Peter Thorne, CSIRAC computer technician
Some users had long sustained projects on CSIRAC, like meteorologist Dick Jenssen. Dick developed early programs on weather forecasting, and incidentally wrote computer games.
One was called ‘The way the ball bounces,’ which attempted to predict, from your pattern of moving a switch up and down, whether you would move it ‘up’ or ‘down’ next time. You had the choice of one or zero – yes or no – and the computer would look at your previous behaviour and make a prediction. This might have come from his work predicting weather patterns.
It was a simple computer game, using the lights and switch on the console. Kay Thorne was most likely to trick the computer and beat it: around the time it seemed to be predicting the pattern of her choices, she would do something quite different.
The real engineers
Using the computer didn’t mean sending in your program or sitting remotely: people would book time days or weeks ahead, then come in and sit at the console to ‘drive’ the machine. If the computer didn’t work at weekends, I would ring the users to tell them, or if it looked like a simple problem I would attempt to get it working and keep it going while they did their work.
I was not very good at fixing it, so if something very complicated went wrong, it inevitably waited until Monday for Ron Bowles or George Semkiw to attend to, because they were the real engineers and I was pretty much an amateur.
The power system for the computer was such that there wasn’t much excess capacity. CSIRAC had a mean time between errors of probably half an hour. It was sensible to write out intermediate results frequently, so that one could recover from a failure without total loss of work to that stage.
We had a tea-room and were all keen tea drinkers. On one occasion somebody plugged in the electric jug on the 59th minute of somebody’s program and the power went off, because the jug overloaded the system. Unfortunately they had written a program in such a way that they didn’t get any intermediate results, so all their work was lost. They had waited a week or a fortnight for the opportunity to use the computer and now all their work was gone – because someone had plugged in the jug!
Quick on his feet
Ron Bowles was a central character in the operation and maintenance of the computer, and he was very quick on his feet. We used to punch out paper tape, particularly later on the narrow channel variety, and take it across the room to put into things called Flexowriters –like electric typewriters with a paper tape reader on them – which would then print out the results from the paper tape. So it was a form of off-line printing.
The thin tape frequently folded over and jammed - you could hear the note of the tape readers change, as it started to happen. But Ron could get from one side of the computer room to the other very quickly and stop them before the paper tape actually tore. He did this by running and then skidding on the polished floors to a halt beside the printer, just in time to switch it off.
Starting the drum
It’s worth recording the technique used to start the drum. The drum was quite a large diameter disc with the heads facing a flat surface, driven by a V-belt coming from an electric motor. This was around the back of the computer.
It turned out that the inertia of the disc, like a stationary flywheel, was too much to get the motor started. So a technique was developed (I think they had previously tried a clutch or flexible couplings) of driving the motor through a V-belt which had a degree of slip on it, so if the motor varied in speed the flywheel effect of the big disc would ride through the variations.
One would go around the back of the computer, turn on the motor (which would come up to speed) and apply a large screwdriver to the side of it, flat on the V-belt, to increase the tension to bring the disc up to speed. Then the disc would stay at that speed.
But if you weren’t very good at this, or the belt was loose, the disc might never get up to speed, and it could stabilise at a lower speed with a bit of slip in the belt. In this case data could be written on the disk in such a way that you couldn’t erase it under program control. You would then have to demagnetise the whole disk. The engineers weren’t too happy if you had an inexperienced applicant of the screwdriver, who didn’t know how to bring the disk up to speed.
Major electric shock
With CSIRAC itself, the main power supplies were high voltage ones for the vacuum tubes, and there were strips of wires – busbars – running along the back of the computer. If you put your hand on those, you would be across a 300 volt power supply and receive a very major electric shock. But people always used to open the cabinets and leave them open, and I didn’t see anyone get a shock from it.
Frank’s famous ‘rubber donger’
The vacuum tubes were relatively unreliable. We did have the problem that little metal particles would get between the electrodes and short them out. So the technique developed was to use Frank Hirst’s valve tester, which was a big rubber stopper on the end of a stick - Frank’s famous ‘rubber donger.’
You would run a diagnostic tape that was checking the computer, then open the cabinet doors and walk down behind the cabinets and go ‘bong, bong…’ Sooner or later you would hit a valve and the program would stop, so you would then start the program again, hit the valve again, and if it stopped again, you would conclude that this was a suspect valve. So you would take that one out and put in a new one. There was also a standard valve tester: you would plug a valve into it if you had any doubt about its integrity.