[Asterisk-Users] Sipura g729 call quality to PSTN

Rich Adamson radamson at routers.com
Fri Feb 18 09:21:08 MST 2005


> > That does not sound right at all. The difference between the two Time=
> > values should have been 10 (milliseconds).
> > 
> > Did you reboot the Sipura after making the change? There are some values
> > in the Sipura that don't take effect until after the next reboot; I don't
> > have a clue whether this happens to be one of them.
> 
> Yes - sipura was rebooted.  Actually, the changes did seem to take
> affect even before the reboot (verified by call quality improvement
> and ethereal traces).
> 
> So in your opinion, instead of 80, it should be a difference of 10? 
> If so - then you are saying that the timestamp is in miliseconds?
> 
> I am as puzzled as you - really does not seem logical, but call
> quality is finally decent and it does not seem to bother asterisk at
> all.  Do you see any potential problems with this?

I did a fair amount of experimenting this morning using a spa3000 with
g711 and g729 codecs. I'm more confused now then ever. I also used
ethereal to inspect timestamps, etc.

 spa3k(fxs) -> asterisk -> IAX(ITSP) -> pstn net -> analog phone

The spa3k is running v2.0.13(GWg) firmware, and * is CVS-HEAD-02/13/05.

The spa3k had a default RTP Packet Size of .030 (30 milliseconds) even
though the User Manual indicated that 20 milliseconds is the default.
Asterisk config is default at 20 milliseconds.

I changed the spa3k rtp from .030 seconds, to .020 seconds for
consistency. Audio quality "seemed" to be better when using g711.

Regardless of whether I used g711u or g729, the rtp timestamps were
always 160 difference between consequtive packets (as observed by
ethereal).

Changing the spa3k rtp to .010 seconds yielded timestamps that were
always 80 difference between consequtive packets (same as you
observed). However, * -> spa3k continued to have 160 difference.
Audio quality seemed to improve another step, and the occasional
echo that we heard seemed to disappear. Pure guess is the smaller
rtp size is impacting the jitter buffer and/or echo canceller in
the spa3k. I'm going to run with these settings for a while to see
what the longer term impact/stability might be.

Rich





More information about the asterisk-users mailing list