Return to On Line Documents
Return to "Early English Computers" table of contents
|Early British Computers, 5 - THE ACE, THE 'BRITISH NATIONAL COMPUTER', starting page 023|
At the close of the second world war there was some feeling amongst scientists at the Ministry of Supply that a National Mathematical Laboratory should be established to coordinate facilities and techniques relating to machine-aided computation. The practical outcome of this feeling was the establishment in the summer of 1945, of a Mathematics Division in the National Physical Laboratory (NPL) at Teddington, Middlesex. Amongst the father-figures associated with this event were Professor D. R. Hartree of Manchester University and (from October 1946) Cambridge University, and Dr L. J. Comrie, founder of the Scientific Computing Service in London. Hartree was the chief link between British and American computing efforts in the immediate post-war years; both he and Comrie had accumulated a great deal of experience on mechanical and electro-mechanical calculators.
As far as building a stored-program computer, the initial enthusiasm came largely from a group of people who had been involved with the COLOSSUS deciphering activity at Bletchley Park. In particular, amongst the Bletchley team which disbanded in the autumn of 1945 were the mathematicians Professor Max Newman and Dr Alan Turing, and the Post Office engineers T. H. Flowers and Dr A. W. M. Coombs. In October 1945 Newman moved to Manchester University, where he wished to set up a 'calculating machine laboratory'. 9 His plan was to construct a stored-program computer similar to one being proposed by John von Neumann of Princeton University, using a special storage device called the Selectron tube. As is explained in Chapter 16, von Neumann's group at Princeton grew out of the American ENIAC development at the Moore School of Electrical Engineering, University of Pennsylvania, Philadelphia. This group was the source of one of the
Early British Computers, 5 - THE ACE, THE 'BRITISH NATIONAL COMPUTER', starting page 024
first formal proposals for a stored-program computer - the 'EDVAC report'.10 The Selectron tube, under development by the Radio Corporation of America from about 1945, was for a time thought to be the most promising digital storage device. Max Newman's plans at Manchester were in the end overtaken by events: the Selectron ran into prolonged technical difficulties and the Princeton computer was not working until 1952, with a different storage mechanism; meanwhile a completely independent computer had been built by the Electrical Engineering Department at Manchester - as described in Chapter 7. (In passing it should be said that, despite the relatively late completion of the Princeton project, there is little doubt that John von Neumann himself was the most influential of all the early computer pioneers.)
Fig. 5.1 Map showing the location of some centres of early computer activity. The National Physical Laboratory is at Teddington, the Telecommunications Research Establishment at Malvern, and the wartime Code and Cipher School at Bletchley.
Early British Computers, 5 - THE ACE, THE 'BRITISH NATIONAL COMPUTER', starting page 025
Of the other Bletchley personnel, Turing joined the new Mathematics Division at NPL where, with characteristic energy, he immediately set about designing a universal computer. Turing had written an important theoretical paper on computers in 193611 and, although familiar through personal contact with von Neumann, he had no need or inclination to copy anyone else's design. On 19 February 1946 he presented to the Executive Committee of NPL, probably the first complete design for an electronic stored-program computer,12, 13 including a cost estimate of E11,200. It is likely that Sir Charles Darwin, the NPL Director, thought of Turing's proposal for an Automatic Computing Engine (ACE) in terms of a single national effort which would result in a computer housed at NPL and serving the needs of the whole country. At that time the NPL establishment at Teddington did not readily have the capability for constructing a large national computer and Darwin was therefore anxious to enlist the cooperation under contract of other relevant organisations such as the Post Office.
It is clear that at the Post Office's Dollis Hill research laboratory both Flowers and Coombs, ex-Bletchley men, had the necessary interest and expertise to implement a stored-program computer, and indeed the immediate post-war intention was for the Post Office to do just that. Flowers had visited the Moore School, University of Pennsylvania - the main centre for US computer activity - late in 1945. He has since said, 'Unfortunately the pressure of telephone reconstruction after the war left so little effort for other projects that eventually the commitment [to build a computer] had to be withdrawn. Some mercury delay lines were constructed but little else.' These particular lines resulted from a development contract placed by NPL in June 1946; a prototype 1000-bit delay line was working at Dollis Hill by 21 January 1947.14 Another Post Office commitment, to the Telecommunications Research Establishment (TRE) at Malvern, involved the construction of a computer called MOSAIC for defence applications; this spanned the period 1947-54 and is described in Chapter 10.
Early British Computers, 5 - THE ACE, THE 'BRITISH NATIONAL COMPUTER', starting page 026
Having failed to secure the help of the Post Office in building a national computer, Darwin and the NPL Mathematics Division turned their attention in the autumn of 194( to two new British groups which had emerged as likely sources of computer electronics expertise. The first was a small team led by Dr (later Professor Sir) F. C. Williams at TRE; the second was a group led by Mr. (later Professor) Maurice Wilkes at Cambridge University. Williams had been working since July 1946 on a novel binary storage system using conventional cathode ray tubes. Wilkes had attended a series of lectures on the American EDVAC proposal for a stored-program computer, held in August 1946 at the Moore School, University of Pennsylvania, and had returned determined to build a similar machine at Cambridge.
Fig. 5.2 Some British computer pioneers and other distinguished guests at the opening of the Science Museum's computing gallery in London, December 1975. Back row, left to right: Donald Davies, Tommy Flowers, Grace Hopper (USA), Jim Wilkinson, Tom Kilburn, Raymond Thompson, Maurice Wilkes, Cecil Marks, Allen Coombs. Front row: Mrs Douglas Hartree, Freddie Williams, Max Newman, David Wheeler, Konrad Zuse (Germany).
As a result of Darwin's approach to THE a party led by Dr R. A. Smith (superintendent, THE Physics Division) and including Williams and Dr A. M. Uttley, visited NPL on 22 November 1946. A report of this meeting survives15 and is worth quoting from because it illustrates the scarcity of expertise at that time:
Early British Computers, 5 - THE ACE, THE 'BRITISH NATIONAL COMPUTER', starting page 027
The director [Sir Charles Darwin] emphasised the extreme importance which he attached to the development of A.C.E. and put it as having the highest priority in his opinion of any work that was being done for D.S.I.R. at T.R.E. [D.S.I.R. was the Department of Scientific and Industrial Research, a predecessor of the present Science Research Council.] He was most anxious that some effort should be set aside for work on this project. Dr Smith explained the difficulties in which T.R.E. found itself at the moment as regards staff knowledgeable in electronic circuit technique. Apart from the small number of staff now working for Dr F. C. Williams most of the able circuit technicians had been transferred to the Department of Atomic Energy. Dr F. C. Williams had himself been appointed to a professorship in the University of Manchester and was leaving in six weeks. Dr A. M. Uttley had staff knowledgeable in computing technique but not expert in valve circuits. These staff were, however, almost completely tied up with important work for the Ministry of Supply on computers for military applications and it was unlikely that much effort would be available from that source. It therefore seemed likely that the only way in which T.R.E. could continue to contribute would be to second a small number of staff, say one Scientific Officer and an assistant, to work under Dr Williams' direction at Manchester University, and for a small team to be constructed from the remainder of his staff, and possibly one or more from Dr Uttley's present staff, under the direction of Dr Uttley.
