l l not vani sh; i t wi l l , however , be ver y smal l ,
and vani sh as the paral l el ogram condi ti on, ~1 = A a, 13Z = A 1, i s attai ned.
I t i s obvi ous that equal changes of X~) and X;) wi l l pr oduce ver y smal l
changes i n the gener ated functi on, and that pr opor ti onal changes of klj ka,
X~O~and X~O)wi l l have a si mi l arl y smal l effect. Conver sel y, certai n smal l
changes i n the gener ated functi on wi l l be obtai nabl e onl y by maki ng such
&EC. 7.5] LARGE VARI ATI ONS OF DI MENSI ONAL CONSTANTS 205
l arge changes i n these parameters that the l i near theor y wi l l not appl y.
I f we excl ude l arge Ag~from consi derati on ther e are i n effect two fewer
degr ees of fr eedom than one mi ght expect; an attempt to establ i sh seven
preci si on poi nts wi l l be l i kel y to fai l , al though fi ve shoul d be obtai nabl e
i f the i ni ti al fi t i s good. Such i s usual l y the case wi th three-bar l i nkages,
whi ch wi l l r ecei ve mor e detai l ed di scussi on i n Sec. 7.7.
704. Smal l Vari ati ons of Di mensi onal Constants.-I t i s usual l y desi r-
abl e to appl y the approxi mate l i near form of the gaugi ng-parameter
method, even when one must r educe the number of preci si on poi nts i n
or der to deal wi th smal l vari ati ons of the di mensi onal constants.
Let the number of i ndependent di mensi onal constants be n + 1, and
the number of speci fi ed preci si on poi nts be k l ess than thi s. I t i s then
possi bl e to sol ve any n + 1 k i ndependent equati ons from among
Eqs. (18) for any n + 1 k of the Ags, i n terms of the other k of these
quanti ti es. For i nstance, sol vi ng for Ago, Agl j . . . Agm_~,i n terms of
Ag~_h+l , . . . Agn, one obtai ns rel ati ons of the form
Ago = COO + Cal Ag_~+l + . . + CO~Ag.,
Agl = C1O+ C,,Agn~+l + + C,kAg.,
/ (19)
. . . . . . . . . . . . . . . . . . . . . . ., .,., . . . .
)
.,
Ag.k = C.k,. + c._k,,Ag.k+, + + Cnk, kfifh.
Any set of small Ags whi ch sati sfi es Eqs. (19) wi l l consti tute a val i d and
practi cal l y useful sol uti on of the gi ven probl em. Such a sol uti on i s not
mathemati cal l y uni que, but i t wi l l be effecti vel y so i f one i s attempti ng to
establ i sh the maxi mum number of preci si on poi nts subject to arbi trary
choi ce.
7.5. Large Vari ati ons of Di mensi onal Constants.-By sol uti on of
Eqs. (18) one can determi ne a set of changes Ago, Ag,, . . Agm,i n the
di mensi onal constants whi ch reduces to zer o the fi rst-order terms i n the
gaugi ng er r or . Equati ons (16) become, exactl y,
nn
L!gp = ;
n
~2gg8)
Agi Agi +...,
8gi 8gi
S=o, l ,p. (20)
asl jxl
When thi s gaugi ng er r or i s negl i gi bl e we say that the Ags are smal l ; the
probl em has been sol ved i n the fi rst step. We turn now to the case i n
whi ch 8g~@)i s not negl i gi bl e, but i s adequatel y r epr esented by the second-
or der terms wri tten out i n Eq. (20). I n such cases the desi gn probl em i s
not sati sfactori l y sol ved by a fi rst appl i cati on of the method of Sec. 7.3,
but i t can often be sol ved by successi ve appl i cati ons of the method, whi ch
pr oduce successi ve i mprovements i n the di mensi onal constants. We shal l
now see how the conver gence of thi s process can be hastened.
206 FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS
[SEC.7.6
Knowing the Ags whi ch sol ve Eqs. (18), and the gaugi ng er r or 6g~@)
after these cor r ecti ons are made, one can easi l y compute al so the gaugi ng
er r or s resul ti ng when proporti onal l y l arger or smal l er changes are made i n
the di mensi onal constants. Let a new set of cor r ected di mensi onal
constants be gi ven by
g;(x) = gi + XAgi ,
i =o, l ,. ..p. (21)
I n the Tayl ors seri es expansi on for the gaugi ng parameters, Eq. (15), the
fi rst-order terms are then changed by a factor A, the second-or der terms by
a factor AZ, and so on. The resul ti ng gaugi ng er r or i s
n nn
2
G~)Agi _ ~ i 2
2s
132g&J
~g;() = go g$) + XAgo A
~ Ag&gi
c?g,tl gi
i = I ;cl i zl
+
. . .
, (22)
or , by appl i cati on of Eqs. (12), (18), and (20),
Jg:( = (1 k) ag$) + X%gr), S=o, l ,. ..p. (23)
The val i di ty of thi s expressi on of cour se depends on the possi bi l i ty of
negl ecti ng hi gher-order terms i n Eq. (20).
When A = O the di mensi onal constants and gaugi ng er r or have thei r
uncor r ected val ues. I ncrease i n A wi l l r educe the gaugi ng er r or 6g{@)
so l ong as the quadrati c term i n Eq. (23) remai ns negl i gi bl e. As A
approaches 1 the quadrati c term wi l l eventual l y (by our assumpti ons)
become appreci abl e, and may even become ver y l arge. I t i s evi dent
that one can obtai n a smal l er gaugi ng er r or by appl yi ng a fracti on of the
cor r ecti on i ndi cated by the l i near theor y (O < k < 1) than by appl yi ng
the whol e cor r ecti on (k = 1). The mor e i mportant the quadrati c terms
the smal l er wi l l thi s fracti on be; i t i s, however , al ways possi bl e to fi nd
some posi ti ve val ue of i whi ch gi ves a better set of constants than ei ther
k= Oor X=l .
I n practi ce one begi ns wi th knowl edge of dg$), and computes the
Ags. As a check on the val i di ty of the cal cul ati on one shoul d then
determi ne the val ues of ~g~), usual l y by di rect cal cul ati on [Eqs. (13)
and (14)] rather than by use of Eqs. (20). I f these quanti ti es are not
sati sfactory smal l , one shoul d make a smal l er change i n the gs; the
appropri ate val ue of X can be determi ned by use of Eq. (23), A bei ng
chosen to make the quanti ti es ~g~()as smal l as possi bl e. The constants
g{ computed by Eq. (21) wi l l then ser ve as i ni ti al val ues for a second
appl i cati on of the method.
706. Method of Least Squares.The desi gners ul ti mate objecti ve i s
to assure that the output er r or
8X2 = X2. X2 (24)
SEC.7.7]
GAUGI NG THE THREE-BAR LI NKAGE 207
shal l be kept smal l throughout the domai n of operati on of the mechani sm.
One way to assure thi s i s so to choose the di mensi onal constants of the
mechani sm, on whi ch 6x2 depends, as to mi ni mi ze the i ntegrated squared
er r or ,
I(go, gl, gz, gn) =
\
(6x,)2 dx,, (25)
or the correspondi ng sum over a di screte spectrum of output val ues,
(26)
Such condi ti ons are most reasonabl e when accuracy i s equal l y i mportant
for al l val ues of the output vari abl e, or al l val ues of r. Mor e general l y,
one shoul d i ntroduce a wei ghti ng functi on, w(x2) or w(r), whi ch i ncreases
wi th the i mportance of accuracy i n the resul t at the correspondi ng X2or r.
One wi l l then so choose the gs as to mi ni mi ze
/
I w= [ I O( Z2) 6X212dx%
or
(27)
(28)
subject to any other condi ti ons whi ch must be i mposed on the di men-
si onal constants.
Least-squares methods sui tabl e for use i n l i nkage probl ems have been
devel oped by K. Levenburg. I t i s, however , the opi ni on of the author
that l east-squares methods are rel ati vel y unrewardi ng. I n parti cul ar,
when the method depends on the use of an expansi on i n whi ch onl y
l i near terms are retai ned ther e i s al ways the danger that a resul t obtai ned
after a l arge expendi ture of l abor may be i nval i dated by thi s approxi -
mati on. I n general the author pr efer s to set tol erances on the output
er r or tol er ances whi ch may vary wi th X2or r-and to appl y the methods
of the precedi ng secti ons to bri ng the actual structural er r or s wi thi n
these tol erances.
7.7. Appl i cati on of the Gaugi ng-parameter Method to the Three-bar
Li nkage. Formulation of the Equations. -I n appl yi ng the gaugi ng-param-
eter method to three-bar l i nkage desi gn we may choose the di mensi onal
constants as fol l ows:
1K. Levenbur g, A Method for the Sol uti on of Certai n Nonl i near Probl ems i n
Least Squares, Quart. Appl. Math., 2, 164(1944).
2!I !? FI NAL .4DJ USTMENT OF LI NKAGE CONSTANTS
[SEC. 7.7
( )
B,
x
B,
Z
A,
x
x~o),
k,,
x~o),
k,.
(29)
As gaugi ng parameter we shal l use go. I n effect, we shal l gauge the
er r or of a desi gn by computi ng the
r equi r ed l ength B~l of the connecti ng
bar as a functi on of the other di men-
si onal constants and the vari abl e pai rs
~ (xY), ZY)); we shal l seek to make con-
stant the rel ated gaugi ng parameter,
()
~
g~) = ~1 .
(30)
k--+
A spectrum of
FI G. 72.-Three-bar l i nkage.
chosen accor di ng
rul e, symbol i zed by
1 = (zY)I s) . ~.
z!
The spectrum of ZZi s then determi ned:
Zy) = (Z21ZI ). zyJ.
Equati ons (3) and (4) become
Xy = g3 + g,(zy Zp)
w = g6 + gdx$) x!)).
. -,
val ues of Z1 can be
to some arbi trary
(31)
(32)
(33)
(34)
Let the hori zontal separati on of the ends of the connecti ng bar, i n terms
of the uni t A 1, be U(), and the verti cal separati on, i n the same uni ts,
be V(). Then by the geometr y of the l i nkage (Fi g. 7.2) we have
u() = 1 + g~Cos Xy) g~Cos X!),
(35)
V@) = g] si n X~) gz cos X~),
g$) = (u()) + (V() )2.
(36)
(37)
An equati on of the form of Eq. (9) coul d be obtai ned by el i mi nati ng
from Eqs. (33) to (37) the quanti ti es X~), X$), U(), and V(8). Thi s,
however , i s not necessary for our purposes.
SEC. 7.8]
GAUGI NG A THREE-BAR LI NKAGE
2@9
The parti al deri vati ves
(38)
wi l l now be gi ven i n a form sui tabl e for numeri cal cal cul ati on:
&&&
2g,
1 Gf) (Zyl_ ~.o)) = Qf)
(43)
(,) _ U(S)si n Xy) = Q$)
(44)
& G:) = V@J COSXz
Zg,
-*Gg) (-+G&))x x)) Q$) 45)
I t i s to be remembered that al l angl es are expressed i n radi ans, and that
ga, gd, g6, and g6 must be i nterpreted correspondi ngl y.
One can use the quanti ti es Q? di rectl y i n the sol uti on of Eqs. (18).
On i ntroducti on of the quanti ti es
(46)
whi ch are si mpl y constant mul ti pl es of the Ags, Eqs. (18) become
n
2
Q~)Aqi = &g~), S=o, l ,. ..p.
(47)
;=O
Havi ng sol ved Eqs. (47) for the Aqs, one can compute the Ags by Eqs.
(46).
7.8. Appl i cati on of the Gaugi ng-parameter Method to the Three-bar
Li nkage. An Exampl e.As an exampl e of the gaugi ng-parameter
method we shal l check and
by the geometr i c method i n
the rel ati on
I
i n the domai n 1 s Z1 S 10.
,5.1,9~ be taken ~
i mprove the l ogari thmi c l i nkage desi gned
Sec. 519. Thi s was i ntended to gener ate
X2 = l ogl o xl (48)
The desi gn constants establ i shed i n Sec.
I
210 FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS [SEC. 7.8
( )
B, 2
go= ~
= 1.055002 = 1.11303,
\
g, = ~, = 0.70700,
g~ = * = 0.55000,
I
(49)
g, = X@ = 2.61798 (= 150.0000),
g, = k, = 0.10666 (= 55.0000/9),
g, = X$ = 2.03330 (= 116.5000),
g, = k = 1.57079 (= 90.000)
I
Al l constants are gi ven to the fi fth deci mal pl ace, or a thousandth of a
degr ee, si nce thi s number of di gi ts wi l l be carri ed through the further
cal cul ati ons.
We have fi rst to choose a sui tabl e spectrum of val ues for xl . A
uni form di stri buti on of val ues i n thi s spectrum woul d yi el d a rel ati vel y
TABLE7,1.CALCULATI ON OF ~g$)
8
0
1
2
3
4
5
6
7
8
9
10
.s
o
1
2
3
4
5
6
7
8
9
10
o.m
0.25892
0,58489
0.99526
1.51189
2.16228
2.98107
4.01187
5.30957
6.94328
9.00ooo
o .89493
O .95372
O .98902
O .99996
O .98629
O .94832
0.88701
O .80386
0.70091
o .58070
O .44620
Xp
degr ees
150. OCI O
148.418
146.426
143.917
140.761
136.786
131.782
125.483
117,552
107.569
95.000
Cos x:)
O .44620
0.30071
O. 14781
0.00873
0.16505
0.31730
0,46175
0.59482
0,71325
0.81412
0.89493
si n X~)
0.50000
0.52371
0.55306
0.58901
0.63256
0.68472
0.74568
0.81428
0.88659
0.95336
0.99619
0.84571
0.89481
0.93497
0.96641
0.98968
1.00567
1.01505
1.01782
1.01232
0.99341
0.94972
Cos XV)
O.86603
0.85189
0.83317
0.80817
0.77451
O.72880
O.66630
O,58046
0.46255
0,30185
0.08716
u(a)
0.63313
0.56310
0.49224
0.42382
0.36164
0.31022
0.27496
0.26246
0.28Q69
0.33883
0.44617
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
9P
1,11608
1.11777
1.11647
1.11357
1.11025
1.10761
1.10593
1.10484
1. 1035R
1.10167
1.10104
x:)
degr ees
116 50
107,50
98.50
89.50
80.50
71.50
62.50
53.50
44.50
35.50
26.50
0.00303
0.00474
0.00344
0.00054
0,00278
0.00542
0.00710
0,00819
0.00945
0.01136
0.01199
I
SEC. 7.8] GAUGI NG A THREE-BAR LI NKAGE 211
poor check i n the range of smal l x,, wher e fracti onal er r or s tend to be
greatest. I t i s better tochoose auni form di stri buti on of spectral val ues
for z,; we shal l take
Zp = 0.1.s
1
S= 0,1,2, . ..1O.
@ = 100.1.
(50)
The cal cul ati on of the gaugi ng er r or of thi s l i nkage i s shown i n Tabl e
7.1. The gaugi ng parameter g$) i s constant to wi thi n one per cent; the
requi red l ength of the connecti ng bar, {g;, i s constant to wi thi n one-
+0.010-
/
/
/
/
/
+0.005
SO
~
o
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 I .c
h!)
-0.005 ~
FIG. 7.3.Gauging error i n the tl rst l ogari thmi c linkage. Solid line, result of direct
calculation. Dashed line, an approximation with slowly varying curvature.
hal f per cent. Fi gure 7.3 shows the gaugi ng er r or ?g$) pl otted agai nst
the homogeneous i nput vari abl e h~). The gaugi ng er r or i s not l arge, but
i t i s evi dent that i t can be made much smal l er; thi s l i nkage has onl y
one poi nt of preci si on, whereas i t shoul d be possi bl e to obtai n fi ve (Sec.
73).
We can pr oceed i n the fol l owi ng way to make a reasonabl e cl oi ce of ,
the poi nts whi ch are to be establ i shed as poi nts of preci si on. Through
the cur ve of 6g&)i s drawn the dashed l i ne of Fi g. 7.3, whi ch fol l ows i t
cl osel y but wi th mi ni mum vari ati on i n curvature, i ntersecti ng i t i n fi ve
poi nts. I f we establ i sh these poi nts aa poi nts of preci si on, we wi l l
212 FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS [SEC. 78
be maki ng a change i n 6gfJ whi ch al so has sl owl y varyi ng curvature, and
whi ch must ther efor e approxi mate cl osel y to the dashed curve; the
resi dual gaugi ng er r or shoul d then be nearl y equal to the verti cal separa-
ti on of the two curves i n Fi g. 7.3. For conveni ence, l et us choose i nstead
to establ i sh poi nts of preci si on at s = 3, 6, 9, and 10. The fi fth poi nt
shoul d l i e between s = O and s = 1, and i t woul d not be enti rel y sati s-
factory to take ei ther of these as poi nts of preci si on. I nstead of taki ng
s = 0.5 as the fi fth poi nt, we can obtai n the same resul t by requi ri ng
that s = O shal l be, not a poi nt of preci si on, but a poi nt wher e ther e i s a
predetermi ned er r or : bgjf) = 0.0019, as read from Fi g. 7.3. That i s,
i nstead of sol vi ng Eqs. (18) or (47) wi th 8g$0J= 0.0031, hi ch woul d
make s = O a poi nt of preci si on, we shal l sol ve them wi th
We shal l choose to sol ve Eq. (47). On cal cul ati on of the Qs by Eqs.
(39) to (45), these equati ons take on the fol l owi ng form, for s = O, 3,6,9,
10, respecti vel y:
1.00000 Aq, 0.12545 Aql 1.03935 Aq, 1.04898 Aq~
+ O.00000 Aq, + 0.18925 Aq, + O.00000 Aq, = 0.00500,
1.00000 AqO+ 0.22671 Aql 0.96267 Aq, 1.03066 Aq3
1.02577 Aq, + 0.43224 Aq, + 0.12967 Aq, = 0.00054,
1.00000 AqO+ 0.57370 Agl 0.77340 AqZ 0.88136 Aq~
2.62740 Aq, + 0.71259 Aqs + 0.42755 Aq6 = 0.00710,
1.00000 Aq, + 0.84480 Aql 0.30102 Aqz 0.62289 AqS (51)
4.32490 Aq, + 1.00551 Aq, + 0.90496 Aq, = 0.01136,
1.00000 Aq, + 0.90721 Aql 0.02447 AqS 0.52725 AqS
4.74525 Aq, + 1.04901 Aq, + 1.04901 Aq, = 0.01199.
Si nce ther e are her e two fewer equati ons than ther e are vari abl es, i t i s
possi bl e to fi x two of the vari abl es arbi trari l y, subject onl y to the condi -
ti on that al l Aqs shal l be smal l . On el i mi nati ng Aq,, Aq,, Aqz, and Aqs
from these equati ons we obtai n
0.00294 Aq, 0.05528 Aq, 0.05390 Aq, = 0.00229. (52)
The coeffi ci ent of Aq, i s smal l ; Aq, can be chosen arbi trari l y wi th l i ttl e
effect on the rel ati on between Agb and Aqe. I t i s ther efor e reasonabl e to
set
Aq, = O. (53)
We can then sol ve Eqs. (51) for each of the Aqs i n terms of Aq,, fi ndi ng,
for i nstance,
AqS = 0.05486 0.8546S AqC, (54)
Aqs = -0.04142 0.97503 Aq6.
SEC. 7.8] GAUGI NG A THREE-BAR LI NKAGE 213
I ~Aq3 and Aq, are to be smal l , we must keep Aq6 smal l , si nce i ts coef-
fi ci ents are l arge. The best val ue of AqCi s approxi matel y zer o; aposi -
ti ve val ue wi l l i ncrease the magni tude of Aqs, a negati ve val ue that of
Aq8. We shal l ther efor e choose
Aqo = O,
and fi nd i n consequence
Aq, = 0.04512,
Aql = 0.04272,
Aq, = 0.01985,
Aq, = 0.05486,
Aq5 = 0.04142.
I
Ago = Aq, = 0.04512,
Agl = & Agl = 0.02136,
Ag, = ; Aq2 = 0.00992,
By Eqs. (46),
1
Ag? = ~1
Aq, = 0.03879,
AgA=+~Aq, =O,
2g,
Ag, = ~ Aq, = 0.03765,
2g,
Ag8=~Aq0 =0.
Zg,
1
Fi nal l y, by Eq. (12),
I g: = 1.06791,
g; = 0.72836,
g; = 0.55992,
g: = 2.65677 (= 152.2220),
g; = 0.1 0666,
(55)
(56)
(57)
(58)
g: = 1.99565 ( = 114.3420),
gj = 1.57079.
1
To check the performance of the l i nkage wi th the new constants g{, ye
compute the new gaugi ng er r or bg$). Thi s i s shown i n Tabl e 72,
together wi th the val ues
6
(fsg~)=. = bg$)
2
Qj@ Aq,
(59)
i =O
predi cted by a theor y i n whi ch onl y fi rst-order terms i n the ~qs are
retai ned. The di fference between these two quanti ti es, denoted by +),
represents the negl ected quadrati c and hi gher terms [Eq. (20)]. Fi gure
74 shows these quanti ti es graphi cal l y, @ appeari ng as the verti cal
separati on of the ful l and dashed l i nes.
214
8
0
1
2
3
4
5
6
7
8
9
10
s
o
1
2
3
4
5
6
7
8
9
10
FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS [Snc. 7.8
TABLE 7.2.GAIAXJLATION OF L@)
Xyl
sill Xp
152.222 0.46605
150.640 0.49030
148.648 0.52029
1445.139 0.55718
142.983 0. Q0205
139.008 0.65595
134.004 0.71929
127.705 0.79117
119,774 0.86799
109.791 0.94093
97.222 0.99207
V(*)
0.84960
0.89708
0,93545
0.96514
0.98688
1.00168
1.01047
1.01349
1.00934
0.99309
0,95337
+0.0057
U(*)
0.58636
0.51333
0.43984
0.36921
0.30529
0.25270
0.21694
0.20477
0.22445
0.28563
0.39829
Cos xy
O .86476
O. 87156
O .85399
o .83039
0.79846
-0.75480
0.69471
0.611W
O .49658
O .33859
O. 12571
9$)
1.06564
1.06826
1.06852
1.06781
1.06713
1.06722
1.06811
1,06909
1,06915
1,06781
1,06755
X!)
114,342
105.342
96.342
87.342
78.342
69.342
60,342
51.342
42,342
33.342
24.342
0.00227
0,00035
0.00061
O,ooo1o
0.00078
0.00069
0.00020
0.00118
0.00124
0.0001o
0.00036
sinx$)
0.91110
O.96436
O.99388
O.99892
o .97937
o .93570
O.86899
O,78089
O.67355
o .54964
0.41218
0.00197
0,00059
o .00041
o.m
0.00083
0,00083
0.00ooo
0.00101
0,00118
o,00ooo
o.m
Cos x$)
0.41218
O .26458
0.11046
0.04637
0.20207
0.35279
0.49482
0.62467
0.73914
0.83540
0.91110
~[8)
0.00030
0.00024
0.00020
0.00010
0.00005
0.00014
0.00020
0.00017
0.00006
0.0001o
0. C43036
I
. 0
9
1
2
-0.005
FI G. 7.4.Gaugi ng er r or i n the fi rst i mpr oved l ogari thmi c l i nkage, Ful l l i ne, resul t of
exact computati on. Dashed l i ne, val ues expected on l i near theor y.
I
sm. 7.8] GAUGI NG A THREE-BAR LI NKA13E 215
We have thus establ i shed preci si on poi nts at posi ti ons shi fted onl y
sl i ghtl y from those i ni ti al l y-and rather arbi trari l y-sel ected. The
7esul t of thi s fi rst cal cul ati on mi ght ver y wel l be accepted as fi nal . I t i s,
on the other hand, easy enough to make a fi rst-order cor r ecti on for the
effects of the quadrati c and hi gher terms. We have onl y to repl ace
@~J on the ri ght-hand si de of Eq. (51) by 7 f), and sol ve for new Ags and
Ags to be added to those al ready obtai ned. As before, we choose
arbi trari l y Agi = Aqd = O. The second-or der cor r ecti ons to the gs are
then found to be
A2g0 = 0.00246,
A2gl = 0.00170,
A2g2=
0.00130,
Azgs = 0.00292,
Ag4 = 0.00000,
A2g, = 0.00328,
A2g6 = O.00000.
1
The new and fi nal constants of the l i nkage are
g( = 1.07037,
g: = 0.72666,
q: = 0.55862,
)
(60)
~~ = 2.65385 ( = 152.0540), ) ,
~; = 0.10666,
g( = 1.99893 (= 114.5300),
gg = 1.57079.
(61)
The fi nal val ues of the gaugi ng parameter and the gaugi ng er r or are
shown i n Tabl e 7.3, together wi th the resul ti ng er r or i n the homogeneous
TABLE 7. 3.CHARACTERISTICS OF TEE SECOND IMPROVED LOGARITHMIC LINKAGE
s
o
1
2
3
4
5
6
7
8
9
10
9:()
1,06846
1,07105
1.07123
1.07039
1.06958
1.06956
1.07035
1.07132
1.07149
1.07036
1.07042
dgfl()
0.00191
0,00068
0.00086
0,00002
0.00079
0.00081
0,00002
0,00095
0.00112
O.00001
0,00005
1

2g*Q~)
4 .7295
3 .3403
2 .5673
2 .0707
-1.7212
1.4594
1.2561
1.0963
O.9742
O.8902
O.8532
~x;(a),
radi ans
o .00903
0.00227
0.00221
0.00004
0.00136
0.00118
0.00003
0,00104
0.00109 ~
0.000ol
0.00004 ~
~xw
degree:
0.517
0.130
0.127
0.002
0.078
0,068
0.002
0.060
0.062
0.000
0.002
iih,
o .00574
0.00144
0.00141
0,00002
0.00087
0.00076
0.00002
0.00067
0.00069
0.00ooo
0.00002
216 FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS
[SEC. 7.8
output parameter, bhz. To compute thi s we note that
whence
(62)
(63)
the conver si on to terms of the homogeneous output vari abl e i s obvi ous.
These resul ts are al so pr esented graphi cal l y i n Fi g, 7.5.
+0.005
s
0
:0
#
- -1
-0.005 -5
FI G. 7.5.Characteri stLcs of the second i mpr oved l ogari thmi c l i nkage. The dashed line
gives the form of a correction to be discussed in Sec. 7.9.
10
w
A
0.5
l-+)=-
Fm. 7.6.-Second improved logarithmic linkage.
The l i nkage i tsel f i s outl i ned i n Fi g. 7.6. The constants go, g,, and g,
determi ne the l engths of the l i nkage arms, whereas g,, and gl determi ne
the nature of the i nput scal e. The l i nkage i s shown wi th the i nput arm
at ei ther end of thi s scal e-that i s, for ZI = 1 and xl = 10. The output
arm i s shown i n the correspondi ng posi ti ons requi red by the geometr y
of the l i nkage. Because of the structural er r or i n the desi gn, these posi -
J_
SEC. 7.9]
THE ECCENTRI C LI NKAGE
217
ti one do not coi nci de wi th the ends of the X2 scal e determi ned by 96 and
gO; the scal e readi ngs are those shown i n ~l g. 76. (These have been
determi ned by exact computati on; hence one fi nds z!) = hi ) = 0.00595
i n Fi g. 7.6, i n contrast to the approxi mate val ue, 0.00574, i n Tabl e 7.3.)
7.9. The Eccentri c Li nkage as a Cor r ecti ve Devi ce.-When speci fi ed
tol erance are ver y cl ose i t may not be possi bl e to meet them by any
choi ce of the parameters of such si mpl e l i nkages as the three-bar l i nkage.
Reducti on of the structural er r or to tol erabl e l i mi ts then requi res i ntro-
ducti on of new adjustabl e parametene i nto the l i nkage. I n many cases
one can i ntroduce smal l addi ti onal cor r ecti ons by a superfi ci al change
4(
Fxm 7.7.Three-bar linkage modified by a double eccentric linkage.
i n the structure of the l i nkage.
Repl acement of an i deal harmoni c
transformer by a noni deal one i s such a change; another i s the i ntroducti on
of eccentr i c l i nkages at the joi nts of three-bar l i nkages or harmoni c
transformers. These modi fi cati ons i n the structure are mechani cal l y
sound, and permi t one to make use of al l previ ous computati ons-an
i mportant economy i n effor t.
Fi gure 7.7 shows a three-bar l i nkage modi fi ed by the i ntroducti on
of an eccentr i c l i nkage at ei ther end of the connecti ng bar.
The movi ng
pi vots of the i nput and output cranks car r y pl anetary gears meshi ng wi th
stati onary gears. The connecti ng bar i s not pi voted to the cranks, but
to the eccentr i c pi vots E, and Ez on the pl anetary gears; the ends of the
connecti ng l i nk do not move wi th the pi vots of the cranks, but about
them i n ci rcl es wi th radi i e, and ezusual l y smal l . Thus the di stance
between the ends of the cranks i s not a constant; i n effect, go can be made
,
218 FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS [SEC. 7.9
to vary, as i s requi red for mor e preci se generati on of the gi ven functi onal
rel ati on.
Each eccentr i c l i nkage may be characteri zed by thr ee constants: the
tooth-rati o tof the stati onary to the pl anetary gear, the eccentri ci ty e
of the pl anetary gear, and the angul ar posi ti on of the crank (denoted by
Xl = alt) for whi ch the eccentr i c pi vot l i es on the center l i ne of the
crank, at a maxi mum di stance from the frame pi vot. The doubl e
eccentr i c l i nkage i n Fi g. 77 thus provi des the desi gner wi th si x addi ti onal
parameters to adjust. Obvi ousl y, for greatest preci si on one shoul d
adjust al l constants of the devi ce si mul taneousl y. Usual l y one can
obtai n ver y sati sfactory resul ts by accepti ng as fi xed al l constants
determi ned i n previ ous desi gn wor k, varyi ng onl y the constants of the
eccentr i c l i nkages. I ndeed i t i s often sati sfactory to use onl y a si ngl e
eccentr i c l i nkage, wi th consequent reducti on to thr ee i n the number of
constants to be adjusted.
To determi ne the constants of the eccentr i c l i nkage one can empl oy
the gaugi ng-parameter method i n a somewhat modi fi ed form, val i d so
l ong as the eccentri ci ty of the l i nkage i s smal l .
I n deal i ng wi th a modi fi ed three-bar l i nkage one can advantageousl y
use the squared l ength of the connecti ng l i nk, go, as the gaugi ng param-
eter . Refer ence to Fi g. 7.7 shows that i ntroducti on of the fi rst eccentr i c
l i nkage has the same effect on g, as a change bg~) i n the l ength of the
i nput crank and a change 13X~)i n i ts angul ar posi ti on, wher e
~g~) = el cos (tl X~) al ), (64a)
()
&yy) = <1
si n (tl X~) al ),
g,
(64b)
to terms of the fi rst or der i n the smal l quanti ty el . Si mi l arl y, i ntroduc-
ti on of the second eccentr i c l i nkage has the effect of changi ng gn and 6XZ
by, respecti vel y,
~g~) = ez cos (tzX~) a~),
(65a)
( )
&yy = %2
si n (t2X~) az).
gz
(65b)
The resul ti ng change i n the gaugi ng parameter i s
I t i s thus the sum of four si nusoi ds mul ti pl i ed by the sl owl y varyi ng Gs,
Combi ni ng Eqs. (64), (65), and (67), one can wri te
SEC. 7.9] THE ECCENTRI C LI NKAGE 219
g) 4 G +( $[ 1X ) X ) 1+ 1( %?91?9
+e, [ ( G +r #i n[ i 2 x ) x ) - , + f % 8)
Her e the contri buti on of each eccentr i c l i nkage to the gaugi ng parameter
i s expressed as a si nusoi d wi th adjustabl e frequency, ampl i tude, and
phase constant, the second and thi rd of these quanti ti es bei ng subject to
sl ow vari ati ons of predetermi ned character. The di fference i n effect of
eccentr i c l i nkages on the i nput and output cranks ari ses partl y from
di fferences i n these vari ati ons, but pri nci pal l y from the fact that the
argument of the si nusoi d i s i n the fi rst case a l i near functi on of Xl , i n
the second case a l i near functi on of X2.
I t i s possi bl e to use the addi ti onal fl exi bi l i ty provi ded by eccentr i c
l i nkages to i ncrease the number of preci si on poi nts, i f al l constants of the
devi ce are adjusted si mul taneousl y. When onl y the constants of the
eccentr i c l i nkages are to be adjusted i t i s usual l y desi rabl e to l eave
undi sturbed the preci si on poi nts al ready establ i shed. One can make
Ag$] vani sh at fi ve previ ousl y establ i shed preci si on poi nts by adjustment
of the fi ve constants tl, tz,CYI, cw,and ez/e~. Then Ag&)wi l l have the same
zer os as the gaugi ng er r or of the ori gi nal three-bar l i nkage, and usual l y
the same general form; by appropri ate choi ce of the remai ni ng constant,
say el , one can gi ve i t roughl y the same magni tude. The compl eted
l i nkage wi l l then have the same preci si on poi nts as before, but smal l er
gaugi ng er r or s.
When a si ngl e eccentr i c l i nkage i s to be used one can
l eave undi sturbed onl y two preci si on poi nts.
Ezample.As an exampl e, we shal l further r educe the structural
er r or of the l ogari thmi c l i nkage of Fi g. 7.6, usi ng a si ngl e eccentr i c l i nk-
age. The desi gn pr ocedur e i s then extr emel y si mpl e, but requi res the
exer ci se of some judgment i f best resul ts are to be obtai ned.
The er r or functi on of the ori gi nal l i nkage, as shown i n Fi g. 7.6, has a
general l y si nusoi dal character. The poi nts of preci si on occur for
h$ = 0.05, 0.3, 0.6, 0.9, 1.0,
h~ = 0.0125, 0.1125, 0.3325, 0.772, 1.0.
(69)
Except for the l ast, they are qui te evenl y spaced i n X2, but unevenl y
spaced i n Xl ; they have about the same di stri buti on as the nul l s i n a
si nusoi d wi th argument l i near i n X2. I f a si ngl e eccentr i c l i nkage i s to
be used i t shoul d be pl aced on the output crank; i t wi l l then be possi bl e
to l eave four, and not just two, of the poi nts of preci si on essenti al l y
unchanged. We wi l l have then, on i ntroduci ng the Qs i n pl ace of the Gs,
[ ( ) 1
Q~)
Ag$) = 2ez[(Q~))z + (Q&))2]~~si n t*Xf) m + tanl
Q$)
. (70)
220 FI NAL ADJ USTMENT OF LI NKAGE CONSTANTS [SEC. 7.9
The nul l s of thi s expressi on occur when
()
Q~)
tzX~) at + tanl = n 180.
Q~)
(71)
Tabl e 7.4 shows the val ues of Q~)/Q&J at the previ ousl y establ i shed
preci si on poi nts, s = 0.5, 3, 6, 9, 10, and the val ues of t2X~) m requi red
i f these poi nts are to be nul l s of Agfj); n has been assi gned the val ues O, 1,
2, 3, 4 at the successi ve nul l s.
~
c$hz
&o.ool
=
/
0.5
0
n
1
#..ol
tig:
-0.CQ5
FIQ. 7.S.Structural error in the final logarithmic linkage.
TABLE 7.4.VALUES OF (t*Xf) aJ REQUIRED FOR VANISHING OF ALI$)
Qp)/Q:)
t,xy~ as,
8
degr ees
1
I
0.5 4.6543 77.9
3,0 2 .2272 245,8
6.0 1.0853 407.3
9.0 O.2994 556.7
10,0 O.0233 721.3
Let us choose to retai n the poi ntss = 0.5 and s = 9 as poi nts of preci -
si on. Taki ng the val ues of X~J i n Tabl e 7.1 as suffi ci entl y accurate, wc
then requi re
tz ( 110) a* = 77.9,
t, (33.5) a, = 556.7,
)
(72)
whence
tz= 6.26,
az = 766.5
7
ffz
.
tz
122.4.
I
(73)
SEC. 7-9] THE ECCENTRI C LI NKAGE

862
221
e2=0.00041 (exaggerated)
F1~.7.9.Fi nall ogari thmi cl i nkage.
The correspondi ng val ues of Ag$)/ez are shown i n Tabl e 7.5, and are
pl otted (dashed cur ve) i n Fi g. 7.5. Thi s cur ve has roughl y the same
form as the resi dual er r or 6g~@Jof the l i nkage whi ch i s to be i mproved;
i nspecti on wi l l show the Ag$) gi ves about the best approxi mati on to
~g~(ajwhen
ez = 0.00041. (74)
TABLE7.5.CALCULATI ON OFSTRUCTURAL ERROR OF FINAL LOGARITHMIC LINRAGE
s
o
1
2
3
4
5
6
7
8
9
10
@
e2
1.070
1.066
2.113
0.975
1.1845
2.0909
0.7207
1.4760
1 .9617
0.0107
1.9882
Ag:)
0,00042
0,00042
0.00084
0,00039
0.00047
0.00084
0.00029
0.00059
0.00078
0.00004
0.00079
L3gfl()
0.00149
0.00026
0,00002
0,00037
0,00032
0.00003
0.00027
0.00036
0.00034
0.00005
0,00084
.3hZ
O .00420
0.00050
o.m
0.00044
0.00032
0.00002
0.00019
0.00023
0.00019
0.00003
0.0W41
222 FI NAL ADJ UASTMENT OF LI NKAGE CONSTANTS ~SEC. 7{
The resul ti ng val ues of Ag$) and of the fi nal gaugi ng er r or
are gi ven i n Tabl e 7.5, together wi th the resul ti ng er r or i n the homo-
geneous output vari abl e; the l ast two quanti ti es are pl otted i n Fi g. 78.
The structural er r or remai ns l ess than 0.05 per cent except i n the i mmedi -
ate nei ghborhood of hl = hz = O, wher e i t abruptl y ri ses to 0.4 per cent.
The l i nkage i s outl i ned i n Fi g. 79 i n i ts confi gurati on for s = 2, ver y
near to one of i ts preci si on poi nts.
CHAPTER 8
LI NKAGES WI TH TWO DEGREES OF FREEDOM
Functi ons of two i ndependent vari abl es are usual l y mechani zed by
three-di mensi onal cams (Fi g. 1.24), whi ch are expensi ve to manufacture,
and rather bul ky; the y are, however , easy to desi gn and have ver y wi de
appl i cati on. Bar l i nkages wi th two degr ees of fr eedom can al so ser ve to
mechani ze functi ons of two i ndependent vari abl es. These l i nkages have
the advantages of bei ng fl at and smal l , of gi vi ng smooth fri cti onl ess
performance al l owi ng appreci abl e feedback, and of bei ng rel ati vel y
i nexpensi ve to manufacture i n quanti ti es. They are, on the other hand,
rel ati vel y di ffi cul t to desi gn, havi ng al ways resi dual structural er r or s
whi ch must be brought wi thi n the speci fi ed tol erances. The mathe-
mati cal desi gn of these l i nkages wi l l be tr eated i n the remai nder of thk
book. Basi c concepts needed by the desi gner wi l l be i ntroduced i n the
present chapter. Succeedi ng chapters wi l l show, partl y by pr ecept and
partl y by exampl e, how to desi gn l i nkage mul ti pl i ers or di vi ders (Chap.
9) and l i nkage generators of mor e general functi ons of two i ndependent
vari abl es (Chap. 10).
8.1. Anal ysi s of the Desi gn Probl em.-Mechani sms wi th two degr ees
of fr eedom have at l east one output parameter X~ functi onal l y rel ated
to two i nput parameters Xi and Xi :
X~ = F(X,, X,). (1)
I f the domai n of defi ni ti on D of thi s rel ati on i s a rectangl e,
Xi n S Xi S Xi W,
Xjm ~ xi ~ xi M,
(2)
the mechani sm i s sai d to be regul ar.
To such a mechani sm one may add functi onal scal es that establ i sh
rel ati ons between the parameters Xi , Xi , X~, and correspondi ng vari abl es
Zi , Zj, z~, respecti vel y. The mechani sm wi l l then ser ve to establ i sh a
functi onal rel ati on
zk = ~(~~,~j) (3)
between these vari abl es; we may say that the devi ce, mechani sm pl us
scal es, mechani zes Eq. (3). I f thi s rel ati on of the vari abl es i s to be
si ngl e-val ued, i t i s necessary that to defi ni te val ues of the i nput vari abl es
ther e cor r espond dei i ni te val ues of the i nput parameters, and that to a
defi ni te val ue of the output parameter ther e corresponds a defi ni te val ue
233
224 LI NKAGES
of the output vari abl e.
form,
WI TH TWO DEGREES OF FREEDOM [SEC. 81
The scal es must then establ i sh rel ati ons of the
xi = ( X, 1 X,) z,,
Xj = (X~lZj) Zj,
1
(4)
Xk = (wI X.J . Xk,
wher e al l thr ee oper ator s (but not necessari l y thei r i nverse operators)
are si ngl e-val ued. I f Eqs. (4) are of l i near form,
xi = x: )+ ki ( zi x$ ) ) ,
Xj = x; + k, ( zj 2$ ) ,
1
(5)
Xk = z~o)+ Kk(Xk X~O~),
the devi ce provi des a l i near mechani zati on of Eq. (3). When a
mechani sm i s to be a component of a mor e compl ex computer, i t i s often,
but not al ways, requi red to provi de a l i near mechani zati on of the rel ati on
between i nput and output vari abl es.
Any mechani sm generati ng a functi on F of two i ndependent param-
D-
Xi
X*
F
x,
FI G. 8,1.Schematic repre-
sentation of mechanism gener-
ating a function Xk of two
independent parameters, Xi
and Xj.
eter s may be r epr esented schemati cal l y as
i n Fi g. 8.1. Thi s representati on i s suffi ci ent
i n the case of three-di mensi onal cams, whi ch
can gener ate i n one step, so to speak, any
wel l -behaved functi on of two i ndependent
parameters. Si mpl e bar l i nkages, on the
other hand, can gener ate onl y a restri cted
cl ass of functi ons; to mechani ze a gi ven rel a-
ti on between parameters one must usual l y
bui l d up a mor e compl i cated structure, a
combi nati on of one or mor e si mpl e l i nkages
of two degr ees of fr eedom and several l i nk-
ages of one degr ee of freedom. I t i s then
necessary to consi der the i nternal structure of the functi on gener ator F.
Let G denote a si mpl e bar l i nkage wi th two degr ees of freedom, gen-
erati ng a functi on of two i ndependent parameters,
Y~ = G(Y;, Y]), (6)
I
of a restri cted cl ass. By combi ni ng such a l i nkage wi th thr ee l i nkages
havi ng one degr ee of freedom, as shown schemati cal l y i n Fi g. 82, one
can gener ate rel ati ons of a much wi der cl ass between parameters X,, Xi ,
Xk. A mor e el aborate structure i s that shown i n Fi g. 83, whi ch consi sts
of four l i nkages, each wi th two degr ees of freedom, so connected as to
make use of feedback. Theoreti cal l y, such structures make possi bl e a
further extensi on of the fi el d of mechani zabl e functi ons. I n practi ce i t i s
usual l y suffi ci ent to use the si mpl er structure of Fi g. 8.2, to whi ch we shal l
henceforth confi ne our attenti on.
SEC. 8.1] ANALYSI S OF THE DESI GN PROBLEM
225
We have then to consi der structures consi sti ng of a l i nkage wi th two
degr ees off r eedom, whi ch establ i shes a rel ati on [Eq. (6)] between i nternal
parameters Yi , Y,, Y~, and thr ee l i nkages of one degr ee of freedom, whi ch
rel ate the i nternal parameters to the correspondi ng external parameters
X{, Xj, X~~
Y, = (Y,l x,) . x,,
Yj = (Yj[Xj) . Xj,
1
(7)
x, = (X,1 Y,) . Y,.
Together , these establ i sh a rel ati on between the external parameters
[Eq. (l )]; the functi onal scal es, i n turn, conver t thi s i nto a rel ati on [Eq.
(3)] between vari abl es zi , Zj, z~, whi ch i s to be made to approxi mate as
cl osel y as possi bl e to some gi ven rel ati on, throughout a speci fi ed domai n.
The l i nkage G, wi th two degr ees of freedom, we shal l cal l the gri d
gener ator , for reasons whi ch wi l l become evi dent l ater. The l i nkages
B @.
Fro. 8.2.Combination of grid generator Fro. 8.3.Feedbrnck linkage with two
and transformer l i nkages.
degr ees of fr eedom.
!7,, Tj, Tk, we shal l cal l the transf ormersj because they transform the
i nternal parameters Y i nto the external parameters X. The di vi si on of a
mechani sm i nto a gri d gener ator and transformers i s to some extent
arbi trary; the breakdown of a gi ven functi onal rel ati on [Eq. (1)] i nto a
gri d-generator rel ati on [Eq. (6)] and transformer rel ati ons [Eq. (7)] i s
compl etel y arbi trary. We shal l ther efor e make use of the general i zed
term gri d gener ator for a gi ven functi on as denoti ng any l i nkage wi th
two degr ees of fr eedom whi ch wi l l ser ve as the l i nkage G i n a mechani za-
ti on of the gi ven functi on.
Transformer l i nkages i ncrease the fi el d of l i nearl y mechani zabl e func-
ti ons, but not the fi el d of functi ons mechani zabl e i n the mor e general
sense. A rel ati on z, = j(zi , z,) mechani zed by a gri d gener ator [Eq. (01,
transformer l i nkages [Eq. (7)], and functi onal scal es [Eq. (4)] can be
mechani zed al so by associ ati ng the same gri d gener ator di rectl y wi th
scal es whi ch establ i sh rel ati ons
Y~ = (Yi l X~) . (Xi l Zi ) Z i= O.($i),
Yj = (YjlXj) . (X, [~j) Xj = #j(X]),
I
(s)
Zk = (Zklx.k) o (X~lYJ Yk = ddIk).
226 LI NKAGES WI TH TWO DEGREES OF FREEDOM
[SEC. 82
Transformer l i nkages i n a desi gn thus ser ve onl y to change the form of the
functi onal scal esusual l y to make them l i near.
I t i s obvi ous that the choi ce of a gri d gener ator i s the central probl em
i n the desi gn of a l i nkage wi th two degr ees of freedom.
When a l i near
mechani zati on i s desi red, one can then pr oceed to desi gn the transformer
l i nkages by methods dk,cussed i n the precedi ng chapters; concerni ng thi s
l atter stage of the wor k, whi ch offers no new theoreti cal probl ems, l i ttl e
mor e need be sai d. I t i s evi dent that a ver y si mpl e gri d gener ator
may ser ve i f the transformers are made suffi ci entl y compl ex, whereas
another choi ce of gri d gener ator may make unnecessary the use of one or
mor e transformers. I t i s i mportant that the transformers not add too
much to the compl exi ty of the desi gn; a good gri d gener ator shoul d be
si mpl e i n structure, and al so adapted to use wi th si mpl e transformer
l i nkages. For i nstance, we shal l see that the common di fferenti al i s a
theoreti cal l y adequate gri d gener ator for an i mportant cl ass of functi ons;
i ts general use i n l i nearl y mechani zi ng these functi ons i s, however , not to
be recommended, si nce the requi red transformers tend to be excessi vel y
compl ex.
I n practi ce one has avai l abl e a rel ati vel y smal l number of types of
l i nkage sui tabl e for use as gri d generators; the avai l abl e gri d-generator
functi ons G bel ong to several restri cted cl asses. Usual l y these wi l l not
i ncl ude an exact gri d-generator functi on for the gi ven functi on; a struc-
tural er r or must be admi tted i n desi gni ng the gri d gener ator . Structural
er r or s must al so be admi tted i n the desi gn of the transformer l i nkages.
Thus i t i s al ways i mportant i n desi gni ng such mechani sms to make a fi nal
adjustment of al l avai l abl e constants, i n or der to mi ni mi ze the over-al l
structural er r or .
I n summary, mechani zati on of a gi ven functi on of two i ndependent
vari abl es i nvol ves the fol l owi ng steps:
1. Choi ce of a sui tabl e type of gri d gener ator .
2. Sel ecti on of the constants of the gri d gener ator .
3. Desi gn of the transformer l i nkages.
4. Fi nal adjustment of al l constants of the mechani sm.
The i deas to be devel oped i n the remai nder of thi s chapter are essenti al for
the fi rst of these steps; they al so form a foundati on for the pr ocedur es
requi red i n the second step, whi ch wi l l be descri bed i n l ater chapters.
8.2. Possi bl e Gri d Generators for a Gi ven Functi on.-I t i s ver y easy
to gi ve a formal characteri zati on of al l functi onal rel ati ons whi ch can be
mechani zed by use of a gi ven gri d gener ator . Combi ni ng Eqs. (6) and
(8), we see that these are the rel ati ons whi ch can be expressed as
~k = f(z,, zy) = ~fi (G[@&vi ), @j(z;)] ] , (9)
wher e G i s the gi ven gri d-generator functi on and ~~, 4i , $k are arbi trary
SEC. 8.2] GRI D GENERA TORS FOR A GI VEN FUNCTI ON
si ngl e-val ued functi ons of thei r arguments. Conversel y, to
a gi ven functi onal rel ati on
~k = j(~i , ~j)
one can empl oy a gri d gener ator wi th parameters rel ated by
Y, = G(Y,} Yj) = 41 {.f[~~(yi )) @T(y~)l ) ,
227
mechani ze
(3)
(l o)
wher e &, O;l , d~l are the i nverse of arbi trary si ngl e-val ued functi ons
d%,ti j, ~k.
The rel ati ons expressed i n Eqs. (9) and (10) can al so be expressed i n
terms of contour l i nes of the functi ons j and G. Let us pl ot contours of
constant Y~ = G(Yi , Yi ) i n the (Yi , Y?)-pl ane and l abel them wi th the cor -
$[~~%::,,,,,
.
(-1)
k
Yi<
*k(-z)
z,+
(a)
(b)
F1rJ.8.4.Topological transformation of contours.
respondi ng val ues of Y~ (Fi g. S.4a). Next l et us i ntroduce a change i n the
i ndependent vari abl es, defi ned by the equati ons
Yi = di (~i ),
Yj = ~~(~~),
1
(11)
wher e +~and @j are si ngl e-val ued functi ons of these arguments. Repott-
i ng the contours of constant G i n the (zi , x,)-pl ane (Fi g. 84b), we obtai n
l i nes of constant j(zi , ~j), as defi ned by Eq. (9). I f these contours are
rel abel ed wi th val ues of x~ gi ven by
Zf) = &( YfJ), (12)
they wi l l r epr esent the functi onal rel ati on
Zk = f(~i , Xi ) (13)
defi ned by Eq. (9), for a parti cul ar choi ce of the functi ons A, @i J dk. It
i s thus cl ear that a gi ven gri d gener ator can be used i n mechani zi ng a
gi ven functi on i f the contours of constant G( Y~, Yi ) can be transformed
i nto those of constant j(zi , xi ), or conversel y, by any topol ogi cal trans-
,
_. ~_.=~=_ -..
228 LI NKAGES WI TH TWO DEGREES OF FREEDOM [8Ec. 8.3
formati on of the form of Eq. (I i ), wi th rel abel i ng of the contours accord-
i ngto Eq. (12).
Formal rel ati ons such as Eqs. (9) and (10) are not of great val ue i n
practi cal desi gn wor k. The graphi c presentati on of these rel ati ons by
means of systems of contour l i nes i s of mor e i nterest, but as an i ndi cati on
of a di recti on of devel opment, rather than as a compl eted i dea.
What i s
real l y needed i s a means of characteri zi ng gi ven functi ons, on the one
hand, and avai l abl e gri d generators, on the other , whi ch wi l l make i t cl ear
at once whether or not a gi ven gri d gener ator can be used i n mechani zi ng
a gi ven functi on. Even mor e val uabl e wi l l be a means of characteri zi ng
a gi ven functi on whi ch wi l l assi st one i n desi gni ng a new and sati sfactory
gri d gener ator . I n both respects the i dea of gri d structure of a func-
ti on i s of fundamental i mportance.
8.3. The Concept of Gri d Structure.-The representati on of a functi on
of two i ndependent vari abl es by a gri d structure i s an extensi on of the
fami l i ar representati on by a set of contours of constant val ue of the
dependent vari abl e. I t wi l l her e be i ntroduced i n a speci al i zed form,
sati sfactory for the cl assi fi cati on of functi ons; i n l ater secti ons i t wi l l be
general i zed and appl i ed i n desi gn wor k.
Rectangular Grid Structure with Respect to a Center S and a Contour C.
We have now to construct the gri d structure of a functi onal rel ati on
defi ned through a domai n D i n the (xi , zj)-pl ane.
Let S be a poi nt i n the domai n D, associ ated wi th val ues of the vari a-
bl es whi ch wi l l be denoted by Zj), Zj), z~); thi s i s to ser ve as the center
of the gri d structure. Through S construct the contour B of constant z~,
Z~O)= J (G, Zj). (15)
(See Fi g. 8.5.) Next choose an adjacent contour C, defi ned by
Thi s, together wi th the poi nt S, wi l l fi x the gri d structure that i s to be
constructed.
Through S construct the verti cal l i ne xi = Zj), i ntersecti ng the con-
tour C at the poi nt (zj), zj-1), Zjl )). Through thi s l atter poi nt construct
the hori zontal l i ne x; = xj1], i ntersecti ng the contour B at the poi nt
(Z.o, 2;-1), ~k
o)). Through thi s poi nt, i n turn, construct the verti cal
l i ne xi = zj), i ntersecti ng the contour C at the poi nt (Z:), Zj-), z~)).
Conti nuati on of thi s process extends the stepl i ke structure of l i nes
between the two contours, both above and bel ow S, and defi nes sequences
of val ues of the two i ndependent vari abl es:
SEC. 8.3] THE CONCEPT OF GRI D STRUCTURE 229
(2), ~~1), ~:o), X:1), .&), . . . ,
. . . , xi
. . . .
&), 2$1), .&, .&, &), . . . .
The rectangul ar gri d of l i nes
Z; = Zr) (17)
and
z, = Z$) (18)
wi l l cover part, but not al ways al l , of the domai n D. Thi s rectangul ar
gri d serves to defi ne a system of contours
ftk = Xf) , (19)
whi ch, together wi th thi s gri d i tsel f, wi l l make up the rectangul ar gri d
structure of the functi on, defi ned wi th r espect to the center S and the
contour C.
FI G. 8.5.I deal mi d structure.
I deal Grid Structure.The rectangul ar gri d
and i ts l i nes so numbered, that a si ngl e contour,
Xk = z~o),
passes through al l gri d i ntersecti ons for whi ch
r+s=O,
and a si ngl e contour,
Zk = X\-l ),
pasws through al l gri d i ntersecti ons for whi ch
T+ S=l .
has been so constructed,
(20a)
(20b)
(21a)
(21b)
230 LI NKAGES WI TH TWO DEGREES OF FREEDOM [SEC. 8.3
Ther e i s an i mportant cl ass of functi ons such that, no matter how the
center S and the contour C are chosen, ther e wi l l be a si ngl e contour,
Xk = Zf) , (19)
passi ng through al l gri d i ntersecti ons for whi ch
r+s=t, (22)
t bei ng any i nteger, posi ti ve or negati ve. Such a functi on wi l l be sai d to
have i deal gri d structure.
An i deal gri d structure (defi ned wi th r espect to a center S and a
contour C) wi l l consi st of the rectangul ar gri d speci fi ed above, pl us al l the
contours of constant zk whi ch pass through the i ntersecti ons of the gri d.
Such a gri d structure wi l l appear as shown i n Fi g. 8.5. Thi s gri d struc-
tur e can al so be descri bed as consi sti ng of thr ee fami l i es of curves, gi ven
by Eqs. (17), (18), and (19), such that through ever y poi nt of i ntersecti on
ther e passes a cur ve of each
fami l y. Thi s descri pti on wi l l r e-
9,
mai n val i d even when the con-
!t~
zk(z:l )
#
I
cept of i deal gri d structure i s
.;;
.~j~~~~~j~~~-~--- :(?c! .->$) en%:;i al G~i d Structum.-
0 ------
s When di fferent contours of con-
--
% ))
=p)
,(1,0)[
stant x,4 pass through gri d i nter-
.-
,(0:1]-!---:- secti ons characteri zed by the same
I I ,
1,
val ue of (r + s), the g;l d struc-
::
,11 tur e wi l l be sai d to be noni deal .
=g) (0) (!) (1)
~, z) ~,
Fi gure 8.6 represents an extr eme
Zi
case of noni deal gri d structure.
Fx~.S,6.Noni deal gri dstructure.
When the gri d structure i s non-
i deal one cannot di sti ngui sh by the si ngl e i ndex (r + s) the contours of
the fami l y defi ned by the gri d; one mi ght i nstead l abel each cur ve wi th
the two i ndi ces, r and s, of the correspondi ng gri d i ntersecti on, as shown
i n Fi g. 8.6.
I t i s not conveni ent to consi der al l these contours as bel ongi ng to the
gri d structure of the functi on, nor woul d thk contri bute to the cl ari ty wi th
whi ch the gri d structure represents the properti es of the functi on.
I t i s
suffi ci ent to i ncl ude onl y one such contour f or each val ue of (r + s), l abel -
i ng i t wi th thi s quanti ty as the si ngl e i ndex. The choi ce of the contours
to be i ncl uded i s to some degr ee arbi trary. We shal l consi der a noni deal
gri d structure to consi st of the rectangul ar gri d defi ned i n the usual way,
pl us the contours of constant xk that pass through the gri d i ntersecti ons
wi th r = s = n, pl us i ntermedi ate contours that i nterpol ate smoothl y

