[Asterisk-Dev] Accept patch to make iax2's calc_timestamp suspicious about the ast_frame.delivery value?

[Steven Critchfield]

On Fri, 2004-06-18 at 07:08, Andrew Kohlsmith wrote:
> On Friday 18 June 2004 03:10, steve at daviesfam.org wrote:
> > This is bad when you use the jitter buffer as the buffer starts tossing
> > away all these "very old" frames and the result is silence...
> 
> Could two * boxes over a ~4ms link manage to bugger up the timestamps for 
> three or four seconds?  I had an issue with my setup in where connections 
> going to the VOIP provider but not through the remote PRI would now and again 
> get three seconds of absolute dead silence in one direction only.  After the 
> "gap" audio returned to normal.
> 
> Path (no NAT, middle * box never got out of the loop since lefthand * box was 
> on private LAN with middle * box's 2nd ethernet adapter):
> 
> Nortel MICS -> Adit600 -> T100P(*1) -> IAX2 -> (*2) -> IAX2 -> Nufone
> 
> *2 also has a TE405P with one port going to a telco PRI for local 
> incoming/outgoing calls.  There was never dead audio on calls going in/out 
> the PRI, only through Nufone.
> 
> Jeremy and friends suggested disabling jitter buffer between *1 and *2, 
> switching to GSM (from iLBC) and enabling trunking.  I had disabled trunking 
> on *2 entirely since with jitter buffer enabled, trunking turned regular 
> audio into what I would call "bursty" audio:
> 
> Normal: "Hi, how are you today?"
> Bursty:  "Hihow....areyouto....day?"
> 
> Nufone's suggestions fixed the problem but I'm curious as to what fixed it.  I 
> hope to turn things back to the original way one thing at a time to see what 
> fixes it but the dropouts were an infrequent thing and the system seems to 
> just work now, so I'm hesitant to bugger around with it on a production 
> system.  :-)

On our system when we had a private T1 to connect to asterisk machines,
we disabled the jitter buffer only to get those gaps fixed. We had a
~4ms ping times, and we thought the jitter buffer over such a short link
was causing trouble with the echo cancel stuff. Since we needed echo
cancel, we dropped jitter. Then everything cleaned up.

-- 
Steven Critchfield  <critch at basesys.com>