Hi all, I know that this was previously discussed on PJSIP threads at: http://permalink.gmane.org/gmane.comp.voip.pjsip/1751 However, I could not see a conclusion to this discussion. A short description of the issue: Currently PJMEDIA stream.c indicates DTMF event payloads as soon as they arrive, audio packets however are sent through the jitter buffer and later delivered to the receive stream via get_frame calls. This makes it difficult to generate DTMF tones locally which are in-synch with the Rx audio. I noted that put_frame2 and get_frame2 were added to jbuf, perhaps to help with this issue. Where, you could *assume* that the size parameter indicates that the frame retrieved was AUDIO/MEDIA (size=X) or DTMF (size=4). Was this the intended purpose? I also noted an early ticket, http://trac.pjsip.org/repos/ticket/478 which dealt with out-of-order/duplicate DTMF packet types. I think a potential solution to both synchronisation and out-of-order/duplicate problems would be to put them through the jitter buffer and perform the DTMF event decode on jitter buffer output side. What are the developers thoughts on this? -- BRgds, Steve King Design Engineer Zetron Australasia
