The EDSAC Replica Project aims to reconstruct one of the most important early British digital computers.
Designed in 1947 by a team lead by Maurice Wilkes, the original EDSAC computer operated for almost 10 years, starting from its first successful program run on 6th May 1949, at the Cambridge University Mathematical Laboratory.
We are now initially targeting to have a working reconstruction of EDSAC as it was in 1951 (when it was in everyday service at the University of Cambridge Mathematical Laboratory) operational by the end of 2019.
The story of the project is being told through a series of short videos.
The EDSAC Replica Project is a registered charity and is affiliated to the UK’s Computer Conservation Society.
The Electronic Delay Storage Automatic Calculator (EDSAC) was originally built in in the Cambridge University Mathematical Laboratory by a team lead by the late Professor Sir Maurice Wilkes immediately following the Second World War.
It is generally accepted that the EDSAC was the first practical general purpose stored program electronic computer. Other, earlier machines were either dedicated to a single task (e.g. Colossus and code breaking) or were purely experimental (e.g. the Manchester University "Baby" Small Scale Experimental Machine).
As head of the Mathematical Laboratory, Wilkes’ brief was to provide 'mechanical' aids that would assist mathematicians, scientists and engineers at the university to perform complex and time-consuming calculations. He had observed research workers doing laborious computations with the aid of mechanical desk calculators and mathematical tables. His prime motive in building EDSAC was to provide them with faster and better facilities.
Wilkes’ vision was to create a computer which, unlike earlier machines, could be used by a wide range of researchers in the university instead of the few highly specialised ‘acolytes’ who tended the other early electronic computers. He wanted to create a computer that was accessible and practical, rather than to push the boundaries of technology. To reach that goal, he adopted very conservative design principles and the result was a reliable machine that did useful and significant work through its life. It typically operated for 35 hours a week. During the day there were engineers on hand to deal with problems that arose. Approved users could work on the computer overnight, but if it broke down they had to wait until morning before it would be investigated.
Like all computers of its time, EDSAC was based on thermionic valves which Wilkes knew well from his wartime work on advanced radar systems. He sketched out the main elements of the design during a five day voyage from the USA to the UK while returning from a seminal 1946 conference of American computing pioneers at the University of Pennsylvania.
Construction was led by Bill Renwick, appointed by Wilkes as Chief Engineer. A team quickly grew around Wilkes and Renwick as they refined the design and gradually brought EDSAC to life. Another important contributor at this stage was David Wheeler, Wilkes' research student, who was responsible for many of the features that made the machine practical for ordinary users.
EDSAC ran its first program on 6th May 1949 and was soon pressed into service to support research in the university. It provided a computing service for over nine years, until it was superseded by EDSAC 2, built by the same team.
During that period a substantial number of users had their research transformed and their horizons extended through the increase in computer power that EDSAC and EDSAC 2 gave them. Among these were future winners of three Nobel Prizes - John Kendrew and Max Perutz (Chemistry, 1962) for the discovery of the structure of myoglobin, Andrew Huxley(Medicine, 1963) for quantitative analysis of excitation and conduction in nerves and Martin Ryle (Physics, 1974) for the development of aperture synthesis in radio astronomy. All acknowledged EDSAC in their Nobel Prize speeches. You can see Professor Huxley’s EDSAC programming manual on display in the gallery.
EDSAC and its many uses won recognition and financial support from outside sources. In particular funds were provided by the catering company J Lyons, which went on to build the LEO I computer, the world's first business computer, based on the EDSAC design,
EDSAC was modest in terms of modern-day computers. There were only 18 operation codes and initially just 512 words of memory, later extended to 1024. Instructions were executed at a rate of approximately 650 per second. Input was by punched paper-tape and output by teleprinter.
Although today’s PCs operate can perform calculations millions of times faster than EDSAC, EDSAC was a huge improvement on what was available at the time - human calculators using mechanical desk machines. It has been estimated that EDSAC introduced a 1,500 times productivity increase, and transformed the progress of research throughout Cambridge University. Researchers became able to solve problems that were previously considered either impractical or impossible.
EDSAC 1951 Film
Maurice Wilkes' 1976 commentary on the 1951 film about how EDSAC was used in practice:
The goal of the project is to build an authentic working reconstruction of the EDSAC computer and to run a program on it as was done when the original machine first went into regular service for users in 1951.
Our target is to have it running fully by the end of 2019.
The project has the following supporting objectives:
To provide a tangible demonstration of the achievements of the Cambridge pioneers.
To celebrate the achievements of British scientific and technological contributions to the early development of computing.
