Return to On Line Documents
Permission to publish from Simon Lavington here
Thanks to John Barrett for helping proof the OCRing :-))


Early British Computers

THE STORY OF VINTAGE COMPUTERS
AND THE PEOPLE WHO BUILT THEM

Simon H. Lavington


Front cover


Back cover
EARLY BRITISH COMPUTERS details the pioneering work on modern computers that took place in the United Kingdom between 1935 and 1955 including such landmark projects as the world's first working stored- program computer, the first commercially available computer, and the first transistorized computer. Beginning with a survey of computing in the 1930s, this book chronicles secret wartime developments, post-war re- search in universities and government establishments, and the work of companies such as Elliot Brothers, English Electric, Ferranti and Lyons in the 1950s. From here the growth of an indigenous computer industry is traced up to the period of merger and consolidation in the early 1960s. Contemporary American developments highlight the interplay of ideas on both sides of the Atlantic.

With its comprehensive bibliography, guide to computer terminology, and extensive illustrations, EARLY BRITISH COMPUTERS is an invaluable book for expert and general reader alike.


Also available in the Digital Press History of Computing Series: PROJECT WHIRLWIND: The History of a Pioneer Computer by Kent C. Redmond and Thomas M. Smith Whirlwind was the first high-speed electronic digital computer able to operate in "real time." This carefully researched and generously illustrated book examines the technological traditions, the research and development policies and practices, the funding crises, the management techniques, and the administrative philosophies that went into this innovative engineering triumph. For more information on this book and other titles, write to:
DIGITAL PRESS
Educational Services
Digital Equipment Corporation
12 Crosby Drive, Bedford, MA 01730
ISBN 0-932376-08-8


Copyright c 1980 by Simon H. Lavington

All rights reserved. Reproduction of this book, in part or in whole, is strictly prohibited.
For copy information in the U.S.A., its territories, and Canada,
contact Digital Press, Educational Services, Digital Equipment Corporation, Bedford, Massachusetts 01730.

Printed in U.S.A.

1st Printing, May 1980
Documentation Number EY-AX012-DP-001
Library of Congress Cataloging in Publication Data:
Lavington, Simon Hugh.

Early British Computers

 
 1. Title.
 TK7885.A5L38 621.3819'5'0941 80-13475
 
 ISBN 0-932376-08-8
 
 Sole distributor for the U.S.A. and Canada:
 DIGITAL EQUIPMENT CORPORATION
 
 Published in England by Manchester University Press, 
 Oxford Rd., Manchester M13 9PL.
 
 DIGITAL
 

CONTENTS

page
1. Introduction 1
2. Computing in the 1930s 4
3. The second world war 8
4. The technology of early computers 13
5. The ACE, the 'British national computer' 23
6. The Cambridge EDSAC 31
7. The Manchester Mark I 36
8. The NPL Pilot ACE 44
9. Transistor computers 48
10. Defence computers 53
11. Elliott Brothers 57
12. Pioneering small computers 62
13. Leo and English Electric 68
14. Ferranti Ltd, ICT and ICL 78
15. Programming an early computer 87
16. Meanwhile in America 98
17. The National Research Development Corporation 102

Appendix 1: the plain man's guide to computer terminology

106
Appendix 2: technical specification of 15 early British computers 114
Appendix 3: technical specification of 9 early American computers 120

Bibliography and references

126
Acknowledgements and picture credits 131
Index 133


Early British Computers, INTRODUCTION, starting page 001


1
INTRODUCTION

The word 'computer' prior to 1940 meant only one thing: a clerk equipped with a hand calculating machine who would 'compute' the standard calculations required for wages, actuarial tables, astronomical predictions, etc. This was of necessity a tedious process. From 1935 to 1945 the application of electronics to 'automatic computers' generally took the form of developing faster calculating equipment related to specific problems. However, in 1945 there arose an interest in the design of universal automatic computers which could be applied to an unlimited range of problems, once the appropriate 'program' had been fed into the machine. These universal computers are the ones we would recognise today.

