Australia’s first Computers : CSIRAC

There were two major cycles of computer development in Australia.1 The first produced the fifth stored program computer in the world. It was originally known as the CSIR Mk1 and was later renamed the CSIRAC. Of the first generation of machines it is the only one that is still intact, though no longer working, and may be seen in the Museum of Melbourne.

This first Australian machine was built during an intense period of development in calculating machines that took place at the CSIR’s Division of Radiophysics and Division of Mathematical Instruments. Although it was built as an experimental machine, it was expected to do much of the calculation work required in the research program of the Radiophysics Division. The second cycle was the development of the SILLIAC, which is the subject of another chapter.

CSIR Mark I

In late 1945 Trevor Pearcey came from the UK, via the US, to the CSIR Division of Radiophysics, then in the National Standards Laboratory building on the grounds of Sydney University.2 He had been appointed to do mathematical work on the propagation of radio waves in the analysis of the formation of clouds and rain, and in the design of antennas for radio-astronomy, for the division. He had worked on microwave radar at the Air Defence Research Development Establishment (ADRDE) in Malvern, UK, and continued to work on the use of radar in various aspects of the analysis of rainfall (under the head of the Radiophysics Division, E.G. Bowen3)

Pearcey’s work mainly involved large-scale calculations and he was familiar with the electro-mechanical analogue computing devices of the time4 having worked with Hartree, and having used the Hartree differential analyser at Manchester University, in the UK. He had also become familiar (through association with Leslie Comrie) with punched card techniques used in big accounting and statistical systems of the day. On his way to Australia he travelled through the US and visited the Harvard Computation Laboratory where he met Howard Aiken and saw the Harvard Mk I digital calculator. He also saw Vannevar Bush’s Differential Analyser at MIT. He did not see ENIAC as it had not been announced yet and he new nothing of the Bletchley Park work (though he must have at least heard of Turing’s theoretical work). However, he was familiar with the work of the Admiralty Research Establishment, at Haslemere, UK, on the use of acoustic delay lines for the storage of pulses in enhancing radar echoes.5

In April 1946, Pearcey had already produced preliminary proposals for a high-speed computing machine. He was influenced by the publication of a paper by McCulloch and Pitts on switching devices and nerve action6 and by what he saw of the Harvard Mark I.7 At Radiophysics, a wide range of calculation techniques proved necessary and various devices based on modified punched-card equipment and relay controlled calculators were developed for specific calculation problems. Pearcey became more interested in the development of computing techniques and, with his engineering colleague Maston Beard, produced a set of logical designs for an electronic digital computer.8 As the division was also involved in the development of electronic computing components (an outgrowth of its work on radar and aircraft distance measuring equipment) it became clear that in order to test out all these devices and Pearcey’s logical designs a computing machine would become necessary.9 Thus was born the CSIR Mk1 project. Maston Beard was in charge of the engineering development and Pearcey did the logical design, aided by Geoff Hill.

By 1948 the designs for the Mk1 were complete enough for the Radiophysics group to start building the machine. The design provided for a simple machine of relatively low numerical precision which made it easy to program. As it was based on Pearcey’s own conceptual designs, it was not particularly similar to any of the existing overseas machines, though Pearcey did return to the UK in late 1948 to look at work being done by others there. He visited the laboratories where EDSAC (Cambridge under Wilkes), the PilotACE (NPL) and the Manchester Mark I were being developed but eventually did not alter any of his own logical designs for the CSIR Mark I.10

CSIRAC organisation
Fig 1: Logical organisation of CSIR Mk.I

The CSIR Mk1 would be assembled from 2000 valves, and was organised around a serial “Digit trunk” (or data buss) over which data and instructions were transferred from the input paper-tape reader to memory and from there to either the instruction register for interpreting (executing) or as data to the arithmetic unit. The instruction in the instruction register would control which parts of the machine (input, memory, logical, arithmetic and output modules) were to be connected to which, to feed data around the machine via the digit trunk so that it could be acted upon, stored or presented as output.11 Apart from the instruction and the sequence (or next instruction) registers there were a set of arithmetic registers (A, B, C and H) that allowed Addition, Subtraction, Multiplication, Division, AND, XOR and NAND functions, 16 additional accumulators (the D register) for temporary data storage and the up to 1024 words of memory that were added to the machine over its operational lifetime. [Fig.1]

The memory storage was of the delay line type (as used in the ACE) because it was already established technology and Pearcey was already familiar with it. Moreover, in 1947 and early 1948, the electrostatic storage that Williams and Kilburn were developing, although being incorporated into the Manchester SSEM, was not yet fully proven and was not announced in the literature that Pearcey would have had access to until 1949. For the Mk1, Reg Ryan designed a memory system of 32 mercury-filled monel-metal tubes which would function as acoustic delay-lines. The contents of each delay-line had to be constantly recirculated to maintain data integrity. This recirculating serial memory of 16 20-bit words (giving 320 “bits” per delay line)12 was later doubled by interleaving another 16 words into the original set giving 1024 20-bit words of serial memory in the 32 delaylines.13 Programs running on the machine had to wait for the appropriate address to cycle through in order to access any particular memory location (or word), giving an operating speed of about 500 instructions per second.

