[Asterisk-Users] Re: Asterisk in production as a fax server, anyone?

[Rich Adamson]

>      >> However, based on the comments you give I'd suspect that you're
>     having
>      >> what people seem to be calling "frame slipping".  There seem to be
>      >> some motherboards that react poorly with Zap cards (or their
>      >> respective drivers) and cause that.  Your zttest results should be
>      >> revealing here.
> 
>     Frame slips are NOT motherboard related!

For those that happen to be following this thread, be careful with the 
above statement as one person is "assuming" T1/E1 interfaces while 
another is "assuming" the statement applies to analog interfaces such as 
the TDM400 card. The statement applies to one assumption but not the other.

>     A Frame slip is due to clocks at opposite ends of a circuit such as a T1
>     running at different speeds. Either a buffer overflows and one frame is
>     thrown away or there is no data when a frame is needed so the previous
>     frame is repeated.
> 
>     The solution is to have one end of the circuit supply the clock and the
>     other end derive the clock from the incoming signal.
> 
>     Don Pobanz
> 
> 
> How would you check clocks speeds at opposite ends of a circuit (T1, E1, 
> BRI, ...) ?
> 
> As it seems frame slips occur from time to time (for instance, on 10% of 
> received faxes), do you imply that Asterisk settings should be changed 
> so that on every fax received, it should adopt opposite clock speed 
> (unlike today where by chance, 90% of circuit clock speeds are the same) ?

The issue on T1/E1's is not clock speed itself but rather the low level 
"synchronization" of the clocks. The timing signals necessary for clock 
synchronization are "always" embedded in the transmit side of every 
T1/E1 data stream. Its part of the T1/E1 design specifications and 
cannot be removed under any circumstance by anyone. Whether you use it 
or not is defined in /etc/zaptel.conf.

The telco's and other T1/E1 service providers "never" listen to your 
synchronization; rather, they have a very well understood hierarchy 
where synchronization is always derived from their upstream providers 
(whoever they happens to be). Therefore, if you synchronize your clocks 
to the T1/E1 provided to you, your clocks will be synchronized to the 
rest of the world (in total).

FWIW, the folks at sangoma have said that one of there design 
verification tests includes ensuring that faxing works since it is one 
of "the" most critical tests that validates overall design.

So, if you are absolutely sure that you've specified the correct T1 
synchronization parameters in your /etc/zaptel.conf and you still have 
fax reliability issues, look elsewhere in your implementation for the 
root cause.