The universal computer has become known as the stored-program digital computer, because it embodies two interesting characteristics not present in other types of calculating machine.

These characteristics are:

  1. An internal store (or memory) whose contents can be selectively altered automatically during computation. This store is used to hold both instructions (the program) and data.

  2. The ability of the machine to vary its actions in a strictly defined manner, according to the value of data items encountered during computation. In practice this is only achieved if data items are represented as numerical, i.e. digital, quantities inside the computer.
(For those unfamiliar with computer terminology, an introduction to the concepts and definitions of the more important technical terms is given in Appendix 1.)


Early British Computers, INTRODUCTION, starting page 002

Fig. 1.1 Part of Charles Babbage's Analytical Engine, designed between 1833 and 1837. This remarkable mechanical calculating machine was planned to have a store (or memory) for 1000 numbers, and was to be controlled by punched cards similar to those used in the Jacquard loom. In developing his engine, Babbage conceived the idea of what we would now call a stored-program computer.

The first person to appreciate the possibilities of stored-program computers was the Cambridge mathematician Charles Babbage (1791-1871), who planned a mechanical Analytical Engine which was to be controlled by punched cards.1, 2 Babbage's implementation was, in the event, defeated by technical difficulties, but he introduced concepts which were remembered a century later once the electronic era had dawned.

Babbage apart, the question 'who invented the computer?' is really unanswerable. By the mid-1940s there were several research groups in Britain, America and Germany who were familiar with the basic concept of a stored-program digital computer. They worked from a basis of common sense, generally coupled with an intimate knowledge of existing special-purpose calculators such as the huge ENIAC (Electronic Numerical Integrator

Early British Computers, INTRODUCTION, starting page 003

and Computer) - an American machine which is described later. The transition from calculators without the stored-program capability to those with it was a gradual though conscious one, related principally to the development of suitable storage techniques. The first machine based entirely on the stored-program principle was a very small computer which worked in Manchester University on 21 June 1948 but, as with the Wright brothers and the history of aviation, Manchester should be regarded as just one of several centres of activity.

This book centres on the emergence of an indigenous computer industry during the decade 1945-55. It is largely a story of individual effort seen against a background of various, usually unsuccessful, attempts at coordination on a national level. The decade in question saw an impressive crop of British ideas in spite of, or more likely because of, the lack of central coordination and lavish funding. All this, however, is to anticipate events. At the start of the period under discussion there were few people who knew (or cared) what a stored-program computer was. Even those who did were uncertain as to whether contemporary technology could be made reliable enough to give useful computing runs. To set the scene we must first consider what 'computing' meant before the second world war.



Early British Computers, COMPUTING IN THE 1930s, starting page 004

2
COMPUTING IN THE 1930s

Before the second world war there were three types of calculating machine in common use. Most numerous were the mechanical and electro-mechanical hand calculators which would add, subtract, multiply and divide two numbers. These machines were generally about the size of a typewriter. Most of them utilised mechanisms based on German, Swedish and American inventions. Secondly, there were electromechanical punched-card machines such as sorters and tabulators, derived from the pioneering work of Herman Hollerith and James Powers in America at the turn of the century. Originally applied to commerce and statistics, punched-card devices began to be used for scientific calculations in the late 1920s. The third and most spectacular class of calculator was the Differential Analyser. This was a large and highly specialised 'equation solver'.3

Fig. 2.1 A popular type of hand calculator known as a Brunsviga, in recognition of its German origins. These machines could add, subtract, multiply and divide, and could be used to 'compute' tables of standard astronomical data, etc.


Early British Computers, COMPUTING IN THE 1930s, starting page 005

Fig. 2.2 Punched card equipment of the 1930s. Tabulators and sorters of the type shown above were widely used in commerce and administration. Note the Queen Anne legs!