The machine ran its first program in 1949 but was not really fully operational until 1951, when it received its first public airing at the Conference on Automatic Computing Machines held in Sydney during August 1951.14 It was used for program development in numerical analysis and for data reduction in the scientific work of the Division of Radiophysics. [Fig.2]

CSIRAC with Pearcey
Fig 2: Shows the Mk I in an early state with Trevor Pearcey at the Flexowriter. (CSIRO photo # 2923 – dated 5 Nov 1952, Melbourne University CSIRAC archive photo #163, Courtesy CSIRO)

Input and output in the original version of the machine were handled by a Hollerith card reader and card punch (by British Tabulating Machines). A “Flexowriter” character printer was used to print out the results of calculations that the Mk I output onto punched cards. A 12-channel paper-tape reader and paper-tape punch were incorporated during 1952-3, replacing the punched cards.15 A magnetic drum storage device, designed by Brian Cooper,16 was installed in late 195217

System monitoring and maintenance were done with CRT monitoring equipment consisting in three CRTs added by Ryan in 1952.18 Because the CSIR Mark I program and data memory used acoustic delay lines, the memory contents had to be continually recirculated to maintain signal amplitude and thus data integrity. It was possible to take a tap off this re-circulation circuit and display the sequence of bits in memory as a trace across the CRT screen. The contents of each succeeding 20-bit word of a memory delay-line could then be lined up one below the other with a suitable vertical triggering ramp circuit. This produced a 20-bit by 16-line display which could at least display lines and characters as used for the noughts and crosses game or for indicating the reaction time in the reaction-time testing game. Although the main use of the CRT displays was for program diagnostics it was also a very good indicator of the state of completion of iterative mathematical routines such as those used in the solution of polynomials. [Fig.3]

CSIRAC with Geoff Hill
Fig 5: Geoff Hill watching the newly installed monitoring CRTs as he sets input values with the switch panel. (CSIRO Photograph 2772-2)

The diagnostic system was extended when the switch panel that was used to start up the machine was incorporated into a proper desk console19 with six 2 inch (50 mm) diameter CRT displays offering monitoring of the arithmetic registers, the contents of the memory delay-line currently in use and a switchable selected delay-line memory display. Additionally, the machine was provided with a speaker so that it could alert the user to programmed events in the program sequence. The presence of the speaker encouraged some of the programmers (particularly Geoff Hill) to produce musical output. By inserting an instruction to write to a particular destination register, to which the speaker was attached, subroutine loops could be made to run at a rate that produced pulses at an audible frequency. With careful setting up of the loops many musical tones could be programmed and Geoff Hill is reported to have programmed a version of Colonel Bogey that was played to the delegates at the first Australian Conference on Automatic Computing Machines, held at the CSIRO Division of Radiophysics, March 1951. The machine was used for making music irregularly over its life. Professor Tom Cherry who was the head of the Department of Mathematics in Melbourne University wrote a program that would allow anyone to produce a musical sequence with CSIRAC by punching the appropriate codes into a “pianola tape” version of the usual punch-tape input. The music was coded with a row of digits indicating pitch (5 digits) and duration of the pitch (the next 5 digits) and a separate row of intensity values.20

———————————————————-

FOOTNOTES

1 The first generation of digital computers in Australia are covered in
Pearcey, Trevor (1988) A History of Australian Computing, Melbourne: Chisholm Institute of Technology;
Bennett, J.M., Broomham, R., Murton, P.M., Pearcey, T. and Rutledge, R.W. (1994) Computing in Australia: The Development of a Profession, Hale & Iremonger in association with the Australian Computer Society Inc., Sydney;
Deane, John (1997) CSIRAC: Australia’s First Computer, Killara, NSW: The Australian Computer Museum Society;
Deane, John, (2003) SILLIAC: Vacuum tube supercomputer, Australian Computer Museum Society, Sydney, Australia;
McCann, D. & Thorne, P. (2000) The Last of the First – CSIRAC: Australia’s First Computer, Melbourne: Dept of Computer Science and Software Engineering, University of Melbourne.

