search

mmccoyd / hillside

Family of split ergonomic keyboards with three rows of five or six keys, aggressive column stagger, generous thumb arc and optional bottom utility keys
MIT License
315 stars 42 forks source link

Slave-side key presses are not registered by master side #1

Closed jurgen-kluft closed 2 years ago

jurgen-kluft commented 2 years ago

Current situation: Left and right board soldered diodes + reset + TRRS + socketed Pro Micro. Both sides have a Pro Micro flashed with default firmware (cooked by 'https://config.qmk.fm') Both the left and right side work by themselves, all keys are working.

When testing for continuity (USB disconnected) and TRRS connected, the 'data' pin of the left Pro Micro is connected to the right Pro Micro 'data' pin. VSS and GND are also showing continuity.

So currently I suspect the firmware to not recognize that it should act as 'slave'.

Any suggestions/pointers are very much appreciated at this point.

jurgen-kluft commented 2 years ago

For my master/slave boards with Pro Micro I had to add the following define into config.h:

define SPLIT_USB_DETECT

Thanks to Drashna for providing the assist.

This solved the issue of the master not receiving slave key presses since the slave side did not recognize itself correctly as slave.

mmccoyd commented 2 years ago

From your other issue #2, it seems you are running a much earlier version of QMK. The split is defined in info.json, which is likely not read correctly by a much earlier version of QMK.

mmccoyd commented 2 years ago

I'll create a test keymap with a keymap.c in the morning. The default keymap.json creates a simple keymap.c in the .build directory, but an explicit test and example is better.

jurgen-kluft commented 2 years ago

Same here, will do a run with and without that define to see if the issue is reproducible now with the new QMK firmware.

jurgen-kluft commented 2 years ago

I have removed/added my fork as well as my installation of QMK so I am now on the latest version QMK Firmware 0.16.7.

I tested as follows: 1) Compile my custom keymap for Hillside, test left-side and right-side 2) Add #define SPLIT_USB_DETECT to my config.h, compile and test left/right side

Only with 2) am I able to get key presses registered from the slave side.

Note: you can find my hillside keymap here: www.github.com/jurgen-kluft/qmk-keyboards/tree/master/hillside/keymaps/jaykay

mmccoyd commented 2 years ago

Your keymap is a tad complex, though worthy of further study. I believe SPLIT_USB_DETECT is old and no longer needed. Turning it on or off did not seem to affect the below keymap.

Here is a simple .c based default that I'll create a PR for in a bit. It builds cleanly with 0.16.7. It includes a commented out line in config.h to set MASTER_RIGHT. I have it in 0_1/keymaps/, but it seems to work fine in keymaps/. As is it works fine with left connected. With the MASTER_RIGHT uncommented it works fine with right connected. default_dot_c.zip

Let me know if this default_dot_c works for you. What stood out to me in yours was the KYRIA_TO_HILLSIDE_LAYOUT mapping.

(My in-process design avoids this left right nonsense with a diode. Something I need to change on this PCB.)

jurgen-kluft commented 2 years ago

From the conversation I had with Drasha, he mentions this "because a lot of the newer pro micro clones don't properly detect vbus levels". That means that indeed the slave will presume that it is master and that is exactly what I am seeing. So it is a hardware dependency, if you have these kind of Pro Micros you might need to define SPLIT_USB_DETECT.

https://docs.qmk.fm/#/feature_split_keyboard?id=hardware-configuration-options

In what way did this stand out to you? What stood out to me in yours was the KYRIA_TO_HILLSIDE_LAYOUT mapping.

mmccoyd commented 2 years ago

Drasha knows far far more than I do, but the board has fixed master behavior. It is master left, as I understand and designed, unless it is told in config to be right. So I'm not sure how a clone hardware issue could confuse that.

On the mapping, just that that looked like an extra layer of complexity that would be easy to get wrong. Though unlikely to make one side not work. So not really an issue for this problem.

mmccoyd commented 2 years ago

Ahh, well it is fixed based on config defaults for the board. If the keymap changes those defaults, then that no longer holds.

jurgen-kluft commented 2 years ago

Yes I understand that, but to get back to basics, when I use the default hillside keymap my slave side doesn't work. Then adding SPLIT_USB_DETECT to the default keymap fixes it.

You can clearly see in the QMK code that SPLIT_USB_DETECT will identify master by actual USB activity and not by reading vbus.

mmccoyd commented 2 years ago

Oh, I thought it did work with the default, but not with your keymap.

What about with the default_dot_c keymap folder, if you set MASTER_RIGHT in config.h? And alternatively explicitly MASTER_LEFT?

If it is a hardware issue, I'm not sure of a solution beyond explicit config, or new hardware.

mmccoyd commented 2 years ago

I'm assuming the code and hardware should be able to distinguish between noise vs data in from USB or TRRS and decide based on fixed master assignment what it should then do.

mmccoyd commented 2 years ago

Did not mean to close it, fumble fingered.

jurgen-kluft commented 2 years ago

I am sure this is not a hardware issue, for me the only suggestion I can give at this moment is for you to add a note in your documentation about this so that people using Pro Micro MCU's can avoid this specific issue.

I will test the zip that you posted and use MASTER_RIGHT although I am sure that this also will work. However this makes it more involved and risky as a user to flash the firmware correctly since you have to do a 'compile/flash/change/compile/flash'.

In your documentation I suggest you add the comment to enable/add SPLIT_USB_DETECT to their own config.h when they use Pro Micro MCU's.

mmccoyd commented 2 years ago

Huh. You flash the same firmware to both sides. You do have to pick which side you want the cable to go on though.

I'll get a pair of Pro-micros to test. But you mentioned bad clones. If the issue is bad clones, that is separate.

In any case, I'll add some trouble shooting notes. Thanks for the feedback and sorry it did not work quite right with your boards.

jurgen-kluft commented 2 years ago

Thanks Michael, to be honest the board is great, like the minimal feel and the low profile. So I should be the one thanking you for putting in the design work, the detailed documentation and making it open-source.

Thank you !

Note: Drashna directed me to read this article if I was interested in the details (and trouble) on how the split code is determining master/slave. https://medium.com/@keebio/the-case-of-the-wayward-elite-c-73f0fd691f88#3bf5

mmccoyd commented 2 years ago

Thanks! Creating it scratched an itch and having other people actually use it is really awesome. I have a version in the works that uses a diode in the key matrix on one side to know which side is which, and which has a slight ring and pinky finger splay. Once that is out I'll update this version with the same handedness detection. Hopefully one of all these works out for you.

I'll read up on that article. Thanks. I do remember a while ago reading about voltage levels detection and what not.

Added: That article is what I had read long ago. Another useful one is this: https://deskthority.net/wiki/Elite-C. (About the Elite-C, but listing the detection issue and workaround.)