Doug Bollen 1968 PEAC

Download as pdf or txt
Download as pdf or txt
You are on page 1of 67

A low cost.

general purpose analogue computer of modem


design,
, intended for the amateur or student.
A usef~l toOl which is capable of solving.complicated problems
at high speed.
,
Can be used as a model to simulate mechanical systems and
electronic networks.
.
Extends enormously the scope of the amateur experimenter.
This series of articles will eX'P'aln In detail the desl,n. construction.
and o'P8rGtion of PEAC.

of the publicity afforded to computers favours


digital equipment. However, digital methods
tend to be disproportionately expensive for sptaU
installations. On the other hand, although analogue
equipment is ideally suited to limited, low-cost applications, it was not until the silicon transistor had become
firmly established, and enough practical published
information was available, that a start could be made
on designing analogue computing equipment to yield
a reasonable standard of performance in the lowest
possible price range.
OST

A WORTHWHILE PROJECT.
No doubt mallY readers will think that construction
of a true computer could involve them in a great deal
of time, money, and effort. They might also believe
that an average understanding of mathematics would
not be sufficient to equip them to operate a computer
effectively. However, the amount of time and money
spent building PEAC need be no more .than is consumed by a home constructed hi fi outfit of normal
proportions and performance, and the computer will
solve even simple problems a great deal faster than the
human mind or slide rule, once if has been programmed
to do so;
In fact, a general purpose computer can find application in almost every sphere of technical activity, and is
particularly useful in the electronic workshop, to the
point of becoming indispensable after a short period
of use.
UNIT CONSTRUCTION
PEAC is arranged in the form of units, and is
organised in such a way that reasonably advanced
computations may commetice upon completion of the
first unit, UNIT "A". The cost of building UNIT."A",
based upon typical retail prices at the time of writing,

will not be much above 25, and yet it will solve


algebraic polynomial equations, simultaneous linear
equations, simple differential equations, and can also
be used to simulate the behaviour of many elementary
mechanisms and electronic networks.
UNIT "A" is designed primarily to satisfy a minimum user requirement, for experimental and educational work, but it also serves as a convenient starting
point for the addition of further units to- expand the
computer to almost any desired degree of capability
and complexity. The additional facilit. pu;vided by
the add-on UNITS "B", "C", and "D" are described
in the specification. See also the block diagram, Fig. 1. 1.
A comprehensive PEAC installation, equipped with
a function generator and multiplier, and with fu))
integrating facilities for the fast solution of a range of
differential equations, might finally cost around 60:
not a lot to pay for an item of workshop equipment
which can solve electronic formulae in IOms,' and
which may also be employed as a variable waveform
generator, 18 input high quality audio mixer, variable
characteristic high Q audio filter, large inductante or
capacitance simulatQr, d.c. or a.c. millivoltmeter, and
many other things besides.

COMPARISON BETWEEN AtJlALOGUE


AND. DIGITAL COMPUTERS
Although popularly regarded as an inaccurate
machine of limited usefulness, the analogue computer
is to be found in the Polaris missile, spacecraft. aircraft,
large scale chemical processes, and many automated
production lines, quite apart from basic research work,
where flexibility and ease or working are often 'consid~red to be more important than extreme accuracy.
The analogue computer is, in most cases, very much
faster than its digital counterpart, and can offer far
more in the way of general facilities for a given outlay.

ANALOGUE METHODS '


The time taken to SQlve a problem on an analOgUe
computer is independent of problem le gth. All
The statement that an aeroplane is a machine for
circuits oPerate in parallel, simultaneously. ~ ty~ical
solving sets oC'Clifferential equations is not very far
solution might be arrived at in 2Oms? nd this solution
removed fronT' the truth. If the aeroplane did not
'solve its equations, correCtly it would not be able to
can then be repeated at the rate, say, of 2~ solutions
per second. In human terms the solution is virtuallf
fly at all. Almost all relationships or events ~ be
described mathematically, or in turn be represented by
immediate and continuo,!s, therefore: a'ny adJlllStments
'tnade to problem parameters (tenm! of an ~uation)
an analogy. A model aeroplane in a wind tunnel
solves, by analogy, roughly the same equations which
while the computer is WOl'king will be immediately
govern the behaviour of the real aeroplane, although
reflected in the solution readout. iThis rapid. response
/lllows the operator to quickly gain an insighj into the
in much simpler and less expensive fashion.
iWQrKings and structure of a roblem.
An analogy of a physical or mathematical process
could be achieved by a system of gears, pulleys, and
In contrast, digital computers perfonn many mathematical operations in a pre-determined and cOmparalevers; or by the controlled flow of gases or liquids.
But in the last couple of decades electronic methods
tively lengthy sequence, which bears little obvious
'relationship to the structure of the problem, but _hey
of simulation and equation solving have become almost
universal, because of the accuracy, ai.'ailability, and
do offer the very high degree of accuracy essential for
calculations involving money or very precise data.
\ adaptability of standard electronic components.
lbe main purpose 'of the analogue computer is to
The computer of the future will undoubtedly combine the best of both worlds with analogue and digital
allow a model to be set up quickly a!)d easily, to
equipment in hybrid fonn.
simulate the behaviour ot a full scale s}}tem, and at

The Practical ElectroniC' Malo,.,.


TIle
basIc equ f:lmell't I. UNIT <tAU.
Computln, facilities mjlf be e.xtencle#
by the addition of further units: "S",
"C" and "D". In accordance witH.
....r rfHJulremenu.
Thl. Pltotoff'flfjh .hows UNIT "A"
'ItGnCfIn, on UNIT 'S"
Compu&er Is of flexible des',n.

STABILISED
POWER
SUPPLY

205V
-245V
50H.-

Fig. 1.1 Block


diagram of

PEAC
INTEGRATOR MODE SWITCHING

51

SPECIFICATION

r!----r:----r!
:J 01
52 ....

S3:l

fl!!!!
L ____ -'_____ J

NUUI

+10
0

I N P--'---1
W UIN:,PUT I

-g>

VOLTS

.. --v

READOUT AND
NULL METER

0----1

c---..., 1"....

Ir"U
It
I

MASTER
POTENTIOMETER

UNIT "A"
POWER SUPPLY
Input 205V-245V 50Hz. Output 125V d.c.
Voltage
regulation better than 1% for loads of 0-200mA. and 2%
for 0-300mA. Total ripple 2mV. Complete short circuit
protection.

o---~

0----1

I
I

............

'>----\)

0----1 I ,,/
'>---;.-y/ ,
OPTIONAL UNREGULATED
POWER SUPPLY FOR
COMPUTER RELAYS

MULTI-PURPOSE OPERATIONAL
AMPLIFIER AVAILABLE WHEN
MULTIPLIER IS NOT IN USE

UNIT "B"
MASTER POTENTIOMETER
' 25 kilohm 300' wirewound; 25 watt. Two-voltage measuring
ranges O-IV and O-IOV. Scale length 14in. Accuracy
better than 05% of full scale.
READOUT METER
Centre zero 1000000-IOOJLA. calibrated O-O3V. O-IV.
0-3V. and O-IOV. Accuracy better' than 2% of full
_
'
scale.

AMPLIFIERS
Three multi-purpose operational amplifiers. each with five
silicon transistors. Open loop voltage gain greater than
5.000. Output 10V at 5mA. Current demand (average)
4OmA. ' Equivalent input drift under normal room conditions better than 05mV per hour. Unity gain frequency
response within 1% for O-IOkHz. and 5% for 0-25kHz.
Typical noise and hum at output 3mV.

'INTEGRATOR SWITCHES
Provision for three or more Integrating amplifiers. Compute
times ranging from 10ms-Is. Single shot,or repetitive
mode with "hold" facility. Current demand around 65mA.

VOLTAGE SOURCE
Five independent outputs. each continuously variable in
three steps giving O-OIV. O-IV. and O-IOV. Dial
setting accuracy better than 3% of full scale between dial
divisions 1-10. Total current demand SOmA.

UNIT "e"
FUNCTION GENERATOR
Diode function generator for parabolic and other functions.
Typical accuracy 2%. Frequency response to several kHz.

COEFFICIENT POTENTIOMETERS
Four 10 kilohm 270 potentiometers. Dial setting accuracy
better than 5% of full scale between dial divisions 1-10.

UNIT" 0"
MULTIPLIER
Four quadrant multiplication of two or more variable voltage
inputs. Also incorporates an operational amplifier which
may be used on its own to supplement the amplifiers of UNiT
"A". Frequency response generally better than 0-50Hz.
Approximate current demand around 75mA.

SUMMING' NETWORKS
Three five-input summing networks provided with voltage
check sockets. and plug-in computing components.

38

the same time solve the equation which represents the


System. Sometimes the computer will be used just for
solving equations or, alternatively, as a working model
only, depending on the nature of the problem. The
advantage of the electronic computer is that it will do
each, or both at the same time, with ease.
The computer is set up, or "programmed", for a
particular task by inserting computing components,
i.e. resistors and capacitor~. into sockets on the front
panel. This procedure will be described in full detail .
in due course.

computing element art arranged to be very close to


earth potential in the absence of an inp\1t voltage, it is
feasible to take the earth rail for granted and regard
all circuits as having only two terminals, instead of
the usual four .
. Although the symbol and function of cac::h of the
elements of Fig. 1.2 are common to all analogue computers, the actual circuit design and' choice of components will naturally vary from one computer to
another. For example, the time-division multiplier of
Fig. 1.2e is only one among many possible circuit
configurations for achieving multiplication of independent variables. Alternative approaches include the
Hall effect, the servo, logarithmic, and quarter square
mUltipliers.

ANALOGUE COMPUTER CIRCUITS


In the electronic analogue computer, the analogy is
created fundamentally by manipulating sets of d.c.
voltages. There is nothing to prevent
a.c. voltages being used-in fact they
often are-except that a.c. measurement techniques are generally less
INPUT~
accurate at low levels than d.c.
_OUTPUT
However, when simulating dynamic
T
.
processes . with d.c. voltages, the
computer will be handling a voltage
POTENTIOMETER
which varieswith time. In this context
it is more appropriate to regard a
waveform, even if it is a pure
OUTPUT
RZ
sinewave, as a d.c. voltage varying
-EZ,o-~,.,......-+-t
with time according to a formula
E3o-R3JWIr--'
Rf=Rl=RZ"'R3
which describes the nature of the
VIRTUAL EARTH
waveform.
The main computing element is the
SUM~ING AMPLIFIER
"operational amplifier". As far as
Eic o--"'NIr--......~"""'--,
operational amplifiers are concerned,
the decibel is much too coarse a
unit to use for the measurement of
frequency response, so amplitude
linearity is usually expressed as a
OUTPUT
percentage variation . over a fairly
restricted range of audio frequencies.
In some cases, for example, an operational amplifier and its attendant
SUMMING INTEGRATOR
circuits will be expected to respond to
inputs from d.c. to 5kHz with an
accuracy ora fraction of 1 per cent,
and up to 10kHz at no worse than 1
per cent.

COMPUTING ELEMENTS
The majority of problems can be
solved by the varied application of
only five analogue elements, but the
size of the problem to be handled will
in turn depend on the qmmtity of
elements available, and hence on the
overall size of the computer.
The five computing elements are
.s.h own in Fig. 1.2, together with their
. conventional symbols and generalised
Junctions. The symbols are used as
a kind of shorthand when drawing
up a computer programme.
. The first thing to note about the
simplified circuit diagrams of Fig. 1.2
is that the <:;ommon earth return is
often completely ignored. Computer
supply voltages are usually positive
and negative in relation to an
earthed centre tap. Since the input
and output terminals of each

xZ
OR iO=f(X)

DIODE FUNCTION GENERATOR

xo----t

I---oxr

r o - - -........
TIME DIVISION MULTIPLIER

Fir. '.2

Ancdorue computlnr element.

39

It is proposed to examine computing elements in


greater detail when they are dealt with individually at .
a later stage, but in the meantime a brief survey
will suffice.
COMPUTING POTENTIOMEl'ER
The potentiometer of Fig. 1.2a may be used ~or
multiplying a variable voltage (often called a machIne
variable) by a constant of less than unity.
Example: potentiometer input l' 5 volts. Slider
set exactly half way along resistance track, corresponding to a constant of 0'5. Output voltage Eo therefore
equals 15 x 0'5, or 075. As set, the potentiometer
will multiply any input voltage by 05.
When incorporated in the feedback loop of an
operational amplifier, the potentiometer will divide a
machine variable by a constant smaller than 1. The
fact that potentiometer constants are less than unity
is no real disadvantage. It is a simple matter to
either increase input voltages by a factor o~ ten, or
increase the gain of an operational amplifier ~en
times, to bring the potentiometer constant abov~ ~ruty.
Like the slide-rule, it is simply a matter of decidIng In
advance where the decimal point should be.
SUMMING AMPLIFIER
The summing amplifier of Fig. 12b uses a high gain
operational amplifier with several inputs . to achieve
addition and subtraction of machine variables. When
the operational amplifier has a voltage gain equal to.
several thousand, input voltages will be accurately
summed together, without unwanted interaction. The
summing junction SJ is at "virtual earth", a :-v~y of
saying that SJ will never be more than a few milhvolts
above or below earth potential, and is also, to all
intents and purposes, shunted by a resistance of only
a few ohms. Compared with input resistors RI-R3,
the SJ shunt resistance is very low indeed, a condition
necessary for accurate summing of voltages.
.
A definite relationship exists between resistors
RI-R3, and feedback resistor Rr, and if these resistors
are arranged to plug into the amplifier, many problem
conditions can be met by "ringing the changes" on
preferred values of fixed resistor, including multiplication by a constant as well as addition.
If a voltage El is applied via resistor RI (in Fig. 1.2~)
ot the summing junction SJ, the output voltage Eo will
This "hotograph shows UNIT "A" being used to simulate
a tuned Le circuit, consisting of an Inductance of 5H In
series with a ca"acltance of 5/LF. The oscilloscope Is displaying "hase shift within the simulated circuit at the
resonant frequency of 31Hz, and the trace also giYes an
Indication of the damping factor or "Q" of the circuit

be - El ;:. The operational amplifier is designed to


invert an input voltage, hence the minus Sign in front
of this expression. The ratio between input resistor
and Rr holds good for each input.
Example: apply three input v~ltag~s ~l = ?,
E2 = - 3'5, and Ea = 2 to the summmg Junct1~m via
RI = 10 kilohm, R2 = 2 kilohm and Ra = 100 kilohm.
Let the feedback resistor Rr = 10 kilohm. The
relationship between voltages and resistances will be

Eo

(El ~ -

2 ;: + E3 ;:)

Substituting

valu~s

(510 - 35!.Q + 2JQ.) = (5 - 35 x 5) + 02),


10
2
100
therefore Eo = 123.
.
In the above example, the summing amplifier has
not only summed negative and positive inputs, but has
also multiplied E2 by 5, and Ea by a constant of 0'1, .
merely by selection of appropriate values of input .
resistor.

Eo

='

SUMMING INTEGRATOR
The summing integrator is used for th~ detailed
investigation of time dependent variables, and for the
solution of problems involving calculus.
The integrator of Fig. 1.2c is based on the inverting
operational amplifier, with capacitor er acting as the
feedback component. The output from a single integrator, in response to a steady voltage input, is a linear
ramp voltage which increases with time at a rate
dependent on choice of input resistor,. feedba~k
capacitor, and input voltage. Once agaIn, precise
relationships must exist between computing components and voltage, but now time is introduced as an
additional analogue variable.
The action of electronic integration is best explained
by a working example, and reference should be made
to the diagram of Fig. I.3a.
Example: a fairly sluggish motor car accelerates
from rest at a steady rate of 20ft/second/second.
Examine the progress of the motor car during the
first four seconds of its motion. The computer is set
up t.o operate in "real time", that is to say, the t!me
actually occupied by the motor car when acceleratIng.
The problem layout of Fig, 1.3a shows a computing
potentiometer "A" coupled to the input of Integrator
"I", which in turn feeds Integrator "2". Voltmeters
are connected into circuit to display the three parameters of interest. Potentiometer "A" is first adjusted
so that its dial reads 2, corresponding to multiplication
by the constant 0'2, to represent 20ft/s2 scaled down
to yield a voltage of appropriate magnitude for t.he
integrators to handle. The output from the potentlOmeter is a steady voltage analogue of a steady rate of
acceleration.
As soon as switch S3 is closed to the +V position,
the velocity and distance meter pointers will start to
move in a manner analogous to the motion of the
motor car. Velocity will increase linearly with respect to
time, while distance will be displayed as an accelerating
pointer movement. Integrator "2" computes distance (s) as a voltage function of the square of time, in
'
.
terms of s = tal 2
With the problem of Fig. 1.3a, acceleration; velocity,
and distance are immediately available to the computer
operator as dial and meter readings. He can vary
acceleration just by turning the dial of the potentio-

D.e. VOLTMETERS
ACCELERATION
a=ft/.ec1

DISTANCE
S=ft

VElOCITY
v=ft/~c

'.la

(left). T"e u of Inte,rator. I.


Illustrated In
dla,ram. In t"'. examp'.
t"e rate of acce'eratlon. ve'oclty. and
dlJtCJnce covered by a motor car are computed and can be read off t"e potentlometer
dla' and meter .ca'e.

Fig.

t"'.

/
-VOLTS

+VOLTS

10

10

::..I~

<:>

...>

...

;:!

II)I~
<.J

OISTAHCE 5 AFTER

O~~----~----------~--~~~~~
~--~Ir-~~------HOLO------~.~I-'I--
COMPUTE
RESET

VI

STEADY VOlTAGE

Ci

FI,. , .3b. Arrestln, a computation to ,Ive a steady


1

TIME SECS.

Z
3
TIME SECS.

meter. If switch S3 is moved to the -V position, the


car will decelerate and stop.
COMPUTE, HOLD AND RESET
It is obviously inconvenient to take readings from

. voltmeters when pointers are on the move, and it is


impossible to do so if time t is very short, as with fast
events, or when the computer is speeded up to some
fraction of real time. The sequence governing switches
SI and S2, in Fig. 1.2c, is therefore arranged to provide
three facilities, called "compute", "hold", and "reset".
The purpese of the "hold" facility is to allow a
steady meter reading to be taken at any point on the
voltage/time curve output Elf an integrator. The high
.gain introduced by the operational amplifier effectively

vo'tmeter reading

multiplies the capacitance of Cr when the integrator


input is disconnected from input resistors and reset
resistan~ R r .
With amplification, Cr becomes the
equivalent of a very large capacitor which is capable of
holding a charge for a relatively long time. In prac- .
tice, the ability of an integrator to "hold~' or store a
voltage will also deper1ti on low amplifier drift.
Fig. 1.3b shows graphically the effect of compute,
hold, and reset modes, when applied to the distance
curve of Fig. 1.3a. In this case, it is necessary to halt
the computation after an elapsed time of 25s, and
obtain a value for distance in the form of a steady meter
reading.

lZ~----------------------'---~

10

)...

VI

I
'/

...........=>

/
/

'"

I /

8
!;:;

BIAS

//I

11

~I

OUTPUT

INPUT
BIAS VOLTAGE

/
SlOPE OETERMIlEO
8YRj

",

l,// ./\/./
3
2

././
./

././
./
'/

DIOOE NETWORK
OUTPUTS

/'
-----Z

INPUT VOLTS X

L BIAS VOLTAGE
FI,. , .4a ('eft). illustrating flow
can be constructed from a .erles

a mathematlca'

function

of stralg"t line cane_nu

F'g. , .4b (above). A .'n,'e diode network and Its output


c"aracterlstlc
41

The compute mode is initiated by opening SI and


closing S2 (Fig. 1.2c). After 25seconds, S2 automatically opens and the amplifier input-is left floating,
with Ct still connected between input and output and
holding a stored charge. A .meter couplep to the
integrator output will show the distance travelled after
25s of acceleration.
.
The "hold" period can occupy several tens of seconds,
and is usually at the discretion of the operator. To
begin a new computer run, SI is closed, discharging
Ct through R r , thus resetting the integrator output to
zero. The input Elc in Fig. I.2c, is to allow an initial
condition to be applied to the integrator, as in the case
of a motor car which does not start from rest, but is
already in motion when it a~lerates. When computing
and resetting times are shorter than about Is, voltmeter answers will appear to be given at the instant of
pressing the button which initiates the SI, S2 cycle. .
The above description relates to a "single shot"
computer run, where the operator adjusts, takes a
reading, adjusts, and so on. In the repetitive mode,
the hold facility is ignored and the computer keeps on
repeating the answer curve, for display on an oscilloscope, chart recorder, or XY plotter..
DIODE FUNCTION GENERATOR

In many computer applications it is necessary to


generate a voltage which varies according to some nonlinear function not provided by normal operational
amplifier techniques. The diode function generator of
Fig. I.2d will allow a mathematical function to be
. constructed from a series of straight I.ine tangents, 'as
shown in Fig. 1Aa.
Each straight line characteristic is obtained from a
single diode-resistor network, and when the outputs
from several networks are summed together a complete
function will result. The shape of the final approximated
curve ' is determined by adjustment of the .network
r~sistors. Apart from powers of x, and other functions, roots are achieved by placing the function
generator in the feedback loop of an operational
amplifier.
A single diode network appears in Fig. lAb, and
the slope of its output characteristic can be varied by
adjustment of RI. The diode breakpoint (the voltage
at which the diode starts to conduct) is dependent on
the value of Rb.
MULTIPLIER

The computing potentiometer will multiply a variable


by a constant, but special techniques must be used to
mUltiply one variable by another variable. The
procesS' employed in modern computers i.s akin to
modulation, where the gain of a circuit is controlled
by an applied voltage.
The multiplier should yield a product of correct
,sign when multiplying negative or positive variables,
and this isrea.dily achieved with the self-excited time
division circuit of Fig. I.2e. The time division multiplier
operates on the principle 'of modifying the mark-space
- and amplitude of a square wave in accord with two
voltage inputs. The filter of Fig. 1.2e extracts the
mean level of d.c. from the square waveform. An
additional advantage of the Fig. I.2e circuit is that it
can be arranged to cater for more than two variables.
For example, inputs Xl, X2, and X3 multiplied by
input Y.

Next month: Commencing the construction of

UNIT ""A".
42

;4
I

embarking. on constructional dl'taili, a few


words must be said concerning measuring and
\.0
test equipment required.

EFORE

VOLTAGE STANDARD

It is necessary, at an early stage of omputer


construction, to establish a voltage standard for setting
up the PEACcircuits.
Si1'lce relative voltage levels are more important than
absolute levels, one particular voltmeter of proven
reliability can serve as a voltage standard, and this
might well be a reputable testmeter which has a large
scale conveniently calibrated in terms of 0-10 volts,
with a d.~. sensitivity of not less than 20,000 ohms per
volt. Even if the testmeter has an error of 2 per cent
of the indicated reading on d.c. ranges, it should be
capable of reproducing a given reading, from day to day
under similar room temperature conditions, with much
greater accuracy.
In addition to use as a voltage st<mdard, the testmeter can, of course, be employed for setting up
problems, answer readout, comparative resistance
checks, and for general testing of all circuits. There is
nothing- to prevent re-calibration of the computer to
laboratory voltage standards at a later date, a1)d this has
been allowed for in the overall design of PEAC.

COMPUTER INSTRUMENTATION
Analogue computer instrumentation has much in
common with electronic workshop equipment. Among
those instruments likely to be of use to the computer
operator are : ap oscilloscope, a small collection of d.c.
voltmeters, an audio oscillator, an a.c. voltmeter, and a
component measuring bridge.

\,.. The oscilloscope need not conform to a modem


specification, and could be a government surplus
However, it is often an advantage to have a large
area, and redund~.mt television sets can be converted
computer readout purposes with excellent results.
limited bandwidth of magnetic deflection is no
advantage at normal computer operating ~.
D.C. voltmeters with centre zero scales are
useful for rough checks on the terms of 'a COInPllteJr::
equation, where, for example, the wish is to see
varies ' in relation to x when manipulating a
taneous equation.
A sine wave oscillator, with attendant a.c. vo'ltn1letc~;
will often be employed for work on transfer fUllLCti~l)nsf '
and for general electronic circuit simulation.
Finally, the component bridge is a help i'~~-k:=~'~
up plug':in computing components, and for
possible SOUl'ceS of ~or.
n is assumed that special classes of equipment, such
as the XY plotter, will not be available to the am,ateur;
and they are therefore excluded from further mention.

