The computer “CSIRAC” (1965)

(Dr Frank Hirst) Trevor,
how did it come that you designed the machine
with 16 D Registers? (Dr Trevor Pearcey) Well we
first mastered the problems associated with a
single register and then realised that we could put in 16 words into one adding unit thereby multiplying the
capacity of the adder. (Dr Frank Hirst) Yes, it
seems a pity really that more modern machines haven’t
got quite so many D Registers or index registers as CSIRAC. (Dr Trevor Pearcey)
Yes, this is so. Only recently when designers have been aiming at running
more than one program at a time have the number of registers
increased markedly. (Dr Frank Hirst) This
is the computer CSIRAC. It was the first fully automatic
electronic digital computer to be built in Australia. It was constructed in the
Radiophysics Division of the Commonwealth Scientific and
Industrial Research Organisation to the designs of Mr Trevor
Pearcey and Mr Maston Beard. Mr Pearcey was responsible for the
logical design of the circuitry and Mr Beard engineered the
electronic components. This computer is of advanced electronic
design for its day and it has a very
flexible command code. We are fortunate to have
Mr Pearcey with us now, its designer, and Trevor, perhaps you could
relate to us some of the ideas that you had when
you designed the machine. (Dr Trevor Pearcey) The automatic
digital computer which… CSIRAC was a very early example, was initially the product of the pooling of ideas
by the mathematician and the electronic engineer who brought the ideas of the
mathematician to physical realisation. Three basic principles are involved
in the automatic digital computer. One, that the fast
computer must be provided with a sufficient internal store so as to be able to
hold its program that is the sequence of operations
which it is instructed to perform. The second is that
data and program are formally identical and are in fact held
within the same store. The third is that program
must consist of a network of sequences of instructions and that the computer traverse this network in a manner which is
determined by the partial answers. In this way the
computer is given the facility for discriminating
between differing conditions, the conditions being those
which it had already computed and to thereby
repeat the program frequently but in slightly
different form. Soon after World War II the need arose in the
Division of Radiophysics of what was then CSIR, for both a more rapid
computing capability and for continued development of the electronic
pulse techniques which had been developed for
radar during World War II. Digital computing it was seen
would serve both these purposes. In 1948 the division undertook the study of automatic
digital computing and in 1951 CSIRAC, this machine, was actually exhibited
publicly in operating order and has since then been
in regular service for the best part of a
period of 13 years. During some of this time
it has been improved and during the last
nine years of its life has been used as a teaching
and research machine in the University of Melbourne. The design, although using engineering methods, which have
now been rendered obsolete by the invention
of the transistor, concentrated upon
logical functions which would render
it easy to use and some of them have been
incorporated in machines to this day. A set of 20 binary digits tells the machine how to move one particular datum from one part of the
machine to another at the same time carrying
out a simple operation upon it such as addition. The program, which it
performs, consists of a set of such simple transfers of data with appropriate transformations during their passage. The main store of CSIRAC and most of the
incidental registers consists of a number of
acoustic delay lines. These take the form
usually of pipes containing mercury, each
one about 5 feet long down which acoustic waves travel taking a time of about
one millisecond. When they reach the far end of
the pipe they are detected, amplified and recirculated to
the starting point. By this means CSIRAC was able to store 756 items of data, that is, six decimal digit
items, or the equivalent, and to be able to
operate upon them a thousand times a second. This is indeed
slow compared with 500,000 operations a second
which are now on current machines. Its main medium for
accepting programs and data is paper tape of two kinds, one a wide kind with 12 channels frequently used for
recording programs and one of five channels identical with paper tape used
in common telegraphic equipment. Program and data are coded automatically
by special devices with typewriter type of keyboard and the machine can
read these tapes at about 100 rows per second or twenty 5 decimal digit numbers. Output onto similar
tape is slower at about six 5 decimal digit numbers
over the equivalent and these are printed out after punching at a later stage on similar keyboard instruments. CSIRAC has now been
rendered obsolete by recent developments
in electronics and problems have
grown too large to be held within it and too time-consuming to run. Problems are common now which can only be
performed on equipment of vastly greater
size and speed, computing and storage capacities being at least 500
times that of CSIRAC with correspondingly fast
input and output devices. These are all now
currently available and are being installed
throughout Australia. Here you see a program being recorded on
wide paper tape. An instruction written by the programmer in a
fairly simple computer language on his sheets of paper is transcribed
through the keyboard and you will notice that
two keys are depressed before a punching
action takes place. An instruction then
consists of two groups of ten binary digits. Thank you. The program has now been
put onto paper tape and we will put it
onto the reader. For this purpose… for the purpose of
this description there is no data
on the paper tape. This tape reads the
12 holes row by row and is operated
photoelectrically. (Dr Frank Hirst) We’re going to
feed the tape now into CSIRAC and I press the appropriate
control buttons on the console and the tape is inched forward until we position it into the
buffer register at the right spot. Now the bootstrap
tape is going in and the tape is moving
into the reader. You can actually see
it go into the memory filling up the cells
of the memory. By switching on the console keyboard we can see the tape in position in store. It’s loaded in the memory
and we now set the data for the problem and I am setting
this on the keyboard registers. This is the number that has
to be fed to the program. I fed that number
into the machine and now the next number goes in and I start the button here and the calculation takes
place and you’ll see the results coming out on
punched paper tape from the paper tape
punch at this stage. Now I can see inside
the memory tubes and watch the arithmetic
registers in action. You can see the counting being done in the D Registers
and you can see the accumulator calculating, adding and
subtracting and so on. This array of cathode
ray oscilloscopes is not evident on
more modern machines. Because of the older
type of machine these display tubes were present so
one could do program testing. Now we’re putting the results from that calculation
into the Flexowriter which is going to print
out the results for us and we feed it into the
reading device here and we press the start
read here and the tape will run into the Flexowriter
and the printing of the codes on the
tape now takes place. This is a loan
repayment schedule. The loan is being amortised over several years. This
is the principal outstanding at the beginning
of the first quarter. This column shows the interest
at 6% payable for the quarter and this is the amount
being paid off the loan. As the loan schedule goes
of course the interest becomes less each quarter and
the repayment is greater. And this is then done by the machine
in a few minutes but on a desk machine of course
it would take quite a long time. The program stays in the machine and
a different loan can be calculated by just pressing the next
parameters in for the loan amount. Now we’re coming to
the final payment and the machine will add
up the complete total of all the interest paid in the first column and the complete amount of
money paid off the loan and this of course balances with the
outstanding amount of the first quarter. (Dr Trevor Pearcey) I have said
that CSIRAC was easy to use. Let me illustrate by mentioning
a few points of its design. From the operator’s
point of view the display of the operations was
comprehensive and convenient. The state of the store and
the arithmetical registers was shown as arrays
of spots or traces on small cathode ray tubes and the state of
the control system was made visible as rows of
lights on the panels in front of the control console. A switchboard provided facilities for manual
control of a program and for insertion
of requisite data while the program was running other than the data, which was
provided on the punched paper tapes. (Dr Frank Hirst) Well that
is the story of CSIRAC. This machine which is
still in operation is perhaps the oldest
at present in the world and it is fitting that this
machine is to be stored in the Applied Science section
of the National Museum. It will be a historic exhibit and lots of people in the
future should gain much information about early
days in computing from the presence of
CSIRAC in the museum. We are very pleased
that it’s going there because this machine has been
used for hundreds of computations in research projects
and been used to teach many students in the
University of Melbourne.

Leave a Reply

Your email address will not be published. Required fields are marked *