Re: [rohc] SID packets (IP and ATM question)

[Stephen Casner <casner@packetdesign.com>]

On Wed, 2 Aug 2000, Alex Tulai wrote:

> During periods of silence, a SID packet is sent every silent frame according
> to the speech compression standards (G.729 Annex B and GSM 06.32).
> Does anybody know if equipment manufacturers do follow the standard and 
> send every SID packet or follow a policy of sending them at a reduced rate ?

I am not an expert on these compression standards, but I believe they
also include annexes for "discontinuous transmission" (DTX) -- at
least some of them do.  If the transmission is over a circuit, then
you might as well send a frame in every time slot.  I suppose there is
not even much to gain by sending a SID frame instead of a voice
frame.  But in the packet world, not sending silent frames makes a big
difference.  The number I recall from papers a while back was that
only 40% of the full duplex bandwidth is used for non-silent frames.


That brings me to a related question about the GEHCO presentation at
IETF last Friday.  I believe I heard the statement that the bandwidth
savings from using CNG (comfort noise) was only "a couple of
percent".  If I heard correctly, why is that the case?  Is it because
the CNG frame is about the same size as the voice frame and a CNG
frame is sent in every slot?

It seems to me that there would be a lot more bandwidth/spectrum to
gain using a packet-oriented scheme and not sending anything 60% of
the time than there would be in using a circuit-oriented scheme and
saving one byte per slot.
							-- Steve