NPS ARCHIVE
1960
MCLAUGHLIN, D.
i
ANALOG SOLUTION OF CENTRAL!
' ■ FORCE PROiUJfM
DEAH H. MctAUG-HLIN
i' 'l!l\>l iltiu'Iiti'il''
■
•
ii!i!iiii!fiHKIii i*>r
' i \ : ill I .'" ' ■ If 1 ; ' ' ' ;' ''•
illillltlwlllliiPnililliHliilf
' . I] llf] Stlfl ■' M ' ' ' i! '■ ' ■ ' :
illilitniii Imliilihlili
IHilimiiitmHwII
'[lyUlia iH!« flulhi'
Rl!ifrl(Kif(l)!lrHif(iil(i fiillffifl]
il?iliii^
jfjjpj! I (Ikiillj |j|i
^IbllillnHllt'llJJIil1 : Iff frill)! »{fl
|i!| ! ! :l!l|jiwi!t!i' 1 i iljlfli
I'llrtwflllmlW'H
»iHHIi'i!l'l^pilij!
iltiiiilltj' P hIII
^ '■' "■■ ' ' . ' "
i1P2 'HiMff if fi tiS- *
IW1S1!
lMH8^ffiflSHV*HH!Mmri
lil.'M.'HW:
3SS»~
DUDLEY KNOX LIBRARY
m«™S22TGRADUATE SCHOOL
MONTEREY, CA 93943-5101
ANALOG SOLUTION OF
CENTRAL FORCE PROBLEM
by
dean n. Mclaughlin
lieutenant, united states navy
/0>(?S Ale '£
ANALOG SOLUTION OF
CENTRAL FORCE PROBLEM
*****
DUDLEY KNOX LIBRARY
NAVAL POSTGRADUATE SCHOOL
MONTEREY, CA 93943-5101
Dean N. McLaughlin
ABSTRACT
Electronic analog computers have been used extensively for the
solution and display of many dynamics problems. The majority of the
problems worked with have been those involving linear differential
equations with constant coefficients. In cases involving non- linear
differential equations fewer solutions have been developed. This ha®
been due mainly to the need of using non- linear elements in the com-
puter circuits when setting up the solutions.
One such problem, that of a mass moving under the action of a
first power central force, is treated in some detail. The differen-
tial equation is derived, the problem is scaled, and the circuitry
developed. Solutions obtained by the use of the electronic analog
computer are displayed and compared with solutions obtained by num-
erical methods and errors and their sources are discussed. Finally
there is an overall evaluation of the usefulness of analog computers
to this sort of problem. In an appendix, a second practical dynamics
problem is discussed, but a solution was not obtained due to lack of
time available.
TABLE OF CONTENTS
Section Title Page
1. Introduction - 1
2„ Background 1
3o The Problem and General Method of Solution 2
4„ The Differential Equation of Motion 4
5c Discussion of Equipment 5
6o Computer Equation and Scaling 7
7o Analog Computer Circuits 8
8c Results 10
9» Discussion of Discrepancies 23
10c Conclusions 24
Bibliography 26
Appendix I 27
Appendix II 37
Appendix III 42
iii
LIST OF ILLUSTRATIONS
Figure Page
1. Force Diagram for Central Force Problem 4
2„ Division Circuit 9
3., Circuit Diagram 12
4, Photograph of Computer Assembly 13
5o Photograph of Problem Board 14
• • •
6o Recordings of, R, Z, R, and R 15
7. Recordings of, 9 , R, R, and O 16
8, Summary of Analog Results 17
9o Radius vs Time Curves 18
10. Angle vs Time Curves 19
Ho Radius vs Angle of Rotation Curve 20
12 0 Z vs Radius Curve 22
A13. Table of Numerical Integration 30
A14, Table of Calculated Values 32
A15. Table of Calculated Values 33
A16. Integration Curve, Part 1 34
A17« Integration Curve, Part 2 34
A18. Integration Curve, Part 3 35
0
A19. & and r vs Time Curves 35
A20. r vs Time Curve 36
A21„ Table of Circuit Elements 41
IV
TABLE OF SYMBOLS AND ABBREVIATIONS
(without subscripts) Constants in differential equation
C, Capacitor (i = f, 1,2, 3, . . . . .)
M Meg ohm
R (without subscript) Voltage representing radius r
Ra Resistor (i = f ,1, 2,3, . . . . .)
X A capital representing the voltage equivalant of a variable x
?■ Output voltage of division circuit
a Coefficient potentiometer value
e Input voltage to an operational amplifier
e Output voltage of an operational amplifier
o
f ^subscript) Element in feedback loop
r Radius
t Computer time
t Problem time
P
oC^ Scaling factor (i » 1,2,3,...)
& Angle of rotation
UJ^. Input voltage coefficient (i = 1,2,3,....)
