Le 08/11/2017 à 12:43, Tanu Kaskinen a écrit : > On Tue, 2017-11-07 at 15:29 +0100, 21naown at gmail.com wrote: >> If you cannot be 100% sure, is any problem replaced by a 0dB sound? If >> so, then I can check for a possible problem with Audacity and >> Sample Data Export. For example, if an underrun or a rewind occurs, is >> it replaced by 0dB by PulseAudio? Or it is a random value? > If an underrun happens inside PulseAudio (i.e. an application doesn't > provide data fast enough), the missing audio is replaced with silence. > Very good news! > The issues with rewinds were about monitor sources and resampling. If I > understood correctly, your source material is in 48000 Hz and Jack is > configured with that rate too, so there's no need for resampling, and > you haven't mentioned recording from a monitor source either, so there > should be no reason to be concerned about rewinds. If rewinds happen, > they shouldn't cause any glitches. > > But if you're curious, the monitor source bug causes short chunks of > audio to be duplicated in the recording stream. I haven't seen concrete > examples of what kind of glitches resamplers can have, but my intuition > is that one sample in the audio stream might have a significant > adjustment towards zero, and a few samples after that might have > slightly wrong values. > I give you my audio chain more precisely: Playing = Audio file (to simplify) â?? PulseAudio (â?? pulseaudio-module-jack) â?? Jack â?? Audacity = Recording â??Audio fileâ?? means the audio source is in the computer, it is not taken by a microphone for example (is this is what you said by â??monitor sourceâ??? Or you talk about the tab in pavucontrol â??Recording â?? Audacity â?? Monitor of SOMETHINGâ???). Settings like the sample rate/number of channels/â?¦ are set to be the same for my entire audio chain, to avoid possible sound modifications like resampling. Yes I think it will not be more than 48000 Hz / 2 channels. As you said, duplicated values (plus you said it is a bug) are related to rewinds. So no rewinds in the log = no possible duplicated values. If I understood, in my case, you cannot confirm 0 dB is the only result I could have in case of problems? (because you think I can have other problems than underruns?)