The purpose of the larger meeting was mainly directed towards elucidating the present position of the development of the A.C.E. It appeared that although an elaborate paper design had been laid down, the fundamental problem of storage of information has not been solved and that, as had been suspected, the experimental work of Dr Williams's at T.R.E. on storing information on a cathode ray tube was considerably in advance of the work which the Post Office were doing on the use of delay lines for storage purposes. There was therefore necessary a considerable amount of basic investigation on storage systems before the computer could actually be brought to the stage of being assembled. Switching techniques were also discussed and it appeared that although the knowledge existed for making electronic switching systems, a good deal of work would have to be done before suitable practical designs were evolved. This knowledge, however, mainly existed with Dr F. C. Williams and with the group in T.R.E. engaged on the development of counters for the Department of Atomic Energy.
Early British Computers, 5 - THE ACE, THE 'BRITISH NATIONAL COMPUTER', starting page 028
Fig. 5.3 Time-chart showing the date on which some of the vintage British stored-program computers became operational. Details of these and many other machines are given in later chapters.
Dr Uttley's group at THE went on to build TREAC,16 a parallel computer using Williams tube storage and completed by mid-1953 (see Chapter 10). Williams, who had an outstanding reputation as an inventor, left THE for Manchester University in December 1946. He took with him Tom Kilburn, who was later to become the driving force behind all subsequent Manchester computer development. As for helping with the ACE project, an NPL proposal and draft contract was sent to Manchester in December 1946 but was declined shortly afterwards by Williams. The reasons for Williams' lack of interest were threefold:
- there was at that time no need for Williams to seek additional financial support for his own work;
- the wording of the NPL draft contract appeared restrictive compared with normal academic practice; and
- Turing's ACE design, and the personality of Turing himself, were incompatible with Williams' own way of pursuing research.
Early British Computers, 5 - THE ACE, THE 'BRITISH NATIONAL COMPUTER', starting page 029
NPL's interaction with Cambridge University was spread over a somewhat longer period, beginning in the autumn of 1945 when Wilkes had been released from wartime work at THE to take up the Directorship of the newly formed University Mathematical Laboratory. An impression of the initial Cambridge activity may be had from the following quotation from a letter by Wilkes to J. R. Womersley, Superintendent of NPL Mathematics Division, dated 2 April 1946:
We are just getting organised and I want to decide what our research programme on new calculating machines shall be. I would like to have a discussion with you so that anything we do may be coordinated with your own activities. You will probably have suggestions to make as to how we can best apply our limited efforts.All these negotiations left Darwin and the concept of a British national computer somewhat at a loss. A new Electronics Division was set up at NPL in August 1947 and it was here that the Pilot Model of the ACE computer was built, eventually running its first program on 10 May 1950; this development is described in Chapter 8. The summer of 1947 also marked Turing's effective departure from NPL. He spent a sabbatical year at his old Cambridge College and then moved, in the autumn of 1948, to the Mathematics Department at Manchester University, run by his friend Max Newman. Turing is remembered as a truly brilliant mathematician who many found very difficult to work with. It is
I am now rather more in the picture as to American activities than when I last discussed the subject with you. Professor Hartree was good enough to write to Philadelphia and get permission to tell me officially about the American projects and he gave me a copy of the report he wrote after his last visit to the States.
Discussions concerning a formal contract between Cambridge and NPL gained momentum in November 1946, by which time Wilkes' group had initiated the building of what was to become the Cambridge EDSAC stored-program computer. By April 1947 the idea of an NPL/Cambridge contract had been dropped, for much the same reasons as had applied in the case of Williams. Wilkes went on to complete EDSAC by May 1949 (see Chapter 6).
Early British Computers, 5 - THE ACE, THE 'BRITISH NATIONAL COMPUTER', starting page 030
significant that his original ACE proposal was radically different from any other contemporary computer design and that, although spending time at Cambridge and Manchester, he did not materially contribute to the design of either of those Universities' machines. At Manchester he eventually devoted most of his energies to developing early programming techniques and solving partial differential equations on the computer in connection with his theory of morphogenesis, the 'growth and form of living things'. He died suddenly in June 1954.
Any possibility of a 'British national computer project' had evaporated by 1947. If personalities and resources had been otherwise, it is quite likely that such a computer would have been completed by the Post Office or NPL, acting independently or jointly, by 1948. Even if this had occurred, it is however most unlikely that the austere post-war economic climate would have permitted any immediate industrial exploitation. It should be remembered that wartime petrol rationing did not end until May 1950, and food rationing of one form or another lasted until July 1954. There were problems enough without the scientific oddity of a universal computer!
In order to preserve the chronological sense of the story, we must now leave NPL pondering Turing's ACE design, and move the spotlight to the Universities.
|Early British Computers, 6 - THE CAMBRIDGE EDSAC, starting page 031|
Maurice Wilkes' Cambridge University group was perhaps the most coordinated computer design team anywhere in 1947. It Included W. Renwick, S. Barton and G. Stevens on the hardware side and later D. J. Wheeler on the programming side. Their objective was to set up a usable and reliable computing service (for university research workers) in a short timescale; they were therefore not necessarily interested in building the 'best possible machine'.17 Their stored-program computer became known as EDSAC (Electronic Delay Storage Automatic Calculator) and, as the nomenclature implies, its design was influenced by the American proposals" for a machine called EDVAC (Electronic Discrete Variable Automatic Calculator).
Fig. 6.1 Part of the Cambridge University EDSAC computer during construction in 1947. The photograph shows the following members of the design team (left to right): G. J. Stevens, J. Bennett, S. A. Barton, P. Farmer, M. V. Wilkes, W. Renwick, R. Piggott. Professor Wilkes is kneeling beside a mercury delay line assembly.The development of a suitable storage unit was the principal problem facing all the early computer designers. It was mainly because of storage difficulties that the American EDVAC did not see the light of day until 1952. In Wilkes' case he chose mercury delay lines for EDSAC because encouraging results had been achieved with this technique towards the end of the war by the Admiralty and others, for storing pulses in a device designed to improve the tactical clarity of radar displays. It was fortunate that a research physicist at Cambridge, T. Gold, had worked on delay lines at the Admiralty Signals Establishment and was able to provide the EDSAC team with accurate constructional details.18 In EDSAC each mercury-filled tube or 'tank' was about 5 ft long and stored 576 binary digits. The main store consisted of 32 such tubes, with additional tubes acting as central registers within the processor.19
Early British Computers, 6 - THE CAMBRIDGE EDSAC, starting page 032
Fig. 6.2 The completed Cambridge EDSAC in 1949. This famous computer quickly settled down to provide an efficient computing service and helped to put Britain well ahead in the implementation of practical computers and their software.
