[Asterisk-Users] Opinions on IAX JitterBuffer in old-school 1.0.0?

[steve at daviesfam.org]

On Thu, 27 Oct 2005, Shane Burrell wrote:

> I have a small issue with some remote users connecting to my primary
> Asterisk server using 1.0 Every few seconds, there is a subtle "tick" and a
> very small amount of jitter. The tick is not consistent i.e. it could be in
> 2 seconds, could be 5, could be 10. This does not affect core functionality,
> and the call is quite usable but my endusers want it eliminated. My primary
> is also running 1.0. I am using the AstShape script on both ends and tos is
> set to 0x18 on both ends. If I connect through IAX to my primary with a
> 1.0.9 (a at h 1.5) there is no tick. GSM codec end-to-end, no transcoding. SNOM
> SIP phones on the remote, latest firmware. CPU on both ends is basically
> nil. No IRQ conflict. The tick only exhibits itself through my firewall but
> note if I use 1.0.9 though the firewall there is no tick. If I bring the
> remote IAX box inside the LAN and plug it in, no problem.  Default jitter
> buffer settings on both ends as follows:
> 
> jitterbuffer=yes
> maxjitterbuffer=500
> maxexcessbuffer=80
> minexcessbuffer=10
> jittershrinkrate=1
> 
> Roundabout average-type values for "IAX2 SHOW CHANNELS" is 20ms lag, 10-20ms
> jitter, 60-80ms jitter buffer
> 
> My theory is somehow the jitter buffer is contributing to this. Before I
> start twiddling knobs, I'd like opinions on whether the jitterbuffer could
> be contributing to this, and whether fooling around with the jitterbuffer
> values would have any effect.  For various reasons, I am staying with 1.0
> for the moment. 


Hi,

You'll hear audible artefacts in two situations:

First is where a frame doesn't arrive, or arrives too late for the jitter 
buffer to absorb.  That will result in a 20msec audio dropout in the old, 
PLC-less jitter buffer.

The second is when the jitter buffer has got too big and starts to shrink 
itself.  So when it is more than "maxexcessbuffer" msecs bigger than the 
jitter being seen.  In that case, you will hear the teeniest effect as 
every 20msec it is shrunk by "jittershrinkrate".  So in your case really I 
don't think you'll notice - the audio is played out 5% too fast.

There were plenty of tweaks post 1.0 and then, of course, Steve Kann's 
buffer replaced the old one.  Its use of packet-loss concealment makes a 
big difference to the subjective audio.

Does sound like you have the fix - upgrade to a newer Asterisk.

Steve