SEC. 8.3] THE CONCEPT OF GRI D STRUCTURE

231
between these and ther efor e pass near the i ntersecti ons wi th
?-=s+l =n
andr+l =s=n. (A preci se method f or choosi ng these i ntermedi ate
contours wi l l be i ndi ca~d i n Sec. 8.6.)
l ?I G.S,7.Gri d structure Of the functi on zk = ~ ( 2; - 1-Zi zi + zi 2) *
4
F@re 8.7 i l l ustrates a typi cal noni deal gri d structure, that of the
functi on
k = * z:+ z+ :)M
(23)
The poi nt xi = xi = X, = 1 has been chosen as the center S, and the
contour C i s that for whi ch m = 0.9 = z~l ). The contours are sym-
metri cal wi th r espect to the dotted l i ne i n the fi gure, and so i s the r ec-
232 LI NKAGES WI TH TWO DEGREES OF FREEDOM [SEC. 8.4
tangul ar gri d. I t wi l l be obser ved that near the contours B and C, and
near the l i ne r = s, the contours pass ver y nearl y through the gri d i nter-
secti ons. Away from these l i nes the noni deal character of the gri d struc-
tur e becomes i ncreasi ngl y apparent, as the contours pass farther and
farther from the gri d i ntersecti ons.
Grid Structui-e in the Neighborhood of a center.The gr eater the
di stance from the center S to the adjacent contour C of the gri d struc-
ture, the mor e coarsel y does the gri d structure r epr esent the properti es
of the functi on. I n or der to defi ne for a functi on the gri d structure i n
the nei ghborhood of a center , one must al l ow the contour C to approach
the center S, and concentr ate attenti on on a ver y smal l regi on about the
center whi ch, neverthel ess, contai ns a consi derabl e number of gri d l i nes.
One can expand i n Tayl ors seri es about the center S any wel l -be-
haved. functi on:
I n the i mmedi ate nei ghborhood of S the quadrati c and hi gher terms i n
Eq. (24) can be negl ected. To thk approxi mati on the contours of
constant ~k are paral l el strai ght l i nes, the gri d consi sts of i denti cal
rectangl es, and the gri d structure i s i deal . Thus one can say that the
gri d structure of any wel l -behaved functi on i s i deal i n the nei ghborhood of
i ts center . The practi cal si gni fi cance of thi s statement, whi ch wi l l be
brought out mor e compl etel y i n l ater secti ons, i s thi s: I t i s al ways easy to
fi nd a gri d gener ator for a functi on i f the domai n of mechani zati on i s
suffi ci entl y restri cted; what i s di ffi cul t i s to fi nd gri d generators useful
throughout extended domai ns.
8.4. Topol ogi cal Transformati on of Gri d Structures.-I t has been
shown i n Sec. 8.2 that the topol ogi cal transformati on
Yi = ~i(~t),
Yj = @~(Zj), (11)
carri es cent ours of the functi on
Yk = G(Y,, Y) (6)
i n the (Y~, Y;) -pl ane i nto cent ours of the functi on
(13)
i n the (zi , x,)-pl ane. Thi s transf ormati on carri es verti cal strai ght l i nes i n
the (Y~, Yj)-pl ane i nto verti cal strai ght l i nes i n the (z~, x,)-pl ane, and
hori zontal strai ght l i nes i nto hori zontal strai ght l i nes. I ndeed, the
r eader wi l l easi l y see that the i dea of gri d structure has been so defi ned
SEC. 8.5] THE SI GNI FI CANCE OF I DEAL GRI D STRUCTURE 233
that i f thi s transformati on carri es a center Sr i n the (Yi l Yj)-pl ane i nto a
center S. i n the (xi , zi )-pl ane, and a contour CF i nto a contour C., then i t
carri es the compl ete gri d structure of the functi on G( Yi l Yi ), defi ned
wi th r espect to SY and CY, i nto the gri d structure of the functi on j(xi , xi ),
I
defi ned wi th r espect to S=and Cs.
The val ues of the vari abl es associ ated
wi th the gri d l i nes and contours wi l l be transformed accordi ng to Eqs.
(11) and (12), but the i ndi ces r, s, t, wi l l be unchanged.
The mai n concl usi on of Sec. 8.2 can ther efor e be restated i n the fol l ow-
i ng terms: A gi ven gri d gener ator can be used i n the ezact mechani zati on
of a gi ven functi on i f, and onl y i f, ther e exi sts a topol ogi cal transforma-
ti on, of the form of Eq. (11), that carri es each gri d structure of the
{
functi on G(Yi 7 Yi ) i nto a correspondi ng gri d structure of the gi ven func-
ti on j(zi , z;). I n practi ce, of course, al l that need be shown i s that some
gri d structure of the functi on G( Yi l Yj), wi th suffi ci entl y smal l meshes, can
be thus transformed i nto a correspondi ng gri d structure of the functi on
~(z,, Xi ), wi th errors wi thi n speci fi ~d tol er~~ces.
The topol ogi cal transformati on cannot change i ntersecti on properti es
of the l i nes of the gr i d structure; i t must then transform an i deal gr i d
structure i nto another i deal gr i d structure, a noni deal gr i d structure i nto
another noni deal one. I t fol l ows that a gi ven functi on wi th an i deal
gri d structure can be mechani zed exactl y onl y by a gri d gener ator wi th
i deal gri d structure, a gi ven functi on wi th noni deal gri d structure onl y by
a noni deal gri d gener ator .
I n Sec. 8.5 i t wi l l be shown that al l functi ons wi th i deal gri d structure
can be mechani zed by any gri d gener ator wi th i deal gri d structure, such as
the common di fferenti al .
I n the case of noni deal gri d structures the si tuati on i s not so si mpl e.
Ther e are many di fferent ways i n whi ch a gri d structure can be noni deal ;
i t i s i n general possi bl e to determi ne whether or not a gi ven gri d gener ator
wi l l ser ve i n the mechani zati on of a gi ven functi on onl y by maki ng a
detai l ed compari son of thei r r especti ve gri d-structure properti es. I n
Sec. 8.6 ther e wi l l be i ndi cated the basi c i deas of a systemati c method for
choosi ng from among a number of gi ven types of gri d gener ator the one
whi ch i s most sui tabl e for the mechani zati on of a gi ven functi on. Unfor-
tunatel y, thi s method cannot suffi ce for the desi gn of noni deal gri d gen-
erators unti l an extensi ve fi l e of gri d structures has been accumul ated.
I n the present state of the art i t i s necessary to desi gn a gri d gener ator ab
i ni ~i o for each gi ven functi on; the way i n whi ch thi s can be done, by a
study of i ts gri d structure, wi l l be i ndi cated i n Sec. 87, and i l l ustrated at
l ength i n Chap. 10.
8.5. The Si gni tl cance of I deal Gri d Structure.-I t wi l l now be shown
that i f a functi onal rel ati on
Zk = f(~i , Zj) (25)
---- -.. . .->--- _... ._
1
234 LI NKAGE8 WI TH TWO DEGREES OF FREEDOM [SEC. 8.5
has i deal gri d stmcture, then ther e exi sts a topol ogi cal transformati on
Yi = @i(Zi),
I
(26a)
Yj = @j(Xj), (26b)
Xk = &(Yk), (26c)
such that
Yk = Y,+ Yj. (27)
I nother words, i fthefuncti onal rel ati on Eq. (25) hasi deal gri d structure
i t can be expressed as
@l (x~) = @i (Z) + ~j(~j) .
(2s)
I t wi l l fol l ow i mmedi atel y that thi s functi on can be mechani zed usi ng a
di fferenti al as gri d gener ator , together wi th transformer l i nkages and
scal es whi ch establ i sh the rel ati ons of Eqs. (26).
B\
Fm. 8,8.Subdivision of an ideal Rid structure.
Let us consi der the i deal gri d structure defi ned wi th r espect to a
center S and a contour C, as shown by the sol i d l i nes of Fi g. 8.8. Asso-
ci ated wi th each i ntersecti on i n thi s gri d structure are val ues of the
i ndi ces r, s, and t,such that
r+s=t. (29)
The i ndex r i s a si ngl e-val ued functi on of the xi -coordi nate of the i nter-
secti on, s i s a si ngl e-val ued functi on of xi , and x~i s a si ngl e-val ued functi on
of t. I n short, the i ndi ces r, s, t have al l the characteri sti cs whi ch shoul d
be possessed by the parameters Yi , Yj, Yk, respecti vel y, except that they
are defi ned onl y for a di screte sequence of val ues, i nstead of as conti nuous
functi ons of xi , xi , Xk. We shal l now show that the defi ni ti on of the i ndi ces
can be extended to appl y to a conti nuum of val ues; the theor em above
wi l l then fol l ow on i denti fi cati on of r, s, twi th Yi , Y7, Yk, respecti vel y.
Let us consi der the porti on of Fi g. 8.8 l yi ng between the contours B
and C, and between the l i nes z~o)and Zjl ). I t i s cl earl y possi bl e to choose
a contour C such that a step structure constructed between the contours
I
I
I
SEC. 8.5] THE SI GNI FI CANCE OF I DEAL GRI D STRUCTURE 235
Z3 and C passes from the center S to the poi nt (zj), Z\-), ZL)) i n two
steps, i nstead of one. Now l et us construct a gri d structure wi th r espect
to the center S and the contour C (sol i d and dashed l i nes of Fi g. 8.8).
I t i s cl ear from the method of constructi on that thi s new gri d structure
i ncl udes the contour C as i ts fi rst contour beyond C. I t fol l ows i mmedi -
atel y that ever y l i ne of the ori gi nal gri d appears i n thi s new one; i n
addi ti on, ther e i s a new l i ne i nterpol ati ng between each pai r of adjacent
l i nes i n the ol d gri d. I nstead of reassi gni ng i ntegral i ndi ces to al l l i nes
I
I
-o-
f the new gri d, we shal l retai n the ol d i ndi ces for the ol d l i nes and assi gn
hal f-i ntegral i ndi ces to the i nterveni ng l i nes. Al l i ndi ces are just hal f as
l arge as they woul d have been i f the constructi on had been begun wi th
the center S and the contour C; Eq. .3
(29) i s sti l l sati sfi ed, but hal f-i ntegral
i ndi ces may occur i n i t as wel l as +2
i ntegral ones.
I n the same way we can construct 1
a new gri d structure i n whi ch C i s the o
second contour (rather than the fi rst) ~1
beyond S, and can assi gn to the l i nes
of thi s structure quarter-i ntegral i n- -2
di ces whi ch sati sfy Eq. (29). Con- -3
ti nui ng to subdi vi de the ori gi nal gri d -3 -2 -1 0 +1 +2 +3 +4
i n thi s way, we can defi ne gri d-struc-
Zi
tur e l i nes cor r espondhg to arbi trary
FI G.8.9.Gri dstructureofz, 1- ~j = u.
val ues of r,s, t,i n a conti nuous range, throughout mai ntai ni ng the val i di ty
of Eq. (29). These i ndi ces appear as functi ons of z,, xi , x~, havi ng the
form of Eqs. (26); i t i s onl y necessary to i denti fy r, s, t wi th Yi , Yj, Y~,
to compl ete the pr oof of the theorem.
As exampl es of functi onal rel ati ons wi th i deal gri d structure we may
take
xk = xi + ~j,
(30)
I
I
wi th gri d structure shown i n Fi g. 8.9,
wi th gri d structure shown i n Fi g. 8.10, and
Xi
~k = ,
Xi
(32)
wi t,h gri d structure shown i n Fi g. 8.13.
An al ternati ve statement of our resul t i s the fol l owi ng: I f a functi onal
rel ati on has i deal gri d structure, i t i s al ways possi bl e to appl y a topo-
1
236 LI NKAGES WI TH TWO DEGREES OF FREEDOM
[SEC. 8.5
l ogi cal transformati on of the form of Eqs. (26) that wi l l transform thi s
gri d structure i nto the form shown i n Fi g. 89, wi thi n some domai n of the
vari abl es. Possi bl e l i mi tati ons of the domai n of thi s transformati on wi l l
be evi dent on compari son of Fi gs. 89, 8.10, and 813. The gri d structures
of Fi gs. 8-9 and 8.10 cor r espond cl osel y i n the fi rst quadrant, and the
general form of the requi red transformati on of hori zontal and verti cal
I ~I G.S.IO.Grid structure of Z,2 + Zjz = z~z.
coordi nates i s cl ear enough; on the other hand, i t i s al so cl ear that a
transformati on whi ch wi l l ser ve to car r y one gri d structure i nto the other
i n the fi rst quadrant wi l l not have thi s effect i h the second quadrant, or
the fourth. Thi s i s due to the fact that the transformati on Eq. (26a), as
defi ned by the gri d structure, ceases to be si ngl e-val ued when the contour
C i s fol l owed through a poi nt of i nfi ni te sl ope; si mi l arl y, Eq. (26b) ceases
to be si ngl e-val ued when C i s fol l owed through a poi nt of zer o sl ope.
Together , these l i mi tati ons restri ct the transformati on from Fi g. 89 to
I
SEC.8.5] THE SI GNI FI CANCE OF I DEAL GRI D STRUCTURE 237
Fig. 810 to correspondi ng quadrants. A ver y di fferent exampl e i s
provi ded by Eq. (33). The transformati on equati ons
Y, = i n Z,,
Yj = i n Xj, (34)
Y~ = i n z~,
whi ch transform Fi g. 8.13 i nto Fi g. 8.9, transform the fi rst quadrant of
Fi g. 8.13 i nto the whol e of Fi g. 8.9; other transformati ons car r y each
of the other quadrants of Fi g. 813 i nto the whol e of Fi g. 89.
We have now pr oved that any functi on wi th i deal gri d structure can
be mechani zed usi ng a di fferenti al as gri d gener ator . Thi s i s by no means
necessary, nor i s i t usual l y desi rabl e.
I t i s, i n fact, possi bl e to use any
gri d gener ator wi th i deal gri d structure i n mechani zi ng any gi ven func-
ti onal rel ati on x~ = j(w, ZJ wi th i deal gri d structure; the choi ce shoul d
depend on the mechani cal desi rabi l i ty of the devi ce as a whol e. I n or der
to make contact wi th the anal ysi s of Sec. 8.1, l et us suppose that i t i s
desi red to establ i sh between external parameters Xi, Xi, X~, a given
rel ati on
X~ = F(Xil Xi)
(35)
wi th i deal gri d structure; thi s i s a probl em equi val ent to that of fi ndi ng a
l i near mechani zati on of a rel ati on of the form of Eq. (35) between
vari abl es xi , xi , x,. Let ther e be gi ven a gri d gener ator wi th i deal gri d
structure mechani zi ng the rel ati on
Y~ = G(Yi , Yj)
(36)
between i nternal parameters Yi , Yj,
Yk. We have seen that thi s rel ati on
can al so be mechani zed using a di fferenti al as gri d gener ator ; Eq. (36) i s
equi val ent to
zk = z, + Zj}
(37)
2, = (2,1 Yi ) Y{,
Zi = (Zjl Yj) Yj,
I
(38)
Y, = (Yjl zk) 2, ,
wi th the i ndi cated transformer functi ons al l si ngl e-val ued. Conversel y,
the gi ven gri d gener ator , Eq. (36), can be used i n mechani zi ng Eq. (37), aa
i ndi cated i n the i nner ci rcl e of Fi g. 8.11; the transformer functi ons
requi red are the i nverse of those i n Eq. (38). We know al so that the
resul ti ng di fferenti al can be used i n mechani zi ng Eq. (35), i n combi nati on
wi th transformer l i nkages generati ng the rel ati ons
Zi = (Zi l Xi ) Xi ,
Zj = (Zjl Xj) Xj,
I
(39)
X~ = (X~l Z~) cZk,
238 LI NKAGES WI TH TWO DEGREES OF FREEDOM [SEC. 8.6
as shown i n the outer ci rcl e of Fi g. 811. I t thus becomes obvi ous that
the gi ven gri d gener ator can be used i n mechani zi ng Eq. (35), by com-
bi ni ng i t wi th transformers mechani zi ng the oper ator s
(YilXi) = (YilZt) o (ZilXi),
(Yjlxj) = (Y~lzJ -(zjlx,),
)
(40)
(X,I Y!J = (xk(zk)(zkl Yk).
These transformers maybe si mpl er i n structure than those requi red
wi th the si mpl er di fferenti al gri d gener ator , and the domai n of operati on
of the compl ete deti ce may be mor e extensi ve. For exampl e, i t i s cer -
tai nl y possi bl e to bui l d a mul ti pl i er wi th l i near scal es, usi ng a di fferenti al
xa
Fm. 8.11.Mechanization of a given function with ideal grid structure
generator with ideal grid structure.
by a given grid
as gri d gener ator , and l ogari thmi c transformers [Eq. (34)] such as that
i l l ustrated i n Fi g. 7.9. Theresul ti ng mechani sm woul d be unnecessari l y
compl i cated, and woul d be operabl e onl y i n a domai n i n whi ch none of
the vari abl es changes si gn. I t i smuchmore sati sfactory touse as gri d
gener ator a star l i nkage (Chap. 9). Thi s can be so desi gned as to have an
al most i deal gri d structure, and, i n combi nati on wi th si mpl e trans-
formers, makes up a mul ti pl i er useful through a domai n that i ncl udes
both posi ti ve and negati ve val ues of z~ and ~. I t i s thus evi dent that
the probl em of desi gni ng new gri d generators wi th i deal gri d structure i s
one of consi derabl e practi cal i mportance; i t wi l l be the subject of Chap. 9.
8.6. Choi ce of a Noni deal Gri d Generator.The number of types of
si mpl e and mechani cal l y sati sfactory gri d generators i s rather l i mi ted, but
the gri d structures of these devi ces can be vari ed wi del y by changi ng
desi gn constants. I t appears to be practi cabl e to set up an atl as of gri d
SEC. 8.6] CHOI CE OF A NONI DEAL GRI D GENERATOR 239
structures from whi ch, by si mpl e compari son wi th the gri d structure of a
gi ven functi on, i t woul d be possi bl e to sel ect a type of gri d gener ator
sui tabl e for a mechani zati on of that functi on, and to determi ne approxi -
matel y the requi red desi gn constants. We may note her e some char-
acteri sti cs of thi s probl em, and some methods of si mpl i fyi ng i t.
The gri d structure of a gi ven functi on may di ffer from the catal oged
gri d structure of a sati sfactory gri d gener ator for any or al l of four
reasons:
1. They maydi ffer bya topol ogi cal transformati on, Eqs. (8), whi ch
i s to be carri ed out by the transformer l i nkages.
2. The contours C of the gri d structures may not correspond.
3. The center s h chosen for the gri d structures may not correspond.
4. The catal oged gri d structure may cor r espond to use of the wr ong
termi nal as output termi nal .
These four factors wi l l be consi dered i n turn.
&gulaTized Grid Structures.I n or der to make possi bl e di rect com-
pari son of the gri d structures of gi ven functi ons and gi ven gri d generators,
i t i s desi rabl e to r educe to a common form al l gri d structures whi ch di ffer
onl y by a topol ogi cal transformati on.
Thi s common form wi l l be ter med
the regul ari zed gri d structure. I t i s possi bl e to mechani ze a gi ven
functi on by a gi ven gri d gener ator i f, and onl y i f, thei r regul ari zed gri d
structures are i denti cal , or can be made so by pr oper choi ce of the el e-
ments menti oned i n I tems 2, 3, and 4 of the precedi ng paragraph.
I n general terms, one may defi ne a regul ari zed gri d structure as that
obtai ned from any gi ven gri d structure by appl yi ng a topol ogi cal trans-
formati on whi ch conver ts the rectangul ar gri d of the ori gi nal structure
i nto a square gri d. Mor e preci sel y, the transformati on to be appl i ed i s
that whi ch maps i nto a square gri d the ver y fi ne gri d structwe formed
i n the l i mi t as the contour C approaches the center S. Let the gi ven
functi onal rel ati on be
zk = ~(~i , ~j). (14)
The transformati on to the pl ane of the new vari abl es (z,, zj) can be
defi ned i n terms of l i ne i ntegral s i n the (z, zj)-pl ane, extendi ng from the
chosen center S = (Xiol Xjot xkO) al ong the contour of constant Sk:
H)
af
~ = @i (zi ) =
.,, z .k~,,
d%, (41a)
H)
af
q = @j ( ~J =
. , ~ . , =. , O
dzi. (41b)
The vari abl e Zk i s then so defi ned, as a functi on of zk, that Eq. (14)
reduces to
zk=zi +~ (42)
240 LINKAGES WITH TWO DEGREES OF FREEDOM [SEC. 86
al ong the l i ne z, = Zi i n the (z,, zj)-pl ane.
Rewri ti ng Eqs. (41) as
x, = 1#/(zi ), (43a)
q = $(zj), (43b)
one may express the requi red rel ati on as
, = [ 4) ) 491
(44)
Al l functi ons wi th i deal gri d structure have the same regul ari zed gri d
structur~that i l l ustrated i n Fi g. 8.&except for possi bl e di fferences
i n the spaci ng of the gri d l i nes.
For i nstance, i f the gi ven rel ati on i s
( )
aj _ Xi
(-)
af Xi
z . Zo ax ~
(46)
%zk ZLO 7 Zk=Zko
Then
(47a)
x; Z:.
Zi=
hkl l -
(47b)
Thus
xi = 4-l (zi ) = (X:. + 2wzi )~4, (48a)
Z; = 41 (Zj) = (X~~+ 2Xk0Zj)i, (48b)
and Eq. (44) becomes
xk = (~~~+ 2~~&~)i , (49a)
or
(49b)
As wi th al l functi ons havi ng i deal gri d structure, the zs thus defi ned
sati sfy Eq. (42) not OrdY when zi = zjl but throughout the domai n i n
whi ch the transformati on Eq. (41) i s defi ned and si ngl e-val ued; the
regul ari zed gri d structure i s that of Eq. (42).
Fi gure 8.12 shows a regul ari zed noni deal gri d structure, that of Eq.
(23). I t i s, i n fact, the regul ari zed form of the gri d structure shown i n
Fi g. 8.7, and has been constructed graphi cal l y by r efer ence to that fi gure,
rather than by anal yti cal worki ng out of the transformati on di scussed
above. We know that the rectangul ar gri d of Fi g. 8.7 woul d be r educed
to an al most square gri d by thi s transformati on. I n Fi g. 8.12 thi s gri d
has been constructed as exactl y square, wi th negl i gi bl e er r or . The con-
SEC. 8.6] CHOI CE OF A NONI DEAL GRI D GENERATOR 241
tour l i nes i n Fi g. 812 must have the same rel ati on tothi s square gri d
as the contour l i nes of Fi g. 8.7 have to the rectangul ar gri d; poi nts of
i ntersecti on are easi l y establ i shed by i nterpol ati on, and the contours
passed through them. Such a constructi on i s qui te accurate enough
for the purposes her e contempl ated i f the mesh of the ori gi nal gri d struc-
tur e i s not too open.
The curves of the structure
thus establ i shed must cor r espond
to equal l y spaced val ues of zi , zjl
zk that sati sfy Eq. (42) al ong the
l i ne .Zi = Zj. One can determi ne
the spaci ng constant a onl y by
r efer ence to the transformati on
equati ons; thi s, h owe ve r, i s a
matter of scal e whi ch i s of no
practi cal i mportance.
I n a regul ari zed gri d structure
the square gri d contri butes noth-
i ng to the characteri zati on of the
functi on; attenti on can be focused
on the system of contours of con-
stant z~. The useful ness of the
-4a -2a
1
+Za +4a
FI G. 8.12.Regul ari zed noni deal gr i d struc-
ture.
i dea of regul ari zed gri d structure i s l argel y due to thi s fact.
EjTect oj Change of Contour C.The transformati on to a regul ari zed
gri d structure conver ts the functi onal rel ati on
~k = ~(~,, ~j) (14)
i nto another,
)
zk = g(zi , Z~, (50)
whi ch defi nes a regul ari zed surface i n (zi , Zj, z~)-space.
Thi s surface i s
tangent to the pl ane Zk= zi y Zj al l al ong the fi ne Zk= O, and i ntersects
i t al l al ong the l i ne zi = Zj; i ts form i s made obvi ous by the contour l i nes
of the regul ari zed gri d structure.
Change i n the choi ce of the contour C
changes the spaci ng of these contours, but not the form of the regul ari zed
surface that they descri be. Thus, i n compari ng regul ari zed gri d struc-
tures of gi ven functi on wi th those of gi ven gri d generators, one shoul d
compare surface to surface, not contour to contour.
Thi s can be done
wi thout knowl edge of the spaci ng constant.
E$ect of Chunge of the Center S.Passage to the regul ari zed gri d
structure al ways transforms the contour B through the chosen center S
i nto the strai ght l i ne zi + Zi = O i n the (zi , z~)-pl ane. When the gri d
structure i s i deal , al l contours of constant Zk are transf or med i nto paral l el
strai ght l i nes; the appearance of the regul ari zed gri d structure does not
242 LI NKAGES WI TH TWO DEGREES OF FREEDOM [SEC. 86
depend on S. When the gri d structure i s noni deal thi s i s not preci sel y
true. Adj scent contours are conver ted i nto l i nes whi ch are onl y approxi -
matel y strai ght, and the speci al characteri sti cs of the noni deal structure
become evi dent i n the form of the mor e r emote contours. Choi ce of the
center S on another contour B (Fi g. 812) woul d make that contour
transform i nto a strai ght l i ne, and i ntroduce a correspondi ng curvature
i nto thetransformed contour B. I f Bi svery nearl y a strai ght l i ne i n the
ori @nal regul ari zed gri d structure, thechange i nform of B, and of the rest
of the gri d structure, wi l l be smal l .
I t i s thus evi dent that the chosen
center of the gri d structure i n Fi g. 812 coul d be changed wi thi n wi de
l i mi ts wi th l i ttl e effect on the appearance of the regul ari zed gri d structure.
on the other hand, a stri ki ng change woul d occur i f S wer e chosen as
the center .
I t i s usual l y suffi ci ent, for practi cal purposes, to r epr esent a gi ven
gri d gener ator by a si ngl e regul ari zed gri d structure. The domai n of
useful ness of the gri d gener ator wi l l be l i mi ted by mechani cal consi dera-
ti ons, and i t wi l l be natural to choose S near the center of thi s domai n.
The domai n of a gi ven functi on to be mechani zed wi l l al so be speci fi ed,
and a center S wi l l be chosen near the center of thi s domai n. I f the gri d
gener ator i s to be useful i n mechani zi ng thi s parti cul ar functi on the
center s S and S must cor r espond at l east roughl y, and the di fference
between them wi l l not cause l arge di fferences i n the appearance of the
regul ari zed gri d structures.
Efl ect oj Choi ce of Output !!ermi nul .-Gi ven a mechani sm sui tabl e for
use as a gri d gener ator , one mi ght choose any of the thr ee termi nal s as
the output termi nal , and mi ght associ ate the i nput parameters wi th the
other termi nal s i n two di fferent ways.
To each of the si x possi bl e ways
of usi ng thi s mechani sm as a gri d gener ator ther e corresponds a di fferent
gri d structure. The appearance of the gri d structure depends pri nci pal l y
on whi ch of the termi nal s i s associ ated wi th the output parameter yk;
i nterchange of the i nput termi nal s, i n thei r associ ati on wi th Yi and Yi ,
merel y produces a refl ecti on of the gri d structure i n the di agonal l i ne
Y, = Yi .
I n an atl as of gri d structures one mi ght then r epr esent each mecha-
ni sm by thr ee regul ari zed gri d structures correspondi ng to the thr ee
choi ces of output termi nal . Al ternati vel y, one mi ght present a si ngl e
r egul ar i zed gr i d structure. For each gi ven functi onal rel ati on i t woul d
then be necessary to construct thr ee r egul ar i zed gr i d structures, wi th
xi , Zi , and x,, i n turn, treated as the output vari abl e. A match between
one of these thr ee structures and a catal oged structure (after a possi bl e
r efl ecti on i n the di agonal ) woul d then show that the catal oged mecha-
ni sm coul d be used, and woul d i ndi cate the way i n whi ch the paramet.e
shoul d be associ ated wi th i ts termi nal s.
SEC. 8.7] USE OF GRI D STRUCTURES I N LI NKAGE DESI GN 2?43
8.7. Use of Gri d Structures i n Li nkage Desi gn.-The concept of gri d
structure i s of fundamental i mportance as an ai d i n desi gni ng gri d
generators for speci al appl i cati ons. For thi s purpose i t becomes neces-
sary to i ntroduce the fol l owi ng general i zati on of the i dea.
Generalized G% Structures.-I n the precedi ng di scussi on the gri d
structure of a functi onal rel ati on
has been defi ned as a system of l i nes i n the (zi , ~j)-pl ane: strai ght l i nes
representi ng constant val ues of z; and zj, whi ch form a rectangul ar gri d,
and a superi mposed fami l y of contours of constant Zk. Such a gri d
structure i s a ver y speci al form of i ntersecti on nomogram representi ng the
gi ven rel ati on.
Now i t i s not at al l necessary to treat xi and zj as cartesi an coordi nates.
I nstead, one can take the pl ane of representati on as the (zi , zj)-pl ane, and
l et w and Zi be any pai r of curvi l i near coordi nates i n thi s pl ane, gi ven
i n terms of the cartesi an coordi nates zi l z~,by
)
Zi = Zi(Zi, Zjj (51)
)
~j = ~~(Zi ,Z~.
The constructi on of the gri d structure then pr oceeds as before. Adjacent
contours B and C of constant Xkare chosen, and between them, begi nni ng
at the center S, ther e i s constructed a step structure consi sti ng of porti ons
of contours of constant xi and xj.
These l atter contours, extended
through the pl ane, make up a curvi l i near gri d, i nstead of the rectangul ar
one previ ousl y obtai ned. The compl ete gri d structure consi sts of thi s
gri d, together wi th contours of constant Xkwhi ch paw through the gri d
i ntersecti ons r = s, or i nterpol ate smoothl y between these contours.
The same gri d structure can be obtai ned i n a di fferent way. Let
Eqs. (51) defi ne a topol ogi cal transformati on between the (~i, xj)-pl ane
and the (~, zi )-pl ane. Thi s transformati on wi l l car r y a center S= i n the
&at pl ane i nto a center S. i n the second, and a contour C. i n the fi rst
i nto a contour C. i n the second.
I t wi l l al so transform the enti re gri d
structure defi ned i n the (a, xi )-pl ane, usi ng S=and C., i nto the gri d stmc-
tur e defi ned i n the (zi , zi )-pl ane, usi ng S. and C..
Such a topol ogi cal transformati on of the gri d structure wi l l , of course,
affect none of i ts i ntersecti on properti es; i n parti cul ar, i t wi l l sti l l ser ve as
an i ntersecti on nomogram representi ng the gi ven functi on.
We shal l ,
i n fact, consi der gri d structures whi ch di ffer onl y by a topol ogi cal trans-
formati on of the form of Eq. (51) as equi val ent representati ons of a func-
ti onal rel ati on.
Mez?uznkal Real~zat~on of a Given Grid Structure.-The authors
teohni que for mechani zi ng functi ons of two i ndependent vari abl es makes
244 LI NKAGES WI TH TWO DEGREES OF FREEDOM [SEC. 8.7
i mportant use of such topol ogi cal transformati ons of gri d structures.
The basi c i dea i s to transform the gi ven gri d structure i nto a form whi ch
suggests a sati sfactory mechani cal form for a gri d gener ator . The tech-
ni que empl oyed i n thi s transformati on wi l l be i ndi cated i n l ater chapters;
her e we shal l merel y take note of the way i n whi ch a gi ven gri d structure
may suggest a correspondi ng mechani zati on of the functi on.
.
*,=Q.-2 XA=cd
l/vl/4-k1
*=-al
I I zi + I
,1\l \l Uz,-1
FIG. 8.13.-Grid structure of zk = 2i/z, (a = U@
Consi der, for exampl e, the gri d structure of the rel ati on
Xi
~k=
Xj
(52)
as shown i n Fi g. 8.13. The spaci ngs of the rectangul ar gri d l i nes change
i n geometri cal sequence; the fi xed rati o i s her e 1.25. The contours of
constant xk are radi al l i nes; the correspondi ng val ue of Xkfor each l i ne i s
the val ue of xi at i ts i ntersecti on wi th the hori zontal l i ne xi = 1. At each
poi nt of thi s fi gure one can read off correspondi ng val ues of xi , xj, and zk
whi ch aati sf y Eq. (52). Now, l et ther e move over thi s fi gure a pi n con-
nected mechani cal y to thr ee cl i ffer ent scal es. I f these connecti ons and
scal es are so arranged that one can read on the fi rst scal e the val ue of Zi
at the posi ti on of the pi n, on the second scal e the val ue of xi , and on the
thi rd scal e the val ue of Zk, then the devi ce as a whol e becomes a mechani -
zati on of the gi ven functi on. I n the present case, the fi rst scal e shoul d
show the hori zontal di spl acement of the pi n from the ori gi n, the second
scal e i ta verti cal di spl acement; the readi ng of the thi rd scal e shoul d be
SECaq USE OF GRI I J STRUCTURES I N LI NKAGE DESI GN 246
FIQ. 8. 14.Transformed grid structure of z~ = Zilzi.
246 LI NKAGES WI TH TWO DEGREES OF FREEDOM [SEC. 8.7
proporti onal to the hori zontal di spl acement of the i ntersecti on of the
radi al l i ne through the pi n wi th a hori zontal l i ne.
The di vi der (or mul ti -
pl i er) of Fi g l "l Oaccompl i shes thi s i navery si mpl e andobvi ousway; i t
i s the natural mechani zati on of the gri d structure of Fi g. 1.12, whi ch
di ffers from Fi g. 813 onl y by a refl ecti on
A topol ogi cal transformati on of thi s gri d structure wi l l car r y i t i nto a
form (Fi g. 8.14) suggesti ng a ver y di fferent type of mechani zati on. The
hori zontal Li nes of Fi g. 8.13 are transformed i nto a fami l y of ci rcl es, aU of
the same radi us, Ll , wi th center s l yi ng on a strai ght l i ne Cl . The verti cal
l i nes of Fi g. 8.13 are tri msformed i nto a second fami l y of ci rcl es, al l of the
same radi us, Lz, wi th center s l yi ng on the cur ved l i ne C2. Fi nal l y, the
radi al l i nes of Fi g. 8.13 are transformed i nto a thi rd fami l y of curves.
These are ver y nearl y, al though not exactl y, ci rcl es wi th the same radi us,
L8; si nce the approxi mati ng ci rcl es i ntersect at a common poi nt thei r
center s must l i e on another ci rcl e wi th radi us L~curve Ca in the fi gure.
On i gnori ng the smal l devi ati on from ci rcul ar form of the curves of the
thi rd fami l y, we are l ed di rectl y to the mechani zati on shown i n Fi g. 8015-
an approxi mate di vi der or mul ti pl i er, but a qui te accurate one. The joi nt
P can be made to l i e on a parti cul ar ci rcl e of the fi rst fami l y by pl aci ng i t
at one end of a bar PA I wi th l ength Ll, and fi xi ng the joi nt A 1i n the center
of thi s ci rcl e, on l i ne Cl . Conversel y, i f the joi nt A 1 k constrai ned to
l i e on the l i ne Cl as by bei ng pi voted to a sl i de-i t wi l l necessari l y be
al ways at the center of the xi -ci rcl e on whi ch P l i es; a scal e pl aced al ong
Cl can thus be cal i brated to gi ve the val ue of xi at the posi ti on of the
pi n P. I n the same way, the val ue of z{ can be read on a scal e,l yi ng al ong
the cur ved l i ne Cz, usi ng as i ndex poi ~t the joi nt A z connected to the pi n
P by a bar of l ength Lz. Fi nal l y, val ues of the quoti ent zk mi ght be read
on the ci rcul ar scal e C~. I nstead of pi voti ng the bar PA 8, of l ength L,,
to a ci rcul ar sl i de, one can constrai n the poi nt A 8 to l i e on the cur ve Ca
by a second bar OAs, al so of l ength R*, pi voted at the center of thk ci rcl e.
As shown i n the fi gure, the i ndex poi nt has been transferred to thi s second
bar i n an obvi ous way.
The shortcomi ngs of thi s parti cul ar devi ce are obvi ous: the use of a
cur ved sl i de, and the nonl i neari ty of the xi - and zI c-scal es. To i mprove i t
me shoul d devi se a mor e sati sfactory way to gui de the poi nt A t al ong the
cur ve CZ, and shoul d l i neari ze the xi - and z~-scal e readi ngs by transformer
l i nkages, such as harmoni c transformers or three-bar l i nkages.
How thi s can be accompl i shed i s i l l ustrated i n Fi g. 8.16, whi ch shows
the fi rst l i nkage mul ti pl i er so desi gned as to be operabl e through a domai n
i ncl udi ng posi ti ve and negati ve val ues of al l vari abl es. The poi nt A ~
is constrai ned to fol l ow the cur ve Cz of Fi g. 8.15 by pl aci ng i t on an exten-
si on of the central bar of a three-bar l i nkage @Y& (The requi red desi gn
techni que i s i ndi cated i n. Sec. 1004.) Moti on of A z al ong CZ produces a
SEC. 8.7] USE OF GRI D STRUCTURES I N LI NKAGE DESI GN
P
247
FI CI .8.15.Mechani zati on of the gr i d structure of Fi g. 8.14.
248 LI NKAGES WI TH TWO DEGREES OF FREEDOM [Sm. 871
FI G. 8.16.Li near mechani zati on of z~ = Z,p$k.
SEC. 8.7] USE OF GRI D STRUCTURES I N LI NKAGE DESI GN 249
correspondi ng rotati on of the bar crj3of the thre~bar l i nkage; a noni deal
harmoni c transformer conver ts thi s rotati on i nto l i neari zed readi ngs on
the x,-scal e. TO l i neari ze the z~-scal e the rotati on of the bar OAS has
l i kewi se been conver ted i nto l i near moti on of a sl i de by means of a non-
i deal harmoni c transformer.
Thi s desi gn of a practi cal l i nkage mul ti pl i er has thus been arri ved at
i n thr ee steps:
1. Topol ogi cal transformati on of a mul ti pl i er gri d structure i nto a con-
veni ent form.
2. Desi gn of a si mpl e devi ce for mechani zi ng thi s gri d structure.
3. Conversi on of the desi gn to a mor e sati sfactory form, by appl yi ng
constrai nts i n a cl i ffer ent way and l i neari zi ng the scal e readi ngs.
An i mportant fourth step i s the fi nal adjustment of l i nkage di mensi ons.
These steps are by no means uni que, and one can desi gn a great vari ety of
l i nkage mul ti pl i ers. For i nstance, i t i s possi bl e to fi nd other transforma-
ti ons of the mul ti pl i er gri d structure i n whi ch the cur ve Cz becomes a
ci rcl e or a strai ght l i ne, and the desi gn of a l i nkage constrai nt for the
poi nt A, becomes tri vi al . To accompl i sh thi s one needs a thorough
understandi ng of the techni ques to be di scussed i n the next chapter.
CHAPTER 9
BAR-LI NKAGE MULTI PLI ERS
A techni que for desi gni ng bar-l i nkage mul ti pl i ers wi l l be devel oped i n
thi s chapter, both for i ts i ntri nsi c i nterest and as an exampl e of a general
techni que. The probl em of mechani zi ng any other functi onal rel ati on
wi th i deal gri d structure i s essenti al l y the same, as regards the desi gn of
the gri d gener ator ; di fferences ari se onl y i n the detai l s of transformer
l i i kage desi gn, whi ch wi l l need no di scussi on her e.
\
.. c~
. /y
FIG. 9.1.Star grid generator.
9.1. The Star Gri d Generator.The gri d gener ator consi dered
throughout thi s and the fol l owi ng chapter wi l l be the star gri d genera-
tor or star l i nkage i l l ustrated i n Fi g. 91. The general pri nci pl es to
be expl ai ned can be appl i ed to other gri d generators, but the detai l ed
constructi ons wi l l of cour se requi re modi fi cati on.
The r eader wi l l
r ecogni ze that i n the case of the star gri d gener ator these constructi ons
are parti cul arl y si mpl e; thi s si mpl i ci ty and the sati sfactory mechani cal
properti es of thi s devi ce gi ve i t a speci al useful ness i n practi ce.
The star l i nkage consi sts of thr ee l i nks, Ll, Lz, LS, joi ned together at
one end by a common joi nt P. The l engths of these l i nks we shal l al so
denote by L,, L,, La. At thei r far ends are fr ee joi nts A 1, As, A~, whi ch
are i n some manner gui ded al ong thr ee curves C,, Cz, C3. I nput and out-
put parameters, Yl , YA, Ys can be read at these joi nts on arbi trari l y
250
8mA 9.2]
DESI GNI NG A STAR GRI D GENERATOR 251
graduated scal es l yi ng al ong these curves. The l i nkage establ i shes
between these parameters a rel ati on
Y, = G(Y., Y,) (1)
whi ch characteri zes i ts behavi or as a gri d gener ator .
I t i s at once cl ear that any functi onal rel ati on that can be gener ated by
a star l i nkage can al so be r epr esented by an i ntersecti on nomogram con-
si sti ng of thr ee fami l i es of ci rcl es, of radi i L1, LZ, La, respecti vel y, r epr e-
senti ng constant val ues of the parameters Yl , Yi , Ys. I n fact, the l i nkage
coul d be used i n drawi ng thi s nomogram. I f the joi nt A, is fi xed at the
poi nt Y1 on the Cl -scal e, the joi nt P can then be made to descri be the
Y]-ci rcl e on the nomogram; ci rcl es correspondi ng to defi ni te val ues of
Yz and Y* can si mi l arl y be traced out by fi xi ng the joi nts A 2 and A ~,
respecti vel y. Each ci rcl e on the nomogram thus represents a cor r espond-
i ng poi nt on one of the thr ee scrdes. To each confi gurati on of the l i nkage
ther e corresponds a poi nt on the nomogram at whi ch thr ee ci rcl es i nter-
sect; correspondi ng scal e readi ngs and nomogram poi nts i ndi cate the
same tri pl et of val ues of the parameters, Y], Yz, Ya, sati sfyi ng Eq. (1).
I t fol l ows that any functi onal rel ati on that can be gener ated by a star
l i nkage must have a gri d structure that consi sts of thr ee fami l i es of ci rcl es
wi th fi xed radi i , or can be brought i nto such a form by the geueral topo-
l ogi cal transformati on, Eq. (8.51).
The gri d structure of a star gri d gener ator may be al most i deal over a
wi de range of parameter val ues, or strti gl y noni deal , dependi ng on the
l i nk l engths and the choi ce of curves Cl , C*, CS; i t i s thus useful i n mecha-
ni zi ng functi ons wi th ei ther i deal or noni deal gri d structure. I n the
present chapter we shal l be i nterested onl y i n desi gni ng star gri d gener -
ators wi th nearl y i deal gri d structure.
9s2. A Method for the Desi gn of Star Gri d Generators wi th Al most
I deal Gri d Structure.-I t wi l l be i nstructi ve to exami ne the gri d struc-
tur e of the star l i nkage shown i n Fi g. 9.1. Thi s can be done graphi -
cal l y as i l l ustrated i n Fi g. 92. We choose the center S of the gri d
structure, and mark the correspondi ng posi ti ons of the joi nts A 1, A 2, As,
on the thr ee scal es wi th the val ues of r, s, t,for thi s center , O, 0, 0. About
these poi nts we draw ci rcl es Cf~J,C$)j C(~),wi th radi i L], LZ, L~; these i nter-
sect at S. Let us choose C(j) as the contour B i n the gri d structure, thus
assi gfi ng to Ys the r ol e of xk i n Sec. 8.3. Near the poi nt t = 0 on the
Y3-scal e we sel ect another poi nt to cor r espond to t= 1. About thi s we
descri be a ci rcl e C(j, of radi us Ls, to ser ve as the contour C of the gri d
structure. Between the contours C$) and C(i ) we can now construct a
step structure consi sti ng of arcs of radi us L1, centeri ng on Cui ve Cl , and
arcs of radi us L2, centeri ng on cur ve CZ. Begi nni ng at S we can fol l ow
the ci rcl e C(!) to i ts i ntersecti on wi th the cur ve C(:). Thi s i ntersecti on
252 BARLI NKAGE MULTI PLI ERS [SEC. 9.2
corresponds to the i ndi ces r = O, t = 1; i n or der to sati sfy the rel ati on
r+s=t (2)
i t must al so be assi gned the i ndexs = 1. (Al l thr ee i ndi ces are i ndi cated
i n Fi g. 9.2, though any two of them woul d be suffi ci ent for i denti fi cati on.)
At a di stance L~from thi s poi nt ther e must l i e the poi nts = 1 on the cur ve
C2. About the poi nt s = 1 we descri be the ci rcl e C(\), whi ch i ntersects
the contour B = Cl? at a poi nt wi th the i ndi ces s = 1, t = O; the other
+4
\
+3
+2
C*
+1
Q
+5
FI G. 9.2.Gri d structure of a gi ven star l i nkage. The arms of the l i nkage are shown wi th
the commcm joi nt at the center of the gr i d Btructure.
i ndex must be r = 1. At a di stance L1 from thi s i ntersecti on ther e
must then l i e the poi nt r = 1 on cur ve C 1. By conti nui ng th~s process
we can bui l d up a step structure between the chosen contours, and estab-
l i sh on the curves Cl and C2 the sequence of poi nw correspondi ng to
i ntegral val ues of the i ndi ces r ands. The fami l i es of ci rcl es C(I ) and C(j)
about these poi nts form the basi c gri d of the gri d structure.
Al l thi s
fol l ows uni quel y from our choi ce of the center S = (O, O, O) and the poi nt
t=l on Cfi .
-.
SEC. 9.2] DEsI GNI NG A STAR GRI D GENERATOR 253
I f the gri d structure of thi s gri d gener ator wer e i deal , al l i ntersecti ons
of the gri d wi th r + s = t woul d l i e on a ci rcl e C\i )wi th radi us L3 and
center on the cur ve C9. Actual l y, as shown i n Fi g. 92, i t i s possi bl e to
construct ci rcl es C$) that pass ver y nearl y, though not exactl y, through
these i ntersecti ons. (Note, for i nstance, the di vergences i n the upper
ri ght-hand cor ner of the gri d structure.) For practi cal purposes the gri d
structure may be consi dered as i deal over the gr eater part of the domai n
i l l ustrated. Wi thi n thi s domai n i t wi l l mechani ze the rel ati on
X3 = Z1 + 22, (3)
i f the scal e cal i brati on i s that establ i shed by thk constructi on, or any other
rel ati on wi th i deal gri d structure, i f the scal es are pr oper l y transformed.
Fi gure 9.2 shows ver y cl earl y the system of curvi l i near tri angl es whi ch i s
the di sti ngui shi ng mark of i deal gri d structure. So nearl y i deal a gri d
structure i s by no means characteri sti c of star l i nkages.
Thi s parti cul ar
l i nkage has been expressl y desi gned to have an al most i deal gri d structure,
by a method whi ch wi l l now be descri bed i n detai l .
Our probl em i s essenti al l y that of constructi ng thr ee fami l i es of ci rcl es,
C~J, C~), C~) (wi th radi i Ll, L2, Lg, respecti vel y) whi ch i ntersect to form a
tri angul ar structure such as that shown i n Fi g. 9.2. We begi n by choosi ng
arbi trari l y si x poi nts i n a pl ane. Thr ee of these, A\), A\), A ~l), wi l l
ser ve as the poi nts r = 1, 0,
1, on the cur ve Cl of the compl eted l i nk-
age; they shoul d l i e on a l i ne of moderate curvature, wi th roughl y equal
spaci ngs, but can otherwi se be chosen at wi l l (cf. Fi g. 93). The other
thr ee poi nts, A $1), A$O), A $1), are to ser ve as the poi nts t = 1, 0, 1, on
the cur ve Ca, and shoul d be chosen subject to si mi l ar condi ti ons. About
the poi nts A \l ), A \ o), A ~1), construct ci rcl es C~, Cjo), C\-l ), wi th arbi trary
radi us LI ; si mi l arl y construct about the poi nts A f), A $0),A !-l), the ci rcl es
C$l ), Cj), C$-l ), wi th radi us L~. Since these ci rcl es are to form the basi s
of the gri d structure, Lg shoul d be so l arge that each C&i rcl e i ntersects
each ci rcl e C~J i n two wel l -separated poi nts.
The i ntersecti ons of the
ci rcl es wi l l then fal l i nto groups of ni ne, wel l separated i n the pl ane. one
of these groups wi l l l i e near the center of the gri d structure, whereas the
other wi l l l i e outsi de the regi on i n whi ch i t i s al most i deal ; i t i s for thi s
reason that the two groups of i ntersecti ons shoul d not be cl ose to each
other .
Choosi ng one of the two sets of i ntersecti ons, we l abel each i ntersec-
ti on wi th the correspondi ng i ndi ces r, s, ~: ( 1, 0, 1), ( 1, 1, O),
(1, 2, 1), (0, 1, 1), (0, 0, 0), (0, 1, 1), (1, 2, 1), (1, 1, o),
(1, O, 1). [The second i ndex i s i n each case chosen to sati sfy Eq. (2).]
These ni ne gri d i ntersecti ons have been chosen wi th a hi gh degr ee of
arbi trari ness; our probl em i s now to bui l d the gri d structure about thi s
nucl eus, mai ntai ni ng i ts i deal character so far as possi bl e by appropri ate
choi ce of the avai l abl e des~ constants.
254 BAR-LI NKAGE MULTI PLI ERS [SEC. 9.2
The thr ee i ntersecti ons ( 1,0, l ), (O, O, O), (1, O, 1) must al l l i e on
the ci rcl e C~O). By constructi ng thi s ci rcl e we can establ i sh i ts radi us,
Lz, and i ts center , the poi nt A j) on the cur ve CZ. The known poi nts,
( 1, 1, O) and (O, 1, 1), and the known radi us L, then ser ve to determi ne
Fm. 9.3.Construction of a star linkage with nearly ideal grid struoture
the ci rcl e C\l J, wi th center A\~); si mi l arl y the poi nta (O, 1, 1) and
(1, 1, O) determi ne the ci rcl e C!-) wi th center A~-u. These new ci rcl es
fi x four addi ti onal gri d i ntersecti ons: ( 2, 1, 1), (2, 1, 1), (1, 1, 2),
and (1, -1, 2)
Ther e are now determi ned the thr ee l i nk l engths, Ll, Ls, L,, and thr ee
poi nts on each of the curves Cl , (?z, C8. By passi ng smooth cuwti s
through these poi nti we can setup a star l i nkage wi th a nearl y i deal gri d
structure i n the nei ghborhood of the center S = (O,O,O). To i mprove the
accuracy of thi s constructi on, and to extend the domai n of nearl y i deal
gri d structure, i t i s nemsaary to determi ne other poi nts on these curves.
We now know that the poi nt A\ 2) must l i e on a ci rcl e Q~? of radiuaLt with
i ts center at (-1, 2, 1), and that the poi nt A\ -a lies on circle Q,-, with
SEC. 9.2]
DESI GNI NG A STAR GRI D GENERATOR 255
the same radi us and center at (1, 2, l ). Si mi l al l y A~2j and A~2)
l i e on ci rcl es of radi us L, about center s (2, 1, 1) and (2, 1, l ),
respecti vel y, and A $] and A $2)l i e on ci rcl es of radi us La about center s
(1, 1, 2) and ( 1, 1, 2). No i urther i nformati on i s to be extr acted
from the known poi nts of the gri d.
Now l et us make a tentati ve choi ce of the poi nt ( 2, 2, O), on the
known ci rcl e C$). Together wi th the known poi nt ( 1, 2, 1) thi s deter -
mi nes the ci rcl e C\*)and i ts center A~2);C\zj,i n turn, compl etes the deter -
mi nati on of the gri d i ntersecti on (O, 2, 2). By extensi on of thi s process,
a tentati ve choi ce of the poi nt ( 2, 2, O) l eads to equal l y tentati ve deter -
mi nati ons of other el ements of the gri d, accordi ng to the fol l owi ng scheme:
(2, 2,0) + (1, 2, l )+l l y, C\)+ (o, 2, 2),
(o, 2, 2) + (1, 1, 2) + A\, Cf + (2, o, 2),
(2, o, 2) + (2, 1, 1) + A\), Cp + (2, 2, o),
(2, 2, o) + (1, 2, 1) + A\-), c!-) + (o, 2, 2),
(o, 2, 2) + (1, 1, 2) +A\-), C$) + (2, o, 2),
(2, O, 2) + (2, 1, 1) -Af), c~-) + ( 2, 2, o).
Thus we arri ve fi nal l y at a constructi on for the poi nt ( 2, 2, O), wi th
whi ch the whol e process was started. rhi s constructi on wi l l , i n general ,
l ead back to the tentati vel y chosen i ni ti al poi nt onl y i f that choi ce was
made cor r ectl y. For exampl e, an i ncor r ect choi ce of the poi nt ( 2, 2, O)
l eads to constructi on of the dashed gri d l i nes of Fi g. 9.3. The curvi l i near
hexagon, traced out i n the cl ockwi se di recti on. fai l s to cl ose. A second
choi ce of the i ni ti al poi nt l eads to a di fferent er r or i n cl osi ng; i nterpol ati on
fi nal l y l eads to a cor r ect choi ce and the constructi on shown i n bol d l i nes.
Ther e are thus determi ned two addi ti onal poi nts on each of the curves
Cl , Cz, C,, and si x addi ti onal ci rcl es i n the gri d structure.
These si x new ci rcl es fi x addi ti onal gri d i ntersecti ons-enough of them,
i n fact, to determi ne i mmedi atel y two mor e poi nts on each of Cl , C2, C8:
A\), l l i -~, A!), A\-), A\), A$-). For exampl e, the i ntersecti on of
C\-a)and C\) determi nes (3, 2, 1), and that of C~-l ) and C~2)determi nes
(3, .1, 2); these poi nts, i n the l ower ri ght-hand comer of Fi g. 9.3, i n turn
determi ne A\s)and C~. I t i s at thi s poi nt that the fl exi bi l i ty i n the desi gn
becomes i nau5ci ent ta permi t constructi on of an exactl y i deal gri d struc-
ture: these new ci rcl es shoul d pass through certai n tri pl e i ntersecti ons, but
the constructi on does not assure that they wi l l do so. For i nstance, the
ci rcl es C~+a), C~-o, and C$) shoul d pass through a common poi nt, (3,
3, O); i n Fi g. 9.3 i t can be seen that they pass ve~ nearl y but not exactl y
through the same poi nt. A si mi l ar fai l ure occurs at (3, O, 3); per fect
tri pl e i ntemecti ons at these poi nts can be obtai ned onl y by changi ng the
Ori gi nal arbi trary assumpti ons. I n the present case thi s woul d hardl y
256
BAR-LI NKAGE MULTI PLI ERS
[SEC. 93
be wor th whi l e, as the l i nkage al ready determi ned has effecti vel y i deal
gri d structure i n a ver y l arge domai n.
9.3. Gri d Generators for Mul ti pl i cati on. -The exampl e of the pr eced-
i ng secti on shoul d make i t cl ear that i t i s a si mpl e and strai ghtforward task
to desi gn a star gri d gener ator wi th al most i deal gri d structure over a l arge
domai n. From a practi cal poi nt of vi ew thi s i s onl y a begi nni ng i n the
wor k of desi gni ng a sati sfactory star gri d gener ator for mul ti pl i cati on.
Other aspects of thi s probl em wi l l now be i ndi cated.
Li ke any other gri d gener ator wi th i deal gri d structure i n an extended
domai n, the star l i nkage of the precedi ng secti on can be used i n desi gni ng a
mul ti pl i er. Cal i brati on of the scal es i n terms of the vari abl es
Z1 = kr, ZZ = ks, Z3 = kt (5)
wi l l conver t i t i nto an adder wi th ver y si mpl e structure, mechani zi ng the
rel ati on
zl +z*=z~ (6)
throughout a domai n wi thi n whi ch each vari abl e maybe ei ther posi ti ve or
negati ve. Recal i brati on i n terms of vari abl es xl , x2, x3, determi ned by
l ogl a Z1 = 21, l ogl a Z2 = 22, l ogl o X3= 23,
(7)
wi l l conver t i t i nto a mul ti pl i er mechani zi ng
2122 = 23 (8)
throughout a domai n i n whi ch all vari abl es are posi ti ve.
The most obvi ous di sadvantage of such a mul ti pl i er i s i ts use of
cur ved sl i des. A mor e general l y useful devi ce coul d be desi gned i f the
curves Cl , C~, Ct wer e of si mpl e, mechani cal l y desi rabl e forms; el aborati on
of the star l i nkage i nto a mul ti pl i er l i ke that of Fi g. 815 or Fi g. 8.16 woul d
then fol l ow the l i nes i ndi cated i n Sec. 8.7. One can i n fact bri ng the
curves Cl , Cz, C3 i nto desi rabl e forms by maki ng appropri ate changes i n
those el ements of the desi gn that wer e arbi trari l y chosen i n Sec. 9.2. A
sati sfactory method for doi ng thi s wi l l be i ndi cated i n Sec. 9.5, wher e i t
can be i l l ustrated i n connecti on wi th a probl em havi ng addi ti onal features
of i nterest.
Mul ti pl i ers desi gned i n thi s way do not permi t change i n si gn of any
factor. I f z1, for i nstance, i s to pass through O, then z, and r must pass
through the cwrespondi ng val ue =. To accompl i sh thi s i n a mecha-
ni sm wi th fi ni te travel one must have a zl -scal e of fi ni te l ength; the
sequence of poi nts A.f) must approach a poi nt of condensati on as r * w.
Even when one has assured the exi stence of such a poi nt of condensati on,
correapcmdmg to Z1 = 0, i t wi l l be neceesa~ to face the probl em of
extendi ng the scal e i nto the regi on of negati ve a.
I
SEC. +31 GRI D GENERATORS FOR MULTI PLI CATI ON 257
Such poi nts of condensati on have thei r equi val ents i n the gri d struc-
tur e of the rel ati on
(9)
an al ternati ve form of Eq. (8). I n Fi g. 9.4 thi s gri d structure i s devel oped
about the center X2 = X3 = 1, wi th the l i ne xl = 1.25 chosen as the con-
tour C. The step structure between the contours B and C approaches the
ori gi n i n an i nfi ni te number of steps; the successi vel y added l i nes of the
rectangul ar gri d wi l l tend to fi l l out the domai n X2 >0, X3 >0, but wi l l
never extend outsi de thi s regi on. The gri d i s of cour se i deal , and i ncl udes
the contours xl = (1 .25)r, wi th T taki ng on al l i ntegral val ues; as r - m
these contours approach the hori -
zontal axi s, and as r - m they
approach the verti cal axi s.
To obtai n gri d structures for
thk rel ati on i n al l four quadrants,
one must use a separate center
for the gri d structure i n each
quadrant. I f these four poi nts
are chosen i n si mi l ar posi ti ons i n
the four quadrants, symmetri cal
wi th r espect to the two axes, and
i f correspondi ng contours C are
used, then the four gri d structures
wi l l approach the coordi nate axes
symmetri cal l y. They wi l l then
Fm. 9.4.Gri d structure of m = z~/zz.
appear to fl ow smoothl y i nto each other i n crossi ng these axes, and the
whol e fi gure wi l l take on the appearance of a si ngl e gri d structure (Fi g.
9.5). I t i s i mportant, however , to remember that thi s resul t i s obtai ned
arti fi ci al l y, and that the coordi nate axes are l i nes of condensati on i n the
gri d structure.
A topol ogi cal transformati on of Fi g. 95 that woul d car r y the thr ee
fami l i es of strai ght l i nes i nto thr ee fami l i es of ci rcl es of constant radi us
woul d conver t i t i nto the i deal gri d structure of a star l i nkage. To
each ci rcl e of thi s gri d structure ther e woul d cor r espond an i ntegral val ue
of r, s, or t,and a cal i brati on poi nt on one of the scal es; to the ci rcl es
obtai ned by transformati on of the coordi nate axes ther e woul d cor r espond
poi nts of condensati on of the scal e cal i brati ons. Such a star l i nkage
woul d thus have the characteri sti cs to be demanded i n a gri d gener ator for
a mul ti pl i er that must al l ow change of si gn i n the vari abl es.
We shal l now use thi s i dea as a gui de i n desi gni ng a ver y sati sfactory
star gri d gener ator for mul ti pl i cati on.
258 BAR-LI NKAGE MULTI PLI ERS [SEC. 9.4
9.4. A Topol ogi cal Tranaformati on of the Gri d Str uctur e of a Mul ti -
pl i er.-Let us attempt to transform the gri d structure of Fi g. 9.5 (car-
tesi an coordi nates z*, XJ i nto an i deal gri d structure i n whi ch each of the
thr ee fami l i es of strai ght l i nes i n the ori gi nal structure i s r epr esented by a
fami l y of congruent ci rcl es (cartesi an coordi nates y,, Ua).
Fi rst, l et us consi der the fami l y of l i nes of constant z1. I n the ori gi nal
gri d structure al l these l i nes i ntersect at a common poi nt O. Such a
pr oper ty wi l l not be changed by a topol ogi cal transformati on; i n the
Fm. 9.5.-Gri d structure of a mul ti pl i er or di vi der permi tti ng changes i n si m of the
vari abl ee.
transformed gri d stmcture the correspondi ng fami l y of ci rcl es of radi us L1
must al l i ntersect at a common poi nt O. (See Fi g. 96.) I t fol l ows that
the center s of these ci rcl es must al l l i e on a ci rcul ar arc ab wi th radi us L1
and center at Ot. The radi us L1 can be chosen at wi l l , as the probl em i s
i ndependent of the scal e of constructi on; i t i s usual l y conveni ent to take
LI as the uni t of l ength.
The rel ati on between a gi ven strai ght l i ne of the ori gi nal gri d stmc-
tur e and the ci rcl e i nto whi ch i t i s transformed may be establ i shed by
exami ni ng the topol ogi cal transformati on i n the nei ghborhood of the
SEC. 9.4] TRANSFORMATI ON OF A GRI D STRUCTURE
259
I
a
\
ao
a-l
% o Q
\
%
/
42
b
-1
/
b.
@
l %. 9.6.Fi rst topol ogi cal transformati on of the gr i d structure of a mul ti pl i er.
.
260 BAR-LI NKAGE MULTI PLI ERS [SEC. 9.4
ori gi n O. I n i ts general form thi s transformati on i s
X2= Xz( yz, ya),
X3= G( gj , yJ.
1
(l o)
I n the nei ghborhood of O, wher e al l vari abl es can be tr eated as smal l
quanti ti es, thi s reduces to
X2 = C22Y2 + C23Y3,
}
(11)
x3 = c32.y2+ c33y3,
on negl ect of smal l quanti ti es of the second or der (Thi s, of course, i s
val i d onl y i f O i s not a si ngul ar poi nt of the transformati on.) Let
us assume that the transformed grid structure is symmetrical with respect
to the horizontal axis, in the nei ghborhood of the ori gi n. Then we must
have
CZ3= C32 = O, (12)
and
Z3
C3aY3
. ,
X2 C22yz
(13)
i n the i mmedi ate nei ghborhood of O.
I n other words, a l i ne of sl ope xl
at the ori gi n O of the ori gi nal gri d structure i s transformed i nto a cur ve
through the ori gi n O wi th sl ope changed by the constant factor (cS81CZ.J.
Si nce the sl opes of the successi ve zl -contours i n Fi g. 9.5 change i n a
geometr i c progressi on, the sl opes at O of the correspondi ng ci rcl es must
al so change i n geometr i c progressi on, and by the same rati o (1.25). Thi s
i s tr ue al so of the sl opes of the radi i from the ori gi n O to the center s of
these ci rcl es, whi ch are the negati ve reci procal s of the sl opes of the ci rcl es
themsel ves. Choosi ng arbi trari l y a val ue of cJc,z, we can then con-
struct, i n the transformed gri d structure, ci rcl es correspondi ng to each
of the l i nes of constant Z1 i n the ori gi nal gri d structure. I n Fi g. 9.6 ther e
are i ndi cated four of these ci rcl es (A O,A- 1, Az, As) wi th center s above
the hori zontal axi s at poi nts ao, al , at, as; the di sti ngui shi ng subscri pts
are the r val ues of the ori gi nal gri d l i nes, whi ch l i e i n the second and fourth
quadrants of Fi g. 9.5. The correspondi ng l i nes i n the fi rst and thi rd
quadrants transform i nto the ci rcl es Bo, B- _l , B-z, B-.3, wi th center s bo,
b-l , b-a, b-s, whi ch are the mi rror i mages of ao, al, az, a-s, i n the hori -
zontal axi s. The sequence of poi nts ao, a_,. a_2, . . . , whi ch l i es i n the
domai n of posi ti ve xl , has a poi nt of condensati on a_w on the hori zontal
axi s; an extensi on of the xl -scal e i nto the domai n of negati ve x, i s
provi ded by the symmetri cal l y pl aced sequence b,, b_l, . . . , whi ch has
the same poi nt of condensati on. Thi s poi nt i s of cour se the center of
the ci rcl e i nto whi ch the verti cal axi s of the ori gi nal gri d structure i s
transformed.
SW. 9-4] TRANSFORMATI ON OF A GRI D STRUCTURE 261
Consi derati on of thi s fami l y of ci rcl es wi l l make i t cl ear that ther e
exi sts no topol ogi cal transformati on of the type under di scussi on whi ch
maps the whol e of the (x2, m)-pl ane onto the (Y2, yJ-pl ane. The ori gi nal
)
contours of constant x] i ntersect onl y at the ori gi n and at i nfi ni ty, but the
ci rcl es i nto whi ch we are attempti ng to transform them may i ntersect
anywhere i n the (yz, yJ-pl ane, i f arbi trari l y l arge val ues of r or m are
admi tted. We can at best hope to establ i sh a topol ogi cal transformati on
that carri es a porf~on of the ori gi nal gri d structure (certai nl y one wi thi n
whi ch the magni tude of m i s l i mi ted) i nto a gri d structure consi sti ng
of arcs of ci rcl es. Thi s i s qui te suffi ci ent for our purposes, si nce r = co,
m = O, are not excl uded from the domai n of the transformati on.
Now l et us consi der the fami l y of ci rcl es of constant x3. Si nce the
ori gi nal l i nes X3 = c wer e symmetri cal wi th r espect to the hori zontal axi s,
i t i s natural to gi ve the transformed ci rcl es si mi l ar symmetry; thei r center s
must l i e on the hori zontal axi s cd. The verti cal axi s m = O i s al ready
known to be transformed i nto a ci rcl e of radi us L,, wi th center at A-=.
I t fol l ows that thi s second fami l y of ci rcl es must have the same radi us as
the fi rst: Ls = L1. Si nce the l i nes X8 = Z$) conver ge on X3 = O as
t + a, the poi nt a-w must be a poi nt of condensati on on the x3-scal e,
as wel l as on the zl -scal e. The l i nes of the ori gi nal gri d i ntersect the
xg-axi s at Z$;); the transformed ci rcl es must i ntersect the y3-axi s at poi nts
determi ned by
z$~)= X3(O, yi~)), (14)
or , i n the i mmedi ate nei ghborhood of O, by
The val ues of z!) go to zer o i n a geometri cal progressi on (rati o 1.25) as
t+m. I t fol l ows that the sequence of val ues yj) approaches zer o, and
the centens of the zs ci rcl es approach a~, i n a si mi l ar progressi on as
t+m, As a fi rst attempt to fi nd a transformati on of the desi red
character, l et us assume that thi s geometri cal progressi on i s exact, rather
than an approxi mati on val i d onl y i n the nei ghborhood of O; that i s, we
assume that Eq. (15) i s val i d for al l t.
Then, after choosi ng arbi trari l y a
val ue of cS3, we can constmct the ci rcl e correspondi ng to any x$-l i ne of
the ori gi nal gri d structure. I n Fi g. 9.6 ther e are shown ei ght of these
ci rcl es, wi th center s CO,c-l , c-2, c--s, and do, d-l, d_z, d_t.
(The subscri pt
i s the i ndex t.) The cs l i e i n the domai n of posi ti ve x3, and have a con-
densati on poi nt at x, = O; the ds provi de, somewhat arti fi ci al l y, an
extensi on of the scal e i nto the domai n of negati ve 2s.
The assumpti ons made up to thi s poi nt [transformati on of the xl - and
z s-contours i nto fami l i es of ci rcl es symrnetri cal to the hori zontal axi s,
val i di ty of Eq. (15), and speci al val ues of CZ2and CM] determi ne two
262 BAR-LI NKAGE MULTI PLI ERS [SEC.9-4
fami l i es of i ntersecti ng ci rcl es, and thereby determi ne compl etel y the
nature of the topol ogi cal transformati on. I t remai ns to be seen whether
thi s transformati on has the desi red character-whether the m-contours
are al so transformed i nto a fami l y of ci rcl es wi th common radi us Lz. I t
i s i mmedi atel y evi dent that thi s i s not the case.
I n Fi g. 9.6 ther e appear
64 poi nts of i ntersecti on of the z,- and x~contours, di sti ngui shed by
smal l ci rcl es. These are the transformed posi ti ons of the i ntersecti ons i n
the ori gi nal gri d, and through them must pass the transformed contours
of constant x2. I t wi l l be obser ved that the i ntersecti ons i n the l ower
hal f of the gri d l i e on curves that are concave upward, whereas those i n the
upper hal f l i e i n curves (not shown) that are concave downward. By
the symmetry of the constructi on, the strai ght l i ne w = Omust be tran~
formed, not i nto a ci rcl e, but i nto the strai ght l i ne yZ = O; the radi i of
curvature of the other x~contours i ncrease as they approach thi s l i mi ti ng
strai ght l i ne. The transformed i deal gri d structure i s not that of a star
l i nkage; i ndeed i t i s evi dent that i f such a gri d structure exi sts i t must be
an unsymmetri cal one. We can, however , approxi mate thi s gri d struc-
tur e by the noni deal gri d structure of a star l i nkage, repl aci ng the system
of m-contours of the i deal gri d structure by a system of approxi mati ng
ci rcl es of the same radi us. Thi s approxi matel y i deal gri d structure,
whi ch does have the other characteri sti cs that we desi re, wi l l l ater be made
much mor e nearl y i deal by readjustment of the desi gn constants.
I t i s possi bl e to pass ci rcl es ver y nearl y through al l the i ntersecti ons
i n the l ower hal f pl ane of Fi g. 9.6, by choosi ng a mean val ue La for the
radi us of curvature and l ocati ng the center s of the ci rcl es i n the upper
hal f pl ane. Thi s wi l l establ i sh the general posi ti on of m-scal e, and
wi l l make i t necessary to pass through the i ntersecti ons of the upper hal f
pl ane ci rcul ar arcs that are concave upward; the fi t ther e cannot be ver y
good, and we must spl i t the er r or s of constructi on as wel l as Dossi bl e.
The best way to do thi s i s to construct a ci rcl e through one set of
i ntersecti ons i n or der to establ i sh a radi us Lz. (I n Fi g. 9.6, Eo i s the
ci rcl e i n questi on, and the radi us chosen i s just equal to L1 and L3.)
Wi th thi s radi us, we construct arcs about each of the known gri d i nter-
secti ons. I f the gri d structure under constructi on wer e to be i deal , the
arcs characteri zed by a gi ven val ue of s = t r woul d al l i ntersect i n a
common poi nt. Thi s i s not the case her e; i nstead of poi nts of i ntersecti on
ther e exi st mor e or l ess di ffuse regi ons of i ntersecti on, wi thi n whi ch we
can l ocate the center s of the gri d ci rcl es wi th some degr ee of arbi trari ness.
Thi s arbi trari ness can be used to good advantage. I f a si mpl e mechani -
zati on of the gri d structure i s to be possi bl e, the zz-scal e must l i e i n a
si mpl e curve, preferabl y a strai ght l i ne or a ci rcl e. I n the present case,
the regi ons of i ntersecti on l i e roughl y on a ci rcl e. I n parti cul ar, the
ci rcul ar arc e, wi th center at Q, passes ni cel y through al l regi ons of i nter-
1
SEC.9,4] TRANSFORMATI ON OF A GRI D STRUCTURE 263
secti on except for thr ee at the extr eme ends. Thi s ci rcl e wi l l be taken
as the m-scal e on thi s scal e, and as near to the center s of the regi ons of
i ntersecti on as i s possi bl e we choose the center s el , eo, e-l , . . . of the
gri d structure ci rcl es El , Eo, E-1, . . . i n the l ower hal f pl ane, and the
center s fl , jo, f-l , . . . , of the gri d structure ci rcl es Fl , FO,F_l , . . . , i n
the upper hal f pl ane. The gri d stmcture ci rcl es must conver ge on a
FIG. 9.7.FirtA approximate multiplier.
through the ori gi n O. The center of the ci rcl e E-. ci rcl e E-m = F_w
i s e~ = f _m, the poi nt of condensati on of the poi nt sequence e. i n the
posi ti ve domai n of x2, and of the poi nt sequence f ,i n the negati ve domai n,
, i t i s the zer o poi nt on the x~scal e. Thi s compl etes the ~s+w.
determi nati on of the constants of the star l i nkage.
The gri d structure of Fi g. 96 i s mechani zed by the approxi mate
mul ti pl i er sketched i n Fi g. 9.7. Thi s consi sts of a star gri d gener ator
264 BARLI NKAGE MULTI PLI ERS
[SEC. 9.5
wi th arms LI , L2, Ls, meeti ng at a common joi nt P. The fr ee end of Ll
i s for ced to move al ong a ci rcl e a wi th center O, and the fr ee end of L2
al ong a ci rcl e e wi th center Q, by arms RI and R2, respecti vel y; the fr ee
end of L~ moves i n a strai ght sl i de c. I t fol l ows from the theor y of the
transformati on that the l engths of L1, L~j and R 1must be equal ; Lz al so
has thi s l ength, but onl y acci dental l y.
To use thi s devi ce as a mul ti pl i er,
the scal es must be cal i brated i n terms of the vari abl es z,, x,, z,, rel ated
to the i ndi ces r, s t, by Eqs. (5) and (7). The scal e poi nts of Fi g. 9.6
thus occur for val ues of z,, x2, x3, whi ch change i n geometr i c progressi on;
these are the scal e poi nts shown i n Fi g. 97. One can easi l y show that
xl = K1 tan O, (16)
X3 = K3h. (17)
Cal i brati ons on the z~scal e fol l ow (though not uni quel y, si nce the
mul ti pl i er i s not exact) from the rel ati on ZS = z MZ.
The r eader shoul d sketch thk mechani sm for z, = O and for Xz = O,
i n or der to see why i n these cases the val ue of zs i s necessari l yy zer o. He
wi l l al so be abl e to show by si mpl e geometr y why the mul ti pl i cati on i s
al most exact for smal l val ues of z I and x2.
9.6. I mprovement of the Star Gri d Gener ator for Muhi pl i cati on.
The er r or s i n the mul ti pl i er of Fi g. 97 appear ver y cl earl y i n i ts gri d
structure, i n whi ch many of the tri pl e i ntersecti ons characteri sti c of
i deal gri d structure have di si ntegrated i nto smal l tri angl es. To i mprove
thi s desi gn we must change one or mor e of the fami l i es of gri d ci rcl es i n
such a way as to r educe the si ze of these tri angl es. Thi s can be done by a
method that i s useful whether the functi on to be mechani zed has i deal or
noni deal gri d structure.
Let us exami ne the possi bi l i ty of i mprovi ng the gri d structure of
Fi g. 9.6 by changi ng the fami l y of ci rcl es CD, whi l e keepi ng fi xed the
fami l i es AB and EF. I n Fi g. 9.8 ther e are shown ei ght of the ci rcl es
AB and 10 of the ci rcl es EF, defi ni ng 80 poi nts of i ntersecti on through
whi ch ci rcl es CD shoul d pass. I t has al ready been noted that the ci rcl es
CD must have the same radi us as the ci rcl es AB; to i mprove the gri d
structure we can change onl y the posi ti ons of thei r centers. Let us
attempt to do thi s by the method of the precedi ng secti on, drawi ng arcs
of radi us LS about the 80 i ntersecti ons of the gri d. As befor e, the arcs
wi th a common i ndex t do not i ntersect i n a common poi nt, as they must
i f an i deal gr i d structure i s to be obtai ned. I nstead, we Obsel -ve an
i nteresti ng and characteri sti c phenomenon: arcs wi th centers i n the same
quadrant of the gr i d structure i ntersect ni cel y i n a seri es of poi nts that
defi ne a cur ved x3-scal e-but a di ffer ent scal e for each quadrant. Thi s i s
due to the fact that the four quadrants of the ori gi nal gri d structure are
actual l y i ndependent, and have been associ ated wi th each other i n a
SEC. 9.5] I MPROVEMENT OF A STAR GRI D GENERATOR 265
symmetri cal , but essenti al l y arti fi ci al , manner. I n Fi g. 9% the poi nts
correspondi ng to posi ti ve ZZ have been marked wi th smal l ci rcl es, the
others wi th dots.
e{
FIG. 9.8.First step in redesigning the multiplier grid structure.
I t i s evi dent that we can add ci rcl es CD to pr oduce a gri d structure
that i s nearl y i deal i n vari ous pai rs of quadrantsthose wi th X2 >0, or
wi th X2 < 0, or xl > 0, or x, < Obut that i t wi l l then be far from
i deal i n the other pai r of quadrants. To state i t di fferentl y, we can