The fundamental mechanism of a Differential Analyser is a wheel-and-disc integrator - see Fig. 2.4. This was invented in 1876 by James Thomson (brother of Lord Kelvin, the celebrated Scottish scientist). The developments necessary to produce a practical calculating machine were carried out in America by H. W. Nieman and Vannevar Bush, who produced the first full-scale Differential Analyser in 1930. Professor D. R. Hartree of Manchester University brought the ideas back to England, and by 1939 four large Differential Analysers had been built in this country - (at Manchester University, Cambridge University, Queen's University Belfast, and the Royal Aircraft Establishment, Farnborough). The great contribution of Differential Analysers to science and engineering was the speed at which they could solve differential equations, compared with other contemporary methods.


Early British Computers, COMPUTING IN THE 1930s, starting page 006

Fig. 2.3 A large differential analyser, designed by D. R. Hartree in 1935. The prototype for this was a small- scale machine built from pieces of children's Meccano construction sets, which actually solved useful equations concerned with atomic theory in 1934.

Although by 1939 most of the calculating equipment available was derived from foreign designs, Britain was well to the fore in the development of computational methods. In particular, the mathematician Dr L. J. Comrie founded Scientific Computing Service Ltd in London in 1937, a company which built up a considerable reputation in the development of new techniques and the production of mathematical and astronomical tables. The Scientific Computing Service's 'human computers' (i.e. mathematical clerks) used a variety of electro-mechanical hand calculators and punched-card equipment. They offered a unique consulting service to the scientific community.

As was indicated earlier, all the calculating devices of the 1930s differed fundamentally from today's computers. Although the true stored-program computer did not become a reality until after the war, two small developments in the late 1930s can now be seen as tentative

Early British Computers, COMPUTING IN THE 1930s, starting page 007

pointers in the right direction.
4 One such pointer was the project undertaken at Iowa State College between 1937 and 1941 by J. V. Atanasoff and C. Berry, who planned an electronic binary digital 'equation solver' (which was never in fact completed). The other was the series of electro-mechanical binary digital 'equation solvers' built in Berlin by Konrad Zuse, of which the first fully operational model (Z3) was working in 1941. Both projects involved a certain amount of internal storage for numbers, but not for instructions. The machines were of a type called sequence-controlled calculators, representing an important steppingstone towards the modern computer. The exigencies of war caused setbacks to both projects. However, Zuse, with great personal tenacity and insight, started building machines again in 1948; in 1950 he founded his own company Zuse KG.

Fig. 2.4 The wheel-and-disc integrating mechanism of a differential analyser, based on an idea of the Scottish scientist James Thomson in 1876. The idea was developed further in America in the 1920s, thereby giving a means for the rapid solution of differential equations. This stimulated interest in 'automatic computation'.



Early British Computers, THE SECOND WORLD WAR, starting page 008

3
THE SECOND WORLD WAR

The second world war brought a host of computational problems which had to be solved in a hurry. The areas for which special-purpose computing devices were designed and built included gunnery control, flight simulation and aircrew training, radar signal processing (e.g. in airborne interception) and code deciphering. The devices produced were, in general, too specialised to have any direct influence on subsequent stored- program computer development, except in one important sense. They created at places such as the Telecommunications Research Establishment (TRE) and the Post Office Research Station, groups of engineers who became very competent at designing ingenious electronic equipment. (TRE is now the Royal Signals and Radar Establishment, Great Malvern, Worcestershire.) British inventiveness was in full flood and many post-war stored-program computer projects owed a lot to the engineering impetus generated during the war years.

There is one particular wartime project which is of considerable historical interest, since it involved the first large-scale use of thermionic valves for digital computation. This was the special-purpose 'computer' named COLOSSUS, produced in great secrecy by the Post Office Research Station for the Government Code and Cipher School, Bletchley Park, Buckinghamshire. The full story of British code-breaking is still shrouded in the Official Secrets Act, but the following may be inferred from the few facts which are available (see, for example, references 5 and 6).