UNIT "A" CONSTRUCTION

The general form of construction adopted for


is based on a series of boxes made with laminates
white Arrnaboard or Formica and hardboard. The
resulting box is rigid and durable, with a surface which
easily takes panel transfers and. lines drawn in Indian
ink. With such a construction, it is possible to achieve
a professional appearance using only simwe WOOd1"'0l~k~i;'"
ing t o o l s . '
,
It is advisable to start with the UNIT '~A"
panel and case. This slightly unusual procedure,
building the box before starting on internal cm::uits.

13/4"

.~
T

S6

ON

10

:Ii

!I

f-5/~'
@

~T.,
1/2

@
S2

- - 13Ja"

@
S3

-tfi
"
3/4

10\.)
~.

314"

""

12

-,,'1

DRILL TO TAKE
6 B.A. SCREWS .

'A ;;

~M~:

ALL MINIATURE
SOCKETS IN 3/16 DIA .HOLES

1"

~
'~C'-1/2

1"

'. 1

DRILL AND FILE


TO TAKE SWITCHES

~"

1/;

+11V;

D.C. VOLTAGE SOURCE


318"

~~

~/4

@+ ., .G-

@1"

:iI+' ;~' ~l- .

10

~~

L@--------~@r--------~@@~---------~
LINES DRAWN WITH"
INDIAN INK

SUMMER 1

o o o
BLACK

@WHITE

RED

O'.,: ./'T' VELLOW

Fig. 2.2. Left-hand portion offront panel. Drilling details,


layout of components, and panel engraving. (Below the
broken line, there are two further sections, each a replica
of "Summer I")

BLUE

@GREEN

/ / - '\

)1

/--"I (

/ - ........ \ / -

........

I1

', _/1 \ _/1 ',,_/ ', _ . /

IT
I

5"

COEFFICIENT POTENTlOMETERS

VOLTAGE SOURCE

~- _+- -O-PE-R-A-Tl-ON-A-L-A-M-PL-'F-IE_R_1 ~~'


~L1LNE_S_!.D_R_AW_N_r--------------------~~1/2"

r-__________SU_M_M_E_R_1______
SUMMER 2

WITH INDIA~ .
.,..

__

OPERATIONAL AMPLIFIER 2

__________________ INj
1SUMMER 1

OPERATIONAL AMPLIFIER 1

11

i~==========================~
~1.~----------------------11~;----------------------~.1

Fig. 2.1. UNIT" A" front panel. Overall dimenslorrs and sectional dlvldln,lInes

106

"'~'" "'~'" "'~'" "'~'"


SK3

SK3

WHITE

SK3

SK3

BLACK

REO

8SK4

8SK4

SK4

SK4

Fig. 2.3. Reverse side of front panel. left-hand portion


(Fig. 2.2). showing components and wiring. Summer 2 and
3 are wired exactly as Summer I shown here. The three
terminals TU. 2 and 3 are mounted on the side of the box

J_ '- . ,..!'!t.,~
.
~
!
. " . e . ' ,;'
~: ~:~
. :.,: .~~ .
t- . ,.~i ~ .f:, .~ . ~e. .. . .:. .--J1
..... ' 0.-_."..
,. ,. ,. ,". ....-- --.
~-tJ~
,.
,.
,.
....
r,w
,..
I
..
. .~ .

t.:.

."."",-,o()'1

f:' i'='

' 1 "- "

"".0;:1

"'"

, 0<1'

i'='

SK3

BLUE

SK4

~ ..,.-~~

YELLOW

SK5~

",~s.

-o. "

"'l

+."'\,1 .

-.-~.-.. Y

.....

, ~-.

CII'''~::;.;.,

,'",-.",

UNIT "A" front panel

may be justified on two counts: firstly, the front panel


really forms a circuit which is designed to be accessible,
and is an important part of the unit; secondly, the
method of construction chosen brings economy by
dispensing with a self-supporting internal chassis
assembly, and much of the internal gear is actually
mounted on the front panel, or to the box itself.
UNIT "A" FRONT PANEL

To prepare the front panel, a sheet of white plastic


laminate, slightly larger than its finished size of
13in x 17i-in, is glued to a sheet of hardboard of the
same measurements with Evostick or a similar adhesive.
When firm, the panel edges can be planed, rasped, or
sandpapered down to size, while making sure that all is
square. Next, taking Fig. 2.1 and the photograph of
the front panel as a guide, mark out the main dividing
lines with a pencil.
The positions of all holes and slots may be found by
referring to panel drawings Fig. 2.2 and Fig. 2.5.
Establish hole centres by first marking with a pencil,
then indenting with a sharp spike. Note that all
drilling should be carried out from the plastic laminate
side of the panel, to avoid chipping the white surface.
It is important to handle tools carefully, and prevent
them skidding across the plastic surface and scoring it .
When all holes have been drilled, deburr them on the
reverse side of the panel with sandpaper, and check that
components will fit correctly before applying a coat of
clear varnish to the hardboard backing.
107

,I
I

YELLOW

1\1 1

SK6

SK5

SK4

SK3

G~~
BLUE

SK14
GREEN

1
/1I
TO OP-AMPS
2 AND3

TO OPERATIONAL
AMPLIFIER PANEL

Fig. 2.4. Right-hand portion of front panel viewed from


rear. showing components and wiring. Operational Amplifiers 2 and 3 are wired exactly as "Operational Amplifier
I" shown here

To finish the panel, draw in all lines and symbols


with a nib pen and Indian ink. If any mistakes are
made, the ink can be removed-when dry-with a
typewriter eraser, and surface shine restored with metal
polish. Lettering can be applied by the "rub-on" or
"stick-on" transfer methods, and should be protected
by a thin layer of clear varnish.
When the panel decor has dried, mount all sockets,
potentiometers, knobs with dials, switches, and the
neon mains lamp. Dials may be lined up on potentiometer spindles later.
UNIT "A" BOX

Rear view of UNIT "A" front panel

108

This time, the box is first constructed of hardboard


on a wooden frame, and is later covered with plastic
laminate. See fig. 2.12.
Cut and finish the four hardboard panels to size, and
cut the various lengths of softwood. The manner of
assembly could be as follows: attach wood lengths
A and C to top and bottom panels with panel pins or
countersunk woodscrews, gluing all joints. Attach
lengths B to side panels, bring panels together and
secure. Next, position D, E, and F. Note that there
is no length D at the back portion of the top panel so the
slotted amplifier mount F should be lined up vertically
with its companion E . All drilling must be left until
the plastic laminate is in place.

o
I
1
/1"-..'

1\ \ \
,-/1

Y
I
I

I
I

t""""-- 11/8- - - 1 - _ I
I

I
I

I
IlOkIl

II

------o------~-----~
--v~
------

------

-0- ---~--~
~

I COEFFICIENT POTENTlOMETERS

COMPONENTS

Fig. 2.5.
Right-hand portion of front panel. Drilling
detalls,/ayout of components, and panel engraving. Below
the broken line there are two further sections, each a
replica of "Operational Amplifier I"

UNtT "A" FRONT PANEL AND BOX


Resistors
RI-R5
9lkO (5 off)
R6-RIO 9100 (5 off)
RII-RI5 1000 (5 off)
All 5%, !W carbon film
Pre-set Potentiometers
VRI-VR5
2500 miniature wirewound slider
type (5 off)
VRIS-VRI7 500 wirewound panel mounting
type (3 off)
Potentiometers
VR6-VRIO IkO 3W linear wirewound, 20% or
better, 270 0 effective rotation (5 off)
VRII-VRI4 10kO 3W linear wlrewound, 20%
or better, 270 0 effective rotation (4 off)
Switches
SI-S6 Double-pole, on/off slide switch (Radio.
spares) (6 off)
Plug
PLI

3 way panel mounting mains plug and cable


connector

Fuse
FS I 15A cartridge fuse and 20mm fuseholder
Lamp
LPI Neon indicator lamp (Radiospares "miniature
200-250V panel neon" with self-contained
resistor)
Sockets
21 Red, 15 Black, 15 Blue, 15 Yellow, 15 White,
12 Green (painted green, see text)
48 miniature sockets, black or red to choice
Terminals
Insulated screw, to take 4mm stackable plugs
(Radiospares). I Red,1 Green, I Blue
Miscellaneous
Material for panel and box: Hardboard: 2 off 13in x
Sin, 2 off 18in x Sin, I off 13in x 171in. White
plastic laminate: 2 off 13in x Sin, 2 off 18in x5in,
I off 13in x 171in. Softwood: 52in x tin square,
4in x _in x lino
20 s.w.g. tinned copper wire. Insulated sleeving
Dials and knobs
Nine 0-10 2700 dial knobs (Bulgin type K400),
. black or grey

109

~~------~~~~------~~~------~~--~------~~~------~)

+125_ TU

~RED

O~~~---o--~~----~~~+---~~~~----~~~~--~~
-125~ TL2
) ~BLUE
TL3
GREEN

10

Fig. 2.6.

Circuit diagram of Volttlge Source section

=:E E'"
10

10

SK2 VR13

VR12
10kll
0

10~:

E~

SK4
8REN

SK4 GREEN

CP1 (RED)

CP2 (BLACK)

E'"

10

VR14
SK2 10kll

SK3

SK2

SK3
SK4 GREEN

SOCKET IDENTIFICATION
The following abbreviations will be used in the
programming instructions for PEAC. Applied
as prefixes to socket (SK) numbers. they clearly
establish the identity of the particular socket
referred to. For example. "VS2/SKI"; "CPI/
SK3" etc.

SK3

VS
CP
S
I
OA

SK4 GREEN

CP3 (BLUE)

CP4 (YELLOW)

Fig. 2.7. Circuit diagram of Coefficient Potentlometers section

SKI

INPUT 1
(RED)

SK2

SKt . SQ

INPUT Z

(BLACK)

SKI

SK2

SK1

SKI

SK2

INPUT 3
(BLUE)

Voltage source
Coefficient potentiometers
Summer
Input (Summer)
Operational Amplifier

SK2

INPUT 5
(WHITE)

Miniatun

Acluol si.e

soekets

Fig. 2.8. Circuit diagram of Summer I, 2, and 3

Fig. 2.11. Method of bending leads to


make ,,'ug-/n "rogramming resistors

INITIAL
CONDITIONS

::lVRI5
50n

SK14
GREEN

7m~

INPUT
SKB
WHITE

Grltn

r-I
I

Bloc\( JTO TI Pow.r


. Puc\( Transformer .

I
I

I
I

,-<)":-:o--~Or::.=::ce

I
I

'SA

I
I
IL ______ JI

Fig. 2.9. Circuit diagrtlm of O"erationtll Am"lifier I, 2, and 3

110

Fig. 2.10.

Circuit diagram of mains sUI>I>'y

WOOD

2 OFF 173,~' X ,,~' X,,~'

2 OFF '2"x,,;'x1f

4 OFF 4"

z'

@ ,

'
'

X',;' X";'

OFF'2 3,~'

X";'X";'

HARDBOARD PANELS

OFF 4" X"; X";' WITH SLOT

2 OFF 13" X5" (SIDES)

OFF 4" X 7,~' X 3,~' WITH SLOT

2 OFF 18" X5" (TOP AND BOTTOM)

Fig. 2.12. Constructional details of UNIT "A" Sox

-Cut plastic laminate to fit hardboard panels wifh


1\-in overlap, and glue to the box sides first. . Reduce
the overlap to size when the laminates are firm, before
fitting the top and bottom surfaces. When trimming
the top and bottom panels down to size, take care not
to scratch and score the side pieces. For economy, the
bottom plastic laminate layer can be omitted.
When satisfied with the laminated exterior, the
1062in dia. hole can be made by a series of small
drillings and finished with a half-round file. The box
interior and wood may be varnished, but the raised lip
at the front of the box is best painted black, or some
dark colour, to contrast with the front panel.
The finished box is quite strong, and will support the
full weight of a normal adult when the front panel is in
place. However, it is recommended that this test
should not be applied too often!

FRONT PANEL WIRING


Attach the front panel to its box, which will act as a
convenient mount when wiring the back of the panel.
The bare earth wire linking all green sockets runs
along the top half of the front panel and down its lefthand side, looking from the back; this should be
soldered in place before embarking on the sleeved
wiring. (No matching green sockets were available
for the prototype, so odd coloured socIs-ets were
painted green with cellulose model aeroplane dope.)
The 4mm red, green, and blue terminal sockets on .
the side of UNIT "A" are designed to take stackable
plugs, and will make available the power supply outputs to external sub-units. Wiring can proceed from
the terminal sockets along the voltage source (see
Fig. 2.3) and then to the rest of the front panel.

Circuit diagrams for all the various "sections"


incorporated in the front panel are given in Figs. 2.6 to
2.10 inclusive. Wiring details are given in Fig. 2.3 and
Fig. 2.4.
The summer and operational amplifier sections are
triplicated-although only one of each of these sections
has been shown in the diagrams Fig. 2.2 to Fig. 2.5
inclusive.
The purpose of the miniature sockets, which appear
in the above mentioned diagrams, is to take the plug-in
programming components; explained by Fig. 2.11.
Resistor leads are preformed in the manner shown.
The distance between miniature sockets is standardised
at 1in, to allow the use of a special made-up two pin
plug to support the bulkier components, such as large
polyester capacitors.
When wiring up the operational amplifier sockets,
ignore for the time being the coloured flexible wires
shown in Fig. 2.4 as these are the flying leads from the
operational amplifier panel, and will be referred back
to when the time comes to mount the amplifiers.
Fit the mains connector PLl and fuseholder for FSl
to the side of the box. Wire up the neon lamp LPl
and the fuse FSI to PLl as shown in Fig. 2.10,
CORRECTION.
In Part 1, Page 40, last line of the equation in the example at
top of right-hand column should read:
Eo

=-

10
1010)
( 510 - 3-5 2 + 2100

=-

.
(5 - (3-5

x 5) + 0-2),

therefore Eo = 12-3.

Next month:
amplifiers

Power supply and operational


111

PEAC
By,
D.BOLLEN'

The current res.erve of the stabilised supply will just be


sufficient to cater for the needs of UNITS" A to 0". If
further expansion of the computer is contemplated,
beyond the inclusion of UNIT "0", a subsidiary
unstabilised supply can be added to the computer at a
late stage of construction, to power the relays of
UNITS, "B" .and "0", and thus make available some
extra current from the stabilised supply.

E main design target for the PEAC power supply


.
was a maximum voltage variarion of not more than
1 per cent under all normal operating conditions.
Several circuit configurations were tried, based upon
either .series or shunt regulation, but it was found that
shunt regulation invariably gave the best performance
for a given cost, plus the bonus of complete short-circuit
protection. It was not considered to be.a disadvantage
for computer work, where nearly everything is switched
on for most of the time, that a shunt regulated supply
would be wasteful of power under no-load conditions.

fig. 3.1. Circuit diagram


stabilised power supply

+30V

IL

DZ
51AR2

zov
12V
0

12V

1
1
1
1

C2
03
SIAR2

20V

04
SIARZ

I
lE

...

1000I'F

200

Cl
1000pF

IN
1
I
I
I

+IZ5V

11
1
1
1 1
1
1
I1
1
I

r-':" " (.--,


I 51
I
I
I

C3
1000pF

r,--'
1

The circuit of Fig. 3.1 is based on a small, standard


type of rectifier transformer, with bridge rectification

of the

245

Pl1

STABILISED POWER PACK

DV

RZ In

-IZ5V
ALL D.C. VDLTAGES MEASURED
WITH RESPECT TO THE DV LINE

125

0%
1%

-\--

100
200
300
LOAD CURRENT mA

2%
3%
4%

OC 29
OC 36

COt
o

@b

AC126

ACY28

f.Ob\

~RED

SPOT

400

Fig. 3.2. Performance curve of


stabilised power supply

CATHODE

Fig. 3.3.

followed by two shunt regulators arranged in series to


give positive and negative outputs relative to a zero
voltage earthed centre-tap. Diodes DI-D4 provide
full-wave rectification of the 40V r.m.s. nominal
transformer output. Capacitors Cl and C2 are wired
in series, with their common connection taken to the
transformer centre-tap, and this doubles the capacitor
voltage rating without the need for bleeder resistors.
RI and R2 achieve some measure of preliminary ripple
smoothing while dropping the unregulated d.c. voltage
to a safe value for C3 and C4.

COMPONENTS
UNIT "A" POWER PACK
Resistors
RI, R2 70. 07A power resistors S ~~ (2 off)
R3 4000. SW wirewound S%
R4 3000. SW wirewound S%
RS. R6 Ik0. 2W carbon 10% (2 off)
R7 600. 0'7A power resistor S ~u (two 30n in
series. see text)
RB. R9 loon IW carbon 10% (2 off)
Potentiometers
VR I. VR2 soon 3W panel mounting. wirewound
(2 off)
Capacitors
CI-C4 I,OOO!-,F elect. SOY d.c.
(4 off)

900mA rippled

Transformer
TI Rectifier transformer. Standard mains primary.
Secondary. 20V- 0- 20V 0'7A (Radiospares)
Diodes
01-04 SIAR2 (Westinghouse) or 002026 (Lucas)
(4 off)
OS.06 ZSOISOBF (STC) or ISSOISR (Texas) (see
t ext) (2 off)
Transistors
TRI. TR2 ACY28 (STC) or ACI26 (Muliard) (2 off)
TR3. TR4 OC29 or OC36 (Muliard) (2 off)

2N3904
2N3906

~
~

~WHITE
SPOT

SIAR2

'~

OAZ02

l:::~

f
o

IB30

COLOURED BAND
OR END

2N697
2N696
2N1893
b

V
o

Z50150BF

"~ ~

CATHODE

Transistor and diode key

SHUNT REGULATORS
To understand the action of the twin shunt regulators,
temporarily assume that the -125V output terminal is
at zero voltage. The centre-tap and the positive outputs
will then be positive in relation to the negative output.
TR3 and TR4 collector-emitter voltages are both
clamped at 12'5V, and the unregulated d.c. voltage is
dropped across R7. Therefore, the voltage appearing
at the junction of R 7 and TR3 emitter is + 25V
relative to the assumed zero rail, with the centre-tap
output at + 12 5V. As all three output terminals are
floating, it is a simple matter to connect the centre-tap
output to an external earth and classify it as the zero
voltage rail, with the other terminals forming positive
and negative regulated outputs.
VRI setting will determine the voltage across TR3,
and VR2 the voltage across TR4. The range of adjustment of VR I and VR2 is sufficient to allow for regulator
diode (D5 and D6) tolerances on nominal voltage of
15 per cent, and will therefore permit the use of
manufacturers' rejects or "bargain" price regulator
diodes. IOW diodes are specified for D5 and D6 in the
Fig. 3.1 circuit, to achieve a low dynamic resistance, and
reduce the short-term thermal changes which are
inevitable when smaller regulator diodes are run at
high temperatures.
Fig. 3.2 will give an idea of the capabilities of the
regulated power supply, and maximum current limits.
If an optional press-button switch is wired across one
half of R 7 (Fig. 3.1) output current can almost be
doubled for short periods, and special purposes. The
prolonged use of this extra current facility will, however,
result in mains transformer overheating.
POWER PACK CONSTRUCTION
Low cost semiconductors were used throughout the
prototype power pack. The diodes DI-D4 should
have a p.i.v. rating of not less than lOOV, and a
maximum current rating of lA or more. It is advisable
to check all diodes with an ohmmeter, for high reverse
resistance and correct polarity. The D5 and D6

Miscellaneous
Four capacitor clips to fit CI-C4
S.R.B.P. panel 4in X 12in X r.in or tin
4 B.A. and 6 B.A. assorted screws. nuts. washers. and
solder tags
Insulated sleeving
20 s.w.g. tinned copper wire
16 s.w.g. sheet aluminium 2 off 4in x 4in. and 2 off
Itin X I~ill..

201

Fig. 3.4 Power supply component layout

'~

-I l

Fig. 3.5 Underside wiring of power supply panel

DRILL HOLES TO
SUIT VRl AND VRZ
IN ONE HEAT SINK

16 SWG
ALUMINIUM
' (Z OFf)

16 SWG
ALUMINIUM
(2 OFF)

Fig~

3.6.

. 202

Power supply heat sink details

measured voltage, when passing a current of about


l00mA, can fall anywhere within the limits 125-175V.
If a choice exists, TRI-TR4 can be selected for
highest beta gain, but matching is not necessary.
Collector-emitter leakage currents of TRl and TR2,
with open circuit base, will preferably be below about
200pA at normal room temperature.
It is seriously recommended that the reader who
intends to build PEAC should adhere closely to the
semiconductor types specified -here. and not consult
other lists of equivalents. A key to transistor and diode
connections appeafs in Fig. 3.3, and this covers all the
semiconductors used in PEAC circuits.
Power pack components are assembled on a -firin or
tin s.r.b.p. panel measuring 4in x 12in. The panel
sits on the wooden framework at the bottom of the
UNIT HA" box. Component layout appears in
Fig. 3.4, with the underside wiring in Fig. 3.5. Heat
sinks for TR3, TR4, D5, and D6 are made up from
16 s.w.g. aluminium sheet, and measurements are given
in Fig. 3.6.
.
First drill the s.r.b,p. chassis panel to accept hardware
and wires, using Fig. 3.5 as a guide. Mount the mains
transformer, capacitor clips, power resistors, and the
three output terminal screws. Attach the regulator
diodes, with their heat sinks and solder tags, to the
panel, taking care not to damage the diode top
terminals. Bolt TR4, VRl, and VR2 to the appropriate
heat sink, solder R9 to TR4 emitter and base pins, and
install the assembly on the s.r.b.p. panel. Similarly,
bolt TR3 to its heat sink, complete with R8, and fix to
panel.

Both power transistors should have a solder tag


attached to their upper mounting bolts to make convenient connection to transistor collectors. ' Without
insulating washers, 'T R3 and TR4 heat sinks will be
"live", but damage is unlikely to result in the event of an
accidental short-circuit.
Insert capacitors CI - C4 in clips, with polarity as
indicated on Fig. 3.4. Also observe correct polarity
when mounting diodes DI-D4. Before wiring up all
components, insert R3 in the panel, alongside TR3
heat-sink.
COLOUR CODED WIRE

Wiring can start at the input end of the panel, with


6in lengths of orange, black, and green multi-stranded
wire soldered to the live, neutral, and screen tags on the
mains transformer. Red and blue wires are reserved
exclusively for 12SV d.c. positive and negative supply
rails, with green wiring as the common earth throughout
the computer.
Wire colour coding is almost essential for computer
circuit interconnection, as it enormously simplifies
fault tracing and assembly. However, the wiring of
individual circuits, such as the power pack panel, can
take the form of single colour sleeved 20 S.W.g. tinned
copper wire.
H will be noticed (Fig. 3.4) that TRl and TR2 are
supported only by their leads, and this is to allow best
positioning for good ventilation, well away from heat
sinks. In the prototype R 7 was made up from two
07 A power resistor sections, to allow for the optional
extra current facility mentioned earlier.
When power pack wiring is completed and checked,
multiple solder tags can be fitted to the three output
terminal screws.
TESTING THE POWER PACK

Connect the transformer input leads to the mains


socket on the side panel of the UNIT "A" box, with
the organge lead taken via FSl (see Fig. 2.10 and
Fig. 3.1), and, also join the neon indicator leads to the
live and neutral mains socket screws.
Turn VRt and VR2 fully anticlockwise and switch
on. A quick check with a voltmeter will show if there
is any serious departure from the voltages shown in
Fig. 3.1. If any overheating of heat sinks or mains
transformer seems imminent, switch off immediately
and locate fault.
To set up the power pack, apply voltmettlr leads to
earth and positive output terminal, and advance VRl
for a reading of 12SV. Repeat the procedure for the
negative output and VR2. If it is impossible to bring
an output to 12'SV, this will indicate a wiring fault or
trouble with a regulator diode.
After the power pack has been left on for some time,
. VRl and VR2 can be finally trimmed for exact outputs
of 12SV. With no external load on the power
supply, TR3 and TR4 heat sinks can be expected to run
fairly warm .
To ensure that power pack regulation conforms to the
curve of Fig. 3.2, positive and negative outputs can be
loaded by a selection of SW resistors in series with an
ammeter, while voltage is still being monitored. A
worst case variation of 2 per cent change in voltage for
300mA change in current should be taken as an acceptable performance limit. When one half of R 7 is
temporarily shorted out, at least SO per cent more
current should be available before voltage drops
beyond 2 per cent.

