Forwarded mail from Alexander Patrakov: "Alexander E. Patrakov" <[EMAIL PROTECTED]> writes:
> Please look at xine-lib-1.1.2/src/xine-engine/resample.c > > This file contains sound resampling functions. They have the property > that the first sample of the output data is the same as the first input > sample, and the last sample of the output data is the same as the last > input sample. And they don't keep any state between resampling chunks, > i.e., the first call resamples the first N samples, then the next call > resamples the next N samples completely independently. The properties > mentioned above guarantee incorrect behaviour at chunk boundaries. > > Namely, suppose that the original sample rate is 22.05 kHz, thus, the > interval between samples is normally 45.351 us. Xine concerts the first > 576 samples to 48 kHz, then it converts the next 576 samples. However, > the input sample 575 (counting from zero, i.e., the last sample of the > first chunk) maps _exactly_ to the output sample 1253, and the input > sample 576 maps _exactly_ to the output sample 1254. The time distance > between these signal values becomes 20.833 us instead of the correct > 45.351 us. In other words, all data that should be inserted between > these output samples are lost. This manifests itself as a click if the > signal has a non-zero derivative at that point. > > The problem is aggravated by the fact that, in the default ALSA dmix > setup, xine always resamples its output to 48 kHz (even though ALSA can > do it better itself with "plug" and libsamplerate), thus, all narrowband > internet radios suffer from clicks in xine. > > If you can't fix the xine resampler, please make sure it is not used in > the default plug:dmix setup. I checked with current cvs, but that file didn't change. Perhaps someone on xine-devel can do something with this suggestion? -- Gruesse/greetings, Reinhard Tartler, KeyID 945348A4 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