The German forces are known to have used two classes of machine for encoding signals prior to transmission: the Enigma series and the Geheimschreiber system. Both machines 'scrambled' the letters of a message by a complicated, virtually non-repeating mechanism of stepping rotors which rendered the encoded signal extremely difficult to decipher - unless of course the recipient had a similar machine to the sender's, adjusted to a similar set of initial rotor positions. The difference between the two systems was partly one of operational convenience: Enigma machines had a three-rotor system (four for the German Navy), and were portable; Geheimschreibers had ten rotors and used the standard 5-bit teleprinter code (Baudot code) for actual transmission. ('Bit' is an abbreviation of 'binary digit', see Appendix 1.) Enigma codes were used for all day-to- day military signals and Geheimschreibers, being regarded as more or less unbreakable, for top secret strategic messages.


Early British Computers, THE SECOND WORLD WAR, starting page 009

Fig. 3.1 The first COLOSSUS code-cracking machine, operational at the British Code and Cipher School at Bletchley Park in December 1943. This special-purpose 'computer' contained 1500 thermionic valves and proved that large numbers of electronic circuits could be made to do reliable calculations at speed. The system of pulleys to the right of the photograph guided the punched paper tape, containing the message to be decoded, through a photo-electric sensing unit. Information about trial cipher keys was generated by COLOSSUS internally during the decoding process. Partial results and the final decoded message were automatically printed out by the electric typewriter, seen to the left of centre.


Early British Computers, THE SECOND WORLD WAR, starting page 010

Fig. 3.2 Two WRNS operators at the controls of COLOSSUS. The deciphering process was interactive, in the sense that an operator was able to alter the sequence of trial cipher keys in the light of meaningful transliterations which might appear on the machine's printer. Full details of COLOSSUS and other wartime code-cracking machines are still secret.

There were two aspects to cracking Enigma and Geheimschreiber codes. One was to try and keep abreast of the developments which took place from time to time in the machines, for example in rotor wiring and plugboard connections and in method of use; the other was to work out the system of initial rotor settings, which were changed three times a day according to tables printed in German code- books. Inspired by pre-war code-cracking activity in Poland, the basic method used by the British at Bletchley Park for decoding was to build 'trial and error' machines which scanned quickly through many combinations of symbols until sensible transliterations were obtained. The decoding machines were, in part, Enigmas in reverse. Since three Enigma rotors had approximately 102' (one thousand billion billion) possible initial states, considerable ingenuity was required at Bletchley Park to reduce the number of trial combinations to a manageable size for each message. It was only the cleverness of Bletchley mathematicians such as A. M. Turing that made decoding of large numbers of messages a practical
Early British Computers, THE SECOND WORLD WAR, starting page 011

proposition. Even so, many tens of electro-mechanical relay machines and many hundreds of WRNS operators were employed full-time on deciphering Enigma signals. Their success contributed very significantly to the allied war effort.

Deciphering the Geheimschreiber codes was much more difficult and relay-based machines proved too slow. Accordingly, in January 1943 a small team at the Post Office Research Station, led by T. H. Flowers and working to functional requirements drawn up at Bletchley Park by the mathematician M. H. A. Newman, designed an all- electronic deciphering 'computer', the COLOSSUS. This was assembled at the Post Office Research Station at Dollis Hill, London, and began useful work at Bletchley Park in December 1943. COLOSSUS contained 1500 thermionic valves - far more than any single electronic device built up to that time. The message to be deciphered was fed into an optical reader as a repetitive loop of punched paper tape in 5-bit teleprinter code, at the amazing rate of 5000 characters per second. (It was said that the hardened steel guide-pins had to be replaced frequently, because paper tape travelling at such a high speed soon wore grooves in them!) Internally, COLOSSUS contained electronics for counting, comparison, simple binary arithmetic and logical operations. Output was via an electric typewriter. The 'program', or strategy for altering trial cipher keys, was controlled from plug-boards and switches. It is believed that the deciphering was interactive, the cryptanalyst making adjustments to the program switches in the light of meaningful letter-combinations which might appear on the printer.