2 This building later became the Madsen Building and the home of the Basser Department of Computer Science of the University of Sydney.
3 Bowen had been one of the leaders of the scientific groups researching radar in the UK.
4 Pearcey, T., (1994) “Australia Enters the Computer Age” in Bennett, Broomham, Murton, Pearcey and Rutledge, 1994, op cit, p.16.
5 Pearcey, 1988, op cit.
6 McCulloch, W. and Pitts, W. (1943) “A Logical Calculus of the Ideas Immanent in Nervous Activity” Bull. Math. Biophysics 5. Also in McCulloch, W. (1965) Embodiements of Mind, Cambridge, Mass.: M.I.T. Press.
7 In a note dated 23 April 1946, Pearcey made some suggestions for designing a Computing Machine as follows (in summary): from Radiophysics archives now in the Australian Archives – file number A1/9 Proposal For Design of High Speed Computor.
Objective.
High speed calculator is needed for the rapid solution of differential equations which will be needed for investigations into meteorological phenomena. The equations will be of the type occurring in potential theory, and in hydrodynamic motion. They will therefore be partial differential equations.
High speed calculation is also required for the rapid tabulation of certain mathematical functions and for the rapid subtabulation of mathematical tables.
The major requirement is for a sequence controlled machine to solve partial differential equations, and this could most easily be done by an economic combination of electric, automatic telephone tape and card elements.
8 Pearcey, T. and Beard, M., (1948) The Logical Basis of High-Speed Computer Design. C.S.I.R.O. Radiophysics report RPR 84; June, 1948. and
Pearcey, T. and Beard, M., (1948) The Organization of a Preliminary High-Speed Computer. C.S.I.R.O. Radiophysics report RPR 84; June, 1948.
9 Pearcey, 1994, op cit, p.18. In a memo dated 1 December, 1947, Bowen (chief of Radiophysics) writes to Cook (Secretary [what we would now call CEO] of CSIR) arguing for the acquisition of punched card machines for the Division and, in paragraph 3,  foreshadowing a proposal to develop a new machine.
3. Contribution to Computer Programme
The Divisions of Electrotechnology and Radiophysics are engaged on a collaborative programme of development of electronic aids to mathematical calculation. We are regarding our part of the work as a development programme which may lead to the design and construction of a new type of calculating machine in a year or two’s time. This investigation would benefit considerably from having alongside a calculating machine in practical use. Furthermore, considerable saving in time and money might result from incorporation of some of the techniques of punched card equipment in the new design. from Radiophysics archives now in the Australian Archives – file number A1/9
10 McCann, & Thorne, 2000, op cit, p.2.
11 Deane, 1997, op cit.
12 Ryan R.D., (1954) “A Mercury Delay-Line Memory Unit” Proceedings Institute of Radio Engineers, Australia, vol.15, 1954, pp.89-95.
13 Pearcey, 1994, opcit, p19.
14 Proceedings of a Conference on Automatic Computing Machines, CSIRO Division of Radiophysics, March 1951.
15 CSIRO Annual Report 1952/53, 30 June 1953, p.150.
16 Cooper, B.F.C. (1951) “A Magnetic Drum Digital Storage System” in Proceedings of a Conference on Automatic Computing Machines, CSIRO Division of Radiophysics, March 1951, and Cooper, B.F.C. (1953) “A Magnetic Drum Digital Storage System” Proceedings of the IRE of Australia, vol.14, 1953, pp.1-9.
17 McCann & Thorne, 2000, opcit, p.3.
18 Ryan R.D., (1954) “Electronic Computer Test and Monitor Equipment” CSIRO Division of Radiophysics Report no. 128, April 1954.
19 It is difficult to know when this occurred but the CSIRO Annual Report for June 30, 1952 mentions the intention to build a control desk in the next year, 1952-53. It was probably later not earlier.
20 Doornbusch, Paul (n.d. 2001?) The Music of CSIRAC Australia’s First Computer, available at and
Doornbusch, Paul (2004) “Computer Sound Synthesis in 1951: The Music of CSIRAC” Computer Music Journal, vol.28, no.1 pp.10-25.
Doornbusch, Paul (2005) The Music of CSIRAC: Australia’s First Computer Music, Melbourne: The Humanities, Common Ground Publishing.


Return to Contents

<span>%d</span> bloggers like this:
search previous next tag category expand menu location phone mail time cart zoom edit close