266 BARLI NKAGE MULTI PLI ERS [sac. ,9.5

combi ne the scal e segments of Fi g. 9.8 i n several ways to form an z~scal e,
obtai ni ng a mul ti pl i er that wi l l be ver y accurate so l ong as one of the
factors has a parti cul ar si gn, but much l ess accurate when i t has the
other si gn. On the other hand, the di fference between the di vergi ng
scal e segments i s too great to permi t a sati sfactory compromi se; no
choi ce of form for the z3-scal e can make the gri d structure nearl y i deal i n
al l quadrants To obtai n an deal grti structure the chcles AB and EF
must be so chosen thut the diverpng zrscale segments coalesce to form a mnglc
z~scale. Thi s method of stati ng the probl em reduces i t to an especi al l y
conveni ent form, whi ch can be sol ved by al ternatel y adjusti ng the
ci rcl es AB and EF. We shal l choose to change fi rst the fami l y of ci rcl es
EF.
I n varyi ng the gri d structure ther e are two i mportant pri nci pl es to be
observed:
1. The gri d structure shoul d not be gi ven mor e than one degr ee of
fr eedom at a ti me.
2. The gri d structure shoul d not be excessi vel y sensi ti ve to changes
i n the vaned parameters. Thi s can often be assured by changi ng
parameters i n such a way that certai n el ements of the gri d structure
remai n unchanged.
For exampl e, to i mprove the gri d structure of Fi g. 9% l et us rearrange
the ci rcl es EF wi thout changi ng thei r common radi us. Let us mai ntai n
the ci rcul ar form of the zz-scal e, but r otate i t about the poi nt e_~ = j_~,
thus keepi ng unchanged the ci rcl e E_~ = F_.. Thi s rotati on gi ves
the one degr ee of fr eedom that we desi re i n the probl em, accordi ng to
Pri nci pl e 1; we must ther efor e r emove al l fr eedom i n the cal i brati on of
the scal e. By Pri nci pl e 2, the rul e for thi s cal i brati on must be such
that the gri d structure changes onl y sl owl y wi th rotati on of the z~scal e;
we shal l ther efor e demand that i t keep unchanged the gri d i ntersecti ons
marked wi th bol d dots i n Fi g. 9.8.
Let rotati on of the x,-scal e, e, about the poi nt e_= car r y i t i nto the
posi ti on e (Fi g. 9.8). The cal i brati on poi nts e:, j: on thi s new scal e
must l i e at di stance Lz from the correspondi ng fi xed gri d i ntersecti ons, and
are easi l y constructed. The new system of gri d ci rcl es I I F can then be
drawn, and fi nal l y, by constructi ng arcs about the new gri d i ntersecti ons,
the ne .Vset of z3-scal e segments. Thus the whol e constructi on does have
one degr ee of freedom, and i t i s easy to study the effect on the form of the
z3-scal e segments of rotati ng the z~-scal e. Tri al wi l l show that rotati on
of the scal e i n a cl ockwi se di recti on bri ngs cl oser together the two zs-scal e
segments to the ri ght of a_w; by an i nterpol ati on or extrapol ati on one
fi nds a rotati on of e whi ch makes the separati on of these scal e segments
ver y smal l . Thk i s the rotati on shown i n Fi g. 9.8, the tentati vel y chosen
SEC. 9.5] I MPROVEMENT OF A STAR GRI D GENERATOR 267
rotati on havi ng been omi tted as of l i ttl e i nterest. The new scal e and
the new ci rcl es EF are shown i n Fi g. 9.9, together wi th the new form
of the m-scal e constmcti on. The i mprovement i n the agreement of the
z~scal e segments on the ri ght i s ver y stri ki ng, but adjustment of the
ci rcl es Al l wi l l be requi red to i mprove the agreement on the l eft,
\
\ /// /
F1~, 9.9.Second step in redesigning the multiplier grid structure.
To i mprove the gri d structure further l et us rearrange the ci rcl es Al ?
Wi thout changi ng thei r radi i . To do thi s we shal l keep fi xed the form
of the xl -scal e, ab, whi l e rotati ng i t about the poi nt a_W. Cal i brati ons
on the new scal e wi l l be hel d at a fi xed di stance L ~from the gri d i nter-
1
secti ons i ndi cated by bol d dots i n Fi g. 9.9. Wi th these c~anges we
I
must make one other change, whi ch has no paral l el i n the step pre~i ousl y
descri bed. Rotati on of the x,-scal e wi l l move i ts center from O to O.
.L
268 BAR-LI NKAGE MULTI PLI ERS [sm. 9,5
The new ci rcl es AB will all i ntersect at thi s poi nt, through whi ch ther e
must al so pass the conver gence l i mi t of the ci rcl es EF, E-. I t i s
ther efor e necessary to keep Lz al ways equal to the di stance fr om e_=
to O as the xl scal e i s rotated. We have then to consi der a si mul taneous
vari ati on of both the ci rcl es AB and EF, but i t i s a vari ati on wi th one
degr ee of fr eedom whi ch offer s no di ffi cul ti es.
F1~. 9.10.Improved multiplier grid structure.
Begi nni ng wi th a tri al rotati on of the zl -scal e i nto a posi ti on ab
(Fi g. 99), one can establ i sh cal i brati on poi nts a:, b: by dr awi ng arcs
about the chosen poi nts of the ori gi nal gri d. The A B ci rcl es can then be
constructed, the new Lz deter mi ned, and the EF ci rcl es constructed.
Fi nal l y, the new x~scal e segments can be establ i shed, and the best
angul ar posi ti on for the xl -scal e deter mi ned by an i nterpol ati on or
SEC. 9.5] I MPROVEMENT OF A STAR GRI D GENERATOR 269
extrapol ati on. I n Fi g. 9.10 thi s constructi on i s shown for an x,-scal e
deter mi ned by such an i nterpol ati on. 1 The four z~scal e segments do not
FrQ. 9.1l .MuMpl i er gr i d structure wi th ci rcul ar z~cal e.
mer ge exactl y, but the groups of arcs make acceptabl y sharp i ntersecti ons,
through whi ch a strai ght l i me can be l ai d by a smal l sacri fi ce i n the fi t at
the extr eme ri ght end of the scal e. The ci rcl es CD constructed about,
I I t wi U be obesr ved that, becauee of the change i n L,, aU gr i d i nterjecti ons have
been shi fted, even those used i n eanetructi ng the new z,-eeal e. (Ol d posi ti ons are
ehown by bol d dots i n Fi g. 9.10.) Thi s i s not a matter of i mportance; afl that i e
requi redof the constmcti on i a that i t shal f not shi ft the gri d etmcture w vi ol entl y.
270 BAR-LI NKAGE MULTI PLI ERS [SEC. 9.5
poi nts on thi s recti fi ed xJ-scal e are shown i n Fi g. 9.10. I n vi ew of the
~mal l number of steps r equi r ed, the resul t can be consi der ed ver y sati s-
factory: the gr i d structure i s so near l y i deal over a wi de domai n that
anal yti cal methods can be empl oyed for i ts fur ther i mpr ovement.
I t i s evi dent that the gr i d structure of Fi g. 9.10 i s not the onl y possi bl e
sol uti on of our ori gi nal pr obl em. The gr i d structure of Fi g. 9.8 coul d
FI CA9.12.Multiplier grid structure ideal through a very large. domain.
have been vari ed i n many ways other than those descri bed above to
obtai n even l arger domai ns of nearl y i deal gri d structure, and cl i ffer ent
forms of the scal es. I ndeed, thi s method i s as useful for control of the
form and extent of scal es as i t i s for the i mprovement of gri d structures.
For i nstance, the l i near z~scal e of Fi g. 9.10 i s easi l y conver ted i nto the
ci rcul ar arc shown i n Fi g. 9.11. Thi s form was obtai ned by rearrangi ng
the ci rcl es EF, thi s ti me by i ncreasi ng the radi us of the zrscal e, whi l e
I
SEC. 9.6]
DESI GN OF TRANSFORMER LI NKAGES
271
keepi ng fi xed the poi nts e__ and e,.
New scal e cal i brati ons wer e so
chosen as to keep fi xed the gri d i ntersecti ons i ndi cated by bol d dots i n
Fi g. 9.11. Thi s fi gure shows al so the new fami l y of ci rcl es EF, and the
resul ti ng ci rcul ar form of the zs-scal e; the curvature of thi s can be
changed at wi l l by choosi ng a new center Q! for the xz-scal e.
A mor e
el aborate seri es of vari ati ons l eads to the gri d structure shown i n Fi g.
9.12, ~th nearl y strai ght scal e. The outstandi ng characteri sti c of thi s
gri d structure i s the l arge domai n wi thi n whi ch i t remai ns effecti vel y i deal .
,
/
o
.
b-l- .Aw
I
I
>6
+o-
~
-%
%&
s
\
%
Fm. 9.13.Improved multiplier, with the grid structure shown in Fig. 9.10.
9.6. Desi gn of Transformer Li i es.-The gri d structure of Fi g. 9.10
suggests the desi gn for a mul ti pl i er shown i n Fi g. 9.13. The structure
of thi s mul ti pl i er i s the same as that of Fi g. 97; i ncreased accuracy has
been obtai ned by changes i n the di mensi ons, but the number and rel ati ons
of the el ements remai n unchanged. The ranges of the vari abl es can
be changed wi th some freedom: the readi ngs on the z+, zr, and zrac~es
carI be changed by factors a, b, and c, respecti vel y, provi ded or dy that
Ob=c.
(18)
272 BAR-LI NKAGE MULTI PLI ERS [SEC. 9.6
Such a mul ti pl i er, wi th nonuni form scal es, i s of l i mi ted i nterest; the
real i mportance of thi s devi ce l i es i n the possi bi l i ty of usi ng i t to dri ve a
computer. I n such an appl i cati on i t may be r egar ded as an i deal gri d
gener ator and used together wi th transformer l i nkages i n the linear
mechani zati on of the rel ati ons
X3 = X1X2, (19)
X3 =fl (zl )jz(zz), (20)
or i ndeed of any functi onal rel ati on mth i deal gri d stmcture. I f a l i near
mechani zati on of Eq. (19) i s requi red, the functi on of the transformer
l i nkages may be r egar ded as that of repl aci ng the nonuni form scal es of
Fi g. 9.13 by uni form scal es; i n other cases the transformer l i nkages ser ve
al so as functi on generators.
Let us consi der fi rst the probl em of desi gni ng a mul ti pl i er wi th uni form
scal es. To descri be the confi gurati on of the gri d gener ator we may use
i nternal parameters Yl , Y2, Ys, defi ned i n Fi g. 9,13. The scal es shown
i n the fi gure establ i sh a nonl i near rel ati on of the vari abl es z to the
parameters Y,
z, = (X,1Y,) Y,, r=l ,2,3, (21)
whi ch can be determi ned, for i nstance, by measurement of the fi gure.
We wi sh now to establ i sh the same rel ati on between the parameters Y
and the vari abl es z as i ndi cated on unijorm scal es. We i ntroduce
transformer l i nkages whi ch present new termi nal s, descri bed by external
parameters Xl , X2, X8. These external parameters are rel ated to the
i nternal parameters by the l i nkage equati ons
x, = ( Xr lY,) Y. I -=1,2,3, (22)
and to the vari abl es xl , z,, x3, by l i near rel ati ons,
z, = Z:o) + K,(x, x$)), (23)
whi ch may be symbol i zed by
z, = (Zrl l x, ). x, . (24)
Our probl em i s to fi nd l i nkages such that the l i nkage oper ator s sati sfy the
rel ati ons
(Glyr) = (Lllxr) (Xrlyr),
r=l ,2,3. (25)
Thi s probl em takes on a compl etel y fami l i ar form when i t i s expr esed i n
terms of homogeneous parameters and vari abl es: o,, I %, oS, correspondi ng
to Yl , Yz, YS; Hl, Hz, H3, correspondi ng to X1, X2, XS; hl, hz, ha, cor -
respondi ng to x,, z*, z~. I n terms of homogeneous parameters and vari a-
bl es a l i near transformati on reduces to the i denti cal transformati on, and
Eq. (25) reduces to
(h,le,) = (Hrlo,). (26)
SEC. 9.6]
DESI GN OF TRANSFORMER LI NKAGES
273
Our probl em i s thus to fi nd l i nkages wi th oper ator s (H,l o,) havi ng the
known form of (h,l 0,), subject to the condi ti on that the i nput parameters
Y, have a gi ven characterthat they are, for i nstance, angul ar parameters
wi th a speci fi ed angul ar range. Thi s i s exactl y the type of probl em di s-
cussed i n Chaps. 4 to 6.
As an exampl e of the use of the same gri d gener ator i n l i nearl y mecha-
ni zi ng another functi onal rel ati on wi th i deal gri d structure, we may con-
si der the probl em of mechani zi ng Eq. (20) wi th l i near scal es i n xl , u, xi .
TABLE9,1.CEAnACTEEI STI CS OF THESCALESOF FI G. 9.13
ala
1.000
0.800
0.640
0.512
O.000
0.512
0.640
0.800
1.000
db
1.000
0.800
0.640
0.512
O.000
0.512
0.640
0.800
1.W3
de
1.000
0.800
0.640
0.512
O.000
0.512
0.640
0.800
l .m
Y,, degr ees
1
h,
I
0,
I
1
22,3
17.6
14.0
11.1
0.0
11,1
13,9
17.4
22.1
0.000
0.100
0,180
0.244
0.500
0.756
0.820
0900
1.000
0.000
0.106
0.187
0.252
0.502
0.752
0.815
0.894
1.000
Y,, degr ees
33.7
27.8
22.6
19.9
0.0
19.9
25.2
32.1
49.9
Y,
0.264
0.198
0.148
0.112
O.cn)o
0.098
0.120
0.147
0.182
O.000
0.100
0.180
0.244
0.500
0.756
0.820
0.900
1,000
h,
O.000
0.100
0.180
0.244
0.500
0.756
0.820
0,900
1.000
e,
0.000
0.071
0.132
0.165
0.403
0.642
0,705
0.787
1.000
O.000
0.148
0.260
0.341
0.592
0.811
0.861
0.921
1.000
.-.
274 BAR-LI NEAGE MULTI PLI ERS [SEC. 9.6
The quanti ti es to be mul ti pl i ed are
21 = jl (zl ),
22 = j2(z*).
}
(27)
The z,- and z,-scal es of Fi g. 913 are then to be i nterpreted as scal es of
ZI and 22; the rel ati on of xl to Yl and X2 to Y2 fol l ows from the obser ved
rel ati ons of ZI to YI and .22to Yz, together wi th Eqs. (27). Except for thk
di fference of detai l i n establ i shi ng the form of the oper ator s (x,1Y,), the
pr ocedur e of the precedi ng paragraph appl i es wi thout change. The com-
pl eted mechani sm may be of exactl y the same type as the mul ti pl i er
1.0r
0.9
0.8
0.7
0.6
h,
0.5 -
(hzI @J
0.4
0.3
/
0.2
0,1
0
0 0.1 02 0.3 0.4
0.5 0.6 0.7 0.8 0.9
1.0
e,
FI G. 9.14.Operator~ (&l @,) for the mul ti pl i er, Fi g. 9.13.
i tsel f, or of even si mpl er form. I t i s, i ndeed, one of the i mportant vi rtues
of bar-l i nkage generators of functi ons of two i ndependent vari abl es that
thei r compl exi ty does not necessari l y i ncrease wi th the compl exi ty of the
anal yti c form of the functi on, as i ts does wi th conventi onal computers.
Thi s fact wi l l appear most cl earl y i n the next chapter.
As an exampl e of the form of the l i neari zi ng operatom, we may con-
si der those needed for the mul ti pl i er i n Fi g. 9 13, i f the rangea of moti on
\
i
I
$
\
I
)
t
I
1
1
i
SEC. 9.6] DESI GN OF TRANSFORMER LI NKAGES 275
are to be those i ndi cated i n that fi gure. Tabl e 9.1 shows the val ue of
each vari abl e at the i ndi cated poi nts of cal i brati on, the correspondi ng
val ue of the associ ated parameter, andthe homogeneous vari abl es h
and o at each poi nt. The oper ator s (h,l d,) are pl otted i n Fi g. 9.14. I t
wi l l be noted that the oper ator (hzl 02) shows an appreci abl e di sconti nui tyy
i n sl ope when hz = 0.5. Thi s i s due to the sti l l i mperfect match between
the four quadrants of the transformed nomogram.
The type of l i nkage used for each transformer wi l l depend both on the
nature of the functi on and on the type of output termi nal desi red. I n
FI G.9.15.Mul ti pl i erpermi tti ngchangei n si gnof onl y one factor.
general i t wi l l be found that the l i nkages di scussed i n Chaps. 4 to 6 offer
al l the fl exi bi l i ty requi red. Exampl es are provi ded by Fi g. 8.16, whi ch
has al ready been expl ai ned, and by Fi g. 9.15. The l atter fi gure shows a
mul ti pl i er desi gned to permi t change i n si gn of m, but not z 1. The
central joi nt of the star gri d gener ator appears a l i ttl e bel ow the center
of the fi gure. The bars have thei r other ends gui ded al ong a hori zontal
strai ght l i ne, whi ch serves as the xl -scal e, and al ong ci rcl es wi th center s
0,0, respecti vel y. Rotati on of the end of one bar about O i s transformed
i nto hori zontal sl i de moti on proporti onal to ZZ by an i deal harmoni c
transformer: rotati on of the end of the thi rd bar about O i s transformed
i nto a paral l el sl i de moti on proporti onal to
three-bar l i nkage and an i deal harmoni c
X3 by a combi nati on of a
transformer. The desi gn
276 BAR-LI NKAGE MULTI PLI ERS [SEC. 96
i nvol ves four di mensi onal constants subject to arbi trary choi ce: A, whi ch
determi nes the scal e of the gri d gener ator and the l ength of the z,-scal e;
B, whi ch determi nes the scal e of the three-bar l i nkage; C and D, whi ch
determi ne the l engths of the x2- and x~-scal es, respecti vel y. I f X,, X,, X3
are di spl acements measured i n the same uni ts, the l i nkage generates the
rel ati on
X3 = ~ 0909 ~, &
~AC
(28)
The fi rst steps i n desi gni ng transformer l i nkages wi l l i nvol ve the use
of tabul ar or graphi cal methods. A graphi cal presentati on of the oper -
ators, as i n Fi g. 9.14, wi l l make i t possi bl e to read off val ues of the
oper ator s for evenl y spaced val ues of d,, when these are requi red. [As a
rul e one shoul d di sregard i rregul ari ti es i n the l i neari zi ng operators, such as
are shown by (hzIL%)i n the exampl e above; the oper ator s shoul d be
repl aced by smoothl y varyi ng approxi mati ons.] Such graphi cal i nter-
pol ati ons are not necessary when the geometr i c method i s to be used i n
desi gni ng a three-bar l i nkage. One then needs to use a geometr i c seri es
of val ues of one of the vari abl es i nvol vedsuch as are provi ded i n
Tabl e 9.1 i n the case of the vari abl es xl , x2, xi . To appl y the geometr i c
method di rectl y to the entri es i n such a tabl e one woul d requi re an overl ay
constructed for the same geometr i c rati o (her e g = 1.25). The overl ay
woul d al so need to be extended i n both di recti ons from the zer o l i ne by
addi ti on of a new seri es of l i nes wi th spaci ng [cf. Eq. (5.92)]
y(f+l) y(t) = _ ~gt,
t=o, l ,.... (29)
The detai l s of thi s extensi on of the geometr i c method wi l l present no
di ffi cul ty to the reader.
To mi ni mi ze the accumul ati on of er r or s i n desi gni ng transformer
l i nkages, the fol l owi ng pr ocedur e i s often useful . When two of the trans-
formers have been desi gned, a graphi cal recal i brati on of the thi rd of the
ori gi nal scal es can be carri ed through befor e i ts l i neari zi ng transformer i s
desi gned. For exampl e, l et new uni form scal es for z, and X2 be con-
structed and accepted as exact. On these scal es l ay down geometr i c
seri es of poi nt~
xy~ = k Cl g,
zp~ = * Czg. (30)
Next, construct correspondi ng cal i brati on poi nts on the ori gi nal non-
uni form z1- and z~scal es, usi ng the known constants of the transformer
l i nkages. Then, usi ng the known constants of the gri d gener ator , con-
struct the poi nts z~4) on the ori gi nal z~scal e that cor r espond to xl = z~~
and xl = z$~, for vari ous choi ces of r and s. I f the gri d gener ator wer e
exactl y i deal , and the transformer l i nkages wer e wi thout structural er r or ,
SEC. 9.7] ANAI I YTI C ADJ USTMENTS OF LI NKAGE CONSTANTS 277
al l the poi nts wi th
r+s=t
(31)
woul d fal l together on two poi nts of the x2-scal e, correspondi ng to
Z$l = * C1C29t. (32)
Actual l y ther e wi l l be some scatteri ng of these poi nts, and i t wi l l be
necessary to choose mean posi ti ons for the new cal i brati on poi nts xf~
on the x,-scal e. Thi s method of constructi ng the x3-scal e i ntroduces a
parti al cor r ecti on for al l desi gn er r or s commi tted up to thi s poi nt; i t
remai ns onl y to desi gn the transformer l i nkage for the xs-termi nal .
9.7. Anal yti c Adjustment of Li nkage Mul ti pl i er Constants.-Fi nal
adjustment of the constants of a mul ti pl i er can be carri ed out by anal yti c
methods si mi l ar to those descri bed i n Chap. 7. From the poi nt of vi ew
of theor y, the present probl em di ffers from the earl i er one pri nci pal l y
i n the necessi ty for control l i ng the structural er r or i n a two-di mensi onal ,
rather than a one-di mensi onal , domai n. From a practi cal poi nt of vi ew,
the l arge number of adjustabl e constants makes a compl ete treatment
of the probl em tedi ous, but assures hi gh accuracy i n the resul t i f enough
car e i s taken. The present secti on wi l l descri be a strai ghtforward
appl i cati on of anal yti c methods to the fi nal adjustment of l i nkage con-
stants; the next secti on wi l l i ndi cate some associ ated or al ternati ve
techni ques used by the author. The r eader shoul d be warned that the
practi cal i mportance of thi s part of the desi gn pr ocedur e, and the l abor
requi red, are out of proporti on to the bri ef di scussi on that can be devoted
to i t i n thi s vol ume.
I f the combi nati on of star gri d gener ator and transformer l i nkages
wer e an exact l i near mul ti pl i er, i t woul d gener ate a rel ati on
RX3 = X,X, (33)
between external parameters X1, X,, X3, at l east wi thh a domai n
Xl m ~ Xl ~ XI M,
X2. s x2 s X2M, X3. s x, s X,.v. (34)
Her e E i s a constant, and the parameters are so defi ned that Xl = O
when z, = O, etc. Because of structural er r or s i n the mechani sm i t wi l l
actual l y gener ate a rel ati on that can be wri tten as
RX3 = X1X2 + 6(X,, X2). (35)
The l ast smal l term i s the structural -error functi on of the mechani sm,
whi ch must be brought wi thi n speci fi ed tol erances by adjustment of the
l i nkage constants.
I n the case of a mul ti pl i er i t i s conveni ent to gauge the er r or of the
mechani sm by the structural -error functi on. The effor t i n the cal cul ati on
can be decreased by computi ng the er r or functi on for spectral va4ues of
278 BARLI NKAGE MULTI PLI ERS
[SEC. 97
Xl and X2 whi ch form geometr i c progressi ons i n the two hal ves of each
i nput scal e:
Xy)+ =
}
-HO 9, ~= . .
x(f)- = _ x~o) . y,
>
2, 1,0,1,2, ..., (36a)
xv+ =
}
x!0)9J ~=... 21012
Xy) = _x~O) . g,,
17 ,1,7 .
(36b)
I f the mul ti pl i er wer e exact, the correspondi ng spectral val ues of X3
woul d form si mi l ar geometr i c progressi ons i n ei ther hal f of the output
scal e:
RX~fE~~ = x\O)x\o)gr+s= ~x~o)gr+n,
(37)
Rx$+o- = _X~O)XjO)#+S= RX$)g?+8.
The spectral val ues of X3 i n ei ther hal f of the scal e woul d then depend
onl y on the val ue of r + s for the correspondi ngg spectral val ues of Xl
and Xt. ~~th the actual mul ti pl i er we have i nstead
RX@*J = X~)*X$J* + 6(Xy)*, Xi )+); (38)
the spectral val ues of X3 do fal l i nto groups accordi ng to the val ue of
r + s, but they are scattered about the i deal val ue for the group, X~O)g+s,
wi th er r or s gi ven by the structural -error functi on J(Xf)*, X~)*).
The spectral val ues of the structural -error functi on are conveni entl y
arranged as a matri x-or, mor e properl y, as an assembl y of four i nfi ni te
matri ces. To si mpl i fy the notati on we shal l wri te
6(X~)*, Xj)*) = E;,*. (39)
Wi th val ue of Xl i ncreasi ng upward, val ues of X, i ncreasi ng to the
ri ght, the matri x takes the fol l owi ng form:
E=
Each r ow or col umn of dots represents an i nfi ni te number of r ows or
col umns of spectral val ues of Xl as r ~ + m, i n the case of rows, or of Xi
as s a t co, i n the case of col umns. I t i s, fortunatel y, not necessary to
gi ve detai l ed consi derati on to these parts of the matri x. I n the graphi cal
process of constructi ng an i deal gri d structure i t was suffi ci ent to con-
si der gri d l i nes wi th smal l posi ti ve and negati ve val ues of r and s, and
SEC. 9.7] ANALYTI C ADJ USTMENT OF LI NKAGE CONSTANTS 279
the ci rcl es of conver gence, r = co, s = co; the same restri cti ons on
r and s can be made i n the present di scussi on, and for the same reasons.
I f the gri d structure i s i deal at Xl = X2 = O--and i t shoul d be kept
so throughout the wor kthe output parameter X3 wi l l be i ndependent
of Xl when X2 = O, and i ndependent of X2 when Xl = O. Al l entri es
i n the central cross of the matri x (40) wi l l then have the same val ue,
E&;___
The el ements of the structural -error matri x are functi ons of al l
structural constants of the mul ti pl i er: the di mensi ons of the star gri d
gener ator and the transformer l i nkages, the ori gi ns from whi ch the
parameters Xl , X,, and X3 are measured, and the constant R of Eq. (38).
Let the i ndependent constants be n i n number: gl , gz, . . . g.. -A
smal l vari ati on Agi of the constant gi wi l l change the matri x el ement
~&*
E~~ by an amount ~ . Agi . I t wi l l then modi fy the matri x E by
addi ng to i t Agi ti mes the i nfi ni te matri x
[ 1
aE$>
G,=
ttgi
(41)
I f smal l vari ati ons are made i n al l the constants gi , the structural -error
matri x wi l l become, to terms of the fi rst or der i n the Agi ,
E+tiE=E+
2
Gag,.
i
(42)
The probl em i s then to determi ne the form of the matri ces Gi and to
choose wrudl (cf. Sec. 9-3) val ues for the Agi whi ch make the fi nal struc-
tural -error matri x E + ~E as smal l as possi bl e-or at l east, to r educe
the er r or s i n certai n regi ons of thi s matri x unti l they meet speci fi ed
tol erances.
The l abor i nvol ved i n sol uti on of thi s probl em i s consi derabl e, and
the wor k must be arranged wi th care. I t i s necessary to consi der onl y
the porti on of the matri x that corresponds to the domai n of acti on of
the mul ti pl i er. The central cross of the matri x must be i ncl uded, but the
cal cul ati ons need not be extended to l arge negati ve val ues of r and s,
parti cul arl y i f vari ati ons of the constants are restri cted to those that
mai ntai n the accuracy of the mul ti pl i cati on for Xl = O and X2 = O.
Anal yti c determi nati on of the deri vati ves (8E~$)/ (dgi) i s advantageousl y
repl aced by l arge-scal e graphi cal constructi ons to determi ne the matri ces
(43)
for smal l changes Agi i n each parameter; these matri ces can be used
dkectl y i n the cal cul ati ons, or conver ted i nto the matri ces Gi , as desi red.
280 BARLI NKAGE MULTI PLI ERS
[SEC. 9;
The matri ces associ ated wi th uhanges i n constants of the termi nal
l i nkages have si mpl e forms. Vari ati on i n a constant g~of the output
transformer l i nkage produces a change i n X3 whi ch depends onl y on X3;
consequentl y the correspondi ng matri x G~ has i denti cal entri es al ong
each l i ne of constant X?. These are l i nes of constant T + s: di agonal s
paral l el to the pri nci pal di agonal s i n the l ower-l eft and upper-ri ght
quadrants (posi ti ve X3), and di agonal s perpendi cul ar to these i n the
upper-l eft and l ower-ri ght quadrants (negati ve X3). Al l entri es i n the
central cross of the matri x wi l l be i denti cal . A matri x Gi associ ated
wi th a constant gi of the Xl -transformer l i nkage has entri es whi ch i n each
r ow are proporti onal to X2; i f g~i s a constant of the Xj-transformer
l i nkage, G~ has entri es whi ch i n each col umn are proporti onal to X,.
I n ei ther case, the entri es i n the central cross wi l l vani sh i f the changes
i n the transformer l i nkages do not affect the accuracy of the mul ti pl i -
cati on for Xl = O and X2 = O.
I n varyi ng the di mensi onal constants of the star gri d gener ator one
shoul d fol l ow the pri nci pl es di scussed i n Sec. 9.5; one shoul d make
changes wi th one degr ee of fr eedom whi ch mai ntai n the i nvari ance of
pr oper l y chosen el ements of the gri d structure, and parti cul arl y the exact
performance of the mul ti pl i er for Xl = O and X2 = O. Such changes
can of cour se be descri bed by a si ngl e parameter g~,whi ch may be ter med
a restri cted parameter. I t woul d be extr emel y di ffi cul t to compute by
anal yti c methods the matri ces Gi associ ated wi th restri cted parameters;
the graphi cal constructi on i s qui te practi cabl e i f the wor k i s done on a
suffi ci entl y l arge scal e. The entri es i n the central cross of these matri ces
are al l zer o.
I t remai ns to make a l i near combi nati on of the matri ces E and Gi l
i =l ,z,...
n, that wi l l have al l el ements as smal l as possi bl e wi thi n
the domai n of i nterest. Fol l owi ng the i deas of Sec. 7.3, one can make n
pr esel ected el ements of the resi dual -error matri x vani sh exactl y. (The
sol uti on wi l l of cour se be spuri ous i f l arge Agi are requi red. ) One can
al so appl y the method of l east squares (Sec. 76). The author pr efer s
to bui l d up the requi red l i near combi nati on i n a successi on of steps, i n
whi ch ther e are formed l i near combi nati ons of the Gi that can be used
to r educe the el ements i n one or another of several regi ons of the er r or
matri x E wi thout i ntroduci ng new er r or s el sewhere.
For si mpl i ci ty,
l et us di vi de the domai n of i nterest i nto two regi ons, A and B. By
l ayi ng out the matri ces Gi , one can see how to make a number of l i near
combi nati ons of these whi ch are smal l i n some part of r egi on A. By
l i near combi nati ons of the resul ti ng matri ces one can bui l d up combi -
nati ons of the Gi whi ch are smal l thr oughout r egi on A but l arge i n
r egi on B; these can be used to r educe the el ements of the er r or matri x i n
regi on B wi thout i ntroduci ng new er r or s i n regi on A. Si mi l arl y, one
SEC. 9.8] GAUGI NG THE ERROR OF A GRI D GENERATOR 281
can bui l d up other l i near combi nati ons of the G~whi ch can be used to
r educe the er r ocs i n regi on A wi thout i ntroduci ng new er r or s i n regi on B.
The probl em has thus been di vi ded i nto two si mpl er and essenti al l y
i ndependent probl ems: that of reduci ng the er r or i n regi on A, and that
of reduci ng the er r or i n r e~on B. I f these probl ems are not si mpl e
enough to be sol ved by i nspecti on, each regi on can agai n be subdi vi ded,
and the process of formi ng new l i near combi nati ons carri ed through as
before. The method requi res some fl exi bi l i ty of approach, and the
desi gner wi l l profi t from exper i ence. The author ti nds i t a compl etel y
sati sfactory method.
9.8. Al ternati ve Method for Gaugi ng the Er r or of a Gri d Generator.
I t i s usual l y sati sfactory to car r y out the fi nal anal yti c adjustment of
di mensi onal constants for the gri d gener ator and transformer l i nkages
separatel y. Thi s greatl y si mpl i fi es the cal cul ati ons by reduci ng the
number of di mensi onal constants that must be vari ed si mul taneousl y.
I n adjusti ng the constants of a gri d gener ator i t i s conveni ent to use
an al ternati ve method for gaugi ng the er r or s i n the al most i deal gri d
structure. We have noted that, i n an i deal gri d structure, systems of
contours of constant z 1, x2, zt meet i n exact tri pl e i ntersecti ons, whereas
i n a noni deal gri d structure these nodal poi nts of the gri d di si ntegrate
i nto l i ttl e tri angl es. To make a star gri d gener ator i deal one woul d have
to make these tri angl es vani sh throughout the domai n of i nterest. The
l i near di mensi ons of these tri angl es have the essenti al characteri sti cs
requi red of a gauge of the er r or i n the gri d structure: they change pr o-
porti onal l y to smal l changes i n the di mensi onal constants of the gri d
gener ator , and vani sh when the er r or vani shes. Si nce they are especi al l y
easy to determi ne by graphi cal constructi on, i t i s ver y conveni ent to use
them di rectl y as gaugi ng quanti ti es i n the fi nal adjustment of l i nkage
constants.
The star gri d gener ator i s i ntended to establ i sh a rel ati on
between vari abl es z,, x,, x,, read on associ ated scal es whi ch are i n general
nonuni form. Wi thout attempti ng to control the uni formi ty of the
scal es, we shal l attempt to make thi s rel ati on as exact as possi bl e by
varyi ng the constants of the gri d gener ator and the cal i brati on of the
scal es. As i n Sec. 9.7, we choose spectral val ues of the vari abl es whi ch
form geometr i c progressi ons:
282 BARLI NKAGE MULTI PLI ERS [SEC. 9.8
Fi rst we may construct the z1- and zti ontours i n the domai n of
i nterest. The i ntersecti ons i n the resul ti ng cuwi l i near gri d can be
l abel ed wi th the i ndex pai r (~~, s A ). Next we construct the z~contours
speci fi ed by Eq. (45). These wi l l not, i n general , pass through the i nter-
secti ons of the ori gi nal gri d, but wi l l ser ve to compl ete smal l tri angl es
wi th one ver tex at each i ntersecti on. Let the si de of the (r ~, s ~ )-
tri angl e opposi te the (r+, s f )-i ntersecti on have l ength L%, taken as
posi ti ve i f the z~contour passes the i ntersecti on on the si de of i ncreasi ng
z3, negati ve i f i t passes on the other si de. The devi ati on of the gri d
structure from the i deal can then be r epr esented by a matri x
L = [L~$], (46)
i denti cal i n structure wi th the matri x of Eq. (40), except that the quan-
ti ti m E~~ are repl aced by the quanti ti es L~$.
A change Agi i n a restri cted parameter 9, of the gri d gener ator wi l l
change the gri d structure; the l engths L~$ will become, to terms of the
fi rst or der ,
L% + 6L% = L;? +
8L$$
Agi .
agi
(47)
The quanti ti es 6L$,~ can be graphi cal l y determi ned for some smal l Agi ;
i t i s then a si mpl e matter to wri te down the matri x
(48)
whi ch corresponds to the matri x Gi of Sec. 97.
I f each of the restri cted parameters of the gri d structure i s changed
by a smal l amount, the matri x of tri angl e di mensi ons wi l l become, to
terms of the fi rst or der ,
(49)
To make the gri d structure i deal one woul d l i ke to choose val ues of the
Agi (necessari l y smal l ) that make the matri x L + 5L vani sh i denti cal l y.
Determi nati on of the Agi can then pr oceed as descri bed i n the precedi ng
secti on, except that one wi l l not i n general attach the same rel ati ve
i mportance to reducti on of the vari ous quanti ti es L~$ as to reducti on
of the cor r espondhg output er r or s E*$; what wei ghti ng factor i s to be
appl i ed wi l l be obvi ous from i nspecti on of the gri d structure.
When the scal es associ ated wi th the gri d gener ator have been deter -
mi ned, i t wi l l remai n to desi gn the transformer l i nkages and to adjust
thei r constants as descri bed i n Chap. 7. Fi nal l y, the performance of
the compl ete mechani sm must be determi ned by exact cal cul ati on. The
pr ocedur e descri bed i n thi s secti on was appl i ed i n desi gni ng the mul ti pl i er
SEC. 9.8] GAUGI NG THE ERROR OF A GRI D GENERATOR 283
i l l ustrated i n Fi g. 9.15. The resi dual er r or s are shown i n Tabl e 9.2,
whi ch gi ves zs z M2 for a seri es of val ues of xl and x2, when the con-
stants A, C, D of Eq. (28) are so chosen that the gener ated rel ati on
shoul d be m = z1z2.
TABLE 9.2.STRUCTUWL ERROR z, ZGU IN MULTIPLIER, FIG. 9.15
x
1.000
0.797
0.637
0.511
0.409
0.000
0.409
0.511
0.637
0.797
1.000
O.000
0,00195
0.00195
0.00195
0.00195
0.00195
0.00195
0.00195
0.00195
0.00195
0.00195
0.00195
0.401
-0,00230
0.00230
0.00390
0.00429
0.00400
0.00195
0.00400
0.00441
0.00419
0,00330
-0,00029
0.511
-0.00060
0.00080
0.00240
0.00289
0.00281
0.00195
0,00320
0.00330
0,00300
0.00179
-0.00179
0.660 0.802 1.000
0.00460 0.00308 0.00790
0.00140 O .00269 0.00080
0,00006 O .00250 O ,00250
0.00070 O .00230 0.00390
0.00070 o .00220 0.00410
0.00195 0,00195 0,00195
0.00099 o .00201 0.00339
0.00150 0.00250 o ,00330
0.00039 0.00310 0.00039
0.00090 o .00330 0.00250
0.00330 0,00320 0.00460
CHAPTER 10
BAR-LI NKAGE FUNCTI ON GENEI UTORS WI TH TWO DEGREES
OF FREEDOM
The precedi ng chapter has descri bed a techni que for the desi gn of
bar-l i nkage mul ti pl i ers-a techni que whi ch i s al so appl i cabl e i n the case
of generators of arbi trary functi ons wi th i deal gri d structure.
The
present chapter wi l l descri be and i l l ustrate a paral l el techni que for the
desi gn of bar l i nkages that gener ate a gi ven functi on wi th noni deal
gri d structure. As i n the precedi ng di scussi on, attenti on wi l l be restri cted
to the use of the star gri d gener ator .
10.1. Summsry of the Desi gn Procedure.I n desi gni ng a star gri d
gener ator for a mul ti pl i er, we began by consi deri ng an i ntersecti on
nomogram for the gi ven functi on,
X3 = X1X2, (1)
i n the for m of an i deal gr i d structure. We woul d have l i ked to carry
out a topol ogi cal transformati on of thi s i nto an equi val ent nomogr am
i n whi ch each fami l y of l i nes woul d be a fami l y of i denti cal ci rcl es. The
transformed nomogr am woul d necessari l y retai n the i deal gr i d structure
of the ori gi nal one; the cor r espondi ng star l i nkage woul d then be an i deal
gr i d gener ator . We saw that such a transformati on can not be found,
but, gui ded by thi s i dea, we succeeded i n l ayi ng out thr ee fami l i es of
i denti cal ci rcl es whi ch had near l y the desi r ed characteri sti cs wi thi n a
restri cted r egi on. Then, usi ng graphi cal methods i n adjusti ng the con-
stants of the cor r espondi ng star l i nkage, we wer e abl e to make the gr i d
structure take on mor e and mor e near l y the desi r ed i deal for m.
The same l i ne of thought can be fol l owed i n desi gni ng a star gr i d
gener ator for an arbi trary functi on. Di ffer ences i n the pr ocedur e ari se
pri nci pal l y fr om the fact that i n i mpr ovi ng the i ni ti al gr i d structure
we cannot concentrate si mpl y on maki ng i t i deal , but must at each step
take account of the speci al functi on that i s to be mechani zed.
To desi gn a star gr i d gener ator for a gi ven functi on of two i ndependent
vari abl es,
z~ = j(q, X2), ,2)
we fi rst r epr esent i t by an i ntersecti on nomogr am consi sti ng of thr ee
fami l i es of l i nes;
2s4
I
SEC. 10.1] SUMMARY OF THE DESI GN PROCEDURE 285
Z1 = Zy), (3a)
Z2 = Xf), (3b)
Zs =
ZOO
(3C)
We desi re to appl y to thi s nomogram a topol ogi cal transformati on that
wi l l transform each fami l y of curves i nto a fami l y of ci rcl es of constant
radi us:
1. Ci rcl es C~J of radi us L I , center s A~), on the l i ne of center s Cl .
2. Ci rcl es C$) of radi us Lz., center s A f), on the l i ne of centers CZ.
3. Ci rcl es C!/) of radi us Ls, center s A$?, on the l i ne of center s Ca.
I f we can fi nd such a transformed nomogram i t wi U be equi val ent to the
ori gi nal one, and from i t we shal l be abl e to determi ne the constants of
the desi red star gri d gener ator : l i nk l engths Ll, L~, L3; gui di ng curves
Cl , Cz, Ca; scal es wi th the x~)-, x(SJ- 2 , x~)-cahbrati ons at poi nts A ~~,A ~),
A!), r especti ve y.
I n seeki ng such a transformati on we can be gui ded by the speci al
characteri sti cs, and especi al l y the si ngul ari ti es, of the ori gi nal nomo-
gram. Under a topol ogi cal transformati on, i ntersecti ons transform i nto
i ntersecti ons and poi nts of tangent y i nto poi nts of tangency; these
features of the ori gi nal nomogram must then appear i n the transformed
nomogram. I f al l l i nes of constant xl i ntersect at one poi nt of the ori gi nal
nomogram, then al l ci rcl es C~) of the transformed nomogram must
al so i ntersect i n a si mi l ar manner. I f a l i ne ZI = z~) i s tangent to the
l i ne X2 = z~) wher e X3 = x$), then the ci rcl es C~) and C!) must be tangent
to each other at a poi nt on the ci rcl e C!).