In 1944 a second version, COLOSSUS Mark II, was designed to keep pace with developments in the Geheimschreiber encoding rotors. COLOSSUS Mark II had conditional branching (defined in Appendix 1), was faster than its predecessors, and contained 2500 valves. The first of ten Mark II machines worked on 1 June 1944.

It is believed that COLOSSUS II contained all the elements of a modern computer except an internal program store. The interactive nature of the problem and the requirement for high computing speeds meant that such a store would have been an expensive and unnecessary luxury. The Bletchley Park machines were built for a specific purpose and their success in breaking the 'unbreakable' German Geheimschreiber ciphers

Early British Computers, THE SECOND WORLD WAR, starting page 012

amply justified their design. On a more general level, they proved that high-speed digital computing could be carried out reliably using thermionic valve circuits. The COLOSSUS team was aware that, once an economic storage technology became available, a universal stored-program computer would be within sight. The only trouble was that they weren't allowed to talk about their experience once hostilities ceased! On a more positive note, there is some evidence to suggest that the COLOSSUS activity did incline official thinking towards digital computers.

Before describing the post-war computing projects themselves, it is appropriate to survey the technology available at the time. We have today become so used to the convenience of the silicon chip that it is sometimes forgotten what electronic effort was required to implement even the simplest computational function in the 1940s. The next few pages also serve to introduce devices whose names have long since vanished from everyday computing vocabulary.



Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 013

4
THE TECHNOLOGY OF EARLY COMPUTERS

One or two of the very early stored-program computers were electromechanical, being based on devices called relays. A relay is a switch which can be opened or closed automatically by appropriate electrical signals. (The activating mechanism is an electro- magnet which can 'pull' the switch contacts together.) Relays were a comparatively cheap way of implementing computing and control equipment, though they were unsuitable for use in constructing realistic storage units. Their main disadvantage for use in processors was their relative unreliability (contactwear and susceptibility to dust) and slowness of operation. A relay took a few milliseconds (i.e. thousandths of a second) to 'switch', whereas a thermionic valve circuit could 'switch' in less than a microsecond (i.e. less than one millionth of a second). In the late 1940s computing speed, along with automatic operation, was regarded as one of the main advantages of the 'new' type of computer over its predecessors. It might thus be surmised that relays tended only to be used by research groups who could not afford anything faster!

In the period 1948-55 most computers used thermionic valves (or vacuum tubes) in their processors. There were many types of valve, giving a variety of performance characteristics. A pentode valve had five electrodes within it, whereas triode and diode valves had three and two electrodes respectively. A decatron was a special ten-state neon-based valve used in counting applications. In essence, a valve is an evacuated glass tube (sometimes inside a metal outer casing), in which some metal electrodes are made to control the flow of electrons, the electrons being produced by heating a 'cathode' electrode. In contrast, modern electronic components control the flow of charged particles inside a solid lump of semiconducting material - hence the term 'solid- state circuits'. The

Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 014

disadvantages of valves mainly concern their cathode heaters. These heaters consume appreciable electrical power and are, therefore, wasteful of energy. Further, the heat they produce is not all put to good use in providing electrons inside the valve, and surplus radiated heat has to be disposed of (e.g. by forced-air cooling). Finally, the heating elements deteriorate with time, especially if the equipment is being turned on and off periodically rather than left running. The early scientific reports on computers are full of statistics on numbers of valve failures per week and it is clear that component reliability was of great concern.

As far as semiconductors are concerned, the rectifying (diode-like) properties of semiconducting materials had been known about for some time. During the second world war components based on crystals of germanium came into service as the so- called 'crystal diodes'. Several early computers made use of these simple solid-state devices in conjunction with valves for logic circuits. The first major impact of solidstate components upon computer technology came in the mid-1950s when transistors were becoming reliable and cheap enough to replace valves altogether. The significant advantages offered by transistors were firstly ones of reduction in physical size and in power consumed. Other advantages such as reliability, cheapness and speed followed on, once the transistor manufacturing processes had evolved to a satisfactory state.