Early British Computers, 6 - THE CAMBRIDGE EDSAC, starting page 033
Fig. 6.3 The direct commercial descendant of the Cambridge EDSAC was a computer named LEO (Lyons Electronic Office) shown here in about 1953. LEO pioneered the use of computers for business data processing, as described further in Chapter 13.
A full technical description of EDSAC is given in Appendix 2, together with a comparison between it and other contemporary computers. Briefly, EDSAC stored 512 36-bit words (a 'word' being a general term for a collection of bits treated as a whole by the computer). The time taken to perform an addition instruction was 1.4 milliseconds. Input was via a 5-bit electro-mechanical paper tape reader and output was to a teleprinter. EDSAC contained about 3000 thermionic valves and filled a large room. The financing of the project was via normal University research channels, plus a (for those days) sizeable donation of £2500 from J. Lyons & Co. Ltd. (The connection between this company and Cambridge is an interesting one, to be described later.)
EDSAC first ran a program on 6 May 1949, and offered a regular computing service from early 1950 until its shut-down in July 1958. This was the first such service in the world to be run on a proper stored-program computer. The EDSAC group wished to make life easy for the inexperienced user, so programming was in terms of an elementary symbolic assembly language, whereby the programmer wrote out his instructions in terms of meaningful alphabetic characters. These were
Early British Computers, 6 - THE CAMBRIDGE EDSAC, starting page 034
punched on paper tape, read into EDSAC, and then converted automatically to the basic binary machine instructions. An example of the resulting convenience to the user is given in Chapter 15, where four early programming systems are compared. A significant feature of the Cambridge computing service was the availability of a well- stocked library of standard subroutines held on paper tape. Whilst not the only advocates of subroutine libraries, the Cambridge group were nevertheless very influential in setting standards for their use, and in pioneering programming techniques in general. The group's book The preparation of programmer for an electronic digital computer (Wilkes, Wheeler and Gill, 1951) was the first textbook on programming a stored-program computer. Stan Gill was at Cambridge from 1949 to 1954.
The active encouragement of computer development in Britain was not only a characteristic of Douglas Hartree but also of Maurice Wilkes and the Cambridge Mathematical Laboratory as a whole. A major international computer conference was organised at Cambridge in June 1949, and a regular series of computing seminars was later initiated by the Mathematical Laboratory. These well-remembered seminars ensured that almost all the British computer teams were in frequent contact with each other. Wilkes and the EDSAC group were thus of considerable inspiration to other pioneers, as well as being outstanding pioneers in their own right. Knowing they had something worth selling, they even made a film of how to use EDSAC - surely a world first! Stills from this epic may be seen in reference 18.
On a more general front, in 1951 Wilkes was the first to expound clearly the principle of microprogram control as a basis for machine design (see Appendix 1, Section C). Microprogramming has had a major influence on the implementation of small and medium-range computers.
In addition to the work on programming and subroutine libraries, an interest in computer design was maintained at Cambridge after 1949. Engineering modifications and enhancements to EDSAC were carried out from time to time until the mid-1950s. A new computer called EDSAC II was then conceived, this becoming operational in 1957. Some technical details of EDSAC II are given in Appendix 2, from which it can be seen that this machine had a very respectable turn of speed.
Early British Computers, 6 - THE CAMBRIDGE EDSAC, starting page 035
Regarding hardware design, the main influence of the original EDSAC on other projects lay in the close informal association with the J. Lyons catering company. Early in 1947, T. R. Thompson of Lyons had begun to investigate the possibility of designing computers for use in offices. Contact between J. Lyons & Co. and Wilkes' team was first suggested by Dr H. H. Goldstine (of Princeton), and the cooperation became a reality in the autumn of 1947. From 1949 onwards, T. R. Thompson and J. M. M. Pinkerton supervised the construction of the LEO (Lyons Electronic Office), which was a re-engineered version of EDSAC with the same instruction set. Construction of LEO units was contracted out to other firms such as Wayne Kerr Laboratories Ltd, and the Coventry Gauge and Tool Co. The 'LEO Chronicle' kept by T. R. Thompson indicates that LEO was able to run simple test programs in the spring of 1951 and was doing regular clerical jobs for J. Lyons by November of the same year. Armed with an entirely new input/output system which at first included the provision of magnetic tape storage, LEO was performing a full business data processing service by the end of 1953. In addition, various scientific jobs were being undertaken for outside users, such as the Ministry of Supply and de Havilland Propellers Ltd. By mid-1954, Pinkerton was submitting proposals for a LEO II and on 4 November 1954 a new company, LEO Computers Ltd, was founded. The wider implications of these industrial developments are mentioned again in Chapter 13. Meanwhile we should look at an independent line of research which was being pursued in Manchester.
|Early British Computers, 7 - THE MANCHESTER MARK I, starting page 036|
As has been noted, F. C. Williams and Tom Kilburn arrived at Manchester University in December 1946, with the intention of developing a novel form of computer storage using conventional cathode ray tubes (CRTs). This became known as the Williams tube. Bits of information were actually stored as very small areas of electronic charge, put on the phosphor-coated screen of the CRT by a controlled beam of electrons. Since the charge would leak away in about a fifth of a second, the pattern of information had to be continually refreshed. The success of Williams tubes was based upon the discovery of a relatively simple method for effecting this regeneration. Williams tubes could, moreover, be built comparatively cheaply from standard components and the ideas were soon taken up by other computer groups in the USA and elsewhere. The significant advantage of this memory over other contemporary storage systems was that Williams tubes allowed random access to word locations, as opposed to the sequential access mechanism inherent in delay-line stores.
Back at Manchester, 1947 was spent perfecting the storage system, Kilburn publishing the results together with the outline design for a hypothetical computer in December of that year.7 The team was joined by G. C. Tootill from TRE, and a very small prototype computer was then built round a 32X32-bit word Williams tube store, in order to subject the new memory system to the 'most searching tests possible'. Though small, this machine contained all the elements of a stored-program computer but with manual input from a keyboard and output to a monitoring display screen. This prototype ran a 52 minute factoring program on the morning of 21 June 1948,9 thus becoming the world's first stored-program computer to operate.