As a fi rst step, we l ay down tentati ve transforms of two of the ori gi nal
fami l i es of l i nes, l et us say the xl - and zz-contours. These wi l l be
fami l i es of ci rcl es that have the i nvari ant characteri sti cs of the ori gi nal
xl - and xz-contours, at l east wi thi n the domai n of mechani zati on; to each
wi l l be assi gned a tentati ve val ue of xl or x2, wi th due r egar d for al l
i nvari ant characteri sti cs of the ori gi nal assi gnment. The tentati ve
choi ce of these transforms i s suffi ci ent to determi ne the form of a tentati ve
topol ogi cal transformati on and a correspondi ng new form of the thi rd
fami l y of curves; to determi ne these curves we have onl y to pl ot, i n the
curvi l i near coordi nate system formed by the xl - and zz-ci rcl es, the curves
X8 = & as gi ven by Eq. (2). I f thi s tentati ve transformati on shoul d
be the desi red one, these x~-contours wi l l then be ci rcl es wi th the same
radi us. Of course, we cannot expect so fortunate a resul t. Usual l y,
however , we can see how to rearrange or renumber the z1- and xz-ci rcl es
so sa to make the xs-contours roughl y ci rcul ar. We thus obtai n a
transformati on of the ori gi nal nomogram i n whi ch two of the fami l i es
of contours have the desi red ci rcul ar form, and the thi rd fami l y has
approxi matel y the desi red character. On repl aci ng the thi rd fami l y of
curves by a system of approxi mati ng ci rcl es wi th the same radi us, we
286 BAR-LI NKAGE FUNCTI ON GENERATORS
[SEC. 10.2
can now conver t i t i nto the gri d structure of a star l i nkage whi ch gener -
ates, at l east approxi matel y, the gi ven functi on.
I t remai ns to modi fy thi s star l i nkage i n such a way as to i ncrease the
preci si on wi th whi ch i t generates the gi ven functi on, and at the same
ti me to bri ng i ts scal es i nto some conveni ent form, preferabl y ci rcul ar.
Thi s can usual l y be accompl i shed by a method of successi ve approxi -
matei ons. Accepti ng the x,- and w-ci rcl es, C~) and C$, as determi ned
above, we can r epl ot the contourszl = z~) as defi ned by Eq. (2). These
wi l l be nearl y ci rcul ar; they can be approxi mated by a new system of
ci rcl es C:), whi ch can often be so chosen that the l i ne of center s has a
si mpl e and conveni ent form. The resul ti ng star-l i nkage nomogram i s
usual l y mor e accurate than the fi rst approxi mate nomogram, and mor e
conveni entl y mechani zed. Next, accepti ng the CT)- and C~l -ci rcl es
thus obtai ned, we can r epl ot the contours X2 = z~) and repl ace them by a
new set of ci rcl es C~l , and so on. Ther e i s no guarantee that thi s method
wi l l conver ge on a sati sfactory sol uti on of the probl em; i f i t does not,
the process must be begun agai n wi th a drasti cal l y di fferent i ni ti al
structure.
An al ternati ve method for i mprovi ng the gri d gener ator wi l l be
i l l ustrated i n Sec. 10.3.
When a graphi cal method i s no l onger adequate for the further
i mprovement of the gri d structure, anal yti cal methods can be brought
i nto pl ay. These, agai n, are essenti al l y the same as those used i n the
desi gn of mul ti pl i ers. The gri d gener ator , consi dered ei ther separatel y
or i n combi nati on wi th transformer l i nkages, generates a rel ati on between
the xl -, x2-, x~-scal e readi ngs whi ch may be wri tten as
x$ = j(zl , U) + 6(21, 24. (4)
The structural er r or 6(z,, z*) i s a functi on of al l di mensi ons of the mecha-
ni sm, as wel l as of ZI and z*. I t can be eval uated for spectral val ues
~~) and ~~) of these l atter vari abl es and brought wi thi n speci fi ed
tol erances by the methods descri bed i n Sec. 9.9. Ther e i s, however ,
no advantage i n maki ng a speci al choi ce of the spectra z~) and Z$J,
except as thi s may be i ndi cated by si ngul ari ti es or i nvari ants of the
gi ven functi on.
10.2. Exampl e: Fi rst Approxi mate Mechani zati on of the Bal l i sti c
Functi on i n Vacuum.As an exampl e, we shal l desi gn a star gri d gener -
ator for the bal l i sti c functi on i n vacuum. The el evati on angl e Z9 of a
gun that i s to send a projecti l e through a poi nt at ground range z],
rel ati ve al ti tude 22, may be obtai ned by sol vi ng
9,
21 si n zs cos zs 22 COS223 = 21,
v
(5)
SEC. 102] THE BALLI STI C FUNCTI ON I N VACUUM 287
wher e v i s the i ni ti al vel oci ty of the shel l and g i s the accel erati on of
gravi ty. We shal l take z, and x, as i nput vari abl es i n the l i nkage and
gener ate the requi red el evati on of the gun, x3, as the output vari abl e.
For si mpl i ci ty i n the cal cul ati on we shal l take v = 500 m/see and
g = 10 m/sec.2 The parabol i c trajectori es of the shel l s then have the
forms shown i n Fi g. 10.1.
I t wi l l be noted that for each target posi ti on wi thi n the boundi ng
envel ope ther e are two possi bl e val ues of X3. The l arger val ue of X3
corresponds to a hi gh tr ajector y, whi ch becomes tangent to the envel ope
of the trajectori es befor e the target i s reached. The smal l er val ue of z3
gi ves a l ower tr ajector y, wi th shor ter ti me of fl i ght to the target; the
shel l reaches the target befor e i t reaches the envel ope of trajectori es.
13,033
12,0Q0
11,000
10,OOO
:
~9000
E8000
: 7W0
~ 6000
% 5000
= 4000
3000
2000
I wo
o
5000 10,OOO 15,000 20,000 25,000
Horizontal ranget, i nmaters
FIG. 10.1.Trajectories of aheUain vacuum.
The dashed ellipse is the locus of maximum
altitudes. These altitudes are indicated for some of the trajectories.
We shal l be i nterested onl y i n the smal l er of the two possi bl e val ues of w
and, correspondi ngl y, onl y i n the porti ons of the trajectori es between the
ori gi n and the envel ope. We shal l al so excl ude from consi derati on the
regi on of ver y smal l sl ant range, i n the nei ghborhood of the si ngul ar
poi nt xi = 22 = O, wher e z~ i s not defi ned.
I t i s to be emphasi zed that the present exampl e i s i ntended onl y to
i l l ustrate a general techni que and does not necessari l y consti tute the
best sol uti on of the stated probl em. Equati on (5), whi ch i s of rel ati vel y
si mpl e anal yti c form, can be mechani zed by a net of standard com-
puti ng mechani sms. Such a devi ce wi U be l ess desi rabl e mechani cal l y
than a bar-l i nkage functi on gener ator , but i t wi Ll be much easi er to
desi gn and fr ee from structural er r or s. These advantages of a computi ng
net are l ost i n the case of read bal l i sti c functi ons, whi ch offer no gr eater
di ffi cul ti es i n bar-l i nkage desi gn than does the present probl em. I n such
288 BAR-LI NKAGE FUNCTI ON GENERA TORS [SEC. 10.2
cases however , the usual practi ce i s to separate the functi on to be gener -
ated i nto two parts, one of si mpl e anal yti cal form, to be gener ated by a
computi ng net, and the other a resi due, to be gener ated by a bar l i nkage.
The resul ti ng probl em i n bar-l i nkage desi gn wi l l then usual l y be l ess
di ffi cul t than that her e consi dered, and the compl ete sol uti on can be
gi ven much gr eater accuracy. The present exampl e, neverthel ess,
offers a number of i nteresti ng poi nts for di scussi on.
Si nce the trajectori es i n Fi g. 10.1 are contours of constant z,, the
fi gure i s actual l y an i ntersecti on nomogram that coul d ser ve for the
sol uti on of Eq. (5). We shal l take i t as the starti ng poi nt of the desi gn
process and attempt to fi nd a topol ogi cal transformati on that wi l l
transform the parabol as i nto a fami l y of i denti cal ci rcl es, the hori zontal
l i nes i nto a second fami l y of i denti cal ci rcl es, and the verti cal l i nes i nto a
thkd.
Determination of the za-scale. The most d@i cul t stage of the wor k
i s al ways the begi nni ng; ever y possi bl e cl ue must be used as a gui de.
We obser ve fi rst that al l the parabol as i ntersect i n a common poi nt,
Z1=Z2 =0. I f the desi red transformati on exi sts, i t must car r y these
parabol as i nto a fami l y of ci rcl es, of radi us Ls, whi ch al so i ntersect i n a
common poi nt, the or i gi n of the transformed nomogr am. The centers
of these ci rcl es must then l i e on a ci rcl e of radi us L ~about the ori gi n;
thi s i s the z~-scal e, thus deter mi ned as to for m and posi ti on, but havi ng
no known cal i brati on poi nts. One cal i brati on poi nt can of cour se be
chosen at wi l l , wi thout l oss of general i ty. We ther efor e begi n construc-
ti on of the transformed nomogr am, Fi g. 10.2, by dr awi ng the z~-scal e
and the ci rcl e X3 = O wi th arbi trari l y chosen radi us L3; the cal i brati on
poi nt Z3 = O on the zs-scal e has been chosen to l i e di rectl y bel ow the
ori gi n.
Next we obser ve i n Fi g. 10.1 that the contour Z2 = O i s tangent to
the trajectory 23 = O at the or i gi n and l i es above i t ever ywher e el se.
The transformed ci rcl e ZZ = O must then be tangent to the transformed
ci rcl e 23 = O at the ori gi n; i n Fi g. 10.2 i ts center must l i e di rectl y above
or di rectl y bel ow the ori gi n. Compari son wi th Fi g. 101 suggests that i ts
center shoul d l i e bel ow the ori gi n, and that i ts radi us, LZ, shoul d be
gr eater than Ls. I t has been so drawn i n Fi g. 102. The choi ce of L,,
whi ch has been made arbi trari l y, can be changed at wi l l i f the desi gn
pr ocedur e shoul d fai l to progress sati sfactori l y. Thi s choi ce of the
ci rcl e Z2 = O al so fi xes the poi nt Z2 = O on the x~scal e at a di stance Lz
bel ow the ori gi n.
Gui ded by the di stri buti on of i ntersecti ons of the trajectori es 23 = x:)
wi th the hori zontal l i ne Zz = O, we are now i n a posi ti on to make a
tentati ve cal i brati on of the zg-scal e. I t i s a fami l i ar fact that projecti l es
shot i n vacuum at el evati on angl es Xa and 90 X8 wi l l have the same
I
SEC. 10.2] THE BALLI STI C FUNCTI ON I N VACUUM 289
hori zontal range; for i nstance, the parabol as x, = 40 and Z3 = 50 of
Fi g. 101 i ntersect on the l i ne Q = O. The tr ajector y X3 = 45 gi ves the
greatest hori zontal range. The transformed ci rcl e m = 45 i n Fi g. 10.2
must then i ntersect the ci rcl e X2 = O at the greatest possi bl e di stance
from the ori gi n-a di stance equal to the di ameter of the x,-ci rcl es.
Thi s i ntersecti on poi nt, S i n Fi g. 102, can be determi ned by use of a
com~ass. The mi d~oi nt of the di ameter OS l i es on the ZJ scal e at the
cal i brati on poi nt X3 = 45; cal i brati on poi nts for x, < 45 wi l l l i e bel ow
thk poi nt, ~al i brati on poi nts for x, >45, above i t.
FI Q. 10.2.Construction of tentative scales in mechanization of the ballistic function in
vacuum.
We shal l now choose other cal i brati on poi nts on the x,-scal e, such
that the ci rcl es m = z!) and 23 = 90 Zf) i ntersect the ci rcl e Z2 = O
at the same poi nt. Speci fi cal l y, havi ng chosen cal i brati on poi nts for
X3 = 10, 20, 30, 40, whi ch i nterpol ate between the known poi nts
x$ = 0 and 23 = 45, we shal f construct the correspondi ng ci rcl es of the
transformed nomogram. Through the i ntersecti on of each such ci rcl e
wi th the l i ne zz = O, and the ori gi n, we shal l construct a ci rcl e of radi us
La. These ci rcl es cor r espond to X3 = 80, 70, 60, 50, respecti vel y;
thei r center s are the desi red cal i brati on poi nts on the x,-scal e. I n pr o-
ceedi ng thus, we are gui ded by properti es of descendi ng porti ons of the
290 BAR-LI NKAGE FUNCTI ON GENERATORS [SEC. 102
parabol i c trajectori es wi thza > 45, whi ch have no i mmedi ate rel evance
tothechosenmobl em. The rel ati ons that wehavee stabl i shedarethus
useful gui des i n the prel i mi nary cal i brati on of the z~-scal e, but they
need not be mai ntai ned throughout l ater devel opments.
Fi rst of al l , we note that the ci rcl ezs = 90, l i ke the associ ated ci rcl e
X3 = 0, must be tangent totheci rcl exa = O; i ts centermost l i e di rectl y
above the ori gi n. The za-scal e must then extend through an arc of 180.
Si nce the scal e from Z8 = 0 to X3 = 45 cover s l ess than 90 of thi s arc,
the spaci ng between consecuti ve zs-cal i brati ons must, on the average:
i ncrease wi th X3; we can reasonabl y assume that thi s i ncrease conti nues
smoothl y thr oughout the l ength of the scal e. We shal l ther efor e choose
cal i brati on poi nts w = 10, 20, 30, 40, whi ch have gradual l y i ncreasi ng
separati ons and whi ch l ead to the determi nati on of poi nts X3 = 50, 60,
70, 80, wi th separati ons that fal l into the same smoothl y i ncreasi ng
sequence. Such poi nts are easi l y found; they are shown i n Fi g. 10.2,
together wi th the associ ated ci rcl es of radi us L3.
Determination of the x,-scal e.Thus far we have establ i shed onl y the
poi nt X2 = O on the zz-scal e. As our pri nci pal cl ue i n the further con-
structi on of thi s scal e we have the poi nts of tangency of the ZQ- and
x3-contours. I n Fi g. 10.1 the hori zontal l i ne m = 375 i s tangent to the
parabol a ~a = 10 at i ts ver tex; the transformed ci rcl e X2 = 375 i n
Fi g. 10.2 shoul d si mi l arl y be tangent to the transformed ci rcl e X3 = 10.
The ci rcl e z, = 10 has al ready been determi ned, but we know onl y the
radi us of the ci rcl e X2 = 375; its center the poi nt X2 = 375 on the xz-
scal e-may l i e anywhere on a circle of radi us Lz Ls, wi th i ts center at
the poi nt za = 10. An arc of thi s ci rcl e near the poi nt X2 = O i s shown
i n Fi g. 10.2. Si mi l arl y, the scal e poi nts X2 = 1460; 3125, 5160, 7335,
9370, 11040, 12130, 12500, must l i e on ci rcl es of radi us L2 LS about
centers on the za-scal e, at the poi nts X3 = 20, 30, 40, 50, 60, TO,
80, 90, respecti vel y. Arcs of these ci rcl es al so appear i n Fi g. 10.2.
Ther e i s another cl ue to the nature of the zz-scal e, but i t i s rel ati vel y
unrel i abl e. I t i s wel l known that the poi nts of maxi mum x, on the
trajectori es l i e on an el l i pse (the dashed cur ve of Fi g. 10.1). These
maxi ma occur at fi rst cl ose to the or i gi n O; thei r z l -coordi nates i ncrease
wi th x3, and then decr ease to O as ~a goes to 90. Now, the transformed
nomogr am under constructi on bears some si mi l ari ty to the ori gi nal one,
Fi g. 10.1, i n whi ch X2 i ncreases fr om bottom to top, m fr om l eft to ri gLt.
I n vi ew of thi s we may expect the poi nts of tangency between the trans-
for med X2- and z~-ci rcl es to move at fi rst toward the ri ght, as ~a i ncreases,
and then as far as possi bl e to the l eft. I f thi s i s to be the c~e, the
m-scal e must then ri se to the ri ght of the ori gi n, wi th i ts upper end 17,
cor r espondi ng to X2 = 12500, at about the same l evel as the scal e poi nt
Z3 = 90. F@re 10.2 shows such a choi ce of E.
SEC. 10.2] THE BALLI STI C FUNCTI ON I N VACUUM 291
A prel i mi nary choi ce of the x~scal e can now be made. For mechan-
i cal reasons i t has been constructed i n Fi g. 10.2 as a ci rcul ar arca
tentati ve choi ce that can be modi fi ed at any ti me. The cal i brati ons on
thi s scal e are deter mi ned by i ts i ntersecti ons wi th the arcs al ready con-
structed. For l ater use, cal i brati on poi nts have been i nterpol ated for
evenl y spaced val ues of X2: 1000, 2000, 3000, . . . 12000. These are
easi l y and accuratel y deter mi ned by pl otti ng on cross-secti on paper a
smooth cur ve of di stance al ong the wscal e agai nst the val ue of X.Z,usi ng
the known cal i brati on poi nts, and r eadi ng off the di stances cor r espondi ng
to the chosen val ues of x2.
Thi s compl etes the determi nati on of our tentati ve topol ogi cal trans-
formati on. Contours of constant X2 can be dr awn i n at wi l l , but are
omi tted fr om Fi g. 10.2.
Determination of the xl -.scal e.-We have next to construct contours of
constant ZI i n the transformed nomogram. Thi s i s conveni entl y done
usi ng computed val ues of X2for a seri es of val ues of xl and x3, as gi ven i n
Tabl e 10.1. Onl y val ues wi thi n our restri cted domai n of i nterest are
tabul ated. -
TABLE 1O.1.VALUES OF z, COMPUTED BY USE OF ECL (5)
\
21
23
80
70
60
50
40
30
20
10
2000 4000
8650 12050
4830 8250
3150 5650
2150 4000
1550 2800
1150 1900
650 1100
280 375
6000 8000 1000O
10300
7500
5400
3800
2500
1350
300
11030 10350
8700 9350
6450 7100
4550 5000
2900 3100
1450 1350
75
12000 14000
I
9250
7330 7200
5150 5050
3100 2830
1100 600
16000 18000
6700 5750
4700 4050
2400 1730
0
20000
4500
3150
850
To construct the cur ve Z1 = 2000 we r efer to the fi rst col umn of
Tabl e 10.1. We construct arcs of radi us L2 wi th center s at the poi nts
X2 = 8650, 4830, 3150, 2150, 1550, 1150, 650, 280, i ntersecti ng, r espec-
ti vel y, the ci rcl es X3 = 80,70,60, 50,40,30,20, 10. The poi nts of
i ntersecti on l i e on the cur ve Z1 = 2000. I n the same way we can deter -
mi ne poi nts on the other contours of constant Q, as shown i n Fi g. 102.
Because of i rregul ari ti es i n the XT and x~scal es, these poi nts do not l i e on
smooth curves, but on rather i rregul ar ones; they have been connected by
strai ght l i nes i n Fi g. 10.2, merel y to bri ng out thei r rel ati ons.
I t woul d have been ver y grati fyi ng i f the contours of constant xl
had turned out to be ci rcl es of constant radi us L1. The actual resul t i s
not bad, for a fi rst tri al , si nce the curves do resembl e arcs of ci rcl es. The
radi i of these ci rcl es are not exactl y equal , but i t i s not di ffi cul t to sel ect an
average radi us LI .
292 BAR-LI NKAGE FUNCTI ON GENERATORS
[SEC. 103
Wehavenow toconstruct thexl -scal e. About eachof theestabl i shed
poi nts of the contour ZI = 2000 we construct arcs of radi us L1. These
i ntersect near the upper margi n of Fi g. 10.2 and thus determi ne roughl y
the posi ti on of the poi nt xl = 2000 on the xl -scal e. Si mi l ar construc-
ti onsare shown fortheestabl i shed poi nts of theother xl -contours. The
i ntersecti ons are rather di ffuse; the form of thezl -scal e i s not determi ned
ver y preci sel y. Fortunatel y, themost di ffuse i ntersecti ons occur for the
l east cri ti cal part of thezl -scal ej the center . For these val ues ofxl , thezz-
and zs-contours are ver y nearl y tangent to each other , and the computed
val ue of ZS i s ver y i nsensi ti ve to the val ue of xl . For i nstance, l et us
consi der a case i n whi ch the central pi vot of the star l i nkage i s at a
poi nt of tangency of the x2- and xs-contours. As l ong as X2i s fi xed, any
di spl acement of the z,-i nputwhether al ong or perpendi cul ar to the
scal e-can move the star pi vot onl y al ong the *z-ci rcl e and thus pr oduce
at most a second-or der change i n x3. We have, ther efor e, to attach l i ttl e
i mportance to the cl i ffuseness of the i ntersecti on i n the central part of the
zl -scal e: we can adjust the posi ti on of that part of the scal e and i ts
cal i brati on poi nts wi t h rel ati ve freedom. The reason for thi s i s al so
apparent on i nspecti on of Fi g. 10.1; for val ues of xl near 12000, a change
i n xl wi th constant X2 carri es one ver y nearl y al ong a tr ajector y wi th
constant x3. (The steepl y descendi ng trajectori es we have al ready
excl uded from consi derati on. )
I t i s evi dent that the z,-scal e wi l l be nearl y ci rcul ar; i n Fi g. 102 i t
has been gi ven an accuratel y ci rcul ar form. I t then becomes cl ear that
the cal i brati ons i n Z1 wi l l be al most equal l y spaced. Thi s fact suggests
that an exactl y even scal e i n xl shoul d be l ai d down-a pr ocedur e that has
been fol l owed i n Fi g. 10.2. We have thus gi ven to the x,-scal e a par-
ti cul arl y si mpl e form, whi ch we can hope to mai ntai n through l ater stages
of the devel opment.
10.3. Exampl e: I mprovi ng the Mechani zati on of the Bal l i sti c Func-
ti on i n Vacuum.I n fol l owi ng the method descri bed i n Sec. 10.1 for the
i mprovement of our prel i mi nary mechani zati on of the bal l i sti c functi on,
we can accept the xl - and xi -scal es al ready defi ned and reconstruct the
z2-scal e. I t shoul d now be suffi ci entl y cl ear how thi s wor k woul d pr oceed.
We shal l ther efor e appl y another useful techni que to thi s probl em.
Let us accept the ver y conveni ent xl -scal e of Fi g. 10.2 and the estab-
l i shed val ues of LI , Lz, and La. I nstead of prescri bi ng the xz-scal e
di rectl y, we shal l keep unchanged the contour X3 = 40, and shal l requi re
that the new l i nkage gi ve an exact sol uti on of the probl em whenever
X8 = 40. TMs requi rement wi l l compl etel y defi ne the xz-scal e for al l
val ues of Zz l ess than 5160. For i nstance, we can l ocate the scal e poi nt
m = 1000 i n the fol l owi ng manner. From the ori gi nal nomogram we
read that when ZS = 40 and X2 = 1000, Z1maybe 1300 or 23300. About
II
SIDc.10-3] I MPROVI NG THE MECHANI ZATI ON 293
the poi nts zl = 1300 and xl = 23300 on the xl -scal e we draw ci rcti l ar
arcs of radi us LI , i ntersecti ng the contour X8 = 40 at poi nts A and B
(Fig. 10.3). These poi nts must both l i e on the contour m = 1000; we
ther efor e construct about them arcs of radi us Lt, and l ocate the scal e
poi nt X2 = 1000 at thei r i ntersecti on. The scal e poi nts X2 = 2000, 3000,
4000, 5000, can be determi ned si mi l arl y by the use of the data i n Col umns
1 and 2 of Tabl e 10.2.
TABLE 1O.2.VALUES OF z, COMPUTEDBY u. (5)
x
1000
2000
3000
4000
5000
6000
7000
8000
9000
1000O
11000
12000
40
1300 23320
2700 21900
4350 20250
6450 18150
1000O14500
50
2850
4000
5300
7000 17600
9600 14950
60
2600
3400
4350
540U
6700 14950
8650 12950
70
1600
2100
2600
3160
3850
4600
5600 10450
7600 8400
80
800
1000
1240
1500
1800
2100
2480
2950
3850
The scal e poi nts thus establ i shed l i e on a ci rcul ar arc wi th center Qz
(Fi g. 10.3) and are equi di stant to wi thi n the accuracy of the constrw?t;on.
We shal l ther efor e compl ete the zz-scal e by extendi ng i t as a ci rcul ar arc,
wi th equi di stant scal e di vi si ons up to ZZ = 124)00.
I t remai ns to reconstruct the contours of constant Za and the
x~scal e. Poi nts on the contours xa = 10, 20, 30 are conveni entl y
l ocated by the use of the data i n the l ast thr ee r ows of Tabl e 10.1; they
l i e at the i ntersecti ons of arcs of radi us LI and LZ about correspondi ng
poi nts on the zI - and zz-scal es, respecti vel y. Poi nts on the contours
28 = 50, 00, 70, 80 can be l ocated si mi l arl y by the use of data gi ven i n
Tabl e 10.2. -Ml these poi nts are shown i n Fi g. 103 as smal l ci rcl es.
Through them we can pass ci rcl es of radi us La, wi th er r or s that are
appreci abl e onl y at the outer extremi ty of a few of the arcs. The center s
of these ci rcl es l i e on a nonuni form xa-scal e that i s onl y roughl y ci rcul ar.
The fact that the Z1- and zz-scal es are even and ci rcul ar makes thi s
sol uti on of the probl em attracti ve, even though the z~scal e i s of the most
general type. No transformer l i nkages wi l l be requi red for the i nputs; i t
remai ns to desi gn a si ngl e bar l i nkage that wi l l both gui de the za-poi nt of
the star l i nkage over the present nonci rcul ar zs-scal e and provi de an out-
put moti on l i near i n Za. How thi s can be done wi l l be shown i n Sec. 10.4.
Fi gure 10.4 shows a schemati c l ayout of the l i nkage i n i ts present state,
I
294 BARLI NKAGE FUNCTI ON GENERATORS
[SEC. 103
0Q,
wi th a cur ved sl i de for the output termi nal . The el evati on scal e i s
restri cted tothe range l OO < X8 < 80, whi ch al one woul d be i mportant i f
themechani zed bal l i sti cs had practi cal si gni fi cance. Thi s functi on gen-
eratorhas averysmal l er r or for sl ant ranges gr eater than 2000m, except
forafew poi nts cl o.seto theenvel ope of trajectori es. (Thesol uti on near
ZI = X2 = O i s poor because no attempt was made to for ce the contours
I
I
o.
I
(~
o
.
~ti
7U
+..
%.
A
,,W
w
/0
I
)
o
4U
B
20 3f
16,CQ2
\-
.,. ,ly Q,
/
y17,003
)
12,GVJ
Il,mn
IOSXO
7QXI
m
.,.,2 lIXO
/
/
19,ml
, 20,ml
/
21,0YJ
22,0ca
23,CC0
FIG. 10.3.Construction of improved scales for mechanization of the ballistic function in
vacuum.
of constant ~a to i ntersect at a si ngl e poi nt.) Thi s desi gn coul d be
i mproved somewhat by graphi cal methods, wi thout sacri fi ci ng the si mpl e
forms of the z,- and x,-scal es; sti l l further i mprovement coul d be obtai ned
by appl yi ng anal yti c methods. I n practi ce, however , maxi mum accu-
racy coul d be obtai ned by reformul ati on of the probl em, i ntroducti on of
new vari abl es, and use of the bar l i nkage to mechani ze a functi on of mor e
sui tabl e character.
SEC. 10.4~ CURVE TRACI NG AND TRANSFORMER LI NKAGES
295
10.4. Curve Tracing and Transformer Linkages for Nonci rcul ar
Scal es.-Practi cal appl i cati on of a gri d gener ator wi th a ri onuni forml y
cur ved scal e requi res sol uti on of two probl ems:
1.
2.
Desi gn of a constrai nt for the gri d-generator termi nal that i s mor e
sati sfactory than a cur ved sl i de.
Desi gn of a transformer l i nkage to provi de a sati sfactory external
termi nal , usual l y wi th an even scal e.
Horizontal rangezl=~
0
0
0
0
05000
0,
0
010,000
0
0
0
0
015,000
0
-J
0
0
020,000
0
0
0
0
Fm. 10.4.Schematic layout of ballistic computer.
These probl ems can often be sol ved si mul taneousl y by a devi ce such as
that sketched i n Fi g. 1005, i n whi ch the termi nal T of the gri d gener ator
i s pi voted to and gui ded by a ri gi d extensi on QTR of the central l i nk of a
three-bar l i nkage PQRS. One of the cranks of thi s l i nkage may i tsel f
ser ve as the termi nal of the transfomner, as shown i n Fi g. 105, or a
harmoni c transformer may be added to gi ve a sl i de termi nal , as i n the
mul ti pl i er of Fi g. 8.16. We shal l consi der her e the fi rst and si mpl er
al ternati ve; the pr ocedur e i s easi l y extended to the second case by use of
the i deas presented i n Chap. 8.
296 BAR-LI NKAGE FUNCTI ON GENERA TOE% [SEC. 10.4
Let us consider that the rigid triangle QTR of Fi g. 10.5 consi sts of
ri gi d bars. The devi ce may then be di vi ded i nto two parts:
1.
2.
A transformer l i nkage, consi sti ng of the l i nk TR and the crank RJ .$.
When the joi nt T i s gui ded al ong the scal e AB, these el ements ser ve
to transform readi ngs on the uneven scal e AB i nto i denti cal read-
i ngs on the even ci rcul ar scal e CD.
A constrai nt l i nkage, consi sti ng of the crank PQ and the l i nks QR
and QT. Thi s, together wi th the el ements of the transformer
l i nkage, gui des the joi nt T al ong the scal e .4B.
I n desi gni ng such a l i nkage these parts are consi dered separatel y, and i n
the or der l i sted.
P
0.0
A
Q
FIG. 10,5.Linkage transformer from an uneven noncircular scale AB to an even circular
scale CD. The joint T will follow AB without further constraint.
The Transformer Li nkage.The transformer l i nkage can be desi gned
by appl i cati on of the geometr i c method for three-bar-l i nkage desi gn, as
descri bed i n Chap. 5.
To understand how thi s can be done wi thout any si gni fi cant change i n
the pr ocedur e, we may consi der the three-bar l i nkage from a poi nt of
vi ew not previ ousl y emphasi zed. A three-bar l i nkage can be used to
gener ate a rel ati on between an i nput vari abl e xl , whi ch can be read on a
uni form scal e associ ated wi th the i nput crank, and an output vari abl e zZ,
whi ch can be read on a uni form scal e associ ated wi th the output crank.
Now, a nonl i near mechani zati on of the same rel ati on can be obtai ned by
use of the output crank al one; the uni form xn-scal e may be suppl emented
by a nonuni form x,-scal e, and the output crank used si mpl y as a poi nter
to i ndi cate correspondi ng val ues of xl and x2. The functi on of the i nput
crank and connecti ng l i nk i n the three-bar l i nkage i s, then, that of
transformi ng thi s ci rcul ar but nonuni form scal e of ZI i nto a uni form
ci rcul ar i nput scal e. I t wi l l be obser ved that the geometr i c method of
three-bar-l i nkage desi gn can be understood from thi s poi nt of vi ew; i t wi l l
be noted al so that the ci rcul ar form of the output scal e pl ays no essenti al
8EC, 10.4] CURVE TRACI NG AND TRANSFORMER LI NKAGES 297
r ol e i n the pr ocedur e. Thi s method can be appl i ed, then, whenever i t i s
desi red to car r y out a transformati on between a ci rcul ar scal e and another
scal e of arbi trary form. (I nterchange i n the rol es of termi nal s as i nput
and output does not affect the pr ocedur e.) For exampl e, i t can be used
i n the desi gn of hi ghl y noni deal harmoni c transformers, i n whi ch one scal e
i s l i near, as wel l as i n the present case, wher e one scal e may have arbi trary
form.
To desi gn the transformer l i nkage one can then pr oceed as fol l ows:
1.
2.
3.
4.
5.
Choose a spectrum of val ues of the vari abl e (for exampl e, x) wi th
di fferences that change i n geometr i c progressi on, wi th rati o g:
(6)
Construct the gi ven scal e AB, and on i t l ay down the poi nts cor -
respondi ng to these spectral val ues of z(). These poi nts cor -
respond to the poi nt P(o), P(l), . . . P(m), of Fi g. 5.21.
About the poi nts P() construct ci rcl es C(r) wi th radi us BZ = ~.
Thi s compl etes a chart correspondi ng to Fi g. 5.21.
Over thi s chart move the characteri sti c overl ay of the geometr i c
method, Fi g. 5.23, constructed for the chosen val ue of g. I f i t i s
possi bl e to fi nd a posi ti on of thi s chart (face up or face down) i n
whi ch successi ve ci rcl es C(o), C(l ), C(z), . . . pass through the i nter-
secti ons of an overl ay ci rcl e wi th successi ve radi al l i nes, the
requi red constants of the transformer l i nkage can be read off at
once. The posi ti on of the pi vot S i s the center of the overl ay. The
ci rcul ar scal e CD must coi nci de wi th the overl ay ci rcl e on whi ch the
fi t was found; the cal i brati on poi nts on thi s scal e correspondi ng to
the chosen spectrum z(r) l i e at the i ntersecti on wi th thi s ci rcl e of the
ci rcl es C(). The l ength of the connecti ng l i nk TR corresponds to
the arbi trari l y chosen l ength Bz..
I f necessary, tr y a successi on of val ues of Bj i n or der to fi nd that
whi ch gi ves the best fi t.
Fi gure 10.6 shows a transformer l i nkage, thus desi gned, f or the z,-scal e
of Fi g.. 10.4.
The Constraint Linkage.The possi bl e paths of the joi nt Tin l i nkages
of thi s type are the three-bar curves di scussed, from a mor e mathemati cal
poi nt of vi ew, by Roberts, Cayl ey, and Hi ppi sl ey. 1 Even as restri cted
by the choi ce of the el ements TR and RS of the transformer l i nkage, thi s
1S. Roberts, On Three-Bar Moti on i n Pl ane Space, R-oc. l fa~h. Sot., Lord., 7,
14 (1875).
A. Cayl ey, On Three-bar Moti on, Proc. Math. Sot., Lend., 7, 136 (1875).
R, L, Hi Pp~l ey, IA ATewMethod of Descri bi ng a Three-bar curve, Proc. M~h.
Sot., L07td., 18, 136 (1918).
!298 BARLI NKAGE FUNCTI ON GENERATORS [SEC. 10.4
i s a l arge fami l y of curves, wi th whi ch a great vari ety of curves A B can
be fi tted accuratel y.
A si mpl e graphi cal method suffi ces for the desi gn of these l i nkages.
We wi sh to choose l engths for the bars TQ and QR such that when the
poi nts T and R move over thei r r especti ve scal es the poi nt Q wi l l descri be
a ci rcl e. I f we then constrai n Q to move on thi s ci rcl e, by means of the
crank PQ, and R to move on the ci rcl e R, by means of the crank SR,
the joi nt T will be constrai ned to move al ong the scal e AB, as desi red.
We ther efor e pr epar e a chart that shows the scal es A B and CD i n thei r
pr oper rel ati on. Over thi s chart we pl ace a transparent over l ay on
whi ch i s marked a l i ne of l ength Bz, r epr esenti ng the bar TR. I f we now
Fm. 10,6.Transformer linkage for the z,-scale of Fig. 10.4.
move the ends T and R of thi s bar al ong thei r r especti ve scal es, the other
poi nts of the overl ay wi l l have the moti ons of poi nts ri gi dl y attached
to the bar TR. The path traversed by any poi nt of the overl ay can
qui ckl y be l ai d out on the chart. Compari son of a number of these paths
wi l l usual l y cal l attenti on to a regi on on the overl ayi n addi ti on to that
near the poi nt Rthat traverses a nearl y ci rcul ar path. Compari son
of the paths of a few poi nts of thi s regi on wi l l then suffi ce for the l ocati on
on the overl ay of the poi nt Q that has the most nearl y ci rcul ar path. The
l ength of the bars TQ and QR can then be measur ed on the over l ay;
the pi vot P wi l l be l ocated at the center of the ci rcul ar path, and PQ wi l l
have a l ength equal to i ts radi us. Thi s wi l l compl ete the determi nati on
of the l i nkage constantsto the accuracy possi bl e by graphi cal methods.
Fi gure 10.7 shows, for the exampl e of the precedi ng secti ons, the
paths of a number of poi nts of the overl ay. Si nce the poi nt T moves
over a roughl y ci rcul ar path AB, itwas to be expected that a ver y near l y
ci rcul ar path, QQ, woul d be found near bythat the bar l ength TQ
woul d be smal l . A sketch of the compl eted transf ormer-and-const rai n:
l i nkage for thi s exampl e i s shown i n Fi g. 108.
SEC. 1004] CURVE TRACI NG AND
D
TRANSFORMER LI NKAGES 299
~
{
CR
+P
FIG. 10.7.Paths of points rigidly attached to the har TR, as its ends move along
scales AI? and CD. The point P is the center of the circular path QQ. The scales are
the same as in Fig. 10.6.
FI G. 10.8.Complete tran8f0rmer-and-c0n8traint linkage for m-scale of Fig. 10.4.
APPENDI X A
TABLES OF HARMONIC TRANSFORMER FUNCTIONS
An extended di scussi on of the structure and use of these tabl es wi l l be
found i n Sees. 4.3 to 47.
A-l . Tabl e A*L Hi as a Function of Oi.Each unit of this table may
be read in two ways, according to the following schemes:
The defi ni ng rel ati ons are
H, =
si n (Xi n + Oi AXi ) (si n Xi)mim
(Sin Xi)- (si n Xi )ti .
(4.12)
Hz =
COS(Xk + @AXi ) (COS Xi )*
(COSXi )= (COS Xi )fi
(4.42)
(COSXi ).- (COSXi )mi JI ,
g = (si n Xi ).M (ti Xi )mi r,
(4.31)
wher e max and ti n i ndi cate, respecti vel y, the maxi mum and mi ni mum
val ues of the functi on i n questi on, when
A*2. Table A,2. @i aa a Function of Hk.In each column are tabu-
lated values of Oi corresponding to equally spaced val ues of H~, for the
val ues of Xi . and Xi M shown at the top of the col umn. Or dy those
vrdues of Xh and Xi ~ are i ncl uded for whi ch Oii s a si ngl e-val ued functi on
of Hh.
301..
TABLE A. 1,H~ AS A FUNCTION OF Oi
w
a
N AZ = 40
m
mKiGd.
El l =.
0.0
0.1
0.2
0,3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0
0,1
0.2
0.3
0.4
0.5
0.6
0,7
0,8
0.9
1.0
,.
0.0 0.0000 0.000o 1.0
0.1 0.0655 0.1286 0.9
0.2 0.1398 0,2523 0.8
0.3 0.2226 0,3705 0.7
0.4 0,3134 0.4826 0.6
0.5 0.4118 0.5882 0.5
0.0 0.0000 0.000o
0.0706 0.1346
0.1489 0,2630
0.2345 0.3847
0.3270 0.4989
0.4260 0.6051
0.5310 0.7028
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.0000 0.000o
0.0750 0.1421
0,1567 0.2763
0.2448 0.4021
0.3388 0.5188
0.4383 0.6259
0.5427 0.7228
0.6517 0.8090
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
-
0,0 0.000o 0.000o 1.0
0.1 0.0788 0.1517 0.9
0,2 0.1636 0,2934 0.8
0.3 0.2539 0.4246 0.7
0.4 0.3492 0.5446 0..6
0.5 0.4491 0.6527 0.5
0.6 0.5531 0.7485 0.4
0.7 0.664380.8315 0.3
0.8 0.7715 0.9013 0.2
0.9 0.8848 0.9576 0.1
1.0 1.0000 1.0000 0.0
100
1.7321 140
0.0000 0.0000
0.0823 0.1647
0.1698 0.3166
0.2621 0.4551
0.3586 0,5795
0.4589 0.6891
0.5625 0,7834
0.6690 0.8620
0.7777 0.9245
1.0
0.9
0.8
0.7
0,1
0.2
0,3
0.4
0.5
0,6
0.5
0,4
0.3
0.2
0.1
0.6 0.5174 0.6866 0.4
0.7 0.6295 0.7774 0.3
0.8 0.7477 0.8602 0.2
0.9 0.8714 0.9345 0.1
1.0 1.0000 1.0000 0.0
0.6
0.7
0.8
0.9
1.0
0.6414]0.7916
0.75680.8710
0.876510.9405
1.00001.0000
0.3
0.2
0.1
0.0
0.7
0.8
0.9
1.0
0.7646 0.8842
0. 8W9 0.9480
1.0000 1,0000
0.2
0.1
0.0
0.88830.9706
1.00001,0000 0.0
-iIiiJEE- m miIil!E
m
n
35
5 0.2737
0.0 0.000 qo,oooo 1,0
0.1 0.08860.2102 0,9
0.2 0,1810~0.3973 0.8
0.3 0.276710.5604 0,7
0.4 0.37540.6988 0.6
0.5 0,4764(0.8116 0.5
0.6 0,5793;0.8984 0.4
0.7 0.6836,0.9588 0.3
0,8 0.7889,0.9924 0.2
30
10 0.1989
I I I
1
25
15 ,0.1375
0.000o 0.000o
0.0943 0,2911
0,1911 0.5336
0,2901 0.7264
0,3906 0.8686
0.4923 0.9593
0.5946 0.9984
0.6970 0.9854
0.7990 0.9205
0.9002 0.8039
1.m,o,6364
2!ziiii_
0,0000 0.0000
0.0970 0.3577
0.1961 0.6377
0.2965 0.8387
0.3980 0.9595
0.5000 1.0000
1.0
0.9
0.8
0.7
0.6
0,5
0.4
0.3
0.2
0.1
0,0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0,8
0.9
1.0
o.om~o.000o
0.0857 0.1837
0.1757 0.3505
0.2697 0.4997
0.3673 0.6305
0.4680 0.7422
0.5712 0.8344
0.6766 0.9066
0.7835 0.9584
0.8915 0.9896
1.0000 1.0000
1.0
0.9
0.8
0.7
0,6
0.5
0,4
0,3
0.2
0.1
0.0
0.0
0.1
0.2
0.0000 0.0000
0.0915 0.2445
0.1861 0.4564
0.2835 0.6347
0.3831 0.7783
0.4845 0.8666
0.5871 0.9591
0.6904 0.9954
0.7940 0.9954
0.8974 0.9591
1.00000.8866
1.0
0,9
0.8
0.7
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0,9
1,0
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.3
0.4
0.5
0.6
0.7
0.8
0.6
0,5
0.4
0.3
0.2
0.1
0.6020 0.9595
0.7035 0.8387
0.8039 0.6377
0.9030 0.3577
1.00000.0000
--l
0,9 0.89450.9992 0.1
1.0 1. OQOCO.97890.0
850
3.6535 125
0,9
1.0 0.0
lEiitEl laiEl LE2LE
., .,.,, .,,. ,., ,,, ,, ,,, ,,,. .,, ,,,, , , , . ,,, ,,, > .., . .
TABLE A.1.H~ AB A FUNCTION OF & (Cd.)
AX< = 40
El
-=la
o
4; 0,3640
.
8{
.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.0000O.6364
0.0998o. 8039
0.2010 0.9205
0,3030 0.9854
0.4054 0.9984
0.5077 0.9593
0.6094 0.8686
0.7099 0.7264
0. S3890.5336
0.9057 0.2911
1.00000.0000
1.0
0.9
0.8
U
0.5
0.4
0.3
0.2
0.0000 0.9789
0.1055 0.9992
0.2111 0.9924
0.3164 0.9588
0.4207 0.8984
0,5236 0.8116
0.6246 0.6988
0.7233 0.5604
0.8190 0.3973
0.9114 0.2102
Ii?
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.0000 0.8866
0.1026 0.9591
0.2060 0.9954
0.3096 0.9954
0.4129 0.9591
0.5155 0.8866
0.6169 0.7783
0.7165 0,6347
0.8139o, 4564
0.9085 0.2445
I.0000 0.0000
1.0
0.9
0.8
0.7
0.0
0,1
0.2
0.3
1.0
0,9
0.8
0.7
0.6
0.5
:::
0.2
0.1
0.0
0.0
0.1
0.2
0.3
0.4
0.000+.0000
0.10850.9896
0.216510.9584
0. 3234~0.9066
0.42880.8344
0.5320 0,7422
0.6327 0.6305
0.7303 0.4997
0.8243,0.3505
0.914310.1837
1.00000.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.6
0.5
0.4
0.3
0.2
0.1
0.4
0.5
0.6
0.7
0.8
0.9
0.5
0.6
0.7
0.8
0.9
1.0
0.4
0.3
0.2
0.1
0.0
0.8
0.9
1.0
0.1
0,0
0.9
1.0 0.0 1.0 jlaoqo.oofq
- l Gd2d-
.
m m liiiJzH
r
Oi
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0

