On Mon, Apr 5, 2010 at 4:25 PM, Mark Brown <broonie at sirena.org.uk> wrote: > On Mon, Apr 05, 2010 at 02:23:50PM -0500, pl bossart wrote: > >> Ideally we should enforce a stronger smoothing on the time estimation, >> there's no reason why such variations should occur. I forced the >> smoothing history to 20s, maybe there's a better way to do this. > > FWIW a lot of impleementations seem to struggle to keep track of time > properly over silence periods so it's worth resetting things whenever > you get one. ?Adapting the jitter buffer depth rather than the sample > rate also helps if the source is having difficulty - it can be due to > network issues, and this does absorb changes up and down without needing > to be too smart. Agree, the sample-rate changes should only be there to take care of long-term differences between local audio time and remote network time reference. Even 1% deviation on sample-rate is huge.... -Pierre
