On Tue, Aug 17, 2010 at 8:07 AM, Jeff Brower <jbrower at signalogic.com> wrote: >> >> As receiver, client is required to be able to decode any packet length >> (within reasonable limit), hence the ptime is not necessary for >> decoding, and that's what pjsip does. > > Yes, but how does pjsip know packet length if it was not in the SDP at call setup time? ?Does pjsip look at type of > codec and packet length and calculate ptime? Yes. > I believe we've faced this problem before also (and had to force ptime > manually). > I don't know what that problem was, but if you think that was the problem, try reproducing it with pjsua. >> The 100ms frame interval may have come from the ptime setting in >> pjsua_media_config, which I assume your app has changed it to 100. > > I think the OP means "20 msec and 160 samples per packet". ?Otherwise, if he really means "160 samples per second", > then I have no idea what voice stream could be intelligible at 160 Hz sampling rate. > I was referring the 100ms in this: > When we analyzed why, we observed that when PJSIP receives their 183, it > tries and creates an audio playback stream assuming it will receive packets > in 100 millisecond intervals and 800 samples per frame (that's what our > dev_create_stream callback logs reports). Cheers Benny