-=$3=

0.0 0.000o 1.0000

0,1 0,1286 0.9345
0.2 0.2523 0.8602
0.3 0.3705 0.7774
0.4826 0.6866
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0,2
0,1
0.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.000o 1.0000
0.1191 0.9480
0.2354 0.8842
0.3483 0.8090
0. 4573[0.7228
0.561710.6259
0.6612 0.5188
0.7552 0.4021
0.8433 0.2763
0.9250 0,1421
1.00000.ooOo
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.0
0,1
0.2
0.3
0.4
0.5
0,6
0.7
0.8
0.9
1.0
0,00001.0000
0.11520.9576
0.2285,0.9013
0,33920.8315
0. 4469;0,7485
0.5509,0.6527
0.65080.5446
0,746110.4246
0,83640.2934
0.92120,1517
1.00000.0000
1,0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.0000 1,0000 1.0
0.1 0.1235 0.9405 0,9
0.2 0.2432 0.8710 0.8
0.3 0.3586 0.7916 0.7
0.4 0.4690 0.7028 0.6
0.5 0.5740 0.6051 0.5
0.6 0.6730 0.4989 0.4
0.7 0.7655 0.3847 0.3
0.8 0.8511 0.2630 0.2
0.9 0.9294 0.1346 0,1
1,0 1.00000.000o 0.0
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0,5882 0.5882
0.6866 0.4826
0.7774 0.3705
0.8602 0.2523
0.9345 0.1286
1,0000 0.000o
0.1
0.0
-m-
1.7321 80
m l a
II
Ax, = 50
TABLE A.1.Hk AS A FUNCTION OF Oi (Cont.)
g
Is
=i=i -m=l
55
5 0.5774

.H==l.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.000o 0.0000
0.0769 0.1801
0.1609 0.3450
0.2512 0.4937
0.3473 0.6250
0.4483 0.7377
0,5535 0.8312
0,6621 0.9046
0.7733 0.9575
0.8862 0.9893
1,0000 1.0000
1.0
0,9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.000o 0.OOOO
0.0559 0.1349
0.1232 0,2643
0.2017 0,3874
0.2905 0.5032
0.3891 0,6109
0.4968 0.7095
0.6126 0.7983
0.7357 0.8768
0.8651 0.9441
1.0000 1.0000
1.0 0,0
0.1
0.2
0.3
0.4
0,5
0.6
0.7
10,00000.0000
0.0726 0.1638
0.1531 0.3160
0.2409 0.4554
0.3355 0.5811
0.4360 0.6920
0.5417 0.7873
0.6518 0.8663
1.0
0.9
0.8
0.7
0.6
0.5
0,4
0.3
0.0
0.1
0.2
0,3
0.4
0.5
0.6
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0,1
0.2
0.3
0.4
0.5
0.6
0.7
0.0000 0.0000
0.0677 0.1518
0.1444 0.2945
0.2295 0.4272
0.3224 0.5488
0.4224 0.6583
0.5287 0.7550
0.6404 0.8381
;:;
0.8
0,7
0.6
0.5
0.4
0.3
0.9
0,8
0.7
0.6
0.5
0.4
0.3
0,2
0,1
0.0
0.?
0.8
0.9
1,0
0.6275 G 8i61
0.7470 0.8903
0.8715 0.9517
1.0000 1.0000
0.8
0.9
1.0
0.75680,9069
0.87700.9611
1.00001.0000
0.2
0.1
0.0
0.8
0.9
1.0
0.7655 0.92&l
0.8818 0.9731
1.0000 1.0000
0.2
0.1
0.0
--l

100
1.4282 150
lEi2!IElm m IizJz?T
40
10 0.2866
.
E!
-=1=1-El
0.0
0,1
0.2
0.3
0,00000.0000
0.081010.2012
0.168110,3825
0.26080.5426
0.35820,6801
1.0
0.9
0,8
0.7
0,6
0.5
0,4
0.3
0.2
0.1
0.0000:0.0000
0.08480.2270
0.17490,4274
0.26970,5995
0.36840. 7422
0.0 0,0000o. 0000
0.0884 0.2592
0.1814 0.4819
0.2782 0.6665
0.3781 0.8116
0.4805,0,9160
1,0
0.9
0.8
0,7
0.6
0.5
0.4
0.0
0,1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1,0
1.0
0.9
0.8
0.7
0,6
0,5
0.0
0.1
0.2
0.3
0.4
0.5
10.00000.0000
0.0919 0,3007
0,1876 0.5498
0.2864 0.7457
0.3876 0.8866
0,4903 0.9716
0.5938 1. OOJ30
1.0
0.9
0.8
0.7
0,6
0.5
0.4
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0,9
1,0
0.0000o. 0000
0.0954 0.3563
0.1938 0,6364
0.2946 0,8379
0.3969 0.9593
0,5000 1.0000
0.6031 0.9593
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.1
0.2
0.3
0.4
0.5
0.4
0.5
0.6
0.7
0.8
0,9
0.4597 0.7941
0.5644 0.8837
0.6716 0.9481
0.7805 0.9870
0.8903 1.0000
1.0000 0.9870
0.4703 0,8544
0.5746 0.9351
0.6806 0.9837
0.7873 1.0000
0.8941 0.9837
1.0000 0.9351
0,4
0.3
0.2
0.1
0.0
06
0.7
0.8
0.9
1.0
0.5844IO,9789
0.6891 1.0000
0.7938 0.9789
0.8977 0.9160
1.0000 0.8116
0.6
0.7
0.8
0,9
1.0
0.3
0.2
0.1
0.0
0.6973 0.9716
0.8000 0.8866
0,9012 0.7457
1.0000 0.5498
0.3
0,2
0.1
0.0
0.7054 0,8379
0.8062 0,6364
0.9046 0.3563
1.00000.0000
0.3
0.2
0.1
0.0 1.0 0.0
- !zitiEd m- - -m- Iiz12Er m
i--
TABLE A. I.Hk AS A FUNCTION OF e, (Cont.)
T+=! -=~-
0.0
0,1
0.2
0.3
0.4
0.5
0.6
0.7
0.00000.5498
0,09880.7457
0.20000,8866
0.3027 0.9716
0.4062 1.0000
0.5097 0.9716
0.6124 0.8866
0.7136 ,0.7457
1,0
0,9
0.8
0.7
0.6
0.5
0.4
0.3
0.0
0.1
0,2
0.3
0.4
0.5
0.6
0.7
0.8
0.0000 0.9870
0,1097 1.0000
0,2195 Q.9870
0,3284 0.9481
0.435610.8837
0.54030.7941
0.64180.6801
0.73920.5426
0.83190.3825
0.91900.2012
1. 0000/0.0000
1.0
0.9
0.8
0.7
0,6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.1
0.2
0.3
0.4
0,5
0,6
0.7
0.000(10.0000
0.0954,0.3563
0,19380,6364
0. 2946)0,8379
0.3969 0,9593
0.5000 I.000o
0.6031 0.9593
0.7054 0.8379
0,8062 0.6364
0.9046,0.3563
1.00000,0000
1,0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.0
0.1
0.2
o,oooo~o.8116
0.10230.9160
0.20620.9789
0.310911.WX3O
0.41560.9789
0,519510.9160
0.62190.8116
0.72180.6665
0.818610.4819
0.911610.2592
1.000010.0000
1.0
0.9
0.0
0.1
o. 00000.9351
0.105910.9837
0.2127 1.OoW
0.3194 0.9837
0.4254 0.9351
0.52970.8544
0. 6316]0.7422
0.73030.5995
0,825110.4274
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.8
0.7
0.6
0.5
0.4
0.3
;:;
0.0
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.3
0.4
0.5
0.6
0.7
0,8
0.9
1.0
0.2
0.1
0.0
0.8
0.9
1.0
0.812410.5498
0.908110.3007
1.0000:0.0000
0.2
0.1
0.0
0.8
0.9
1,0
0.9 0.915210.2270) 0.1
1.0 1.000010.0000 0,0
0.9
1.0
.
60 T
4.6027 105
. .
45
2.7118 95
m
I
1 A
50
3.4897 100
.
16.28491 110~
Oi
/
5$ 0,4663

-I=E=l
-m
-r 1
20-0
70 1.0000
.
1.0
0.9
0.8
0.7
0.6
0.5
0,4
0.3
0.2
0.1
0.0
0.ooOo 1.0000
0.1182 0,9731
0.2345 0.9284
0.3482 0.8663
0.4583,0.7873
0.56400.6920
0.664510.5811
0.0000:1.0000
0.11380.9893
0. 2267;0.9575
0,3379 0.9046
0.4465 0.8312
0.5517 0.7377
0.6527 0.6250
0,7488 0,4937
0.8391 0.3450
0.9231 0,1801
1.00000.0000
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.0000 1.0000
0,1285 0.9517
0.2530 0.8903
0.3725 0.8161
0.4861 0. 72!38
0.5930 0.6321
0.6924 0.5236
0.7834 0.4052
1.0
0.9
0.8
0.7
0.6
0.5
04
0.3
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0,000011.0000
0.13490.9441
0. 2643~0.8768
0.38740.7983
0. 5032!0.7095
0.61090.6109
0.709510.5032
0. 7983~03874
1,0
0.9
0,8
0.7
0.6
0.5
0.4
0.3
0.0
0.1
0.2
0.3
0.4
():;
0.7
0.8
0.9
1,0
1.0
0.9
0.8
0.7
0.6
0,5
0,0 0.0 O.OOOojl.mo
0.12300.9611
0.2432)0.9069
0.3596 0.8381
0.4713 0.7550
0.5776 0.6583
0.6776 0.5488
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.1
0.2
0.3
0.4
0.5
0,1
0.2
0.3
0.4
0.5
0.6 0.4
0.3
0.2
0.1
0.0
0,6
0.7
0,8
0,9
1.0
0.7591 0.4554
0.8469 0.3160
0.9274 0.1638
I .000010.0000
0.7
0.8
0.9
1,0
tO7705t0.427Zl 0.3
0.2
0.1
0.0
0.85560.2945
0.93230.1518
1.Oom,o .0000
0.8
0.9
1.0
0.8654 0.2779
0.9378 0.1424
1.0000 0.0000
0.2
0.1
0.0
0.8 0.8768 0 2643
0.9441 0.1349
1.0000 0,0000
0.2
0.1
0.0
0.9
1.0
=I__2td1
~
30
1.4282 80
m IL
.
25
1.1918 75
m..
AXi = 60
TABLE A.1.Hh AS A FUNCTION OF 0; (~on!. )
D
70 I
10 0.8391
0.0 0.0000 0.0000 1,0
0.1 0.0534 0.1499 0.9
0.2 0.1196 0.2923 0,8
0.3 0.1980 0,4257 0.7
0.4 10.28770.5486 0.6
0.5 10.38760,6597 0.5
0.6 10.496710.7577 0,4
0.7 0.61380.8415 0.3
0.8 O.7377IO.91O3 0.2
0.9 0.86690.9634 0,1
1,0 1,00001.0000 0.0
100
1.131s 160
0.0
0.1
0,2
0.3
0.4
0.5
0.6
0.7
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.0
0.1
0.2
0,3
0.4
0,5
0,6
0.7
0.8
0.000o o. 0000
0.0658 0.1756
0.1419 0.3383
0.2274 0.4863
0.3213 0.6180
0.4227 0.7321
0.5303 0.8271
0.6432 0.9021
0.7599 0.9563
0,8793 0.9890
1.0000 1,0000
1.0
0.9
0.8
0,7
0.6
0.5
0.4
0.3
0,2
0.1
0.0
0.1
0,2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.9
0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0,oooo 0.0000
0. 060Q0.1611
0,1314 0.3124
0.2136 0.4522
0.3055 0,5790
0,4062 0.6913
0.5146 0.7880
0.6294 0.8680
0.7495 0.9305
0.8735 0.9746
1.0000 1.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
03
0.2
0.1
0.0
0.7
0.6
0,5
0,4
0.3
0.2
0.1
0.8
0.9
1.0
0.2
0.1
0.0
0.8
0.9
1.0
0.9
1.0 0,0 1.0 ll.OOOC!O.991110,0
.
105
1.0000 165 I
u
o
2.1254 1%
m Td2Er
-m m-
- 1=+4 -m-
45
1.5 0.3032
lo.000o~o.oooo
0.076010.2143
0. 160C10.4205
0.25130.5746
0. 3487~0 7~67
0.4512,0.8012
1,0
0.9
0.8
0.7
0.6
10.CQoa10,0000
0.0849 0,2693
0.1760 0.4997
0.2723 0.6887
0.3728 0.8344
0.4764 0.9351
0.0
0.1
0.2
0.3
0.4
0.5
1.0
0.9
0.8
0.7
0.6
0.5
0.0
0.1
0.2
0.3
0,4
0.5
0.6
0.000o o .0000
0.0891 0.3067
0.1836 0.5604
0.2823 0.7584
0.3842 0.8984
0.4883 0.9789
0.5933 0.9992
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.000o 0.0000
0.0933 0.3547
0.1910 0.6347
0.2921 0.8369
0,3955 0.9591
0.5000 1.0000
0.6045 0.9591
0,7079 0.8369
0.8090 0.6347
0.9067 0.3547
1,0
0.9
0.8
0,7
0.6
0.5
0.4
0.3
0.2
0.1
0.5
0.55780.9169 0.4
0.6671 0.9728 0.3
0.7781 0.9983 0.2
0.8894 0.9932 0.1
1.0000 0.9575 0.0
0.6
0.7
0.8
0.9
1.0
10.581910.989610.4
0.3
0.2
0.1
0,0
0.6881 0.9974
0.7940 0.9584
0.8984 0.8731
1,0000 0.7422
0.7
0.8
0.9
1,0
0.6981 0.9588
0.8016 0.8583
0,9026 0.6988
ll.oom 0.4819
0.2
0.1
0,0
0.9
1.0 Ip.000olo.oocq 0.0
m
1
&l l m
Li12E1 m
....,, .,,,,. .
TABLE A.1.Hk AS A FUNCTION OF ei (Cod.)
AX, = 60
3=E=l
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0. 0000 0.0000
0.0933 0.3547
0.1910 0.6347
0.2921 0.8369
0.3955 0.9591
0.5000 1.ooOo
0.6045 0.9591
0.7079 0.8369
0.8090 0.6347
0.9067 0.3547
1.00000.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
I i i di m
3a=l
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.OOOO 0.9911
0.1154 0.9997
0.2306 0.9825
0.3444 0.9399
0.4554 0.8722
0.5625 0.7803
0.6645 0.6650
0.7602 0.5278
0.8487 0.3700
0.9289 0.1934
1.0000 0.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0

35
2.1254 95
m
i i i LEr
[
6$ 0.5774
0.0 0.0000 1.0000 1.0
0.1 0.1207 0.9890 0.9
0.2 0.2401 0.9563 0.8
0.3 0.3568 0.9021 0.7
0.4 0.4697 0.8271 0.6
0.5 0.5773 0.7321 0.5
0.6 0.6787 0.6180 0.4
0,7 0.7726 0,4863 0.3
0.8 0.8581 0.3383 0.2
0.9 0.9342 0,1756 0.1
1.0 1,0000 0.0000 0.0
30
1,7321 90
-m-
l z
-=T=l
0.0
0.1
0.2
0.3
0,4
0.5
0.6
0.7
0.8
0.9
1,0
0.0000 1.0000
0.1265 0.9746
0.2505 0.9305
0.3706 0.8680
0.4854 0.7880
0.5938 0.6913
0,6945 0.5790
0.7864 0.4522
0.8686 0.3124
0.9400 0.1611
11.000010.00001
1,0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
ml
- =P=l
0.0 0.000o 0.8837 1.0
0.1 0.1060 0.9580 0.9
0.2 0.2138 0.9953 0.8
0.3 0.3221 0.9953 0.7
0,4 0.4299 0,9580 0,6
0.5 0.5359 0,8837 0.5
l a
-m-
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1,0
0.000o 1.0000
0.1331 0.9634
0.2623 0.9103
0.3862 0.8415
0.5033 0.7577
0.6124 0.6597
0.7123 0.5486
0.8020 0.4257
0.8804 0.2923
0.9466 0,1499
1.0000 0,0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-FF=l-
0.0
0.1
0.2
0.3
0,4
0.5
0.6
0.7
0.8
0,9
1.0
0.0000 0.9575
0.1106 0.9932
0.2219 0.9983
0.3329 0.9728
0.4422 0.9169
0.5488,0.8312
0.65130.7167
0.748710.5746
0.8400 0.4065
0,9240 0.2143
1.0000 0. Ocoo
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1 1 I
40
2.6306 100
-m
0.0
0.1
0.2
03
0.4
0.5
0.6
0.7
0,8
0.9
1.0
0.0000 1.0000
0.1408 0.9543
0.2760 0.8940
0.4042 0.8200
0.5240 0.7330
0.6340 0.6340
0.7330 0.5240
0.8200 0.4042
0,8940 0.2760
0.9543,0.1408
1.00000.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0,3
0.2
0.1
0.0
m Iiiziar
w
G
w
Axi = 70
TABLE A.1.Hk AS A FUNCTION OF 6; (Cont.)
r
750
5 10.83911

11-
70.
65
Ai0u732
5 O.58121
... ,
Zi=El=a.
FFLa.
0.0 ~o.OOOoo.oooo 1.0
0,1 0.06480.2036 0.9
0.2 10.1411 O.3893 ~0.8
0.3 0.22770.5543 0.7
0.4 0.32330.69611 0.6
0.5 0.42650.8126 0.5
0.6 0.53580.90211 0.4
0,7 10.649410 9633: 0.3
0.8 10.7659 ~0.99511 0.2
0.9 O.8833O.9973I 0.1
1.0 1.0000,09696 0.0
0.0
0.1
0.2
1,0
0,9
0,8
0.7
0.0
0.1
0.2
0.3
O.OOOoo.oood
0.05180.1702
0.1178,0.3301
0.19680.4773
0. 2879!0.6096,
1.0 0.0 O.OOOOIO. OOOO1.0
0.9 0.1 ,0,05860.1858 0.9
0.8 0.2 0.13000.3580 0.8
0.7 0.3 0.2130 !0.5139 0.7
0.0
0.1
0.2
0.3
o. OOOOJO .0000
0,0441,0.1570
0.1039,0.3065
0.17860.4461
1.0
0.9
0.8
0.7 0.3
0.4
0.5
0.6
0.7
0.8
0.9
0,6
0.5
0.4
0.3
0.2
0.1
0.4
0.5
0.6
0.7
0.8
0.9
O.2613J;0 .5739
0.36770.6879
0. 4794!0.7864
0.6003,0.8679
0.72870.9313
0. 8626IO.9755
0.6 0.4
0.5
0.6
0,7
0.8
0.9
0.6
0.5
0.4 0.30650.6513 0.6
0.5 0.4090~0.7680 0.5
0.6 10.51900.8623 0.4
0.7 10.634810.93290,3
0.8 ~0.754810.9787 0.2
0.9 0,877110.9989 0.1
1.0 ;l.OOOO:O.9934 0.0
0.5
0.4
0.3
0.2
0.1
]0.38960.72511
10.50040.8221,
10,61860.8991
10.74260.9549/
FJ?~~l
0.4
0.3
0.2
0.1
0,0 1.0 11.till.00001 0.0 i.o 11.0000il.000010.0 1,0

100
1.0000 170
I I I
95
1.1918 165
I I
r
1.4282 1%
I ,%O=. _
,3.0794 150
I 85 I
1. 7206; 155

-m
T
~-4Y~2047

=35
35 o. 1577
I ~1
, -----
0.0000 0.0000
0.0810 0.2762
0.1700 0.5123
0.2659 0,7048
0.3671 0.8508
0.4722,0.94811
0.5796,0,99531
0.0 0
0.1 10
.00000.0000
.0859!0.3104
1.0 : 0.0
0.3 0.1
0.2
0.3
0.4
0.5
0.6
07
0.2 0.8
0.1 10.9
0.0 1,0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
1,0
0.9
0.8
0,7
0.6
0.5
0.4
0.0 0,0000o. 0000
0.0758 0.2479
0.1609 0.4654
0.2538 0.6490
0.3533 0.7961
0.4578 0.9046
0.5657 0.9728
0.6756 0.9996
0.7857 0.9847
0.8944 0.9282
1.0000 0.8312
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0,0
0.1
0.2
0.3
0.4
0.5
0.6
1.0
og
0.8
0.7
0.6
0.5
0.4
0.3
10oooo~o.0000
~0;090810.3528
10 18760.6327
10:289110.8358,
0.39380.9588
i O.500011 .00W
~0 6062~0.9588
0.71690.8358
~0.8124O.6327
1,0
0.9
0,8
0,7
0,6
0.5
0.4
0,3
0.2
0.1
0.2
0.3
0.4
0.5
0.2 0.17890 .567410.8
0.3 0,27760.7671 0.7
0.4 0,380510.0066~0.6
0.5 \0,4862~0. W371 0.5
0.6 0,5930,0.9974 0.4
0.7 0.6!)94i 0.947410.3
0.6
0,7
0.8
0.9
1.0
0.7
0.8
0.9
1.0
0.3
0.2
0.1
0.0
0.3
0.2
0,1
0.0
0.7
0.8
0.9
1.0
0.68770.9917
0.79480.9373
0.899410.8329
1,00000,6801
0.7761 0.9996
0.8890 0.9772
1.00000.9201
0.2
0,1
0.0
0.8 10
0.9 ,0
1.0 1
.80370, 8344(
.90440.66021
0000~0.4272
10.90920.3528!
1 Oooop .00001 ~.
01
0,0
ml
1 1
70
,3.1020 140 1
,-- ._.l-mo..l_ .
;3.8571, 135 ~
.
I-----
,,
~l
[ 60 ~
4.8846 130
... . .
6.3432 l ;!: I
Axi = 70
TABI,E A.1.H~ AS A FUNCTION OF 0; (Cont.)
m=i==l--
25
45 0.2593
I ml
-R-
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1,0
0.0000 0.4274
0.0956 0.6602
0.1963 0,8344
0.3006 0.9474
0.4070 0.9974
0.5138 0.9837
0.61950.9066
0.722410.7671
0.821110.5674
O,9141IO.31O4
l.OOOoo. oooo
1.0 1.0 0.0
0.9 0.1
0.8 0.2
0.7 0.3
0.6 0.4
0.5 0.5
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.0000
0,0908.0.3528
0.1876 0.6327
0.2891 0.8358
0.3938 0.9588
0.5000 1.0000
0.6062 0.9588
0.7109 0.8358
0.8124 0.6327
0.9092 0.3528
1.0000 0.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
D.O
D.1
0.2
0.3
0.4
0.5
0.0000 0.6801
0.1006 0.8329
0.2052 0.9373
0.3123 0.9917
0.42040.9953
0.527810.9481
0 .6329)0. 8508
0.734110.7408
0.830C[0,5123
0.9190,0,2762
l.OOOoo. oooo
0,Oooo 0.8312
0.1056 0.9282
0.2143 0.9847
0.3244 0.9996
0.4343 0.9728
0.5422 0.9046
0.6467 0.7961
0.7462 0.6400
0.8391 0.4654
0.9242 0.2479
1.0000 0.0000
1.0
0.9
0,8
0.7
0,6
0.5
0.4
0,3
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.0
0.9
0.8
0.7
0.6
0.5
0.4
(3:;
0.9
0,8
0.7
0.6
0.5
0.4
0,3
0.2
0.1
0.0
0,6
0.7
0.8
0.9
1.0
0.4 0.6
0.3 0.7
0.2 0.8
0.1 0.9
0.0 1.0
0.2
0.1
0:0
0.848710.4248
0.92950.2241
I .OOOc:o, oooo
0.1
0.0
-M m l a
l l~lLl lhiid
lk5zi!!m
r
Oi
0.0
0,1
0.2
0.3
0,4
0.5
0.6
0.7
0.8
0.9
1.0
m-
-=i==l-
0.0 O.OOOol. oooo 1.0
0,1 0.129710.9887 0.9
0.2 0.2574,0.9549 0.8
0.3 0.38140.8991 0.7
0.4 0.4996~0.8221 0.6
0.5 0.6104 0.7251 0.5
0.6 0.7121 0.6096 0.4
0.7 0.8032 0.4773 0.3
0.8 0.8822 0.3301 0.2
0,9 0.9482 0.1702 0.1
1.0 1.00000.0000 0.0
0.000o 0.9696
0.1167 0.9973
0.2341 0,9951
0.3506 0.9633
0.4642 0.9021
0.5735 0.8126
0.6767 0.6961
0.7723 0.5543
0.8589 0.3893
0.9352 0.2036
1.0000 0.ooOo
0.0 0.00000.9934 1.0
0.1 0.122910.9989 0.9
0.2 0.24520.9787 0.8
0.3 0.3652,0.9329 0.7
0,4 0.481010 .8623 0.6
0.5 0.59100.7680 0.5
0.6 0.69350.6513 0.4
0.7 0,787010.5139 0.3
0.8 0.870010.3580 0.2
0.9 0.94140.1858 0.1
1.0 1.00000.0000 0.0
1,0
0.9
0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0,7
0.8
0,9
0.000o 1.ooOo
0.1374 0.9755
0.2713 0.9313
0.3997 0.8679
0.5206 0,7864
0.6323 0.6879
0.7331 0.5739
0.8214 0.4461
0.8961 0.3065
0.9550 0.1570
1.0000 0.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.1
0.2
:::
0.5
0.6
0.7
0.8
0,9
1.0
0.Oooo 1.0000
0.1464 0.9648
0.2874 0.9122
0.4209 0.8426
0.5449 0.7574
1.0
);
0:6
0.5
0.7
0.6
0,5
0.4
0.3
0.2
0.1
0.6576 0.6576
0.7574 0.6449
0.8426 0.4209
0.9122 0.2874
0.9648 0.1464
1.(MOOO.0000
0.4
0.3
0.2
0,1
0.0
0.0 1.0
u-
._.
10
1.0000 80
l a-
I &
. .
25
1.7206 95
m
.
AX, = 80
TABLE A.1.Hk ASA FUNCTION OF 9i (Cont.)
l =T=- El - -=$=
-R
=m=d
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.0000 0.0000
0.0240 0.1516
0.0689 0.2984
0.1337 0.4374
0.2173 0.5662
0.3180 0.6820
0,43380,7827
0. 5626!0.8663
,0.701610.9311
1.0
0.9
0.0
0.1
0.2
0.00000 .0000 1.0
0.03420.1638 0.9
0.08740.3203 0,8
0.7
0.0
0.1
0.2
0.3
0,4
O.OOOCO. OOCHI
0.051010.1937
0.1174 0.3735
0.1979 0.5359
0.2910 0.6778
0.3949 0.7964
0.5076 0.8893
0.6267 0.9549
0.7501 0.9917
0.8754 0.9991
1.00000.9769
1.0
0,9
0.8
0.7
0.6
0.5
0.0
0.1
0.2
0.3
0.4
0.5
0.6
U
0.9
1.0
1.0
::;
0.7
0.6
0.5
0.4
0.3
0.8
0.7
0.6
0.5
0.4
0.3
0.2
03
0.4
0.5
0.6
0.7
0.8
0.9
0.1582 0.4666
0.2454 0.5996
0.3473 0.7169
0,4619 0,8161
0.5870 0.8954
0.7201 0.9531
0.8587 0.9882
1.0000 1.0000
0.6
0.5
0.4
0.3
0.2
0.1
0.5
0.6
0.7
0.8
0.9
1.0
0.4
0.3
0.2
0.1
0.0
0.76290.9997
0.88250.9871
1. OOOOO.941O
0.2
0.1
0.0
0.9
1,0
0.84840.9760
1.00001.0000
0.1
0.0 1.0 0.0
-kiiizd- -ciiitiE- m m
55
25 0.3434
I I I
l=E=l
-m-
El .
0.0
0.1
0.2
0.3
0.4
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.00.Oooco. 0000 1.0
0.1 0.0766 0.2809 0.9
0.2 0.1634 0.5214 0.8
0.3 0.2588 0.7167 0.7
0.4 0.3610 0.8630 0.6
0.5 0.4679 0.9575 0.5
0.6 0.5775 0.9983 0.4
0.7 0.6876 0.9847 0.3
0.8 0.7962 0,9169 0.2
0.9 0.9010 0.7961 0.1
1.0 1.0000 0,6250 0.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.000o
0.0646 0.2313
0.1420 0.4387
0.2303 0.6180
0.3283 0.7659
0.4338 0.8794
0.5448 0.9563
0.6591 0.9951
0.7746 0.9951
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.0
0,1
0.2
0.3
0.4
0.5
0.6
0,7
0.8
pxll&l mc&
0.1529 0.4774
0.2449 0,6650
0.3450 0,8136
0.4512 0,9201
0.5615 0.9825
0.6737 0.9997
0.7856 0.9711
1.0
0,9
0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.Oooo 0.0000
0.0822 0.3125
0.1736 0,5719
0.2723 0.7732
0.3765 0.9125
0.4841 0.9870
0.5930 0.9953
0.7011 0.9373
0.8064 0.8140
0.9066 0.6279
1.0000 0.3825
1.0
0.9
0,8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.7
0.6
0.5
0.4
0,3
0.2
0.1
0.0
0.5
0.6
0.7
0.8
0.9
1.0
0.3
0.2
0.1
0.0
088900.9563
1.00000.8794
0.1
0.0
0.9
1.0
0.895i 0.8975
1.00000,7803 1.0
- I i Ei kEr -liziM
l lkmd
m IziiLE1 kMi-Mt-
. .. ., . . ...
AXi = 80
TABLE A.1.H, AB A FUNCTION OF & (Cont.)
=Ta -=i=4-
-n
25
55 0.3434

0.0000 0.7803
0.1049 0.8975
0.2144 0.9711
0.3263 0.9997
0.4385 0.9325
0.5488 0.9201
0.6550 0.8136
0.7551 0.6650
0.8471 0.4774
0.9293 0.2543
1.00000.000o
-Po=l-
0.0
0,1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.0000
0.0878 0.3505
0.1836 0.6305
0.2856 0.8344
0.3918 0.9584
o.50001.0000
0,6083 0.9584
;:;
0.8
0.7
0.6
0.5
0.4
0.0
0.1
0.2
0.3
0.4
0.5
0,6
0.0000 0.3825
0.0934 0.6279
0.1936 0.8140
0.2989 0.9373
0.4070 0.9953
0.5159 0,9870
0.6235 0,9125
1.0
0.9
0,8
0,7
0.6
0,5
0,4
0.0
0.1
0.2
0.3
0.4
0.5
1.0
0.9
0.8
0.7
0.6
0.5
0.0
0.1
0.2
0.3
;::
0.6
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.0000 0.8794
0.1110 0.9563
0.2254 0.9951
0.3409 0.9951
0,4552 0.9563
0.5662 0,8794
0.6717 0,7659
0.7697 0.6180
0.8580 0.4387
0.9354 0.2313
1.oOoo 0.0000
;::
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0,0
0.6
0.7
0.8
0.9
1.0
0.4
0.3
0.2
0.1
0.0
0.7144 0,8344
0,8164 0.6305
0.9122 0,3505
,1.00000,0000
0,3
0.2
0,1
0.0
0.7
0,8
0.9
1.0
0.7277 0.7732
0.8264 0.5719
0.9178 0,3125
1.0000 0,0000
0.3
0.2
0.1
0,0
0.7412 0.7167
0.8366 0.5214
0.9234 0.2809
1.00000.0000
0.7
0.8
0.9
1.0
0.3
0.2
0.1
0.0
0.7
0.8
0.9
1.0
-kiik!El- m
1 1 I
40
3.5442 120
TEiiLa
El El - -m- - l =i =l
0.0
0.1
0.2
0.3
0.0000 69410 1.0
0.1175 0.9871 0.9
0.2371 0.9997 0.8
0.3564 0.9787 0.7
0,0
0.1
0.2
0.3
0,0000 0.9769
0.1246 0.9991
0.2499 0.9917
0.3733 0.9549
1,0
0.9
0.8
0.7
0.6
0,0 0.0000 0,9949
0.1 0.1325 0.9982
0.2 0.2640 0.9752
0.3 0.3919 0.9265
1.0
0.9
0.8
0.7
0.0 0.000o 1.0000 1.0
0.1 0.1413 0.9882 0.9
0.2 0.2799 0,9531 0.8
0.3 0.4130 0,8954 0.7
0.4 0.5381 0,8161 0,6
0.5 0.6527 0.7169 0.5
0.6 0.7546 0.5996 0.4
0.7 0.8418 0.4666 0.3
0.8 0.9126 0,3203 0.2
0.9 0.9658 0.1638 0.1
1.0 1.00000.0000 0.0
0.00.0000 l.oOCwl 1.0
0.1 0.1516 0.9760 0.9
0.2 0.2984 0.9311 0.8
0.3 0.4374 0.8663 0.7
0.4
0.5
0.6
0.7
0.8
0.9
1,0
0.4731 0.9243
0.5849 0.8377
0.6896 0.7206
0.7852 0.5752
0.8699 0.4043
0.9419 0.2113
1.0000 0.0000
0.6
0.5
0.4
0,3
0.2
0,1
0.0
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.4924 0,8893
0.6051 0.7964
0.7090 0.6778
0.8021 0.5359
0.8826 0.3735
0.9490 0.1937
1.0000 0.0000
0.4
0.5
0,6
0.7
0.8
0.9
1.0
0.5139 0.8530
0.6274 0.7560
0.7304 0.6375
0.8207 0.4999
0.8967 0.3457
0.9569 0.1780
1.00000.0000
0.6
0.5
0.4
0.3
0.2
0.1
0,0
0.5
0.4
0.3
0.2
0.1
0.0
-t&iiE1 -hiiiiElm liziJ2El LiIiEE1
AX; = 90
TABLE A1.-Hk A8 A FUNCTION OF Oi (Coni.)
m.
-Em
1
h 90
0 1.0000
. .
0.0 0.0000 0.000o
0.1 0.0123 0.1564
0.2 0.0489 0.3090
0.3 0.1090 0.4540
0.4 0.1910 0.5878
0,5 0.2929 0.7071
0,6 0.4122 0.8090
0.7 0.5460 0.8910
0.8 0.6910 0.9511
0.9 0.8436 0.9877
1.0 1.0000 1.0000
_r2iizza.
0. 0
0.1
0.2
0.3
0.0000 0.0000
0.0508 0.2170
0.1183 0.4159
0.2010 0.5917
0.2969 0.7402
0.4034 0.8576
0.5181 0.9411
0.6381 0.9887
0.7604 0.9991
0.8820 0.9721
1.00000.9083
:::
0.8
0.7
0.6
:::
0.3
0.2
0.1
0.0
0.0000 0.000o
0.0339 0.1838
0.0879 0.3580
0.1607 0.5181
0.2505 0.6604
0.3550 0.7811
0.4717 0.8775
0.5977 0.9471
0.7300 0.9882
0.8652 0.9998
1.0000 0.9816
0.000o 0.000o
0.0428 0.1996
0.1038 0.3856
0.1819 0.,5536
0.2748 0.6993
0.3804 0.8192
0.4961 0.9104
0.6189 0,9705
0.7459 0.9982
0.8740 0.9926
1.0000 0.9540
0.0
0.1
0.2
0.3
1.0
0.9
0.8
0.7
0.0
0.1
0.2
0.3
1.0
0.9
0.8
0.7
1,0
0.9
0.8
0.7
0.6
0,5
0.4
0.3
0.2
0.1
0.0
0,4
0.5
0.6
0.7
0.8
0,9
1.0
0.6
0.5
0.4
0.3
0.2
0.1
0.4
0.5
0.6
0,7
0.8
0.9
1.0
0.6
0.5
0,4
0,3
0.2
0.1
0.0
0.4
0.5
0,6
0,7
0.8
0.9
1,0 0.0
l a LL
.
90
1.0000, 180
Oi 65
25 0.4345

L
I
85
1.1868 175
Ire IL-!
75
1.6524 165
F50 I I
El
1-55 I I I 45 I I
35 0.3062 4010. 2536[
.
I I I I
I 4510.20711
l l l
0,0 0.0000 0.000o 1,0
0,1 0.0651 0.2586 0.9
0.2 0.1441 0.4863 0.8
0.0
0.1
0.2
0.3
0.0000 0.0000
0.0844 0.3479
0.1790 0.6279
0.2815 0.8329
0.3894 0,9580
0.5000 1.0000
0.6106 0.9580
0.7185 0.8329
0.8210 0.6279
0.9156 0.3479
1,0000 0.0000
A:;
0.8
0.7
0.6
0.5
0,3
0.4
0.5
0,6
0.7
0.8
0.9
0,2353 0.6773
0.3362 0.8271
0.4445 0.9319
0,5575 0.9890
0.6723 0.9973
0.7862 0.9563
0.8963 0.8672
1.000CIO.7321
0.7
0.6
0.5
0.4
0.3
0.2
0,1
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.4
0.3
0.2
0.1
0.0 1.0 0.0
m LiiiLir
.
TABLE A.1.Hk AS A FUNCTION OF e, (Cd.)
AXi = 90
I =l =i
.E1 I=m=l
-P%=
0; 45
45 0.2071
1
1.0 0.0
0.9 0.1
0.8 0.2
0.7 0.3
0.6 0.4
0.5 0.5
0.000o 0,5759
0.0971 0.7627
0.2019 0.8975
0.3119 0.9772
0.4244 0.9997
0.5365 0.9644
0.6456 0.8722
0.7489 0,7255
0.8440 0.5278
0.9283 0.2839
1.0000 0.0000
0.0
0.1
0.2
0.3
0.4
0,5
0.6
0.7
0.0000 0.7321
0.1037 0.8672
0.2138 0.9563
0.3277 0.9973
0.4425 0.9890
0.5555 0.9319
0.6638 0.8271
0.7647 0.6773
0.8559 0.4863
0<93490.2586
1.0000 0.000o
;::
0,8
0.7
0.6
0.5
0.4
0.3
0.2
0.0 0.0000 0.0000 1.0
0.1 0.0844 0.3479 0.9
0.2 0.1790 0,6279 0.8
0.3 0.2815 0.8329 0.7
0,4 0.3894 0.9580 0.6
0.5 0.5000 1.0000 0.5
0.6 0.6106 0.9580 0.4
0,7 0.7185 0.8329 0.3
0.8 0.8210 0.6279 0.2
0.9 0.9156 0.3479 0.1
1.0 1.0000 0.0000 0.0
0,0
0.1
0.2
0.3
0.4
0.5
0.0000 0.3450
0.0907 0.6002
0.1904 0.7961
0.2966 0.9282
0.4067 0.9932
0.5181 0.9893
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.0
0.1
0.2
0.3
0.4
0.5
0,6
0.7
0.8
0, 0000 0.8377
0.1106 0.9329
0.2263 0.9871
0.3443 0.9989
0.4616 0.9682
0.5754 0.8955
0.6828 0.7828
0.7813 0.6328
0.8684 0.4492
0.9419 0.2365
1.00000.0000
;::
0.8
0.7
0.6
0.5
U
0.2
0.6
0.7
0,8
0.9
1.0
0.6280 0.9169
0.7336 0.7774
0,8324 0.5746
0.9219 0.3133
1.0000 0.0000
0.4 0.6
0.3 0.7
0.2 0.8
0.1 0.9
0.0 1.0
0.2
0.1
0.0
0,8
0.9
1.0
0.1
0.0
0.9
1.0
0.1
0.0

45
4.8284 135
--l
.
25
2.3016 115
I
~ lln3d 40
3.9440 130
(
1-l
-3r
,2.7321 120
15
75 0.6052
.
El.
-m
JELia.
0.0 0.0000 0.9540 1.0
0.1 0.1260 0.9926 0.9
0.2 0.2541 0.9982 0.8
0.3 0.3811 0.9705 0.7
0.4 0.5039 0.9104 0.6
0.5 0.6196 0.8192 0.5
0.6 0.7252 0.6993 0.4
0.7 0.8181 0.5536 0,3
0,8 0.8962 0.3856 0.2
0.9 0.9572 0.1996 0,1
1.0 1.0000 0.0000 0.0
0,0
0,1
0.2
0.3
0.0000 0,9816
0.1348 0.9998
0.2700 0.9882
0.4023 0.9471
0.5283 0.8775
1.0
0.9
0.8
0.7
0.6
0.5
0.0
0.1
0.2
0.3
0.4
0.5
0.6
:::
0.9
1.0
0.0000 1.0000
0.1564 0.9877
0.3090 0.9511
0.4540 0.8910
0.5878 0.8090
0.7071 0.7071
0.8090 0.5878
0.8910 0.4540
1.0 0.0 0.000o 0.9083
0,1 0,1180 0.9721
0.2 0.2396 0,9991
1.0
0,9
0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.000o 0.9958
0.1448 0.9973
0.2881 0.9719
0.4262 0,9202
0.5559 0,8435
0.6738 0.7437
0.7771 0.6232
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.9
0.8
0.7
0,6
0.5
0.7
0,6
0.5
0.4
0.3
0.2
0.1
0.4
0.5
0.6
0,7
0.8
0.9
1.0
o.6450{o.78ii
0.7495 0.6604
0.8393 0.5181
0.9121 0.3580
0.9661 0.1838
1.0000 0,0000
0.4
0.3
0.2
0.1
0.0
0.6
0,7
0.8
0.9
1.0
0.4
0.3 0.8632 0.4850
0.9301 0.3326
0.9761 0.1695
1.0000 0.0000
0,3
0.2
0.1
0,0
0.95110.3090
0.98770.1564
1.00000.0000
0.2
0.1
0.0 0.0
I
_
1.0000 9$
-lizitiId =1
Axi = 100
TABLE A. 1.H~ AS A FUNCTION OF Oi (Cont.)
-- -l=T=l
TT=lPza
rll
-75
25 0.5338
.
0.0 0.0000 0.0000 1.0
0.1 0.0429 0.2210 0.9
0.0
0.1
0.2
0.3
0.4
0.0000 0.0000
0.0241 0.1881
0.0716 0.3675
0.1411 0.5329
0.2303 0.6791
0.3367 0.8019
1.0
0,9
0.8
0.7
0.6
0.5
0.4
0.0 0.0000 0.0000 1.0
0.1 0.0130 0,1736 0.9
0,2 0.0514 0.3420 0.8
0.3 0.1142 0.5000 0.7
0.4 0.1994 0.6428 0.6
0.5 0.3044 0.7660 0.5
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0,9
1.0
0.000o 0, OCQO
0.0340 0.2038
0.0895 0.3949
0.1649 0.5677
0.2578 0.7169
0.3654 0.8379
0.4844 0.9270
0.6114 0.9816
0.7422 1.0000
0.8731 0.9816
1.0000 0.9270
1.0
0.9
0,8
0.7
0.6
0.5
0.4
0.2 0
0.3 0
0.4 0
0,5 0
0,6 0
0.7 0
0.8 0
0,9 0
1,0 1
.10570,4247
,18640.6048
,28260.7560
.39130.8736
,50920.9540
.63290.9949
0,8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.5 0.2694 0.7306 0,5
0.6 0.3922 0.8337 0.4
0.7 0.5311 0.9138 0,3
0.8 0.6818 0.9687 0.2
0.9 0,8397 0,9965 0.1
1.0 1,0000 0.9965 0.0
0.5
0.6
0.7
0,8
0.9
1,0
0.6 0.4260 0,8660 0.4
0.7 0.5606 0.9397 0.3
0.8 0.7041 0.9848 0.2
0.9 0.8520 1.0000 0.1
1.0 1,00000,9848 0.0
0.4570 0.8974
0.5875 0.9627
0.7243 0.9958
0.8632 0.9958
1.0000 0.9627
0,3
0.2
0.1
0.0
0,3
0.2
0.1
0.0
.75840.9949
1.88200.9540
.00000.8736
iiiJ2El Bl l a
F

-R-
BE=l
.!=ii=l.
1.0
0,9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.0000 0.0000
0.1 0.0734 0,3131
0.2 0.1609 0.5759
0,3 0.2598 0.7803
1.0
0.9
0.8
0.7
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.0000 0.0000
0.0512 0.2401
0.1206 0.4571
0.2062 0.6444
0.3054 0.7964
0.4151 0.9083
0.5321 0.9769
0.6527 1.0000
0.7733 0.9769
0.8903 0.9083
1.0000 0.7964
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0,3
0,2
0.1
0.0
0.0
0,1
0.2
0.3
0.4
0.5
0.6
0,7
0.8
0.9
1.3
0.000o 0. 000(
0.0589 0.261<
0.1346 0.492;
0.2248 0, 686t
0.3268 0.837{
0.4375 0 .941(
0.5535 0.9934
0.6713 0. 993/
0.7873 0.941(
0.0 O.OCHJ O 0,0000 1.0
0.1 0.0663 0,2856 0.9
0.2 0.1480 0,5321 0.8
0,3 0.2426 0.7321 0.7
0.4 0.3473 0.8794 0.6
0.5 0.4589 0.9696 0.5
0.6 0.5740 1.0000 0,4
0.7 0.6891 0,9696 0,3
0.8 0.8007 0.8794 0,2
0,9 0.9054 0.7321 0.1
1.0 1,Oooo0,5321 0.0
0.4
0.5
0,6
0.7
0.8
0.9
0.3671 0.9201
0.4796 0.9911
0.5938 0.9911
0. 7063!0.9201I
:::%:%
0.6
0,5
0.4
0.3
0.2
0.1
0.8
0.9
1.0
0.89800. 83?7
1.00000. 686[ 1.0 0.0
m m m m l Gt i 2El
.
I
I
O
d
m
m
+
m
a
b
r
m
a
o
,
.
.
.
.
.
.
.
.
.
.
0
0
0
0
0
0
0
0
0
0
+
o
m
m
b
p
~
+
m
m
d
o
:
o
o

o
o
o
o
&
o
o

I
o
a
l
m
r
-
w
m
*
m
m
d
o
*

0
0
0
0
0
0
0

0
0
o

I
3
1
5
o
A
m
c
o
d
4
1
0
w
b
m
m
o
.
.
.
.
0
0
0
0
0
0
0
0
0
0

I
O
q
c
m
b
o
m
-
$
m
m
-
o
~

o
o

o
o
o
o
o
o
o
o
I
-

1
!
0
0
0
0
0
0
0
0
0
0
*

I
X
]
o
.
m
m
+
u
a
a
b
c
a
a
o
0
0

0
0
0
0
0
0
0
0
*

o
m
K
l
t
-
w
u
3
-
#
m
3
e
4
.
o
4

0
0
0
0
0
0
0
0
0

O
+
e
l
c
m
+
m
w
t
-
m
a
o
O
o
o
o
o
o
o
o
o
o
-
i
o
c
n
m
t
-
w
*
*
m
w
.
o
~

o
o
o
o
o
o
o
o
o
o

O
t
i
m
m
+
m
c
o
t
-
m
m
o
O
o
o
o
o
o
o
o
o
o
i
o
m
c
o
t
.
w
m
~
m
w
.
o
A
o
o
o
o
o

o
o
o
o
o
I
O
+
o
l
m
.
$
m
a
t
.
c
a
a
o
0
0

0
0
0
0

0
0

0
0
+

!
I
o
p
c
a
r
-
m
m
~
m
w
-
o
4
0
0
0
0
0
0

0
0

0
I
I
o
+
t
x
m
+
m
w
h
c
o
a
o
O

o
o
o
o
o
o

o
o

i
I
1
O
m
c
o
r
-
w
u
a
+
l
m
.
x
.
o
A
o

o
o
o
o
o
o
o
o
0
I
I
AXi = 110
TABLE A.1.Hk AS A FUNCTION OF Oi (Ootd.)
El
-l=F=l-
0.0
0,1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.0030 0.0000
0.0044 0.1763
0.0347 0.3491
1.0 0.0
0.9 0.1
0.8 0.2
0.7 0.3
0.0000 0.0000
0.0137 0.1908
0.0543 0.3746
0.1202 0.5446
;:;
0.8
0.7
0.6
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.0129 0.0000
0.0001 0.1628
0.0186 0.3251
0.0677 0. 48Q9
0.1457 0.6244
0.2496 0.7504
0.3756 0.8543
0.5191 0.9323
0.6749 0.9814
1,0
0.9
0.8
0.0
0,1
0,2
0.3
0.4
0,0000 0.0000
0,0246 0.2065
0.0741 0.4019
0.1467 0.5790
0.2397 0.7313
0.3497 0.8532
:::
0.8
0.7
0.6
0.5
0.4
0.7
0.6
0.5
0.4
0.3
0.0930 0.5120
0.1770 0.6590
0.2838 0.7848
0.4093 0.8847
0.5489 0.9550
0.6976 0.9931
0.8498 0.9978
1,0000 0.9687
0.6 0.4
0.5 0.5
0.4 0.6
0.3 0.7
0.2 0.8
0,1 0.9
0.2091 0,6947
0.3178 0.8192
0.4421 0.9136
0.5775 0.9744
0,7191 0.9994
0.8617 0.9877
1.0000 0.9397
0,5
0.4
0.3
0.2
0.1
0.0
0.5
0.6
0.7
0.8
0.9
1.0
0.4726 0.9403
0.6040 0.9893
0.7390 0.9985
0.8726 0.9675
1,0000 0.8974
0.3
0.2
0.1
0.0
0.2
0.9
1.0
0.83720.9999
1.00000.9871
0.1
0,0
0.9
1.0 0.0 1,0
%i i i ! 231
l ITI l lTl
lEili!El liELi!Et
-m-
l=i=l
rl
60
50 0,3064