Capacitor
Resistor
Operational amplifier
— ■ @ Coefficient potentiometer
~^|^ f-» Function Multiplier
1, Introduction
This thesis presents the electronic analog solution to a non-
linear dynamics problem which leads to the differential equation
*" A fix I f 1
An example problem is taken and the steps in reduc-
ing it to a form suitable for an electronic analog computer^ hereafter
referred to as an analog computer, are shown. The results are then
compared with two solutions obtained by numerical methods. In Appen-
dix III an equation of the form; >< =AX f 8X is
discussed and the problems encountered in trying to obtain an analog
computer solution are delineated.
The writer wishes to express his appreciation for the assistance
given him by Professor John E. Brock and to the Professors, particular-
ly Professor 0. H» Polk, and the technicians of the Electrical Engineer-
ing Department. The numerical solutions in Appendix I were contributed
by Professor Brock.
2. Background.
Solutions for many dynamics problems have been established using
analog computer and references can be found in the technical litera-
ture. One such reference for a non- linear problem^, Analog Computer Solu-
2
tion of a Non- Linear Differential Equation, by H. G. Markey, (2) , was
found but was only applicable in a general way to this investigation.
It was considered that if a means could be found to display some
of the classical problems encountered in early college dynamics on the
i
?d/dt is denoted by a dot placed above the variable operated on,
"vers in parentheses refer to references in Bibliography.,
analog computer the following advantages would be obtained;
(a) the general usefulness of the analog computer could be
made more readily apparent;
(b) in dealing with these problems attention could be focused
on the dynamic principles leading to the governing differential equations
and upon the mechanical significance of the results and not upon the mathe-
matical difficulties in obtaining an analytical solution;
Cc> in the case of those problems where analytical solutions
have been obtained for certain particular parameter values and which
therefore seem to be separated into many different cases the dynamical
significance of which is not apparent, the general problem could be
dealt with directly;
(d) it would be possible to include normal dynamical in-
fluences fsuch as energy loss due to pivot friction or windage) with-
out so complicating the mathematics of the solution that the modified
problem appears to be entirely different from the idealized problem,,
In addition to the above it was desired to obtain these results
using only the analog computers and their associated equipment, normal-
ly available in an analog computer laboratory.
3o The problem and general method cf solution.
The problem considered was that of determining the subsequent-
motion of a body weighing 1930 pounds, attached to a spring having a
free length of five inches and a scale of ten pounds per inchj, when
released with the following initial conditions,, At the initial instant
the radius is four inches and its rate of change is zero; the polar
anglej » ^s zero and its rate of change is three radians per second.
end of the spring is attached to a fixed point and the body
■srmitted to slide without friction upon a horizontal plane.
We will discuss the sequence of steps necessary for the solution
a problem of this type9 and then we will proceed with the solution.
One might of course proceed with a full scale experimental program as
a method of solution^ but eliminating this possibility we would s
a. Using the principals of Mechanics arrive at one or more
differential equations describing the motion.
b„ Solve these equations9 incorporating the starting condi-
tions. This solution may be analytic^ numerical^ experi-
mental (dealing with, possibly, scaled down mechanical
variables) 9 or by means of an analogs in which one deals
experimentally with variables of another type (such as
electrical) which satisfy similar differential equations,
Co Interpret the mathematical, experimental 8 or analog results
in the proper mechanical light so as to arrive at a mean-
ingful solution to the original problem.
In this thesis8 we are investigating the practicability of pro-
ceeding immediately from the first step to a solution by use of stand-
ard analog computer equipment. We do not have an analytical solution
to the problem stateds although it is likely that one might be obtained
in terms of elliptic functions and integrals. However^ in an appendix
we will develop two different numerical solutions to the problem with
which we can compare our analog solutions.
The differential equation of motion.
Figo 1 shows the body in a general position. The solid arrow re-
presents the spring force F = 10(r-5), where r represents the radial
s
distance from the fixed point 0, The dotted arrows represent D'Alembert
forces necessary for dynamic equilibrium. We see that
Fs +• Yn <X,7 r: O
Yn o*,q ~ O
Now by kinematics, a^-r?-n& and <^-& --L sL. (n*& ) . From the
second equation we see that ft & z C = Ccfi$£t . This can also be seen
from the fact that the angular momentum of the system about 0, namely
Yn ft & , is invariable. From the first equations, we have
10 (* -s) +F * 3 7j.(ji -jygfy-Q t and from this we iit-ih-lo+b-hd^z-O .
Substituting & =. C/m^- we finally get ft = .£ Q.h +■ /O
In our cases evaluating C at the initial instant we have C = 48, Thus
we have as our set of differential equations:
Now it is possible to perform some mathematical manipulations which
simplify this system. In particular, it is easily possible to obtain a
first integral of the first equation of the system. However 9 w© regard
it as contrary to the spirit of this thesis, the purpose of which deals
with the ready feasibility of making an analog computer solution of this
system, to perform any such manipulations, and it is
.<£?