Locate the power pack inside the UNIT "A" box,


and wire outputs to the main terminals TLI, TL2, and'
TL3. Voltage source dial alignment and setting up
details will be discussed later, but a few [ouglf checks
with power on are in order, to see that all yoltage
source sockets and switches are functioning correctly.
OPERATIONAL AMPLIFIER

The most important analogue computing circuit is


the operational amplifier; so named because it will
perform a number of mathematical operations, such as
addition, subtraction, change of sign, multiplication by
a constant, division by a constant, and integration. All
the thinking behind "op-amp" design is concerned
with making the circuit as unobtrusive as possible, so
that it can be regarded purely as an operational "black
box".
An analogue computer d.c. operational amplifier
should comply with the following general requirements.
(a) Direct coupling between all stages to handle d.c.
signals. Input and output terminals at earth
potential in the absence of a signal, with 180 degree
phase change (inversion) between input and output.
Output voltage swings both positive and negative
in relation to earth, and as large as the computer
reference voltage ( 10V).
(b) Large voltage gain in the open-loop configuration.
(c) Low output impedance.
(d) High input impedance.
(e) Very low input current.
(f) Sufficient bandwidth to cause negligible phase shift
or attenuation of a signal up to the highest frequencies encountered.
(g) I nsignificant output voltage drift over several
hours.
.
(h) Good margin of stability when subjected to a wide
range of different input, output, and feedback
conditions.
Performance figures for UNIT "A" operational
amplifiers are given in the Table 3.1, but to fully
understand how some of the design problems are solved
it is necessary to consult the actual "op-amp" circuit of
Fig. 3.7.
OPERATIONAL AMPLIFIER CIRCUIT

The input stage of circuit Fig. 3.7 consists of a longtailed pair (TRI, TR2), offering the advantages of high
voltage gain, near zero input .offset voltage relative to

+125
RED
WIRE

RB
4700

TR4

2N3906
INPUT
YELLOW
WIRE
OUTPUT
WHITE
WIRE
R7
15kO

-12'5

'"

COMPONENTS

. ..

EXTERNAL FEEDBACK
COMPONENTS

..-

UNIT HA" OPERATIONAL AMPLIFIER


The following items are for a single amplifier. and are
~equired in triplicate to cover the three amplifiers employed
in UNIT HA".
Resistors
RI 39Mfl 5';{, ca rbon film
R2~ R4
27kfl 5% carbon film (3 off)
R5
~2kn
.
R6. R7 15kn (2 off)
R8
470n
R9
39kn
RIO 330n 2W
RII loon
All
10% . tW carbon composition. except where
otherwise stated

Potentiometer
.
VRI 5kn vertical skeleton pre-set
Capacitors .
Cl 0'0221-'F miniature polyester 250V d.c.
C2 39pF polystyrene 125V d.c.
Transistors
TR 1- TR3 2N2926 orange (General Electric) or
2N3904 (Motorola) (3 off)
TR4
2N3906 (Motorola)
TRS
2N 1893 (Bentron). 2N696. or 2N697
(Gene raJ Electric).
.M iscellaneous
S.r.b.p. panel 2in >: 2!in
Eight small turret tags
TO-S transistor cooler Type BCIOSB (Bentron)
6 B.A. screws. nuts. and spacers
Stranded core p.v.c. wires; red. green. blue. orange.
mauve. grey. yellow. and white
12in x 4in s.r.b.p. amplifier mount
Note: All transistors and cooler can be obtained from
Rastra Electronics ltd .. 275-281 King Street. Hammersmith. W.6.

204

BLUE
WIRE

Fig. 3.7. UNIT uA" Operational


amplifier circuit

earth, and low drift with change in temperature when


TRI and TR2 are closely matched. The long-tailed
pair also gives good rejection of drift induced by changes
in supply voltage, and has a reasonably large input
impedance at low collector current levels.
An input signal will undergo a phase change of
180 degrees between the base and collector ofTRl, and
the voltage datum level is shifted away from earth
towards the positive rail voltage. Ignoring for the
moment Cl, the signal is passed straight t.o the base of
TR3.
VRl, R5, and R7 form an adjustable potential
divider across positive and negative supply rails, and the
VRI setting determines the working points of direct
coupled stages TR3, TR4, and TR5. Front panel
control VRI5 sets the amplifier input at zero volts,
while VRl does the same for the output.
TR3, while contributing some voltage gain. also
introduces another 180 degree change of phase, to
bring the overall phase difference between the amplifier
input and TR3 collector to zero. Obviously. the
voltage at the collector of TR3 will be even closer to
positive rail voltage than the collector of TR I, but this
cumulative voltage shifting can be virtually eliminated
by using a pnp transistor for TR4. At the same time,
TR4 common emitter stage brings more voltage gain
and another and final phase change of 180 degrees.
So, the situation at the collector of TR4, when VRl5
and VRI are at correct settings, will be no overall
voltage shift, a total phase difference of 180 degrees, and
a total voltage gain in the region of 5,000.
Finally, the addition of an emitter follower stage
provides the low input impedance requ\J)ed for driving a
variety of useful loads, without unwanted circuit
complications. TR5 causes negligible further voltage
shifting, adds no change of phase, and with a voltage
gain very close to unity, will simply reduce the output
impedance of the operational amplifier without modifying its other characteristics.

IMPORTANCE OF HIGH OPEN-LOOP GAIN

SOr---------~----~~~----,_~------,__,

ideal operational . amplifier would have an


infinite voltage gain when no feedback resistor was
present, but since this is unattainable in practice, the
effect of a finite open-loop gain on amplifier accuracy
must be examined.
In Fig. 3.8, selected values of open-loop gain - A are
plotted against closed-loop gains - G, and percentage
amplifier error. Closed-loop gains are normally
restricted to 0,1-50 as this caters for almost all operational conditions, and it is seldom required to extend
these limits. A different set of circumstances apply
when the op-amp is used for integration, and these will
be considered in detail later.
Very high - A gains bring attendant drift and
stability problems, and this in turn demands a larger
number of components and more complicated circuitry
to keep drift and stability within acceptable limits. At
the opposite extreme, very simple amplifier circuits can
Th~

OPEN-LOOP
GAIN -A
~10~----~~~------~~~~--~~------~
I

:cC>

a.

.....
I

~ lr-~--~--r-~~f-~r--f----~r--f-----;
o
.....

0~0~0~1---------::""'....L------'1U----"""'-L-----~10

"10

AMPLIFIER ERROR

Fig. 3.B. Open-loop ,ain plotted alainst closed-loop rain,


and percentage amplifier error
Rfl00kll

Rfl0kll

100kfi

~
A

10kll

10kfi
LOAD

_LOAD

(a)

-...

...

"

-2,5
0
""

""""w

if?

-5

1\

.1;1

Co'
.p

j!lf'

-75

.'%;

. &~

...
;

-10

l'

'.Ii%

100Hz 200Hz
150Hz

.' '*..

,. . .i

L>II,.,'/
.~

(it
10Hz

!J?'

I!o

I~I

w .....
1'\

10kHz 20kHz
15kHz 30kHz

1kHz
FREQUENCY

TABLE 3.1
UNIT "A" OPERATIONAL AMPLIFIER. TYPICAL
PERFORMANCE
Supply voltage 125V 0'5%
Open-loop voltage gain 5,000 d.c.-IOOHz. 200 at
10kHz
Maximum output voltage 10V for loads > 2kO
5V for loads > 3000
Input impedance 4OkO approx.
Input current 0005,..A for I V out
Closed-loop frequency response O-l'OkHz within
1% when Rr =
100kO
Equivalent Input drift 05mV per hour
Input' offset voltage and current almost zero when
amplifier
correctly
balanced .
R.M.S. noise, referred to input with input open
clr.cuit 200,."V
_
Normal maximum range of plug-in components
RID 2-IOOkO
Rr 100100kO
.
Cr I-OOI,."F
Stability. unconditional with all normal problem
layouu

Fig. 3.9.

Amplifier frequency response

be built to yield -A gains in the region of 100-1,000,


but when -G approaches 50 the errors of such
amrlifiers would be near 10 per cent. Thus, if a low
value for - A was chosen, for the sake of simplicity, the
range of available closed-loop gains would have to be
restricted if the error was not to exceed one or two per
cent, and this would place severe limitations on the
operational flexibility of the amplifier.
It was assumed that PEAC operators would not wish
to employ plug-in computing components with a
selection tolerance better than, say, 1 per cent.
Therefore, the error contributed by the amplifier will
preferably be less than external component errors, but
not so small as to call for ridiculous extremes of
. circuit sophistication. The thickened .curve of Fig. 3.8,
corrc;:sponding to - A = 5,000, shows that the
maximum error contribution of UNIT "A" amplifiers
is 1 per cent or less for - G gains of less than 50.
BANDWIDTH AND STABILITY

A direct coupled amplifier of the Fig. 3.7 type will


display an almost constant phase change of exactly
180 degrees over a range of frequencies from d.c. to
205

about 20kHz. Thereafter, with increasing frequency,


.the phase angle will begin to shift until, at several
hundred kHz, and especially when the amplifier has a
high gain, sufficient positive feedback is present to cause
sustained oscillation. To counteract this instability,
small capacitors are suitably situated in the op-amp
circuit to reduce gain at critical frequencies, and it
follows that the use of such capacitors will place a
limitation on the available frequency response of the
amplifier.
Cl of Fig. 3.7 will block the unwanted high frequency
content of incoming signals, and plays a major role in
determining the bandwidth of the amplifier. If Cl is
reduced in value, bandwidth will be increased, but so
will the likelihood of instability. Needless to say, any
form of instability will be highly detrimental to
accuracy, and must be avoided at all costs. C2 works
in a different way, by introducing negative feedback and
consequent loss of gain at very high frequencies. Both
capacitors act together to combat instability under the
very varied conditions of operational amplifier use.
The measured frequency response of a representative
UNIT "A" amplifier is given in Fig. 3.9, and is very
linear up to the well-defined break frequencies of (a)
open-loop, (b) with feedback resistor of 100 kilohm, and
(c) when RI = 10 kilohm.
DRIFT
If a d.c. amplifier is adjusted so that its output
voltage is zero when there is no input signal, over an
interval of minutes, hours, or days-depending on the
amplifier, its power supply, and its surroundings-a
spurious voltage will begin to appear at the output. A
poor amplifier in adverse conditions will require
frequent manual adjustments to keep its output at zero.
Fortunately, drift errors are very small wher- an
operational amplifier is used for summing and sign
changing, due to the presence of a feedback resistor, and
!l0 adjustment of the amplifier will be called for during
mtervals of perhaps several hours, except in applications
requiring a very high degree of accuracy. However,
when the operational amplifier is being used as an
integrator, with a capacitor in its feedback loop, it is
quite possible for drift errors to exceed I per cent
within a space of less than an hour if suitable
precautions are not taken.
The figure quoted in Table 3.1 for drift is the amount
of input voltage, either positive or negative, required at
the amplifier summing junction to reset the amplifier
output to zero after it has been allowed to drift fcir one
hour following a preliminary computer warm-up
period. In practical terms, a drift of about 05mV

RI
820kn

IIIA-PIOO

~2'.

"\ '.......TRANSISTOR
SOCKET

8'
\

"- .........

__

..-

/
E

9Y

I BATTERY

-L.

I
I

R,. 3.10. Transistor test-rl,. Note: re.,erse battery


connections ancl milliammeter for pnp

206

per hour is not likely to prove to be too troublesome


with most PEAC applications. Full scale analogue
computers are sometimes installed in a temperature
controlled computing room, and this considerably
improves drift performance.
TRANSISTOR SELECTION
Several prototype UNIT "A" amplifiers were
constructed us.ing non-selected transistors, and about
one third of the amplifiers failed to meet the
specification of Table 3.1. Defects were due entirely to
"spreads" in semiconductor characteristics, and disappointment will be avoided if all amplifier semiconductors are tested before use.
It has already been mentioned that the long-tailed
pair input stage transistors (TRl and TR2 in Fig. 3.7)
should be matched. In all nine transistors of the same
type will be required for TR1, TR2, and TR3 in the three
operational amplifiers, and it wiII assist the matching
and selection process if, say, one dozen transistors are
purchased at the same time. No wastage will be
involved as "spare" transistors can later be used up in
other PEAC circuits.
A simple test-rig circuit is given in Fig. 3.10 to
facilitate the matching of TRI and TR2, and the circuit
can also be quickly adapted for checking other transistors. The test-rig could take the form of a transistor
socket and resistor mounted on an odd piece of s.r.b.p.,
or Veroboard, with a testmeter employed as a miIliameter.
Select each TR1-TR2 pair for near identical betas of
100 or more; this will dispose of six transistors. Do
not attempt to pair off transistors of different types even
if they do have the. same beta. From the remaining
transistors, choose three with the highest beta for TR3.
Although TR4 is a pnp transistor, it must be of
silicon construction for low leakage drift. The
majority of pnp silicon types at present on the market
are unsatisfactory for use in the op-amp circuit because
they exhibit almost no gain at all at very low collector
current levels. Of all the types so far tested only the
2N3906 was found to be consistently good at low
currents, therefore a suitable equivalent cannot be
quoted. To check TR4, reverse the battery leads to the
Fig. 3.10 test-rig, and switch connections to the milliameter before plugging in the pnp transistor. TR4
should display a beta of about 50 or more.
When handling plastic encapsulated transistors,
which tend to look alike, take note that lead conbections do not necessarily conform to a common
pattern. In particular, notice the lead differences
between types 2N2926 and its equivalent 2N3904, and
remember that the 2N3906 is pnp. To avoid mishaps,
always refer to Fig. 3.3 before applying current to the
transistor.

By
D.BOLLEN

IPEAC
&~&[L@@(1D[

@[RJ1J~(WU~~
constructional details for UNIT "A" were
completed last month. Ul'olIT "A" is, itself, a
T
complete, self-contained computing equipment, and the
HE

method of operation, with practiqll examples, is


described in this article.

PATCHING LEADS
The best plugs to use for patching the computer are
those of " split-pin" construction, as they can quickly be
attached to wires without the' aid of a screwdriver. It
is a help if plugs are obtained in various colours, and
are mated to different coloured wires to allow easy
identification.
,
. For the majority of problems capable of solution by
UNIT "A", certain patching leads may be left in
position on the front panel. For example, coefficient
potentiometers are almost always used with the "0"
end of their reSistance track connected to earth (link
SK3 to SK4 for CPI, CP2, CP3, and CP4, Fig. 2.7). .
Similarly, until such time as integrator mode
switching is brought into use, the integrator sockets
depicted in Fig. 2.9 are joined together by means of a
special three-way patching lead consisting of two short
lengths of wire joined by a plug, with a plug at each end.
Looking at Fig. 2.9, OAI/SK4, SK9, and SKIO are
linked, and repeat for OA2 and OA3. Three more
semi-permanent patching leads are made up to link
each operational amplifier to its companion summer
network. Connect OAI/SK8 to SI/SK5, and do the
same for 0A2/SK8-S2/SK5, and OA3/SK8-S3/SK5.
The rearrangeable patching leads should be of
assorted lengths and colours, the longest to patch from,
say, CP4/SK2 to S3/Il/SKI, diagonally across the
UNIT "A" front panel, and the shortest to link nearly
adjacent sockets.

COMPUTING RESISTORS
If a comprehensive range of 1 per cent high
stability computing resistors was purchased all at
once, to meet every requirement, the cost would
probably exceed 20. There are after all 101 preferred
values in a 1 per cent range covering resistors from
only 10 kilohm to 100 kilohm. Nevertheless, in the
period when the computer operator is learning how to
handle PEAC, and a high degree of accuracy is not
essential, the majority of ordinary problem set-ups can
be catered for by a small number of 1 per cent and
2 per cent plug-in resistorS. A reSistor selection
list, with suggested values of Rr and Rin for standard
op-amp closed-loop gains, is given in Table 4.1. Also,
a component list included in this article sets out
minimum quantities, with tolerances, of computing
resistors.
Computing capacitors will be discussed later, in
connection with integration.
SETTING UP THE VOLTAGE SOURCE
To set up all voltage source outputs, first remove the
dials from VR6 to VRIO (Fig. 2.2), and turn the
potentiometer spindles fully anticlockwise. If the
potentiometers have fiats on their spindles, make up
blanking pieces consisting of small segments of hardwood or plastic, so that control knobs can be conveniently located at a selected position on each spindle.
Connect the positive lead of a sensitive d.c. voltmeter
(O-lV, 20 kilohm/V) to VSl/SKI, and the negative
voltmeter lead to VSl/SK4 (Fig. 2.6), then set slide
switch SI for a positive voltage output. Switch on the
computer power supply and 86.
Carefully rotate VR6 spindle clockwise until a very
small voltage appears, just sufficient to slightly defiect

289

the meter pointer away from zero. Now place a dial


knob on VR6 spindle, without disturbing the potentio-
meter Setting, and align so that the "0" division on the
dial is vertical and opposite the pointer mark on the
surface of the front panel. Tighten the dial knob
grub screw.
Switch off S6 and replace the O-IV meter with the
O-IOV d.c. meter which has been chosen to serve as a
voltage standard for the computer, while retaining the
same meter lead polarity. Rotate VR6 dial until the
"10" division is opposite its pointer, and switch on S6.
Now adjust slider resistor VRI from the back of the
UNIT "A" box, for a precise reading of IOV on the
"standard" meter. Repeat the above procedures for
outputs VS2, VS3, VS4, and VS5, and remember to
adjust only the particular slider (VRI-VR5) which is
associated with the output being set up.
When all the voltage source dials are aligned, return
to VSI and make sure that its output is still + IOV.
Switch off S6, reverse the "standard" voltmeter leads,
and set SI for a negative output. Switch on S6 again
and check the voltmeter reading; if it is not exactly
tOv, go to the back of the UNIT "A" box and trim the
power pack control VR2 (Fig. 3.4), this ensures that
voltage source negative and positive outputs are equal.
SETTING UP THE COEFFICIENT
POTENTIOMETERS
Insert a patching lead to link CPl/SK3 to CPI/SK4
(Fig. 2.7), and do the same for CP2, CP3, and CP4.
Take a long patching lead from VSI/SKI to CPI/SKl.
Remove the dial from VRII (Fig. 2.5) and rotate
spindle fully clockwise. With the negative lead
connected to any earth socket, insert the "standard"
meter positive lead into CPI/SK2 after first setting SI
for a positive output. Adjust VSI dial for a meter
reading of IOV. Rotate CPI spindle carefully anticlockwise until the meter pointer just beings to drop
below the IOV division. Replace CPI dial knob on
VRII spindle, align the "10" division with the pointer,
and tighten the grub screw. Repeat for CP2, CP3, CP4.
With a tOv input to CPI/SKI, and a O-IOV meter
connected to CPI/SK2, it is a simple matter to check
the agreement between dial divisions and voltage
output from the coefficient potentiometer. If there are
serious discrepancies between voltage output and dial
reading this will indicate that the effective electrical
rotation of the potentiometer differs from the 270 degree
dial calibration. Errors can often be minimised by
slight readjustment of the dial knob on its spindle, to
spread the error over the entire scale. Generally

TABLE 4.1
SUGGESTED VALUES OF- COMPUTING RESISTOR FOR
STANDARD CLOSED-LOOP GAINS

Op-amp gain

All resistors 2% unless otherwise


stated

~--G
R 01
02
03
04
05
06
07
OS
0'
10
20
30
40
50
'0
70
SO
'0
100
20'0
300
400
SOO

Rr

Rin

ln

IOOkO
IOOkO
33kO
4OkO
20kO
33kO
13kO
20kO
20kO
10kO
IOOkO
10kO
HkO
33kO
4kO
4OkO
20kO
HkO
13kO
2kO
2kO
10kO
5kO
HkO
4kO
2kO

1%

1%
1%

1%
1%

10kO
20kO
10kO
10kO
10kO
20kO
'lkO
16kO
ISkO
10kO
IOOkO
20kO
10kO
IOOkO
10kO
IOOkO
IOOkO
20kO
91kO
I6kO
ISkO
IOOkO
IOOkO
IOOkO
IOOkO
IOOkO

speaking, the dial setting error should not be worse than


5 per cent at all settings between "I" and "10" dial
divisions. The whole question of computing potentiometer accuracy will be raised later, in connection with
the Master Potentiometer of UNIT "B".
SETTING UP THE OPERATIONAL
AMPLIFIERS
It is usual 't o check operational amplifiers either
before the start of a computation, or at the beginning of
the day, but the computer builder may wish to assure
himself that his amplifiers are all that they should be
when first brought into' service. The zero-setting
procedure given at the end of Part 3 of this series will
have eliminated all but obscure faults. The front
panel balance controls (VRI5, VRI6, and VR17,
Figs. 2.4 and 2.9) are deliberately designed to have a
limited range of adjustment, so that an amplifier fault
will be clearly indicated as an inability to zero-set from
the front panel.
.
To quickly check each amplifier, insert 10 kilohm
feedback resistors into miniature sockets SKI 1 and
SKI2 for OAI, OA2, and OA3 (Fig. 2.9), anq ensure
that the operational amplifiers are already linked to their
summing networks. Insert 10 kilohm input resistors
into
SI/Il/SK3-SK4,
S2/Il/SK3-SK4, and
S3f1I/SK3-SK4 (Fig. 2.8). Patch VSl/SKI to
SI/Il/SKI (Figs. 2.6 and 2.8) and connect the negative
lead of a voltmeter to OAI/SKI3, with the positive lead
going to any convenient earth socket.
Check that OAI output is exactly zero when S6 is off.
If not, zero-set by means of balance control VRlS.
Obtain a positive voltage from VSI by switching on S6
and setting SI and VR6, and monitor VSI output with a
second voltmeter connected to SNl/SK2 red, and an

COMPONENTS
Rf=Rin AND

UNIT "A" COMPUTING


RESISTORS
AND PATCHING LEADS

Rin

+ao--/h'

Plugs
I dozen of each colour: red.
blue. yellow. and white.
front panel . sockets (see
I dozen miniature plugs,
miniature sockets

black.
to fit
text).
to fit

["in

:: INPUT a
BECOMES
OUTPUT - 0

-0

SIGN CHANGE
RI

Resistors
30ff2kO2%
3 off 33kO 2%
3 off4kO I%
3 off SkO I%
3 off9lkO 2%
5 off IOkO 1%
5 off IOkO 2%
3 off 13kO 2%
3 off ISkO 2%
3 off 16kO 2%
3 off lakO 2%
3 off 20kO 2%
3 off 33kO 2%
3 off 4OkO 1%
3 off91kO 2%
5 off lOOk!) 1%
5 off lookO 2%
(All metal oxide or carbon. film.
IW)

Eo =-1

R,

:::!J 1 ~

Rr=R 1=R z

+o~ -(a+b/

+b

ADDITION

R,

RI

:::!J ~

Rf=R,=R Z

+a~ -(a-b)

-b

SUBTRACTION

Rf

Rf=b
Rin

Rin

+0

-A

-ab

+a~-ab
+o~_abC

MULTIPLICATION

+a~_.!..
R

Wire
Stranded core single p.v.c. wires in
assorted coloursl(14Joo76in).

~Il~t IJR:'_~:::;

+ao.....-.JN.-;.'.

Fig. 4.1. (right) These diagrams indicate


how the operational amplifier can be
used to solve various algebraic
equations

RZ

DIVISION

earth socket.- Remember that a positive input voltage


results in a negative operational amplifier output voltage.
Since input and feedback resistors are both 10 kilohm,
the operational amplifier gain will be unity, and both
voltmeters should give precisely the same readings.
Double check by interchanging voltmeters. Now see
that the operational amplifier will faithfully "track"
any input voltage of IOV or less when a temporary
output load of 2 kilohm is connected from OA I jSK 7 to
earth.
The above tests are repeated for OA2 and OA3 by
transferring the patching lead from VSI-SljlljSKl to
S2jIljSKl, and then to S3jIljSKl, and at the same
time reconnecting voltmeters to the appropriate
summer and operational amplifier sockets.
SOFTWARE

; ; : . - K=b

Under the heading of "software" comes all the paper- .


work associated with drawing up a programme for the
computer. The time spent on preparing a programme
for PEAC can vary from a few minutes to several days,
depending on the skill of the programmer and the
nature and complexity of the problem.
The intention is to give a few typical programme
examples as an introduction to using the computer.