0.00.000o 10.0000 1.0
0.1 0,0436 0.2422 0.9
0,2 0.1087 0,4628 0.8
0.3 0,1928 0.6534 0.7
0.4 0.2929 0.8073 0.6
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.0000 0.000o
0.0603 0.2861
0.1390 0.5347
0.2333 0.7368
0.3396 0.8850
0.4541 0.9737
0.5726 0.9997
0.6906 0.9622
0.8039 0.8623
0.9083 0.7039
1.0000 0.4927
1.0
0.9
0.8
0.7
0.6
0.5
0,4
0.3
0.2
0.1
0.0
0.7
0.8
0.9
1.0
I
i!
45
12.7944 155
i i RLETTiiiLEE- m i i i i i l a-
.
TABLEAl .H~ AS A FUNCTION OF 6{ (ConL)
Ax, = 110
-- -i~W-
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0,9
1.0
0.0000 0.0000
0.0760 0.3418
0.1676 0.6217
0.2713 0.8292
0.3835 0.9569
0.5000 1.0000
0.6165 0.9569
0.7287 0.8292
0.8324 0.6217
0.9240 0.3418
1.0000 0.0000
1.0
:::
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
i
.
35
3.8420 145
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
O. omo0.8379
0.1175 0.9341
0.2430 0.9382
0.3720 0.9983
0.4997 0.9641
0.6214 0.8866
0.7326 0.7689
0.8293 0.6152
0.9079 0.4311
0.9656 0.2235
l.omo O.0000
:::
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-Giiiizd-
=i==l
0.0
0.1
0.2
U
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.8974
0.1274 0.9675
0.2610 0.9985
0.3960 0.9893
0.5274 0.9403
0.6503 0.8532
0.7603 0.7313
0.8533 0.5790
0.9259 0.4019
0.9754 0.2065
1,oo1110.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
m m
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.s
0.9
1.0
I=i=l
0.000o 0.4927
0.0917 0.7039
0.1961 0.8623
0.3094 0.9622
0.4274 0.9997
0.5459 0.9737
0.6604 0.8850
0.7667 0.7368
0.8610 0.5347
0.9397 0.2861
I.0000 O.0000
1,0 0.0
0.9 0.1
0.8 0.2
0.7 0.3
0.6 0.4
0.5 0.5
0.4 0.6
0.3 0.7
0.2 0.8
0.1 0.9
0.0 1.0
-m
O.0000 0.6444
0.0998 0.8094
0.2109 0.9256
0.3292 0.9887
0.4503 0.9963
0.5698 0.9482
0.6832 0.8462
0.7865 0.6940
0.8758 0.4971
0.9478 0.2630
1.0000 0.0000
:::
0.8
0.7
0,6
0.5
0.4
0,3
0,2
0.1
0.0
-m
12.4051/130 I
.-.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.9397
0.1383 0.9877
0.2809 0.9994
0.4225 0.9744
0.5579 0.9136
0.6822 0.8192
0,7909 0.6947
0.8798 0.5446
0.9457 0.3746
0.9863 0.1908
1.0000 0.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
m
-m
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.7560
0.1084 0.8834
0.2265 0.9654
0.3499 0.9992
0.4742 0.9834
0.5947 0.9186
0.7071 0. %373
0.8072 0.6534
0.8913 0.4628
0.9564 0.2422
1.00000,0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
liE!LiEl
-l=i=l

o

1.1579 1;:
lLL!
=m=
1.0000 100
AX; = 120
TABLE A. I.Hk AS A FUNCTION OF @i ( Cont . )
1
95
25 0.7642

0.0027 0.0000
0.0052 0.1923
0.0383 0.3796
0.1004 0.5539
0.1888 0.7075
0.2997 0.8336
0,4283 0.9269
0.5688 0.9831
0.7152 0.9999
0.8611 0.9764
1.0000 0.9138
.1=.
-El
T%Lid
lll
0.0
0.1
0.2
:::
0,5
0.6
0.7
0.8
0.9
1.0
0.0 0.0113 0.0000 1.0
0.1 0.0005 0.1777 0.9
0.2 0.0221 0.3541 0.8
0.3 0.0754 0.5215 0.7
0.4 0.1580 0.6725 0.6
0.5 0.2662 0.8007 0.5
0.6 0.3953 0.9003 0.4
0.7 0.5398 0.9670 0.3
0.8 0.6932 0.9979 0.2
0.9 0.8488 0.9917 0,1
1.0 1.000o 0.9486 0.0
0.0
0.1
0.2
0.3
0.0000 0.0000
0.0146 0.2079
0.0576 0.4067
0.1273 0.5878
0.2206 0.7431
0.3333 0.8660
0.4606 0.9511
0.5970 0.9945
0.7364 0.9945
0.8727 0.9511
1.0000 0.8660
1.0
0.9
0.8
0.7
0.0
0.1
0.2
0.3
0.4
0.0271 0.000o
0.0011 0.1640
0.0098 0.3299
0.0528 0.4903
0.1282 0.6383
0.2327 0.7673
0.3617 0.8718
0.5097 0.9472
0.6701 0.9902
0.8360 0.9989
1.000o 0.9729
RJ
0.8
0.7
0.6
1.0
0.9
0.8
0.7
1.0 0.000o 0.Oooo 1.0
0.1 0.0254 0.2248 0.9
0.2 0.0774 0.4356 0.8
0.3 0.1538 0.6232 0.7
0.4 0.2512 0.7794 0.6 0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.6
0.5
0.4
0.3
0,2
0.1
0.0
0.5
0.6
0.7
0..8
0.9
1.0
0.5
0.4
0.3
0.2
0.1
0.0
0.5 0.3654 0.8974 0.5
0.6 0.4913 0.9719 0.4
0.7 0.6235 0.9998 0.3
0.8 0.7562 0.9799 0.2
0.9 0.8835 0.9129 0.1
1.0 1.0000 0.8019 0.0
m iIJiIEJ m IEiiLEEl m
- F3=l El - El - l =E=l
Ri 80
40 0.6077
_ .
0.0
0,1
0.2
0.3
0.4
0.5
I!)::
0.8
0.7
0.6
0.0 0.0000 O.ocml 1.0
0.1 0.0448 0.2633 0.9
0.2 0.1128 0.4999 0.8
0.3 0.2012 0.6993 0.7
0.4 0.3060 0.8530
0.0
0.1
0,2
0,3
0.0000 0.0000
0.0537 0.2856
0.1292 0.5359
0.2231 0.7402
0.3313 0.8893
0.4491 0.9769
0.5714 0.9991
1.0 0.0
0.9 0.1
0.8 0.2
0.7 0.3
0.6 0.4
0.0000 0.0000
0.0624 0.3104
0.1450 0.5752
0.2443 0.7828
0.3558 0.9243
0.4747 0.9934
0.5959 0.9871
0.7139 0.9056
1.0
0.9
0.8
0.7
0.6
0.0
0.1
0.2
0.3
0,4
0.5
0.6
0.Oooo 0
0.0710 0
0.1606 0
0.2652 0
0,3800 0
0.5000 1
0.6200 0
0,7348 0
.Oooo
.3383
.6180
.8271
.9563
.Oooo
.9563
,8271
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.4
0,5
0,6
0.7
0.8
0.9
1,0
0.5
0.4
0.3
0,2
0.1
0.0
0.5
0.6
0.7
0,8
0.9
1.0
0.4226 0.9540
0.5461 0.9982
0.6708 0.9834
0.7916 0.9104
0.9029 0.7823
1.0000 0.6048
0.5 0.5
0.4 0.6
0.3 0.7
0.2 0.8
0,1 0.9
0.0 1.0
0.5
0.4
0.3
0.6
0.7
0.8
0.9
1.0
0.6927 0.9549
0.8079 0.8462
0.9118 0.6778
1.0000 0.4571
0.7
0.823810.7526 0.2
0.920.0.5347 0.1
1.00000.2615 0.0
0,8 0.8394 0.6180 0.2
0.9 0.9290 0,3383 0.1
1.0 1.0000 0,0000 0.0
lizii2El
1 I
40
1 1 1 I
35
2.5924 160 2.9884 155
-l a ! (
..-. . .
TABLE A.1.Hk AS A FUNCTION OF I% (Cont.)
Ax< ==m
1
i% 60
60 0.2887
.
0.0 0.0000 0.0000 1.0
0.1 0.0710 0.3383 0.9
0.2 0.1606 0.6180 0.8
0.3 0.2652 0.8271 0.7
0.4 0.3800 0.9563 0.6
0.5 0.5000 1.ooOo 0.5
0.6 0.6200 0.9563 0.4
0.7 0.7348 0.8271 0.3
0.8 0.8394 0.6180 0.2
0.9 0.9290 0.3383 0.1
1.0 1.0000 0.0000 0.0
-LEiEig
mm=l
0.0
0.1
0.2
:::
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.8019
0.1165 0.9129
0.2438 0.9799
0.3765 0.9998
0.5087 0.9719
0.6346 0.8974
0.7488 0.7794
0.8462 0.6232
0.9226 0.4356
0.9746 0.2248
1.0000 o.oocKl
u
0.8
0.7
0.6
:::
0.3
0.2
:::
-m
- mEl -
1
0.00.0000 0.2615 1.0
0.1 0.0795 0.5347 0.9
0.2 0.1762 0.7526 0.8
0.3 0.2861 0,9056 0.7
0.4 0.4041 0.9871 0.6
0.5 0.5253 0.9934 0.5
0.6 0.6442 0.9243 0.4
0.7 0.7557 0.7828 0.3
0.8 0.8550 0.5752 0.2
0.9 0.9376 0.3104 0.1
1.0 1.0000 O.ooilo 0.0

25
2.9884 145
_m.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.8660
0.1273 0.9511
0.2636 0.9945
0.4030 0.9945
0.5394 0.9511
0.6667 0.8660
0.7794 0.7431
0.8727 0.5878
0.9424 0.4067
0.9854 0.2079
1.0000 0.0000
;::
0,8
0.7
0.6
:::
0.3
0.2
0.1
0.0
0.00.0000 0.4571
0.1 0.0882 0.6778
0.2 0.1921 0.8462
0.3 0.3073 0.9549
0.4 0.4286 0.9991
0.5 0.5509 0.9769
0.6 0. 66&70.8893
0.7 0.7769 0.7402
0.8 0.8708 0.5359
0.9 0.9463 0.2856
1.0 1.00000.000o
1 1
20
2.5924 140 I
::!
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
El
- I i zi t i zl IGL!ZEI
43=4 - El
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.0000 0.6048
0.0971 0.7823
0.2084 0.9104
0.3292 0.9834
0.4539 0.9982
0.5774 0.9540
0.6940 0.8530
0.7988 0.6993
0.8872 0.4999
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.9 0.95520.2633 0.1
1.0 1.00000.0000 0.0
lEaBEl
El
0.5 0.7338 0.8607 0.5
0.6 0.8420 0.6725 0.4
0.7 0.9246 0.5215 0.3
0.8 0.9779 0.3541 0.2
0.9 0.9995 0,1777 0.1
1.0 0,9887 0.0000 0,0
l a
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0,7
0.8
0,9
1.0
0.0000 0.7169
0.1065 0.8584
0.2256 0.9531
0.3521 0.9970
0.4804 0.9882
0.6051 0.9270
0.7205 0.8161
0.8217 0.6604
0.9043 0.4666
0.96460.2432
1.00000.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0,2
0.1
0.0
-m
0.0
0.1
0,2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.9729
0.1640 0.9989
0.3299 0.9902
0.4903 0.9472
0.6383 0,8718
0.7673 0.7673
0.8718 0.6383
0.9472 0.4903
0.9902 0.3299
0.9989 0.1640
0.9729 0.000o
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
.. -. .-
TABLE A.1.Hk AS A FUNCTION OF & (Cod.)
AXi = 130
l1 l
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.04490.0000
0.0055 0.1640
0.0041 0.3327
0.0406 0.4975
0.1132 0.6497
0.21820.7818
0.350310.8868
o.5025~o.9594
0.667310.9959
0.83600.9945
1.mlo.9551
1.0
0.9
0,8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0,0
-M-
=
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.0000
0.0265 0.2430
0.0816 0.4687
0.1625 0.6655
0.2651 0.8233
0.3841 0.9339
0.5133 0.9918
0.6462 0.9940
0.7760 0,9403
0.8960 0.8335
1.0000 0.6791
1.0
0,9
0.8
0.7
0.6
0.5
0,4
0,3
0.2
0,1
0.0
- El
0.00.0239~0.0000 1.0
0.1 0.0004,0.1779 0.9
0.2 0.0129,0.3572 0.8
0.3 0.06080.5287 0.7
0.4 0.1415O.683710.6
0.5 0,2511 0.8141 0.5
0.6 0.3838 0.9134 0.4
0.7 0.5329 0.9764 0.3
0.8 0.6907 0.9999 0.2
0.9 0.8491 0.9826 0.1
1.0 1.OOOCIO.9256 0,0
-ldld-
l=E=
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0,8
0.9
1.0
o.oocdo. Oooo
0.03670.2627
0.10040.5012
0,18780.7032
0.294310.8584
0 .4146~0,9588
10.5425 0 9993
0.6715 0.9778
0,7948 0.8954
0.9062 0.7563
1.0000 0.5677
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0,1
0.0
IiiiEEl--l=-
100
300.7824
_ _
0,0
0.1
0.2
0.3
0.4
0.5
0,6
0.7
0.8
0.9
1.0
0.0101 0.000o
0.0009 0.1926
0.0258 0.3828
0.0836 0.5610
0.1712 0.7181
0.2843 0.8459
0.4169 0.9380
0.5624 0.9895
0.7132 0.9979
0.8616 0.9628
II
1.00000.8858
1.0
0.9
0.8
0,7
0.6
0.5
0.4
0.3
0.2
0,1
0.0
nzi ! I Er
75
55 0.4152

0.00.000o 0.000o 1.0

0.1 0,0465 0.2842 0.9
0.2 0.1183 0.5360 0.8
0.3 0.2118 0.7423 0.7
0.4 0.3222 0.8927 0.6
0.5 0.4438 0.9795 0.5
0.6 0.5704 0,9982 0.4
0.7 0.6955 0,9477 0.3
0.8 0.8127 0.8308 0.2
0.9 0.9160 0.6534 0.1
1.0 1.0000 0.4247 0.0
.L1
35
2.4084 165
95
I
35 0.6909
I
m

0.0
0.1
0.2
0.3
0.4
0.5
0.6
0,7
0.8
0.9
1.0
0. @32410.0000
0,0062 0.2082
0.0422 0.4098
0,1086 0.5945
0.2021 0.7529
0.3178 0.8768
0.4497 0.9598
0.5912 0.9978
0.7349 0.9887
0.8736 0.9330
1.0000 0.8337
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0,2
0.1
0.0
iczL!lE
- l w
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.0000 0,0000
0.0559 0.3079
0.1357 0.5735
0.2352 0.7829
0.3493 0.9256
0.4721 0.9942
0.5975 0.9852
0,7189 0.8991
1.0
0,9
0,8
0.7
0.6
0.5
0.4
0.3
0.8 0.8301 0.7402 0.2
0.9 0.9254 0.5167 0.1
1.0 1.0000 0.2401 0.0
m
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0.0000
0.0156 0.2250
0.0616 0.4384
0.1356 0.6293
0.2340 0,7880
0.3515 0.9063
0.4822 0.9782
0.6193 0.9998
0.7560 0.9703
0.8851 0.8910
1.0000 0.7660
1.0
0.9
0.8
0,7
0.6
0.5
0.4
0,3
0.2
0.1
0.0
l a
-l =
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0,oooo 0.
0,0653 0.
0.1528 0.
0.2582 0.
0.3759 0.
0.5000 1,
0.6241 0.
0.7418 0.
Oooo
3343
6140
8247
9556
0000
9556
8247
1.0
0.9
0.8
0,7
0.6
0.5
0.4
0.3
0.8 0.8472 0.6140 0.2
0.9 0.9347 0,3343 0,1
1.0 1.0000 0.0000 0.0
25
3.1394 155
t
Axi = 130
TABLE A. I.H, AS A FUNCTION OF Oi (Con. !.)
-m - I =l =am
=Ea
0.0 0.0000 0
0.1040 0
0.2240 0
0,3538 0
0.4867 0
0.6159 0
0.7349 0
0.8375 0
.6791
.8335
. 9403
.9940
,9918
. 9339
.8233
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.0
0.1
0,2
0.3
1.0
0.9
0.8
0.7
0.0
0.1
0.2
0.3
0.4
0.0000 0.5677
0.0938 0.7563
0.2052 0.8954
0.3285 0.9778
0,4575 0,9993
0.5854,0,9588
1.0
0,9
0.8
0.7
0.6
0.5
0.4
0.1
0.2
0.3
0.4
0.5
0,4
0.5
0.6
0.7
0.8
0.9
0.4025 0.9852
0.5279 0.9942
0.6507 0,9256
0.76480.7829
0.86430.5735
0.944110.3079
1.0000,0.0000
0.6
0,5
0.4
0,3
0.2
0.1
0,5
0.6
0,7
0.8
0.9
1.0
10.705710,858.i 0.6
0.7 0.81220.7032
0.899610 5012
10.9633~0:2627
,1.0000,0.0000
0.3
0.2
0.1
0.0
.66551
0.8 0.9184 0.4687 0.2
0.9 0.9735 0.2430 0.1
1.0 1.0000 0.0000 0.0 1.0 0.0
-ii!iLi3 iGiLm
=-
l a
m
-l=Y=lEl -
.ZEiiL
0.0 0.000o 0.9551 1.0
0.1 0.1640 0.9945 0.9
0.2 0.3327 0.9959 0.8
0.3 0.4975 0.9594 0,7
0.4 0.6497 0.8868 0.6
0.5 0,7818 0.7818 0.5
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.1
0.2
0.3
;::
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.0000 0.8858
0.1384 0.9628
0.2868 0.9979
0.4376 0.9895
0.5831 0.9380
0.7157 0.8459
0.8288 0.7181
0.9164 0.5610
0.9742 0.3828
0.9991 0.1926
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.1
0.2
0.3
0.4
0.5
0,6
0.7
0.8
0,9
1,0
0.0000 0.9256
0.1509 0.9826
0.3093 0.9999
0.4671 0.9764
0.6162 0.9134
0.7489 0.8141
0.8585 0.6837
0.9392 0.5287
0.9871 0.3572
0.9996 0.1779
0.9761 0.000o
1.0
0,9
0.8
0.7
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.6 0,8868 0.6497 0.4
0.7 0.9594 0.4975 0.3
0.8 0,9959 0.3327 0.2
0.9 0.9945 0.1640 0.1
1.0 0.9551 0.0000 0.0
0.984410.2250
1.00000.0000
0.1
0.0 0.0 1.0 10.989910.00ti l 0.0
lo
1.2781 120
I
I d
I I I
5
1.4474 125
(
15
1.1301 115
1 I
20=
1.0000 110
L.,.
Ax, = 14W
Tom A.1.Hb AS A FUNCTION OF Oi (Cont.)
w
N
IQ
-=F= -l=F=l
0.0
0.1
0.2
0.3
0.4
0,5
0.6
0.0402 0.000o
0.0037 0,1770
0.0065 0.3586
0.0485 0.5340
0.1273 0.6928
0.2381 0.8257
0.3744 0.9246
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.00.0092 0,0000
0.0015 0.2074
0.0298 0.4113
0.0925 0.5995
0.1859 0.7609
0.3044 0.8859
0.4409 0.9670
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
:::
0.8
0.7
0.6
0.5
0.4
0.0
0.1
0.2
0,3
0.4
0.5
0.6
0,0217 0.000o
0.0001 0.1917
0.0163 0.3843
0.0693 0.5663
0.1559 0.7268
0.2710 0.8563
0.4078 0.9472
1.0 0.0
0.1
0.2
0,3
0.4
0.5
0.6
0.7
0.0022 0.0000
0.0072 0.2241
0.0466 0.4396
0.1180 0.6337
0.2171 0.7950
0.3382 0.9138
0.4740 0.9831
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.9
0.8
0.7
0.6
0.5
0.1
0.2
0.3
0,4
0.5
0.6 0.4
0.3
0.2
0.1
0.0
0.4974 0.9693
0.6661 0.9990
0.8371 0.9871
1.00000.9341
0.3
0.2
0.1
0.0
0.7
0.8
0,9
1.0
0.5281 0.9837
0.6899 0.9995
0.8504 0.9711
1.00000,9002
0.3
0.2
0.1
0.0
0.7
0.8
0.9
1.0
0.5580 o.994i
0.7129 0.9941
0.8632 0,9472
1.00000,8563
0.7
0.8
0.9
1,0
0.5875 0.9995
0.7353 0.9814
0.8756 0.9138
1.00000.8007
0.6164 0.9987
0.7571 0.9598
0.8875 0.8686
1,00000.7306
0.3
0.2
0.1
0.0
0.8
0.9
1.0
m
3X==
E!IJzE
E
m
ml
1 1 !
50
1.3997 190
1 1
450
1.5703 185
.m.
-m
0.0
0,1
0.2
0.0000 0.00043
0.0488 0.3049
0.1253 0.5710
0.2250 0,7823
1.0
0,9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.00
0.1 0
0.2 0
0.3 0
0.4 0
0.5 0
0.6 0
0.7 0
0.8 0
0.9 0
1.0 1
1.0
0.9
0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.0000 0.000o
0.0280 0.2612
0.0868 0.5012
0.1731 0.7057
0,2817 0.8626
0.4062 0.9627
0,5392 0.9998
0.6727 0.9719
0.7989 0.8806
0.9102 0,7313
1.0000 0.5329
1.0
0.9
0.8
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.000o 0,0000
0.0385 0,2821
0,1063 0.5349
9.1994 0.7435
0.3123 0.8954
0.4383 0.9816
0,5698 0,9970
0,6990 0.9408
0.8184 0.8161
0.9207 0.6305
1.0
0.9
0,8
0.7
0.6
0.5
0.4
0,3
0.2
0.7
0.6
0.5
0.4
0,3
0.2
0.7
0.6
0.5
0.4
0,3
0.2
0.1
0,0
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.4 0.3713 0.9549 0.6
0.5 0.5000 1.0000 0.5
0.6 0.6287 0.9549 0.4
0.7 0.7498 0.8221 0.3
0.8 0.8560 0,6096 0.2
0.9 0.9411 0.3301 0.1
1.0 1.0000 0.0000 0.0
0.1
0.0
0.9
1.0
0,1
0.0 11.Oooo/o.39491 1.0 11.000 do.22101
1
25
2.5260 165
-la
T
2.8563 160
l a l a
.. ,,,. ,, ,7,7--T, , ,,, ,,, ,, , ,,, ,,, . .,,,,,,,., .,, ,,,, ,.,, ,,,, ,,,, ,,, ,,,, .,,,,, ,.. ,
.- ..
TABLE A.1.Hk AZ A FUNCTION OF e, (ConL)
Axi = 140
lma
.Emid
El -=@=l-
0.0000 0.3949
0.0793 0.6305
0.1816 0,8161
0.3010 0.9408
0.4302 0.9970
0,5617 0.9816
0.6877 0.8954
0.8006 0.7435
0.8937 0.5349
0.9615 0.2821
1.0000 0.0000
0,0
0.1
0.2
0,3
0,4
1.0
0.9
0.8
0.7
0.6
0.5
0.0
0.1
0,2
0,3
0.4
0.5
0.0000 0.5329
0.0898 0.7313
0.2011 0.8806
0.3273 0.9719
0.4608 0.9998
0.5938 0.9627
0.7183 0.8626
0.8269 0.7057
0,9132 0.5012
0,9720 0.2612
1.0000 0,0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0,3
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.0000 0.6428
0.1009 0.8090
0.2216 0.9272
0.3550 0.9903
0.4930 0.9945
0.6274 0.9397
0.7504 0.8290
0.8546 0.6691
0.9337 0.4695
1,0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0,5
0.6
0.7
0.8
0.9
1.0
0,4
0,3
0.2
0,1
0.0
0.6
0.7
0.8
0.9
1.0
0,2
0.1
0.0
0.9
1.0
o.98320.24i9
1.00000.0000
I I I
1.9887 14$
m-
. _ .
15
2.5260 155
IEJzz l a
IGELE
- l w
m==l.
0.0
0.1
0.2
0.3
0,4
0.5
0.6
0.7
0.8
0.9
1.0
0.0000 0. 73Q6
0.1125 0.8686
0.2429 0.9598
0.3836 0.9987
0.5260 0.9831
0.6618 0.9138
0.7829 0.7950
0.8820 0.6337
0.9534 0.4396
0.9928 0.2241
0.9978 0.0000
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0
0.1
0.2
0,3
0.4
0.5
0.6
0.000o 0.9341
0.1629 0.9871
0.3339 0.9990
0.5026 0.9693
0.6591 0.8996
0.7941 0.7941
0.8996 0,6591
0.9693 0.5026
0.9990 0.3339
1,0
0.9
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.000o 0.8007
0.1244 0.9138
0.2647 0.9814
0.4125 0.9995
0.5591 0,9670
0.6956 0.8859
0.8141 0,7609
0.9075 0.5995
0.9702 0,4113
u
0.8
0.7
:::
0.4
0,3
0.2
0.0 0.00000.8563
0.1 0.13680.9472
0.2 0.28710.9941
1,0 0.0
0.9 0.1
0,8 0.2
0,oooo 10.9002
0.1496 0,9711
0.3101 0.9995
0,4719 0.9837
0.6256 0.9246
1.0
0.9
0.8
0.7
0.6
0.5
0.s
0,7
0.6
0.5
0.4
0.3
0.3
0,4
0.5
0,6
0.7
0.8
0.9
0,4420 0.9941
0.5922 0.9472
0.7290 0.8563
0.8441 0.7268
0.9307 0.5663
0.9837 0.3843
0.9999,0,1917
0,7 0.3
0.6 0.4
0.5 0.5
0.4 0.6
0,3 0.7
0,2 0.8
0.1 0.9
0.0 1.0
0.7619 0.8257
0.8727 0.6928
0,9515 0.5340
0.9935 0.3586
0.9963 0,1770
0. 9598i0.0000
0.4
0.3
0.2
0.1
0.0
0,7
0.8
0.9
1.0
0,2
0.1
0.0
0.99850.2074
0.99080.0000
0.1
0.0
0.98710.1629
0.93410.0000
1.0 10.978310.OWIOI
m
1 I I
25
1,0000 115
324 APPENDI X A
TABLE A.2.t), AS A FUNCTION OF Hk
AX; = 40
H,
x;. = 90 85 80 75 70 65 60
550 5o0 I 45
X,,u= 50 45 40 35 30 25 20 15 10 5
. . - . .
0.0 0.0000 0,0000 0,0000 0.0000 0.0000 0.000o o.cWo 0.0000 0.0000 0.0000
0,1 0.3104 0.2424 0.2021 0.1769 0.1600 0.1478 0.1387 0.1315 0.1256 0.1206
0.2 0.4399 0.3805 0.3398 0.3111 0.2899 0.2736 0.2607 0.2500 0.2410 0.2332
0,3 0.5398 0.4892 0.4523 0.4245 0.4030 0.3858 0.3715 0.3595 0.3490 0.3398
0.4 0.6246 0.5822 0.5501 0.5251 0.5050 0.4884 0.4743 0.4622 0.4514 0.4417
0.5 0.6998 0.6652 0.6383 0.6167 0.5990 0.5840 0.5710 0.5596 0.5494 0.5400
0,6 0,7682 0.7410 0.7194 0.7018 0.6870 0.6742 0.6630 0.6530 0.6439 0.6355
0.7 0.8314 0.8114 0.7952 0.7817 0.7702 0.7602 0.7513 0.7432 0.7357 0.7287
0.8 0.8906 0.8775 0.8667 0.8576 0.8497 0.8428 0.8365 0.8308 0.8254 0.8202
0.9 0.9467 0.9402 0.9348 0,9302 0.9262 0.9225 0.9192 0.9162 0.9133 0.9106
1.0 1.0000 1,0000 I .00001,00001.0000 1.0000 1.0000 1.0000 1.0000 1.0000
II
J
x,, = 40 35 30 25 20 15 10
Hk X,M = (-J. ~.
; 0 5
10 15 20 25 30 35 40 45

0.0 0.0000 0.0000 0.0000 o.m,o. oOOoO.00000.0000

01 0,1164 0,1126 0.1091 0.1060,0.1030 0.1002 0.0975 0.0948 0.0921 0.089
0.2 0.2264 0.2201 0.2144 0.2090,0.2040 0.1990 0.1942 0.1894 0.1846 0.179
03 0.3315 0.3238 0.3167 0.3099 0.3034 0.2970 0.2908 0.2844 0.2780 0.271
0.4 0.4328 0.4246 0.4168 0.4093 0.4020 0.3947 0.3874 0.3801 0.3725 0.364
0.5 0.5313 0.5231 0.5152 0.5076 0.5000 0.4924 0.4848 0.4769 0.4687 0.46
0.6 0.6275 0.6199 0.6126 0.6053 0.5930 0.5907 0.5832 0.5754 0.5672 0.558
0.7 0.7220 0.7156 0.7092 0.7030,0.6966 0.6901 0.6833 0.6762 0.6685 0.660
0.8 0,8106 0.8058 0.8010,0.7960 0.7910 0.7856 0.7799 0.7736 0.766
X,m= 10 15 20
X,M = 500
25 30 35 40 45 50
55 60 65 70 75 80 85 90

0.0 0.0000 0.0000 0.ooOo 0.0000 0.0000 o.OOOqo. m,o. Ooooo. Oooo
0.1 0.0867 0.083810.080so. 07750.0738 0.0698 0.0652,0. 0598;0.0533
0.2 0,1746 0.1692 0.1635 0.1572 0.1503 0.1424 0.1333 0.1225 0.1094
0.3 0.2643 0.2568 0.2487 0.2398 0.2298 0.2183 0.2048 0.1886 0.1686
0.4 0.3561 0,3470 0.3370 0.3258,0.3130 0.2982 0.2806 0.2590 0.2318
0.5 0.4506 0.4404 0.4290 0.4160 0.4010 0.3833 0.3617 0.3348 0. 3002
0,6 0.5486 0.5378 0.5257 0.5116 0.4950 0.4749 0.4499 0.4178 0. 3754
0.7 0.6510 0.6405 0.6285 0.6142 0.5970 0.5755 0.5477 0.51O80.4602
0.8 0.7590 0.7500 0.7393 0.7264 0.7101 0.6889 0.6602 0.6195 0.5601
0.9 0.8744 0.8685 0.8613 0.8522 0.8400 0.8231 0.7979 0.7576 0.6896
1.0 1.0000 1.ooOo 1.0000 1.0000 1. Ooooll. ooOo l.oOOoll. ooWll. ooOo
d
APPENDI X A 32:
Tom A.2.@i AS A FUNCTION OF H~ (Cont.)
Ax, = 50
Hk
x,.= 90 85 80 75 70 65 60 55 50
Xill = 40 35 30 25 20 15 10 5 o
. . . . . .
0.0 0.0000 0.0000 0.0000 0.0000 0.oOOO0.0000 0.0000 0.0000 o.oOoc
0.1 0.3072 0.2492 0.2111 0.1855 0.1673 0.1539 0.1435 0.135210.1283
0.2 0.4358 0.3853 0.3479 0.3198 0.2980 0.2806 0.2664 0.2544 0.2441
0.3 0.5354 0.4922 0.4585 0.4318 0.4100 0,3919 0.3766 0.3632 0.3515
0.4 0.6201 0.5839 0.5546 0.5306 0.5105 0.493210.4782 0.4648 0.4527
0.5 0.6955 0.6658 0.6413 0.6206 0.6029 0.5874 0.5736 0.5611 0,5496
0.6 0.7643 0.7409 0.7212 0.7043 0.6895 0.6763 0.6644 0.6534 0.6431
0.7 0.8283 0.8109 0.7960 0.7831 0.7716 0.7612 0.7516 0.7427 0.7343
0.8 0.8884 0.8769 0.8669 0.8582 0.8503 0.8430 0.8363 0.8339 0.8237
0.9 0.9455 0.9397 0.9347 0.9303 0.9262 0.9224 0.9L89 0.9154 0.9121
1.0 1.00Q0 1.0000 1.0000 l.cKloo 1.0000 1.0000 1.0000 1.0000 1.0000
0.0
0.1
0.2
0.3
0,4
0.5
0.6
0,7
0.8
0.9
1.0
[
H,
0.0
0.1
0.2
0,3
0.4
0.5
0.6
0,7
0.8
0.9
1.0
Xim = 45 40 35 30 25 20 15 10 5
X,,u = 5 10 15 20 25 30 35 40 45
11l l l l l1~
O.0000
0.1224
0.2350
0.3408
0,4416
0.5388
0.6334
0,7262
0.8178
0.9089
1.0000
0.0000 0.0000 0.0000 0.0000 o.OOOIJo. oOoco, oood o.
1
0.1173 0.1127 0.1086 0.1048 0.1012 0. 0978~0.09440.0911
0.2269 0.2195 0.2127 0.2062 0.2000 0.1940 0.1881 0.182
0.3311 0.3221 0.3135 0.3054 0.2974 0.2896 0.2818 0.273
0.4313 0.4215 0.4122 0.4030 0,3940 0.3850 0.3759 0.366
0.5287 0.5189 0.5094 0.5~10,490610.4811 0.4713 0.461
0.6241 0.6150 0.6060 0.59700.58780.578510. 56870.55
0.7182 0.7104 0.7026 0.694610.686510.67790.6689 0.659
0.8119 0.8060 0.8000 0.7938 0.7873 0.7805 0.7731 0.765
0,9056 0.9022 0,8988 0.8952 0.8914 0.8873 0.8827 0.877
1.0000 I.cnoo 1.0000 1.0000 I.0000 1.0000 1.0000 1,
1
Xim= o 5 10 15 20 25 30
350
j
40
Xw = 50 55 60 65 70 75 80 85 90
. .
0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0
0.0879 0.0846 0.0811 0.0776 0.0738 0.0697 0.0653 0,0603 0.0545
0.1763 0.1661 0.1637 0.1570 0.1497 0.1418 0.1331 0.1231 0.1116
0.2657 0.2573 0.2484 0.2388 0,2284 0.2169 0.2040 0.1891 0.171
0.3569 0.3466 0.3356 0.3237 0,3105 0.2957 0.2788 0.2591 0,235
0.4504 0.4389 0,4264 0.4126 0.3971 0.3794 0.3587 0.3342 0.3045
0.5473 0.5352 0.5218 0.5068 0.4895 0.4694 0.4454 0.4161 0.3799
0.6485 0.6368 0,6234 0,6081 0.5900 0.5682 0,5415 0.5078 0.4646
0.7559 0.7456 0.7336 0.7194 0,7020 0.6802 0.6521 0.6147 0.5642
0.8717 0.8648 0.8565 0.8461 0.8327 0.8145 0.7889 0.7508 0.692
1.0000 1.0000 I.0000 1:0000 1.0000 1.0000 1.0000 1.0000 1.
326 APPENDI X A
TABLE A.2.o; AS A FUNCTION OF H, (Cont.)
AXi = 60
b
H~
Xirn = 90 85 80 75 70 65 -60 55 50
X;M = 30 25 20 15 10 5 0 5 10
.
h14-
0.0 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.1 0.3032 0.2524 0.2166 0.1910 0.1722 0.1579 0.1465 0.1373 0.1295
0.2 0.4307 0.3863 0.3517 0.3243 0.3024 0.2843 0.2691 0.2561 0.2447
0.3 0.5298 0.4917 0.4605 0.4347 0.4129 0.3943 0.37s0 0.3637 0.3507
0.4 0.6145 0.5823 0.5551 0.5319 0.5118 0.4941 0.4782 0.4638 0.4506
0.5 0.6902 0.6635 0.6406 0.6206 0.6029 0.5869 0.5723 0.5588 0.5462
0.6 0.7595 0.7384 0.7199 0.7034 0.6885 0.6749 0.6622 0.6503 0.6389
0.7 0.8243 0.8085 0.7944 0.7817 0.7701 0.7592 0.7490 0.7392 0.7297
0.8 0.8855 0.8750 0,8655 0.8568 0.8487 0.8411 0.8338 0.8267 0,8197
0.9 0.9439 0.9386 0.9338 0.9293 0.9251 0.9211 0:9172 0.9133 0.9095
1,0 1.0000 1.0000 1,0000 1.0000,1,0000 I.0000 1.0000 1.000o 1.0000
Xi~ = 45 40 -35 -30 /-25 -20 -15 -10
Hk
Xi.11= 15 20 25 30 35 40 45 50 55
0.0 0.0000 0.0000 0.0000 0.000o 0.0000 0.0000 0.0000 0.0000 0.
0.1 0.1229 0.1170 0.1118 0.1070 0.1026 0.098$0.09440.0905 0.0867
0.2 0.2346 0.2254 0.2169 0.2090,0.2015 0.19430.187310.1803 01733
0.3 0.3389 0.3279 0.3176 0.3077 0.2982 0.2888~0.279610.27030.2608
0.4 0.4382 0.4265 0.4153 0.4044 0.3936 0.3829 0.3721 0.3611 0.3497
0.5 0.5341 0.5225 0.5112 0.5000 0.4888 0.4775 0.4659 0.4538 0.4412
0.6 0.6279 0.6171 0.6064 0.5957 0.5847 0.5735 0.5618 0:5494 0.5362
0.7 0.7204 0.7112 0.7018 0.6923 0.6824 0.6721 0.6611 0.6493 0.6363
0.8 0.8127 0.8057 0.7985 0.7910 0.7831 0.7746 0.7654 0.7553 0.7439
0.9 0.9056 0.9016 0.8974 0.8930 0.8882 0.88300,8771 0.87050.8627
1.0 I.0000 1.0000 1.0000 1.0000 I. OOOcll.oooo 1.0000 1.000@sloo
Hk
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
X,m= 0 5 10 15 20 25 30
Xi~ = 60 65 70 75 80 85 90
0.0000 0.0000 0.0000 0.000o 0.0000 0.0000 0.0000
0.0828 0.0789 0.0749 0.0707 0.0662 0.0614 0.0561
0.1662 0.1589 0.1513 0.1432 0.1345 0.1250 0.1145
0.2510 0,2408 0.2299 0.2183 0.2056 0.13150.1757
0.3378 0.3251 0.3115 0.2966 0.2801 0.2616 0.2405
0.4277 0.4131 0.3971 0.3794 0.3594 0.3365 0.3098
0.5218 0.5059 0.4882 0.4681 0.4449 0.4177 0.3855
0.6220 0.6057 0.5871 0.5653 0.53950.5083 0.4702
0,7309 0.7157 0.6976 0.6757 0.648310.61370.5693
0.8535 0.8421 0.8278 0.8090 0.783410.74760.6968
1.0000 I.OQOOI.0000 I.0000 1.00001.0000 1.0000
APPENDI X A 327
TABLE A.2.8, AS A hNCTION OF H, (Cd.)
Axi = 70
H~
xi. = 90
Xw = 20
0.0 0.0000
0.1 0.2986
0.2 0.4247
0.3 0.5231
0.4 0.6077
0.5 0.6836
0.6 0.7537
0.7 0.8194
0.8 0.8819
0.9 0.9419
1,0 1.0000
H,
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
X;. = 50
xiAf = 20
0.0000
0.1294
0.2430
0.3472
0.4452
0.5394
0.6313
0.7222
0.8131
0.9053
1.0000
X,m = 10
XW = 60
0.0000
0.0858
0.1709
0.2563
0.3431
0.4322
0.5251
0.6236
0.7308
0.8521
1.0000
85 80
75
70 65 -60 55
15 10 5 0 5 10 15
.
O.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.2530 0.2193 0.1942 0.1750 0.1600 0.1479 0.1379
0.3847 0.3522 0.3257 0.3037 0.2851 0.2692 0.2553
0. 4*5 0.4592 0.4341 0.4124 0.3934 0.3764 0.3611
0.5782 0.5525 0.5299 0.5097 0.4915 0.4749 0.4596
0.6591 0.6372 0.6175 0.5997 0.5832 0.5678 0.5533
0.7339 0.7161 0.6997 0.6846 0.6704 0.6569 0.6439
0. 8Q45 0.7908 0.7781 0.7661 0,7547 0.7437 0.7328
0.8718 0.8625 0.8537 0.8453 0,8371 0.8291 0,8211
0.9368 0.9320 0.9274 0.9230 0.9186 0.9142 0,9098
I.0000 1.0000 1,0000 1.0000 1.0000 1.0000 1.0000
45 40 35 30 25 20 15
25 30 35 40 45 50 55

O.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0,0000
0.1221 0.1156 0.1098 0.1045 0.0995 0.0947 0.0902
0.2320 0.2218 0.2124 0.2036 0.1951 0.1869 0,1789
0.3342 0.3221 0.3105 0.2994 0.2886 0.2778 0,2672
0.4316 0.4186 0.4059 0.3935 0.3811 0.3687 0.3561
0.5260 0.5129 0.5000 0.4871 0.4740 0.4606 0.4467
0.6189 0.6065 0.5941 0.5814 0.5684 0.5548 0.5404
0.7114 0.7006 0.6895 0,6779 0.6658 0.6528 0.6389
0.8049 0.7964 0.7876 0.7782 0.76843 0.7570 0.7447
0.9005 0,8955 0.8902 0.8844 0.8779 0.8706 0.8621
1.0000 1,0000 1.0000 1.0000 I.000o 1.0000 1.0000
5 0 5 10 15 20
65 70 75 80 85
90

O.0000 0.000o 0.0000 0.0000 0.0000 0.0000
0.0814 0.0770 0.0726 0.0680 0.0632 0.0581
0.1629 0.1547 0.1463 0.1375 0.1282 0.1181
0.2453 0.2339 0.2219 0.2092 0.1955 0,1806
0.3296 0.3154 0.3003 0.2839 0.2661 0.2463
0.4168 0.4003 0.3825 0.3628 0.3409 0,3164
0.5085 0.4903 0.4701 0.4475 0.4218 0,3923
0.6066 0.5876 0.5659 0.5408 0.5115 0,4769
0.7149 0.6963 0.6743 0.6478 0.6153 0.5753
0.8400 0.8250 0.8058 0.7807 0. 747(I 0,7014
1.0000 1.0000 1.0000 1,0000 1.0000 1,0000
328 APPENDI X A
TABLE A.2,e, AS A FUNCTION OF H, (Cont.)
Ax; = 80
0.0
0.1
0.2
0.3
0,4
0.5
0.6
0.7
0.8
0.9
1.0
Xi. = 90
X;M = 10
0.0000
0.2932
0.4176
0,5154
0.5997
0.6758
0.7465
0.8133
0.8774
0.9394
1.0000
H, I ~:: -50
, 30
0.0 0.0000
0.1 0.1281
0.2 0.2393
0.3 0.3410
0.4 0.4369
0.5 0.5294
0.6 0.6204
0.7 0.7113
0.8 0.8036
0.9 0.8991
1.0 1.0000
Hk
x,. = lo
X,M = 70
*

0.0 0.0000
0.1 0.0803
0.2 0.1601
0.3 0.2404
0.4 0.3221
0.5 0.4065
0.6 0.4953
0.7 0.5910
0.8 0.6976
0.9 0.8241
1.0 1.0000
85 80
75
70 65 60 55
5 0 5 10 15 20 25
0.0000 0.0000 0.0000 0,0000 0.0000 0.0000 0.0000
0.2517 0.2198 0.1952 0,1759 0.1604 0.1478 0.1372
0.3809 0.3502 0.3244 0.3024 0.2835 0.2670 0.2524
0.4832 0.4552 0.4307 0.4090 0.3896 0. 3720~0.3559
0.5720 0.5472 0.5250 0.5047 0.4860 0.4686 0,4524
0.6525 0.6313 0.6117 0.5935 0.5764 0.5601 0.5446
0.7276 0.7100 0.6936 0.6779 0.6630 0.6485 0.6344
0.7988 0.7852 0.7722 0.7596 0.7474 0.7354 0.7234
0.8675 0.8580 0.8488 0.8399 0.8310 0.8220 0.8130
0.9343 0.9294 0.9245 0.9197 0.9148 0.9098 0.9046
1.0000 1.000o 1.0000 1,0000 I,moo 1.0000 1.0000
45 40 35 30 25 20 15
35 40 45 50 55 60 65
,__
O.0000 0.0000 0.0000 0,0000 0.000o 0.0000 0.0000
0.1202 0.1132 0.1068 0.1009 0.0954 0.09021 0.0852
0.2274 0.2164 0.2062 0.196410.1870 0.1780t 0.1690
0.3270 0.3138 0.3010 0.2887 0.2766 0.2646 0,2526
0.4221 0.4077 0.3936 0.3796~ 0.3656 0.3515 0 3370
0.5145 0.5000 0.4855 0.4706 0.4554 0.4399 0.4236
0.6064 0.5923 0.5779 0.5631 0.5476 0.5314, 0 5140
0.6990 0.6862 0.6730 0.6590 0.6441 0.6280, 0.6104
0.7938 0.7836 0.7726 0.7607~ 0.7476 0.7330 0.7165
0.8932 0.8868 0.8798 0.8719 0.8628, 0,8522 0.8396
1,0000 I.0000 I.000o 1.0000 1.000C, 1.000Q 1.0000
5 0 5 10
75 80 85 90
0.0000 0.0000 0.0000 0.000c
0.0755 0.0706 0.0657 0.0606
0.1512 0.1420 0.1325 0.1226
0.2278 0.2148 0.2012 0.1867
0.3064 0.2900 0.2724 0.2535
0.3883 0.3687 0.3475 0.3242
0.4750 0.4528 0.42801 0.400
0.5693 0.5448 0.5168 0.4846
0.6756 0.6498 0.6191 0.5824
0.8048 0. 7MJ2 0.7483 0.7068
1.0000 1.0000 1.0000 1.0000
J
APPENDI X A
329
TABm A.2.o; AS A FUNCTIONOF Hk (Cod.)
AX~ = 90
H,
xi. = 90 85
80. 75= _70.
65 60 55 50 45
, x,x = o 5 10 15 20 25 30 35 40 45
ll llllllllll
0.0
0.1
0.2
0.3
0.4
0.5
0,6
0.7
0.8
0,9
1.0
0.0000
0.2871
0.4097
0.5064
0.5903
0.6667
0.7380
0.8060
0.8718
0.9362
1.0000
0.00000,0000 O.00000.0000 0.0000 0.0000 O.0000().0000 O.0000
0.2488 0.2185 0.1944 0.1751 0.1593 0.1462 0.13520.1256 0.1172
0.3754 0.3460 0.3207 0.2988 0,2797 0.2627 0,247510.2337().2211
0.4760 0,4490 0.4249 0.4031 0.3833 0.3651 0.3482 0.3324 0.3174
0. 5639/0.5397 0.5175 0.4969 0.4777 0.4596 0.4423 0,4257 0.4097
0.6441 0.6230 0.6033 0.5846 0.5667 0.5495 0,5328 0,5163 0,5000
0.7194 0.7018 0.6849 0.6685 0.6527 0.6371 0.6216 0.6061 0,5903
0.7916 0,7776 0.7641 0.7507 0.7375 0.7242 1),71070.6969 0.6826
0.8618 0,8519 0.8422 0,8324 0.8225 0.8123 0,8018 0.7907 0.7789
0.9310 0.9257 0.9204 0,9150 0,9094 0.9034 0,8971 0.8903 0.8828
1,0000 1.0000 1.0000 1.0000 1.0000 1,0000 1,0000 1,0000,1.0000
~
I (
I
1 I { I I
I I
J
X<m= 40 35 30 25 20 15 10 5 0
k Xl,lf = 50 55 60 65
70 75 80 85
90
. . .
0.0 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0,0000 0.0000 0.0000
0.1 0.1097 0.1029 0.0966 0.0906 0.0850 0.0796 0,074310.0690 0.0638
0.2 0.2093 0.1982 0.1877 0.1775 0.1676 0.1578 0.1481 0.1382 0.1282
0.3 0.3031 0.2893 0.2758 0.2625 0.2493 0.2359 0,2224 0.2084 0,1940
0.4 0.3939 0.3784 0.3629 0.3473 0.3315 0,3151 0.298210.2806 0,2620
0.5 0.4837 0.4672 0.4505 0.4333 0.4154 0.3967 0.3770 0.3559 0.3333
0.6 0.5743 0.5577 0.5404 0.5223 0.5031 0.4825 0.4603 0.4361 0.4097
0.7 0.6676 0.6518 0.6349 0.6167 0.5969 0.5751 0.5510 0.5240 0.4936
0.8 0.7663 0.7525 0.7373 0,7203 0,7012 0.6793 0,654010.62460,5903
0.9 0.8744 0.86480,8538 0.8407 0.8249 0,8056 0,781510.75120.7129
1.0 1.0000 1.0000 1.0000 1.0000 l.lxloo 1,0000 1,0000 1.0000 1,0000
AX; = 100
Hk
x;.= 90 85 80 75 70 65 61)0 55 50 ;
X,M = 10 15 20 25 30 35 40 45 50 ~
0.0 0.0000
l
l
O.0000o.OOOqo,ooQoO.00000.0000 o. Ooooo.OOOoo. oo(i~
0.1 0.2804 0.2446 0.21560.1920 0.1728 0,156810.1434O. 1320~0.1221,
0.2 0.4007 0.3682 0.3399!0.31510.2933 0,2739 0.2565 0.24080,22641
0.3 0.4962 0.4671 0.4407 0.4168 0.3948 0.3746 0.3558 0.338210.3216
0.4 0.5796 0.5538 0.5300 0.5076 0.4866 0.4667 0.4478 0.4296 0.4119
0.5 0.6559 0.6337 0,6125 0.5924 0.5730 0.5542 0.5359 0.5179 0.5000
0.6 0.7279 0.7093 0.69120.6737 0.6565 0.6395 0,6225 0.6054 0.5881
0.7 0.7972 0.7824 0.7679!0.75350.7390 0.7245 0.7096 0,6943 0.6784
0.8 0.8650 0.8545 0,8440 0.8333 0.8224 0.8111 0.7994 0,7869 0.7736
0.9 0.9323 0.9266 0.9208 0.9148 0.9086 0,9019 0.8947 0,8868 0,8779
1.0 1,0000 1.0000 1.0000 1.0000 1.0000 1,0000 1.0000 1.0000 1,0000
-!
330 APPENDI X A
TABLE A.2.& AS A FUNCTION OF H~ (Cont.)
Axi = 100
H,
Xi. = 45 40 35 30 25 20 f
15 10
X,~ = 55 60 65 70 75 80 85 90
. . . .
0.0 0. Oooo 0.0000 0.000o 0.0000 0.0000 0.0000 0.0000 0.000C
0.1 0.1132 0.1053 0.0981 0.0914 0.085210.0792 0.0734 0.0677
0.2 0.2131 0.2006 0.1889 0.1776 0.1667 0.1560 0.1455 0.1350
0.3 0.3057 0.2904 0.2755 0.2610 0.2465 0.2321 0.2176 0.2028
0.4 0.3946 0.3775 0.3605 0.3435 0.3263 0.3088 0.2907 0.2721
0.5 0.4821 0.4641 0.4458 0.4270 0.4076 0.3875 0.3663 0.3441
0.6 0.5704 0.5522 0.5333 0.5134 0.4924 0.4700 0.4462 0.4204
0.7 0.6618 0.6442 0.6254 0.6052 0.5832 o. 5593 0.5329 0.5038
0.8 0.7592 0.7435 0.7261 0.7067 0.6849 06601 0.6318 0.5993
0.9 0.8680 0.8566 0.8432 0.8272 0.8080 0.7844 0.7554 0.7196
1.0 1.0000 p.oOOo 1.0000 1.0000 l .ol n)o 1.0000 1.0000 1.0000
AX; = 110

