Australia’s first Computers : SILLIAC

In 1953, the newly appointed and very dynamic Chair of the School of Physics at the University of Sydney Professor Harry Messel wanted to take Physics teaching and research into the areas of Nuclear Physics and Cosmic Ray physics. He appointed a significant group of lecturers and researchers led by Dr. John Blatt from the University of Illinois. Blatt argued to Messel that his work in the physics of the atomic nucleus and Messel’s own work in Cosmic Rays would produce large amounts of data and would require complex mathematics to handle and thus an electronic computer would be required for the efficient processing of all this data. To raise the funding for the ambitious program that he initiated, Messel established the Nuclear Research Foundation (NRF), a philanthropic body of local businessmen and organisations.¹ The Foundation contributed the operating costs of the teaching program and the fundamental research work, but extra funding would be needed to finance the computer.

Messel used his contacts with the American Atomic Energy Commission (AAEC) chairman Paul McDaniel to obtain the blueprints for the AAEC computer, a derivative of John von Neumann’s “IAS Computer”,² being developed at the University of Illinois and known as the ILLIAC³ with which Blatt had already had experience.⁴ At the same time, working with the general enthusiasm he had created for physics with the NRF, Messel began a campaign to raise funding for the computer. He did not receive much official support, but one member of the Foundation, Mr Alexander Armstrong, a grazier and member (MLC) of the NSW Government, said he knew someone who might be interested. A luncheon was arranged and Armstrong introduced Messel to a Mr. Adolph Basser, a local businessman who owned the Saunders Jewellery store. Basser must have been very impressed by Messel’s explanation of what a computer was. He owned a racehorse which had won the 1951 Melbourne Cup and at a meeting on February 12, 1954, just before the inaugural meeting of the Foundation, he donated the winnings (£50,000) to Messel’s NRF. He was made a governor of the NRF, and subsequently accorded an honorary Doctor of Science degree from the University, for this donation which was used to set up the Adolph Basser Computing Laboratory in the School of Physics and fund the construction of the SILLIAC.

The CSIR Mk1 was still in Sydney at the time, and in what became the Madsen Building just up the road from the School of Physics. However, there seems to have been some distaste among the School of Physics people for the CSIR Mk1 group. This is probably largely due to the fact that not only was the Mk1 almost entirely taken up with Radiophysics Division and Aeronautical Research Laboratories projects, it was also unfinished and still somewhat unreliable. Another difficulty was Pearcey’s allegedly difficult personality and a general lack of helpful attitude.⁵ As Messel had indicated in an address to the Nuffield Foundation while still searching for funds:

“Ours will not be the first electronic brain in Australia. The C.S.I.R.O. at Sydney has an electronic computing machine, the Mk1, in operation. However, in the words of the builders of the Mk1, “the computer is not intended to provide a computing service but to provide experience for the design of a final model”. We do need a practical computing service, and hence we propose to get our own machine.”⁶

A further, and in many ways far more important issue was that the Mk1, being a serial machine was just too slow to be useful for the amount of work Blatt and Messel were contemplating and the Sydney version of the ILLIAC (thus SILLIAC), being a parallel machine, would be an order of magnitude faster.⁷

So the computer project was on the way, but it needed an engineer to look after its construction, and a numerical analyst (early computerese for a programmer/systems analyst) to continue the development of programs for it. Messel and Blatt hired two engineers, both with experience on the CSIR Mk1, John Algie to be assistant engineer and Brain Swire to be chief engineer. Swire was sent to Chicago to get experience on the ILLIAC. Meanwhile Standard Telephone and Cable (STC) were contracted to build SILLIAC. The original ILLIAC design was slightly modified for its Sydney functions and the engineering was supervised by Swire. It was built in Sydney, the component chassis assemblies and power supplies were done in the STC factory and the frames and final assembly within the Basser Computing Laboratory. [Figs 2 – 4] Messel then lured John Bennett — an Australian mathematician who had worked for Maurice Wilkes at Cambridge on the development of the software libraries for the EDSAC and was currently working for Ferranti on the logical organisation of the Mark I* — back home. He arrived in the Computing Laboratory in February 1956 to became the Senior Numerical Analyst for SILLIAC and to run the Laboratory for many years to come.