1bey will consist of a short written routine, plus


programme layouts. The layouts will be in a duplicated
form, of symbolised diagram and patching circuit, so
that the reader can compare analogue computer
symbols with actual circuits and patching procedures.
A newcomer to analogue computers will best learn
programming techniques by working with PEAC, and
this will also help to increase his knowledge of more
advanced mathematics.
ROLE OF THE OPERATIONAL AMPLIFIER
IN EQUATION SOLVING

Now that the time has come to consider UNIT "A"


as a computer, instead of as' a collection of circuits
handling voltages, it is appropriate to adopt a slightly
different approach.' Voltages will now be replaced by
the letters or numbers of an algebraic equation,
a. b. c. d. x. Y . 2, 3, .. , 5, and so on. Computing
resistors loose their individual identity and are considered only as ratios ;:' ~:, etc., which are also
denoted by equation letters or numbers. The same
applies to coefficient potentiometer settings.
Sign change. In the circuit of Fig. 4.la, an input
voltage classified as term a, reappears at the op-amp
293

-10

+1 0

SYMBOLISED DIAGRAM

.
Flg.4.2

. .IS Unity.
.
output as term - a, when t he -Rc ratIO
Rin

Ja-2b

Programme layouts for - - c -

0 ne

way of looking at this operation, which is common to


all single operational amplifier configurations, is to
assume that a has been multiplied by -], hence
:r

In effect, to mUltiply by - ] is to move a


.
mathematical term from one side of its equation to the
other, so sign change can be used to transpose.
The operational symbol of Fig. 4.la avoids the bother
of inserting resistors and their values when drawing up a
programme layout on paper. The figure inside the
triangle-in this case "1" -merely indicates that the
computing resistor ratio, or alternatively the operational
amplifier gain, is unity.
=

1.

Addition. In Fig. 4. 1b, positive terms a and b


are added to yield an output - (a + b), which can also
be written - a - b. If - (a + b) is applied as an input
to a second unity gain operational amplifier, to give '
two sign changes, it will be converted to a + b. Note
that the figures in the operational symbol triangle show
h Rc
Rr
t at ~ = ], and R2 = 1.
Subtraction. The only difference between Fig. 4.2b and
Fig. 4.2<': is that term b has been made a negative
quantity. The operational amplifier output is therefore -(a - b) or - a + b.
294

Multiplication. In Fig. 4.1 d, Rc and Rin are adjusted


Rc
so that - . = b. Hence, a is multiplied by factor b to
Rin

become an output

- ab. . The letter inside the

operational symbol triangle shows that the Rr

Rin

ratio is b.
Fig. 4.1 e gives an alternative method of achieving
multiplication. A computing potentiometer is connected to the opamp input to multiply a by a factor b.
Rr
adjusted to equal
Therefore, with an input ab, and
R in
c, the result is an output - abc.
Division. When a computing potentiometer is
wired as in Fig. 4.lf, with Rr connected to its slider,
term a will be divided by constant b when Rr = Rin.
Note that Rr is written inside the symbol triangle to
show that b is a divisor.
It can sometimes happen that a feedback resistor is
inadvertently left plugged into an operational amplifier
when it is re-programmed for a division operation, and
this will result in the circuit of Fig. 4.lg. Instead of an
output

-'ba

- (b ~ I)'

the

operational

amplifier

will

yield

COMBINED OPERATIONS

The configurations of Fig. 4.1 have many similarities,


which lead naturally to the combination of several
operations. In fact, it is possible to perform, say, ten
additions or subtractions, three multiplications, and one
division operation all at once using a single operational
amplifier with several inputs and coefficient potentiometers.

+10V~--..-o

COEFFICIENT
POTENTIOMETER

PROBLEM EXAMPLE 1.
SOLVING A SIMPLE EQUATION

UNIT "A" can solve a linear algebraic equation


consisting of more than ten unlike terms, but a simple
example with only four terms will serve as an adequate
practical introduction to programming.
Take
3a - 2b = d
(Eq. 4.1)
c
the letters a, b, and c are regarded as known quantities,
and d is the unknown, but the equation can be transposed to solve for any unknown.
Eq. 4.1 is implemented on the computer as shown in
the Fig. 4.2 patching circuit. Two voltages corresponding to a and -b are taken from the voltage source to
summer SI, where a is multiplied by ~t = 3, and -b
is mUltiplied by

~: =

2.

The machine equation for the problem is,

!ita -l3!b
RI

R2

(Eq.4.2)

and if Rr is made 100 kilohm the equation will take the


form of
100
100
33 a - 50 b
...:..::.--:..-=- = d
(Eq.4.3)
c
Computing resistor values could equally well be

Rr = 10 kilohm, RI = 33 kilohm, and R2 = 5 kilohm,

to yield the same multiplication ratios. Since a 50


kilohm resistor is not included in the short list of
Table 4.1, two 100 kilohm resistors are patched
together in parallel in the patching circuit Fig. 4.2.
Routine. To set up Eq. 4.1 on UNIT "A", first of all
ensure that the voltage source switch S6 is off. Insert
computing resistors into the positions shown in Fig. 4.2
patching circuit, and connect the computing elements
together with patching leads. Set VSl and VS2 dials
to zero, and CPl to "10", corresponding io a divisor of
1. Wire a voltmeter lo OA1/SK13 and zero-set the
operational amplifier by means of VR15. Next
connect a voltmeter to Sl/Il/SK2, and switch on S6.
Set VS1 dial for a trial value of a = 2V. Transfer the
voltmeter from Sl/Il/SK2 to Sl/I3/SK2, and set VS2
dial for a trial value of b = - 2V.
UNIT "A" will now be computing
(3 x 2) - (2 x 2)
1

(Eq.4.4)

with a = 2, b = - 2, c = 1, and therefore d = 2.


When a voltmeter is linked to OA1/SK13 it will be
discovered that the output voltage d is actually - 2V,
due to the operational amplifier sign change. Remedy
by reversing the readout meter leads. If the output
voltage is not exactly -2V, recheck voltages for a and
-b. To check the exact setting of CP1 dial for any
value of c, temporarily remove the patching lead from
CP1/SKI. Patch CPl/SKI to a precise + lOV from a

COMPUTING
RESISTOR

Fig. 4.3. Voltmeter method


potentiometer. settings

of determining coefficient

spare voltage source output, and connect a voltmeter to


SI/I5/SK2. The voltmeter will then indicate the
potentiometer coefficient while taking into account the
loading effect of Rr (see Fig. 4.3). A voltmeter reading
of 475V is equivalent to a coefficient of 0'475. CPl
can now be patched back into the problem set-up. With
a 100 kilohm resistor for Rt, CPl will be dividing by
numbers equal to or less than unity. If Rt is changed to
10 kilohm, the range covered by CP 1 will become
0-10. Therefore, increasing c by a factor of 10 can be
seen quite clearly to be the same as decreasing computing resistor ratios by a factor of 10.
With UNIT "A" now programmed for Eq. 4.1, it is
possible to investigate fully the problem for all reasonable values of a, b, c, and d, and for any unknown
without the need for transposing terms or altering the
problem set-up. For example, to find a when b, c, and
d are known, set band c and adjust a for an operational
amplifier output equal to d. Always monitor an input
voltage with a voltmeter when it is being adjusted.
To see how serious computing errors can occur at
extreme limits, set VSl and VS2 so that terms 3a and
- 2b are virtually equal, and d === O. Also, set CPl to
near zero and observe that d will pass beyond the lOV .
operational amplifier maximum output swing.
PROBLEM EXAMPLE 2.
ANALYSIS OF VOLTAGE DIVIDER CIRCUIT

The voltage divider of Fig. 4.4a is often encountered


in electronic circuits. At first sight, a network consisting of only two resistors might be considered far too
simple to merit investigation by means of a computer,

but it does involve at least six variable quantities


Vh V 2 Ih 12 , Rh and R 2 , and to solve a problem for
any unknown, one of six equations would be required,
based on
(Eq.4.5)
and
(El:}. 4.6)

(a)

RZ

vl =o

I l=O

(e)

(d)

Thus. although it would be ridiculous to use the computer to find one specific answer to one particular
voltage divider problem, the paperwork involved in
solving six equations for several sets of variables could
become surprisingly laborious. What the computer
does in fact allow is the solution to literally any voltage
divider problem under any conditions, without the
need for re-programming.
To solve Eq. 4.5 and Eq. 4.6 simultaneously o n
UNIT "A', the equations are first transposed for
terms V 2 l!-nd 12 which are common to both.

vl

Vz = VI - Rill

+ I z)

(Eq.4.7)

and

(e)

(Eq.4.8)

Il

Rg.4.4. (a) voltage divider circuit; (b) direct simulation of


(a); (c). (d) and (e). t hree variations on (a)

Next, both equations are linked to give a self-enforcing


systems, shown diagrammatically as,
Vz - - + V
-2 = 12
t
R2
I
where the answer to Eq. 4.5. is one of the terms of
Eq. 4.6 (V2 ), and the answer to Eq. 4.6 is one of the
VI -

+10

R1(Il

+10

Fig. 4.5. Programme layouts


for voltage divider analysis
(a) (right) symbolised diagram.
(b) (below) patching circuit

CHECK VOlTAGES ANO POT. SETTINGS SHOWN THUS

298

BU....

+ 12) =

terms of Eq. 4.5 (12)' To see how the problem is


set-up on the computer, refer to Fig. 4.5, and note
the changes of sign involved.
Routine. Switch off S6 and insert all computing
resistors and patching leads, except the link between
OA3 output and OAI input, which carries the voltage
analogue of 12 , Zero-set OAI, OA2, and OA3 in that
order, using a voltmeter applied to each operational
amplifier output socket in turn. Now patch the link
between OA3 output and OA 1 input into circuit. Set
VS I to "0" , and VS2 to " + to" . The voltmeter method
of Fig. 4.3 is employed to set CPI and CP2 both for a
coefficient of 05. Temporarily remove the patching
leads from CPl/SKI and CPZ/SKI, and connect the
"top end~' of the potentiometer tracks to a 10V reference
voltage. Adjust CPl and CP2 for outputs of 5V.
Exactly the same procedure is adopted when it is
necessary to "read off" values for RI and R2, although
approximate readings can be taken from CPI, CP2 dials.
The check voltages in the diagram of Fig. 4.5
correspond to the above voltage source and coefficient
potentiometer settings, and provided that there is
general agreement with Ohm's law, any desired values
can be given to the voltages, currents, and resistances in
Fig. 4.4a. The check voltages could apply to actual
voltage divider quantities of, say, VI = tov, V 2 = 5V,
It = OmA, 12 = lmA (l machine volt = lmA),
RI = 5 kilohm, and R2 = 5 kilohm, where VSI covers
the range O-tomA, VS2 O-tov, CPI O-to kilohm, and
CPZ 0-10 kilohm. Suppose instead that VI had been
assigned the value of I,OOOV, when RI and R2 were both
only 5 ohms. One machine volt would now be equivalent to l00A, and V2 would equal 500V. The ranges
covered by computing potentiometers in the latter case
would then be VSl O-IOOA, VS2 O-I,OOOV, CPI 0-100,
and CP2 0-101 ohms.
Unless informed otherwise, the computer assumes that
VI is an ideal voltage which originates from a source of
infinitely small resistance. Hence, if VI = 0, this
corresponds to a short-circuit, and gives the variation
of Fig. 4.4c. Alternatively, if 12 is made equal to
nought, the voltage divider circuit is transformed into a
load resistor R2 in series with a source resistor R"
given by Fig. 4.4d.
One further variation will serve to show the flexibility
of the programme. In Fig. 4.4e the resistance network
RI and R2 is made to couple two sources of voltage VI
and - V 2 and this occurs when 11 is made larger than
11 + 12 , or in other words, when 12 swings negative.
The layout of Problem Example 2 is an instance of
indirect simulation, where the computer solves
equations and imitates the behaviour of the simulated
circuit. In this indirect "model" of a voltage divider,
relationships between governing equations and actual
circuit parameters are made obvious, and the abstractions of mathematics are brought to life as tangible
voltmeter and dial readings.
Another way of simulating the Fig. 4.4a circuit is by a
direct "model", shown in Fig. 4.4b, which employs
coefficient potentiometers for Rh RL, and RL, voltmeters lor VI and V 2 , and current meters for 11 and 12 ,
Although feasible, the direct model is less elegant, is
not so adaptable to extreme cases, and is subject to
errors which do not occur when the voltage divider is
simulated indirectly.

Next month: Using UNIT "A" to solve a second


order differential equation. Indire'et simulation
of LC circuits, spring pendulums, and servomechanisms by means of integrators.

REGULATED POWER SUPPLY


continued from page 284

No load
urrfttt

lA
Lood

1/01.15
0

10
288
58

o.t7
0145
012

142

099
286
578

0165
0143
0119

\1-4

70
14.\
"3
69

068
05
036
06\
041
023

E
0.47
038
0 28
040
030
017

F
2 .
6
12
2
6
12

IIA
1 _
2
65
40
\8
36
18
505

..It2.7
2\

1-4
VI

13
032

+
550.

Fig. '0. Test measurements for fault-finding


,

CALIBRATION OF OUTPUT VOLTAGE


After checking the voltmeter accuracy against an
A VO or similar instrument known to have good
accuracy itself, the Regulated Volts dial should be
adjusted as follows.
Switch S2 to "Regulated" and S3 to "Volts" and turn
VRl until IV is obtained at the output (best seen on the
AVO). Loosen the knob and rotate to indicate I V on
the calibrated dial. Lock the pointer knob grub screw
while indicating the correct IV. Rotate VRl until the
dial indicates 12V output. Now adjust VR2 until 12V
output (measured) is obtained.
VOLTAGE CHART
Fig. 10 gives typical voltages at six points in the d.c.
amplifier circuit for three different output voltages.
Reference to these voltages and to the currents of the
super-alpha pair TR2, TR3 should assist in any faultfinding.

INDEX
An index for volume three (January 1967
to December 1967) is now available price
Is 6d i ncl usive of postage.

BINDERS
Easi-bi nders are available price 14s 6d
inclusive of postage. State whether "Vol.
I ","Vol. 2","Vol. 3" or "Vol. 4" is required.

Orders for Binders and Indexes should be


addressed to the Binding Department.
Orders for copies of the Index only should be
add ressed to the Post Sales Department. George
Newnes Ltd., Tower House, Southampton
Street, London, W.C.2.

detailed explanation concerning the operation of


UNIT "A" is continued in this month's article,
T
with further practical examples.
HE

We resume by considering the use of the operational


amplifier as an integrator.
An operational amplifier will be handling time as well
as voltage when acting as an integrator, so some means
must be found of inserting intervals of time onto the
computer. One method is to employ external oscillators to provide known functions of time in terms of
frequency. An input to an integrator might consist
of a steady d.c. voltage which is switched on for a time t
(step function or square wave), or alternatively, a sinusoidal voltage of frequency f and period 1If.
If a graph is drawn of the resulting integrator output
function, and this is the form that answers to problems
involving change or motion will usually take, the X axis
of the graph will be calibrated in intervals of time, with
voltage on the Y axis. It follows that an oscilloscope,
which also uses time on the X axis and voltage on the
Y axis, can provide a convenient .form of output display, especially when an integrator is operating at high
speed.
.
The operational amplifier is converted to an integrator when a capacitor Cr is inserted, in place of a
resistor, in the feedback path; see Fig. 5.1. When an
input voltage - Ein is applied to the integrator by
means of a simple switch S for a time t, the output Eo
wiJI take the form of an increasing ramp voltage proportional to t with slope
1
-Ein R in Cf
Note that the operational amplifier will continue to
invert an input voltage even when used as an integrator.

THE INTEGRATOR IN EQUATION


SOLVING

The electronic analogue computer does provide a


powerful technique for obtaining rapid solutions to
problems involving calculus, which cannot be equalled
either by numerical methods or by a digital computer.
If differentiation and integration are regarded as
straightforward mathematical operations, it will be
found that the terms of, say, a second order differential
equation can be manipulated on the computer in much
the same way as the terms of a "steady state" algebraic
equation.
For example, when an equation term y is differentiated against time its derivative dyldt is obtained, and
a second differentiation yields the second derivative
d 2Yldt 2 The reverse process is where integration of the
second derivative d 2y/dt 2 produoes the first derivative
dy/dt, and another integration ~ives y as the result.
Fig. 5.2 shows how a simple integrator can handle
equation terms. Combined operations are made
possible by cascading integrators, while using coefficient
potentiometers and computing component ratios for
summation, multiplication, and division (Fig. 4.1).
The process of differentiation, although feasible if
care is taken, is generally avoided on analogue computers because it gives rise to unstable operational
amplifier configurations, but this imposes only a slight
limitation since integration can be employed-in the
majority of cases-in place of differentiation.
INTEGRATOR ACCURACY

The transfer accuracy of an operational amplifier,


when it is used as an integrator, wiII be theoretically
limited by its finite value of open-loop gain. However,

&~ill[L@@(QJ[g

@ U\':'AJ ~ l1D LrT~~


PEAC
By
D.BOLLEN

INPUT
FUNCTION

TABLE 5.1
Ct

RiB

I,.F

IOOkO
IOkO

28sec

800ms

OI,.F

IOOkO
IOkO

280ms

IOOkO

28ms

o-OI,.F

;:ItT.
I

Cr

-Eift

-~>:..Iw.....~
S

Rift

SOm.

EO--iftR: 'C
1ft f

VOLTS/SEC

Flg.S.1. The operational amp/lf/er as an integrator

Maximum value of t for an error of 1%

the situation is much more complicated than with, for


.example, a summing amplifier (Fig. 3.8) since the
amplifier error can no longer be defined in terms of the
simple relationship between closed-loop and open-loop
gains.
As a guiding principle, integrating amplifiers may
have very large values of closed-loop gain provided
that the time t of an input function remains small.
Closed-loop integrator gains of 1,000 or more are not
uncommon in transistor computers, since low voltages
and low impedances discourage the use of computing
resistors of more than 1()() kilohm, and capacitors of
more than IpF are too bulky. Table 5.1 is calculated
for UNIT "A" amplifiers, and sets out the maximum
allowable interval t for selected values,of Cr and Rin,
where the amplifier transfer error must not exceed
one per cent.
Errors due to unwanted drift voltages also become
significant when t is long and Cr is small. The greatest
care must be exercised when zero-setting integrators to
eliI!linate offset voltages, for good accuracy at long
time intervals. Also, ' the computer should not be
subjected to fluctuations of ambient temperature when
computations cover several hours of integrator use.
COMPUTING CAPACITORS
The computing capacitors used for PEAC will
normally lie within the range O'OI-lpF, and the three
values most commonly employed are O'OlpF, O'lpF,
and IpF. Polystyrene is the preferred capacitoI:
dielectric, for high insulation resistance, but polyester
makes an acceptable second best. Mica, paper, and
ceramic capacitors should be avoided.
Small value polystyrene capacitors of I per cent
and 2 per cent tolerance are easily obtained, but
O'lpF ;md IpF precision components are rare and
expensive. To get around this difficulty, the bridge
circuit of Fig. 5.3 was devised to allow computing
capacitors to be made up from specially selected low
cost 20 per cent capacitors.
The circuit of Fig. 5.3 can be constructed in breadboard form on Verob6Elfd or s.r.b.p., with miniature
sockets to take Cx and RI. If an audio signal generator is not available to supply the bridge wittt about
lOV r.m.s. at 1kHz, a signal could be obtained from a
transistor multivibrator powered by the 25V computer
power supply. Headphones serve to detect the null
point when the bridge is in balance, and should have
an impedance of about 2 kilohms.
The method of making up a computing capacitor of,
say, IpF is as follows. A capacitor panel of plain or
perforated s.r.b.p. is fitted with small turret tags as in
Fig. 5.4. A 20 per cent capacitor of about O'68pF is
wired into position on the capacitor panel before it is
plugged into the bridge Cx sockets, and a I kilohm

BASIC OPERATIONS IN CIRCUIT FORM

SYMBOL

RfCf =0

yo--Wv-......-t -A

-ol~
0

OR

dt

y~
-01;
0

dt

_'_=1
Rin C,
-A

-y

:~..,y
1

INTEGRATION
Fig. 5.2.

integrator

The handling of equation terms by a simple


1kHz
r-------.....
- - - o 10Y
R.I4.S.
Cl

RI

Cx

10,OOOpF

O"jlF IOktl tl'.


IjlF

lo,o

IkIltl,.

r-------l

I
Cx- 1000 X RI

RI

IL _______
Cx
....I

Fig. 5.3.

capacitors

Bridge circuit used for making up computing

"

20 S.W.G. TINNED
COPPER WIRE

Fig. 5.4. Computing capacitor plug-In pGnel

363

resistor is inserted for RI. Assorted polystyrene or


good quality polyester capacitors of lower value are
then temporarily connected across the capacitor panel
to increase Cx by small increments, while listening on
the headphones for a drop in the level of the 1kHz
tone as Cx approaches 1pF.
A typical computing capacitor might finally consist of
a parallel combination of the following values, 0'68pF,
0'22pF, 0'02pF, and 0OO5pF.
If the required value of Cx is exceeded, the note in
the headphones will increase in volume when the null
point is passed. Allow capacitors to cool off after
soldering, and before making a measurement, as heat
can cause a temporary or perm'a nent change in capacitance. With the Fig. 5.3 bridge circuit it is possible
to detect increments of less than O'OlpF in a nominal
1pF capacitor. .
DIFFERENTIAL ANALYSIS WITH
UNIT "A"
A second order linear differential equation with
constant coefficients has become firmly established as
the "classic" introduction to differential analysis on
the analogue computer.
The equation describes an oscillatory system with
variable damping which can 'be used to simulate
indirectly many physical systems, such as the spring
pendulum, a tuned LC circuit, or a servomechanism.
Also, the equation is easy to set up on the computer,
and does not necessarily demand the use of integrator
mode switching,
In general form the equation is,
d 2y
dy
a dt 2 + b dt + cy = /(t)
(Eq. 5.1)

where a, b, and c are the constant coefficients, y is


unknown, and f(t) represents some function of time.
Equation 5.1 can be rewritten to suit a particular
system by substituting appropriate terms.

Furthermore, as the computer will allow operation at


almost any fraction or multiple of real time, a spring
. pendulum and a tuned LC circuit can be simulated
simultaneously, and interesting electro-mechanical
parallels can be seen to exist between the properties of
inductance and mass, resistance and friction, and capacitance and elasticity.
The only real difference between the analogous
behaviour of a weight on a spring, a servo shaft, and a
tuned LC circuit is that the LC combination will normally resonate at a much higher frequency.

PROBLEM EXAMPLE 3.

TUNED CIRCUIT ANALYSIS


UNiT " A" will simulate any series tuned circuit by
solving Equation 5.2, and will give answers in the form
of a.c. meter readings or oscillograms. Tuned circuits
resonating in the MHz region are catered for by slowing
down the problem to some convenient decadal fraction
of real time, so that a simulated circuit on the computer
which is, for example, resonating at 300Hz, will serve
as a model for a real circuit resonating at 30MHz, with
suitable rescaling ofL, C, and t.
To initially determine the relative values of L, C, R,
voltage V, and current I, without too much paperwork,
it is helpful to start with a representative tuned circuit
which allows computer operation in real time, at
frequencies convenient for display by an a.c. voltmeter
or an oscilloscope. 50Hz is a good frequency to
employ as a datum because it can be readily obtained
from the mains supply, and rounded values of L = IH
and C = lOpF will also offer resonance at 50Hz.
Taking the circuit of Fig. 5.6a as a starting point,
from the knowledge that a series tuned circuit will
exhibit an impedance equal to R at resonance, the
r.m.s. current flow at 50Hz will be El/ R, or 20mA when
El = 2V r.m.s. and R = 100 ohms.
It is necessary to rearrange the basic equation ,
Equation 5.2, for the computer by dividing through by
L, and solving for the second derivative.

