On Wed, 11 May 2016, at 06:22 PM, Laszlo KERTESZ wrote: > Back with more data. > I used this option since then, but it turns out that even if Pulseaudio > doesnt do it,some other apps will still abuse the gain adjustments. > > For example i just finished a Google Hangout session (working from my > home > desktop) and with "aec_args="analog_gain_control=0\ > digital_gain_control=0"" - the mic boost (element "Rear Mic Boost") was > going back to 53 (mid level) causing loud noise when there was silence on > my side (used headphones with microphone). > After i set "volume = zero" > in /usr/share/pulseaudio/alsa-mixer/paths/analog-input-rear-mic.conf, > "[Element Rear Mic Boost]" section and restarted Pulseaudio, AGC > adjustments were still made by Chrome, but only within the "Capture" > control and nobody complained anymore. Indeed, it might be interesting to explore an option to disable using boost controls via UI. At least for me, that's not something I see myself immediately having the time for. > So Hangouts does this, and it seems Chrome has AGC built in for "plain" > Webrtc too with no option to disable (i at least couldnt find one). I > tested Jitsi Meet (Webrtc MUC server, locally deployed) and Chrome uses > AGC > boost adjustments with the same Pulseaudio setup. FWIW, with the upcoming PA 9.0, things may get better with built-in AGC. On Firefox, I see a config option to disable AGC, haven't tested if it does. > Skype on the other hand always worked just fine with echo-cancel - either > doesn't do AGC on its own or does it better (this point isnt that > important > now since it will likely stop working for audio/video in the immediate > future and the browser version will replace it with Webrtc sometimes, > hopefully before it, returning us to the previous point). Thinking about > it, desktop apps in general are echo-cancel-friendly, even the > java-powered > Jitsi desktop client. > > It seems that in cases like this the "aec_args" are insufficient and some > option should be provided to disable the gain adjustments server-wide > since > the editing of the config files are overwritten on Pulseaudio update. > > Also note the fact that some sound cards (like most? USB headsets) that > don't have variable mic boost options are immune to this issue. I have > one > USB headset wit mic (actually tried multiple models, all with the same > results) at work and they work just fine with echo-cancel regardless of > Pulseaudio's settings. We currently don't have the ability to do that. We might get this at some point as part of ongoing efforts to enable sandboxing. -- Arun