Early British Computers, 7 - THE MANCHESTER MARK I, starting page 037
Encouraged by this success, the Manchester group doubled its size in September 1948 by taking on two research students (D. B. G. Edwards and G. E. Thomas), and began to expand the computer into a useful facility. An initial stimulus arose from some problems in number theory which had been suggested by Professor Max Newman. The first realistic problem to be solved, an investigation into Mersenne prime numbers, was run in early April 1949, by which time several improvements had been added to the machine. In fact the engineers' enthusiasm for improvements tended to outweigh any call for a stable computing environment, and it is significant that 42 computer patents emanated from Manchester during the period 1948-50. Although continual enhancement obviously affected reliability, an overnight error-free computing run of 9 hours was recorded on 16/17 June 1949.
In an age when acronyms were popular, the Manchester Mark I was sometimes referred to as MADM (Manchester Automatic Digital Machine) or MUC (Manchester University Computer). However, the designers usually refer to it simply as the Mark I. A technical description of the various stages of Mark I development between June 1948 and October 1949 is given in Appendix 2. In summary, the Mark I contained storage for 128 40-bit words in fast Williams tube storage, backed up by 1024 words on a slower magnetic drum store. An addition instruction was performed in 1.8 milliseconds. Input and output was via a 5-bit paper tape reader and a teleprinter. The Mark I was mostly built out of war surplus thermionic valves supplied by TRE. The cost of the project was therefore small.
Both the geographical location of Manchester and the personal inclination of Williams and Kilburn lent a certain independence to the research, with the team 'inventing things as the need arose'. In the wider context, two aspects of the Mark I design stand out. The first is the Manchester invention of index registers, a feature now seen on every modern computer (see Appendix 1, Section C). The second interesting aspect of the Mark I was the early combination of a small, but fast, random-access store backed by a slower (but larger capacity) sequential store. Observations on the information flow between these two levels were to produce significant Manchester inventions incorporated in the ATLAS computer, as mentioned in Chapter 9.
Early British Computers, 7 - THE MANCHESTER MARK I, starting page 038
Fig. 7.1 Part of the Manchester University Mark I computer as extended in 1949. The prototype, operational in June 1948, occupied the centre six racks in the photograph and is believed to be the world's first stored-program computer. The Mark I was built out of war-surplus components with an enthusiasm that left little time for tidyness! Incredibly, it worked for long enough at a time for some useful computation to be carried out. It also contained some novel ideas, for example index registers.
The University Mark I was dismantled in August 1950 to make room for further computer developments. One of the last users was Dr D. G. Prinz of Ferranti Ltd, who computed Laguerre functions in connection with the control of guided weapons.
The credit for ensuring that the Mark I's light was not hidden under an academic bushel probably goes to Professor P. M. S. Blackett (later Lord Blackett), an influential scientist and government adviser. He suggested that Sir Ben Lockspeiser, the then government Chief Scientist,
Early British Computers, 7 - THE MANCHESTER MARK I, starting page 039
should see the prototype computer whilst paying an informal visit to Blackett in October 1948. G. C. Tootill, Kilburn's colleague at the time, was able to give Lockspeiser a convincing demonstration of the Mark I which so impressed him that within a few days he initiated a government contract with the Manchester firm of Ferranti Ltd to make a production version of the machine 'to Professor Williams' specification'. (This is surely an all-time record for administrative speed and brevity!) This resulted in a five-year contract running from November 1948, involving an estimated £35 000 per annum. Perhaps more importantly, the contract established a fruitful link between Manchester University and the computer industry which has been maintained through five projects up to the present day.
Early British Computers, 7 - THE MANCHESTER MARK I, starting page 040
Ferranti Ltd delivered nine production Mark I and Mark I Star computers between 1951 and 1957, the 'Star' developments subsequent to November 1951 receiving financial support from the National Research Development Corporation (NRDC). The first Ferranti Mark I20 was installed at Manchester University in February 1951, thereby becoming the world's first commercially available computer to be delivered. It was identical in design to the October 1949 University machine, except in the matters of detail given in Appendix 2. The machine contained about 4050 thermionic valves and consumed about 25 kilowatts of power.
Fig. 7.2 Some of the Manchester Mark I design team. Left to right: D. B. G. Edwards, F. C. Williams, T. Kilburn, A. A. Robinson, G. E. Thomas. Professor Williams is seen peering at the operator's monitoring and display screen. In an age when acronyms were popular, this computer was sometimes called MADM - Manchester Automatic Digital Machine.
Early British Computers, 7 - THE MANCHESTER MARK I, starting page 041
Fig. 7.3 Symbol of a fruitful collaboration between university and industry: Brian Pollard (left) and Keith Lonsdale of Ferranti Ltd, seated at the console of a Ferranti Mark I with Tom Kilburn in the background. This machine, the production version of the University prototype, was delivered in February 1951 and is thus believed to be the world's first commercially available computer. To the right of the photograph some of the machine's panels have been removed to reveal the same type of circuit-carrying door as is shown in figure 4.3.
Fig. 7.4 The Manchester University Mark II computer, nicknamed MEG, was 20 times faster than Mark I yet consumed less than half the electrical power. It first worked in May 1954.
Early British Computers, 7 - THE MANCHESTER MARK I, starting page 042
The Ferranti Mark I provided a computing capacity far in excess of the University's own needs at that time, and so outside users were actively encouraged. Most of these users came from industry or other Universities, though Dr Alec Glennie, from the government's Fort Halstead establishment, carried out extensive calculations for atomic weapons. (So secret were the calculations that he was required to lock himself in the computer room, and even destroy the printer inking-ribbon upon leaving!) As a hobby Glennie developed in 1952 his 'autocode', and what is believed to be the world's first compiler.21 Glennie's system was not available to the general user, who had to wait until March 1954 when R. A. Brooker introduced the Mark I Autocode language - a distinct improvement on the rather primitive machine-code conventions which Manchester programmers had had to endure. (These conventions, instituted by Alan Turing, required users to write their programs in terms of the alphabetic symbols of the 5-bit teleprinter code. To Turing, who had spent countless hours at Bletchley Park battling with Geheimschreiber 5-bit ciphers during the war, the teleprinter code must have seemed very natural. To lesser mortals it was painful! An example of the agony is given in Chapter 15.) As in most projects, there were moments of light relief. The following is a love letter, 'created' by the Ferranti Mark I in an unpredictable way by making use of an electronic random number generation facility which the machine possessed:Darling Sweetheart,
You are my avid fellow feeling. My affection curiously clings to your passionate wish. My liking yearns to your heart. You are my wistful sympathy: my tender liking. Yours beautifully, M. U. C.