Spring pendulum
d 2y
m dt 2

(Eq. 5.5)

dy

+ p dt + ky

f(t)

(Eq.5.2)
where m is the mass of a weight suspended on a spring
of constant k, which is damped by friction p . The
weight is displaced by an amount y when subjected to a
force dependent onf(t).
Tuned Le circuit
dQ ' I
d 2Q
L(ft2 + R dt +CQ = f(t)

(Eq. 5.3)

where L is an inductance tuned by a capacitance C, and


damped by a series resistance R. Q is the charge in
coulombs on C at any instant of time. The current
flowing in the tuned circuit is given by dQ/dt, and f (t)
represents an input function.
Servomechanism
d 200
dt 2

dO o
21:;(1)

dt +

(1)

00

0i

(Eq. 5.4)

where 0 0 is the angular displacement of the output


shaft, I:; the damping factor, (1) the angular velocity, and
0 1 the angular displacement of the input shaft.
The obvious similarity between the above equations is
emphasised when, in Fig. 5.5, it is seen that they all have
virtually the same problem layout on the computer.
364

Substituting known values from Fig. 5.6a,


d 2Q

df2

100R dQ
IH dt

/(t)

- IH x lO- SC Q + IH

(Eq.5.6)
f(l) in the present case represents a sine wave input of

2V r.m.s. In other circumstances the input function


could be a square wave of amplitude Ein and period 21.
Equation 5.6 is solved on the computer by successive
integration. Looking at the symbolised diagram of
Fig. 5.6b, it can be seen that there are two closed-loops,
one linking the output of OAI via CPI to OAI/Input I,
and the other passing through OAI, OA2, and OA3,
via CP2, and thence back to OAl/Input 3. The
coefficient of CPI will be multiplied by the gain factor
associated with OAI /Input 1. "CP2 coefficient is
multiplied by the product of gains OA1/Input 3, OA2,
and OA3, i.e. 1,000 x 100 x 1 = 100,000.
d 2 Q/dt 2 , obtained from the sum of the voltages
present at the inputs of OAI , is initially assumed to be
present. After one integration OAI provides an
output dQ/dt, and from this all the terms on the right
hand side of Equation 5.6 are assembled. So, dQ/dt is
multiplied by R/L = lOO, using CPI set for a coefficient
of 0'1, and is taken back to OA 1/Input 1 where it is then
added to/(t)/L = 2V r.m.s.

Moving in the other direction on the symbolised


diagram of Fig. 5.6b, dQ/dt is integrated byOA2 to
obtain +Q. Inverting amplifier OA3 changes the
sign of Q before passing it on for mUltiplication by
l/LC = 100,000 (CP2 coefficient of I). - (I/LC)Q is
then added, at OA 1/Input 3, to
R dQ

f(t)

- Zdt + Y

and the sum of all OAl input voltages yields the required d 2 Q/dt 2 Because there are two closed-loops
, in the computer set-up the equation will be selfenforcing.
Routine. Switch on UNIT "A" power supply and
allow a warm-up time of at least 15 minutes. Ensure
that the three operational amplifiers are disconnected
from their summer networks, and have no feedback
components. Apply JOV d.c. voltmeter leads . to
OAl /SKJ3 and an earth socket, and zero-set OA 1 for

an output voltage of less than 1V from the back of the


UNIT "A" box, by means of VRI (Fig. 3.7). Repeat
for OA2 and OA3.
.
Set up the problem according to the patching circuit
of Fig. 5.6b, but omit the feedback capacitors and the
patching link between OA3/SKI3 and CP2/SKl.
Set CPI dial to approximately "1". Connect the voltmeter to miniature socket OA1/SK6 (Fig. 2.9) and
zero-set OAl again, but this time using the front
panel control VRI5.
Next, zero-set OA2 using YR16, and OA3 using
YR17. Insert O' lpF computing capacitors into OAl /
SKII and SKI2, and OA2/SKII and SK12, and make
good the link between' OA3 output and CP2. Set CP2
for a dial reading of "10". Apply the voltmeter to
OA2/SK7 and zero-set the complete assembly of
amplifiers by adjustment of VRI5(OAl) only.
The problem layout will now be ready for dynamic
checks and should not need to be re-zeroed for several
hours if UNIT "A" is being operated in stable ambient
temperature conditions.

b dy

-;; Tt'

+'(t) _ _....;::=-I
a

COMPUTER LAYOUT FOR SOLUTION OF SECOND ORDER EQUATION

_l!. dy
m dt
, (t) _ _ _=-I

+-;;;

-mY
DASHPOT

LAYOUT TO SIMULATE THE MOTION OF A SPRING PENDULUM

R dO

-Ldt
f(t) _ _ _~

+T

-...Lo
Le

LAYOUT TO SIMULATE A TUNED CIRCUIT

- ...16.+ ...9j

LAYOUT TO SIMULATE OVERSHOOT IN A SERVO SYSTEM


Fig. 5.5. A second order differential equation applied to physical systems

367

TABLE 5.1
SHOWING HOW COMPUTER OPERATING FREQUENCIES ARE
RELATED TO CPl SETTING AND AMPLIFIER CLOSEDLooP GAINS

Resonant
Frequency

Typical
Values

CP2
Coefficient

re

Amplifier Gains
OAI
OAl
OA)
Input)

O'()SHz
to

I,OOOH

10,OOO,.F

01

0:1

10

10

01

100H

r;OOO,.F

10

10

10

10

01

10H

100,.F

001

103

1.000

100

10

IH

10p.F

10

105

1,000

100

10

500Hz

100mH

I,.F

10

10 7

1,000

1,000

10

1kHz

100mH O2p.F

10

5x 107

1,000

1,000

SO

~5Hz

5Hz
to
50Hz

I C'OpF

TUNED
CIRCUIT
RESONATES '
AT 50Hz

ZVtlll...

50Hz

R-1001l.

REPRESENTAtiVE TUNED CIRCUIT USED FOR SETTING UP


THE PROBLEM LAYOUT IN "REAL TIME"
B

SYMBOLIZED DIAGRAM

DYNAMIC CHECK VOLTAGES,


~
AND POT SETTlN~S SHOWH THUS ~

r-------,
1

~------~~
1
1

.1 CPt

_
...

Fig. 5.6. Programme layouts


for tuned circuit analysis

1
1

,
1

~----- -.J

r---------,I
,

R,10kQ

,
1

>-~~~I~~~
N
rL

~,=~~3

,I
I

I.~

368

1
I

_____.J1

'SUMMER lAND

SIMPLIFIED PATCHING CIRCUIT

, II

_____ .J

Fig. 5.7. Response of a simulated tuned circuit

Apply a 2Y r.m.s. 50Hz signal to OA1/Input 2, and


monitor by means of a reliable 10Y a.c. meter of not
less than 1 kilohm/volt sensitivity. The input function
should preferably come from a Iow impedance source
to avoid serious loading errors when the voltmeter is
removed. Next, connect the a.c. voltmeter to the output. of OAl and adjust CPl so that OAl input and
output voltages are exactly equal. CPl could alternatively be set by the reference voltage and d.c. voltmeter
method mentioned earlier, for a coefficient of 01. If
the CP2 setting is altered it will be discovered that the
simulated circuit goes off resonance, and can be tuned
by CP2 between approximately :;Hz and 50Hz.
UNIT "A" will now be ready for analysis of the
Fig. 5.6a tuned circuit, and will also cover a useful
range of other values for L, C, and R in real time.
When handling sinusoidal or step functions, an
amplifier will still have a maximum output voltage
swing of 10Y, but this will be the peak voltage value.
To check for overloading with an a.c. meter, ensure
that amplifier output voltages do not exceed 707Y
r.m.s. for a sine wave function, and 5V mean for an
equal mark-space square wave. '
RESCALING PROBLEM EXAMPLE 3.
To rescale the problem for larger or smaller values
of Land C, beyond the coverage of CP2, and by
abandoning real time operation, note that a tenfold
increase in tuned circuit frequency corresponds to a
hundredfold increase in l/Le. For most applications,
where the series resistance R will lie between zero and
just beyond critical damping (R> 2y'[L/C)), the scaling
of R/L can stay as it is for all reasonable values of
Land C, but should anyway only be changed byadjustment of the gain factor at OA1/Input 1. Similarly, the
j(t)/L gain of 100 at OA1/Jnput 2 can remain fixed.

It is not necessary to use inconveniently large or


small input functions when rescaling for new voltages
and currents. 2V r.m.s. could equally well represent an
input fun::tion of, say, O2V r.m.s., and from Ohm's
Law the current I will automatically become 2mA,
instead of the former ' 20mA, even though it is still
represented by 2 computer volts.
If it is desired to extend the computer operating time,
by adjustment of integrator and inverting amplifier
closed-loop gains, refer to Table 5.2, while rememberihg
that integrator closed-loop gains are calculated on the
basis of l/RlnCr where R is in ohms and C is in farads . .
For reasons of reduced accuracy, it is not advisable
to 'use computer operating frequencies above 1kHz or
below 0'05Hz in connection with Problem Example 3.
It should be mentioned that although frequencies in tbe
region of O'05Hz are too Iow for display on an a.c.
coupled oscilloscope, the behaviour of a system can be
demonstrated in slow motion by the oscillating movement of a d.c. voltmeter pointer (centre-zero).
Some typical oscillograms are given in Fig. 5.7 to
show the response of a simulated tuned circuit. If the
computer oscilloscope is provided with a good graticule,
and has a linear response, amplitude and time measurements which are accurate to within approximately
5 per cent may be obtained straight from the trace.
The behaviour of a real tuned circuit can be evaluated
by comparison with a simulated circuit. A tracing is
made of the real circuit oscilloscope display, and is then
superimposed on the readout given by the simulated
circuit. The computer is adjusted so that time scales
are related by a known factor, and tracing and readout
display are identical, then quantitative measurements
are taken from the computer voltages and dial settings.
Next month: The construction and operation
of UNIT "B"
371

'~I~@@(W~

~(!lJlJ~~
PEAC
By
D.BOLLEN

THE

PEAC basic equipment has now been dealt with

I and this month we commence a detailed description of

the chief ,ancillary unit. Subsequent articles will cov:er


the remaining two ancillary units.
Perhaps it should be repeated at this stage that the
three ancillary units are purely optional add-on items.
The additional facilities they each provide, are indicated
in the PEAC Specificatiori (January 1968, page 38).
PEAC UNIT "B"

UNIT "B" reinfon;es the facilities of UNIT" A", but


does not introduce new computing circuit elements.
A masterpotentiometer and a suitably scaled readout
meter improve the accuracy and ease of handling of
UNIT "A", while the integrator mode switching circuit
opens up further possibilities in the solution of Calculus
problems. '
UNIT "B" FRONT PANEL
It may not be necessary to use hardboard for the

front panel if a thick grade of plastic laminate is used,


since the wooden surround in the box front gives plenty
of support.
Prepare a 17iin x 8iin white laminate panel and
establish hole centres with a sharp spike, from the
drawings Fig. 6.1 and Fig. 6.2. Next, drill only the
holes for all sockets, S7, S8; the meter mounting studs,
and cut out a hole for the meter body with a fretsaw.
Beginning with the master pot,entiometer dial, draw a
300 degree arc of radius 21iin with a pencil compass
(refer to Fig. 6.2). Divide the arc into 3-degree divisions with protractor and pencil. The accuracy of the
master potentiometer will benefit from careful preparation of the dial. Draw in the dial arc and divisions
with Indian ink.
Rub-on transfers are suitable for the dials of VR 18
and VRI9, and will save time, but make sure that the
transfer gives main divisions spaced at 30-degree
intervals, for a 1-10 calibration.
When dials are complete, drill holes to take the

spindles of VRI8-VR20, S9 and SIO. Draw in all ink


lines, add transfer numerals, and varnish.
BOX CONSTRUCTION

Commence building the UNIT "B" box by cutting


out two side panels from hardboard; they are shown
in Fig 6.3. Fix lin square softwood lengths A, B, C,
and D to the inside of the side panels. Join the side
panels together by means of horizontal lengths E, and
F, using countersunk woodscrews and glue. Square
up with the assembly placed on a flat surface.
Cover the box framework with hardboard top,
bottom, and front strip panels, and, when firm, reduce
overlapping edges with a rasp and sandpaper. Finish
off the box with a layer of white plastic laminate, and
paint exposed hardboard edges to match the UNIT
''cB'' box.
MASTER POTENTIOMETER AND NULL
METER

A d.c. voltmeter connected to the slider of a computing potentiometer will impose a small load, and when
the voltmeter is removed the measured coefficient will
increase slightly, to the extent of about It per cent in
the case of a IOV 20,000 ohms/volt meter, and a 10
kilohm potentiometer set with its slider near mid-track.
One way of avoiding the error is to leave the voltmeter
connected to the potentiometer after a coefficient
reading has been taken, but this is seldom convenient.
Ideally, the instrument used to measure coefficients
or computer voltages should impose no load at all,
and this condition can be satisfied fairly easily by
employing
an
accurately
calibrated
master
potentiometer.
In Fig 6.4, a permanent load is placed on the coefficient potentiometer CP by the computing resistor
Rln, thus causing a significant dial setting error. To
find the true coefficient of CP, both potentiometers are
supplied with a reference voltage of + 10V, so that
potentiometer coefficients of 0-1 will be multiplied
by 10 to conform to a 0-10 dial. calibration. When
429

COMPONENTS

UNIT "B" FRONT PANEL

NOTE: AI/ front PQnel controls are numbered


consecutively, fol/owing on from UNIT" A", but internal
sub-assemblies have individual component numbering.
Potentiometers
VRI8 100kn carbon linear
VRI9 100kn carbon linear
VR20
25kn
wirewound,
3in,
25W
instrument potentiometer. (G. W. Smith
& Co. (Radio) Ltd., 3 Lisle Street,
London, W.C.2)

RESET TIME

50

Switches
57
Miniature push button, push to make,
one pole
58 Toggle, single pole changeover
59
4 pole, 3 way rotary
510 2 pole, 6 way rotary
Sockets
5 red, 3 -black, 5 blue, 4 yellow, 4 white, and
2 green

100-O-100JLA,

UNIT' "B" MASTER POTENTIOMETER

Resistors
RI 200n
R3 47n
R5
R2 820n
R4 47n
R6
All 5%, tW carbon film or metal oxide

820n
2000.

Pre-set potentiometers
VRI-VR4 loon wirewound (4 off)
panel mounting type
Miscellaneous
16 s.w.g. aluminium sheet 6in x 4in. Tag strip with
three tags.
UNIT "B" READOUT METER

Resistors
RI 82kn 10%
R2 22kn 10%

R3
R4

75kn 5%
12kn 10%

All tw, carbon


composition

Pre-set potentiometers
VRI 22kn
All miniature horizontal
VR2 10kn
mounting skeleton conVR3 2-2kn
struction
VR4 Ikn
Meter protection diodes
DI, D2 OC71 or similar "inverted" germanium
transistor (2 off)
Miscellaneous
S.R.B.P. panel 2}in i< 2in.
-UNIT "B" INTEGRATOR MODE SWITCH

Resistors
RI
10kn
R2
10kn
R3
Ikn

430

R4
R5
R6

Hkn
27kn
Ikn

R7
R8
R9

27kn
47kn
10kn

60

___3~omV
19 70 80
~20
30'

'y

-I.

~2
_

TRIGGER

Fig. 6.1.

1:--";

59

o
RED

.";

58

INTEGRATOR SWITCHING

WHITE

10-100

, 1HOLO~REPEAT

t---9
7,; ';'

'----

100

I-30'

~S7

-I"i~

',:

BLACK

90

~C:7

--elk

I..

to

3~

SYNC

r.

internal

Miscellaneous
Plastic laminate (thick) for front panel, I off,
17tin X 8tin. Rub-on dial transfers and
letters, black (Radiospares)

50

~O

Knobs
One Bulgin K403 , 2jin knob with 3in 'skirt.
Three Radiospares type PK I tin knobs with
pointers
Meter
MI "Sew" MR85P,
resistance 1,000 n

COMPUTE TIME

60

YELLOW

5'1;

1':t '.

BLUE

GREEN

UNIT "B" front panel, integrator switching section

RIO 10kn
RI2 J.3kn
RI4 Ikn
RII Ikn
RI3 10kn
RI5 560kn
All 10%, tW carbon composition
Pre-set Potentiometers
VRI 10kn
VR2 5kn
Both vertical mounting

Capacitors
Cl I,OOOJLF elec. 15V
C2 I fLF polyester 250V d.c.
C3 HfLF polyester 250V d.e.
C4 14fLF elec. 25V
CS 01 fLF polyester 250V d.c.
(The values of C3, C4, C7, and

C6 IfLF polyester 250V d.c.


C7 14fLF polyester 250V d.c.
C8 14JLF elec. 25V working.
C9 0068JLF polyester 250V d.c.
C8 are approximate-see text)

Transistors
TRI-TR6 ACY28 or AC126.
Diodes
D I, D2 OA95 (2 off)
'D3-D14 IB30(Radiospares) (120ff)
Reed Coils
RLA, RLB Miniature triple 12V Osmor type MTI2V (2 off)
Reed Switches
RLAI - RLA3{Hamlin MRG2, 20-40AT (R.T.S. Ltd.,
RLBI-RLB3 P.O. Box 11, Gloucester St. Cambridge) (6 off.)
Miscellaneous
S.R.B.P. panels:
Small tu rret tags.

off 6tin x 2tin; I off 3in

~<

2in.

METER
MOUNTING HOLES

3//;' I

3'86"
PEAC

1
1
1
1-

+1

3'46-

1
I
1
I
1
I

10

....

MASTER POTENTIOMETER
t - - - - - - - - - - - - - - - - - - - - - 111t;'-------------------~.1
Fig. 6.2, UNIT "S" front panel. readout meter and master potentlometer

I,; x 1,~')
Z OFF 51,;
WOOD (ALL

Z OFF 61,;'
Z OFF n"
Z OFF 81t;
2 OFF 163'4"
Z OFF 173'4"

2 OFF

lZ1t~' X 71,~' (SIDES)

1 OFF 18"X 6" (TOP)


1 OFF IS"X 10V; (BOTTOM)
I OFF IS"X 21/; (FRONT STRIP)

Fig.

6.3

UNIT "S" box


constructional details
4 31

AT NULL 1=2 AND K IS THE


TRUE COEFFICIENT OF CP

+10V

the voltage at the slider of CP is identical to the slider


voltage of MP, no current flows through the null
meter, and the true coefficient of CP can be read straight
off the dial of MP.
Since no current flows at null point, no load 'is
imposed, and the input resistance of the measuring
circuit is virtually infinite. Meter protection diodes
are included to preserve good meter sensitivity without
allowing damaging currents to flow through the meter
when the circuit is off ba1ance.

+10V O+w

o-10V o-w
CP
COEFFICIENT
. POTENTIOMETER

"

METER PROTECTION
DIODES

Rin

READOUT METER AND MASTER


POTENTIOMETER CIRCUITS
One meter movement serves fot null indication and
voltage measurement. Considering first the readout
meter circuit Fig 6.5a, miniature pre-set resistors
VRl-VR4 will permit calibration of each meter range
to an external voltage standard, and also help to
eliminate discrepancies between ranges .
The way in which meter protection diodes Dl and
D2 are wired may "be unfamiliar to the reader, so some
explc,mation is called for. If a transistor is operated
"inverted", that is with collector-emitter polarities
reversed, it will exhibit a very low "on" resistance
when the base is near emitter potential. With base
connected straight to emitter, the transistor therefore
becomes a diode with lower than normal forward
resistance, and yet will still offer a high resistance

VIRTUAL
:
EARTH .J._

Fig. 6.4. Master


. coeffl dents

potentiometer circuit

for measuring

Fig. 6.5 (below). Circuit diagram of readout meter and


"master potefftiometer

II "iNVERTED
TRANslSTOiiull
"
DIODE

I~
1i
I~ T

VOLTS
INPUT

L_____

SKl

__--'

I
I

I
I

II

10

METER EARTH

,-------,
Rl

'lR20
2SkQ

SKl

SK2 SK3

SK4

...--'\Nv-+--- +lZ'SV
SK6
+l,-O,V--.-T'o.

SKS

I
I

VOLTAGE
DIVIDER

I
I
I
I

SK7

+1V

t-----~--~~------~--~---~---EARTH

S10A

I
I
I
I
I
I
I
I
I

NULL 10V

VR122kO
METER RESISTOR
I
PANEL

VRZ10kO

VR3 z.2kO

UNIT "8" SOCKET


IDENTIFICATION
The following abbreviations
will be used in programming
instructions
for
PEAC.
RM Readout Meter
MP Master Potentiometer
IS Integrator Switch

SK8

-lV....,--~ro....

R6

L___2~~-l

L ____ ~~~ ____ -.J


READOUT METER
(a)
432

I
I

MASTER POTENTlOMETER

(h)

-12SV

reverse characteristic. The arrangement eliminates theneed for a meter series resistor while still giving adequate protection.
,
'
In Fig. 6.5b, VR20 is a 3in instrument potentiometer
of good linearity. The voltage divider network,
composed of RI-R6 and VRI-VR4, taps off four
standard voltages from'- the computer power supply,
so that the master potentiometer will measure inputs
of 0 to + IV,Oto -IV,Oto + 10V,andOto -IOV
on its 0-10 scale. The accuracy of the master potentiometer, bearing in mind the 14in scale length, approaches
that of a laboratory voltmeter.
FRONT PANEL AND MASTER
POTENTIOMETER ASSEMBLY

Mount all sockets, potentiometers VRI8-VR20,


switches S7, SIO, and meter, on the UNIT "B" front
panel. Make up an aluminium bracket from the
measurements given in Fig. 6.6, and glue it to the front
panel, along with the small tag strip, in the position
shown in Fig. 6.7. A hot solderipg iron applied to the
aluminium bracket will solidify the epoxy resin glue in a
matter of minutes, sufficient to hold the bracket in
place until the joint sets hard.
Rest the front panel inside-out on , the UNIT "B"
box front, to protect panel markings during assembly.
Mount pre-set voltage divider potentiometers VRI-VR4
to' the afuminium bracket, and then proceed with the

FI,. 6-6.

Mountln, bracket for pre-set potent/ometers

MASTER POTENTlOMETER

R6

GLUE -8RACKET AND TAG


TO FRDNT PANEL WITH
EPOXY RESIN

~
Rear .,/ew of UNIT "S" front pane~. s~ow'ni
ma e(, potentlometer and readout meter wlrln, .
435

3"

1:.-

r~

.,. ffi).

DRILL 4" DIA. TO


TAKE METER
TERMINAL SCREWS

When taking the photograph, ensure that the camera


lens .is in line with the centre of the scale card, and that
the mm plane is parallel to the surface of the oversize
drawing, to prevent optical distortion.
Another tip, use white Formica for the drawing,
as then mistakes in ink can be erased without leaving
unsightly grey areas.
To remove the existing scale from the "meter, prise
off th~ transparent meter front, and carefully remove
the scale card by undoing the! two holding screws.
Measurements can then be taken for preparing the
oversize drawing.
To fit the new scale, cut out the photographic reproduction and paste it over the old scale, with edges and
mounting holes of both scales properly registered.

~I
I

c!5

L
,0 ,0 '-0 ,0

Z"

MATERIAL:
S.R.B.P. BOARD

SETTING UP MASTER POTENTIOMETER


AND READOUT METER

With red, green, and blue p.v.c. covered wires,


connect the master potentiometer tag strip (Fig 6.7)
to the solder tags on the power pack output terminals.
Also, temporarily link the rear of MP/SK2 to the green
earth wire. Rotate VR20 spindle fully clockwise
and patch MP/SK2 to SK3, MP/SK5 toSK6, MP/SKI
to SK4, and link RM/SK2 to VS1/SKl. Switch on the
computer and S6, and adjust VS1 for an exact + 10V.
Now obtain a null on the readout meter by setting VR1
on the voltage divider bracket, from the back of
UNIT "B" box.
'
Repeat for VR2 with an input of + 1V by transferring
the patching lead plug from MP/SK6 to SK7, and again
for VR3, SKS, with an input of ~ IV, and VR4, SK9,
with an input of -10V.
After that, while still nulling with a .- toV input,
rotate VR20 spindle slightly clockwise, until the meter
pointer just begins to" move away "from zero. Place
the large knob on VR20 spindle, with the transparent
plastic cursor aligned with the "to" division, and tighten
the grub screw. Set VR20 cursor to the "5" division
and check for null with an input of -5V. It may be
necessary to slightly re-position VR20 knob on the
spindle, and trim VRI-VR4 again to minimise errors.
Calibration of the readout meter is straightforward.
Apply a selection of known voltages to RM/SK1 and

Fig. 6.8. Meter resistor panel, underside view

wiring of master potentiometer components, using


20 s.w.g. tinned copper wire and sleeving.
READOUT METER ASSEMBLY

" Make up the meter resistor panel shown in Figs 6.7


and 6.S, and attach to the meter terminals. Solder D1
and D2 to MP/SKI and RM/SK2, then complete Sto
and resistor panel wiring.
As centre-zero voltmeters with to-0-10 and 3-0-3
scale calibrations are not readily available, a scale will
have to be made. Perhaps the most satisfactory way
of fabricating a new and really accurate meter scale is to
draw it two to four times full size, photograph it, and
then have the resulting negative enlarged back to the
original size on glossy photographic paper. The
enlarging can" be done commercially if the oversize
drawing carries a thick black line to represent a length
of 1in on the finished scale, just outside the scale
perimeter.

