Badly Designed Hardware

Despite the name, Walton Radar Systems did more than radar — it did all sorts of custom systems (1 offs and more) for all sorts of applications. One of the more interesting areas was replacement peripheral systems for obsolete legacy computer systems at various nuclear power stations.

That little picture above shows the extent of the interface timing information Walton was given to help it design replacement equipment for the Ferranti Argus 400 computers at Hartlepool and Heysham nuclear power stations. I've re-drawn it of course, but it's pretty accurate (it's not as though there was a lot to forget) — even down to having the appearance of a slightly blobby photocopy.

Apart from the nominal width of the two strobe signals  (measured where, top, bottom or half-way?) there was no other information at all — no rise/fall times, no timing tolerances, no voltage tolerances (I don't think there was even a voltage scale, you just had to know that the nominal voltage was 4.5V). No nothing... and from this Walton was supposed to design a display replacement system.

The existing computer displays in the control room were cursive displays, which is to say that information was drawn on the screen by moving an electron beam as one would move the tip of a pen, lifting the tip as necessary between repositionings. The problem (the problem?) was that the monitors required careful and frequent calibration lest the inevitable drift in the gains and offsets of the various analogue amplifiers turn Roman text into Elvish. (The borderline between Roman and Elvish script was a little hazy; I distinctly recall seeing reactor operators reading screens illegible to me without any apparent difficulty.) Another problem was that the writing was spidery, unpleasant to look at, and, of course, everything was monochrome.

The Central Electricity Generating Board's R&D division had investigated the possibility of replacing these displays with modern raster graphics — gasp — and maybe even adding colour. They decided it was impossible.

Walton's technical director, Eddie Richardson, disagreed and for a modest consideration offered to build a prototype display replacement. In hindsight, Eddie's trick was blindingly obvious: whilst there was no way to decode the signals that drove the monitors and turn them into raster graphics, if the graphics instructions were intercepted, the mapping from cursive stroke programming to characters and lines etc. was pretty straightforward. All we had to to was intercept the data from the Argus data highway (an old fashioned  term that lost out to bus in the evolution of English, even though a highway is the thing along which vehicles travel and a bus is a vehicle; not doubt the computer usage of bus is due to the American verb to bus).

Given the paucity of the reference information available the scope for design error was enormous, so Geoff Allan (Walton's Chief engineer) and I decided that we should work separately so as not to compound  our inevitable errors: he would design the interface hardware and I would design an Argus 400 emulator; we would both take our own measurements from a real system and interpret documents independently. — and we wouldn't talk to each other about the details of each our designs. The idea was that, where there were ambiguities in specifications or measurements, being rather contrary people, our conflicting interpretations would hopefully lead us to effort directed at reconciling our differences and clarifying how the Argus actually worked.

Hence The QARGIF — the Q-bus to ARGus InterFace. Walton's conservative design principles were thrown largely thrown out the window as far as the QARGIF was concerned, at least at the actual interface level: I made the bus drivers push and pull too much current, completely failed to provide sufficient 0V connections and refused to allow any noise reduction measures in the cabling — no twisted pairs, no shielding. It was the noisiest, most potentially unreliable and nasty interface I could make.

When Geoff's prototype interface was ready we hooked the two up and of course the system as a whole failed to work beautifully... but we resolved the issues and when the finished prototype was taken to site and installed in an Argus it worked first time. 

Subsequently Walton then replaced all the displays, and with that success under its belt set about replacing the aged Burroughs disk drives with solid state memory... and thereby hangs another tale. But, the QARGIF was there through it all, and despite the occasional apparently bizarre behaviour of a system under test, in about 20 years of use it has, apparently, never gone wrong.

Not bad for a piece of badly designed hardware.