Re: [AVT] RE: <draft-ietf-avt-rtp-vmr-wb-03.txt>: sampling rate

[Colin Perkins <csp at csperkins.org>]

On 24 Sep 2004, at 13:00, Magnus Westerlund wrote:
> Qiaobing Xie wrote:
> > I'd think that the saving (in terms of reduction to the system design 
> > and signaling complexity) by avoiding performing RTP timestamp rate 
> > switches (as Magnus pointed out above) far overweighs the ability of 
> > using the sampling rate to drive the media clock.
>
> Also, nothing in the case of VMR-WB prevents it from driving the media 
> clock. The larger issue is that VMR-WB has built in resampling 
> functions, and the fully information carrying sampling rate needed is 
> 16kHz, not 8. However resampling before the codec, or resampling in 
> the codec does not make any difference from the handling, only 
> somewhat in regards to quality, due to encoder optimization.

This is why RTP uses the timestamp rate to indicate the sampling rate 
of the input, and why using a fixed 16kHz rate would be a mistake.

> As the native out-put is 16kHz using a 8kHz RTP clock prevents a 
> receiver from driving its media clock with the appropriate rate. If a 
> receiver is using 8kHz output, then it easier to scale down the rate, 
> then to scale it up for the normal receiver behavior.
>
> Thus I see the least problem of driving the media clock when we only 
> uses a 16kHz clock, rather then a clock that reflects the encoders 
> front-end.

If the codec needs another parameter to indicate the preferred output 
rate, I don't see that as a problem.  However, the RTP timestamp rate 
should match the sampling rate of the input.

Colin


_______________________________________________
Audio/Video Transport Working Group
avt at ietf.org
https://www1.ietf.org/mailman/listinfo/avt