<rfi*
Fig. 1 Force Diagram for Central Force Problem
4
this system with which we shall be directly concerned when we attempt
the analog solution* The numerical solutions for this problem will be
found in Appendix I.
5* Discussion of equipment.
Before taking up the solution of the problem, a description of
the equipment used in the solution of this problem will be presented*
It is assumed that the reader is already acquainted with the basic theory
of the analog computer and with the usual circuitry used, such as summers,
integrators, etc. The Handbook of Automation, Compution and Control,
Vol, 2, E. M. Grabbe, (1), is a good reference for this information as
well as for additional information on the equipment discussed below,
A, Donner Analog Computer, Model 3000,
This analog computer, which can be used for the quantita-
tive solution of linear (and certain classes of non- linear) differential
equations and transfer functions, contains ten DC operational amplifiers,
any one of which can be used for addition, subtraction, multiplication or
division by a constant, sign changing, or integration. Problems express-
ed as differential equations are entered in terms of electrical components
on a detachable problem board. Stability and accuracy are satisfactory
for problem solution times up to 100 seconds or more which permits ac-
curate recording with conventional pen recorders. (5)
B. Donner Function Multiplier, Model 3730.
This function multiplier consists of two multiplier
channels and a regulated power supply. Each multiplier channel produces
an output voltage which is accurately proportional to the instantaneous
produce of two independent input voltages, where each input is either
constant or varying with time Either input may be positive or negative9
so that four quadrant multiplication is provided,, The range of output
and input voltages is plus and minus 100 volts; this being necessary to
stay within the linear range of the operational amplifiers of the com-
puter. To maintain the output voltage at 100 volts or less the Function
Multiplier is designed to give an output voltage which is ,01 the product
of the input voltages, f6)
For the solution of the problem of this thesis two of these multi-
pliers were used. They gave accurate results when used for straight
multiplication although they do tend to drift over a period of time and
have to be readjusted; this is a minor operation, however. When used in a
division circuit, which is discussed in a later section^ the results obtain-
ed were not as accurate^ however. It is believed, however, that this was a
fault of the circuit and not of the ^unction multiplier bacause, as mention-
ed above, the function multipliers gave quite accurate outputs when used
for multiplication alone.
C. Donner Function Generator, Model 3750.
This variable base function generator is designed for use in
conjunction with two external operational amplifiers to produce an output
voltage which can be adjusted to approximate a desired single valued
function of the input voltage. One operational amplifier is required for
operation of the function generator and the other is used to accept the
output signal at its summing junction. This amplifier may also be used
for additional summing^ inverting, integrating or other operations. The
function generator operates on the principal that the function can be
approximated by a series of straight line segments, each segment being
limited to a slope between plus and minus two volts per volt. The input
and output voltages may vary between plus and minus 100 volts. (4)
For the solution of the thesis problem it was desired to use
3
this function generator to generate the function 2304/r but it was found
that the slope of curve for this function exceeded the capability of the
equipment. This is duscussed further in Sec. 9. If it had been possible
to use this function generator the two function multipliers would not have
been required.
6. Computer equation and scaling.
To reduce our problem to a form suitable for the computer it is
necessary to apply scaling factors. This was done using the methods out-
lined in Basic Theory of the Electronic Analog Computer, by R. C. H. Wheel-
er> (9). A brief summary of this process is presented here.
The differential equation to be solved is first arranged so as to
o* A.
solve for the derivative of the highest order. In our case K- = H-
n3
-BX-f-C . The equation is then scaled so that maximum value of each para-
meter is represented by a voltage, close to but not exceeding 100 volts,,
To do this scaling factors are assigned as shown by the following ex-
ample:
X = oc.
X
Here X is the computer voltage representing the variable x, and oC^
is its scaling factor. After being calculated the scaling factor is
usually rounded off to facilitate computations. After suitable scaling
factors are found, the equation is put into the form:
«*R A-., - s*„R ,c
.■'0'
To develop the applicable circuits for the problem solution it is
necessary to determine the values of resistance and capacity needed
for each component of the circuit. Using the procedures in Wheeler' s
books (9), pages 2-10 we find, for example, that an operational amplifier
when used as a summer has an output voltage ^0 -~LU-J,G/ -f-w^C.^ -4 )
or in our case Rs-(wi SL _ ^K^i^H . If we now let 6^ - c*-^ Kf-
where a. is a coefficient potentiometer setting and R_ and R. are re-
x r f i
sistances, we can establish the relationship C*--^ _ UJj, '\ ,_
It should be noted here that an R with a subscript, R„. refers to a resis-
tor and . R. without the subscript refers to the voltage representing the
variable r, the radius of the problem. Now the above relationship can
be solved for a . For integrators the relationship is Co^ - ^Z^=- »
where Cf refers to a capacitor.
