[Bug] Matrix scan issues on Planck Rev 6 and Rev 6.1 on versions 0.7.6 and up

vomindoraan commented 4 years ago

Describe the Bug

Pressing certain keys on Planck Rev 6 and Rev 6.1 results in multiple presses being registered, as if the board had ghosting issues. This is not a hardware/electrical issue since it works fine on older versions of QMK.

For example, pressing Enter on the default keymap outputs tg'. It seems to be a matrix scanning issue.

After some investigation and testing, I've determined that the broken behavior was brought about by #6140. This PR removed the custom matrix.c files for Planck Rev 3 and Rev 6 since the DIP switch functionality contained therein was made a core feature. However, this apparently also messed with the scan/debounce timing (or something of the sort), which resulted in broken matrix scanning.

System Information

Keyboard: Planck
- Revision: Rev 6, Rev 6.1
Operating system: Windows
ARM GCC version: 6.3.1 20170620
QMK Firmware version: 0.7.6 and up

Additional Context

This issue was also experienced by the OP of this thread on the OLKB subreddit. Setting DIP_SWITCH_ENABLE = no as @drashna suggested in the thread didn't fix the problem.

Other affected boards may include Planck Rev 3 and Preonic Rev 3.

alex-ong commented 4 years ago

(I don't have a planck)

Old debounce code; debounc_timer is part of scanning.

scan:
    scan cols:
        scan col
             compare col to previous scan / set debounce timer.
             store as columns in buffer

    check debounce timer:
        propagate to matrix, converting from columns to rows.

new: Debounce timer only updated at end of scan.

scan cols:
    scan col, store as rows in buffer, compare col to previous scan (set bool if changed)
debounce(pass in bool):
   if diff, set timer.
   if timer expired, propagate to matrix

While they are not exactly the same, to my eye they should be equivalent. The scanning code also looks equivalent-ish to qmk-cores default implementation. Only difference i can see is maybe how long it waits between pin-sets, working code uses 20us, new code (matrix.c) uses matrix_io_delay() which defaults to 30us. A longer delay shouldn't result in any weird noise chatter.

You could restore the custom_matrix and see if that works first; That way you have latest qmk_core but using old custom_matrix, hence excluding changes in qmk_core as part of the problem.

DonnieWest commented 4 years ago

Just chiming in here that I am also experiencing this problem. Fresh clone of the qmk project resulted in this issue. If there's anything I can do to troubleshoot just let me know

DonnieWest commented 4 years ago

I can confirm that switching back to the old matrix file for Planck Rev 6 resolves the issue for me

timshearer commented 4 years ago

Think I've also had this issue on my Preonic Rev 3. When pressing a particular key ("U" in my case) I would sometimes get no output and sometimes get multiple keypresses. I switched back to the custom matrix and that seems to fix it (or at least it has so far).

speedRS commented 3 years ago

@timshearer Having similar problems myself. Old post I know but any guidance on how to switch back to the old matrix?

timshearer commented 3 years ago

@speedRS I managed to fix my issue but unfortunately I'm not 100% sure how! There is a possibility that it was hardware related though as I think it may have resolved after I reseated a switch. So perhaps an intermittent connection.

jsebdev commented 2 years ago

Could anyone tell me how to switch back to the old matrix file for Planck Rev 6? Please and thank you very much

samjonester commented 2 years ago

I found this thread after chatting in discord. My new planck/rev6 from Drop was registering phantom keypresses and behaving erratically. Old firmware worked flawlessly with the board. I was directed here and to PR 6140 which removed a custom matrix file for the planck/rev6.

With the approach below, I was able to qmk compile -kb planck/rev6 -km default && qmk flash -kb planck/rev6 -km default. I've also been able to compile and flash with minimal customizations to the default keymap with these steps.

From PR 6140, I added the file keyboards/planck/rev6/matrix.c back into my qmk project.
I removed an include on line 8 for backlight.h which no longer exists.
I reversed the diff to reference and enable the custom matrix.c file.
I ran qmk compile -kb planck/rev6 -km default && qmk flash -kb planck/rev6 -km default
I tested with https://www.keyboardtester.com/
I ran qmk new-keymap -kb planck/rev6 -km <GH_USER> and made a few customizations.
I ran qmk compile -kb planck/rev6 -km <GH_USER> && qmk flash -kb planck/rev6 -km <GH_USER>
I tested with https://www.keyboardtester.com/
I rejoiced that RSFT_T(KC_ENT) gave me a right shift key AND an enter key without moving any keycaps!

This isn't a permanent solution. The custom matrix file is out-of-date and radically different from others in the project. However, it enabled me to successfully compile and flash my planck with qmk v0.14.14. I think this suggests that the default firmware created with the default matrix file is incompatible with the latest hardware from Drop.

jackhumbert commented 2 years ago

Thanks for all the insight here, folks. @samjonester, thanks for outlining your solution as well - @drashna has opened #14488 that adds an additional revision for the Planck & Preonic with these changes. I've also uploaded the binaries here for quickly testing this solution. I don't have a whole lot of insight into why this has become a bigger issue right now, but it may have something to do with lower tolerances somewhere along the supply chain/manufacturing?

qmk / qmk_firmware