Fig. 4.1 Early computer components. Back row left to right: an electro-mechanical relay of the 1940s; a more recent relay; a pentode valve of the 1940s in a metal outer casing; a miniature pentode of the 1950s. Foreground: a 50 pence coin for size comparison; a thermionic diode of the 1940s.
Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 015

Fig. 4.2 left to right: a germanium semiconductor diode of the 1950s; the thermionic diode of figure 4.1; a silicon semiconductor diode of the 1970s.

The basic properties of a transistor were discovered in 1947, at the Bell Telephone Laboratories in America. The earliest experimental use of transistors in a computer was in 1953. Commercially available transistorised computers did not begin to appear until 1956, and it was not until the mid-1960s that valves finally became obsolescent. By 1956 the early form of point-contact germanium transistor was being replaced by the more reliable surface-barrier and junction types, to be followed shortly afterwards by the replacement of germanium by silicon as the basic semiconducting material for most computer transistors. The planar manufacturing process then replaced the 'junction' geometry in many solid-state devices, and by the end of the 1960s the planar silicon integrated circuit or 'chip' had essentially arrived. All these developments brought spectacular reductions in cost and size, whilst reliability and speed increased. The accompanying photographs illustrate these changes better than any technical description. The resulting improvements in central processor technology have given today's computer designers a freedom undreamt of by the pioneers of the 1940s.

Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 016

Fig. 4.3 Flip-flops through the ages. The three units in the photograph all do approximately the same job within a computer, but represent the state of technology in about 1951, 1961 and 1971 respectively. The large 'door' is typical of early thermionic valve computers; the printed circuit board in Alison's right hand uses germanium junction transistors; the smaller board in her left hand contains small-scale silicon integrated circuit chips. The units, which all consist of flip-flop registers, come respectively from a Ferranti Mark I, a Ferranti ATLAS, and the Manchester MU5 computer. The units represent a speed improvement of about 100 times, at a decrease in contemporary cost of about 20 times.

As far as computer storage is concerned, the technological changes over the years have produced equally dramatic results, though the evolutionary path is not so clear. This is because computer storage

Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 017

covers a variety of ingenious devices, witness to the fact that the search for a suitable memory unit was the prime preoccupation of many early research groups. It was, of course, possible to construct a storage unit entirely from thermionic valves, using two valves per stored binary digit - the so-called bistable or flip-flop circuit. However, this was far too expensive a method, except for the limited application of temporary storage within the central processing unit itself. There were two types of high-speed storage in common use in early computers: delay lines and electrostatic tubes.

In a delay-line store, information was held as a train of impulses continuously circulating round a special closed path. The time taken for an impulse to circulate once round the special path was arranged to be very much longer than the time taken by electrical impulses to travel round the wires in the rest of the computer. One way of arranging the suitable 'long path time' was to take advantage of the fact that sound waves travel very much more slowly than electrical waves, and that it is possible to convert acoustic impulses into electrical impulses and vice versa. An acoustic delay-line store therefore had information travelling as acoustic pulses between a transmitter and receiver placed at either end of an acoustic path. The path itself had to have suitable physical properties, a column of mercury frequently being used. For example, in the Cambridge EDSAC computer each delay line consisted of a 5 ft. steel tube filled with mercury and terminated at each end by a quartz crystal. Electrical pulses representing digits were applied to one crystal which converted them into ultrasonic waves. These travelled 'slowly' through the mercury to be picked up by the second crystal, amplified and reshaped, and sent back to the first crystal for recirculation. Close temperature control of the mercury tank was important, in order to keep the delaying properties of the system constant. A magnetostrictive delayline store was also acoustic in nature, but the resulting unit was cheaper, smaller and more robust than the mercury variety. In the magnetostrictive delay line, electrical pulses were converted into stress waves which travelled down a length of (nickel) wire. The conversion from electrical pulses to stress wave and vice versa depends upon the fact that the sudden application of a magnetic field to a length of wire causes the wire to change dimension. Thus, coils similar to those found in an electro-magnet were used for inserting and recovering digital information from the delay line. Although the magnetostrictive principle was widely known, the first computers to employ it for storage purposes were those produced by the firm of Elliott Brothers in England, as described in Chapter 11.