7. Analog computer circuits.
In Appendix II the calculations for scaling the differential equa-
tions of our problem are presented. After scaling we have the follow-
ing equations: ft = __, _ ^ g ■> ^ 0QQ
K3
Before solving this problem on the analog computer two main decisions
3 2
have to be made: first how to calculate R and R , and second how
3 * 2
to develop the terms Z = 576,000/R and & = 30,000/R . It was
hoped at first that the terms for Z and 0 could be developed using func-
tion generators but as mentioned previously this proved unsatisfactory,,
Thus it was expedient to use the division circuit shown in Fig. 2„
8
H
«<
<-
■
h
\AAA-
Division Circuit (6)
Fig. 2
With this circuit Z = 100R| X
The factor 100 results from the
2 3
output of the multiplier being .01 Y£. If we now let Y = R and X
= constant, using the above relationship we should be able to develop
Z = 576,000/R3.
We know from the parameters of the problem that when r is 4S
R should be 20 volts. If we then put this value through two function
2 3
multipliers we come up with K R . As this value is small, .8 volt, we
multiply it by a factor of two using an operational amplifier and then
put it into the function multiplier of the division circuit. Also using
this value of the voltage for R we can calculate the value Z should
have, in this case 72 volts. With these values we can now solve for a
value of X so that with an input of 1.6 volts for Y and the calculated
value of X , Z will be 72 volts. Solving for X:
X -
- / !\
>
/oo Rt
Now letting R ~ equal 10M and R equal 1M, we find that an X of
11.5 should be used. (It was found that resistances of 10M and 1M worked
better than resistors of 0.1M and 1M) . This same procedure was applied
to & and the corresoonding voltage, X. was found to be 30 volts.
It should be noted here that another method for determining Q
presents itself, that of multiplying Z by R/19.2. By doing this the
second division circuit could be eliminated and only another multiplica-
tion3 with its more accurate results, required. This method was tried
and it was found that for some unexplained reason Q passed through
zero and became slightly negative. As a result of this S oscillated
instead of increasing smoothly from zero to a maximum value. For this
reason the division circuit for developing Q was used.
After the above determinations were made, the circuit of Fig. 3 was
assembled and computations made. In assembling the circuit the values of
the a's calculated in Appendix II were adjusted for the actual values of
resistances and capacitors used, e.g., 1.005 M actual resistance vs.
nominal value of 1M. Figs. 4 and 5 are photographs of the setup used
and shows the relative simplicity of the final setup for solution of the
problem.
8. Results.
After assembling the circuit of Fig. 3, it was found that to obtain
the desired values of voltage for Z and 0 the values of the input vol-
tages calculated for X and X- had to be adjusted. For Xs a value of
20 volts, and for X a value of 33 volts was required. Once these ad-
justments were mades the computing runs were made and the results are
shown on the Brush Recordings of Fig. 6 and Fig. 7, These recordings were
all made using a paper speed of 5/mm/sec and with varying voltage scales as
shown on each trace. These results are also summarized in the table of Fig.
8. From these results curves were plotted and then compared with the re-
sults obtained by the numerical solutions^ as shown in Figs. 9S 1Q» & il„
10
In analysing the results each term will be considered separate-
ly„ Considering r first it is seen that the maximum value of 38 ob-
tained agrees with the maximum value of the numerical solution but that
the minimum value of - 12c5 is lower than the - 15.26 of the numerical
solution. This latter discrepancy is attributed to the actual values
obtained for Z and will be discussed later.
11
'
<ir j> <z.
E
CO
u
60
CO
3
O
u
en
60
°i-t
•r-<
CO
U-l
CO
•r4
3
1— 1
CJ
CU
CO
E
°rA
CCJ
"V
C
o
o
4J
1-1
•u
*t3
CO
c
J-l
o
(U
a
a
CO
o
0)
u
CD
u
J2
o
4-)
o
(U
o
.
l-l
u
M
3
M
60
-o
•rl
a)
K
<4-l
m
•r-l
CU
T3
CO
JO
a
•i-l
E
CU
£1
3
a.
4J
C
a
<
c
CD
M
u
e
CO
•^
12
X
h
<s^>
13
cr^
IH
Summary of Analog Results
*0
V
z
R
tn1
r
0
R
r
<9
R
1 1
&
0 o
1
76
76
38
0
0
74
2o96
20
^ 1
0 0
0 1
62
69
34*5
35
4 = 2
68
2o72
3
2O06
1
.2
■ 3
49
38
47.5
27o5
23o7
13 o7
57 o5
71.5
6o9
52
2.08
2k
!