0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
xi. = 90
X,.u = 20
0.000o
0.2730
0.3907
0.4847
0.5673
0.6436
0.7161
0.7866
0.8566
0.9272
1.0 ~ 1,0000
[
H,
0.0
0.1
0,2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Xim = 50
X,~ = 60
0.ooOo
0.1175
0.2174
0.3087
0.3954
0. 4&36
0.5664
0.6556
0.7514
0.8605
I.000o
85 80
750
70 65 60 55
25 30 35 40 45
50 55
.
0.0000 0.0000 0.000o 0.0000 0.000o 0.0000 0.0000
0.2392 0.2113 0.1883 0.1692 0.1531 0.1395 0.1278
0.3597 0.3322 0.3077 0.2859 0.2663 0.2486 0.2324
0.4565 0.4305 0.4066 0.3845 0.3638 0.3444 0.3261
0.5420 0.5181 0.4955 0.4740 0.4534 0.4336 0.4143
0.6212 0.5998 0.5790 0.5588 0.5390 0.5194 0.5000
0.6970 0.6783 0.6598 0,6415 0.6231 0.6046 0.5857
0.7712 0.7557 0.7401 0.7243 0.7081 0.6913 0.6739
0.8453 0.8338 0.8219 0.8095 0.7964 0.7826 0.7676
0.9210 0.9144 0.90741 0.8998 0,8916, 0.88251 0.8722
I.000o 1.0000 1.000o~ l,ml 1,0000 1.0000 1.0000
45 40 35 30 25 20
65 70 75 80 85 90
0.OOOO 0.0000 0.0000 0.0000 0.0000 0.0000
0.1084 0.1002 0.0926 0.0856 0.0790 0.0728
0.2036 0.1905 0.1781 0.1662 0.1547 0.1434
0.2919 0.2757 0.2599 0.2443 0.2288 0.2134
0.3769 0.3585 0.3402 0.3217 0.3030 0.2839
0.4610 0,4412 0.4210 0.4002 0,3788 0.3564
0.5466 0,5260 0.5045 0.4819 0.4580 0.4327
0.6362 0.6155 0.5934 0.5695 0.5435 0.5153
0.7337 0.7141 0.6923 0.6678 0.6403 0,6093
0.8469 0.8308 0,8117 0.7887 0.7608 0.7270
1.0000 l .oOoc, 1.Oooo
1.0000 1.00001 1.0000
I
I
I
!
I
I
APPENDI X A
TABLE A.2.tIi ASA Fu~cmo~ OF Hk (C*.)
Axi ~ 120
331
H,
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
x
,m=
90
X1.u = 30
0.0000
0.2649
0.3798
0.4719
0.5535
0,6294
0.7022
0.7739
0.8461
0.9207
1.0000
H,
x,. = 55
X,~ = 65
0.0 0,0000
0.1 0,1226
0.2 0,2225
0.3 0.3121
0.4 0.3966
0.5 0,4792
0.6 0,5624
0.7 0.6489
0.8 0.7428
0.9 0.8517
1.0 1,0000
Axi e 130
H,
X,m = 90
X,M = 40
0.0 0.0000
0.1 0.2562
0.2 0,3678
0.3 0.4579
0.4 0.5381
05 0.6132
0.6 0.6860
0.7 0.7586
0,8 0.8332
0.9 0.9122
1.0 1.0000
85
35
0.0000
0.2329
0.3498
0.4443
0.5283
0.6067
0.6824
0.7574
0.8337
0.9135
1.0000
80
40
0.0000
0.2059
0.3230
0.4186
0.5042
0.5846
0,6627
0,7407
0.8208
0.9057
1.0000
50 45
70 75
0.ooOo 0.0000
0.1121 0,1027
0.2071 0.1928
0.2939 0.2763
0.3766 0.3569
0.4582 0.4370
0.5410 0.5188
0.6278 0.6054
0.7230 0.7012
0.8356 0.8167
1.0000 1.0000
75
45
0.000o
0.1833
0.2988
0.3946
0.4812
0.5630
0.6431
0.7237
0.8072
0.8973
1.000o
40
80
0.0000
0.0943
0.1792
0.2593
0.3373
0.4154
0.4958
0.5814
0.6770
0.7941
1.0000
85 800
45 50
0.0000 0,0000
0.2256 0.1995
0,3388 0.3124
0,4306 0.4049
0.5128 0.4882
0.5899 0,5671
0.6652 0.6444
0.7408 0.7226
0.8192 0.8044
0.9036 0.8940
1.0000 1,0000
70
50
0,0000
0.1644
0.2770
0.3722
0.4590
0,5418
0.6234
0.7061
0.7929
0.8879
1,00Q0
35
85
0.000o
0.0865
0.1663
0.2426
0.3176
0.3933
0.4717
0.5557
0.6502
0.7671
1.0000
75
55
0.Oooo
0.1774
0.2884
0.3808
0.4647
0,5446
0.6233
0.7037
0.7886
0.8834
1.0000
.
65
55
0.0000
0.1483
0.2572
0.3511
0.4376
0.5208
0.6034
0.6879
0.7775
0.8774
1.0000
30
90
0.000o
0.0793
0.1539
0.2261
0. 297&
0.3706
0.4465
0.5281
0.6202
0.7351
1.0000
70
60
0, Oooo
0.1586
0.2666
0.3580
0.4419
0.5222
0.6020
0.6841
0.7717
0.8713
1.0000
60
60
0.000o
0,1346
0.2391
0.3311
0.4169
0.5000
0.5831
0.6689
0.7609
0,8654
1.0000
65
65
0.0000
0.1425
0.2466
0,3365
0.4197
0.5000
0.5803
0.6635
0.7534
0.8575
I,oooo
332 APPENDI X A
T-m A.2.& AS A FUNCTION OF H, (Cwt.)
Axi = 130
Hk
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Xi. = 60
X,M = 70
0.0000
0.1287
0.2283
0.3159
0,3980
0.4778
0.5581
0.6420
0,7334
0.8414
1,0000
AX; = 140
H,
0,0
0,1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0,9
1.0
xi.= 90
Xi Jf = 50
0.0000
0,2470
0.3550
0,4425
0,5209
0.5949
0.6673
0.7405
0.8170
0.9008
1.Oooo
55
75
0.0000
0.1166
0.2114
0.2963
0.3767
0.4554
0.5353
0.6192
0.7116
0.8226
1.0000
50
80
o.oOoc
0.1060
0.1956
0.2774
0.3556
0.4329
0.5118
0.5951
0.6876
0.8005
1,0000
85
55
0.0000
0.2174
0.3265
0,4154
0.4952
0,5708
0.6452
0.7210
0.8010
0.8901
1.0000
45
85
0.0000
0.0964
0.1808
0.2592
0.3348
0.4101
0,4872
0,5694
0.6612
0,7744
1,0000
40
90
0.0000
0.0878
0.1668
0,2414
0.3140
0.3868
0.4619
0.5421
0.6322
0.7438
1.0000
80
60
0,0000
0.1921
0,3005
0,3897
0.4704
0.5471
0.6230
0.7008
0.7838
0,8780
1.0000
H,
I
Xi m = 65
I
(joo
I
55
X,,u = 75 80 85
0.0 0,0000 0.0000 0.0000
0.1 0.1359 0,1220 0.1099
0.2 0.2348 0.2162 0.1990
0.3 0.3203 0.2992 0.2790
0.4 0.3996 0.3770 0.3548
0,5 0.4765 0.4529 0.4292
0,6 0.5538 0,5296 0.5048
0.7 0.6346 0.6103 0.5846
0.8 0.7233 0.6995 0.6735
0.9 0.8296 0,8079 0,7826
1,0 1,oooo 1.000o 1.0000
75
65
0,Oooo
0.1704
0.2767
0.3654
0.4462
0.5235
0,6004
0,6797
0.7652
0.8641
1.0000
500
90
0,0000
0.0992
0.1830
0.2595
0.3327
0.4051
0.4791
0.5575
0.6450
0.7530
I .000o

70
70
0.000o
0.1518
0.2549
0.3423
0.4226
0.5000
0.5774
0.6577
0.7451
0.8482
1.0000
APPENDI X B
PROPERTI ES OF THE THREE-BAR-LI NKAGE NOMOGRAM
Thk appendi x i ncl udes a mathemati cal di scussi on of the contours of
the three-bar-l i nkage nomogram, and a tabl e of cur ve coordi nates for use
i n the constructi on of a nomogram sui tabl e for accurate wor k. The
nomogram i tsel f appears as Fi g. B.1, i n a fol der i n the back of the book.
B-l . Contours of Constant h-I n the (p, q)-pl ane the contours of con-
stant b are gi ven by Eq. (5.44):
(1)
q = COSl (cosh p &MP).
Si nce the functi on COS-l z i s mul ti pl e val ued, q i s a mul ti pl e-val ued
functi on of p, for any gi ven b. I f (p, q) is a poi nt on the contour of con-
stant b, so is (p, ~q f 2hr), for any i ntegral val ue of k. When b >0,
these poi nts al l fal l on a si ngl e conti nuous contour; when b <0, the
contour consi sts of a system of i sol ated cl osed curves. I n any case the
compl ete contour has an i ni i ni te set of hori zontal axes of symmetry:
~ = kr, }=0,+1,+2,. (2)
Other symmetry properti es depend upon the si gn of b.
Contours for b <0 have a verti cal axi s of symmetry. When b <0,
l e~>O,
(3)
and one can defi ne a real constant T by
T=~l n(l e*), (4a)
e27 = 1 e2b.
(4b)
I n terms of the parameter T, Eq. (1) becomes
~ = COS-I [e-~ cosh (p + T)].
(5)
Thus q(p, b) i s unchanged by change of si gn of p + T; the contour i s
symmetri c to refl ecti on i n the verti cal l i ne
p= T= Jl n(l e%).
(6)
The contours of constant b >0 have no vertical axis of symmetry; the
above argument does not apply because T as defi ned by Eq. (4) i s no
l onger real . One can, however , defi ne a real parameter t by the rel ati on
334 APPENDI X B
I n terms of the parameter t, Eq. (1) becomes
~ = COS-L[e- si nh (p + t)].
(8)
Change i n si gn of (p + 1) changes the si gn of the argument on the ri ght;
q can then be repl aced by (2k + 1)~ q, wher e k = O, f 1, +2, . . .
I t fol l ows that the contours of constant b >0 have an i nfi ni te sequence of
center s of symmetry at
pk = ~ = *l n (eb 1),
?k = (k + +)7r,
k=o, +1, +2,....
)
(9)
The l i mi ti ng contour, b = O, has no verti cal axi s or center of symmetry
except at i nfi ni ty. I ts equati on i s
q = COSl (+ep). (lo)
Thi s curve i ntersects the axi s q = O at p = i n 2, and has no poi nts for
whi ch p > i n 2. I t has hori zontal asymptotes
q = of+4)T,
k= O?l , &2,.... (11)
B02. Contours of Constant X.To study the contours of constant X
i t i s necessary to el i mi nate b from Eq. (1) and Eq. (5.45):
p=~l n(2cos X+2coshb)+~b. (12)
These equati ons may be r ewr i tten i n an i nteresti ng and symmetri cal
form:
2 cos q = e + e-p(l ezb), (13)
2cos X = eb+e-(l e). (14)
Substi tuti on i nto Eq. (14) of d, as gi ven by Eq. (13), l eads to the rel ati on
er cos q-l
Cos x = (1 + ep 2ePcos q)~$
An equi val ent but si mpl er rel ati on,
cos q ep
cot x =
si n q
(si n q si n X > O)
(15)
(16)
fol l ows from thi s by tri gonometri c rearrangement.
For the anal ysi s at hand i t i s conveni ent to r ewr i te Eq. (16) as
si n X cos q si n ~ cos X = si n Xcp,
(17)
or
si n (X q) = si n Xcp,
(si n q si n X > O).
(18)
As noted, onl y that porti on of thi s cur ve i s to be consi dered for whi ch
si n q has the same si gn as si n X.
APPENDI X B 335
Let O < X, < 180. The contour for whi ch X = XOmust l i e onl y
i n the regi on for whi ch si n q, l i ke si n Xo, i s posi ti ve. Thi s contour i s
then
si n (XO q) = S* XOe-p, (si n q > O). (19)
On the other hand, the contour X = X, 180 must l i e onl y i n the
regi on for whi ch si n q, l i ke si n (XO 180), i s negati ve; i t i s the cent our
si n (XO 180 TI )= si n (XO 180)eP,
si n q <0 (20a)
or
si n (XO q) = si n xoe-~, (si n q < O). (20b)
These two contours joi n smoothl y at the ori gi n, formi ng a conti nuous
cur ve whi ch approaches the hori zontal asymptotes q = XO and
~ = XO 180
asp+~.
Si nce si n XO >0, we may wri te the compl ete cur ve for any XO as
si n (XO q) = e(fi l n&x). (21)
Thi s i s the cur ve
cos q = ep, (22)
transl ated upward by
Aq = XO 90, (23)
and to the l eft by
Ap = i n si n XO. (24)
Thus, al l of the curves defi ned by Eq. (21) have the same form, what-
ever the val ue of XO. Equati on (22) gi ves di rectl y the form of the cur ve
for XO = 90, consi sti ng of the contours X = 90 and X = 90.
I t wi l l be noted that Eq. (22) di ffers from Eq. (10) onl y by a refl ecti on
i n a verti cal l i ne and a transl ati on paral l el to the paxi a. I t f ol l ows that the
two contours X = XOand X = XO 180 form (for any O < XO < 180)
a cur ve of the same form as the contour b = Or efl ected i n a verti cal l i ne.
B-3. Expl anati on of Tabl e B. I .Tabl e B-l gi ves the coordi nates i n
the (p, q)-pl ane of the i ntersecti on of the contours of constant b and the
contours of constant X, for
x = 0, 5, 10, . . 0 , 180,
#b = 0.50, 0.49, . . . .
0.49, 0.50.
Readi ng the coordi nate pai rs from associ ated verti cal col umns, one can
pl ot any contour of constant b; readi ng them from a si ngl e r ow, one can
pl ot any contour of constant X.
I
0*m
wwa
,,
,.
.
,.
.,,,
,..
,,,

,,,
,=,
-,
.,,,
,,
,,,
,,

vI
I
I
02
I
0$
ww-
1
338 APPENDI X B
TABLE B.1.COORDINATES OF POINTS ON THE THREE-BAR-LINKAGE
NOMOGRAM (Cont.)

pb = 0.38 0,37 0.36 0.35
I
0.34 0.33
x,
de-
grees
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
30
85
90
95
100
105
110
115
120
125
130
135
140
145
150
155
160
165
170
175
180
7s 7. 73 n! ?, ?!
tp + 10 de- PP + 10 de- *P + 10 de- PP + 10 de- PP + 10 de- !J P+ 10 de-
g.ees greea g,ee9 g,ees g.ees
g,ees
. . . . .
0.1513 0.00 10.1543 0.00 10,1573 0.00 10.1604 0.00 10.1635 0.00 10.1667 0,00
0.1509 1.47 10.1540 1.49 10.1570 1.52 10.1601 1.54 10.1631 1.57 10.1662 1.59
0.1499 2.94 10,1529 2.99 10.1559 3.03 10.1590 3.08 10.1621 3.13 10.1652 3.18
0.1482 4.40 10.1511 4,47 10.1541 4,54 10,1572 4.62 10.1603 4.69 10,1634 4,77
0.1458 5,85 10.1486 5.95 10.1517
6.04 10,1547 6.14 10,1578 6.24 10.1609 6.34
0.1427 7.29 10.1455 7.41 10.1485 7.53 10.1515 7.65 10.1546 7.78 10.1577 7.90
0.1389 8.71 10,1417 8.85 10.1446 9.00 10.1476 9.15 10.1506 9.30 10.1537 9.45
0.1344 10.11 10,1371 10.28 10.1400 10.45 10.1429 10.62 10.1459 10.8O10.1489 10.98
0.1292 11.48 10.1318 11.68 10.1347 11.87 10.1375 12.07 m. 1405 12.28 10.1434 12.48
0.1232 12.8310.1258 13.05 10.1286 13.27 m. 1314 13.50 10.1343 13.73 10.1371 13.96
0.1165 14.1410.1191 14.38 10.1217 14.63 10.1245 14.89 10.1273 15.14 10.1301 15.40
0,1090 15,41 10.1116 15.6810.1141 15.96 10.1168 16.24 10.1194 16.52 10.1222 16.81
0.1008 16,6310.1032 16.9310.1057 17.24 10.1082 17.5510.1108 17.86 10.1135 18.17
0 0918 17.8110.0941 18.14 10.0965 18.47 10.0989 18.8110.1014 19.15 10.1040 19.49
0,0820 18.92 10.0842 19.28 10,0865 19.64 10.0888 20.01 10.0912 20.38 10.O936 20.75
0.0714 19.97 10.0734 20,36 10.0756 20.75 10.0778 21.14 10.0800 21.54 10.0823 21.95
0.0600 20.95 10.0620 21.36 10.0639 21.78 10.0659 22.21 10.0680 22.64 10.0701 23.08
0.0478 21.84 10.0495 22.28 10.0513 22.73 10.0531 23.19 10.0550 23.65 10.0570 24.12
0,0348 22.63 10.0363 23.10 10.0379 23.58 10.0395 24.07 10.O412 24.56 10.0429 25.07
0.0209 23.31 10.0222 23.82 10.0235 24 33 1o.0249 24.84 10.0264 25.37 10.0279 25.91
0.0062 23,87 10.0072 24.40 10.0083 24.94 10.0094 25.5o 10.0106 26.06 10.0119 26.63
9.9907 24,29 9.9914 24,85 9.9922 25.42 9.9930 26.01 9.9939 26.60 9.9949 27.20
9.9744 24.55 9.9748 25,14 9.9752 25.74 9.9757 26.35 9.9763 26.98 9.9769 27.62
9.9573 24,64 9.9573 25.25 9.9574 25.88 9.9575 26.52 9.9576 27.18 9.9578 27.85
9.9395 24.52 9.9391 25.15 9.9387 25.81 9.9384 26.48 9.9380 27.16 9.9378 27.87
9.9212 24.17 9.9202 24.83 9.9194 25.50 9.9186 26.19 9.9177 26.91 9.9170 27.64
9.9024 23.57 9.9009 24.24 9.8995 24.93 9.8981 25.64 9.8967 26.37 9.a953 27.13
9.8834 22.68 9.8813 23.36 9.8792 24.06 9.8771 24.78 9.8751 25.53 9.8730 26.30
9.8642 21.49 9.8615 22.16 9.8588 22.86 9.8560 23.58 9.8532 24.33 9.8505 25.10
9.8454 19.96 9.8421 20,61 9.8386 21.29 9.835o 22.00 9.8314 22.74 9.8279 23.50
9.8274 18.07 9.8233 18.69 9.8190 19.34 9.8147 20.01 9.8102 20.72 9.8057 21.46
9.8107 15.s1 9. eo57 16.38 9. 80Q7 16.97 9.7955 17.60 9.7902 18.25 9.7847 18.94
9.7958 13.19 9.7899 13.69 9.7842 14.20 9.7781 14.75 9.7719 15.33 9.7655 15.93
9.7831 10,24 9.7768 10.64 9.7703 11.05 9.7637 11.49 9.7566 11.96 9.7493 12.45
9.7737 7.00 9.7669 7.28 9.7598 7.57 9.7526 7.88 9.7450 8.21 9.7370 8.56
9.7678 3.56 9.7606 3.70 9.7532 3.85 9.7455 4.01 9.7375 4.18 9.7285 4.37
9,7658 0.00 9.7585 0.00 9.7509 0.00 9 7430 0.00 9.7347 0.00 9.7261
0 00
t
a
www
340 APPENDI X B
TABLE B1. -COORDINATES OF POINTS ON THE THREE-BAR-LINKAGE
NOMOGRAM (ConL)
pb- 0.26
I
0.2,3
I
0,24
I
0.23
I
0.22
I
--0,21
x,
de-
&r*
.
(
:
1(
15
2C
2E
3G
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
135
140
146
154
155
160
165
170
176
180
I I 1 I - I
..
I I
v, v. v. 73 v,
71
*P + 10 de- Itp + 10 de- pp + 10 de- *P + 10 de- Mp+ 10 de- Wp+ 10 de-
grees green grew grees grees grees
. . . . . . . . . __
10.1902 0.00 10,1938 0.00 10.1974 0.00 10,2011 0.00 10.2048 0,00 10,2086 0,00
10.1898 1.77 10.1935 1.80 10.1970 1.83 10.2006 1.85 10,2045 1.88 10.2082 1,91
10.1886 3.54 10.1923 3.60 10.1959 3.65 10.1995 3.70 10.2033 3,76 10,2071 3,31
[0. 1868 5,31 10,1903 5.39 10.1940 5.47 10,1976 5.55 10,2013 5,63 10 2051 5 71
Lo. 1841 7.07 10.1876 7.17 10.1912 7.28 10.1949 7.39 10,1985 7.50 10.2023 7,61
10.1807 8.81 10,1842 8,95 10.1878 9.08 10.1914 9,22 10.1951 9.35 10.1988 9.43
:0.1785 10.55 10.1800 10.71 10.1835 10.87 10,1871 11,03 10.1908 11.20 10.1944 11.36
0.1715 12,26 10,1749 12,45 10.1784 12.64 10,1819 12 84 10,1856 13.03 10.1892 13.22
0.1656 13.96 10.1690 14,18 10.1725 14.40 10.1760 14.62 10.1796 14.84 10.1.932 15.07
0.1590 15.63 10.1623 15.843 10.1657 16.13 10.1692 16.38 10.1728 16,63 10.1763 10,89
0.1515 17,28 10,1547 17.56 10.1581 17,84 10,1615 18.12 10.1650 18.40 10,1686 18.69
0.1430 18.89 10.1463 19.30 10.1496 19.51 10,1529 19.83 10.1563 20.14 10,1598 20.46
0.1338 20.47 10.1370 20.81 10.1401 21.16 10.1434 21.50 10.1467 21.85 10,1502 22.20
0.1236 22.01 10.1267 22.38 10.1298 22,76 10,1329 23.14 10.1362 23.52 10.1396 28.91
0,1124 23.49 10.1154 23,90 10.1184 24.31 10.1215 24,73 10,1246 25.15 10.1279 25.67
0 1002 24.93 10.1031 25.37 10.1060 25.82 10.1091 26.26 10,1120 26.73 10.1151 27.19
0.0870 26.29 10,0897 26.77 10.0925 27.26 10.0954 27.75 10.0983 28.24 10.1013 28.74
0.0727 27.58 10.0753 28.10 10.0779 28.63 10.0806 29,16 10.0834 29.70 10.0862 30.24
0.0573 %.79 10.0597 29.351 0.0621 29.92 10.0646 30.49 10.0673 31.07 10. O7O4J 31 66
0.0407 29.90 10.0428 30,501 0,0451 31.11 10,0474 31,73 10,0498 32.36 10 0524 32.99
0.0229 30.89 10.0248 31,541 0.0268 33,191 0.0289 32.86 10,0311 33.53 10.0334 84.22
0. oa37 31.751 0.0054 32.441 0.0071 33.151 0.0089 33.861 0.0108 34,59 10.0128 35.31
9.9833 32.45 Q. 9846 33.20 9.9860 33.95 9.9875 34.72 9,9891 35.50 9.9908 36.29
9.9615 32,97 9.9624 33.76 9.9634 34,57 9.9645 35,39 9,9667 36,23 9,9670 37.08
9.9383 33.27 9,9386 34.12 9.9392 34.98 9.9398 35.86 9.9406 36.75 9.9414 37.67
9.9136 33,32 9.9135 34.21 9.9134 35.13 9.9135 36.07 9,9137 37,03 9.9140 38. o1
9.8875 33.06 0.8867 34.00 9.8859 34,98 9,8854 35,97 9.3849 36,99 9.8846 38.04
9.3600 32,44 Q. 8534 33.43 9.8569 34.45 9.8555 35.51 9,8542 36.59 9.8530 37.71
9.8314 31.39 9.8288 32.43 9.8264 33.48 9.8240 34.59 9.8217 35.73 9.8195 36.91
9,8020 29,82 Q,7983 30.38 9.7946 31.98 9,7911 33.12 9.7876 34.31 9.7841 35.65
9.7722 27.67 9.7671 28.73 9,7622 29.84 9,7571 31.00 9.7523 32.21 9.7472 33.49
9.7428 24,82 9.7364 25.86 9.7208 26.94 9.7231 28, o9 9,7165 29.29 9.7096 30.57
9.7130 21,24 9.7069 22.19 9.6986 23,20 9.69o1 24.27 9,6815 25.41 9.6725 26,63
9 6906 16,86 9,6808 17,67 9,6707 18.54 9,6603 19.46 9,6495 20.46 0.6332 21.53
9.6708 11.75 9.6599 12,34 9,6480 12.99 Q.6359 13,68 9.6231 14.43 9.13c07 15.25
9.6582 6.04 9.6459 6.36 9.6333 6.70 9,6197 7,08 9.6052 7,49 9,5905 7.93
9,6537 0,00 9.6411 0.00 9,6279 0.00 9.6140 0.00 9,5993 0.00 9,5887 0,20
APPENDI X B
TABLE B.1.COORDI NATEEI OF POINTE ON TEE THREE-BAR-LINKAGE
NOMOGRAM (Cont.)
341
pb -
x,
de-
sr0e8
0
s
10
15
20
25
80
w
40
4b
50
66
60
66
70
76
60
65
w
06
Ilw
105
110
116
124
125
180
18s
MO
145
lm
153
166
165
170
176
lk)

0.20 -0.19 -0,1.s -0.17 -0.16 0.15

m! v, v? v, ?,
.P + 10 d- *P+10 de- PP + 10 de- FP + 10 d- IJP + 10 d+ MP + 10 dq:
OOed
greea ale=?a greea greed mean
. . . . .
10.2124 0.00 10.2163 0.60 10.2203 0.60 10.2243 0.60 10.2264 0.00 10,2325 O.oa
10.2121 1.9a 10.2166 1.96 10.2199 1.99 10.2238 2.02 10.2280 2,04 10.2320 2.o7
10.2109 a.w lo. a147 a.92 10.2189 3.98 10.2227 4.oa 10.2268 4.09 10.2369 4.14
10.ww 6.79 10.2128 5.86 10.2167 5.9G10.2307 6.64 10.2247 6,]a 10.2288 6.21
LO.2001 7.72 10.WW 7.S3 10.2139 7.94 10.2179 8.05 10.2219 S.16 10.2200 8.27
10.2026 o.aa 10.2005 9.77 10.2103 9.91 lo.214a 10.05 10.2183 10.19 10.2224 lo, aa
10.1962 11.62 10.2021 11.70 10.2059 11.87 10, 20U8 12.64 10.2139 12.21 10.2179 12, a8
10. MO la,43 10.1968 13.62 10.2060 la.81 10.2045 14.01 10.2665 14.21 10.2126 14,41
10.1869 15.29 10.1967 15.62 10.1945 15.75 10.1984 15.98 10. 202a 16.21 10.2064 M. 44
[0. 1600 17.15 10,1628 17.41 10.1875 17.66 10.1914 17.92 10.1952 18.19 10.1992 18,45
[0. 1722 la,98 10.1759 19.27 10.1796 19.56 10.1634 19,85 10.1872 20. lb 10,1912 20,44
[0. 1624 20.78 10.1670 21.11 10.1767 21.43 10.1744 21.76 10.1782 22.08 10.1821 22,41
[0. 1526 22. E8 10.1572 22.91 10.1009 2a,27 10.1045 2a, 6a 10.1084 24.06 10.1721 24.36
10.1429 24.26 10.1404 24.69 10.1561 25.09 10.1526 25.46 10.1574 26.68 10,1611 26.28
10. lalz 26,66 10.la46 26.42 10.laal 26.86 10.1416 37.39 10.1452 27.73 10.1490 28.17
[0.1164 27.66 10.1216 28.12 10.1256 26.59 10.1385 29.07 10.la20 29.54 10.1357 ao, 02
10,1644 28.25 10.1075 29.76 10.1109 30.27 10.1142 ao. 79 10.1177 al,al 10.1212 31.84
10.6W2 20.79 10.0922 al.34 10.6954 al. w 10.OWO 32,46 10.1020 aa. 02 10,1054 aa. 59
10.0726 a2.25 10.6750 32.85 10.0780 3a.45 10.0817 24.00 10.0649 a4.08 10.0862 35.ao
[0.0549 22,64 10,0576 24,28 10,0665 a4.93 10.6oa4 a5. 59 10.6065 36.26 10.0096 a6,9a
10.0258 24,91 10.62a2 a5.61 10,O4OUa6. a2 10,04a6 a7. oa 10.0465 a7. 7610.6494 38.48
10.ollw aO.67~0.0172 36.82 10.0196 37,b910.0221 aa. a6 10,0247 39.16 10.0274 ae. 94
9.W26 37.69 9. W45 a7.91 9.W60 a8.73 9.W88 a9,57 10.6612 40.42 10,06ae 41,28
9.3664 a7. 95 Q.9700 aa. aa 9.9718 39.72 9. 07a6 46.62 9.9757 41.54 9.9778 42.47
0,942.4 aa. oo Q.9426 39.55 9.9449 40.61 9.9463 41.49 9.9479 42.49 9,9497 4a,60
9.91U 39.01 9.91s1 46.62 9.9158 41.07 9,9106 42.la 9.9178 42.22 9.9191 44,a2
0.6642 a9.12 9.8643 46,21 9.6644 41.34 9.6647 42.49 9.8851 43.67 9.8859 44.86
9.8520 26.66 9.8512 40.02 9.8566 41.24 9.8409 42.49 0. &96 43.76 9.8495 45.07
9.8175 aa.la 9.8166 ao.29 9.81a9 40,68 9.812a 42.62 9.allo 43.41 9.8698 44,84
9.7806 20.64 9.7776 66.17 9.7746 a9.56 9.7717 40.W 9.7691 42.48 9,7666 44.oa
9.7424 24.82 9. 7a75 20.22 9. 7a20 37.67 9.7232 29.20 9. 72a9 40.79 9.7197 42.46
9.70W 31.92 O.woo aa.24 9.0892 24.64 0.6824 a6,43 9.67$6 aa.09 9.6692 a9.86
9.6624 27.9a 9.6543 29.al 9.0448 26.79 9.6262 a2.aa 9.6257 34.67 9.6160 a5,86
9.6267 22.09 9.6147 za. 94 9.0622 25.26 9.5896 28,76 9.,6763 28.36 9,5029 26.69
9.5967 16.14 s. 5610 17.11 9.5656 la. 18 0.5492 19.26 0. 5a22 26.66 9.514a 22.10
0.5746 8.42 0.5579 8.96 9.5460 9.56 9.5268 10.2a 9.5065 10.98 9.4789 11.82
9.5671 0.00 9.5494 0.00 9.62W 0.00 ?.6104 0.00 !3.4686 0.00 Q,4666 0,00
342 APPENDI X B
TABLE B. L-COORDINATEB OF POINTS ON THE THREE-BAR-LINKAGE
NOMOGRAM (ConL)
~b- -0.14
I
-0.13
1
0.12
I
-0.11
I
0.10
1
-0.09

x,
de
gree

{
1(
1!
2(
2:
31
3:
4(
4:
5(
5:
6C
6:
7C
75
3C
85
90
95
100
105
110
115
120
125
130
135
140
145
150
155
160
165
170
175
180
7. v, 79 v, v. v,
UP+ 10 de- PP + 10 de- +P + 10 de- lJ P+ 10 de- PP + 10 de- J IP+ 10 de-
grees grees gree9 gree. greea grees
. . . .
10.2367 0.00 10.2409 0.00 10.2452 0.00 10.2495 0.00 10, 26a9 0.00 10 25&
0.00
10. 2a63 2.10 10.2405 2.la 10.2447 2.16 10.2491 2.19 lo.25a5 2,21 10,25W 2.24
lo.2a50 4.20 10. 229a 4.26 10. 24as 4.31 10.2479 4.37 10. 252a 4.4a 10.2567 4 48
LO.2aao 6.ao 10.2372 6.a8 10.2415 6.47 10.2458 6.55 10.2503 6.64 10.2547 6.72
LO.2aol 8.39 10.2344 8.50 10.2386 S.61 10.2429 8,73 10,2474 8,84 10.25L3 8.96
10.2265 10.47 10.2308 10.61 10.2350 10.76 10.2393 10.90 10,2437 11.04 10,2480 11.19
[0. 2220 12.55 10.2262 12.72 10.2305 12.89 10.2347 la.07 lo,2a91 13.24 10.2435 13.42
10,2166 14.62 10.2208 14.82 10.2250 15.02 IO. 229a 15.23 10.2336 15.43 10.2380 15.63
10.2104 16.67 10.2145 16.91 10.2167 17.14 Io. zzao 17,37 IO. 227a 17.61 Io,2a16 17.85
10.2032 18.71 10,2073 18.9.9 10.2115 19.25 10.2158 19.51 10.2200 19.7.s 10.2244 20.05
10.1951 20.74 10.1992 21.04 lo.2oa4 21.34 10.2076 21.64 10.2118 21.94 10,2162 22.24
10.1861 22,75 10.1901 23.08 10.1942 2a.41 10.1984 23.75 10.2026 24.09 10.2070 24.42
[0.1761 24.73 10. MJO 25.10 10.1840 25.47 10.1882 25.84 10.1924 26.21 10.1967 26.59
10, 1650 26.69 10.1688 27.09 10.1729 27,50 10.1770 27.91 10.1811 28.32 10.1853 28.74
!0, 1528 28.62 10.1565 29.06 10.1605 29.51 10.1646 29.96 10.1687 ao.41 10.1728 aO.87
.0.1394 ao.51 10. 14al at. oo 10.1470 31.49 10.1510 al.98 10.1550 a2. 47 10.1592 32.97
0,1248 a2. a6 10.1285 32.89 10. laza 3a. 43 10.1362 a3.97 10.1402 a4.51 10.1443 35.05
.0.1089 a4. 17 10.1125 a4.75 10.1162 35. aa 10.1200 35.92 10.1239 36.50 10.1279 37.10
0.0916 35.921 0.0951 a6 .55 10.6087 37.18 10.1024 37.81 10.1062 38.46 10.1101 39.11
0.0728 37.611 0.0762 a8.291 0.0797 38.981 0.0832 a9 .67 10.0869 40. a7 10.0908 41.07
0.0524 39.221 0.0557 a9.96 10.0590 40.711 0.0624 41.461 0.0660 42.22 10.0697 42.99
0.0303 40.741 0. 03a4 41.541 0.0364 42.361 0.0397 4a. wl 0.0431 44.66 10.0467 44,84
0. oo6a 42,141 0.0691 4a.02 10.0120 42.911 0.0150 44.801 0 .01s2 45.70 10.0216 46.61
9.9801 43,42 9.9826 44. a7 9.9853 45.34 9.9880 46. a2 9.9910 47. ao 9.9942 48.29
9.9516 44,5a 9.9538 45.57 9.9561 46.62 9.9586 47.69 9.9612 48.77 9.9642 49.86
9.9266 45.44 9.9223 46.57 9.9242 47.73 9.9263 48.90 9.9286 50.08 9.9312 51.28
9.8867 46,09 9.8879 47, 3a 9.3892 48.60 9.8908 49.88 9.8926 51,19 9.8947 52.52
9.8496 46.41 ,9. 85CKJ 47.77 9,8507 49.16 9,8516 50.59 9.8528 52.03 9.8543 5a. 50
9.8090 46.30 9.3084 47.86 9.8681 49.34 9.8082 50.91 9.8085 52.52 9.8692 54.16
9.7645 45.62 9.7626 47.27 9.7610 48.97 9,7598 50.72 9.7596 52.52 9.7586 54.37
9.7157 44.19 9.7121 46.60 9.7088 47.87 9.7059 49. 8i 9.7oa4 51.84 9.7014 53. 9a
9,6629 41.72 9.6568 43,67 9.6510 45.73 9.6457 47.89 9:6407 50.16 q 6a64 52.53
9.6064 a7 ,82 9.5970 39.89 9.5877 42.10 9.5787 44.46 9.5701 46.97 9.5620 49.65
9,5490 31,99 9.5349 34.05 9.5207 36. aO 9.5064 38.76 9.4921 41.46 9.4782 44 39
9.4955 23.70 9.4758 25.49 9.4552 27.51 9. 43a6 29.78 9.4113 32.35 9.3861 35.28
9.4555 12.78 9.4303 13.88 9.4034 15.14 9.3739 16.62 9. a422 18.35 9.3076 20.42
9.4402 0.00 9.4128 0.60 9.3828 0.00 9.3498 0.00 9.3132 0.00 9.2722 0.00
APPENDI X B 343
TABLE B. I. COORDINATES OF POINTS ON TEE THREE-BAR-LINKAGE
NOMOGRAM (Cunt.)
~b=
-008 I -007I -006I -005I -004I -003
x,
de-
grees
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
135
140
145
150
155
KQ
165
170
175
180

v, v, ?9 v,
1P + 10 de- PP + 10 de- *P+ 10 de- *P + 10 de- Up+ 10 de- pp+ 10 de-
grees grees gree~ gree~ I g:es gress
. _ _ _ _ _ _
102629 0.00 10.2674 0,00 10,2721 0.00 10.2768 0,00 10,2815 0,00 10. 2!363 0.00
10.2625 2.27 10.2670 2.30 10.2717 2.33 10.2764 2.36 10.2811 2.39 10.2860 2,41
!0.2613 4.54 10.2658 4.60 10.2704 4.66 10.2751 4.71 10,2799 4 77 10. 2S46 4.83
.0.2592 6.81 10.2637 6.89 10.2683 7.98 10.273o 7.07 10.277,9 7.15 10.2825 7.24
.0.2563 9.07 10.2608 9.19 10.2654 9.30 10.2701 9.42 10,2748 9..54 10.2797 9.65
0 2526 11.33 10,2571 11.48 10.2617 11.62 10.2664 11.77 10.2710 11.92 10.2760 12,06
0.2480 13.59 10.2525 13.77 10.2571 13.94 10.2617 14.12 10.2665 14.29 10.2713 14.47
0.2425 15.84 10,2470 16.05 10.2516 16.25 10.2562 16.46 10.2610 16.67 10.2657 16.87
0.2361 18.09 10.2407 18.32 10.2452 18,56 10.2499 18,80 10,2545 19.04 10.2593 19,28
0.2288 20.32 10.2333 20.59 10.2379 20.86 10.2425 21.14 10.2472 21.41 10.2520 .21 6S
0.2205 22.55 10.2250 22.85 10.2296 23.16 10.2342 23.46 10.2389 23.77 10.2437 24.08
0.2112 24.76 10,2158 25.10 10.2203 25.44 10.2249 25.79 10.2295 26.13 10,2843 26.47
0 2010 26.97 10.2055 27.34 10,2099 27.72 10.2146 28.10 10,2191 28.48 10,2239 28.86
0 1897 29.15 10.1941 29,57 10.1985 29.99 10.2031 30.40 10.2077 30.82 10.2124 31.24
0.1771 31.32 10.1815 31,78 10. 1s59 32.24 10.1904 32.69 10.1951 33.16 10.1998 33 61
0.1634 33.47 10.1677 33,97 10.1721 34.47 10.1766 34.97 10.1812 35,48 10.1859 35.98
0,1484 35.59 10.1527 36.14 10.1570 36.69 10.1615 37.24 10.1661 37.79 10 1708 38,34
0.1320 37.69 10,1362 38.29 10.1405 38.89 10.1449 39,49 10,1495 40.08 10.1641 40.69
0.1142 39.75 10.1183 40.40 10.1225 41.06 10.1270 41.71 10.1314 42.37 10,1360 43,02
0.0947 41.78 10.0988 42.49 10.1029 43.20 10.1073 43,91 10.1117 44.62 10.1163 45,34
0.0735 43.76 10.0775 44,53 10.0816 45.30 10.0858 46.08 10.0902 46.86 10.0948 47.64
0.0504 45,67 10.0542 46.52 10.0582 47.36 10.0624 48.21 10.0667 49.07 10.0712 49 92
0.0252 47.53 10.0289 48.45 10.0328 49.37 10.0368 50,31 10.0410 51.24 10.0454 S2.18
9.9976 49.30 10. OD11 50,31 10. OG48 51.32 10.0087 52.34 10,0128 53,37 10.0172 54,40
9.9672 50.96 9.9706 52.07 9.9741 53.19 9.9778 54.31 9.9818 55.44 9.9860 56,68
9.9340 52.49 9.9370 53.72 9.9403 54.95 9.9438 56,20 9.9476 57,45 9.9517 58,71
9.8971 53,86 9.8998 55.22 9.9028 56.59 9.906a 57.97 9.9096 59.31i 9.9135 60,77
9.8562 55.00 9.8582 56.51 9.8609 58.05 9.8638 59.60 9. S670 61.16 9.8706 62.74
9.8103 55.84 9.8119 57,54 9.8138 59,27 9.8162 61.02 9.8190 62,79 9.8223 64,58
9.7687 56.26 9.7593 58.19 9,7604 60.16 9.7620 62.17 9.7642 64.20 9.7670 66.25
9.7000 56.08 9.6992 58.29 9.6990 60.57 9.6996 62, s9 9.7008 65.26 9,7029 67.66
9.6326 55.00 9.6297 57.56 9.6275 60.22 9.6263 62.96 9.6261 65.77 9.6269 68.65
9.5546 52.49 9.5481 55.48 9.5426 58.64 9,5381 62.02 9.5354 65,37 9.5341 W.92
9.4647 47.60 9.4520 51.08 9.4404 54.85 9.4204 58.91 9.4222 63.23 9.4164 67,81
9.3645 38.61 9.3408 42.41 9.3174 46.75 9.2951 51.68 g, q74g 57.25 9,2581 63,46
9.2697 22.93 9.2285 26.00 9.1824 29.84 9.1349 34.71 9,0834 40W 9,0315 4g ,16
9.2259 0.03 9.1728 0.03 9.1107 0.02 9,0264 0.00 8,9445 0.60 8.8244 O.w
344 APPENDI X B
TABLE B. I.COOIUMNATES OF PoIms ON ~EE TEREE-BAR-LINKACIE
0.02
x,
o
6
10
15
20
25
30
35
40
45
50
65
00
65
70
75
30
85
90
95
100
105
110
115
120
125
130
135
146
145
1.54
155
166
165
170
175
lea
1P+ 10
degrees
10.2912 0.00
10.2908 2.44
10.2895 4.88
10. 2S74 7.33
10.2844 9,77
10.2808 12.21
10.2761 14.65
10.2706 17.08
10.2641 19.52
10. 256S 21.95
10.2485 24.39
10.2391 26.81
10.2287 29,24
10.2172 31.66
10.2046 34. 0s
10.1907 36.49
10.1755 38.89
10.1339 41.29
10.1407 43,68
10,1209 46.06
10,0994 48.43
10,0758 50.78
10,0500 32.12
10.0216 55,43
9.9904 57.72
9.9560 59,97
9.9177 62.17
9,8747 64.32
9,8261 66.38
9.7705 68.33
9,7058 70.09
9.6289 71.57
9.5344 72.56
9.4134 72.58
9,2466 70.26
8.9841 59.70
8.6532 0.00
NOMOGRAM (ConL)
0.01
7s
P + 10 ~egrw
10.2960 0.00
10,2956 2.47
10.2944 4.94
10.2923 7.41
10.2894 9.8a
10. 2S57 12,35
10.2s10 14,82
10.2755 17.29
10.2691 19.76
10.2617 22.23
10.2533 24.69
10.2440 27.16
10.2336 29.62
10,2221 32.08
10.2094 34,54
10.1946 36.99
10.1803 39.45
10.1637 41.90
10.1456 44.34
10.1258 46.78
10.1042 49.21
10.0797 51,64
10.0547 54,06
10.0264 56.47
9.9951 58.86
9.9606 61,23
9.9221 63.59
9,8791 65.91
9,8203 68.19
9.7745 70.41
9.7095 72.54
9.6320 74.53
9.5367 76.27
9.4134 77,50
9.2401 77,51
8.9502 72,73
8.3572 0.00
0.00
I
v!
p + 0 d.grew
10.3010 0.00
10.3006 2.50
[0. 2994 5.00
IO. 2973 7.50
[0. 2944 10.00
10.2906 12,50
10.2800 15,00
[0.2805 17.50
10.2740 20.00
10.2667 22.50
10. 25S3 25,00
10.2490 27.50
10. 23S6 30.00
10,2271 32.5o
10.2144 35,00
10.2005 37.50
10, 1s53 40.00
10,1687 42.50
10,1505 45.00
10,1307 47.50
10,1091 50.00
10. 0s55 52.50
10.0596 55,00
10.0313 57.50
lo, ooml 60.00
9,9654 62.50
9.9270 65.00
9.8829 67.50
9.8351 70.00
9.7792 72.50
9.7140 75. cm
9.6364 77.50
9.5407 20,00
9.4167 S2.30
9.2413 S5.00
s ,9407 87,50
.
90.00
0.01
I
10.3060 0.00
10.3056 2 53
10.3044 5,06
10.3023 7.59
10.2994 10.12
10,2957 12.65
10.2910 15.1s
10.2855 17.71
10.2791 20.24
10,2717 22.77
10.2633 25.31
10.2540 27. S4
10,2436 30.38
10.2321 32,92
10,2194 35.46
10.2046 3s.01
10.1903 40.55
10.1737 43.10
10.1556 45.66
10. 135s 4S.22
10,1142 50,79
10,0897 53,36
10.0647 55.94
10.0364 5s .54
10.0051 61.14
9.9706 63.77
9.9321 66.41
9,8891 69,09
9. S403 71. s1
9.7s45 74.59
9.7195 77,46
9.6420 80.47
9.5467 S3.73
9.4234 S7,50
9.2501 92,49
s. 9603 102 27
S, 3672 180.00
0.02
I
P+1O
degreem
0.3112 0.00
0,310s 2.M
0.3095 5.12
0.3074 7,67
0.3044 10,23
0.3008 12.79
0.2961 15.35
0.2906 17,92
0.2841 20. 4s
0. 276S 23.05
0.2685 25.61
0.2591 2s .19
0.24S7 30,76
0.2372 33.34
.0.2246 35.92
0.2107 3s,51
0,1955 41.11
.0.1789 43.71
0.1607 46.32
.0. 1409 4s ,94
.0.1194 51,57
.0.095s 54.22
.0.0700 56.83
0.0416 59.57
.0.0104 62. 2S
9.9760 65.03
9.9377 67.83
9. S947 70.68
9.8461 73.62
9.7905 76.67
9.725S 79.91
9,6489 S3 .43
9.5544 S7 ,44
9.4334 92.42
9.2660 99,74
9.0041 115,30
8.6732 180,00
APPENDI X B 345
TABLE Fbl ,COOE01NA~E8 OF POINTS ON ~EE TEEEE-BAR-L1NUC3E
pb=
*
x,
ds-
greea
0
5
10
15
20
25
80
.35
40
45
50
55
60
65
70
75
30
85
90
95
100
105
110
115
lm
125
130
135
140
145
150
155
160
165
170
175
180
0.03
7,
,p + 10 de-
greefl
.0.3163 0,0(
.0.3160 2.5!
.0.3146 5.1
.0.3125 7.71
.0.3097 10.3
.0.3060 12,9.
.0.3013 15.5:
.0.2957 18.1:
.0.2893 m.7:
,0.2820 23,3
,0,2737 25.9
.0,2643 28.5:
10,2539 31.1
10.2424 33.7
.0.2298 36.3$
10.2159 39.0:
10.200s 41.61
10.1341 44.3
10.1660 46.9
10.1463 49.61
10.1248 52.31
!0. 1012 55.0)
10.0754 57.8
10.0472 60,0
10.0160 63.4
9.9817 66.2
9.9425 69.2
9,9006 72.2
9.8523 75.4
9.7970 78.7
9.7320 82.3
9.6369 36.3
9.3641 91.0
9.4464 97.1
9.2881 106.5
9.0615 125.8
8.3544 130,0
NOMOGRAM (Cd.)
0.04 0.05 0.O6 0.07 0.03
78 7. v! m. ?,
~P + 10 de- PP + 10 d- fiP + 10 d- WP + 10 de- PP + 10 de-
grees green greea green green
. . . . . .
10,3215 0.00 10,3268 0.00 10,3321 0.00 10.3374 0.00 10.3429 0.00
10,3211 2.61 10,3264 2.64 10,3317 2.67 10,3370 2.70 10.3426 2.73
10.3199 5.23 10.3251 5.20 10.3204 5,34 10.3358 5.40 10.3413 5.46
10.3178 7,85 10,3220 7,93 10.3283 8.02 10.2337 8,11 10,3392 s. 19
10,3148 10.46 10,3201 10.58 10,3264 10.70 10.3308 10.81 10.3363 10.93
10.3110 13.08 10.3164 13.23 10.2.217 13, a8 10.3271 13.52 10.3326 13.67
10.3065 15.71 10.3117 15.83 10,3171 16.06 10.3225 16,23 10.32W 16.41
10.3010 18.33 10.3062 18.64 10.3116 18.75 10.2170 18.95 10.3226 19.16
10.2945 20.96 10,2W9 21,20 10.3052 21.44 10.3107 21.68 10.3161 21.91
[0.2872 23.69 10.2925 23.36 10.2979 24.14 10.3033 24.41 10.3033 24.63
[0.2739 26.23 10..2342 26.54 10.2896 26.34 10.2950 27.15 10.2005 27.45
10.2695 28.87 10.2749 29.21 10.2303 29.56 10.2853 29.90 10.2912 30.24
10.2591 31.52 10.2646 31.90 10.2699 32.28 10.2755 32.66 10.2810 22.03
10.2477 34.18 10.2531 34.60 10.2335 35,01 10.2641 35.43 10.2697 35.35
10.2351 36.84 10.2404 37.31 10.2459 37.76 10.2515 38.22 10.2571 38.68
10,2212 39.52 10.2266 40.02 10.2221 40.53 10,2377 41.03 10.2434 41,52
10.2061 42.21 10.2115 42.76 10.2170 43.31 10,2227 43,80 10.2234 44.41
10.1895 44.92 10.1949 45.51 10,2005 46.11 10.2062 46.71 10.2120 47.31
10.1714 47.62 10.1770 48.29 10.1825 48.94 10.1333 49.60 10.1942 30,2s
10.1517 50,38 10.1573 51.09 10.1629 51.30 10.1638 52.51 10.1747 52.22
10,1302 53.14 10.1353 53.92 10.1416 34.70 10,1475 55.47 10,1535 66.24
10.1067 55.93 10.1124 56.79 10.1182 57.64 10.1242 33.43 lo.13M 59.83
[0.0810 58.76 10.0863 59.69 10.0928 60,63 10.0939 61.55 10.1062 62.47
10.0528 61.62 10.0537 62,66 10.0648 63.68 10.0711 64,69 10.0776 65,70
10.0218 64.36 10.0278 65.69 10,0341 66.81 10.0406 67.93 10.0472 69,04
9.9876 67.55 9.9938 68.30 10.0003 70,05 10.0070 71.28 10.0140 72.51
9.9496 70,64 9.9360 72.02 9.9628 73.41 9.9608 74.78 9.977 1 76.14
9.9070 73.84 9.9133 75.40 9.9209 76.95 9.9232 78.49 9.9362 30.00
9.8590 77.21 9.3662 78.98 9.8738 30.73 9.8819 82.46 9.8m2 &.16
9.3042 80.30 9.81m 82.33 9.82o4 34.34 9.8293 86.81 9.8387 a8.74
9.7493 84.74 9.7496 87.11 9.7590 39.42 9.7692 91.71 9.78m 98.92
9.6661 39.22 9.6763 92.04 9.8875 04.78 9.6997 97.44 9.7126 100.00
9.5734 94.63 9.3831 97.98 9.6026 101,36 9.6181 104.52 9.6346 107.61
9.4622 101.77 9.4804 106.09 9.6004 110.15 9. 52m 113.92 0.544 7117.40
9.3149 112.75 9.3451 11S,32 9,3774 123.25 9.4103 127.69 9.444 5131.39
9.1234 124.01 9.1349 140.29 9.2434 145.16 9.2935 149.00 9.2497 152.07
8.9845 180.00 9.0364 180.00 9.1707 180.00 9.2428 180.00 9.3058 130.00
346
APPENDI X B
TABLE B.1.COOEDINATES OF POINTS ON THE THREE-BAR-LINKAGE
NT0140C2RAM (COW.)
pb= 0.09
I 010 I 011: 012! 013! 014
x, 1 i-
de- PP+ 10 d:- UP+ 10 d:- !JP + 10 d%- PP + 10 d:- PP+ 10 d:- pp+ loi J-
grees grees grees grees grees Erees g,em
. . . . . . .
01 10.3484 0.00 10.3539 0.00 10.3595 0.00 10.3652 0,00 10.3709 0.00 10.3767 0.00
5. 10 #480 2.76 10.3535 2,79 10,3591 2.s1 10.3647 2.84 10.3705 2.87 10.3763 2.90
10 10.3467 5.52 10.3523 5.57 10.3579 5 63 10.3635 5,69 10.3693 5.74 10.3750 5.80
15 10,3447 8.28. 10.3503 8.36 10.3558 8.45 10.3615 8.53 10.3672 8.62 10,3730 8.70
20 10.3418 11.04 10.3474 11,16 10.3529 11.27 10.3586 11.39 10.3644 11.50 10.3701 11.61
25 10.3380 13.81 10.3437 13.96 10.3493 14.10 10.3550 14.24 10.3608 14.39 10.3665 14.53
30 10.3335 16.58 10.3391 16.76 10.3447 16.93 10.3505 17.11 10.3562 17.28 10.3620 17.45
35 10.3280 19.37 10.3336 19.57 10.3393 19.77 10.3450 19.98 10.3508 20.18 10.3566 20.38
40 10.3216 22.15 10.3273 22.39 10.333O 22.63 10.3387 22.86 10.3445 23.09 10.3504 23.33
45 10.3144 24.95 10.32OO 25.22 10.3258 25.49 10.3315 25.75 10.3373 26.02 10.3432 26.29
50 10 3062 27.76 10.3118 28.06 10.3176 28.36 10.3234 28.66 10.3292 28.96 10.3351 29.26
55 10.2970 30.58 10.3026 30.91 10.3084 31.25 10.3142 31.59 10.3201 31.92 10.3261 32.25
60 10.2867 33.41 10.3924 33.79 10.2982 34.16 m. 3040 34.53 10.3100 34.90 10.3161 35.27
65 10.2753 36.26 10.3811 36.68 10.2870 37.09 10.2929 37.50 10.2988 37.91 10.3050 38,31
70 10.2628 39,13 10.2687 39.59 10.2746 40,04 10.2805 40.49 10.2865 40.94 10.2928 41.38
75 10.2492 42.03 10.2550 42.53 10.2610 43.02 10.2670 43.51 10.2731 44.00 10.2794 44.49
80 10.2843 44.95 10.2402 45.49 10.2462 46.03 10.2523 46.57 10.2585 47.11 10.2648 47.64
85 10.2179 47.90 10.2239 48.50 10.2300 49.08 10.2362 49.67 10.2425 50.25 10.2489 50.83
90 10.2001 50.89 10.2062 51.54 10.2124 52.19 10.2187 52.82 10.2251 53.45 10.2316 54.08
95 10.1808 53.93 10.1869 54.63 10.1932 55.33 10.1997 56.02 10,2062 56.71 10.2128 57.39
100 10.1597 57.01 10.1660 57.78 10.1724 58.54 10.1790 59.29 10.1857 60.04 10.1924 60.78
105 10.1367 60.161 0.1431 61. Oo 10.1497 61.82 10.1564 62.64 10.1634 63.46 10.1703 64.26
110 10.1116 63.391 0.1182 64.30 10.125O 65.20 10.132O 66.09 10.1391 66.98 10.1463 67.86
115 10.0842 66.711 0.0910 67.70 10.0980 68.68 10.1053 69.66 10.1126 70.63 10.1201 71.58
120 10.0542 70.141 0.0612 71.23 10.0686 72.31 10.0761 73.38 10.0838 74.43 10.0916 75.47
125 10.0212 73.721 0.0286 74.92 10.0363 76.10 10.0442 77.27 10.0523 78.43 10.0606 79.56
130 9.9847 77.48 9.9926 78.81 10.0008 80.12 10.0092 81.40 10.0179 82.67 10.0267 83.91
135 9.9443 81.50 9.9528 82.97 9.9616 84.41 9.9707 85.84 9.9800 87.23 9.9896 88.59
140 9.8992 85.84 9.4085 87.48 9.9182 89,09 9.9281 90.66 9.9384 92.20 9.9490 93.70
145 9.8486 90.63 9,8590 92.48 9,8698 94.28 9.8810 96.03 9.8926 97.73 9.9045 99.38
150 9.7914 96.07 9.8034 98.16 9.8159 100.18 9.8288 102.13 9.8421 104.00 9.8557 105.81
155 9.72641 02.47 9.74071 04.84 9.7557 107.11 9.7710 109.27 9.7868 111.33 9.8029 113.28
160 9.65201 10.35 9.67011 13.03 9.6887 115.54 9.7077 117.90 9.7270 120.11 9.7464 122.18
165 9.56821 20.61 9.5921 123.54 9,6164 126.24 9.6407 128.70 9.6649 130.95 9.6890 133.01
170 9.4781 134.72 9.51131 37.65 9.5436 140.22 9.5752 142.49 9.6058 144.51 9.6355 146.30
175 9.39761 54.58 0.44221 56.65 9.4839 158.38 9.5234 159.86 9.5603 161.12 9.5955 162.22
180 9.36221 80.00 9.4132 180.00 9.4598 180.041 9.5028 180.00 9.5428 180.00 9.5802 180.00
APPENDI X B 347
TABLE B.1.COOEDINATES OF POINTS ON IHE THREE-BAR-LINKAQE
NOMOGRAM (Cont.)
Lb = 0.15
I
0.16
x,
de-
grees
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
135
140
145
150
155
160
165
170
175
180