SINGLE SHOT
~

,.t

TRIG6ERO--O
INPUT

Si till
i: lI! n81

MONOSTABLE

'"

"'I

HOLD

: S1

::52i

~:::ED
Rt.
Rr

ico-J'V'V\~-""'--""'-r--+-""-"""'---.

S1

REPETITIVE
52 . S1

I
I t:

-.I i!i
.,

11 ... 111

II=>

E2o-~~""~"",--_",,,~~! 0--4~1~

R3 "

S1

II~II

t;;

lIull

.,

II~IIII

CLOSED
OPEN

11
11 II t;;
111;11 ~

IItsll '"
~

:>-.....+-00..

-J.- YIRTUAl.
- EARTH

436

110 11

TIME
R2

52

FTI---uTI

R1

YRl9

Fig. 6.9. Diode clamp circuit.


showlnl principle of operotlon

r-------

S9C

11

+1z.5V

r-------------~----------------~--~I~E~D~
~
VII
~uu

ro~

INTEGRATOR.SWITCH
CIRCUIT PANEL

!i~

R7 Z7kll

Sl(ft

EXT. TRIG.

V
GREEN

GlEEN

SK1Z
EARTH

Cl
1000"F

115
560kll

__ ...L_______ _

--T------t
I

!-UNSTAallISED
TO-1ZV
SUPPLY

TO -1ZV UNSTABlllSED
SUPPLY (OPTIONAL)

L..-_ _..::(OP::,TIONAL)

WHITE

598

DIODES: D3-014 lB30 GOlD BONDED

SKl

SK6

SK9

Fig. 6.10.

SK7

SK4

SKI

The complete circuit diagram of the integrator switch

adjust VRl-VR4 on the resistor panel for optimum


accuracy on each range.
INTEGRATOR MODE SWITCHING

The simplest type of integrator switch employs a


mechanical relay with several sets of contacts, driven
by an astable multivibrator, and this system is used for
small demonstration and educational analogue computers. The relay is arranged to "gate" 'the inputs of
.several amplifiers simultaneously.
The PEAC integrator switch goes a stage further, with
reed relays for a "clean" switching action at high speeds,
full initial condition facilities, and a circuit based on
two independently timed monostable multivibrators.
Referring back to the basic integrator switch shown
in Fig 1.2c, two changeover switches SI and S2 are
opened and closed in a pre-determined sequence,
.governed by an external timing circuit. It is important
to ensure that integrating amplifier input resistors are
not left floating when they are disconnected from the
virtual earth summing junction, as this could seriously
disturb input and other computer voltages, hence the
presence of SI and S2 earthed contacts.
DIODE CLAMPS

To eliminate the need for expensive reed switches


with changeover contacts, diode clamps can be used
instead of an earthed contact, see the alternative

amplifier circuit of Fig 6.9. The diodes do not interfere with the normal working of the integrator, but
will nevertheless hold resistor junctions close enough
to earth to prevent load variations when SI and S2
are open, and this modification more than halves the
cost of switching components.
In the block diagram of Fig 6.9, the 1st monostablecontrolled by VRl8-determines the period of closure
of SI. When SI opens after a timed interval, a pulse
is delivered to the input of the 2nd monostable, thus
closing S2. S2 will remain closed for an interval
controlled solely by VRI9.
For "single shot" operation, a trigger pulse applied
, to the 1st monostable input, when S8 is switched to
"hold", will initiate the closure of SI (reset) and bring
the integrating amplifier to its initial condition.
As soon as SI opens, S2 closes (compute) and
connects input resistors to the summing junction. At.
the end of the compute period, SI and S2 are both
open (hold), the monostables are quiescent, and the
amplifier output voltage is held steady by the action of
capacitor Cr. The next computer run is started by
another trigger pulse applied to the Istmonostableinput.
Repetitive operation is achieved by passing the
output pulse from the 2nd monostable back to the input
of the 1st monostable, when S8 is switched to "repeat".
SI and S2 are then made to open and close alternately,
and the "hold" facility is deleted.
439

The method of inserting an inital condition voltage


is as follows. When SI is closed the reset resistor Rr is
connected between the amplifier output -and summing
junction, and can therefore be regarded as a feedback
resistor in parallel with Cr.
As long as SI remains closed, Ric will be acting as an
input resistor, so that

and Eo = - Eic when Rr = Ric. Ric and Rr are disconnected from the a!l1plifier summing junction when
SI opens, but Cr will "remember" the initial condition
voltage and hold the amplifier output steady prior to the
a-pplication of compute voItages when S2, closes:
INTEGRATOR SWITCH CIRCUIT

The complete circuit of the integrator switch 'is shown


in Fig 6.10. The 1st monostable consists of TR2 and
TR3, with RLA actuated by emitter follower TRl.
VR18 continuously covers two ranges given by C3
(IO-lOOms), and C4 (OI-ls). Components associated
with the 1st monostable input are C2, RI, and Dl.
The 2nd monostable' is almost identical to the 1st.
TR4 drives RLB, C7 and C8 offer the same timing range
coverage as C3 and C4, and input components are C6,
R8, R9, and D2. However, more care is taken to
establish the correct values for 2nd monostable timing
capacitors C7 and C8, and VR2 allows precise calibra, tion of the "fast end" of the VR19 timing scale, so that
compute intervals can be determined by a reasonably
accurate dial setting.
VRl establishes the working point of both monostables, to achieve reliable operation at all dial settings.
S7 'is a push button on the front panel for starting a
"single shot'~ computer run. Full control of an oscilloscope trace, from UNIT "B" front panel, can be
realised by suitable connection to the integrator switch
circuit. With S8 switched to "hold", the mode
sequence can be triggered repetitively, with a variable
hold interval, by the oscilloscope timebase output or
by a separate oscillator. Consistant syncronisation of
the trace, with continuous or single-sweep timebases,
is made possible by linking IS/SKI0 to an appropriate
oscilloscope input
A SEPARATE SUPPLY

The load capacity of the existing stabilised power


supply can be improved by wiring the collectors of
TRl and TR4 (shown dotted in Fig. 6.10) to a-separate
-12V unregulated supply, which can be housed inside
the UNIT "B" box, and in this event Cl could be
omitted from the Fig. 6.10 circuit, as it merely serves
to prevent current pulses from flowing in the negative,
stabilised supply line during relay switching.
RLA and RLB consist of two triple-switch coils,
catering for the needs of three integrating amplifiers.
A duplicate relay panel could be added I later, by
wiring relay coils in parallel, to increase the switching
capacity to six amplifiers.
CORRECTION
In Fig. 5.7, the captions for the first and second oscillographs (top row, left and centre) should be transposed.

Next month: Assembly and setting _up of the


Integrator Switch; practical examples in the use
of this section. I ntroduction of UNIT "C"
'Function Generator.

440

on its blocks, and, following Fig. 6.10 and Fig. 7.3,


wire all controls and sockets to the turret tags on the
two sub-assembly panels, again with p.v.c. covered
flexible wire, long enough to allow the switching circuit
panel to be turned over for underside inspection. Run
.red and blue wires from S9, and a green wire from
JS/SKI2, to the power pack output solder tags, and
fit knobs to S9, VRlS, and VR19.
SETTING UP THE INTEGRATOR SWITCH

Time intervals can be measured with fair accuracy


when an operational amplifier is employed to integrate
known voltages, and this method is useful for setting
up the integrator switch.
Begin by temporarily soldering SpF electrolytic
capacitors in the C4 and CS positions, with IpF polyester capacitors for C3 and C7 (circuit Fig. 6.10).
Set VRl and VR2 with sliders at mid-track, on the
integrator switch panel.
Connect integrating switch to the operational
amplifier by linking IS/SK7 to OA3/SK9, IS/SKS to
OA3/SK10, and IS/SK9 to OA3/SK4. Fit 100 kilohm
computing resistor in S3/Il/SK3 and SK4. Join
S3/Il/SKl to VS1/SK2 and switch off S6. Insert a
2 kilohm reset resistor in OA3/SK5 and SK6, and
join S3/SK5 to OA3/SK13.

&OO~[L@@M[g

@ U\"liJ [p M lJ~~
PEAC
T

o COMPLETE the construction of UNIT "B", we have


now to deal with the integrator switching section,
the circuit diagram for which has already been given,
see Fig, 6.10.
INTEGRATOR MODE SWITCH ASSEMBLY

Cut and drill the 6tin x 2tin s.r.b.p. panel shown in


Fig 7.1, and rivet turret tags in the positions shown.
From six transistors select two with the highest current
gain for TR2 and TR5. Mount all components,
except range capacitors C3, C4, C7, and CS, on the
s.r.b.p. panel and wire up.
Prepare the 3in x 2in relay panel, from Fig. 7.2.
Fix turret tags and mount RLA and RLB reed coils.
Next, insert miniature diodes D3-D14, with alternating
polarities along the row of diodes, and complete
underside wiring. To finish off the relay panel, place
three reed switches in each coil and secure by soldering
the lead out wires to appropriate turret tags.
Wooden blocks are glued to the rear of the UNJT
"B" front panel to serve as mounts for switching
circuit panel and relay panel (see Fig. 7.3). Note
that the relay panel is fitted end-on into slots cut in its
mounting blocks, and the switching circuit panel is
secured by two woodscrews.
After first attaching lengths of black and white p.v.c.
covered multi-strand wire to the terminals of VRlS
and VR19, screw the switching circuit panel in position
500

By D.BOLLEN
. Switch on the computer and allow a warm up' period
before zero setting OA3 from the back of the UNIT
HA" box, by means of VRl on the OA3 amplifier
panel. Insert a 1pF computing capacitor into
OA3/SKll and SKI2.
With SS switched to "hold", S9 on the 0'1-1s range,
and VRlS and VR19 rotated fully clockwise, press S7
to run the integrating amplifier through reset, compute,
and hold sequence.
.
Listen for two clicks from the reed relays, and observe
that the readout meter pointer will move close to zero.
If the relays click more than twice, or not at all, adjust
VRl on the integrator switch panel.
To obtain a true zero output from the amplifier,
when integrating a zero input voltage, adjust VR17
(OA3 balance control) while repeatedly pressing S7.
If there is a slow drift away from zero 9utput several
seconds after S7 was last pressed, retrim V,"1 on the
OA3 amplifier panel.
.
As the gain of OA3 is set at 10 (lpF for Cr and 100
kilohm for Rin), an input of -0'9V "gated" by the
integrator switch. for an interval of Is should give rise to
an amplifier output of exactly +9V. Switch on S6
and adjust VSl for -0'9V, monitored at S3/Il/SK2 by a
voltmeter.
Now when S7 is pressed, and with VR19 still rotated
fully clockwise, the readout meter reading should rise
to somewhere beiow + 9V and stay there.

r-I~o---o---H-O-UN-T-IN-G--HO-L-E---

14

o
OTT
OTT

00

og
0

000

00

0
0 00
o 0
0

0
TT

0
TT

TT

o 0
0

0 00
0 0

0
0
0

ggTToo

if 0 0 ~~Att ~~I~~E~O T~W 0 0 T~ 0 0 0


0 0

TTO

OTT

013

TTO

ggTToOj
OTT

"-0---o-T-T-----TT-o--.ja'1

00
0

TT

TT TT

HO~~[~N~
on

-Ix. 0;
0

m~

00

00

00

000000 0

J1
~_.

h Z

OTT

o
OTT

OTT

OTT

OTT

OTT

OTT
v

x
WHITE

aLA

Rt

Fig. 7.2. UNIT "S" integrator switch


relay panel. (a) drilling template; (b)
component arrangement; (c) underside
wiring

It

GREY
MAUVE

BROWN
TRZ

SKI

SK4

SK7

Slt3

SK6

SK9

BLUE

BLACK
(b)

RlB

RlA

.le)

Fig. 7.1. UNIT "S" integrator switch circuit panel, (a)


drilling template; (b) component arrangement; (c) underside wiring

Build up the value of CS timing capacitor by adding


more capacitors in parallel, until the +9V output is
obtained when S7 is pressed.
To check the "fast" end of VRl9 scale, set VSl for
-9V and rotate VR19 fully anti-clockwise. Adjust
VR2 on the integrator switch panel to obtain the desired
amplifier output of +9V for a compute interval of
O1s.

.BROWN

BROWN

GREEN

GREEN

(c)

(c)

Section of UNIT "S" pane'


,(viewed from rHr) showing
Integrator switch relay assembly
501

vi

SK9

I
SLOTT E0
.--------,------,WCIODI;N BLOCKS

@
SK1

@
INTEGRATOR
SWITCH
CIRCUIT

PANEL~

FI 7.3. Rear view of


UNIT "S" front panel
showin.
InterrCltor
switch wirln.

TO S10

PROSLEM fXAMPLf 4

STRAIGHT PATH MOTION OF AN OBJECT


CALIBRATING THE SECOND RANGE

To calibrate the 100looms S9 range, repeat the above


procedures in just the same way, but this time use a
O'lpF capacitor for et in sockets {)A3jSKll and SKI2,
and adjust the value of timing capacitor C7 for correct
compute intervals.
Ist monostable timing capacitors C3 and C4 need
not be precise, as VRl8 has no effect on the accuracy
of computations, and is mainly used to control the
switch cycle frequency when integrator output waveforms are displayed by oscilloscope. Therefore, and
merely for the sake of conformity, build up C3 and C4
capacitor values until the cover~ge of VR;18 is ~ppr~xi
mately as indicated by the reset mterval dJal calibration.
CIRCUIT ADJUSTMENTS

The Fig 6.10 circuit should operate reliably at all


switch and dial 'settings, with no noticeable relay
bounce or overlap between the closure of reset and
compute switches. However, it may be found that the
integrator switch will stop running during repetitive
operation, when reset and compute intervals approach
IOms, despite the fact that VRI has already ~n
trimmed for optimum performance. If so, try reducmg
the value of R8.
At the opposite extreme, if the. igtegrator switch
suddenly goes into repetitive operation when S8 is at
"Hold" and VR18 and VR19 settings are near Is,
increase'R8, and also try the effect of doubling the value
of Cl to improve decoupling.
502

Problem Example 4 is primarily intended as a


comprehensive introduction to the use of integrator
mode switching, but the programme is sufficiently
flexible to allow many experiments in dynamics to be
performed.
Several factors can combine to influence the overall
motion of an object, and some are shown in the ball
problem of Fig. 7.4. A ball thrown vertically into the
air will be subject to an initial upward velocity iv,
retardation or negative acceleration due to gravity
- a and air resistance. The situation is further
complicated if the ball is projected upwards from an
initial height is, and is arrested at some height other
than zero.
Ignoring for the moment air resistance, the equations
which govern the motion of the ball are,

and

J:
J:

a dt

+ iv

(Eq. 7.1)

v dt

+ is

(Eq. 7.2)

Clearly, integration of a yields v, and a further integration of v will give s.


The formulae used to calculate velocity or distance
when acceleration is constant are,
v = iv + at
(Eq. 7.3)
s = ivt + -!-at 2 + is
(Eq. 7.4)
and
Eq. 7.3 and 7.4 will not apply if, for example, acceleration is proportional to time. A discussion. of the
implications of variable acceleration lies outSide the

scope of this series, but time varying voltage analogues


of acceleration are fairly easy to generate on the
computer.
.
The drag on a body moving through air or a fluid
conforms to an exponential law, and is prowrtional to
velocity when there is litt1e or no turbulence. Viscous
friction should not De confused with the friction
resulting from solid surfaces in contact, as the latter is
independent of velocity except at very low speeds. A
general solution to an equation which describes the
motion of an object through a viscous medium-where
composite velocities are involved-is often unwieldy
and can demand extensive calculations.
However, an exponential decay can be set-up on the
computet to simulate true viscous friction, in terms of a
coefficient value p which remains constant for all
velocities. Nevertheless; as p will be dependent on
such factors as the surface area, shape, and relative
smoothness of an object,.it can only be determined by
practical experiment, or by comparison between the
computer solution' and the timed motion of an actual
object.
Looking at the symbolised diagram of Fig 7.5, OAI
is employed to integrate a known voltage against time,
so that t tan be conveniently and accurately displayed
as a meter reading. OA2 integrates a to give an output
v, and at the same time handles the initial velocity iv.
The exponential decay e-(Pjm)t is introduced by CPl.
Resulting velocity v is then integrated by OA3 and
initial distance is is included to give distance or height
s at any time t.
Routine. Set-up the problem according to the
simplified patching circuit of Fig. 7.5 but omit for the
time being all Cr capacitors. The integrator switch
is linked to the three operational amplifiers by connecting ISjSKI to OAljSK9, IS/SK2 to OAI/SKlO,
IS/SK3 to OAljSK4, IS/SK4 to OA2jSK9, IS/SK5 to
OA2jSKlO, ISjSK6 to Oi\2jSK4, IS/SK7 to OA3/SK9,
ISjSK8 to OA3jSKlO, and ISjSK9 to OA3jSK4.
Allow the computer to warm up before zero-setting
the amplifiers, alsQ make sure that S6 is oft. Using
the readout meter "on its IOV range, zero-set amplifier
outputs (OAljSK13, S3jI5/SK2, and OA3jSK13) by
means of VRI on each amplifier panel, from the back
of the UNIT "A" box.
Next insert the Cr computing capacitors into amplifier feedback loop sockets (SKI I and SKI2) and set the
int~grator switching controls to give reset and compute

RETARDATION DUE TO

GRAVITY a=-32FT/SEc

I
t __ ------------ +S MAX
I
1

I
I

...

I
I

.it 1JIt
1

AIR "RESISTANCE I'

MASSm

INITIAL VELOCITY

+.IV

tall
I

DIST~NCE
(HEIGHT)
~
+,S
-l-----------1

...
1

~r_:;>__"7__"7"_r,-_...:---------5=0
I
"
I

I
I
I

I
1

I
I

Fig. 7.4. An experiment In dynamics with a ball

times of approximately 01 second. Put S8 in the


"hold" position. With the readout meter on its I V
range, applied to the output of OAI, press S7 and
adjust VRl5 for a zero voltage reading. Repeat
for OA2 output and VRI6, and OA3 output and VRI7,
in that order. The amplifiers should now be
balanced for near zero input offset voltage.
+

Fig. 7.5. Symbolised


dlogram of the ball
problem illustrated
in Fig. 7.4.

"

INITIAL

10
W=1 SEC REAL TIME.
OA1 OUTPUT CAN BE
USED FOR TIME VARYING
ACCELERATION
REAL TIME=COMPUTE TIME X10

STATIC CHECK VOLTAGES AND POT SETTING SHOWN THUS


DYNAMIC CHECK VOLTAGES WHEN COMPUTE TlME=01 SEC

lIEl...

e.

5
t

I'

TYPICAL
UNITS
FT/SEC 2
FT/SEC
DISTANCE
FT
TIME
SEC
MASS
LB
VISCOUS FRICTION COEFF.

r'

503

To enable static and dynamic checks to be made,


trial values are given to the ball problem of Fig. 7.4, as
follows: treal = 1 sec, a = - 32ft/sec2, iv = 25ft/sec,
is = 10ft, v = - 7ft/sec,' "s = 19ft, and ft/m = O.
The problem scaling is such that 1 computer volt = 10
units in all cases. For example, 1V = 1 sec for t at
the output of OAl (10 x compute time), and 19V =
19ft for s at OA3 output. Calculation from the formula
Eq. 7.4 shows that the ball will have travelled just
beyond Smax after a time of 1 sec, when air resistance is
zero.
The next stage is to establish all computer static
voltages shown in the Fig. 7.5 symbolised diagram,
starting with VS1. Set the dial of the master potentiometer to "10" and patch MP/SKI to SK4, MP/SK2
to SK3, and MP/SK5 to SK8 . . Connect RM/SK2 to
Sl/11/SK2. Switch on S6, set switch SlO to "null"
and adjust VS1 dial for a null meter reading, corresponding to a voltage source output of - 1V. Remove
the null input patching lead completely, and use it to
link RM/SK1 to OA1/SK13 ..
With the readout meter on its IV range, press S7,
;tnd trim compute time control VR19 for an integrator
output of IV; this will ensure that the compute interval
is exactly 0'1 sec. Set up VS2, VS3, and VS4 check
voltages, preferably by nulling with the master potentiometer to avoid loading, and rotate CP1 fully anticlockwise. Switch off S6 and press S7 to reset the
amplifiers. Check that amplifier outputs are zero.
To obtain dynamic check voltages, switch on S6 and
press S7, while applying the readout meter to the outputs of OA1, OA2, and OA3 in turn. For greater
convenience, three separate voltmeters can be left
connected as shown in the patching circuit of Fig. 7.5
to give simultaneous readouts of t, v, and s. Before
altering other problem variables, introduce air resistance
by means of CPl and arrest the travel of the ball at
selected positions along its path by adjusting the compute time. It is instructive to compare the velocity
and distance of the ball when a = - 32ft/sec2 and
friction is present, with a ball projected upwards.under
moon . gravity conditions (approximately a =
- 5'3ft/sec 2) in a vacuum.
The existing scaling of layout Fig. 7.5 will provide
the following coverage: .. VR2 O- 100ft/sec2 , VR3
O- IOOftjsec, VR4 O- 100ft, with amplifier outputs
of OAl O'I-lOsec, OA2 O-lOOft/sec, and OA3
o

O- 100ft. The coefficient of CPI covers the range


0-10 for ft/m.
If at any instant during a computer run velocity
exceeds 100ft/sec, or distance is greater than 100ft,
this will . result in amplifier overloading, and a false
problem solution. Spot checks of velocity or distance
voltage trends can be made at selected compute times,
using the single shot facility, and Smax will correspond
with v = 0 at a particular time t. Alternatively,
during repetitive integrator switching, an oscilloscope
will serve to show amplifier overloads as a flattening or
clipping of an output waveform, but this should not be
confused with the short "hold" interval which separates
the opening and closing of reset and compute switches.
RESCALING PROBLEM EXAMPLE 4

The programme of Problem Example 4 need not be


confined to the vertical motion of an object in air, but
could equally well apply to mQvement up and down an
inclined plane in water, or else the horizontal progress
of a fast wheeled vehicle being decelerated by braking
forces, for example.
There are several ways of tescaling Problem Example
4, the most obvious being the adoption of other unit
systems, such as miles/hour, centimetres/sec, or even
inches/year. Providing that compatible units are
employed, and computer voltages are correctly interpreted, there are no serious barriers to unit system
rescaling. Probably the most straightforward way of
verifying a new problem scaling is to set up a simple
check problem, where known values of t, a, v, and S are
computed for an object in a vacuum, to establish the
relationships between static and dynamic voltages.
Where it is desired to extend the range of an existing
unit system, increasing the value of computing capacitors by a factor of ten will reduce real time by ten.
Similarly, a tenfold increase in real time is achieved
when Cr values are divided by ten.
When employing large computing capacitors at
short compute times, always ensure that the reset
.resistor Rr is small enough to completely discharge
Cr during the reset interval. It is also possible to alter
the computer voltage scaling so that, for example,
1 computer volt will equal 100 units instead of 10 units,
but care should be taken to make sure that all voltages
and potentiometer settings conform to the new scaling.
Finally, a word or two about variable acceleration.
If the input to OA2 is transferred from the VS2 source
to the output of OA1, acceleration will then be zero
Fig. 7.6. Simplified patching circuit for the ball problem

INPun

RllOOkn

SUMMER 1

",O,,-P-"U.'--'----J--'

CPl
TIME
VOLTMETER

504

VELOCITY

VOLTMETER

"":'"

DISTANCE ~
VOLTMETER

When employed for squaring an


input voltage, with both networks
operating in parallel, the function
generator will accept input voltages
of O- 10V, 'and yields amplifier outputs of up to 10V. Accuracy can
be within 2 per cent of the indicated
value, depending on the care taken in
setting up a function, for input
voltages between 02V and 9V.