5o5
37oil
806
kl
I06U
7
Wol
2 01|
30
12 oS
6o7
80
9o6
31
lo2i|
34
6.6
8oi
58 oj
o5
26
2
1
81
9o7
23
o92
9o5
65 S
3
06
23
- 7
- 3o5
30
9c6
20
o80
kk
8o3
L0
68o7
o7
22
-12
- 6
75
9o0
18
c72
11
75.5
k
.8
21
= 15
- 7o5
67.5
8.1
15
06O
53
10o£
a:
75o5
o9
20
-17
- 8.5
60
7.2
13
o52
11
75<>5
75o5
5 loO
19
-20
-10
SO
6o0
12
.43
60
12oC
11
1.1
18
-21
-10o5
42.5
5.1
11
*44
11
75 .5
6 il.2
18
= 22
-11
32o5
3o9
11
>hk
6k
12 0^
11
75o5
1 3
13
-23
-11 *
dc. 0^.
2.7
11
*44
11
75o5
7o0
7
1.4
lo)4.3
18 ■
13
-24
-25
-12
-12.5
12.5
0
io5
0
11
11
■44
°44
66
67
13 o2
13 oL|
11
11
75o5
75.5
-\
>7 s
«%=./2
1
Kj-'M
r.x
!
Fig. 8
1?
30
20
10
ution
Solution #2
1+6 8
KadiuSj inches
ii+
Fig 11, Radius vs Angle of Rotation
20
Fo-r r the maximum value obtained was 9,7 and the minimum
zero when r was a maximum and a minimum. This agrees well wi
the numerical solution where the maximum value was 9,6, Considering r
w® see from Fig, 9 that the analog values are slightly higher at all vain
than the r°s of the numerical solution. This error is not considered
excessive.
The largest discrepancies appear when we consider ©„ As can be
seen in Figs„ 89 10s and 11 the analog value reached its maximum for
the first apse (point of greatest distance from the center of attraction'
rapidly and then remained constant for a period of time. Here as with
r the discrepancies are considered to be caused by the values obtained
for ©„
Considering the problem overall^ the more significant results ob-
tained appear to agree rather well with the values obtained by the
numerical solutions. The major discrepancies appear when the part of
the circuit handling the division is considered. As can be seen from
FigSo 6 and 12 for Z, and Fig, 7 for ©9 the outputs of these division
circuits change rapidly to a small negative voltage and then remain re-
latively constant for a period of time. We can also see from Fig, 10
that the division circuit does not do what theoretically it should,, Thus
for either parameter the minimum voltage desireds when r is a maximv?'.-
is never obtained. With Z9 this term is small when compared witli
others in summing for r and the effect is not pronounced. With (
however this defect has a more pronounced effect and % is not dev«
in the smooth curve desired.
21
9, Discussion of discrepancies.
The discrepancies found in the above problem solution were attri
ed to the division circuits used. No satisfactory answer could be found
as to why the desired divisions could not be obtained „ It is known
that a circuit such as this develops a certain amount of noise. That is,,
the function multiplier has a certain amount of noise inherent in its
output and that if this is put through an operational amplifier this
noise is amplified. The Handbook of Automations, Computation and Control,,
Vol, 2S by Grabbe (1) discusses this briefly and mentiors that a small
capacitor placed in parallel with the multiplier will help to alleviate
this problem. This was tried but did not give satisfactory results.
As mentioned previously, if a function generator could have been
used8 the circuitry could have been simplified, i,e„9 no function multip-
liers would have been required. With the Donner function generator the
3
slope of the function 576 8 000/ R s for low values of Rs exceed the maxi-
mum of two volts per volt permitted by the device. One other type of
function generator was tried. This was an Autograft XY plotter converted
to a function generator by replacing the recording pen with a' pick-up
coil and plotting the desired function with a conducting ink„ However 8
with this arrangement the desired range of voltages could not be obtain-
ed.
Still another type of function generator that might have proved
satisfactory s if it had been available, is the photo- former type. This
type of function generator operates as follows. The basic piece of
ment is a cathode -ray tube. An input voltage is applied between
23
.. 1 deflection plates of a cathode ray tube th* stable
amplifier. The voltage between the vertical deflect i< is pla
as the output voltage. This voltage is made to vat i funct .
of the input voltage by a feed-back arrangement which forces -. lectrom
beam to follow the boundary of an opaque mask placed over the lower por-
tion of the cathode-ray screen. Thus as the spot on the eathode-ray tube
screen emerges from behind the mask a photocell in front of the tube ap-
plies an error voltage across the input terminals of the vertical de£! i -
tion d-c amplifiers so phased that the beam is forced downward toward the
face of the mask. Therefore if the mask is shaped to represent the functis
being generated the spot will follow this curve and deliver an output
voltage proportional to the input voltage. This type of function genera-
tor is said to be very accurate in developing many functions. (3)
10. Conclusions.
Considering the results obtained from this problem (keeping in mind
that indeed it is but a single problem) 9 it was found that a "typical" non-
linear dynamics problem can be set up on an analog computer. However this
type of set-up is not done rapidly or easily. Considerable thought has
to be given as to what type of equipment shall be used and what kind of
-cults are necessary. Because they require the use of various types
of non- linear computer accessories the circuits become very sensitive and
results accurate to the degree normally expected from the analog computer
may not be obtained. Care has to be taken in selecting scaling faetorss
where powers and roots are involved,, to avoid over- loading the operation-
al amplifiers. It was found, however 8 that the function multipliers used
24
square and cube R gave quite accurate results^ even at It-
they were kept balanced.
