Model F/XT: Non-working keys - Matrix Details?

[Coeus]

I have a Model F from a PC/XT. It had been working well for quite a while but some keys were beginning to get unreliable, needing extra force to get the keystroke recognised. My theory was that some dirt had got in between the flappers and the PCB so the flapper wasn't getting close enough to be the PCB. Sure enough, when I took it apart, there were signs of dirt on the PC.

So, I cleaned the PCB, cleaned the metal plate, cleaned the flappers and cleaned the keycaps, dried everything and put in all back together. Unfortunately, though the affected keys are now different, there are still keys that don't work and, as they are grouped, for example in a diagonal line, I am now beginning to wonder if there is a fault with one or more the matrix lines, especially the column lines.

Does anyone know if the matrix is documented anywhere, i.e. which keys are on which rows and columns and, preferably, which pin on the microcontroller each row and column connects to? This seems the sort of thing that might have been on the Wiki but I gather that has been down for a while.

Any other ideas for fault-finding this? One idea I had was to probe each of the lines of the matrix, both rows and columns, with an oscilloscope and compare working ones with non-working. I can also, obviously, check for shorts etc. Anything else?

[Coeus]

I got the oscilloscope and a multimeter out and sat down to start testing, connected the keyboard and checked which keys were not working, and the number was now reduced to just the 'F' key. I think there may have been something affecting one of the columns but it may have been residual moisture, even though everything looked completely dry.

Unfortunately, removing the keycap, wiggling the flapper and putting it back didn't fix the 'F' key and I had to take the sandwich apart again, make sure the flapper was correctly place and then put it back together. All keys now working.

[Falkenroth]

Figured you'd end up having to tear it apart again. Been there done that. Glad its working for you now.

[AndyJ]

Every time I've taken an F apart, some of the keys didn't work when I put it back together. It usually took two or three tries for all the keys to work.

I'm pretty sure IBM did all the assembly and testing by hand, judging from the initialed labels on some boards. That would explain why an F/AT board cost the equivalent of $650 2025 microbux back in 1984.

The Model M has fewer pieces due to the barrel plate instead of individual barrels, but thermoforming the plastic rivets meant if it didn't work properly, all you could do was break off all the rivets and try again with a new barrel plate. But actual manufactuing cost on the plates may have been low enough to mean a net savings on labor, even if they had to be destructively disassembled after failing QC.