Larry Atherton's work on my 9100Bs continues apace. The biggest problem we've had is that the earlier of the two units would fail to run a program (the standard diagnostic) until it had been turned on for 5-10 minutes, after which it would work perfectly.
Board-swapping with the other 9100B isolated the problem to the "left" board of the two-board "C Core set", which controls access to the core memory.
Selectively heating individual components on the board revealed two problems: first, a set of 3.01 Ohm (yes, that's correct: 3.01 Ohms!) resistors were out of spec. Second, one specific transistor would cause the problem when cooled and fix it when heated.
Since the transistors HP used in the 9100 have odd part numbers that don't correspond directly to any modern parts, the only way to find a replacement that works is to experiment. First, Larry removed the problem transistor and measured its characteristics when warm, then installed Tektronix sockets so he could swap in new parts without having to solder and unsolder each time. The image above shows the new transistor installed.
The new part reduced the "warm up time until proper operation" to about 45 seconds, but the machine would still fail if you turned it on and immediately tried to run a program in memory.
Looking at the -15V supply, Larry noticed noise on the line. Not a lot, but still. So he added the large blue filtering capacitor you can see in the first image...and the problem went away completely! You can now bring the first 9100B in from an overnight cool down in an unheated garage, turn it on, immediately press END, CONTINUE, and it works!
Now onto cosmetics. The backlights for the register labels had failed years ago, and heat from the bulbs had, over time, slightly warped and scorched the piece of photographic film that comprises the actual label. Larry cleaned the film as best he could and used LEDs for the backlights. Clear plastic tubes added act as light pipes to make the illumination more even, and a thin gold film at the bottom colors reflected light to more closely match the original orange color.
While the final appearance is not identical to the original, it's still amazingly good.
As you may recall from a previous posting, the black plastic socket at the neck of the CRT was cracking. Larry fabricated an extra support from perforated aluminum, visible here at the very back of the CRT, to support it. The front of the CRT is supported on new rubber buffers for shock protection.
OK, everything works. Now it's time to do something about the beat-up chassis. This machine was bought in 1969 by a trucking company and obviously used heavily for many years. Time to paint!
The first 9100B is now complete. The only sign of age is the slightly-dimmer-in-the-center CRT. With a fully functioning card reader, adjusted CRT, completely rebuilt keyboard, and other repairs, this represents several weeks of work on Larry's part...
Edited: 23 Nov 2011, 7:32 p.m. after one or more responses were posted