In setting up a problem of this type it will usually be found
3
that there will be one key term to be developeds such as the A/R of
this problem,. Once a way is found to develop or represent this term
the remaining computer setup is routine and with patience and luck a
solution can be obtained.
2y
PHY
1, -. M. Grabbes S„ Ram© and D. E. Wooldridges Handbook of Auto-
mat ion, Computation and Control^, Volume 29 Computers and Data
Processings John Wiley & Sons,, 1959.
2, H. Mo Paynter9 A Palimpsest -~ frhe Electronic Analog Arts Geo., A.
Philbrick Researches,, Inc„j, 230 Congress St.9 Boston,. Mass^ 1955.
3, Go A0 Kom and T. M„ Korng Electronic Analog Computers 9 2nd Edition^
McGraw-Hill Book Co., 1956.
4» Operating Handbook Donne r Model 3750 Variable Base Function Genera-
tor 8 Donner Scientific Co.
5, Operating Handbook Donner Model 3000 Analog Computer,
6. Operating Handbook Donner Model 3730 Function Multiplier,
7o J, L. Synge and B. A, Griffith, Principals of Mechanics » 2nd Edition,,
McGraw-Hill Book Co.8 1947,
8o N. C. Riggs9 Applied Mechanics, The MacMillan Co., 1930,
9o R0 Co H. Wheeler, Basic Theory of the Electronic Analog Computer,,
Donner Scientific Co., 2829 - 7th St . , Berkeley,, Calif., 1955.
26
APPENDIX I
Numerical Solutions for Central Force Problem
The statement of the problem is given in Sec 3 on page 2„ Re-
presenting this information in mathematical termss we have F
spring
2
lQCr-5) lbs. and m = 1930/386 = 5 lbs sec /in.s and initially (at
time = 0) we have r = 4 inchess r = Oj, 0 = 0S and 0-3
radians / sec o Since energy is conserved 8 we have
where E>, Ts and V are expressed as energy per unit mass in units
of it. */ sec2. Here we have used V- L. '/*-«-(
Using the initial conditions to evaluate E9 we have
u
13
i
Ac apses n Apsidal radii are
given by n -$v - Substituting
and rationalizing w@ gets
7y -/Oft3 -j-nS'/j1-- • 3~
/ 3 - yr/7'4 -f-zJb
one root is 4 we obtain (/?-</) (
This can have only one positive root„ Synthetic division indicates a
root of approximately 13.2 and using Newton's method;
2 7
<x, ^ /3 - £6l) -_ )3 _ (-") _^ /3,/j-
To find the apsidal angle and r and 0 as functions of time,, we
resort to a numerical procedure since the integral involved is not
elementary. Returning to fundamentals we have:
,/7 — ftp* - ~a^-*~3
;o /,* ^ 03£T _^ y-/0
We also know r1 = 4 and r„ = 13.144, Now using an iterated
«
Integration, a curve of r = r(t) is assumed such that r = 0 at
the end points (apses') „ The apsidal time "<Z is divided into n equal
intervals /r\ ; C being as yet unknown. We will use n = 6S although
a larger n will give a more accurate result „
CO „
Assumed values of v* are selected for each epoch. Values of r
are calculated and integrated with the condition/?-^ at £T *■ °
This should yield/7 -0 at t-^ 8 but there is an error €L „ We remove
this error by using a correction curve which is essentially A ft- (2.^ -i-l ^j^-
expressed however in appropriate form for and obtained by numerical
m
integration. This arises from assuming that the error in r is due
to an error in r which must be essentially parabolic in nature9 vanish-
28
Ing at the end points since the apsidal distances are known . The rest
of the calculations are self explanatory and lead to the curved
shown in Fig, 9»
29
1
0
<j
■ <IJ
^
5
, \
<
.<
_
v V
JJ
^?
^.
r
—
- -
?
■:■
■
V
H
;^
sS
>
^
^
Of
n
< -
o
\ -
5
>
"t-
1 -
—
6a
«5»
>•
15
1
c
c
•r
4-
a
t
i
)
e
s
H
■
\
>
0--
it
r>
r1
~^v^.
t
3
^
Pi
■a*
>
<<
5k
Ok'
— >
Q
4-
C
a:
O
\
q)
>
?1
1—
r~
a
c
>
■«=>
£
■a-
0
^5
5r
a
i
,<■*
v»
•>o
<~-
\?
o
< 1
4
■ •!
>-
<
'^>
■^
>
-^.
