[asterisk-dev] Jitterbuffer and FAXes with E1

Martin Vít vit at lam.cz
Thu Jun 28 09:23:33 CDT 2007

>> So i've tried simulate fixed latency on my router (hold data for 200ms 
>> from ATA to asterisk) and disable jb at all. FAX are OK. If jbenable = 
>> yes - problem.
>> There is something wrong with jbenable or am I missing something?
>> Maybe there is somewhere activated PLC or some another filter? 
>> whenever jbenable is turned on?
>> Codec is alaw
> It is correct that a perceptual jitter buffer will play with the timing, 
> and that will break things like faxing. 

I need to investigate this, because without jitter there is no chance to 
send fax on a little jittery network. (and there is no way to enable or 
disable jitter or change jitter on the fly yet)

On production system I have asterisk 1.2 patched with slavs jitterbuffer 
implementation and digium 4port PRI. I've enabled SIP adaptive 
jitterbuffer and with the same ATA reregistred to this production 
system, i've send fax to my 1.4 with success.  (Jun 28 16:16:04 
VERBOSE[8475] logger.c:     -- ***[JB LOG]*** adaptive jitterbuffer 
created on channel Zap/32-1)

Complete scenario:

FAX -> ATA SIP -> Asterisk 1.2 adaptive jitter -> Zap PRI ---> TELCO 
SWITCH ---> Zap PRI asterisk 1.4 ---> iaxmodem.

this work even though jitterbuffer is enabled on patched asterisk 1.2 
which is almost the same as in 1.4. But there have to be some 
differences which broken fax transmitions.

> You need to disable clever 
> jitter buffering algorithms when you need an unbroken stream. Typically 
> VoIP equipment will do this automatically when they hear FAX tones.
This is true for my testing ATA which is switching from adaptive to 
fixed jitterbuffer.

So the question is, why are frames modified when jitterbuffer is active 
in 1.4 even though there is NO jitter (100mbit eth in lab environment). 
I've disabled PLC in codecs.conf.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.digium.com/pipermail/asterisk-dev/attachments/20070628/cae735d7/attachment-0001.htm 

More information about the asterisk-dev mailing list