On Sat, 2015-05-30 at 10:51 +0800, golden wrote: > Hi all, > > Recently I have found a noise problem when using module-loopback. > I connected a alsa-source to a alsa-sink with module-loopback > the data from alsa-source is actually sine-wave (rate=44100, > format=s16le, channels=2) > > I found that in the adjust_rates function, if I comment > "pa_sink_input_set_rate", > there is no noise, but if I uncomment "pa_sink_input_set_rate", > problem appears. > > Here is the setup log: > > D/NMAudio ( 1797): [tid=1816] D: Negotiated format: pcm, > format.sample_format = "\"s16le\"" format.rate = "44100" format.channels =" > D/NMAudio ( 1797): [tid=1816] D: Resampler: > D/NMAudio ( 1797): [tid=1816] D: rate 44100 -> 44100 (method trivial) > D/NMAudio ( 1797): [tid=1816] D: i_ss.format=s16le, o_ss.format=s16le > D/NMAudio ( 1797): [tid=1816] D: format s16le -> s16le (intermediate > s16le) > D/NMAudio ( 1797): [tid=1816] D: channels 2 -> 2 (resampling 2) > D/NMAudio ( 1797): [tid=1816] D: memblockq requested: > maxlength=33554432, tlength=0, base=4, prebuf=0, minreq=1 maxrewind=0 > D/NMAudio ( 1797): [tid=1816] D: memblockq sanitized: > maxlength=33554432, tlength=33554432, base=4, prebuf=0, minreq=4 maxrewind=0 > I/NMAudio ( 1797): [tid=1816] I: Created input 2 "(null)" on audioio.22 > with sample spec s16le 2ch 44100Hz and channel map front-leftt > I/NMAudio ( 1797): [tid=1816] I: media.role = "abstract" > D/NMAudio ( 1797): [tid=1816] D: Negotiated format: pcm, > format.sample_format = "\"s16le\"" format.rate = "44100" format.channels =" > D/NMAudio ( 1797): [tid=1816] D: memblockq requested: > maxlength=33554432, tlength=0, base=4, prebuf=0, minreq=1 maxrewind=0 > D/NMAudio ( 1797): [tid=1816] D: memblockq sanitized: > maxlength=33554432, tlength=33554432, base=4, prebuf=0, minreq=4 maxrewind=0 > I/NMAudio ( 1797): [tid=1816] I: Created output 2 "(null)" on audioio.17 > with sample spec s16le 2ch 44100Hz and channel map front-left > I/NMAudio ( 1797): [tid=1816] I: media.role = "abstract" > D/NMAudio ( 1797): [tid=1816] D: memblockq requested: > maxlength=33554432, tlength=33554432, base=4, prebuf=0, minreq=0 maxrewind=0 > D/NMAudio ( 1797): [tid=1816] D: memblockq sanitized: > maxlength=33554432, tlength=33554432, base=4, prebuf=0, minreq=4 maxrewind=0 > I/NMAudio ( 1797): [tid=1816] I: Adding 35280 bytes of silence to queue > I/NMAudio ( 1797): [tid=1844] I: Max request changed > D/NMAudio ( 1797): [tid=1844] D: Requesting rewind due to uncorking > D/NMAudio ( 1797): [tid=1844] D: Requested to rewind 5292 bytes. > D/NMAudio ( 1797): [tid=1844] D: Processing rewind... > D/NMAudio ( 1797): [tid=1844] D: Processing rewind... > D/NMAudio ( 1797): [tid=1844] D: Rewound 4140 bytes. > I/NMAudio ( 1797): [tid=1816] I: Loaded "module-loopback" (index: #7; > argument: "loopbackid=1722 source=audioio.17 sink=audioio.22 > > -----------> before this line, the rate of sink-input was not adjusted, > everything is fine, no noise appears > > D/NMAudio ( 1797): [tid=1816] D: Loopback overall latency is 19.46 ms + > 182.97 ms + 2.67 ms = 205.02 ms (at the base rate: 205.02 ms,) > D/NMAudio ( 1797): [tid=1816] D: Latency difference: 5.02 �± 1.00 ms, > rate difference: 0 Hz > D/NMAudio ( 1797): [tid=1816] D: [audioio.22] Updated sampling rate to > 44200 Hz. > > -----------> after 10 second, the rate of sink-input was adjusted, noise > comes out. > > I have been playing around all the new rate variables, any value > that do not equals to 44100 will make audio noise. > > That looks very strange to me, so I`m asking if anyone have met this > problem before ? What kind of noise is it? Is the signal a bit distorted or totally messed up? What version of pulseaudio do you have? And what hardware? There is a bug report[1] which mentions noise when receiving audio over bluetooth and playing it to an alsa sink, and using the ffmpeg resampler. The ffmpeg resampler can't actually be used with module-loopback, so it probably falls back to the trivial resampler which you are using too. That's on pulseaudio 5.0 and raspberry pi 2. The src-sinc-medium-quality resampler reportedly doesn't have such noise, but causes issues with high cpu load. [1] https://bugs.freedesktop.org/show_bug.cgi?id=90210 -- Tanu
