Testing and commissioning
As mentioned earlier, the University of Manchester had offered a site to build the replica in the big Computer Hall, and the project would rub shoulders with the huge Cray and ICL service mainframes and disk farms. One whole wall of the hall is plate glass alongside a major internal walkway, and our work took place in good view of passersby. This provided good publicity during construction and enabled anybody to see what was happening even when members of the project team were not present.
The power supply system, the racks, and the feasibility rig were installed in the Computer Hall in early 1996. As team members finished a chassis, they brought it to the university, usually on a Tuesday Team Day, and added it to the growing replica. Chassis were interconnected by wiring between the terminals on the panels. In the early stages, the feasibility rig could be used as a source of signals so that a chassis could be partially tested. The pattern of the team’s working together on the machine one day a week was successful, giving a weekly target for the many offline activities as well as affording people time to attend to the rest of their lives.
The most critical items in the whole machine were of course the three CRT memories. The feasibility rig had given us confidence, but to paraphrase Kilburn’s words during one of our conversations, What about the hurly-burly of the replica? Our most experienced engineer in the team, Charlie Portman, took on the task of achieving working memories. A couple of special chassis were made so that the memories could be interfaced to the printer port of a PC (see the “Mixing Old and New Computer Technology” sidebar). A series of special programs were written to run on the PC in order to exercise and diagnose the memory operation. At this stage it was a great joy to have Kilburn and Edwards come in and sit with us in front of the CRT boxes, helping to get those to work just as they must have done 50 years before. It was chastening to realize that they had achieved the task then without the benefit of our modern test equipment, and especially without the benefit of a personal computer. Although the description of the CRT memory operation given elsewhere seems straightforward, as an analog electronic device it was tricky to adjust. Controls for brilliance, focus, astigmatism, high-voltage supply, deflection voltages, amplifier gain, threshold level, dash width, dot width, and strobe pulse timing all interacted. The secondary emission behavior of the screen phosphor is not uniform in the early tubes used in both the SSEM in 1948 and in the replica, and furthermore they are susceptible to minute areas of zero emissivity, known as phoneys, where a bit cannot be stored. (Phoney is a sort of slang word used by World War II pilots meaning a radar echo that was not a real target. The term carried over to the SSEM team in 1948.) By 1949, the original tubes had been replaced by tubes manufactured in especially clean conditions to eliminate those problems. The team made an early decision to separate the work on the CRTs from the rest of the machine, and to run the two activities in parallel. To enable the main machine to progress, dummy stores (that is, working memories that functionally took the place of the CRT storage subsystem) were built, using semiconductor components, so that the whole machine could be made to work without CRTs (see the “Mixing Old and New Computer Technology” sidebar). These dummy stores have proved to be extremely useful and remain in the replica— in small boxes that hang in the back wiring—to be switched in to help with fault diagnosis and isolation.
By summer 1997, the whole machine was beginning to work, and attempts were made to run programs. One team member, Keith Wood, had written a program that moved a pattern across the store, viewable on the monitor CRT. This stunning program, which we believe had not been attempted in the original machine, vividly shows a modern audience that this really is a universal computing machine. The structure of the program was so different from any program we know about from 1948, all of which solved arithmetical problems, that it seemed to nicely illustrate Turing’s perception of a universal symbolmanipulating machine.
As part of the lead-up to the 50th anniversary celebrations, the university organized a worldwide competition to see who could write the “best” program for the SSEM, with suitable prizes to be awarded. To support the competition, a programming manual was written, along with a simple simulator for entrants to use to test their entry prior to submission. 14 Nearly 130 entries were received from about 20 countries. Almost all the entries worked on the replica, and many exploited the graphical nature of the monitor CRT to show some sort of moving graphic. The winning entrant was a student from Japan, who personally ran his program on the replica during the afternoon of 21 June 1998.
Over the winter of 1997–1998, several months were occupied with just running the computer and making observations and adjustments to improve the reliability. The feasibility rig, no longer needed as a source of signals, could be removed from the scene. The PC interface was tidied up and adapted so that we could instantly load a program into the SSEM memory from a database of programs in the PC without having to always go through the slow and errorprone procedure of entering individual digits from the switches and push buttons.
Preparations were made to move the machine to the MSIM, where it would be put on display and where the anniversary celebration would take place. The machine was to be installed in the 1830 Warehouse, the world’s first railway goods warehouse and actually built in 1830. The brick walls and wooden board floors were reminiscent of the old Magnetism Room and so entirely suitable to house the replica (see Figure 6). The transportation plan was to separate the system into individual racks by labeling and disconnecting one end of each interconnecting wire. The racks would then be moved using an air-cushion vehicle for transport. It was chastening to see the machine completely dismantled at the university ready for moving, and with only four months to go to achieve our goal.