LINEAR FUNCTION
NLR-NON-LlNEAR

Rf

Rf

(\

RESI~STOR
NLR ~ f Clnj
.

-tanoc

NON-LINEAR FUNCTIONS

Quite often some non!linear


function of an applied voltage is
needed in analogue computer work,
NLR
Co=-f (cin)
two simple instances being the square
or square root of a number. An
FUNCTION PRODUCED BY NON - LINEAR INPUT RESISTOR
arbitrary function may also be enNLR
countered, perhaps arising from
experimental data for which no
analytic expression is available.
Cin o--.JW-.......-t [>-.....--<>Co
Servo driven potentiometers and
Co= -fZ (cin)
circuits consisting of biased diodes
are widely used for generating nonFUNCTION PRODUCED BY NON-LINEAR FEEDBACK RESISTOR
linear ' functions, but the latter is
deservedly popular because it can
ffZ
R fZ
....-NY.......... RI=1
be adjusted to cater for a range of
functions, and does not suffer from
a severely limited frequency response.
' To show how a diode function
generator can give rise to non-linear
functions, when allied to operational
FUNCTION SIGN CHANGE
amplifiers, use is made here of the
parallel which_exists between the discontinuous behaviour of a biased
diode network, and the smooth
response of a voltage dependent
resistor. Both can display a fall in
resistance with an increase in applied
voltage.
VARIABLE NON - LINEAR FUNCTION
Consider first of all the circuit and
generalised curve of Fig. 7.7a. Input
and feedback resistors Rin and Rt
,
I
are not influenced by applied voltage,
~J~-+--".,....
A>--4---oc.,
-f(cll- c z "',
therefore a straight line function is
co--f(cln) :!:cZ
generated, while amplifier gain and
RI
\+cZ
tan Cl: remains constant. However, if
r.
VOLTAGE SHIFTING OF NON - LINEAR FUNCTION
some form of non-linear resistor,
or biased diode network, is
Fig. 7.7. Generating non-linear functlom with a yoltage dependent resistor
substituted for Rin (NLR in Fig.
7.7b) the gain of the amplifier
tends to grow with an increase of Ein, and the tangent
when t = 0, and increases linearly to IOft/sec 2 when
to the curve will vary according to some function
t = I sec real time. VSI can be used to adjust the
f(El n), arising from the characteristic of NLR. A
magnitude of a when t > O. Also, if OAI initial
related function f2(Ein) results when NLR is exchanged
conditions are inserted, in a similar manner to OA2
for Rr, as in Fig. 7.7c, but here the amplifier gain falls "
and OA3, many other time functions of a can be
off with an increase of Ein. The curves of Fig. 7.7b
generated.
and Fig. 7.7c only occupy two of four possible quadrants, but four quadrant operation can be achieved if
UNIT "C" FUNCTION GENERATOR
the function is inverted by a sign changing amplifier,
UNIT "C'" contains two diode-resistor networks, one
depicted in Fig. 7.7d.
for positive input voltages, and the other for negative
inputs. The characteristics of each network can be
Fig. 7.7e shows how curves, of widely differing slope
adjusted separately by means of miniature pre-set
and magnitude, may be generated if the characteristic
potentiometers to give a wide range of possible functions,
of NLR is alterable. Finally, any fixed function will
and optimum accuracy. The function generator is
find wider application if its Ein = 0 datum is shifted,
designed to be used in place of a ,normal computing
as in Fig. 7.7f. Moreover, as a voltage shift can also be
resistor, at the input or in the feedback loop of an
applied to the Eo axis, it becomes a simple matter to
operational amplifier.
locate any portion of a curve in any quadrant.
~n

505

...-_-OONEGATIVE BIAS

VR1

POSITIVE
BRANCH

VR4
02
-cbZ

INPUT
c .in

+t

VR3

VRS

Computing ctlpacitors mounted inside


small plastics boxes
SUMMING
JUNCTION
03

VR6
NEGATIVE
BRANCH

'VR7

04
VR8
L-..6--00 POSITIVE BIAS

Fig.7.8a. Circuit of a simple function generator

BIASED DIODE NETWORK

The next step is to see how biased diode networks are


used to achieve an increase of resistance with applied
voltage, and thus imitate the behaviour of an ideal
voltage dependent resistor. Unfortunately, currently
available silicon carbide, selenium, and copper oxide
resistors are far from ideal in many respects, and are
not sufficiently accurate for serious use with operational
amplifiers.
.

~REAKPOINT

VR6

--.I
I

--.L..

...,.L..

BREAKPOINT
VR4

BREAKPOINT
VR2

BREAKPOINT
VR8

--.I

+Cin

Fig. 7.8b . . Adjustable characteristic of simple function


generator
506.

The UNIT "C" function generator is based on the


simple circuit of Fig .. 7.Sa. In the absence of an input
voltage all diodes are biased off, and the network can
be represented by a very high value of resistance in
serie~ with !he operational amplifier input, giving an
amplIfier gam of almost zero. If a positive voltage is
gradually applied to the input terminal, there will be
virtually no output until a point is reached where
Ein is slightly larger than -Eb, whereupon Dl conducts
and connects VRl to the operational amplifier summing
jU!1cti~n. Further increase of Ein, beyond -Eb,
will produce a straight line ~utput of slope determined
by the amplifier gain Rr/VRl.
When Ein reaches approximately the level of -Ebz,
D2 conducts and places VR3 in parallel with VRI, thus
reducing even more the effective' resistance of the
network. It can be easily imagined that where a
nUinber of diodes and variable resistances are cascaded,
the resistance of the network will continue to fall as
Ein becomes larger stilI.
.
Bias voltage - Eb is determined by the relative resistances of VRl and VR2, and the same applies to
- Ebz, VR3 and VR4. Furthermore, the setting of
VRl will obviously affect the combined slope of VRl
and VR3 (see Fig. 7.Sb), and it follows that all the
resistance settings associ(!ted with DI and D2 must be
interrelated.
Considerations applying to the positive branch of
circuit Fig. 7.Sa are also pertinent to the negative
branch formed by D3 and D4, and VR5-VRS, except
that input and bias voltage polarities are reversed.
There is no interaction between the resistance settings
of the positive branch and the negative branch, and the
two can be separated when required for independent
use.
The output characteristic curve of Fig. 7.Sb identifies
slopes and breakpoints with VRl-YRS. As there are
only two diodes in each branch, the result is a very
rough approximation to a smooth curve. Generally
speaking, the accuracy of a diode function generator is
proportional to the number of diodes employed, but a
natural rounding at the junction of straight lines does
occur at low input voltage levels, due to the dynamic
resistance of the diodes (not shown in Fig. 7.8b), so the
deviation from a smooth curve is not as great as might
be expected. Commercial diode function generators
sometimes use more than 20 diodes to achieve accuracies
of better than 1 per cent.
Next month: Construction of UNIT "C" and
some practical applications of this Function
Generator.

The natural forward voltage drop of Dl and D2


furnishes self-bias, and bias conditions for D3 are
satisfied by a fixed resistor R I. The values of slope
adjusters VRI, VR2, YR3, YR4, YR6, VR8 and VRIO
were selected to give a parabolic function approximating
to Eo = El02 when all sliders are at mid-track, and
appropriate bias values for that function are provided
by mid-track settings ,o f breakpoint adjusters VR5,
YR7, YR9 and VRll. The combination VRl2 and R3
serves to eliminate offset voltages resulting from diode
leakage currents, and VR 12 is therefore used for zerosetting.
With so' many possible adjustments, including
amplifier closed-loop gains determined by Rr or Rio
computing resistors, it is obviously impossible to cata- .
logue the coverage of the Fig. 8.1 circuit. As a rough
indication though, powers of EIIl ranging from about.
El 01 . l' to beyond Eln3 are available. If both branches
are cascaded in series with operational amplifiers,
the upper limit will extend beyond Eln6 Corresponding
root functionsl.1y1Eln to 6y1Eln may also be generated.
It is sometimes possible to use the UNIT "c" function
generator for certain trigonometrical functions, and
. logs to the base 10 or e.
UNIT "C" BOX

PEAC

A wood and plastics laminate box, of small


dimensions compared with other PEAC units, will serve
to house the two function generator circuit panels.
The suggested form of construction is shown in Fig. 8.2.
Softwood blocks are glued to a 9!in x 4in x -kin
plywood frame, which has its centre cut out, and white
plastics laminate side pieces are then glued to the blocks.
The front panel sits on the wooden blocks and is
recessed.

!~![L@@(W~

D\"liJ [)(bD Lr ~ ~

ByD.BOLLEN
month the Function Generator UNIT "C" was
introduced. The principle of operation and some of
the uses of the function generator were explained. We
are continuing with a description of the practical circuit,
constructional details, and application information.
LAST

FUNCTION GENERATOR CIRCUIT

The function generator circuit of Fig. 8.1 is designed


to display a nominal resistance of 100 kilohm when the
input voltage is I Y. A typical resistance variation
with applied voltage is from 500 kilohms at O'2V to
10 kilohms at IOV. In the Fig. 8.1 circuit, components
forming the positive branch are identified by the
letter A after a component number, and the letter B is
appended to negative branch numbering. As both
branches are identical, except for diode and bias
polarities, it is not necessary to describe them separately.
DI is a gold-bonded diode, for a low voltage drop
with small input voltages. All other diodes (D2-D7)
are of silicon construction to keep reverse leakage low.
564

UNIT "C" FRONT PANEL

The only items to be mounted on the 9tin x 4in


plastics laminate front panel are eight coloured sockets;
the layout is given in Fig. 8.3. A series of -kin holes are
drilled in the front panel to allow screwdriver access to
slope, breakpoint, and set-zero controls. Panel markings are similar to previous PEAC units.
FUNCTION GENERATOR
CONSTRUCTION

Two 3tin x 3tin s.r.b.p. panels are drilled and


shaped according to the Fig. 8.4a diagram. Before
inserting turret tags, lay the prepared panels out as
shown in Fig. 8.5, so that one panel is turned over in
relation to the other, and components are clearly
seen to be mounted on opposite sides. The underside
wiring of the positive branch panel is shown in Fig.
8.4b, and the wiring ofthe negative branch is in Fig. 8.4c.
All diodes are mounted on turret tags to allow them
to be disconnected for special purposes, where for

.....
....

VRIA
INQ

i.

D2A
OA202

VR2A
lZOkQ

....

"i".

OlA
OA202

VRSA
220kll

..

_..."1:~

RIA
l'SMIl
.t

VII4A
lOOkn

..."1:.
POSITIVE
IIIdeH

VII6A
lOOkll

"i",
VillA
47kll

..

.~.

"
SKI SK2

'"'

'"'

VRIOA
22kn
:":

+INI'UT
VlllZA
41kll

VIISA
INQ

'
VillA

4~o.."..Q

. T'
VII9A
220kll

'TO
Villi A
41kn

'TO

,"1: ...

,:.INI'U!.

'TO
VlltOB
22kll

..

VIII8
41kQ
NEGATIVE
BIIAItCH

.....
.....
I
.....
~; .....
.....
....

-'r
YII6B
100kll

- 'TO

example it is desired to reduce the number of breakpoints, or combine a curved and straight line function .
It is advisable to check the polarity of all diodes with a
meter before mounting them on the circuit panels.
After completing the underside wiring, bolt the two
circuit panels on the plywood frame, as in Fig. 8.5, and
make sure that the front panel holes are aligned with the- pre-set miniature potentiometer slots.

D4A
OA202

R2A
410kll

DSA
OA202

....

""
~~
~..,

c~

!O! ...
!=: ...

~z

:!:5!
-~
t;~

",=>

~~

~6:

I "c,

OA202

.....

D1A
OA202

VIIIIB
41kll

~.

VII9B
220kn

.1".
VR7B
470kn

..~.
VIISI
IMIl

1............

018
OAZ02

SKS SK4
~

+OUTI'UT

.....TO

:w

VIISB
220kn

';:'
VII28
ZZOkQ

-OUTPUT

,...

,...

SKl sKi

I .......

068
OA202

"

OSB
OA202

1 ......
.....

~,1I21
470kll

....
.....

VII41
lOOkll

Fig. 8.1.

Function generator circuit fJanel

RSA INn
...1S8IMll

VRI28
41kQ

SkS SK6

-12'SV

OIA IIlO

041

1111
BNA

OA~02

----~;;

OAZ02

......
~;:

..

OAZ02

'To

L ...

Villi
INn

01: ;'0

+IZ'SV

UNIT "C" fu.,nction generator circuit diairam

SETTING UP THE FUNCTION


GENERATOR
Patching leads for the function generator should
preferably be terminated at one end by miniature
plugs, to permit connection to the UNIT HA" computing component sockets. As the generation of powers
and roots is the main area of interest, functions related
to the square or cube of a number are used in the
following setting-up instructions.
To patch the function generator to OA I, join
FGjSK5 to SljIljSK3, FGjSK8 to Sl jIl/SK4, SIjSK5
to OAl/SK8, and link together OAl jSK9, SKIO, and
SK4. Insert a lOO kilohm computing resistor i!lto
OAl jSKll and SKI2. Take a patching lead from
SljIljSKI to VSl/SK2, and ensure that S6 is off.
The task of setting up the function generator is made
easier if two voltmeters are used, one for Eln connected
to SI/Il/SK2, and the other for Eo to OAljSK13.
ihe Unit " B" readout meter is ideal for monitoring
Eo because it can indicate voltages down to OOIV .
Switch on the computer power supply and zero OA I
by means of its balance control VR 15. Set all function
generator slope and breakpoint potentiometer sliders
to mid-track, and connect the red and blue wires from
the function generator to the power supply terminals
on the side of the UNIT "A" box (TU and TL2).
Adjust VR12B (zero-set) for zero output from OAJ.
Because of the interdependence of slope and breakpoint adjustments, a systematic approach is called for
when setting up a function. Start with the lowest
Eln and VRl and proceed in an orderly fashion towards
VRII and the maximum Eln value. It is a help to
tabulate specific input and output voltages and relate
them to particular slope or breakpoint controls.
To assist the reader, two tables have been prepared
covering square and cube functions, Table 8.1 and
Table 8.2.
If a square function is to be set up on the function
generator, switch on S6 (Voltage Source) and set VSI
for an output of -O'2V, then adjust VR I B for an OA)
output of O04V. Next set VS) for -O'5V and adjust
VR2B for an output of O'25V, and so on, according to
Table 8.1. After application of Eln == - 2'OV, and
adjustment of VR4, change the 100 kilohm computing
resistor in the feedback loop of OA I to )0 kilohm, to
prevent the amplifier overloading when Eln exceeds

v'1O.

565

1/; PLYWOOD

V; XiIi'

WOODEN BLOCK (6 OFF) 1/2" X

Fig. S.2. - Details and measurements of UNIT

generator case

"e" function

TABLE 1

Diode

E'n

I
2
3

-0-2V
-O-SV
-I-OV
{-I'SV
- 2-0V

."
5

6
'7

f'''

-3'SV
-"OV
-6-0V
-6'SV
-90V

Adjust
slope

VRI
VR2
VR3
VR4
VR6
VR8
VR-IO

Eo

Adiust
breakpoint

+- O-04V
+ 0-2SV
+ IOV
-+- 22SV
-+- 4 0V

VRS

-+- 0'62SV
+ 121SV
+ 1'6V
-+- 36V
+ "22SV
-+- alv

VR7
VR9
VRII

}
}

Eo = Ein'
R, = IOOkO

Eo = EJn'
10
RI = 10kO

COMPONENTS .
UNIT "c" BOX
Plywood 9tin X 4in X !in
Softwood tin X tin X 3tin
White plastics laminate 9iin
4tin X fin (2 off)
Rubber grommet !in X !Iirt
UNIT "c" Front Panel
White plastics laminate 9tin
2 yellow. 2 black. 2 blue.

