skylark-qmk/quantum/debounce
Alex Ong 17e7762de7 Eager Per Row Debouncing added (added to Ergodox) (#5498)
* Implemented Eager Per Row debouncing algorithm.

Good for when fingers can only press one row at a time (e.g. when keyboard is wired so that "rows" are vertical)

* Added documentation for eager_pr

* Ported ergodox_ez to eager_pr debouncing.

* Removed check for changes in matrix_scan.

* Added further clarification in docs.

* Accidental merge with ergodox_ez

* Small cleanup in eager_pr

* Forgot to debounce_init - this would probably cause seg-faults.
2019-04-03 14:45:55 -07:00
..
eager_pk.c Clean up debounce a bit (#5255) 2019-03-04 07:44:46 -08:00
eager_pr.c Eager Per Row Debouncing added (added to Ergodox) (#5498) 2019-04-03 14:45:55 -07:00
readme.md Eager Per Row Debouncing added (added to Ergodox) (#5498) 2019-04-03 14:45:55 -07:00
sym_g.c Clean up debounce a bit (#5255) 2019-03-04 07:44:46 -08:00

Debounce algorithms belong in this folder. Here are a few ideas

  1. Global vs Per-Key vs Per-Row
  • Global - one timer for all keys. Any key change state affects global timer
  • Per key - one timer per key
  • Per row - one timer per row
  1. Eager vs symmetric vs asymmetric
  • Eager - any key change is reported immediately. All further inputs for DEBOUNCE ms are ignored.
  • Symmetric - wait for no changes for DEBOUNCE ms before reporting change
  • Asymmetric - wait for different times depending on key-down/key-up. E.g. Eager key-down, DEBOUNCE ms key up.
  1. Timestamp vs cycles
  • old old old code waits n cycles, decreasing count by one each matrix_scan
  • newer code stores the millisecond the change occurred, and does subraction to figure out time elapsed.
  • Timestamps are superior, i don't think cycles will ever be used again once upgraded.

The default algorithm is symmetric and global. Here are a few that could be implemented:

sym_g.c sym_pk.c sym_pr.c sym_pr_cycles.c eager_g.c eager_pk.c eager_pr.c //could be used in ergo-dox!