Qmk_firmware: Bad Lag sometimes happens with I2C on split boards

[nooges]

On some of my builds, if USE_I2C is used, that half and the whole board lags a bunch in registering keypresses.

Describe the Bug

I’ve tried to trace things down a bit, like erasing EEPROM, compiling with and without EE_HANDS and RGB, but no luck so far. Some of my halves work fine.

I’ve heard someone else report a similar issue on Discord as well.

System Information

• Keyboard: Levinson Rev 2 & 3
  
  
  
  • Revision (if applicable):
  
  
• Operating system: Mac
• AVR GCC version: 8.3
  
• ARM GCC version:
  
• QMK Firmware version: Latest
  
• Any keyboard related software installed?
  
  
  
  • [ ] AutoHotKey
  
  
  • [ ] Karabiner
  
  
  • [ ] Other:
  
  

Additional Context

[drashna]

I've actually seen this myself.
Usually make clean and recompiling fixes the issue. Heck, even just recompiling will fix it. So it's an intermitting issue.

[smyjpmu]

It is indeed happening on my Iris 2.8 slave side only and recompiling doesn't seem to fix it.

[nooges]

TODO For self (or anyone trying to fix): Connect I2C lines to oscilloscope and check signal to see if things like rise times look alright.

[nooges]

Traced issue down to the default I2C clock rate being changed to 400kHz from 100kHz for split_common boards on refactor of split_common to use main I2C driver when this change was done: https://github.com/qmk/qmk_firmware/pull/4974

[ethanspitz]

I found this myself as well. I'm building a handwired, and I didn't yet wire the I2C on the left side (no pull ups) and the typing speed was terrible. Read this, and it all made sense. I put the pullup resistors on I2C and now it's perfectly fine.

[nooges]

Somewhat related: https://github.com/qmk/qmk_firmware/issues/7522