SKS

SLOPE

~~~..~

SK8

Diodes
01 IB30 (2 off) (Radiospares)
02-07 OA202 (12 off)

FUNCTION GENERATOR

Fig. S.3.

generator
566

Front "anel 'ayout of UNIT

4in . Sockets: 2 rlid.

Pre-set ,Potentiometers
VRI. VR5
IMO
(4 off)
VR2. VR3. VR9 220kO (6 off)
VR4. VR6
IOOkO (4 off)
VR7
470kO (2 off)
VR8
47kO
(2 off)
VR I0
22kO
(2 off)
VRII. VRI2
47kO
(4 off)
All miniature horizontal mounting

__________ 4N __________

I in (2 off).

UNIT "C" Function Generator Components


Resistors
RI HMO (2 off)
R2 470kO (2 off)
R3 I MO (2 off)
All 10%. tW carbon composition

~1~~

-;.~1

function

Miscellaneous
S.R.B.P. 3iin X 3iin (2 off), Small turret tags
4mm stackable plugs. one -red. one blue
(Radios pares)

Z15~

3,;
0
0
0
0
0
0
0
0
0
0
0
0
0
0

OTT

OTT

OTT

OTT

0 OTT
0

OTT

OTT

OTT

OTT

OTT

0 TT0

00
0
00
0
00
0
0
0

0
0
0
0

OTT

OTT

I..

TT- TURRET TAGS

3 314"

-0

I
2
3
4

-0-3V
-0-5V
-0-75V
{-I-OV
-1 -25V
-I -SV
-2-0V

VRI
VR2
VR3

{-l-SV
-3-0V
-3-SV
-4-64V

VRB

6
7

VR4

0
0

OTT
(b) lJositive branch underside wiring

Adjust
b reakpoint

VRS
VR7

VR6
VR9
VRII

VRIO

(c) negative branch underside wiring _

Fig. 8.4. Fu~ctlon generator clrcu'ft lJanels (2 off)

TABLE 1.1
E'n

33'4N------;.~1
(a) drilling temf>late (2 off)

Diode

OTT
0

TTO 0

Adjust
slope

.........

OTT
O

OTT

Eo

+ 0{1l7V
+ O-I25V
+ 0 -421V
fo = f tn'
+ I-OV
Rt = IOOkO
+ 1-953V
+ H7SV
+ B-OV
+ 1I-S6V
Eo = E' n'
+
-7V
10
+ 4-lB7V
+ 10-oV J Rt = 10kO

}
1

TO

SKI & SKZ

TO

SK5 & SK6

4"

Interior view of UNIT

nc" function generator

Fig. 8.5. TOIJslde and Interconnecting wiring of function


.generator panels. _ Tile circuit boards are shown In pasltlon
inside the UNIT nc" eerse

567

">---4I>---oy

y=x 2 FOR-x
OR y=-{x 2) FOR +x

y=Ff'OR-x

fx=x2

OR

y=-(F) FOJl.+x

>-__- - O y
y=x 2 FORx

Rf 100kA

I1

y=x2+x FOR-x
OR
(x 2tx) FOR+x

y=-

[)>--+-1

ay

.=0 I I~>---+------<
:1
Ri.100kA

Rf l00kA

y= x 2 t a FOR-x AND-a
OR y= - (x Z + a) FOR + x

ay

AND + a

f(xJ=x2
y= (x+a)2 FOR + x
AND t a OR

y= _(Xh)2

FOR -x AND-a ALSO

y=x 2 + 2xa + a 2

y=x 2 _a 2 FOR-x AND


ALSO y=(x-a)(x+a)

+a

NO OUTPUT FOR + x AND-a

Yl=-(x2-3) AND
Y2= x 4 2 FOR + x
(x2-3 X x 1-9=x 23 +1-9)
NO OUTPUT FOR-x

f(x)=x 2

100kA

This photograph. shows PE.AC being 'used to solve simultaneous equations

After the entire range of input voltages listed in


Table 8.1 has been covered, return to Ein = -O2V
and go through the procedure again, to achieve optimum accuracy. The positive branch can be set up
for the same function as the negative branch by transferring patching leads from FGjSK5 to SKI, and
FGjSK8 to SK4, but this time trim VRI2A for zero-set,
and apply positive values of Ein. It may be necessary
to slightly re-adjust slope controls VRI-VR3 when the
two branches are connected in parallel, if there is some
small bias voltage imbalance.

THE FUNCTION GENERATOR IN


EQUATION SOLVING
The fact that an analogue computer can produce
and handle imaginary numbers will be particularly
evident when the function generator is applied' to
equati'o n solving, see Fig. 8.6. One type of function
generator circuit configuration will produce consistant
outputs for, say" the cube of a number, but not for its
square, or vice versa, because x 2 = + y, but
+ xs = + y , and - x s = - y . The computer operator
must therefore choose, or devise, the appropriate
. circuit for a given task.
Output y in Fig. 8.6a will be of the required sign
when the input is - x, but the sign of y with an input
of + x cannot be reconciled with mathematical convention. However, the circuit of Fig. 8.6a does
provide a consistant output when the function is
x S , with inputs of x. Much the same applies to the
Fig. 8.6b circuit, which shows the function generator
arranged for square root operations. Circuit Fig.
8.6c reverses the above situation and gives consistent
outputs for a square function, but not for a cube
function, by employing an extra sign reversing amplifier.
Getting away now from the complexities of square
roots of ~negative numbers and other mathematical
anomalies, Fig. 8.6d can be made to give outputs of
y = x 2 + x, or some other combination such as
y = X 2 . 5 - 3x, depending on the choice of function,
voltage polarities, and computing r.esistor values. The
purpose of other circuits E- H will be self-evident in
Fig. 8.6. Fig. 8.6i gives the symbolised layout for
solving a quadratic equation, where x is unknown and
a, b, and c are constants. The function generator can
also be introduced into problem set-ups where integrating amplifiers are used, as its frequency response is
well in excess of any frequency likely to be encountered.
Next month: The final item of the PEAC
equipment, UNIT "0". will be described.

&~&[L@@(W[g

PEAC'

@ U\':'iU [p (lDLr~ ~

ByD.BOLLEN

month's article deals with UNfT "OH-the


multiplier, which is the final piece of PEAC
equipment. After a technical description, details of
the' construction and setting, up are given.
HIS

The servo driven potentiometer has been widely


employed in the past for multiplication of one variable
voltage by another, but its frequency response, in most
cases, is seldom better than 0-5Hz. Modern analogue
computers now tend to use all solid-state multiplier
circuits, which have a frequency response extending
into the kHz region, but they are both complex and
expensi've. Taking the quarter-square multiplier as an
example, it needs five operational amplifiers and two
diode function generators to produce an accurate
product voltage from two inputs. It follows, therefore.'
that analogue multiplier circuit design can be expected
to present 'considerable difficulties when cost is an
important consideration .
UNIT "on-THE MULTIPLIER
Working on the premise that even a multiplier of
restricted performance can make a worthwhile contribution to an analogue computer which lacks such a
facility, an accuracy of :t.: 25 per cent and a frequency
response of 50Hz under the .most favourable conditions was considered to be an acceptable specification
for the UNlT uD" multiplier. Although 0-50Hz
seems rather limited by ordinary electronic standards,
in the context of "parallel" computer circuit operation
it represents a useful compute time which compares
favourably with the servo multiplier.
UNJT "D" contains three distinct circuits, two
operational amplifiers and a bistable reed relay driver.
One of the amplifiers is identical to those used with
UNIT "A", and is available as a multi-purpose
operational amplifier when the multiplier is not in
service.

640

TIME DIVISION
With the time division multiplier, a square wave is
modulated in sllch a way that the mark/space ratio is
proportional to one input voltage, while the amplitude
of .the waveform is proportional to another input
voltage. The mean value of the resulting waveform is
then proportional to the product of the two input
voltages.
Looking at Fig. 9.1, which sets out the simplified
multiplier circuit with associattod waveforms, a voltage
E2 is compared with a fixed voltage E3 at the input of the
integrating amplifier. A bistable relay is arranged to
switch ~I and S2 when the integrator output reaches a
pre-determined value, conveniently a\>out two thirds of
the maximum available amplifier output swing. . If the
sign of E3 at the SI contacts is corn.~ct, the feedback will
be positive, and a self-sustained oscillation at a frequency determjned mainly by E2 and er will result.
When E2 ,= 0 the output from the integrator will
consist of a saw tooth or symmetrical ramp waveform,
with identical rising and falling slopes, which is
generated by Ea.
'
Assume now that a voltage E2 is applied; . this will be
added to, or subtracted from E"a, depending on the
position of the Sl switch. The ramp waveform is
ther!!fore modified to an asymmetric form where tbe
rising and falling slopes become dependent on the level
and sign of '2'
.
Waveform (a) in Fig. 9.1 depicts the asymmetric
ramp for +E2 and -E2 , while waveform (b) shows the
square wave generated by the switch, of mark/space
dependent on the magnitude of E 2 As S2 is
synchronised with SI, so the input resistor RI will be
alternately switched to the inverting and non-inverting '
inputs of the product amplifier, and will remain at each
contact for a time dependent on the frequency and
mark/space of the switching waveform.
The , amplitude of the product amplifier output is

Rr

Rl

COMPONENTS

SZ

EO

UNIT "D" FRONT PANEl.. AND BOX


Potentiometers
VR2S 1000 wirewound
VR26 500 wirewound
(both panel mounting type)

NON-INVERTING
INPUT

Switches
S II
S 12

3 pole. 4 way rotary


Double-pole slide switch (c/o contacts)

Sockets
2 red. 2 blue. I black. 2 yellow. 3 white.
I green. and 6 miniature sockets

Miscellaneous
Material for front panel and box. Hardboard.
2 off 12fin x 4tin. 2 off 4tin x 3,\in.
White plastic laminate. 2 off 12fin X 4iin.
2 off liin x 4fin. I off 12in x 3iin .
Softwood. 2Sin x fin X t in. Knob. one
Radiospares Itin type PK with pointer.

-.

INTEGRATING
AMPLIFIER

UNIT "D" BISTABLE RELAY AND


PRODUCT AMPLIFIER
Resistors
RELAY
- - - OPERATES

+E2
i

I
I

--

:
RELAY
-r--OPERATES

JUl JlFtr::
I

I I

I I

MARK-SPACE DETERMINED BY MAGNITUDE OF 2

RI ' lkO
*R14 10kO 1%
R2
43kO
RIS IkO
R3
43kO
RI6 8200
R4
43kO
RI7 8200
RS
IkO
RI8 IkO
R6
1000
RI9 82kO
"R7
IlkO 1'0
R20 22kO
R8
10kO
R21 22kO
R9
27kO
R22 82kO
. "'R23 20002.
RIO 22kO
RII 1000
*R24 Ik02~ ~Q
*R2S 12kO 1%
"R12 10kO I ~!o
"'R13 9 lkO 1%
*R26 30001 %
(All 10 % 'f watt carbon composition except
"= I W metal oxide)

Potentiometers
VRI
VR2

100kO vertical skeleton pre-set


2200 miniature horizontal pre-set

Capacitors
Cl
C2
C3
C4
CS

I/-LF polyester 2S0V d.e.


02S/-LF polyester 2S0V d.c .
I/-LF elect. ISV
8/-LF elect. ISV
100/-LF elect. ISV

Transistors
TRI. TR2 2N2926 (orange) or 2N3904 (2 off)
TR3 2N3906
TR4 2N3904
TRS. TR6 ACY28 or AC 126 (2 off)

Diodes

rl

I I

I I

I I

I I

--+-t--j-I--I I
I I
LJ

I I

-+--f-r----t~--

L.J

r-,
I I

L_.J L_-l L_

(-E2) (-El)

--+~--t+-I I
I I
-EO
__ ...J L_-.J

L__

DI-D4

rl

(+E2) (+E1) --..., r--, r-- (-Ez) (+El)--'


I I
I I
+EO
I

r-.., r--, rI

I:

I I

_.J

LJ

LJ

-}-+!--~+I
I I
I I

MEAN D.C. LEVEL EXTRACTED BY FILTER

OA202 (4 off)

Choke
Lt

-EO

SH (Radiospares "Midget" type)

Reed coils
RLA. RLB Miniature triple 12V
Osmor type MTI2V (2 off)

+EO

Reed switches
RLAI. RLA2
RLBI. RLB2

Hamlin MRG2 20- 40AT (4 off)

Miscellaneous
S.R.B.P. I off 3in x 3iin. I off 3in x 4iin.
Small turret tags. Baseboard 12 in X 4in
s.r.b.p. or plastic; laminate

Fir. 9.1. Time division multiplier with associated waveforms


641

R6 loon

GREY
TO RLAl

CZ

R7 11kn

0' Z5pF

51Z
R8
10kA

R9
27kA

VRl

.BROWN
TO RLBl

TR3

. ZN3906

100kA
OUTPUT
GREY,

L1 5H

~
,,
,

I1

RI
IkA

1
1
1

~ :""<i"~

SKI

Cl

R4

IpF

43kn

RIO
ZZk1l.

r,---------i

et

I
I

R15
Ik1l.

OPERATIONAL AMPLIFIER

"

'

IZ ' 5V

11

BLUE

PRODUCT AMPLIFIER CIRCUIT 'PANEL


________
....:.:__________ ...11

RZO
Z2kA

RZI
Z2Ul.

R23
200n

RED- -1
tiNY

I
0

TRS

TR6

AC126

1
I

AC126

I
1

GREEN

YRZ6

"

"

,,','

r------ - ---- ---- - - - - -- - - - - -

I !
I :

RH
10011.

0-- -- . . .... -- .. -- -- ...... ...... -- . ... .. -- . . . . . j/


5K3

I :

'"

R5
Ikn

I
L
________________

5KZ
,
:

,
,

OV

L ___ _ ~~Ng .Q~RQh. -.J

I
I
I

RI9 ,
8'ZkA

R22
82tA

-12V .... - - - - - - - - - ,
UN5TABILI5ED
1 I
BLUE

I
I

-12'5Y

BLUE

'
BLUE

GREY:

BROWN
Rt+)

RH

10kn

I-----:.,-----....l
. RLAZ

YELLOW ...

t
I

. RLBZ .

YELLOW

Fig. 9.2.

Multiplier circuit, comprising product amplifier panel and bistable relay panel

wholly dependent on Eh but whatever the value of El it


wjJJ be divided by 10/ E2 (time division), which is the
same thing as (El x E 2 )/IO, assuming of course that
appropriate values for R I-R3, Rr and E3 are chosen.
Waveforms (c) shows what happens to different signs
of E l - and E 2 , in terms of the square wave. If now the
mean voltage level of the output from the product
642

amplifier is extracted by a suitable filter (see waveform


(d) ) it can be seen that four quadrant multiplication has
been achieved. When El and E2 are both positive, or
both negative, thy product voltage will be positive, but
when El and E~ are of opposite sign, the product
becomes negative.
The multiplier circuit will now be described.

rod

~Iu.

~
VBALANCE

~~==~r-~~ +===0~~=itt~~~~~==~~

:-

____

OPERATIONAL AMPLIFIER ..

e>

________

~
~

31~
'11

~'- ~Ie>~J
__

f - - - - - - - - - - - - - - - - - - - - - - - l z ' l - - - - - - - - - - - - - , - - - - - - - - - - - - - -,i1
Fig. 9.3. Dimensions and engraving details for UNIT "0" front flanel

UNIT "0" MULTIPLIER CIRCUIT


As the operational amplifier circuit has already been
given in connection with UNIT "A", it appears in
symQolised form only in the multiplier circuit of
Fig. 9.2, with VR26 as the front panel balance control,
and a fixed value of input resistor RI2 provided
internally for use with the multiplier. As the feedback
capacitor Cr only affects the integrator waveform
frequency, wjthout altering other multiplier characteristics, it is useful to leave it as a plug-in component,
,so that the multiplier carrier frequency can be adjusted
easily.
The output from the integrator, which it will \be
remembered from Fig. 9.1 carries information as to the
magnitude and sign of input E 2 , is fed via SllB to a
diode resistor network composed of D I, D2, R 15-R 18,
and VR2, the purpose of which is to allow the following
bistable relay driver to be switched at precisely determined voltage levels. VR2 establishes the working
point of the diode resistor network.
A conventional cross-coupled multivibrator is
utilised as a relay driver, with reed coils RLA and RLB
forming the respective collector loads of TR5 and TR6.
D3' and D4 are used to ensure a "cleaner" switching
action at high repetition rates, and the bistable circuit
will function satisfactorily at frequencies in excess of
100Hz without undue relay contact bounce. The
reference voltage, which was shown as Ea in Fig. 9.1,
is extracted from a resistor network R23-R26 and VR25
in Fig. 9.2. VR25 allows positive and negative values
of Ea to be made equal. E3 voltages are then fed, via
RLA2 and RLB2 switches, and resistor R 13, back to the
summing junction of the integrator, thus completing
the closed-loop to maintain oscillation.
SIGN CHANGE
The square wave switching cycle is presented to the
. input of the product amplifier by RLAl and RLBl,
with R14 acting as the input resistor. Changeover ,
switch S12 is included to allow the sign of the multiplier
output voltage to be changed to suit a particular
problem set-up.
A product amplifier open-loop gain of about 1,000,
which is the gain of the Fig. 9.2 circuit, is quite
satisfactory for good accuracy when working with a
fixed, closed-loop gain close to unity. Long-tailed pair
TRl and TR2 provide inverting and non-inverting
inputs, while TR3 is the output transistor, and TR4
forms a constant current load for TR3, in place of a
fixed resistor, thus enabling larger loads to be driven
without excessive dissipation. VRl serves to zero the
amplifier output.

The ratio of resistors R7 and R14 gives a product


' amplifier gain (closed-loop) of 1'1, while R13/RI2 yields
an equivalent gain for the integrating amplifier of 0'91.
The lower value of gain for the integrator enables E2 to
equal Ea without stopping the integration cycle, and yet
the overall gain of the multiplier is still unity because
I t' x 091 = 1.
FILTER CIRCUIT
The purpose of the filter circuit Lt, C2-C5, R6, and
SI lA, is to remove the square wave carrier without
distorting the product waveform when input voltages
are time varying. Bearing in mind that computer
waveforms are extremely diverse; it is almost impossible
to achieve near perfect results with one filter circuit,
especially when the carrier frequency is not far removed
, from input frequencies. To allow compromise, therefore, the cut-off frequency of the Fig. 9.2 filter can be set
by switch SI lA to suit the circumstances of a particular
problem set-up.
'
The three switch positions, 1Hz, 10Hz, and 50Hz,
represent approximately the roll-off points given by the
filter, and the bandwidth handled by the multiplier. In
the 1Hz position the filter will virtually eliminate carrier
ripple when input voltages are of very low frequency, but
the 50Hz setting is used with fast integrator waveform
inputs, where ripple may be less objectionable.
CONSTRUCTION OF UNIT "0" FRONT
PANEL AND BOX
Details of the UNIT "D" front panel and box appear
in Fig. 9.3 and Fig. 9.4. Note that the operational
amplifier (OA4) socket positions and panel markings

General ,,/ew of
, , the multiplier
Gnem"/y

643

WOOD (ALL
6 OFF

"i X "2') .

4Y8 LONG

HARDBOARD PANELS (lfeNTHICK)


2 OFF

1~/:

4YZ

2 OFF

4"2

3346
Fig. 9.4.

Construction of the box for UNIT "0"

are the same as for UNIT "A" operational amplifiers.


SIl, VR2S, VR26, and al\ sockets may be mounted
after the front panel has been marked and dr~led.
INTERNAL LAYOUT OF THE MULTIPLIER

The internal layout and interconnecting wiring of the


multiplier are shown in Fig. 9.5. Operational amplifier,
bistable relay driver, and product amplifier circuit
panels are bolted with stand-off spacers to a 12in x 4in
s.r.b.p. or plastics laminate baseboard, which rests on
the wooden bearers at the base of the UNIT "D" box.
Component placement positions for the bistable relay
circuit panel, and the product amplifier panel, also
appear in Fig. 9.5, together with a rear view of the front
panel assembly. The operational amplifier (OA4) is
made up in accordance with instructions given in the
May issue of PRACTICAL ELECTRONICS (pages 209-210).
BISTABLE RELAY CIRCUIT
CONSTRUCTION

Drill the bistable relay circuit panel according to


Fig. 9.6, and insert turret tags. Then mount alI
components and complete underside wiring, leaving the
reed switches RLAl, RLA2, RLBl, and RLB2 until

Close-up
of the bistable
relay panel

644

last. A triple reed coil is specified for the Fig. 9.2


circuit, to allow the addition of an extra. pair of reed
switches if the multiplier is to be enlarged to cater for
three input voltages; this modification will, of course,
also involve the construction of another product
amplifier.
PRODUCT AMPLIFIER CIRCUIT
CONSTRUCTION

Drilling details and underside wiring of the product


amplifier panel appear in Fig. 9.7. Accurate matching
of input transistors TRI and TR2 may not be necessary
with this low gain circuit. A 2N2926 transistor should
not be employed in the TR4 position~ in place of the
2N3904, as its maximum Vce will be exceeded.
After inserting turret tags, mount resistors and
transistors first, then follow with Lt, and capacitors
C2-CS. Cl is soldered into position last of all, across
the amplifier input turret tags, as shown in Fig. 9.5.
FINAL ASSEMBLY AND SETTING UP OF
UNIT "0"

Mount the three circuit panels on the baseboard and


complete all interconnecting wiring between the circuit
panels and the front panel, including S12 which can be
left floating for the time being. The resulting assembly
can be set-up and tested out of its box.
Connect red, green, and blue flexible wires from the
bistable relay panel to the UNIT "A" power supply
solder tags, or alternatively to TLl, TL2, and TL3 with
stackable plugs.
Place SII in the "off" position and zero-set the
operational amplifier (OA4) folIowing instructions given
earlier for UNIT "A" amplifiers, after allowing the
usual warm-up period. When adjusting the VR26
balance control connect M/SK2 to any earth socket
with a patching lead. . Next, attach a sensitive d.c.
voltmeter (0-1 V) to M/SK3 and zero-set the multiplier
output by adjustment of VR I on the product amplifier
circuit panel.

Fig. 9.5.

x
oTT

00

0
00

00
0

OTT
OTT
0

0
0

OTT
OTT
o TT

0
00

TTO

0
0

OTT
OTT

~+:;".",.,.,..;:~
0--..0 no 0

MOUNTING
HOLE

M~UNTlNG~
HOLE
no
o

0
0

000
0

0
00

00

T~~

JTO

OTT

-T- fT/,

.1

TT

TT

;;--0

TT

TT

No.46 DRILL TO TAKE


SMAll TURRET TAGS

o--FIXI~;R HL~LES---o

000

0
0
0

of UNIT "0" multiplier


MOUNTING HOLE

.X

0
______
OO::..:..TT.:.....J_

3~"

Internal layout and wiring

0
0

000
0

0
0

TT

TT
0

f--------~-

0
000

_ o--MOUNTING HOLE

000

TT TT
0
0

TT
0

Fig. 9.6 (far left). Top and


underside views of bistable
relay panel

.i

4Vi.

Fig. 9.7 (left). Top and


underside views of product
amplifier circuit panel

o
o

___
645

Insert a O 25,uF capacitor into OA4/SK 11 and SK 12,


and switch Sll to 10Hz. A "buzz" from the relays
should now be heard, which mayor may not sound
errati~. Transfer the d.c. voltmeter to OA4 output
while the relays are still working and adjust VR2 on the
bistab1e relay panel for zero volts; this should produce
an even note from the relays. Return the voltmeter
lead to the multiplier output M/SK3 and this time zeroset with VR25.
Apply an input of + 5V to M/SK2; the relay "buzz"
will drop in frequency, but no output should be
observed at M/SK3. Transfer the + 5V patching lead
to M/SKI and again no output should be ~een . Finally,
apply + 5V to both inputs, M/SKI and SK2., to produce
a multiplier output of 52 /10 or 25V.
Throw switch S 12 to change output polarity and
experiment with inputs of differing sign. If all is well,
the product voltage should retain its value of 25 for any
sign combination of input voltages and S12.

For best accuracy it is advisable to go over all


adjustments again to obtain optimum settings, and also
verify that the multiplier will handle a fulLrange of input
voltages.
Due to the fact that the power supply may be working
close to its maximum current limit, there could be some
fall-off in multiplier accuracy because of switching
transients, this can be checked by employing the extra
current facility, SI in Fig. 3.1. . The optional -t2V
. relay power supply should obviate the difficulty if it
occurs. ,
To use the operational amplifier. (OA4) on its own,
merely switch Stt to the "off" position and patch the
amplifier sockets in the normal way.
Next month: The final article in the PEAC series.
This will complete the operational details of
UNIT "0", and will give some examples of
special circuits to represent mechanical phenomena, and some general notes.

We now consider the use of the multiplier UNIT "D';


in solving equations.
THE MULTIPLIER IN EQUATION
SOLVING

Fig. lO.1 sets out four multiplier configurations to


show how equation terms may be handled . As a selfcontained computing element, UNIT "D" will mUltiply
input voltages X and Y to give a product XY/lO, see
Fig. lO.la. Note that arrows are normally used with the
multiplier symbol to identify input and output terminals.
Division of two variable voltages is achieved, in
Fig. lO.1 b, by placing the multiplier in the feedback loop
of an operational amplifier. However, with division,
certain limitations are imposed. The Y input must be
of single polarity, which rules out a.c. waveforms unless
they are d .c. biased above or below Y = 0, but ramp or
step functions will be accepted if they do not change

their sign. With the X input, voltages can be 0 to


tOY d.c., or a.c. peak.
Because an extra filter capacitor (shown dotted in
Fig. lO. lb) is needed to prevent amplification oflow-level
carrier ripple by the open-loop, high gain amplifier,
frequency response is restricted to lOHz for the division
operation, when switch Sll is in the 50Hzposition. It
is sometimes possible to arrange a problem so that the
reciprocal is multiplied,. and thus avoid the limitations
of Fig. lO.lb division. A related configuration in
Fig. lO.1c gives an output XY/(l + X), for inputs of
Xand Y.
In the final example of Fig. IO.Id, the multiplier is
combined with integrators, and therefore handles time
varying voltages.
By
solving
the
equation
dA/dt ~ 21TR x dR/dt, which describes the rate at
which the area of a circle changes with a growth of
radius, the layout of Fig. lO.ld can be used to investigate,

&~&[L@@(1D[g

@~~(UJlY~~
PEAC
By
D.BOLLEN

The Practical l:/ectronlcl


Ana/Olue Computer In Its
complete and comprehenll'le
form. The whole of thl,
eqUipment hal been fully
delcrlbed In chi, lerl.. of
artlclel which Is concluded
this month

720

say, the build-up of tape on a spool, the expansion or


contraction of metal discs and cylinders when heated, or
the surface area of a liquid in a conical reservoir.

SPECIAL ANALOGUE COMPUTER


CIRCUITS
'
Apart from the analogue computing elements already
covered are a few specialised diode circuits which are
used for simulating various mechanical phenomena.
Ordinary silicon' diodes, such as the OA202, can be
employed with the circuits of Fig. 10.2, and are inserted
into the computing component sockets of UNIT "A".
Dead Zone. Amplifier gain in Fig. 10.2a is zero until
the limits
Ein =

RI

-' -

RBI

x 10

or
Ein =

R2 x 10

RB2

are reached, thereafter gain will depend on the slope


given by Rr/RI and Rr/R2.
Limiter. In Fig. 1O.2b, amplifier gain is constant
between the limits set by
, RI
Eo= x lO
RBI

and

Eo

R2 x 10

RB2

When the limits are exceeded, the gain falls to zero.


Friction. A frictional force generated by moving ,
surfaces in contact is virtually constant for all values of

Z=
OR

xy
10

Z=.J.
10

'
WHEN Y=X

VALID FOR x AND Y INPUTS


MAXIMUM FREQUENCY RESPONSE 50Hz

MULTIPLICATION OF TWO VARIABLES

Z=_(I~X)
VALID FOR x BUT Y RESTRICTED
100kll
100kA

TO EITHER + Y OR - Y, DEPENDING
ON S12 SETTING.

025I'F
r-~~-l

ALSO' Z--(jIOX) WHEN Y=X

THEN X RESTRICTED TO +XOR-X

MAXIMUM FREQUENCY
RESPONSE 10Hz

DIVISION OF TWO VARIABLES


Z=~

1+X
X2
OR Z~ I+X WHEN Y=X

VALID FOR x AND Y


MAXIMUM FREQUENCY
RESPONSE 50 Hz

100kll

SPECIAL CASE OF COMBINED MULTIPLICATION AND DIVISION


dA

"it =

dR

2'1T,R X dT

VALID FOR

. J.tl
...'A=

dR
dt

0 \27TR

dR) d t
XdT

COMPUTE TIME;. 50ms

721

...L-....lI.-+-~-...,...+in

Rr

tanOGZ=,RZ

-0

-Ein

+in
Rr

tanOG=-

Rin

+Fr

SIGN OF Fr DEPENDENT

~X 10

ON SIGN OF : :

RB1

dX

+dX

-V

Fr

dt

FRICTIONAL FORCE -Fr


VELOCITY-g
dt

R8Z

X10
-Fr

WHENX<XS FC=O

+FC

WHEN X> Xs FC = -(X-Xs)


DISPLACEMENT X
STOP SITUATED AT Xs
CONSTRAINING FORCE FC

X7XS
I

-X

+X

FC

-FC

-in--~'---+---\'---+in

ADJUST K1 AND KZ FOR BACKLASH


ADJUST er FOR HYSTERESIS

-0

+0

-in - - - - l l f - - + - - - - + in
I

in=-.

velocity, but will change sign when the direction of the


velo~ityisreversed. CiicuitFig.1O.2c satisfies the above
conditions and generates a voltage proportional to a
frictional force Fr.
Elastic stop. When an object makes contact with an
elastic stop, the resulting constraining force is proportional to the penetration of the object into the stop.
In Fig. 1O.2d, term Xs represents the position of the
elastic stop, while X is the displacement of the object.
When X;;;. X s , the amplifier provides an output Fe
which represents the constraining force.
Backlash and hysteresis. Mechanical linkages, gear
trains, and some electrical circuits will often exhibit
backlash and hysteresis, which are simulated by the
circuit of Fig. lO.2e, using a dead zone and an integrator.
Apart from Kt. K 2 , and Cr, adjustments to R2, R3, and
R4 will allow a wide range of characteristics.
Comparator. As its name suggests,'the comparator of
Fig. 1O.2f compares one voltage with another, and
enables some action to be taken at a pre-arranged input
level. The comparator can be applied to the simulation
of impatt forces, where the constraining force is
proportional to the rate of penetration; when
Ein = - Ec, the relay contacts will close and insert a
voltage representing velocity into an equation.
CONCLUDING NOTES
A brief mention should be made of those aspects of
analogue computer usage which were considered to be
beyond the scope of the present series. It would have
been difficult to include the more complex Calculus
problems which PEAC is capable of solving, and also
transfer function techniques were avoided because they
would have demanded some knowledge of Laplace
transforms and the like.
A very important field is the use of analogue computers in controlling processes and evaluating data, so
called "In-plant" applications, but here fairly elaborate
sensing equipment and servomechanisms are called for,
to act as intermediaries between the external process and
the computer.
An important omission, brought to light by a reader's
letter, concerns the use of a temporary feedback
resistor when checking the coefficient of a potentiometer

Product amplifier circuit pane'

which is employed for division (Fig. 4.lf). If the feedback resistor is not present, the operational amplifier
summing junction will no longer be at virtual earth
when the potentiometer is disconnected for measurement purposes, and this can "lead to serious errors.
Therefore, when checking a division potentiometer
coefficient, always insert a 10 kilohm feedback resistor
jnto OA/SKII and SKI2.
If difficulty is experienced in zero-setting a UNIT "A"
operational amplifier after construction, by adjustment
of VRI on the amplifier panel, it may be that transistor
"spreads" are greater than has been allowed for in the
design. The simple cure is to increase RI (Fig. 3.7) to
47 megohm if the amplifier output is fixed close to the
negative supply rail voltage, or, when the output
remains clamped near to the positive rail, decrease
RI to 33 megohm.

UNIT "0" front #HInel


arranlement and cab'net

723

You might also like