To assemble archive material relating to this historic machine and recapture the knowledge of the pioneers who developed and those that used it.
To build as authentic a reconstruction as possible taking into account the availability of components and materials and modern safety standards.
To undertake the work within the spirit of the designers and technology of the time.
To produce a working artefact of exceptional educational value to students and the general public.
To be an exemplar of British engineering and encourage new students to take up computing and engineering.
To demonstrate the working machine to the public as often as is practical once it is built and commissioned at The National Museum of Computing.
To acquire sufficient spares to keep the machine running for the next 25 years.
To train a new generation of volunteers, unfamiliar with 1940s valve technology, to run, demonstrate and maintain it for the foreseeable future.
The EDSAC Replica Project is a UK registered charity number 1145823. Its trustees are Professors Andy Hopper and Peter Robinson (appointed by the University of Cambridge), David Hartley and Kevin Murrell (appointed by the BCS, Chartered Institute for IT) and Dr. Hermann Hauser and Dr. Andrew Herbert (appointed by the Hauser-Raspe Foundation).
The project is affiliated to the UK Computer Conservation Society (CCS).
In October 2011, Dr Andrew Herbert was appointed Project Manager. Andrew had just retired as Chairman of Microsoft Research for Europe, Middle-East and Asia. He has a PhD in Computer Science from the University of Cambridge, and worked with Maurice Wilkes on the Cambridge CAP Computer in the 1970s.
Sponsors And Supporters
While he was Chairman of the Computer Conservation Society (CCS) in 2010, David Hartley suggested marking the significance of the early computer development done at Cambridge University by creating a working replica of EDSAC. This followed the success of the Manchester `Baby' rebuild project at the Manchester Museum of Science and Industry.
David gained support for this project from Hermann Hauser, who later provided initial project funding through the Hauser-Raspe Foundation. Additional funding was provided by Google and a number of private individuals.
We would also like to acknowledge the following companies for their cooperation and assistance.
Teversham Engineering, Cambridge - sheet metal design and production
Marshall Amplification, Milton Keynes - component donations and building almost 20 of our memory unit chassis to a very high standard
The EDSAC Replica Project Team
The EDSAC reconstruction is being built entirely by volunteers, apart from some specialist work such as sheet metalwork which is being done semi-commercially.
The project's technical leader is Chris Burton who also lead the team that built the replica of the Manchester 'Baby' machine.
The initial meeting of volunteers was held in March 2012, although substantial research work, to locate archive material and establish the feasibility of the project, had already been carried out by some key members of the CCS.
The Initial Meeting of the EDSAC Volunteers
Information about EDSAC was found in the Cambridge University Library and the Cambridge University Computer Laboratory library, as well as in the possession of individuals.
The original EDSAC was scrapped in the late 1950s and only three identical examples remain of one specific ‘chassis’ type, out of a total of about 65 different chassis types, and 140+ individual chassis in the original. And no definitive record exists of the final ‘as built’ design, either in terms of the circuitry or the physical layout of the machine. The archive material that we have found so far spans the whole timescale of the original EDSAC project, from Wilkes’ first thoughts to final decommissioning.
EDSAC was repeatedly modified over its almost 10-year lifetime. Our aim is to create a replica of EDSAC as it was sometime in 1951, about two years after it ran its first program (on 6th May 1949) when most of the initial problems had been ironed out, and it was providing a reliable computing service for Cambridge University.
The project has required complex detective work in order to come up with both the logical and physical designs that we can be reasonably confident reflect EDSAC as it was in mid-1951.
Where possible, we have built the replica using authentic materials and components - with one major exception. EDSAC’s memory was constructed from ‘delay line tubes’ filled with mercury; these exploited the relatively slow speed of a sound wave in the liquid metal. Expense and operating concerns will prohibit an authentic reconstruction of these tubes. Our alternative is to use magnetostrictive steel wire delay lines, a technology employed in computers that followed EDSAC, and which follow similar physical principles and method of operation as acoustic delay lines, and connect to identical circuits to those used by the original mercury delay lines.
In the 70 years since EDSAC became operational, technology has changed out of all recognition. Because of this, original components (or even their later replacements) have become very hard to find. In the “You Can Help” section, we will post details of the less common components that we still need to obtain if we are to build a close replica of the original machine.
Location and Construction
The EDSAC replica is being built at The National Museum of Computing (TNMOC) located on Bletchley Park, the UK’s WWII code-breaking centre and undoubtedly the spiritual home of computing in the UK. The new EDSAC gallery was home to a Colossus code-breaking machine during WW2.