— -
£3
•^
r*
)
o
t-1
\
3-
■3
•3
C
D
o
bo
*
<->
\
0"
Q
<>
>
S3
' 0-
q
ex
0.
n
O
^J
QZ
>•
O
5>l
'.A
O"
H
rV
<>
9
<>>
<<
I*
•A
d
^
V
/
\
•
bi
•H
^1
}
s
^1
.3
n
^>
r"1
o
o
>
<~
o
<T-
<r-
o
(-■>
a
O
—
fe)
l^
<
Vc
' &.
>
< J
i3
O
o
o
-C5.
Cm
-)
&
■'h^
>
1 c*
V-
u
0-
JZ.
•O
**
^
**"
tV
fh
""^s
o
o
o
o
r>
o
O
o
O
O
>
v>
g
r>
o
r0
0
*c
o
<3>
o
o>
o
0
■
i
^J-
^
>
H
/
i
.
)
1 1
|S
^
e°
c
?C
■^
<3
•^
>
1
1
c
^
'^
<^
*
^
zo
I9fiq QNZI
Second Numerical Method
From the expression for E, /f^ & +"A~ +-%(p~ v^
Upon substituting & /A , we get f^ as a function of A 2
and thus can construct a curve of 77 as a function of Jr\ (We take
the positive branch of the square root so as to deal with the period
during which r is increasing from 4 to 13.145 inches.). Also we have
J~1 ~ £3oV _ ^ /j -f~ J 0 so that we can construct a curve of yy
as a function of r, and this relation shows that fi - O when r is
& a.
approximately equal to 7,62. Having curves of both /? and f) as
functions of r, we can construct a curve of h as a function of A
The differential of time may be written in either of two w«ys
cJLu. c^> <$JX t and this permits us to write
ft rr
Aid) ^ 7 A Oj./vst)
^ r 4^ ^ r ^ ^ r ^i
so as to avoid infinite values for the integrands. These calcula-
tions can be carried out by numerical methods as shown on the follow-
ing pages and & is found to be 1.4863 secondss which agrees with the
value of calculated in the first numerical solution.
31
-
~*>.
O
c
1/
\
3
0
' ,
<5
CO
<3
e*
\
v\
sa
o~.
^^>
6»
-')
.V.
V
(
o
• >
L
^A
I'
l"
0s-
1*
O
r
o
J-
—^
_>v
£5
be
bo.
o
\ -
> **
d
■■ 5
s~*.
0
*<
©
O
co
DO
>
bo
>
< v
Ik
5^
0
0
0
U
C^
<S
to
CX
te
Ov
\»
<s.
M
1
Cjf
o
<r>
(V,
SS
O
>*
^
•->
i
r
CO
5
o
>■
C">
r>
o
o
if*
•a
o
Cl-
to
O
£
£
rs
.
0
v"
0
iH
CD
►*
TJ
0)
fi
ffl
"V
■P
CO
e
0
O
U
o
O
o
CI
5
o
i.J
n
1
o
8
U
0
S3
CO
a
0
0
0
■'...
1
v 1
*
H
O
H
O
1
o X
X
0
o
^
"0
>
«C1
>
0
"O
O
.-1
^
O
>
v3
OS
c
Vo
>-
■>
>
>
ro
0
r-l
00
O
O
o*
£
S2
O-
->
5v
EH
J*
o
>
o*«
rt^
"^
'O
-M
*K
°5
o*
C">
>
to
1
1
1
OS
•O
CV
1
^<
"Ns,
1
o*
•
hD
t
Q
s
o
o<
OJ
<c\
o
\o
§
S?v
c->
v'X
^
CO
6a
>
do
13
.
CO
n
rO
CO
O
>
0
\
\.1
CM
CM
CO
>
CO
CK
6^
0
0
en
0
ex
'(
p
T
Vo
<3
r-
00
CT5
0
"V.
■*••»
CM
n
>
Yc
'
n
i-QNIi 0<J«0 AA
1969 QN2I
'
s
0
0
o
CO
bo
v3
>
a-
\ -
o
•
— «
0
o
bo
■A
*->
$
o
to
o
$
cr
<0
>
CI
a;
r-
■V
0
Oi
■^
4s
•^
V
O
k(
>
**>
—
H
C
H
a
o
•in
0
H
X)
CO
Eh
■
•<&
<2>
O
n
o
OS
re
o
->
o
—
V
e ■
-
o
\
Q
< »
>>
fc»s
■
H
»
•
Ph
)
$
■■
bo
O
P 1
15
o
<5
e
o
o
P
*0
•o
r -i
.JV
0
Oi
>
^
i -
: i .
►
or
■ >
O
^
>
' •/»
o
1
0
\a
to
1 '
'. j
■ -
-
1'
•-
-
>>
$
i1 '
be
- 1
bo
1
i
33
9
/
8
/
7
6
5
3
o
1
j
1
1
n
U 5
7 8 9 10 11
Radius, inches
12 13 l!+
Fig„ A20 - r vs radius
36
APPENDIX II
Scaling Equations for Central Force Problem
The basic equations for this problem are:
a -- *%*-
The initial conditions at time t = 0 are:
m~--/ J w - oJ a-- o J 3 --* j
and we also know from the numerical solution that the approximate
maximum values of each of the parameters are:
Also from the numerical solution we know that the time from apse to apse
is about 1.46 seconds. For our solution we will select 0~-2>Tf » (1.5
revolutions) as we are only interested in the initial aspect. Knowing
the maximum values and using the relationship x - oC^ X s as explained
in Sec. 6, we can now solve for the scaling factors.