Early British Computers, 7 - THE MANCHESTER MARK I, starting page 043
From 1951 the Williams/Kilburn design team was working on a Mark II computer22 nicknamed MEG (Megacycle engine). This had a similar structure to the Mark I from the user's view, except that it avoided many of the problems associated with numbers overflowing the 'space' allocated (i.e. MEG used floating-point arithmetic). It was possibly the first stored-program computer to include this facility. Electronically, MEG was some 20 times faster than the Mark I, yet was more compact and consumed less than half the power. MEG first ran a program in May 1954. It was the prototype for the Ferranti Mercury computer, of which the first one was delivered to the Norwegian Defence Research Establishment in August 1957 as described in Chapter 14.
It is now necessary to go back in time a few years to pick up the threads of computer development at the National Physical Laboratory, Teddington. It will be remembered that the mathematician A. M. Turing produced a design proposal for the ACE stored- program computer as early as February 1946, but that the project had run into organisational difficulties. Turing had effectively left NPL in the summer of 1947.
Fig. 7.5 The Ferranti MERCURY, the production version of MEG, shows the wardrobelike impersonality of the large modern computer - with the difference that in 1957 each programmer still had complete control of the machine. This accounts for the clock and the comfortable chair!
|Early British Computers, 8 - THE NPL PILOT ACE, starting page 044|
The construction of a pilot version of Alan Turing's ACE proposal had got under way in earnest by mid-1948. The design team was drawn from both the Mathematics and Electronics Divisions at NPL. The leading lights were Dr J. H. Wilkinson (Mathematics) and E. A. Newman (Electronics). The former was an NPL colleague of Turing's; the latter had been recruited from EMI Ltd when the Electronics Division was formed.
The Pilot ACE had a complicated instruction format,23 and was quite unlike any other contemporary machine. An impression of Turing's approach may be inferred from his barbed comment of December 1946, on being shown an outline proposal for Wilkes' Cambridge EDSAC: 'The "code" which he [Wilkes] suggests is however very contrary to the line of development here [at NPL], and much more in the American tradition of solving one's difficulties by means of much equipment rather than by thought.' (This is rather unfair on the rest of the world: Turing's design may have been economical in equipment but it certainly caused the programmer to work hard, as can be seen in Chapter 15.)
As is shown in Appendix 2, the basic internal clock frequency of the Pilot ACE was, at 1 megacycle, the fastest of the early British computers. Instruction times were highly dependent on the position in store of the instruction and an addition could take anything between 64 microseconds and 1.024 milliseconds. The main store consisted initially of 128 32-bit words in mercury delay lines. This was extended to 352 words by the end of 1951, and a 4096 word drum store was added in 1954. Since the NPL already had a large Hollerith punched-card calculator it was sensible to make cards the medium for both input to and output from the Pilot ACE. The computer contained 800 thermionic valves and some help with the construction of the chassis was obtained from the English Electric Company. In concept the machine was similar to Turing's original 1946 proposal, but with some interesting changes, as noted in Appendix 2.
Early British Computers, 8 - THE NPL PILOT ACE, starting page 045
Fig. 8.1 Alan Turing in 1951, on being elected a Fellow of the Royal Society at the age of 39. This brilliant mathematician went too far and too fast for many of his contemporaries, thus making the long-term influence of his ideas on computer design rather less than perhaps it should have been. His highly original proposals for the Automatic Computing Engine (ACE) date from late 1945. Turing died in tragic circumstances in 1954.
The Pilot ACE first ran a simple program in May 1950 and was successfully demonstrated to the press in December of that year. After
Early British Computers, 8 - THE NPL PILOT ACE, starting page 046
some modification it went into full-time use as a service computer in February 1952, and it can be inferred that unreliability caused some headaches before that date. In its structure the Pilot ACE was a remarkable machine, achieving very high performance from a relatively small amount of electronics. The design, which was essentially created in 1945, is so obviously original that it belies the oft-expressed view that all stored-program computers stem from the pioneering EDVAC report. It is clear that Turing's ACE design was, in a word, unique.
Fig. 8.2 The Pilot ACE computer being demonstrated to the Press at the National Physical Laboratory in December 1950 when, to the delight of its designers, it performed with unprecedented reliability. A mixture of apprehension and relief may perhaps be detected on the faces of (left to right) G. G. Allway, E. A. Newman and J. H. Wilkinson. Much of the Pilot ACE can be seen today in the London Science Museum. (Note: DSIR stands for Department of Scientific and Industrial Research, the government body responsible for the National Physical Laboratory.)
Computer design continued at NPL between 1953 and 1957 on the full (final) version of ACE.24 This had a 48-bit word, employed delay-line storage, and had a multiplication time of about 448 microseconds. It contained about 7000 valves and was working by late 1957. Meanwhile,
Early British Computers, 8 - THE NPL PILOT ACE, starting page 047
cooperation with the English Electric Company resulted in a properly engineered version of the Pilot ACE. This was appropriately named the English Electric DEUCE. The first production DEUCE was delivered in 1955, one of the early installations being at the NPL itself in March of that year. This and other English Electric developments are described in Chapter 13. Although NPL are today continuing to make contributions to particular areas of computer science, 1957 saw the conclusion of their excursions into the field of building large computers.
The completion of ACE marked the end of one line of influence emanating from the wartime deciphering work at Bletchley Park. The other line, nurtured by T. H. Flowers and A. W. M. Coombs at the GPO's Dollis Hill research laboratory, will be described after we have first caught up with some developments in the application of transistors to computers.
|Early British Computers, 9 - TRANSISTOR COMPUTERS, starting page 048|
All of the stored-program computers described so far used thermionic valves. Although the physical operation of a transistor was discovered at Bell Telephone Laboratories before the first computer actually worked, the devices were for some years only of interest to electronic research groups. This was because of the difficulties experienced in manufacturing reliable transistors of the point-contact variety then in use. Bell Telephone Laboratories naturally experimented with the use of transistors in computing circuits and this led to an Air Force contract to build a special-purpose computer called TRADIC, whose 'program' was set up manually on a plug-board. TRADIC25 contained 700 point-contact germanium transistors and was working under test conditions in the spring of 1954.
Despite the American technological lead, it was actually in Britain that the first proper transistor computer came into operation. Concurrent with the MEG development mentioned previously, Kilburn's group at Manchester University built a small transistor research computer to gain some experience with the new devices. R. L. Grimsdale took a leading part in the design. Two versions of the computer were completed - in November 1953 and April 1955 respectively. Both versions had a one-plus-one address instruction format and a drum store; the second machine had an extended instruction set and more storage.26 Although somewhat unreliable and slow (the average instruction time was 30 milliseconds), the November 1953 computer is believed to have been the first transistorised machine to run a program. The Metropolitan-Vickers Company adapted the prototype's design to form the basis of their MV950 computer. The first production model was completed in 1956 and six MV950S were made - mainly for internal use within the Metropolitan-Vickers organisation. The production MV950 used the more reliable junction type of transistor.