Fig. 4.4 Storage (or memory) devices for early computers. In the background is a cathode ray tube used in the Williams electrostatic storage system. The long rods in the centre of the photograph are filled with mercury and form an acoustic delay line unit from a DEUCE computer. The plug-in board in the foreground contains a coil of nickel wire for a magneto-strictive delay line in a PEGASUS computer. The small integrated circuit module to the left of centre contains a modern silicon chip which could store more information than the other three units combined.


Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 018

Another class of early storage device depended on the ability of certain materials, notably phosphor, to retain an electrostatic charge for some time. Information was therefore stored as a pattern of charged-up areas on, for example, a phosphor cathode ray tube screen which had been selectively bombarded by a beam of negatively-charged electrons. (The phenomenon of charge-storage in phosphor is also utilised in 'screen persistence', which makes flicker-free television pictures possible.) The difficulty with electrostatic storage devices was that the stored charge leaked away - i.e. was neutralised - unless steps were taken either to retain it or refresh it. Between 1945 and 1950, various inventions, including the selection, barrier grid and holding-beam tubes, attempted to prevent charge leakage. Another approach, used successfully in the Williams tube, was to refresh or regenerate the charge pattern periodically before it had time to leak away significantly. In about September 1946
Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 019

F. C. Williams discovered the 'anticipation pulse' effect,
7 which made charge- regeneration, and hence long-term storage, a relatively simple matter. In particular, the timing of the anticipation pulse gave an early warning that the scanning electron beam was about to arrive at an area of charged phosphor, and the shape of the pulse determined whether this area was currently storing a binary 1 or binary 0; appropriate regeneration could then be arranged in time. Williams cathode ray tube stores were by far the most widely used form of electrostatic storage.

Fig. 4.5 Binary information stored as a pattern of bright or dim dots on the phosphor-coated screen of a Williams tube. Each dot corresponds to an area of electrostatic charge which has coincidentally been made visible by much the same property that produces areas of brightness on a television screen. Although the emission of light was not inherently part of the storage mechanism, it did mean that a Williams tube could give a primitive form of visual display for special-effect computer outputs. The actual display in the photograph shows 32 40-digit binary numbers in a Ferranti Mark I, with an extra 20-digit identification line or 'page address'. (This page address was the germ of an idea which later led to the invention of virtual addressing.)
Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 020

Fig. 4.6 A magnetic drum storage unit from a DEUCE computer (1955). The magnetically coated drum - the dark cylinder in the centre of the container - spins rapidly and information is recorded on its surface by read/write heads at either side. In this particular system the head-structures could be moved vertically to give the effect of many recording tracks.


Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 021

Although delay-line storage was more popular for use in early British computers, Williams tubes narrowly outweighed delay lines as a choice for storage device in early American computers.
8 An extremely desirable property of a fast store is that it should be random access - that is, the time to access information at any one location should be the same as the time taken to access any other location. This 'independence of location' is very important when accessing programs, because of the out-of-sequence branches (control transfers) which they contain. A disadvantage of early delay-line stores was that they were sequential access devices, though this inconvenience could be minimised by a combination of actions taken at the hardware and software level. Williams tubes were random-access devices, and were also more suitable for the design of parallel (as opposed to serial) computers.

