[Asterisk-Dev] audio packet encoding

Derek Smithies derek at indranet.co.nz
Thu Apr 28 13:54:57 MST 2005


Hi,
 The iax2 spec should state that when sending audio packets
the sender may decide to place several audio frames in one packet.

the spec should state, 
  the audio frames are placed in the packet in the order they are 
  collected.Thus, the oldest audio frame is placed immediately after 
  the iax2 header information.

The spec should state that either party of the call have no control 
over the size of the packets they receive. In other words, one party in 
the call cannot ask that the remote node buffer 3 (say) audio frames 
per packet.


Indeed, this is a matter for discussion. In the New packet, which 
currently containss IEs describing the available codecs and preferred 
codecs, we could add an IE which requests 80ms of audio per frame. Adding 
such a feature would reduce the bandwidth used by iax2 signficantly. 

Yes, at the expense of latency, but an extra 20 (or 40) ms is not going to
"break the bank"

Consider
1 gsm frame/packet ==> 31.6 kbit/sec
2 gsm frame/packet --> 22.4 kbit/sec
3 gsm/frame/packet ==> 15.3 kbit/sec

Yep, for those who have huge bandwidth costs, 15.3 kbit/sec is much better 
than 31.6 kbit/sec.


Derek.
-- 
Derek Smithies Ph.D.                         
IndraNet Technologies Ltd.                
Email: derek at indranet.co.nz         
ph +64 3 365 6485                          
Web: http://www.indranet-technologies.com/  




More information about the asterisk-dev mailing list