Is this why Robinson didn't work?
But then we had a problem: the original circuit diagrams suggested that the output from the photo-cell amplifiers were connected to the logic which did the double-delta comparisons. But, - and it is a big but - there was a serious mismatch between the output voltages and current abilities of the photo-amplifiers and the delta logic. We tried to run and and this confirmed that there was a problem.
Much head-scratching followed and a plan was devised: a new chassis with 20 valves on it (one for each channel) was designed and tested. Partial success but still more drive was needed. A re-work using heftier valves showed that this would now stand a good chance of working. See photo on right.
This raises some very interesting ideas. It is well known that Robinson didn't really work very well even on a good day (and it was Tommy Flowers who understood this and started thinking about what would become Colossus). Was this the actual reason that it didn't work properly? The photo-amplifiers and the delta logic were designed by different groups of people. Could it be that in the heat of battle the interface was not done properly; some things were overlooked; and because it seemed to work sometimes that was good enough? We shall never know of course but let's see how well our machine works.
Problems are to be expected with a project like this and we have not been disappointed! The new chassis appears to work well, but this pointed to another problem that we already knew about.
The mechanical alignment of the paper tapes has to be fairly exact but our previous rig was just not good enough. A kind engineering company, The Taylor Kellar Partnership has designed and built a superb new tape-drive and guiding system which looks very promising. Many thanks to Paul Kellar and Mike Hillyard for their tireless efforts. See photo left.
The light sources we were using were a little too bright and light was spilling between channels. This has been fixed by using less bright tungsten bulbs and culimating lenses which steer the light into narrow beams, just wider than the tape. The mechanical drive of the tape was always questionable but the new solution seems pretty good. The main difficulty now is how to join the ends of a tape into a loop. In the 1940's these used a Bostik glue and a heated block. We have found that PritStick works pretty well - although it still causes chuckles when the tape join comes apart and the tape flies off.