Early British Computers, 9 - TRANSISTOR COMPUTERS, starting page 049
Fig. 9.1 Transistors old and new. Upper group, left to right: two germanium junction transistors of about 1957 and 1959 respectively; two silicon planar transistors of about 1963 and 1968 respectively. Lower left: a germanium point contact transistor of about 1953; lower right: a group of three silicon planar integrated circuit modules of the 1970s. The far integrated circuit module has its cover removed to reveal a silicon chip measuring about 0.5 centimetres square. This chip contains the equivalent of about 700 transistors and 1000 resistors, and is typical of large-scale integration of the late 1970s. The mysteries of all the semiconductor devices shown in the photograph derive from the original American invention of the transistor at Bell Laboratories in 1947.
The early transistor circuit experiments at Manchester were very similar to those carried out at the same time by the UK Atomic Energy Research Establishment, Harwell. At Harwell the experiments led to the design of the CADET computer, which first ran a simple test program in February 1955.27 From August 1956 CADET was offering a regular computing service, by which time the machine contained 324 point-contact transistors and 76 junction transistors.
Meanwhile in America the surface barrier transistor was in vogue, being used in the MIT Lincoln Laboratory's TX-0 computer (first working in 1956) and the Philco Corporation's TRANSAC S-1000
Early British Computers, 9 - TRANSISTOR COMPUTERS, starting page 050
computer (first working in 1957). Both these machines were considerably larger and faster than any British transistor computer, and more was to follow. At about this time news began to come through of two very high-performance American transistor projects known as LARC and STRETCH. Clearly, Britain was falling behind.
By the autumn of 1956 Tom Kilburn and his team at Manchester had begun work on another transistor computer called MUSE - a 'microsecond engine'. This was an ambitious project28 which aimed at computing speeds approaching 1 microsecond per instruction. At about the same time the Brunt Committee and Lord Halsbury of the National Research Development Corporation were also trying to initiate a British high- speed computer project, spurred by reports of the massive scale of the Univac LARC and IBM STRETCH computer developments in America. Like many committees before and since, the NRDC was unable to find agreement on exactly who should design such a computer for Britain. From Manchester University's view, neither government nor industrial support was forthcoming for their own MUSE proposal, and so Kilburn decided to go ahead independently with a limited prototype, using internal resources including a Mark I computer Earnings Fund. Amongst Kilburn's team, D. B. G. Edwards was responsible for the hardware side and R. A. Brooker for software. By good fortune, Ferranti Ltd decided at the end of 1958 to support the project. By 1959 the computer had been re-named ATLAS and was thereafter developed as a joint University/Ferranti venture under Tom Kilburn. Perhaps the main Ferranti design contribution to the project lay in the stalwart work of David Howarth on the Atlas Supervisor - considered by many to be the first recognisable modern operating system.
Fig. 9.2 The Manchester University experimental transistor computer, operational in November 1953 and therefore the world's first transistor stored-program machine to work. The magnetic drum store on the left of the photograph was re-cycled from an earlier computer project (as may be seen by comparison with figure A2.1).
Early British Computers, 9 - TRANSISTOR COMPUTERS, starting page 051
Fig. 9.3 The Metropolitan-Vickers MV950, thought to be the first transistor computer to become commercially available (1956).
Early British Computers, 9 - TRANSISTOR COMPUTERS, starting page 0512
Fig. 9.4 M. J. Lanigan and Tom Kilburn with a section of the transistorised MUSE computer in 1959. This University prototype was transformed into the Ferranti ATLAS, believed to be the world's most powerful computer at its inauguration by Sir John Cockroft in December 1962. ATLAS introduced several novel concepts into computer design.
ATLAS was not an 'early' computer so a full description is out of place here. Suffice it to say that at its official inauguration on 7 December 1962 it was considered to be the most powerful in the world. Its fastest instructions took 1.59 microseconds and, due to features called virtual Storage and paging, each one of many simultaneous users could imagine he had up to one million words of storage space at his disposal. ATLAS pioneered many concepts which are in common use today.
ATLAS was what might be called a 'supercomputer', and consequently had a limited market. At the other end of the spectrum several successful small and medium British transistor computers had meanwhile been developed. Amongst these were the Elliott 803, EMI EMIDEC 1100 and the Ferranti SIRIUS, first delivered in 1958, 1959 and 1960 respectively. There was a long overlap period between valves and transistors, and we should now go back in time a few years to catch up with concurrent valve-based projects.
|Early British Computers, 10 - DEFENCE COMPUTERS, starting page 053|
The Government's Bletchley Park code-cracking operations have been alluded to earlier. The need to keep ahead in the cryptanalysis stakes must have continued after the war, though details are not available to the general public. Successors to COLOSSUS were built, but it is not thought that these developments contributed in any significant way to the design of general-purpose stored-program computers.
Other fields of defence activity involving digital techniques were the tracking and telemetry problems associated with guided weapons, etc. The principal stored-program computer development here was the MOSAIC (Ministry of Supply Automatic Integrator and Computer) project, which was implemented between 1947 and 1954 by a Post Office team led by Dr A. W. M. Coombes.29 NPL contributed to the mathematical specification and the All-Power Transformer Co. helped with the manufacture and assembly. Parts of the MOSAIC project are still secret, it having been used for processing radar tracking data in experiments on aircraft. Briefly, the work included building two towable 'data recorders' which produced 3-inch-wide paper tape, together with a very sophisticated display and tracking system. The computer itself had storage for 1024 40-bit words in mercury delay lines, involving nearly a ton of triple- distilled mercury. The physical layout of the delay-line tanks was copied from the scheme used at Cambridge University for the EDSAC. The man who had helped to build COLOSSUS was not afraid of thermionic valves: Coombs used 6000 of them in MOSAIC which, together with 2000 germanium semiconductor diodes, gave a total power dissipation of 60 kilowatts. Although not the earliest of early British computers, MOSAIC was arguably the largest.
Early British Computers, 10 - DEFENCE COMPUTERS, starting page 054
The MOSAIC was housed at Malvern and, whilst it was no doubt very valuable in defence circles, its design had little influence on the mainstream of computer development. In view of the price of mercury, perhaps this was just as well.
From 1947 to 1953 Malvern also housed Dr A. M. Uttley's Telecommunications Research Establishment team developing the TREAC computer.30 TREAC was the only parallel computer being designed in Britain at the time. It had a 512-word Williams tube store backed by a drum, and achieved an addition time of 40 microseconds. Its instruction repertoire was similar to that of the Manchester Mark I, except that there were no index registers. Another difference from the Manchester computer was that the CPU was synchronised to the drum and not vice versa (see Appendix 2). TREAC was notable for the care taken to ensure reliability: exhaustive component tests were carried out during the design phase and the CPU was initially planned with a scheme of 'self-checking' logic.
