[asterisk-biz] Affects of jitter and packet loss in fax transmission

Lee Howard faxguy at howardsilvan.com
Sat Jan 21 13:04:40 CST 2017


On 01/21/2017 02:07 AM, Tahir Almas wrote:
> I  like  to  share  an investigation  that we made recently  to 
> find-out  how badly  jitter  and  packet loss affects fax  
> transmission  resulting  high  number  of dropped  faxes  also  will 
> be  interested  in your feedback in this regard
>
> following  page you will find details
>
> http://www.ictinnovations.com/affect-of-jitter-and-packet-loss-in-fax-transmission-using-t38-protocol

What feedback are you expecting to get?

Essentially your investigation has revealed that packet-loss/jitter are 
problematic for fax.  This is not a revelation to the industry at large...

http://soft-switch.org/foip.html
http://hylafax.sourceforge.net/docs/fax-over-voip.pdf
http://www.mainpine.com/blog/fax-voip-and-the-perils-of-poor-audio-quality.php

Your investigation reveals that some minimal level of packet-loss/jitter 
is generally tolerable by fax.  Again, this is not a revelation to the 
industry.

There are tolerances in ITU T.30 (fax protocol) and especially so with 
ITU T.30 Annex A (error correction mode for Phase C) which enable fax 
receivers (and senders) to cope with a nominal amount of corruption to 
the audio stream.  Furthermore, ITU T.38 permits packet "redundancy" 
(which causes packets to be sent in duplicate, triplicate, 
quadruplicate, etc. in hopes that at least one of those packets makes it 
through).  However, if the audio stream quality is poor enough then the 
tolerances in T.30, ECM, and even T.38 redundancy will be exhausted, and 
faxes will fail much more regularly in those conditions.

A lot of it also has to do with the coincidental timings of the audio 
corruption and the equipment involved.  If one 20ms moment of jitter 
occurs at the very outset of a long image page of fax (Phase C) when ECM 
is not being used then chances are very good that the fax will fail.  
The receiver cannot reliably anticipate how long it must wait for the 
end-of-page signal (Phase D).  Even so, some fax receivers simply are 
not programmed well-enough to tolerate early Phase C jitter - even with 
ECM when it's relatively simple to calculate how long it must wait for 
Phase D.

The solution - whether T.38 is being used or G.711 - is to eliminate 
packet-loss/jitter.  If that cannot be done reliably then the faxing 
must be done via proxy somehow (through a fax service provider).

Thanks,

Lee.




More information about the asterisk-biz mailing list