The machine was completed [Fig.4] substantially ahead of time, running its system diagnostics program (known as Leapfrog⁸) without failure on July 4th 1956. The School of Physics’ Theory Group, led by John Blatt, then carried out the first calculations on SILLIAC. The code was checked on the afternoon of the 4th July and the morning of the 5th July. The first “production run” began at 4:30pm that afternoon and took half an hour. The computer calculated, in the space of that half hour, what would have otherwise taken them a year to do. The resulting data points were hand-plotted to graph paper the next morning. The work was a theoretical calculation on the theory of liquid Helium superfluidity and showed very close correlation with experimental data.⁹

Like all computers, even Babbage’s analytical engine, SILLIAC had a set of necessary modules to carry out its functions. These were the:

• Arithmetic Unit – the equivalent of a calculator. SILLIAC, like all modern digital computers, stored numbers in a binary format, using electronic devices which are able to remain in either of two states, on or off. On is generally represented as a “one” and off as a “zero”. The states of truth or falsity of propositions in logic can be similarly represented, with true being a “one” and false being a “zero”. So SILLIAC, like any other digital computer, was able to carry out logical as well as numerical operations. The Arithmetic Unit could perform additions, subtractions – by complementing the number and adding, division – by successive left shifts or divisions by two, multiplication – successive right shift or multiplications by two, and a clear or reset to zero of the accumulator register. Multiplications and divisions by an odd number were achieved by a series of successive left or right shifts in the accumulator followed by an add or subtract of the original number. Logical operations were made up from basic arithmetic operations as needed.
• Memory – SILLIAC’s memory consisted in 1024 (2¹⁰) words each of 40 bits. That is, it had 1024 locations, or addresses, in memory each of which was 40 bits deep. The words were stored on 40 CRTs of the type known as Williams-Kilburn Tubes, which were an electrostatic memory that stored a “bit” as a spot of brightness stimulated by the effect of the electron beam in the tube on the phosphor coating on the inside of its faceplate. As the phosphor could only hold the electrons for a short period before they leaked away, the memory depended on the continual refresh or regeneration of the spot by a special circuit that had to refresh the spot at least once every tenth of a second (and preferably many more times than that). This kind of memory is very similar in use to dynamic ram in contemporary computers, though implemented in vacuum tube technology rather than monolithic silicon. However, as distinct from the acoustic delay-line type of memory the Williams-Kilburn tube can directly represent a bit-map image if used as a display. [For more detail see the description of the Williams-Kilburn tube in the section on Display Devices.]
• Input/output – Information was transferred to and from SILLIAC in the form of a coded pattern of holes punched in a paper-tape.¹⁰ The paper tape was prepared from a program by punching the tape with the pattern of holes and placing it in a reader under the automatic control of the machine. On completion of a program the machine would then punch a tape with a pattern of holes representing the results. The output tape could then be read and converted to text with the aid of (what was then) a standard Post Office teletypewriter. In 1959 a machine for converting punched cards of the Hollerith type into paper-tape and for converting the paper-tape to punched cards was added.¹¹ Over 1958-9 a magnetic tape “backing store” was developed and subsequently incorporated into the machine for the interim storage of numbers being used in a calculation and for the longer term storage of programs, data, and results which would be regularly used in research or other work carried out on the machine.¹²
• Internal buss – the means by which data were transferred to and from various parts of the machine.
• Sequence control – the machine would proceed to execute instructions (“orders” as they were known to the early programmers) in sequential order as they were drawn from memory and placed on the buss. An “order counter” in the control unit of the machine would contain the location of the next instruction to be executed. In a very simple program, the orders were carried out until the procedure reached a conclusion and a “Stop” instruction was executed. Here the value in the Order Counter would be incremented after each order was carried out. However, in most programs numerous iterations of some section of the program would need to be carried out, eg, in many large calculations, and a conditional transfer of control, or branch, order would be carried out. There were two conditions under which these transfers took place. The first was whether the sign of the number in the A register (accumulator) was positive or zero, in which case a branch would occur, or if negative no branch would occur. The second condition was whether or not a calculation had produced an overflow in the A register. With the execution of a conditional transfer the new location of the next instruction would be directly entered into the “Order Counter” register.¹³