<• - /3./V ,/3/V /, U^<X--,X Ky^-G^V
/0O
h -Si—— —
p/.- - _3_ ^ ,03 / , LcL (*-£ - /Cy &y^ j 0*V
^ " /oo
For time scaling we wish to slow the problem time down so we assume
~C . - S'^p s giving a period for the computer of 7.3 seconds,, apse to
apse.
37
I
!
Now using the above scaling factors we can proceed to scale
basic equations for the computer:
(xA3 «»)
.'>
/•^ TIP
The equations are now in the form f% - —(JaJ. S — dJ., ^ ^~CU,/^°J
re Z = 576, 000/ R , and ^ = 30S000/R . Because Z and & are
developed by the division circuits and 20 is a constant voltage their a*
values (coefficient potentiometer settings) are each 1. Thus their
corresponding resistors are all 1M. To determine the value of a- we
use the relationship c<^j; :_ °^LJ&- . Equating 6*4, to c>^-£2-~ / ^
and letting R„ and R _ equal 1MS we find a to be .8.
38
To obtain R and R we must integrate ^j ** cx-Ls> and
We therefore scale these equations as follows:
'C
,tt'y
, 2 ' w
Now using the relationship ^<' \ - — ^~ we solve for a,.
as shown below. The resulting values are .833 and .48 respectively
using the resistors and capacitors shown.
To complete our scaling we must now determine tie a value
for C7 . cr is found by integrating ( &cx<=p . Scaling this
equation we find
1 fit, ^ - SilA Ja
*&
^---^1 CjJt
39
o/^ -
The problem is now ready for the computer. The table on the
following page summarizes the values for all resistors and capacitors
used for this problem.
40
1
'able of Circuit Elements
Amplifier
Circuit Element
Remarks
(See Fig. 3)
^and Value
1
R7I = 1M
R72 - 10M
Forms -Z
R1 "= 1M
r2 : im
r3 : im
*,'-
Rf2 : im
a^ -1.0
a2 s 0.8
83 = 1.0
* *
Sums + R
Rl ■ = IM
3
Cf^ = ly-f
a^ - 0.833
- f +R dtp = -R
R^ = 2M
k
ac; = O0I4.8
- f -R. dtp = +R
Rp = 5M
5
Cf5 = 1/jf
ag a 0.33
-j-*aV0
7
R81 = IM
R32 = 10M
*
Forms — @
R6 = 0.1M
9
R - 0.2M
f9
a£ =1.0
2K2R-*
Fig. A21
41
APPENDIX III
The Spinning Top
During this investigation, some attention was also directed to
the problem of the spinning top; that is a symmetrical top with one
pc int fixed under the action of gravity. Using the notation of Synge
a
nd Griffith, (7), the equation * ^ % O^^fr - **7*P0 '>l) -^'A^
was reduced to the form of }< ' #X3 ~ C XZ -h c* +" "^
A sample problem was then selected from Applied Mechanics by N, C
Riggs (8), which consisted of a gyroscope being released with & equal to
60 degrees and the subsequent motion being an oscillation between 60 and
82 degrees while precessing at a variable rate,,
With this problem, as with the central force problem, the powers
of x could be obtained using function multipliers. The key to the
solution however consists of developing the square root of X . This
should be possible using either a function generator or a division cir-
cuit similar to that of the central force problem. Using the function
generator was tried first. A curve of Y = 10 X l/29 X and Y
being arbitrary voltages from zero to 100,, was calculated and set into
the function generator. It was then found on scaling the problem that
to prevent the term CX from exceeding 100 volts it was necessary to
•2
assign to X a maximum voltage of approximately nine volts. Then
•n the circuit was assembled it was found that this voltage was too
small for the function generator to operate.
When the use of the function generator proved to be unsatisfactory
the use of a division circuit was attempted. To use this it was assumed
42
that the following relationship was valid., We know that for this
circuit Z - J_0C R^ X. Now letting Y equal Z and then so
R2Y 2 \/Sl
for X we find that X = R^ Z , or Z equals 10 X " , with
100^
R - R = 1M. This arrangement was then put into the circuit and
computations attempted. It was found that the operational amplifier
generating X went from its initial value through zero and then over-
loaded, i.e., the voltage X (representing x = Co<, cr )s did not
vary as it should since the circuit was unstable.
After the above attempts failed it was decided that this particular
problem could not be scaled satisfactorily for an acceptable analog
computer solution within the time available.
43
\