Fig. 10.1 Some central units of the MOSAIC computer, built by the Post Office for a secret Ministry of Supply defence project at Malvern. MOSAIC was working in about 1953 and in terms of electronic components was the largest early British computer.
Early British Computers, 10 - DEFENCE COMPUTERS, starting page 055
Fig. 10.2 The Telecommunications Research Establishment's TREAC computer, also operational in 1953. This was the first parallel electronic computer in the country. The boxes along the lower right side of the machine contain Williams storage tubes. Note the soldering iron conveniently to hand on the left!
Early British Computers, 10 - DEFENCE COMPUTERS, starting page 056
It should be mentioned that research establishments such as THE had for some years been developing other 'computers'. These were mainly analogue devices for simulating aircraft behaviour, etc. and in the context of our story they have little bearing on the history of stored-program digital computers. It is necessary to make the distinction in order to focus attention on the type of machine we now simply call a 'computer'. Analogue devices were, and are, very valuable for certain classes of problem, although it is true to say that their importance is diminishing.
All the early computer projects described so far have been associated with Universities or government research establishments. There was another independent line of development going on in the firm of Elliott Brothers, inspired originally by their work for the Admiralty on antiaircraft real-time fire control.
|Early British Computers, 11 - ELLIOTT BROTHERS, starting page 057|
The first British company to become seriously involved with digital computer technology was Elliott Brothers, a London-based firm which had been manufacturing scientific apparatus since 1801. During the war Elliott Brothers had been supplying a great deal of electro-mechanical gunnery control equipment for the Navy. In 1947 Elliott started its Borehamwood research laboratory under J. F. Coales with several naval contracts, including one for the MRS5 advanced digital real-time fire control system. This work involved the construction of a special-purpose computer or on-line digital control system, designated the '152' (Reference 31). The 152 included a fixed-program store based on the principle of the flying- spot scanner and Williams tubes for the working store. The number representation was serial, but parallel arithmetic processing was employed to give a very fast multiplication time of 60 microseconds. Another naval contract involved a special-purpose machine called the 153 which employed nickel magneto-strictive delay lines backed by a fixed-head disc. An interesting feature of the 152 and 153 was the use of miniature thermionic valves and modular circuits mounted on glass printed circuit boards. The logic circuits, developed by C. E. Owen, were to become the foundation for the Elliott 400 series and subsequently the Ferranti PEGASUS computer.
The MRS5 did not involve a truly general-purpose computer. In the end the contract was terminated by the Navy in favour of an analogue system which was being developed concurrently elsewhere. The project did, however, stimulate interest within Elliott Brothers and from the related research came a general-purpose stored-program computer called NICHOLAS. This had a 1024-word nickel delay-line store, an add instruction time of about 10 milliseconds, and first ran a program in
Early British Computers, 11 - ELLIOTT BROTHERS, starting page 058
December 1952. It was used successfully at the Borehamwood Laboratories to carry out trajectory calculations, etc. for a number of years, during which time a very convenient programming system was developed. Technical details of NICHOLAS are given in Appendix 2.
Fig. 11.1 An Aladdin's cave of electronic wizardry: the type 152 naval gunnery control computer, begun at the Borehamwood laboratories of Elliott Brothers in 1947. This special-purpose real-time computer had a fixed program store, and its initial task was to analyse the performance of the radar set via data recorded on photographic film. Five of the six digital cameras for reading film cassettes may be seen in the foreground of the picture. Out of a related Admiralty contract came the basic modular design for several subsequent general-purpose computers.
Meanwhile the company's Computing Division under W. S. Elliott received an NRDC contract in September 1950 to study the application of printed circuits and other Borehamwood technology to general-purpose computers. This culminated in a report dated January 1952 which, in April of that year, led to NRDC placing a contract with Elliott Brothers for the construction of a small prototype machine. This became the Elliott 401 computer, for which the detailed design team was led by A. St Johnston. It was completed in April 1953 and exhibited that month at the Physical Society Exhibition in London. The 401 had a 1000-word disc store and nickel delay line central registers. It inherited the modular plug-in philosophy from the MRS5 project. Elliott Brothers then developed the 402 production version (first delivered in 1955) and other machines in their successful 400 series. The 405 (first delivered 1956) was a general business machine featuring bulk storage on magnetic film and other interesting features relevant to commercial data processing. The prototype 401 was owned by the NRDC, who moved the machine temporarily to Cambridge and continued the development there until March 1954, when it was handed over to the Rothamstead Experimental Station.
Early British Computers, 11 - ELLIOTT BROTHERS, starting page 059
Fig. 11.2 The Theory Laboratory at Borehamwood, receiving a contract to compute trajectories, realised that the easiest and most interesting way to solve the tedious calculations was to build themselves a computer. NICHOLAS was the result. It was in parts literally home-made, and the wooden frame and consequent fire extinguisher may be seen in this photograph. NICHOLAS performed valuable service from 1952 to 1958.
Early British Computers, 11 - ELLIOTT BROTHERS, starting page 060
In 1953 the 401 team split in two, with W. S. Elliott and others leaving to join Ferranti's London branch. Here they put their modular packaging experience to good use in the design of the Ferranti PEGASUS computer, first delivered in 1955 as described in Chapter 14. The design emphasis for both the 400 series and PEGASUS favoured the use of standard, interchangeable circuit modules. This obviously made sense for volume production, and contrasted with the 'hand-built' methods used for the larger Ferranti Mark I. Modularity in varying degrees was quickly adopted by all computer manufacturers.
Fig. 11.3 The Elliott 401 computer (1953), built under a National Research Development Corporation contract to promote modular circuit techniques. The resulting plug-in packages here give the appearance of books in a book-case. The magnetic disc store and its motor may be seen in the upper right compartment in the photograph. The 401 was the progenitor of a family of successful Elliott and Ferranti modular computers.
Early British Computers, 11 - ELLIOTT BROTHERS, starting page 061
Besides the 405, the team which remained at Elliott Brothers later developed the 800 series of transistorised machines which were in many ways forerunners of the modern mini computer. The Elliott 802 was first delivered to customers in about 1958. (The American PDP/8, manufactured by the Digital Equipment Corporation, is generally regarded as the first widely used minicomputer proper. In Britain PDP/8s began to arrive in about 1965 and it was not until 1969 that the number of PDP/8 installations overtook the number of Elliott 8036 - see reference 50.)
|Early British Computers, 12 - PIONEERING SMALL COMPUTERS, starting page 062|
With the exception of machines such as the Elliott 402 and the MV950, the computers discussed so far were all considered 'large'. That is to say, the financial outlay and manpower training invested in each computer installation was relatively large for the organisation concerned. From the beginning there were also some intentionally small projects.
