How could you be sure that it's not audio device performance? Audio device burst level could be checked using 'device test' of pjsystest[1]. Also, please try making call using lightweight codec such as G711, as CPU load spikes may disturb audio clock. The audio latency section in FAQ, you've mentioned before, may describe more. --- [1] http://trac.pjsip.org/repos/wiki/Testing_Audio_Device_with_pjsystest --- BR, nanang On Fri, Dec 11, 2009 at 1:05 AM, Nuno Costa <nuno.costa at wit-software.com> wrote: > Hi, > > I've been taking a deeper look into the media flow implementation and I > think the problem for the high latency might be in the network management > (maybe at the device level). > Beyond the JB summary from the previous message, check the following logs: > > 17:49:30.000?? strm01D5B674? Jitter buffer empty (prefetch=40) > 17:49:30.000???? ec01CF73F0? Underflow, buf_cnt=0, will generate 1 frame > 17:49:30.000?? Master/sound? 140 samples reduced, buf_cnt=815 > 17:49:30.000?? Master/sound? Buffer size adjusted from 955 to 815 > (eff_cnt=689) > 17:49:31.000???? ec01CF73F0? Underflow, buf_cnt=0, will generate 1 frame > 17:49:31.000?? strm01D5B674? jb updated(2), prefetch=40, size=49 > 17:49:31.000?? strm01D5B674? jb updated(2), prefetch=40, size=41 > 17:49:31.000?? Master/sound? 143 samples reduced, buf_cnt=832 > 17:49:31.000?? Master/sound? Buffer size adjusted from 975 to 832 > (eff_cnt=689) > 17:49:32.000???? ec01CF73F0? Underflow, buf_cnt=0, will generate 1 frame > 17:49:32.000?? strm01D5B674? Jitter buffer empty (prefetch=40) > 17:49:32.000?? strm01D5B674? Jitter buffer empty (prefetch=40) > 17:49:32.000?? strm01D5B674? jb updated(2), prefetch=40, size=49 > 17:49:33.000?? Master/sound? 141 samples reduced, buf_cnt=851 > 17:49:33.000?? Master/sound? Buffer size adjusted from 992 to 851 > (eff_cnt=676) > 17:49:33.000?? Master/sound? 140 samples reduced, buf_cnt=711 > 17:49:33.000?? Master/sound? Buffer size adjusted from 851 to 711 > (eff_cnt=676) > 17:49:33.000?? strm01D5B674? Jitter buffer empty (prefetch=40) > 17:49:33.000???? ec01CF73F0? Underflow, buf_cnt=0, will generate 1 frame > 17:49:34.000?? Master/sound? 136 samples reduced, buf_cnt=735 > 17:49:34.000?? Master/sound? Buffer size adjusted from 871 to 735 > (eff_cnt=667) > 17:49:34.000?? strm01D5B674? jb updated(2), prefetch=40, size=41 > > The jitter buffer is becoming many times empty which means that the audio > latency experienced is not related with the audio device performance. > I've rechecked the network latency and it is very low. Could it be a network > device issue? > > Have anyone experienced this before? > Any ideas on a possible solution? > > Nanang: do you have any suggestion, looking to the JB summary? > > Thank you very much for your support. > > Cheers, > Nuno Costa > > > On 09-12-2009 19:31, Nuno Costa wrote: > > Nanang, > > Thank you very much for your reply. > The JB summary below is from a test call: > > ? JB summary: > ??? size=44 prefetch=40 > ??? delay (min/max/avg/dev)=20/500/362/128 ms > ??? burst (min/max/avg/dev)=2/98/33/29 frames > ??? lost=86 discard=781 empty=427 > > The average delay (362ms) is very high and has impact on the overall > latency. > Any clue on how to optimize this value? > > Best regards, > Nuno Costa > > > On 09-12-2009 17:41, Nanang Izzuddin wrote: > > Hi Nuno, > > Problem behind ticket #783 is high latency issue on WM :) > > Agree, 300ms different is quite strange/high. It is usually useful to > examine the JB statistics at the end of call (printed by > pjmedia_jbuf_destroy()), e.g: if JB was used to get full, clock-drift > or high burst of audio device might take place. > > BR, > nanang > > > On Thu, Dec 3, 2009 at 5:30 AM, <ncosta at wit-software.com> wrote: > > > Hi all, > > I've developed an application using PJSIP for Microsoft Windows Mobile 6.x > devices and I've been testing it on some devices. > > Unfortunately the audio latency is very high on both devices (Samsung > Omnia II and LG LM730). Values are typically as high as 700ms, measured > from end-to-end. > > I've checked your wiki: > http://trac.pjsip.org/repos/wiki/FAQ#audio-latency > > And also this ticket: > http://trac.pjsip.org/repos/ticket/783 > > I've done several tests adjusting the parameters but results haven't > changed. Running 'pjsystest', the average value is 400ms for both devices. > > Has anyone faced this problem already? > > I've done these tests on a local network with a low network latency so I > do not quite understand the 300ms difference between the end-to-end test > and pjsystest. > > Any clue on the problem behind ticket #783? > > Thanks! > > Best regards, > Nuno Costa > > > _______________________________________________ > Visit our blog: http://blog.pjsip.org > > pjsip mailing list > pjsip at lists.pjsip.org > http://lists.pjsip.org/mailman/listinfo/pjsip_lists.pjsip.org > > > > _______________________________________________ > Visit our blog: http://blog.pjsip.org > > pjsip mailing list > pjsip at lists.pjsip.org > http://lists.pjsip.org/mailman/listinfo/pjsip_lists.pjsip.org > > > > -- > ========================================= > Nuno Costa > Senior Engineer > WIT Software S.A. > Coimbra, Lisboa, Leiria, Porto (Portugal) > San Jose (California, USA) > Phone : +351 239 801030 > Mobile: +351 91 9821825 > Email : nuno.costa at wit-software.com > Web: http://www.wit-software.com > ========================================= > > -- > ========================================= > Nuno Costa > Senior Engineer > WIT Software S.A. > Coimbra, Lisboa, Leiria, Porto (Portugal) > San Jose (California, USA) > Phone : +351 239 801030 > Mobile: +351 91 9821825 > Email : nuno.costa at wit-software.com > Web: http://www.wit-software.com > =========================================
