#
✕ nick:waltmck
#asahi-dev · 2026-09-08 21:06
waltmcksofus: is that including the display backlight?
#asahi-dev · 2026-09-08 20:39
waltmckI guess that would explain it. I opened a PR correcting the weights which contains some of my measurements: https://github.com/AsahiLinux/asahi-audio/pull/98
#asahi-dev · 2026-09-08 20:39
waltmckAh cool, so this was just an educated guess for j413 based on other models?
#asahi-dev · 2026-09-08 19:58
waltmckso I'm fairly sure that the 1.5x thing is just wrong at least for j413. However I have no idea for other devices
#asahi-dev · 2026-09-08 19:58
waltmck(this is comparing two different reconstructions of the same voice recording)
#asahi-dev · 2026-09-08 19:57
waltmckI just tested this changing the gains to be the same for all microphones, and it reduces background noise by rought 0.8dB without changing speech volume
#asahi-dev · 2026-09-08 19:48
waltmckI took a 30 speech sample raw recording from all 3 mics, and they only differed by ~0.1 decibels. If this 1.5 constant were right (and mic 3 is less sensitive), we would expect to see a difference of 3.5 dB
#asahi-dev · 2026-09-08 19:47
waltmckmic 3 gets a 1.5x amplitude boost relative to the other mics before the beamformer ever sees the output
#asahi-dev · 2026-09-08 19:46
waltmckYeah that's true, but isn't that what the beamformer is supposed to do?
#asahi-dev · 2026-09-08 19:42
waltmcki.e. https://github.com/AsahiLinux/asahi-audio/blob/687d8cd0f6dc4b12fc0d9c3aebe360656de2882e/firs/j413/mic.json#L27
#asahi-dev · 2026-09-08 19:42
waltmckWhy does mic 3 on j413 have 3/2 the gain of the other 2 mics?