A very early pioneer in small computers was A. D. Booth of Birkbeck College, London University. He had spent six months with von Neumann's computer group at Princeton University in 1947 and had returned to England to design ARC, the Automatic Relay Computer. He was supported for a time by the British Rubber Producer's Research Association. ARC was envisaged as a parallel stored-program computer, but development of a suitable store caused some problems. Although the arithmetic section of ARC was working in the spring of 1948, Booth's magnetic drum store containing 256 20-bit words was not fully in action until some time later.33, 33 ARC was eventually equipped with an electromechanical store for 50 numbers, together with a pluggable sequence unit for 300 instructions, and operated successfully as a sequence-controlled calculator. (It should be added that Booth was short of funds and was obliged to work more or less as a one-man-team for much of the time, with assistance on the programming side from Miss Kathleen Britten whom he subsequently married.)
By 1949 Booth had moved to thermionic valves, where his emphasis was on the design of low-cost computers which could be, and were, attractive to smaller scientific organisations. After a brief experiment with SEC, a Simple Electronic Computer, he designed the APE(R)C series of All-purpose Electronic (Rayon) Computers. (The 'Rayon' signified the sponsorship of the British Rayon Research Association; other sponsors were indicated by inserting their initials in parenthesis.) The APE(R)C had a re-designed magnetic drum built by S. J. Booth, A.D.'s father, and contained 415 thermionic valves. APE(R)C was operating with limited storage in July 1952, and with a complete store some months later.34
Early British Computers, 12 - PIONEERING SMALL COMPUTERS, starting page 063
Fig. 12.1 A. D. Booth, an early pioneer of small computers, seen here with some pattern recognition apparatus at Birkbeck College in 1959. Booth's ingenious ideas transcended the comparatively modest resources at his disposal.
The British Tabulating Machine Company became interested in Booth's work and derived their HEC (Hollerith Electronic Computer) from his APE(R)C project. The nomenclature arose because BTM punched-card equipment had always been known as 'Hollerith machines',
Early British Computers, 12 - PIONEERING SMALL COMPUTERS, starting page 064
in recognition of their historic link with Herman Hollerith and to distinguish them from 'Powers machines'. A prototype HEC was exhibited at the Business Efficiency Exhibition in London in 1953. HEC was first marketed as the BTM 1200 computer in 1954, in which year five were sold. The first business data processing version, the BTM 1201, had an enlarged 1024-word drum store and was first delivered in 1956; 70 such machines were eventually sold.
Fig. 12.2 The APE(R)C computer at Birkbeck College, London University, in about 1952. To the left of the photograph is the teleprinter input/output equipment; the right-hand tall 'post office' rack contains a small drum store about two thirds of the way down; the box on the right of the photograph is the power supply. This project was the inspiration for the British Tabulating Machine Company's HEC computer, later marketed as the BTM 1200 series.
Early British Computers, 12 - PIONEERING SMALL COMPUTERS, starting page 065
Fig. 12.3 The ICCE relay computer at Imperial College, London University, in about 1952. The electro- mechanical relays are contained behind the horizontal rows of covers. ICCE was the largest of the early British relay computers.
Besides Booth's ARC, there were three other relay computer projects of note, all described in appropriate chapters of reference 16. In philosophy, they belonged to the field between calculators with and without the stored-program feature. This in- between area included machines called sequence-controlled calculators which generally held their instructions on some form of semi-permanent external 'storage' such as punched paper tape. Of the three relay computers under discussion, one was at Imperial College, London, under K. D. Tocher and S. Michaelson, another at the Royal Aircraft Establishment (RAE) Farnborough under Dr S. H. Hollingdale and E. J. Petherick, and the third at the Atomic Energy Authority's Harwell establishment under E. H. Cooke-Yarborough and R. C. M. Barnes. The Imperial College machine (called ICCE) employed 20-bit words, with the main storage being external to the main computer, on paper tape. The Farnborough machine was decimal and also envisaged a certain amount of external storage. Originally based on relays, the Farnborough machine eventually went over to a drum store and decatron valves. The Harwell machine
Early British Computers, 12 - PIONEERING SMALL COMPUTERS, starting page 066
was also decimal and initially had 40 8-digit decatron registers for internal storage. Although it could on occasions act as a true stored-program computer, this was not its normal mode of operation. It had a multiplication time of between 5 and 10 seconds, and first worked in April 1951. The Farnborough machine was eventually known as RASCAL (RAE Sequence Calculator) and was still under construction in 1953, by which time the Plessey Company were involved in the development. RASCAL was never actually completed, being overtaken by events at RAE such as the arrival of an English Electric DEUCE. The Imperial College machine was the largest and fastest of the three relay computers, and was working by about 1950. Neither of the three machines seems to have had any direct influence on subsequent industrial developments.
Returning to valves, a small machine with a Dutch pedigree was built by Standard Telephones and Cables Ltd, Monmouth, in the mid-1950s. This was the Stantec ZEBRA, based on a design of W. L. van der Poel from the University of Delpht in Holland. ZEBRA had a main store of 8192 33-bit words on a magnetic drum. When first delivered in about 1957 it cost approximately £23 000, thus being amongst the cheapest general-purpose computers then on sale.35 A total of about 20 ZEBRAS were installed in Britain.
We have so far been concerned only with digital computers. Throughout the period under discussion analogue computers were being built and used, especially for problems related to aircraft design. One of the earliest examples of a hybrid computer, that is to say a machine in which a stored-program digital computer is combined with an analogue section, was built between 1953 and 1955 by Smiths Aircraft Instruments Ltd of Cheltenham. This computer was developed by a team under L. Dilger to cope with the large amounts of simulation data encountered in certain problems associated with blind landing systems, etc. The digital section had an accuracy of four decimal digits, using decatrons and a drum for storage. Input/output was via a teleprinter. The project, which was named SECA (Smiths Electronic Calculating Analyser), was intended to complement an analogue computer of advanced design called SEDA (Smiths Electronic Differential Analyser). The development of both SEDA and SECA was undertaken to provide simulator facilities for guided weapons. Unfortunately, certain key contracts were cancelled and
Early British Computers, 12 - PIONEERING SMALL COMPUTERS, starting page 067
SECA was never fully operational. SEDA, however, continued to be used for about ten years for the development of automatic flight control systems, including those for automatic landing.
Fig. 12.4 A section of the relay computer designed at the Atomic Energy Authority's Harwell establishment in 1951. This machine gave faithful service for many years, being donated to Wolverhampton College of Technology in 1957. It was then re-named WITCH (Wolverhampton Instrument for Teaching Computation from Harwell) and continued to initiate the uninitiated into the mysteries of computing until its retirement in 1973.
Return to On Line Documents
Return to "Early English Computers" table of contents