[asterisk-users] fax / t38 gateway

[Steve Underwood]

JD wrote:
> Gordon Henderson wrote:
>   
>> but it's very do-able, given good Internet connections.
>>     
> [...]
>   
>> I think your statements were just a bit too strong - I agree 
>> wholeheartedly about the V. protocols and copper, but I've found in 
>> practice that faxing over IP is not just theoretically possible, but quite 
>> do-able, and is someone wanted to send me a fax (I have no fax machine, 
>> they're so 80's :), then I'd not hesitate to setup a VoIP DID and give 
>> them the number...
>>
>> Gordon
>>   
>>     
>
> Fair enough. However, my experience has been that defining "good 
> Internet" for faxing over G.711 is not entirely self-evident.
>
> I've seen some really lagged environment work with FAX over G.711 like a 
> dream. And some fast ones break badly:
>
> While I don't have full explanations for this, it _seems_ like the 
> following holds true:
>
> good ping time sequence for faxes--no jitter (all times in ms):
>
>    40, 40, 40, 40, 40, 40, 40, 40...
>
> A bad one -- high jitter low latency:
>
>   5, 10, 5, 20, 18, 7, 12, ...
>
> A _really bad_ one -- "shifting" jitter:
>
>   20, 20, ... , 20, 20, 30, 30, 30,...
>
> No provable explanation as to why.
>
> John
>   
Many ATAs, on detecting FAX tone, switch their buffering mode. They try 
to turn off all the clever adaptive stuff, and work with a fixed buffer 
length, so avoid manipulating the time scale (a complete disaster for 
modems). The fixed buffer they use it typically quite short - like 40ms. 
If you have jitter even approaching 40ms, they fall over. Another 
important thing is packet ordering. Jittery routes can reorder packets. 
If they arrive like 1, 2, 4, 3, 5, will 3 ever be processed? If it 
arrived early enough you would hope that it will be inserted in the 
packet queue at the right point. Don't count on it, though. I've found 
some pretty quirky behave with some equipment when packets are reordered.

Regards,
Steve