I've bought a Model F XT keyboard last year and really enjoyed typing on it. I'm using an external handmade Atmega32u4-based Soarer's Converter. The board itself is in a good shape. The original foam is intact. I had to disassemble it once when I received it in order to reattach the spacebar.
Unfortunately, there's a major "glitch": After some time of using it - there is really no way to predict when - I would suddenly get spooky keypresses. Either, a keypress results in the wrong keycode or the key keeps repeating more than once (ie. is registered as "pressed" even though it is physically released) or even indefinitely. It can even happen without touching anything at all!
What I've already tried to fix it:
- Made sure that the backplate is grounded. The shield is connected all the way to USB connector's shield now. The MCU appears to connect USB Shield and GND, though.
- Tested all connector pins end-to-end (up to the converter) to make sure there is no broken cable or missing contact.
- Exchanged the converter MCU. I tried two Atmega32U4 breakouts.
- Tried several pull-up resistors to CLK and DATA. I remember trying at least 1KOhm and 4.7KOhm.
- Tried Soarer's firmware and several TMK converter firmwares.
- Tried debugging the signals via a logic analyzer. I wasn't able to reproduce the problem as long as I had the logic analyzer attached. And even if it occurred, what would I expect to see? If signal strength/shape is responsible, I'd need an oscilloscope to see that.
- 2x"124611", 56pF/6V, Electrolyte (not sure, but it's axial mounted)
- 1x"124501", 22pF/15V, Electrolyte (sot sure, but it's axial mounted)
- 1x84pF +/-20% 25V (This is a color-coded axial-mounted component and I'm not even 100% sure it is a capacitor. You can see it right next to the MCU in the photo.)
- 1x50pF +/-2% 300V, Mica, CORNELL DUBILIER
- 2x200µF +/-0.5%, 100V, Mica, CORNELL DUBILIER
I can also find 500V +/-1% Micas for d). I guess that the tolerance can actually be below the original component's tolerance.
But for the 200µF Micas (e), I really cannot find any replacement whatsoever. At least not with such a low tolerance.
Do you have any thoughts on this? Perhaps, there is something else I could try to do. Perhaps it makes sense to attach the converter internally in order to avoid the cable length.
If everything else fails, I will try to somehow install an Xwhatsit controller. But those are hard to find here in Germany. Or I'd have to solder my own. And there's the problem of how to attach it to the keyboard matrix. I think it might be possible to solder a custom-built TH-Xwhatsit on top of the original controller, connecting to the matrix via pin headers. But there is really little space in the housing.
PS: Interestingly, I have a Model M122 which had very similar symptoms and I have long used it with the same external converter I'm trying to use with the F XT. However, it turned out, I didn't properly ground the PCB. I've now installed an internal properly grounded converter and the problems with this particular M122 board appear to be gone.