SILLIAC was on display well before its official opening on September 12, 1956. For example it was shown off during the University Open Day of July 22, 1956 when games like NIM (fiddlesticks) and Noughts and Crosses were played on it. It was officially dedicated – “switched on” – by Sir John Northcott, the then Governor of NSW, on the 12th September 1956, in a ceremony conducted between the Computation Laboratory and an adjacent lecture theatre. Formal proceedings began at 2:30pm. Prior to the arrival of the Governor “the audience was presented with a televised description of SILLIAC and its laboratory” being taken on a conducted tour of the machine by Barry de Ferranti (an assistant engineer). The audience were shown how a problem was prepared and run and could listen to the sounds of the machine as it did its calculation. The closed circuit TV presentation was produced with the ABC’s newly acquired Outside Broadcast Unit. [Fig.5]

The dedication of SILLIAC took place at 3:30pm: the official party moved into the Adolph Basser Computing Laboratory to perform the “switch on” ceremony, while the ceremony was relayed to the remainder of the audience via the closed-circuit TV system. Adolph Basser set the “first” tape in the tape reader and the Governor started the machine which then printed out:

THANK YOU, YOUR EXCELLENCY FOR DEDICATING ME TODAY,
AND THANK YOU, DR. BASSER, FOR MAKING ME POSSIBLE.
SILLIAC

Messel then flicked a switch on the control panel and the machine displayed the words “Welcome to SILLIAC” as large characters in glowing green dots on its monitor CRT. The guests were could then play the “day of the week of your birthday” game and NIM as well as hearing SILLIAC play “Happy Birthday to you”.¹⁴

———————————————————-

FOOTNOTES

1 McCaughan, J.B.T. (1987) “The Era Begins”, chapter 2 of Millar, D.D. (ed.) (1987) The Messel Era. The story of the School of Physics and its Science Foundation within the University of Sydney, 1952-1987, Sydney: Pergamon Press, for The Science Foundation, University of Sydney, and Pearcey, 1988, op cit, p.30.
2 For the genealogy of the IAS machines see Deane, John, (2003a) The IAS Family Scrapbook, Killara, NSW: Australian Computer Museum Society.
3 McCaughan, 1987, op cit, p.16.
4 Deane, 2003, op cit, p.22.
5 Deane, 2003, op cit, p.23.
6 From a draft of a letter from Messel to the Nuffield Foundation, January 1954 contained in the University of Sydney Archives, Box 54, and quoted in Deane, 2003, opcit, p.26.
7 Pearcey, 1988, op cit, p.30.
8 Leapfrog was designed to copy itself into the next available sector of memory, then leap into that version of itself and do the process again until all memory locations had been tested. This process put all of the machine to the test. We shall come back to Leapfrog later. See Deane, 2003, opcit, p.44-5.
9 Deane, 2003, op cit, p.44-5. These calculations, which were intended to compare results derived from theory with actual experimental results, produced a major breakthrough in the theory of Superfluidity in liquid Helium. The results were published very shortly afterwards in Butler, S.T., Blatt, J.M. and Schafroth, M.R., “Nature of the l-Transition in Liquid Helium” Il Nuovo Cimento, vol.4, p.674 on the 1^st September 1956 and caused a major stir in the scientific world with Blatt and Butler travelling to the US to give talks about the work. It was also reported with great pride in The Nucleus, vol.2, no.5, p.9, where Barry de Ferranti describes this calculation as being of immense importance in physics.
10 SILLIAC Programming Manual, A Guide to the Preparation of Problems for Solution by the Automatic Digital Computer in the Adolph Basser Computing Laboratory within the School of Physics of the University of Sydney. The Adolph Basser Computing Laboratory, School of Physics, The University of Sydney, 2nd edition, Sydney, 1959, pp.1-3. The first edition was a copy of the University of Illinois’ ILLIAC Programming Manual.
11 The Nucleus, vol.5, no.1, p.17.
12 Deane, 2003, opcit, p.65 and p.70; and The Nucleus, vol.5, no’s 1 & 2.
13 See the SILLIAC Programming Manual or Deane, 2003b, op cit, for much more detail.
14 The Nucleus, vol.2, no.5, September 1956, pp.4-5, “Opening Day”.

---

Return to Contents