I I
v. 7,
.p + 10 de- pp + 10 de-
.srees greem
10 3825 0.00 10.3884 0.0(
10,3820 2.93 10.3880 2.9(
[0,3809 5.86 10.3868 5.91
10 3788 8.79 10.3847 8.w
10.3760 11.73 10,3819 11.8<
10.3724 14.67 10.3783 14.81
10 3679 17..62 10,3739 17.7!
10.3626 20.59 10,3685 20. 7!
10.3564 23.56 10,3623 23.7:
10 3492 26,55 10.3552 26.81
10.3412 29.56 10.3472 29.85
10.3321 32.59 10.3383 32.92
[0.3221 35.64 10.3284 36.0(
[0.3111 38.72 10.3174 39.11
LO.299o 41,83 10.3053 42.27
LO.2857 44.98 10.2920 45. 4e
LO.2712 48.16 10,2777 48.69
10.2554 51.41 10,2620 51,9E
.0.2382 54,70 10.2449 55.32
0,2196 58.07 10,2265 58.74
0.1994 61.52 10.2065 62.25
0.1774 65.06 10,1847 65.85
0.1536 68,721 0,1612 69.58
0.1278 72.531 0.1357 73.46
0.0997 76.501 0,1079 77.51
0,0691 80.681 0,0778 81,78
0.0359 85.141 0.0451 86.33
9. 9!035 89.931 0.0096 91.24
9.9598 95,16 9,9710 96,69
9.91661 OC.97 9.92911 02.52
9.86971 07.54 9,88391 09,21
9.81921 15.14 9.83581 16.91
9.76601 24,12 9.78571 25.93
9.71291 34,91 9.73631 36.64
9.66431 47.90 9,69221 49.34
9,62891 63.18 9.66051 64.02
9.61551 80,00 9.64881 80. OC
0,17
I
0.18
I 019 I 020
I
I I II I-T
1% l,! Id in
P+lo de- PP+ 10 de- MP+ 10 de- PP + 10 dn-
greea grees ~reea
&w
D.3943 0.00 10.4003 O.oa 10.4063 0.00 10.4124 0 00
D.3938 2.98 10.3999 3.01 10.4060 3.04 10.4121 3.07
3.3927 5.97 10.3989 6,02 10.,4047 6.08 10.4109 6.13
D,3907 8,96 10.3967 9.04 10.4028 9.12 ,10. 40R9 9.21
). 3879 11.95 10.3939 12,0G 10.3999 12. IT 10.406 L2,28
). 3843 14.95 10,3903 15.09 10.391M 1$.2% ao. a
~
15,37
1.3798 17,96 10.3859 18,13 10,,3921 18,30 10.39
18.47
1,3745 20,99 10.3806 21,19 10.3868 21. ~ 10.393C) 21.58
). 3684 24,02 10.3745 24.25 m. 3807 24. 4@dO .3869 24,71
).3614 27.08 10.3675 27.34 LO-3738 27,59 10, 38w ~7 .85
1,3534 30.15 10,3596 30,44 10,3659 30.73 10.3722 31. o2
). 3444 33.24 10.3507 33.57 10.3570 33.89 10.3634 34.22
). 3345 36.37 10.3409 36.73 10.3472 37.09 10.3536 37,44
).3236 39.52 10.3301 39.91 10,3364 40.31 10.3429 40.70
).3116 42.71 10.3181 43.14 10.3246 43,57 10.3312 44,00
). 2985 45.93 10.3050 46.41 10.3116 46.88 10,3184 47.35
), 2842 49.21 10.2909 49.73 10.2975 50,24 10,3044 60,75
). 2686 52.54 10.2754 53.10 10,2822 53.66 10.2892 54,21
).2517 55,94 10,2586 56.55 10.2656 57.15 10.2728 57.75
). 2334 59,41 10,2405 60.07 10.2476 60,72 10.2549 61,36
).2136 62.97 10.2209 63.68 10,2282 64.39 10.2358 65.09
),1921 66.64 10<1996 67,41 10.2072 68.18 10.215O 68.93
) .1688 70.43 10.1766 71,27 10.1845 72.09 10.1926 72,91
). 1436 74.38 10.1518 75.28 10,1600 76.17 10,1684 77,05
).1163 78,51 10.1249 79.49 10.1336 80,45 10,1424 81 40
). 0868 82,87 10.0958 83.93 10.0151 84.97 10.1144 85.99
). 0547 87.51 10.0644 88.66 10.0743 89.79 10.0843 90.,88
),0199 92.51 10:0305 93,76 10,0412. 94,97 10,0520 96.15
) .9823 97,97 9,9939 99.32 10.0056 100.61 10,0175 101.87
).9417 104.01 9.9546 105.44 9.9676 106,83 9.9808 108.16
) .8983 110,80 9.9129 112.33 9.9275 113,78 9,9424 11.5.18
). 8524 118.58 9.8692 120.16 9.8860 121.66 9,9028 123.08
).8052 127,62 9.8248 129.21 9 8442 130.69. 9,8634 132.07
) .7596 138.24 9.7822 139.70 9,8047 141.013 %8267 142.31
).7192 150.64 9.7456 151,82 9.7710,152.89 9.7957 153.86
). 6908 164,77 9,7200 165.44 9.7479166.04 9..7746 166.58
). 6804 180,00 9,7106 180.00 9.73941180.00 9. 7G7J 18Q.00
348 APPENDI X B
TARLE B.1.COORDINATES OF POINTS ON THE THREE-BAR-LINKAGE
NOMOGRAM (Cont.)
pb= 0.21 0.22 0.23 0.24 0.25 0.26
x,
?, v. 78 ?1 v,
de. pp + 10 de- )@+ 10 de. pgl + 10 de- pp+ 10 de. @p + 10 de. tip + 10 d:-
grew green gree8 greem
greea
grees greem
. . . . _ _ _ _

0 10.4186 0.00 10.4248 0,00 10,4311 0.00 10.4374 0.00 10.4438 0.00 10 4502 0.60
5 10 4182 3.09 10.4245 3,12 10.4306 3.15 10.4370 3.17 10.443.5 3 20 10 4498 3.23
10 10.4171 6.19 10.4233 6.24 10.4295 6.3o 10.4359 6.35 10,4423 6.40 10 4486 6,45
15 10.4151 9,29 10.4213 9,37 10.4276 9.45 10.4340 9.53 10.4403 9.61 10, 446!3 9,69
20 10.4123 12.39] 10,4185 12.50 10,4249 12.61 10.4312 12.72 10.4376 12.83 10.4441 12.93
25 104088 15.51 10.4151 15.65 10.4214 15.78 10.4278 15.92 10.4342 16.05 10.4407 16.19
30 10.4044 18.64 10.4108 18.80 10.4171 18.97 10.4235 19.131 0.4300 19.29 10.4365 19.45
35 ,10.3992 21.781 0.4056 21.971 0.4119 22.161 0.4184 22.361 0,4249 22.55 10 4315 22.74
40 10.3932, 24 931 0.3996 25.161 0.4060 25.38.1 0.4125 25.601 0.4190 25.821 0.4256 26.04
45 10.3863 28.111 0.3928 28.371 0.3992 28,621 0.4057 28.871 0.4123 29.121 0.4190 29.37
50 /10.3786 31.31 10.3850 31.601 0,3915. 31.881 0.3981 32.161 0.4047 32.441 0.4115 32.7z
55 10 3698 34.541 0.3763 34.861 0,3829 35.171 0.3896 35.491 0.3963 35.801 0.4030 36.11
60 10 3602 37.801 0.3667 38.151 0,3734 38,501 0.3801 38.841 0.3870 39.191 0.3938 39 53
65 10.3496 41.091 0.3562, 41.481 0.3629 41.861 0.3698 42.241 0,3767 42.621 0,3836 42,99
70 10.3379 44.431 0.3446 44.851 0,3515 45.271 0.3584 45.691 0.3654 46.101 0 3724 46,51
75 10.3251 47.811 0.3320 48.271 0.3391 48.741 0.3460 49.181 0.3531 49,631 0.3602 50.07
so 10.3113 51.261 0.3183 51.761 0.3254 52.251 0.3325 52.741 0.3397 53.231 0.3470 53 71
85 ~10.2462 54.761 0,3034 55.301 0.3106 55.841 0.3179 56.371 0,3253 56.901 0.3327 57.42
90 10,2800 58.341 0.2873 58.931 0.2946 59,51 10.3021 60.0/3 10,3097 60.651 0.3173 61,21
95 10.2624 62.011 0.2698 62,641 0.2774 63.271 0.2851 63.891 0.2929 64501 0.3007 65,10
100 10 2434 65.781 0.2511 66.471 0.2589 67.141 0, 266s 67.611 0,2748 68.461 0.2829 69,31
105 10.2228 69.691 0,2308 70,41 10,2389 71.141 0.2471 71,851 0,2554 72 561 0,2637 73,25
110 10.2008 73.711 0.2091 74.501 0.2175 75,281 0.2260 76.051 0.2346 76, sol 0.2433 77.55
115 10.1770 77.92 10.1857 78.77 10.1945 79.61 10.2034 W.43 lo.21z4 81.24 lo. z215 82,03
120 10.1514 82.33 10.1606 83.25 10,1698 84.14 10.1792 85, oz 10.1396 85.38 10,1983 86,73
125 10,1240 86.99 10.1337 87.97 10.1435 8E.93 10,1534 89,87 10,1635 90.79 10,1736 91.68
130 10,0946 91.96 10.1049 93.01 10.1154 94.03 10.1259 95. o2 10.1367 96.00 10.1475 96,94
135 10,0630 97.26 10.0742 98,41 10.0855 99,49 10. I3969 100.55 ]0, 1084 10].57 10. 12M 1o2,56
I
140 10.0295 103,09 10.0417 104.27 10,0540 105.41 10,0664 106.52 10,0788 107,58 1!3.09]4 108,131
145 9.9941 109.45 10.0076 110,69 10.0211 111.88 10,0346 113.02 10.0483 114,12 10,0620 115.18
150 9.9572 116.51 9.9723 117.79 9,9871 119.00 10. CKI22120.16 10,0171 121.27 10. O322 122,33
155 9.9196 124.43 9.9365 125.71 9,9531 126.91 9.9698 128.06 9,9864 129.14 10. I3I3$!813(3. 17
160 9.8825 1$3.37 9.6015 134.59 9.9201 135,73 9.9386 13(3.80 9,9569 137,81 9,9750 138,76
165 9.8483 143.47 9.8695 144,54 9.8903 145,54 9.9107 146,46 9,9308 147.33 9.9506 148,14
170 9.8197 154.75 9.8431 155.57 9,8659 156.32 9.8880, 157,01 9.9099 157,66 9,9308 158,25
175 9.8005 167,07 9.8252 167,51 9.8497 167.92 9.8733168.30 9.8959 168.64 9.9182 168.96
MO 9.7937 180,06 9,8193 180.00 9.8440 180.6Q 9.8679180.00 9.8911 180,00 9.9137 lW,00
I
1 ) 1 1 1 I I I I I I
APPENDI X B 349
TABLE B.1.COORDINATES OF Ponwa ON TEE TEREE-BAR-LINUGE
NOMOGEAM (Cont.)
@-
0.27 0.28 0,29 o.ao 0.31 0.82
I
x,
de- #p + 10 d 14p+ 10 dq: #p + 10 d #p + 10 dm: IJp+ 10 ::. @p+ 10 d%
grew Creea
m
#rem grew greed areea
. . . . _
o 10.4567 0.00 10.4632 0.00 10.4698 0,00 10.4764 0.00 10.4831 0.00 10.48W 0.00
5 10.4562 3.25 10.4629 3.28 10,469.5 3.31 10.4760 3.22 10.4828 3.36 10.4895 3.23
10 10,4562 6.51 10.4618 6.56 10.4684 6,61 10.4749 6.66 10.4816 6.72 10,4234 6.77
15 10.4533 9,77 10.4598 9.85 10.4664 9.93 10.4731 10.00 10.4798 10.08 10.4866 10,16
20 10.4506 13.64 10.4572 13. lb 10,4628 13.2510.4706 13,2510,4773 13.46 10,4341 13,56
25 10,4472 16.3210.4528 16.45 10.4605 16.5810.4672 16.7110.4740 16,& 10.4808 16.97
30 10.4420 19.61 10.4497 19.77 10.4504 19.9310.4632 20.09 10.4699 20.24 10,4763 20.40
35 10,4380 22.93 10.4447 23.11 10.4515 23.30 10.4382 23.48 10.4651 23.66 10.4720 23. M
40 10,4323 26.26 10.4290 26.47 10.4457 26.63 10.4526 26.90 10.4595 27.11, 10.4664 27.31
45 10.4257 29.61 10.4325 20.86 10.4393 30.10 10.4462 30.34 10.4321 30.57 10.4601 26.81
30 10.4182 32.99 10.4251 33.27 10.4320 33.64 10.4238 22.81 10.4459 34.07 10.4530 34.34
.55 10.4099 26.41 10.4168 36,72 10,4238 37.02 10.4308 37.31 10.4279 37.61 10.445a a7.80
60 10.4007 a9.87 10.4077 40.20 10.4147 40.52 10.4218 40.86 10.4290 41.18 10.4262 41.51
65 10.a906 43.a6 10.3977 42.7a 10.4048 44.OB10.4120 44.45 10.419a 44.81 10.4266 46.16
70 10.3795 46.91 10.2367 47.31 lo. a9a9 47.71 10.4013 48.10 10.4687 4a. 49 10.4161 48.87
75 10.a675 30.51 10.3747 50.95 lo. aazl 51.a8 10.a896 61.80 10.a971 52.22 10.4647 52.64
30 10.3544 54.18 10.3618 64.65 10.a694 6.5.12lo. a769 55.53 10.2346 36.03 10.2822 56.4a
83 10.2403 67.93 lo. a478 58.44 10.3536 63.04 lo.362a 58.43 10.3711 58.92 10.2780 60.41
90 10.3250 61.76 10.3328 62.al 10.2407 62.85 10.2487 63.22 lo. awa 63.91 10.2047 64.42
95 10.3086 65.70 10.3166 66.2a lo. a2.47 66.85 lo. aaae 67.43 lo. a412 67.90 10.2496 6a,5a
lcm 10.2911 69.75 lo.2e9a 70. a8 10. ao77 70. ae 10.3161 71.60 10.3247 72.2o 10.2232 72.79
105 10. 272a 73.93 10.2308 74.60 10.%95 75.26 10.2982 7s.91 10.3670 78.65 Io. abw 77.18
110 10.2522 78. 2a 10.2611 78.99 10.2701 70.70 10.2792 ao. a9 lo. zaaa al.oa 10.2976 al.7s
116 10. 2a07 82.80 10.2400 83.67 10.2494 34.31 10.2389 86.05 10. 2W6 36.76 10.2781 86. M
Ma 10.2079 87.55 10.2177 aa. a6 10.2276 89.15 10.2375 89.92 10.2476 90.68 10.2576 91.41
125 10. maa 92.55 10.1942 93.41 10.2046 94.24 10.2151 96.06 10.2257 95.34 10.2262 96.61
lao 10.1534 97.86 10.1694 98.76 10.1805 88.62 10.1916 100.47 10.2028 101.26 10.2141 102.10
las 10.1317 102.52 10.1425 104.46 10.1553 106.36 10.1672 106.24 10.1791 107.09 10.1911 107.91
140 10.1641 108.61 10.1167 110.67 10.1294 111.60 10.1421 112. ag 10.1549 lla.26 10.1677 114.@4
145 10.0757 116.19 10.0894 117.17 10. loal 118.10 10.1168 119.00 10.1266 119.37 10.1442 120.70
km 10.6471 12a.24 10.0620 124.31 10.6769 125.22 10.0917 126.17, 10.1064 126.86 10.1211 127.77
153 10.0192 131.15 10.0332 132.07 10.0515 132.96 10.0675 122.79 10.0824 124.32 10.0091 las. w
160 9. wao 139.66 10.0105 140.&J 10.0179 141.29 10.0452 142.06 10.0622 142.76 10.0789 142.42
165 9.9698 148.89 9.9882 149,60 10.0076 150.26 10.0260 15).89 10,0429 151.47 10. O6U3 152.02
170 9.9517 138.81 9.0717 169, a2 9.9916 159.30 10.0109 160.26 10.0298 160.67 10.0486 161.07
175 9. 9a96 169.25 9.9604 169.62 9.981a 169.78 10.0010 170,01 lo.021a 170.24 10.0400 170.44
184 9.9356 126,00 9.9369 180.00 9.9777 180.00 9.9979 Iao. cm 10.0178 130.00 10.0372 180.00
350 APPENDI X B
I
TABLE B.1.COORDINATES OF POINTS ON THE THREE-BAR-LINKAGE
NOMOGRAM (Cont.)
I
7 ~
@ = 0.33 0.34 0.35 0.36 0.37 0.38
x,
7. v, 7. 73 7, 7<
de- pp + 10 de- PP + 10 de- pp + 10 dc- fip + 101de- ,up + 10 d,- J,p + 10 de-
grees grees gre.s grces
srees
grees gree,
0 10 4967 O.otl 10.5035 0.00 10,5104 0.00 10.5173 0.00 10,5243 0,00 10.5313 0.00
5 10,4962 3.41 10.5031 3.43 10.5101 3.46 10,5170 3.4s 10.5240 3 51 10,5309 3.53
10 10.4952 6.S2 10.5021 6.87 10.6090 6.92 10.5159 6.97 10,5229 7.01 10.5299 7.06
15 10.4934 10.23 10.5OQ3 10.31 10.5072 10.3s 10.5141 10.46 10.5211 10,53 10.5282 10.60
20 10.4909 13.6610.4978 13 76 10.5047 13 86 10.5117 13,96 10.5186 14.05 10 5258 14.15
25 10. 4s77 17.1010.4946 17.22 10.5015 17.35 10.5085 17.47 10.5155 17.59 10.5227 17.71
30 10.4837 20.5510.4906 20.70 10.4976 20.85 10.5046 21.00 10.5117 21.15 10.51s9 21,29
a5 10.4789 24.0210. 4S59 24.20 10.4929 24.38 10,5000 24.55 10.5071 24.72 10.5144 24. S9
40 10.4734 27.5210.4805 27.72 10.4875 27.93 10.4947 28.13 10,5018 2S. 32 10.5092 28.52
45 10.4671 31.04 10.4743 31.27 10.4814 31.50 10.4886 31.73 10.4958 31,95 10,5032 32.17
50 10.4601 34.60 10.4673 34. S6 10,4745 35.11 10.4817 35.37 10.4s91 35.62 10.4965 35. S6
55 10,4522 38.19 10.4594 3S .48 10. 466S 3S.76 10.4741 39.04 10.4816 39,32 10.4s90 39 59
60 10 .4435 41.83 ,10.450s 42.14 10.4582 42.45 10.4657 42.76 10,4732 43,07 10. 4S08 43.37
65 10.4340 45.51 10.4414 45,85 10.4489 46.19 10.4505 46,53 10.4641 46, S6 10.4718 47,19
70 10.4236 49.25 10,4312 49.62 10. 43s8 49,99 10,4465 50,36 10.4542 50.72 10.4620 51,0s
75 10.4123 53.05 10.4200 53.46 10.4278 53. S6 10.4356 54.25 10.4434 54.64 10.4514 55 03
80 10.4001 56.92 10.4080 57.36 10.4159 57.79 10.4239 5s. 22 10.4320 58,64 10,4400 59,05
85 10. 3s70 60. SS 10.3950 61.35 10.4031 61.81 10.4113 62.27 10.4195 62,72 10.4276 63 16
90 10.3729 64.93 10.3812 65,44 10.3s95 65.93 10.3979 66.42 10.4063 66.90 10.4148 67.37
95 10.3579 69.09 10.3664 69.63 10.3749 70.16 10,3835 70.67 10.3922 71.18 10.4009 71 69
100 10.3419 73.37 10.3506 73.94 10.3594 74.50 10.3683 75.06 10.3772 75,6o 10, 3S62 76 13
105 10.3249 77.80 10.3339 7s.40 10.3430 78.99 10.3522 79. 5s 10.3614 80.15 10.3707 80,71
110 10,3069 82. 3S 10.3163 83,02 10,3257 83,65 10.3352 84.26 10.3448 S4 .86 10.3544 85,45
115 10.2878 87.15 10.2976 87,82 10,3075 8S .48 10.3174 S9.12 10.3272 89.75 10.3373 90 36
120 10.2678 92.13 10.27S0 92.84 10. 28S4 93.52 10. 29s7 94.19 10.3091 94.85 10.3195 95, 4s
125 10.2470 97.36 10.2577 98.09 10, 26S6 98.81 10.2794 99.30 10.2902 100,17 10.3012 10&83
130 10,22531 02.87 10,2367 103.63 10.2481 104.36 10.2595 105.07 10.2709 105.76 10.2824 lo&43
135 10,20301 08,70 10.2151 109,47 10.2271 110.22 10.2392 110.94 10.2513 111,64 10.2634 112.32
140 10.18051 14.90 10.1932 115,67 10.2060 116.42 10.21s8 117.14 10,2315 117.841 0.2442 118.51
145 10.15791 21.5a 10.1714 122.26 10.1850 123.00 10.1986 123.71 10.2121 124,391 0.2254 125.04
130 10.13571 28.54 10.1502 129.28 10.1647 129.99 10.1790 130.661 0.1933 131.311 0.2074 131.93
1s5 10.11471 36.06 10,1302 136.751 0.1455 137,401 0,1607 13s.03 10.1757 138.621 0.1907 139.19
160 10.09551 44.07 10.1119 144.671 0.12s1 145.251 0.1442 145.801 0.1599 146.311 0,1758 146.81
165 10.07931 w .55 10.0966 153.041 0.1137 153.511 0.1303 153.95 10.1468 154,361 0,1631 154,76
170 10.00701 61.44 IO. 0850 161.791 0.1026 162.121 0.1198 162.43 10.1369 162.721 0.1537 163.00
175 10.05851 70.63 10.0775 170.821 0:0955 170.991 0.1132 171.15 10.1306 171.301 0.1478 171.44
180 10.05611 80.00 10.0747 180.001 0.0930 180. OJ 10.1109 180. WI 10.1285 180. C$310.1458 180.00
APPENDI X B 351
TARLE B.1.COOIWINATES OF POINTS ON THE TIIREE-BAR-LINKAGE
NOMOGRAM (Cont.)
p>- 0.39 0,40 0.41 0.42 0.43 0.44
x, ?. 79 7s 7? ?9
de- pp + 10 d:- pp + 10 de- pp + 10 de- Pp+ 10 de- PP + 10 de- PP + 10 de-
grees grees grees grees greea grees grees
0 10.5384 0.00 10,5455 O.OQ10.5527 0.00 10.5599 0.00 10.5672 0.00 10.5745 0.00
5 10.5381 3.55 10.5452 3.58 10.5524 3.60 10.5596 3.62 10.5669 3.65 10.5742 3.67
10 10.5371 7.11 10.5442 7.16 10.5514 7.2o 10.5586 7.25 10.5659 7.30 10.5732 7.34
15 10.5354 10.67 10.5425 10.74 10.5497 10. s1 10.5570 10.88 10.5643 10.95 10.5716 11.02
20 10.5330 14.25 10.6401 14.34 10.5474 14.43 10.5547 14.53 10.5620 14.62 10.5694 14.71
25 10.5299 17.83 10.6371 17.95 10.5444 18.07 10.5517 18.19 10,5590 18.30 10.5664 18.41
30 10.5261 21.44 10.5334 21.58 10.5407 21,72 10.5480 21.86 10.5554 22.00 10.5628 22.14
35 10.5216 25.06 10.5290 25.23 10.5363 25.4o 10.5437 25.56 10.5511 25.72 10.5586 25.88
40 10.5164 28.71 10.5238 28.91 10.5312 29.10 10.5387 29.28 10.5462 29.47 10.5537 29.65
45 10.5105 32.39 10.5180 32.61 10.5254 32.82 10.5330 33.04 10.5405 33.25 10.5481 33.45
50 10. W39 36.11 10.5115 36,35 10,5190 36.59 10.5266 36.83 10.5342 37.06 10.5419 37.29
55 10.4965 39.86 10.5042 40.13 10.5118 40.40 10.5194 40.66 10.5272 40.92 10.6349 41.17
60 10.4884 43.66 10.4961 43,96 10.5039 44.25 10.5116 44.54 10.5194 44.82 10.5273 45.10
65 10.4795 47.52 10.4873 47.84 10.4952 48.15 10.5031 48.47 10.5110 48.77 10.5191 49,08
70 10,4700 51.43 10.4778 51,78 10.4858 52.12 10.4938 52.46 10.5019 52.79 10.5101 53.12
75 10,4595 55.41 10,4675 55.78 10.4756 56,15 10.4838 56.47 10.4920 56.87 10.5003 57.23
so 10.4482 59.46 10.4564 59.86 10.4647 60.26 10.4731 60.65 10.4814 61.03 10.4899 61.41
85 10.4362 63.60 10.4446 64.03 10.4530 64.45 10.4616 64.87 10.4701 65.28 10.4787 65.68
90 10,4234 67.84 10.4320 68.29 10.4406 68.74 10.4493 69.18 10.4581 69,62 10,4669 70.05
95 10.4097 72.18 10.4185 72.66 10.4274 73.14 10.4363 73.61 10.4454 74.07 10.4544 74.52
100 10.3953 76.65 10.4043 77.16 10.4135 77.66 10 4227 78.16 10.4220 78,64 10,4412 79.11
105 lo. Wm 81.26 10.3894 81.79 10.3988 S2 ,32 10.4684 82.84 10.4179 83,34 10,4274 83.84
110 10.3640 86.02 10.3738 86.58 10.3835 87.13 10.3934 87.67 10.4032 88,20 10,4131 88.72
115 10.3473 90.96 10.3575 91.55 10.3676 92,12 10.3778 92.68 10.3880 93.23 10.3982 93.76
120 10.23IXI 96.10 10.3405 96.71 10,3511 97.30 10.3617 97. ss 10.3723 98.44 10.3829 98.98
125 10.3122 101.47 10.3231 102.09 10.3341 102,69 10.3452 103.28 10.3563 103.85 10.3673 104.41
130 10.2939 107.08 10.3054 107.71 10,3169 108.33 10.3284 108,92 10.3400 109.50 10.3515 110, O6
135 10.2754 112.97 10.2875 113.61 -10.2995 114.22 10.3115 114.81 10.3236 115.39 10.3356 115.94
140 10.2570 119.16 10.2697 119.79 10. WJ23 120.39 10.2949 120.98 10.3075 121.54 10.3201 122.08
145 10,2389 125.68 10.2522 126.28 10.2655 126.86 10. n,88 127.43 10.2919 127.97 10,3050 128,49
130 10.2216 132.52 10.2336 133.10 10.2495 133:65 10.2625 .124.18 10.2771 134.68 10.2908 135.17
155 10.2055 139.73 10.2202 140.25 10.2247 140.75 10.2493 141.22 10.2635 141.68 10,2778 142.12
lea 10.1912 147.28 10.2066 147.73 10.2217 14$.15 10.2365 148.66 10.2515 148.95 10.2663 149,33
165 10.1794 155.14 10.1953 155.50 10.2109 155. & 10.2264 136.16 10.2418 156,47 10.2569 156,76
170 10.1702 163.26 10. 1.%7 163.51 10.2028 162.75 10.2190 162.98 10.2345 1s4. 19 10.7,500 164,40
175 10.1647 171.53 10.1814 171.71 10.1977 171:23 10.2142 171.95 10.2300 172,0S 10.2456 172.16
180 10.1628 180.00 10.1795 lso. om 10.1960 180.60 10.2123 130.00 10, 22s3 lSO. IXI 10,2441 180,1J3
352 APPENDI X B
TABLE B.1.COORDINATES OF POINTS ON TTIE THREE-BAR-LINKAGE
NOMOGRAM (Cd.)
Mb-
0.45
I
0.46
~
0.47
I
0.48
xl I.ll l! h In.
de- Wp+ 10 di- p + 10 de- pp + 10 de- Wp + 10 di
m
gre.m green grees grem
o 10.5819 0,00 10.5893 0.60 10.5967 0.00 10.6042 0.01
5 10.5816 3.69 10.5890 3.71 10.5964 3.74 10.6039 3.71
10 10.5806 7.39 10. 58s0 7,43 10.5955 7.47 10.6030 7.5:
15 10.5790 11.09 10.5864 11.15 10.5939 11.22 10,6015 11.21
20 10.5768 14.80 10.5842 14.89 10.5917 14,98 10.5993 15.0(
25 10,5739 18.53 10.5814 18.64 10.5889 18,75 10.5965 18..9
30 10.5703 22.27 10.5779 22.41 10.5855 22.54 10.5931 22.6
35 10.5$61 26.04 10.5738 26.20 10.5814 26.35 10.5891 26.51
40 10,5613 2s.83 10.5690 30.01 10.5767 30.19 10.5844 30, 3(
45 10.5558 33.66 10.5635 33.86 10.5713 34.06 10.5791 34, 2(
50 10.3496 37.52 10.5574 37,75 10.5653 37.97 10.5731 38. 1!
55 10,5428 41.42 10,5507 41.67 10,5586 41.92 10.5665 42.11
60 10.5352 45.37 10.5433 45,65 10.5512 45.91 10.5593 46,11
65 10.5270 49.38 10.5352 49.67 10,5432 49.96 10.5514 50.2!
70 10 5181 53 44 10.5264 53,76 10,5346 54.07 10.5429 54,31
75 10.3085 57.57 10,5169 57.92 10.5253 58.25 10.5338 58. 5!
80 10.4983 61 78 10.5068 62.15 10.5154 62.51 10,5240 62,8(
85 10.4873 66,07 10.4960 66.46 10.5048 66.85 10.5136 67.2:
90 10,4757 70.46 10.4846 70.8s 10.4936 71,28 10.5026 71.61
95 10,4634 74.96 10.4726 75.40 10.4818 75 82 10 4910 76.2,
100 10,4505 79.58 10.4600 80.03 10.4694 80,48 10,4738 W.g
105 10 4370 8.,32 10.4468 84.80 10.4564 85.26 10.4661 85,7
110 10 4230 89.22 10,4330 89.71 10,4430 90.19 10,4530 90, 6(
115 10,4086 94.28 10.4188 94.79 10.4291 95.28 10.4395 95.7(
120 10.3935 99.52 10.4042 1C4,04 10.4149 100.54 10,4256 101.U
125 10,3783 104.95 10.3894 105,48 10,400.5 105.99 10.4116 106,4!
130 10 3630 110.60 10.3745 111.13 10.3860 111.64 10.3976 112.1
135 10.3476 116.48 10.3597 117.01 10,3716 117.51 10.38.36 118.01
140 10 3326 122.61 10.3451 123.12 10.3575 123.61 10.3699 124.01
145 10,31s0 128.99 10.3311 12!3.48 10.3440 139,94 10,3568 130.3!
15JI 10,3043 135.63 10.3179 136.08 10.3313 136.52 10,3446 136,9:
155 10 2918 142.54 10.3059 142,94 10.3198 143.32 10.3335 143.6!
160 10,2809 149.68 10,2954 150.02 10,3097 150.35 10.3238 1,50,6(
165 10,2719 157.05 10.2869 157.32 10.3015 157.57 lIJ.311Y3 157. s
170 10.2652 164.59 10.2806 16478 10.2954 164.95 10.3102 165.1:
175 10.2610 172.26 10.2766 172.36 10.2918 172.45 10.30S6 172, S
180 10.2597 leo.oo 10.2751 130.00 10.2903 180,00 10,3053 180.11
0.49 0.50
l. [ 1.
.P + 10 de- fiP+ 10 d:
m pees
10,6118 0,00 10.6193 O,m
10.6115 3.78 10.6190 3,80
10.6106 7.56 10;6181 7.60
10.6091 11,35 10.6166 11,41
10.6069 15.15 10.6145 15.23
10.6041 18.96 10.6118 19,07
10.6007 22.80 10.6084 22,92
10.5967 26.65 10.6045 26.80
10.5921 30.54 10.6000 30.71
10.5869 34.46 10.5948 34,64
10.5810 38.40 10.5890 38,62
10.5745 42.40 10,5826 42.63
10.5674 46.44 10 5756 46 70
10.5597 50.53 10.5680 50.81
10.5513 54.69 10.5597 54.99
10.5423 58.91 10,5508 59 23
[0.5327 63.21 10.5413 63,55
10.5224 67.59 10,5313 67,96
10.5116 72.07 10.5207 72,45
10.5002 76,65 10,5095 77.05
10.4883 81.34 10.4978 81.76
10.4759 86.16 10.4857 86.60
[0.4631 91.12 10.4732 91.57
10.4499 96.24 10.4603 96.70
10.4363 101.51 10.4471 101.98
10.4227 106.97 10.4338 107.44
10.4091 112,62 10.4206 113.09
10,3955 118,47 10.4074 118.93
10.3823 124.54 10.3946 124.98
10.3697 130.83 10.3824 131.25
10,3579 137.33 10.3711 137.72
10.3472 144.05 10,3609 144.39
10.3380 1.50.96 10.3520 151.25
[0. 3304 158.05 10.3447 138.28
10.3248 165.29 10.3394 165.44
10.3214 172.62 10.3360 172.70
10.3202 130,00 10,3349 WO.oo
I
I ndex
A
Accuracy, of bar-l i nkage computers,
27, 41
of bar-l i nkage mul ti pl i ers, 39
of camoi ds, 23
of computi ng mechani sms, 1, 4, 166
of doubl e-bal l i ntegrator, 26
of graphi cal methods for l i nkage desi gn,
versus numeri cal methods, 199
of sauari rur cam. 22-23
(~ee aZ-s;specific mechanism)
Adder, bar-l i nkage, 36
star l i nkage i n desi gni ng of, 256
Addi ti ve cel l s, 6-12
di i 7erenti al bel t, 10
bevel -gear , 6-7
cyl i ndri cal -gear, 7
wi th spi ral gear , 9
l oop-bel t, l G1l
screw, 9-10
spur-gear, 79
dMerenti al wor m gear i ng, 9, 11
B
Backl aah of cyl i ndri cal -gear di fferenti al , 7
of bar l i nkages, 27
of pl ane cams, 19
BackJaah er r or of bar l i nksgea, 21, 33
Bal l i sti c functi on i n vacuum, mechani za-
ti on of, 286-299
k-l i nkage adder , accuracy of, 41
appr o-te, 41, 42
er r or , 37, 41, 42
structural er r or of, 41
B8r-l i nkage computers,
integrators, 27
complex, 40-42
ecmceptaof, 43-57
design of, al-32
terminology of, 43-57
compared wi th
Bar-l i nkage mul ti pl i ers, 3740, 2.50-283
desi gn of, 3740, 250-283
er r or of, 37, 3839, 40
Bar l i nkages, 5, 2742
versus camoi ds, 23
characteri sti cs of, 32-33
advantage of, 5-6, 3233
di sadvantages of, 33
compar ed wi th cams, 33
de6ni ti on of, 27
di mensi onal constants of, 31, 32
ehl ci ency of, 28
er r or of, 27, 28, 3133
back l ash er r or , 21, 33
el asti ci ty, 33
mechani cal , 33
structural , 31-33
fri cti onal l oeaes of, 28
hi story of use of, 28-31
mechani cal features of, 28
strai ght-l i ne moti on by, 29
wi th one degr ee of fr eedom, 34-36
wi th two degr ees of fr eedom, 3740,
223-249
adder , 37
mul ti pl i ers, 3740, 250-283
as substi tute for three-di mensi onal
cam, 37
Bel t cam, 20-21
Bel t Merenti al , 10
Bevel -gear di t7erenti rJ, 6-7
Bl ock di agram i n computer desi gn, 34
Bl okhj Z. Sh., 31n.
c
Camoi ds versus bas I i nkagea, 23
Cams, 2.4, 19-23
bel t, X21
compar ed wi th bar l i nkages, 33, 3*,
eompenaated bel t, 2123
cyl i ndri cal , 20
defi ni ti on of, 19
353
354 COMPUTI NG MECHANI SMS AND LI NKAGES
I
I
I
Cams, l i nkage, 34
pl ane, 19
squari ng, 2123
three-di mensi onal , 23, 37
Cayl ey, A., 30n., 31n., 297
Cel l s, 3, 6-40
addi ti ve (see Addi ti ve cel l e)
of bar-l i nkage computers, 41
cams (see Cams)
i ntegrators (see I ntegrators)
l i near, 6
l i nkage, 40
mul ti pl i ers (see Mul ti pl i ers)
resol vers, 1&19
Compensated bel t cam, 21-23
Compensati ng di fferenti al , 16
Compl ementar y i denti fi cati on of param-
eter and vari abl e, 49
Component sol ver, 2, 26
Components for computer , sel ecti on of;
4
Computati onal errors, 201
Computer desi gn, 2-5
anal yti cal method of, 3
bar-l i nkage mul ti pl i ers, 25&283
components for , 4
constructi ve method of, 2
er r or of, 2, 4, 223
sel ecti on of components i n, 4
functi on generators wi th two degr ees
of fr eedom, 204300
harmoni c transformer l i nkages, 58-106
l i nkage combi nati ons, 16&222
l i nkagee wi th two degr ees of fr eedom,
223-249
as mechani zati on of equati on, 3
model of, 5
three-bar l i nkages, 107165
Computi ng mechani sms, accuracy of, 1,
4, 166
conti nuousl y acti ng, 1
di fferenti al -equati on sol vers, 12
component sol vers, 2, 26
i ntegrators, 2
pl ani meter, 2
speedometer s, 2
el ementar y (see Cel l s)
i deal functi onal , 43
typee of, 12
(fl ee al so Computers)
Constrai nt l i nkage i n bal l i sti c computer ,
297-299
Crank, 27
Crank termi nal , 44
Cyl i ndri cal cam, 19
Cyl i ndri cal -gear di fferenti al , 7
backl ash of, 7
fri cti on of, 7
Cyl i ndri cal rack, 10
D
Dawson, G. H., 30n.
Desi gn, bar-l i nkage computer , pr obl em
of, 3132
doubl e-har moni c transformer, 101-106
Di fferenti al anal yzer, 4
Di fferenti al -equati on sol vers, 2
Di fferenti al wor m gear i ng, 9, 11
Di fferenti al s, bel t, 10
bevel -gear , &7
compensati ng, 16
cyl i ndri cal -gear, 7
l oop-bel t, 1(P11
i n mechani zati on of functi ons, 233, 238
screw, 910
wi th spi ral gear , 9
spur-gear, 79
Di mensi onal constants, 32
vari ati ons of, 201202, 205
Di rect i denti fi cati on of parameter and
vari abl e, 48
Di vi der , bar-l i nkage, 40
Di vi si on by mul ti pl i ers, 40
Domai n, of parameters, 44
of vari abl e, 44
Doubl e-bal l i ntegrators, 24-26
Doubl e harmoni c transformers (see Har-
moni c transformers, doubl e)
Doubl e three-bar l i nkage, 36
i n seri es, 195-198
successi ve approxi mati ons i n desi gn of,
196, 197
E
Eccentri c l i nkage, 35
as cor r ecti ve devi ce, 217
Effi ci ency of bar l i nkage, 28
El asti ci ty er r or of bar l i nkages, 33
Emch, A., 31n.
Error, of bar-l i nkage adder , 37, 41, 42
of bar-l i nkage mul ti pl i ers, 37, 38, 39, 40
of bar l i nkages, 27, 28, 31, 32, 33
I NDEX 355
Error, of computi ng mechani sms, 2, 4
of doubl e-bal l i ntegrators, 26
of l i nkage mul ti pl i er, 4
maxi mum, 2
of noni deal harmoni c transformer, 59,
60, 67-75
of preci si on squari ng devi ces, 4
of sl i de mul ti pl i er, 4
of squari ng cam, 23
(See al so speci fi c mechani sm)
F
Feedback, 34
Fol l ower (of three-di mensi onal cam), 23
Fri cti on, of bar l i nkages, 32
of camoi ds, 23
of cyl i ndri cal -gear di fferenti al , 7
Fri cti on-wheel i ntegrators, 24
Functi on, mechani zati on of, 4647
Functi on generators, 1
bar-l i nkage computers, 2742
compar e wi th i ntegrators, 23
wi th two degr ees of fr eedom, 284-299
factori zati on of, 168-174
gener ated by three-bar l i nkages, 122-
127, 146
gr i d structure of, 228-233
wi th i deal gr i d structure, 233
of two i ndependent parameters, 224
of two i ndependent vari abl es, mechan-
i zed by bar l i nkages, 223
mechani zed by three-di mensi onal
came, 23, 223
G
Gauge of preci si on of l i nkage, 201
Gaugi ng constant, 202
Gaugi ng er r or , 201, 202, 203, 205, 211-
215
Gaugi ng parameters, 200-217
i n adjusti ng l i nkage constanta, 201-205
i n eccentri c l i nkage, 218
i n three-bar l i nkage desi gn, X)7-217
Geometr i c method, i n mechani zati on of
l ogari thmi c functi on, 156-165
for three-bar l i nkage desi gn, 145-165
Gr@i c.d methods of l i nkage desi gn, 199
Gr eenhi U, A. Cl ., 30n.
Gr i d gener ator , defi ni ti on of, 225
gaugi ng er r or of, 281283
for given function, 225
mechanization of function by, 238
nonideal, 23%242
star, with almost ideal grid structure,
25&234
structural error of, 226
Grid structure, function of, 228233
generalized, 243
ideal, 22%230, 23>238, 240
star grid generators with, 250-284
nonideal, 23&233, 241, 28+299
regularized, 239-242
transformation of, topological, 232
233, 239, 244, 246, 249, 251
use of, in linkage design, 243249
H
Harmonic-transformer functions, tables
of, 63%7, 301332
Harmonic transformers, 27, 34, 36, 58-
106
doubl e, 36, 77106
desi gn of, 95-106
i deal , 58-67
i n homogeneous parameters, 6263
i deal doubl e, 77+5
monotoni c functi ons mechani zed by,
78, 8288
mechani zati on of functi on by, 61+33
for monotoni c functi ons, 78, 8288
noni deal , 58-60, 6777, 249
desi gn of, 75-77
er r or of, 59, 60, 6775
structural er r or of, 59, 60, 6775
noni deal doubl e, 95-102
for nonmonotoni c functi on, 89-91
parameters i n, 61
i n seri es, 77106
two, wi th three-bar l i nkage, 16&195
use of, 61-62
Hart, H., 30
Hart i nversor, 30
Hi ppi sl ey, R. L., 29n., 31, 297
Homogeneous parameters, 43,4749, 62-
63, 6S-71
i deal harmoni c transformer expr essed
i n, 62-63
Homogeneous vari abl es, 43, 47-49, 62,
64-67, 78,79, 87, 89, 171
COMPUTI NG MECHANI SMS AND LI NKAGES
1,
1;
356
I
I deal doubl e harmoni c transformer (see
Harmoni c transformer, i deal doubl e)
I deal harmoni c transformer (see Har-
moni c transformer, i deal )
I denti fi cati on, compl ementar y, 49
di rect, 4%49
I nerti a of bar l i nkages, 33
I nput parameter (see Parameter i nput)
I nput scal e, 46, 51
I nput vari abl e, 45
I ntegrators, 2, 4, 2*26
compar ed wi th functi on generators, 23
component sol ver, 26
defi ni ti on of, 23
doubl e-bal l , 2426
fr i cti on-wheel , 24
I ntersecti on nomograms, 40, 251
I nver sor , Hart, 30
Peaucel l i er, 29, 30
I nver ted functi on, 184
mechani zati on of, 138-139, 143145,
155, 177
J
Johnson, W. W,, 31n.
K
Karpi n, E. B., n.
Kempe, A. B., 3(M.
L
Laverty, W. A., 30n.
Leaet-square method i n l i nkage probl ems,
206-207
Levenbur g, K., 207
Lever , 27
Li near cel l z (see Addi ti ve cel l s)
Li near mechani zati on, 46, 47, 48
Li near termi nal , 46
Li nk, 27
Li nkage came, 34
Li nkage computers, defi ni ti on of, 44
Li nkage constants, fi nal adjustment of,
eccentri c fi nkage i n, 217222
gaugi ng parameters i n, 200-217
Li nkage desi gn, graphi cal versus numer -
i cal methods, 199-200
Li nkage desi gn, numeri cal methods i n,
167198
numeri cal versus graphi cal methods,
199-200
structural er r or i n, 199
use of gr i d struct!re i n, 243249
Li nkage i nversors, 34
Li nkage mul ti pl i er constants, adjust-
ment of, 277281
Li nkage mul ti pl i ers, 4, 28-29, 223283
Li nkages, bar, 5
wi th one degr ee of fr eedom, 5%198
star, 238
wi th two degr ees of fr eedom, 3740,
223250
desi gn pr obl em for , 22*226
Logari thmi c functi on, 36
Logari thmi c l i nkage, 238
check of, 209-217
i mpr ovement of, by eccentri c l i nkage,
21%222
by gaugi ng-par ameter method, 209
217
Loop-bel t di fferenti al , 10-11
M
Mathemati cal desi gn, 45
Mechani cal er r or , of bar l i nkage, 33
Mechani sm, regul ar, 45, 223
Mechani zati on of al i gnment nomogr am
wi th thr ee paral l el strai ght l i nes, 37
Mechani zati on of bal l i sti c functi on i n
vacuum, 286-299
Mechani zati on of a functi on, 46-47
by combi nati on of three-bar l i nkage
and two i deal harmoni c trans-
formers, 171, 186
wi th a di sconti nui ty i n deri vati ve, 114
117
by harmoni c transformer, 61-63, 166
by homogeneous parameters and vari a-
bl es, 62-67
by i deal doubl e harmoni c transformer,
78-95
wi th i deal gr i d structure, 250, 272
by l i nkage combi nati ons, 166
by method of I eaat squares, 62
by noni deal doubl e harmoni c trans-
for mer , 95-101
wi th one degr ee of fr eedom, 166-198
by three-bar l i nkage, 166, 184
I NDEX 357
Mechani zati on of i nver ted functi on by
nomogr aphi c method, 138-139, 143
145, 155
Mechani zati on of l ogari thm functi on
by doubl e three-bar l i nkage, 36
by geometr i c method, 156-166
Mechani zati on of monotoni c functi ons,
78
Mechani zati on of rel ati on between vari a-
bl es, 2
Mechani zati on of tangent functi on, 171,
175, 177, 178, 180, 182, 183, 186
by i deal harmoni c transformer, 165
Model of computer , 5
Monotoni c functi on, mechani zati on of,
78, 82-88
Mul ti pl i ers, 1215
bar-l i nkage, 3740, 25@283
defi ni ti on of, 12
ful l -range, 40
hal f-range, 40
l i nkage, 4
nomogr aphi c, 14-15
quarter-range, 40
r esol ver , 15-19
sl i de, 4, 1214
star l i nkage i n desi gni ng of, 256-283
transformati on of gr i d structure of,
25%264
wi th uni for m scal es, 272
N
Nomogr am, 13, 14-15, 37, 176, 184
three-bar-l i nkage, 120-145, 333-352
three-bar-l i nkage functi ons r epr esented
by, 120-127
for transformer l i nkage, 275
Nomogr aphi c chart, i n three-bar-l i nkage
cal cul ati ons, 120-122
Nomogr aphi c method, 178, 181, 183
wi th three-bar I i &age, 174
for three-bar-l i nkage desi gn, 118-145
Nomogr aphi c mul ti pl i er, 14-15, 40
Noni deal doubl e harmoni c transformer
(see Harmoni c transformer, non-
i deal doubl e)
Noni deal harmonic transformer (we
Harmoni ctransformer,noni deal )
Nonl i nearmechani sati on,47
Nonmonotoni c functi on, 89-91, 169, 170,
172
Numeri cal methods of l i nkage desi gn, 199
0
Oper ator , i nverse, 50-51
Oper ator formal i sm, 4%51
Oper ator notati on, 49-57
Oper ator symbol i sm, 185
Operators, 4%57
for doubl e harmoni c transformer, 79,
81, 86, 93, 95, 96
equati on of, 50, 51, 53
graphi cal representati on of, 5154
pr oduct, 5153
square of, 54
square-root of, 5457
Output parameter (see Parameter, out.
put)
Over l ay, constructi on of, 133137, 13&
137, 143
exampl e of constructi on of, 176
use of, 13=136, 141142, 150, 151, 152,
153, 154
Output scal e, 46, 51
Output vari abl e, 45
P
Paral l el ogram l i nkage, 28, 29, 34, 113,
115, 127
Parameters, 6
defi ni ti on of, 44
domai n of, 4+45
of harmoni c transformer, 61
homogeneous (see Homogeneous pa-
rameters)
i nput, 10, 19, 44, 45
of component sol ver, 26
of l i nkag~ wi th two degr eee of
fr eedom, 223
output, 11, 19, 44, 45
of component sol ver, 26
of fi nkagec wi th two degr eee of fr ee.
dom, 223
Peaucel l i er, 29
Peaucel l i er i nver eor , 29, 30,95
Pi n gear i ng, 19
Pl aue cams, 19
P18ni meter, 2
358 COMPUTI NG MECHANI SMS AND LI NKAGES
Precti l on, of camoi d, 23
of l i nkage, gauge of, 201
of squari ng cam, 23
Preci si on squari ng devi ces, 2, 4
er r or of, 4
(See al so Cams)
R
Regul ar mechani sm, 45
Resi dual er r or of bar l i nkages, 32
Redver s, 15-19
Restri cted parameter, 28o, 282
Roberts, S., 297
s
Scal e, defi ni ti on of, 45-46
i nput, 46, 51
output, 51
Scal e factor, 61
screw di fferenti al , 11
Sel ecti on of components of computer , 4
Sel f-l ocki ng, of multipliers, 12
of Peaucel l i er i nversor, 30
of pl ane cams, 19
Sl i de mul ti pl i er, 4, 12-14
Sl i de termi nal , 44, 166
Speedometer s, 2
Spi ral gear , di fferenti al wi th, 9
Spur -gear dti erenti al , 7-8
Square-root oper ator , use of, 197-198
Squari ng cam, 21-23
Star l i nkage, 37, 238
i n desi gni ng of adders, 256
i n desi gni ng of mul ti pl i ers, 256
(See a180Star gr i d gener ator )
Star gr i d gener ator , 250-299
for bal l i sti c functi on i n vacuum, 286-
299
as mul ti pl i er, 256-257, 264
i mpr ovement of, 264-271
Strai ght-l i ne moti on, 29
of bar-l i i kage adder , 37, 41
Structural er r or , of bar l i nkages, 32, 33
de6ni ti on of, 32
of gr i d gener ator , 226
of l i nkages wi th two degr ees of fr ee-
dom, 223
i n l ogari thmi c l i nkage, 216, 219, 222,
223
Structural er r or , of noni deal harmoni c
transformer, 59, 64, 6775
r educti on of, 199, 207
of star gr i d gener ator and transformer
l i nkages, 277-283
Successi ve approxi mati on, 81, 91-95, 98
i n desi gn of functi on generators, 286292
i n doubl e three-bar l i nkage desi gn, 196,
197
i n termi nal harmoni c desi gn, numer i -
cal method, 187
graphi cal method, 187
wi th three-bar l i nkage components,
174, 177, I fMl , 184
i n three-bar-l i nkage desi gn, 129, 137
140, 143-145, 146, 154, 177
Svoboda, A., 198
T
Tangent l i nkage, 19>195
Termi nal s, 43-44
crank, 44
defi ni ti on of, 43
i nput, 11, 43
l i near, 46
output, 11, 43-44
sl i de, 44
travel of, 45
Termi nal harmoni c transformer, desi gn
of, 18&192
Three-bar l i nkages, 35, 36, 107-166
cl assi fi cati on of, 108-112
combi ned wi th two harmoni c trans-
formers, 166-195
equati ons for , 107108
fi el d of functi ons of, 118
functi ons gener ated by, 122127
i n seri es, 195-198
speci al cases, 112117
Three-bar l i nkage component, desi gn of,
174-195
geometr i c method, 145-156
nomogr aphi c method, 118-145
pr obl em of, 117118, 127128
Three-bar-l i nkage nomogr am, pr oper -
ti es of, 3X+352
Three-bar moti on, 31
Three-di mensi onal cams, 23
Tol erances, speci fi ed, to be met by com-
puter, 2
(See ako Speci fi c. mechani sm)
I NDEX 359
Topol ogi cal transformati on of nomo-
gram, 40
Transferred r egi on, 137
Transferred poi nts, 137
Transformer l i nkages, 225
of bal l i sti c computer , 29&297
desi gn of, 271277
geometr i c method, 276
graphi cal methods, 276
for nonci rcul ar scal es, 29%297
Transformers, defi ni ti on of, 225
l ogari thmi c, 238
Travel of a termi nal , 45
Two harmoni c transformers, properti es
of, 166
assembl y of l i nkage combi nati on for ,
193-195
desi gn of three-bar-l i nkage component
of, 168186
factori zati on of functi on i n, 168-174
r edesi gn of termi nal , 18&193
Two harmoni c transformers, pr obl em of,
166-168
v
Vari abl es, compl ementar y, 48
defi l ti on of, 45
homogeneous (see Homogeneous vari -
abl es)
i nput, 45
output, 45
r ange of, 46
w
Watt, James, 28
Worki ng model of computer , 5
Worm gear i ng, 9
di l l erenti al , 9, 11
z
Zer o posi ti on, 6, 22