A third class of computer storage device, still widely used, is based on the ability of certain materials to retain (i.e. store) their state of magnetism. Thus magnetic tapes, films (complete with sprocket holes), discs and drums all basically record information on a magnetic surface ready for subsequent 'playback'. ('Writing' and 'reading' are the terms normally used in place of 'recording' and 'playback', but the principle is much the same as that used in an ordinary audio cassette tape recorder.) Storage units based on magnetic surface recording inevitably depend for their access-time on the speed at which the read/write head moves past the surface. This speed has always proved much too slow for the potential rate of information flow demanded by a computer's central processor. Thus in some early machines, as in most present-day ones, magnetic surface recording is useful mainly as a back-up to a faster unit. In this role its advantage lies in its comparatively low cost per stored bit. This economic relationship is of course relative. In absolute terms the actual capacity (quantity of information) contained in both the fast and the slow section of a computer's store has typically increased at least a thousandfold between 1950 and 1980.

Historically, the demise of delay lines and electrostatic tubes for fast storage units came between about 1954 and 1958 with the introduction of the magnetic core store, an important American invention associated with J. W. Forrester of the Massachusetts Institute of Technology. In this store, each binary digit is remembered as the magnetic state of an individual ring or 'core' of ferrite material. An array of many such cores, suitably threaded by selection and sensing wires, goes to form the complete core store. These stores had access times much closer to those required by central processors. For general- purpose computers, core stores lasted well into the 1970s. Nowadays, stores based on semiconductor integrated circuits have replaced cores as the common form of fast storage device, except in certain military applications where cores are preferred for their relative insensitivity to power supply interruptions.

Early British Computers, 4 - THE TECHNOLOGY OF EARLY COMPUTERS, starting page 022

Having gained an idea of the electronic techniques in the 1940s, let us return to the trail of the first proper stored-program computer.

Fig. 4.7 A 'stack' of magnetic cores, compared in size with a match box. This particular stack is made from 48 planes, each plane wired with a mesh of 32X32 tiny rings (or cores) of ferrite material, enabling the complete unit to store 1024 48-digit binary numbers. This core store dates from about 1959.


Go to next chapter

Return to On Line Documents
Return to "Early English Computers" table of contents



Permission to publish from Simon Lavington here
August 8, 2001
Dear Ed,

Thanks for your interest in my book 'Early British Computers'. This was published in 1980 by Manchester University Press and, because of personal links with Gordon & Gwen Bell, Digital Press co-published the book for the American market. The book was also translated into Japanese & published in Japan a couple of years later.

The book is out of print in the UK. The Science Museum (London) keeps encouraging me to write a second, much enlarged, edition and I am at last getting down to the task. As part of this exercise, and at the request of the Science Museum and the Computer Conservation Society, I recently wrote a booklet on the Ferranti Pegasus computer to mark the installation of the last remaining working Pegasus in a special gallery in the Science Museum in May of this year. See:

http://cswww.essex.ac.uk/pegasus/

I am currently doing a bit of research on UK defence-related computers in the period 1945-54, to fill in a few more gaps in the first edition of

Early British Computers.

I own the intellectual rights to the book. The original publishers, Manchester University Press, have agreed to my plans for a second edition. They will probably not be interested in publishing the new book, so have given their blessing to my approaching other publishers - (eg the Science Museum in London).

Having said all that, in principle I like the idea of making the original version of Early British Computers available from your On-Line Document page of:

http://ed-thelen.org/comp-hist/on-line-docs.html

I was impressed that my favourite architecture book, Bell & Newell, is available from your site. This contains my favourite Pegasus quote on page 170, namely:

"Pegasus has the nicest ISP processor structure discussed in this section-perhaps in the book. It is included because it is probably the first machine to use an array of general registers as accumulators, multiplier-quotient registers, index registers, etc."

So ..... How do we procede? What are the arrangements for copyright? Do we have to re-negotiate permission to reproduce photos? ...

Regards, Simon

Professor Simon Lavington,
Department of Computer Science,
University of Essex,
Colchester CO4 3SQ,
UK
Direct dial: +44 (0) 1206 872677
Fax: +44 (0) 1206 872788;
e-mail: lavis@essex.ac.uk