Re: [Linphone-developers] G722

[Simon Morlat]

Hey Vadim,

Can you please send me the final patch for msrtp.c ? so that I can immediately 
merge it on svn.
Thanks a lot

Simon

Le Friday 30 January 2009 17:11:01 Vadim Lebedev, vous avez écrit :
> Simon,
>
> Silly me,  you're right of course....
> For some misterious reason i've  assumed that msg722.c HAVE to increment
> by 320....
>
>
> Thanks
> Vadim
>
> Simon Morlat wrote:
> > Hi Vadim,
> >
> > I think some small modifications like:
> >
> > static int sender_set_session(MSFilter * f, void *arg)
> >
> > {
> >
> > SenderData *d = (SenderData *) f->data;
> >
> > RtpSession *s = (RtpSession *) arg;
> >
> > PayloadType *pt =
> >
> > rtp_profile_get_payload(rtp_session_get_profile(s),
> >
> > rtp_session_get_send_payload_type(s));
> >
> > if (pt != NULL) {
> >
> > if (strcasecmp(pt->mime_type,"g722")==0)
> >
> > d->rate=8000;
> >
> > else d->rate = pt->clock_rate;
> >
> > } else {
> >
> > ms_warning("Sending undefined payload type ?");
> >
> > }
> >
> > d->session = s;
> >
> > return 0;
> >
> > }
> >
> > static int receiver_set_session(MSFilter * f, void *arg)
> >
> > {
> >
> > ReceiverData *d = (ReceiverData *) f->data;
> >
> > RtpSession *s = (RtpSession *) arg;
> >
> > PayloadType *pt = rtp_profile_get_payload(rtp_session_get_profile(s),
> >
> > rtp_session_get_recv_payload_type
> >
> > (s));
> >
> > if (pt != NULL) {
> >
> > if (strcasecmp(pt->mime_type,"g722")==0)
> >
> > d->rate=8000;
> >
> > else d->rate = pt->clock_rate;
> >
> > } else {
> >
> > ms_warning("Receiving undefined payload type ?");
> >
> > }
> >
> > d->session = s;
> >
> > return 0;
> >
> > }
> >
> > Assuming the g722 encoder increments its timestamp by 160 each time it
> > processes 320 samples of pcm data, this should work.
> >
> > Do you agree ?
> >
> > Simon
> >
> > > The desire to keep stable ABI is very good reason to avoid changing
> > >
> > > payload type.
> > >
> > >
> > >
> > > However it seems that modifciations of of msrtp.c becomes somewhat more
> > >
> > > complicated than
> > >
> > > you think.
> > >
> > >
> > >
> > >
> > >
> > > First of all when sending (line 223 at msrtp.c) we need to do something
> > >
> > > like:
> > >
> > >
> > >
> > > / if (pt == g722_pt)
> > >
> > > timestamp /= 2;/
> > >
> > >
> > >
> > > Then when receiving, in addition fo 'reading' 8000 instead of 16000
> > > from
> > >
> > > the payload descriptor we need to
> > >
> > > do something like in receiver_process:
> > >
> > >
> > >
> > > / while ((m = rtp_session_recvm_with_ts(d->session, timestamp)) !=
> > >
> > > NULL) {
> > >
> > > uint32_t pkttstamp = rtp_get_timestamp(m);
> > >
> > > int pt = rtp_get_payload_type(m);
> > >
> > >
> > >
> > > if (pt == g722_pt)
> > >
> > > pkttstamp *= 2;
> > >
> > >
> > >
> > > mblk_set_timestamp_info(m, pkttstamp);
> > >
> > > mblk_set_marker_info(m, rtp_get_markbit(m));
> > >
> > > mblk_set_payload_type(m, pt);
> > >
> > > rtp_get_payload(m,&m->b_rptr);
> > >
> > > ms_queue_put(f->outputs[0], m);
> > >
> > > }
> > >
> > > /
> > >
> > >
> > >
> > > Any comments?
> > >
> > >
> > >
> > > BTW the g722_pt could have different values for send and receive
> > >
> > > directions, but this
> > >
> > > is a general problem in mediastreamer --- even if oRtp is able to
> > >
> > > function with different profiles for each direction
> > >
> > > mediastreamer2 and linphone does not support it
> > >
> > >
> > >
> > > Thanks
> > >
> > > Vadim
> > >
> > > > Simon
> > > >
> > > > Le Wednesday 28 January 2009 19:30:56 Vadim Lebedev, vous avez écrit :
> > > >> Hello,
> > > >>
> > > >>
> > > >>
> > > >> We've a following interop problem with G722 codec:
> > > >>
> > > >>
> > > >>
> > > >> Because of historical reasons the relevant RTP RFC speicifies that
> > > >>
> > > >> when using G722 payload
> > > >>
> > > >> RTP TIMESTAMP should be incremented with 8KHZ frequency even if the
> > > >>
> > > >> REAL sampling rate
> > > >>
> > > >> is 16KHZ.
> > > >>
> > > >>
> > > >>
> > > >> As you understand msrtp.c filter is unable to handle this situation,
> > > >>
> > > >> so we've been thinking about possible enchancements.
> > > >>
> > > >>
> > > >>
> > > >> One idea that comes to mind is to add a 'rtp_rate' field to
> >
> > Payloadtype
> >
> > > >> structure , and if it is non zero and different from sampling rate
> > > >> to
> > > >>
> > > >> compute adjustement (divider or multiplier) to rtp time stamp.
> > > >>
> > > >>
> > > >>
> > > >> Any comments on this approach?
> > > >>
> > > >>
> > > >>
> > > >>
> > > >>
> > > >> Thanks
> > > >>
> > > >> Vadim
> > > >>
> > > >>
> > > >>
> > > >>
> > > >>
> > > >>
> > > >>
> > > >> _______________________________________________
> > > >>
> > > >> Linphone-developers mailing list
> > > >>
> > > >> address@hidden
> > > >>
> > > >> http://lists.nongnu.org/mailman/listinfo/linphone-developers