RE: [rohc] ROHCv2: Default ts_stride?

[pelle at cdt.luth.se]

Rohit,

I think that NB codecs will remain for a while; and we know that they are 
prevalent today, so it makes sense to use 160 as default value in RoHCv2 - it 
will be useful right away and for some time ahead. Furthermore, this is anyway 
a small optimization as you point out, so I don't think it will be a problem 
for other codecs in the future to send a new value with the IR header.

Cheers,

///Ghyslain

Quoting "Kapoor, Rohit" <rkapoor at qualcomm.com>:

> Kristofer,
> 
> I think this is a good idea, even though the gains are kinda small
> (since only typically IR packets will see a reduction in size of ~2
> octets).
> 
> The only question is: assuming WB vocoders are more prevalent in the
> future, should we optimize for that (delta Timestamp = 320) or for the
> more prevalent NB codecs of today?
> 
> Thanks,
> Rohit
> 
> > -----Original Message-----
> > From: Kristofer Sandlund [mailto:kristofer.sandlund at ericsson.com]
> > Sent: Friday, September 14, 2007 3:56 AM
> > To: rohc at ietf.org
> > Subject: [rohc] ROHCv2: Default ts_stride?
> > 
> > Hi all,
> > 
> > we've done some work on the v2 draft and have come up with
> > a change that we feel would improve the RTP profiles. As usual,
> > object if you think this is a bad change.
> > 
> > - In 3095, the TS_STRIDE has a default value of 1 (no scaling), so
> that
> > every RTP context has to explicitly send a new scaling value. To us,
> it
> > seems like the default should be the most usual stride we expect to
> see,
> > 
> > so that those codecs do not have to update the stride explicitly.
> > Since voice will be the place where the gain is largest, and it seems
> > like the most frequent stride to use will be 160 (e.g. AMR and EVRC
> > with 8000Hz sampling rate, 20ms frame period), we suggest to set the
> > initial  value for ts_stride ROHCv2 to 160.
> > Effects of such a change would be:
> >   * Codecs such as AMR & EVRC would not have to send ts_stride at all
> >     (saving about 2 octets * optimistic approach)
> >   * Codecs with other stride values that want to scale the TS would
> >     behave exactly as today (i.e. they have to set up the ts_stride)
> >   * A flow where the compressor would use no scaling (in 3095, it
> >     does not have to send ts_stride at all) would have to send the
> >     ts_stride * OAA
> > 
> > So, IMO a value of 160 will be more frequent than TS with scaling
> > factor 1, so I think this is an improvement to the compression ratio.
> > 
> > Objections?
> > 
> > /Kristofer
> > 
> > _______________________________________________
> > Rohc mailing list
> > Rohc at ietf.org
> > https://www1.ietf.org/mailman/listinfo/rohc
> 
> _______________________________________________
> Rohc mailing list
> Rohc at ietf.org
> https://www1.ietf.org/mailman/listinfo/rohc
> 
> 





_______________________________________________
Rohc mailing list
Rohc at ietf.org
https://www1.ietf.org/mailman/listinfo/rohc