It is being built entirely by volunteers, many of them in their 60s or 70s (some even older!) many of whom have experience of old valve-based technology. In time, some of them will train much younger volunteers, so that the EDSAC replica can be maintained in an operational state for decades to come.
As is TNMOC’s policy, it will not be a mere static exhibit. Once completed, it will be powered up and running as often as is practical. Particular efforts will be made to involve and inspire schoolchildren – the next generation of technologists and engineers!
Right now, if you visit TNMoC you may well see our volunteers at work debugging and commissioning EDSAC; we encourage you to talk to them, ask them questions and find out more about the project and the history of this very important computer.
Helping the EDSAC Project
There are several ways that you may be able to help.
Support the project financially
Volunteer to help with the project
Locate or donate components
We have not yet raised all of the money needed to complete the project. It is high profile and regularly attracts very positive national and international media coverage.
If you or your organization are interested in becoming a financial sponsor of this unique venture, please contact Andrew Herbert by emailing email@example.com
Volunteer - Join The Team!
The EDSAC replica is now close to 100% physically complete, and we are in a very active commissioning phase – debugging the individual chassis and testing sub-systems. This inevitably involves making changes to some circuits and individual components. Probably the most challenging issue revolve around shaping and timing of pulses.
Most of our team of volunteers are well past retirement age, as only people of that vintage have the necessary background in designing and building valve-based circuits. We now need to recruit a younger generation of volunteers with electronics skills who can work alongside the current team to “learn EDSAC” and be able to maintain it in the longer term and perhaps also to demonstrate it to the public as a living, working historical artefact.
We also need.
Someone familiar with mechanical CAD and/or circuit schematics CAD to pull together all of the work that has been done so far and ensure that we have is fully and consistently documented as a record for future generations.
A technical author to produce operating and safety manuals etc would be extremely useful.
If you have these skills, and have sufficient spare time and enthusiasm to devote to the project, please email firstname.lastname@example.org
Help Us To Find Components
If you have components which you think the project could use and wish to donate them, please email email@example.com - but please check the lists below first to see if we need them. Dealers with stock to sell are also welcome to contact us.
When you look at a photograph of EDSAC, the overwhelming impression is of rack after rack of thermionic valves (vacuum tubes) - about 3,400 of them! However, sourcing valves was not our main problem. Even though valves were superseded by solid-state electronic devices from the early 1960s, valve production continued for decades, as spares were needed for existing equipment. And indeed many valves are still produced today. Specialist dealers hold large stocks of the common types of valve, often over 60 years old and still unused!
Our main problem was finding authentic ‘passive’ components such as resistors and capacitors and which are still 'in spec'. Manufacturers made much progress in the 1950s and later in miniaturising these, and today’s components are minute compared to those with equivalent performance characteristics from 70 years ago. It is just not feasible to use old recycled resistors and capacitors as their rated values tend to drift over time, and they also become much less reliable. Just as Maurice Wilkes did in 1949, we have used the best modern day components we can obtain. If there is a choice of physical size, we use the largest ones we can get, as these look closer to the originals. This does not compromise the authenticity of the EDSAC replica in any way – a modern day 100 ohm resistor performs exactly as a 100 ohm resistor form 1949.
Another problem is getting hold of the physical components which allowed the electronic components to be interconnected, such as terminal strips. We had to have over 1,000 feet of these made by a specialist manufacturer to very close to the original specification. Valveholders are also hard to find; while many types are still manufactured they are most often not of the authentic dimensions. We are still trying to source the following components:
B9G valve-holders. Maximum diameter of raised part of socket 1.5 inches (38.1mm). Spacing of fixing hole centres 1.8125 inches (46.04mm). These dimensions are important as the metal chassis are being made to authentic dimensions. We have had to use some modern B9Gs of Chinese origin, but originals are much more authentic and suitable. We would still like to obtain more original B9Gs if possible.
Although we now have most of the valves we need, donations of additional valves are very welcome as, once commissioned, we intend to keep the EDSAC replica running until at least 2040. The following are the main valve types used in EDSAC. Each type has several different designations. The number in brackets indicates approximately how many of each type EDSAC uses. The larger the number, the more spares we would like to have
EF54, CV1136, VR136 (1150)
EB34, CV1054, VR54 (750)
EA50, CV1092, VR92 (600)
EF55, CV173 (200)
CV1116, VR116 (30)
EF50, CV1091, VR91 (20)
ECR60, CV1097, VCR97 (6) (6" diameter cathode ray tubes)
12E1, CV345 (10)
CV685, VR150/30 (5)