[Asterisk-Dev] has somebody ever tested faxdetection ?

Steve Underwood steveu at coppice.org
Mon Nov 7 09:36:18 MST 2005


Thomas Häger wrote:

> Hi all,
>
> lately we had a little bug with misdn echo cancel. If a fax was 
> detected we got a kernel freeze.
> This was caused by NULL pointer access inside of our echocanceler 
> code. In this context we wanted to test this issue by calling a fax 
> machine to effect the kernel freeze. But nothing happend. So we 
> debugged a little bit, and we saw that the tone of 2100 Hz was never 
> detected ... respectively the amount of 3 cyles were never reached.
>
> In debugging we saw several values for "det->tone_cycle_duration", but 
> no value was between   (425*8) 3400 and (475*8) 3800.
> This must happens more than twice, so that ec will be disabled.
> Has someone an idea what exactly is the meaning of these values 
> (3400-3800) ?
> You can find the function in "zaptel/ecdis.h".

The detector you are looking at does not detect continuous 2100Hz tone. 
It detects a 2100Hz tone with phase reversals every 450ms. This is the 
echo canceller disable tone specified in G.164/G.165. It is not the FAX 
answer tone. The numbers you quote allow a suitable tolerance for the 450ms.

> We have tested it as follows:
>
> SIP --> Asterisk --> chan_msidn (TE) ---> PSTN / ISDN(PtP) ---> 
> faxmachine other side.
> How customers reported us, they got a kernel freeze in this case. But 
> we didn't get a kernel freeze, because the 2100Hz was never detected, 
> and EC was enabled all the time.
> We tested also fax detection with playing back a reference audio file 
> of 2100 Hz (5 sec.), but nothing happened.
>
> Has someone enough experiences with fax detection and can give us a 
> hint why the fax detection didn't work for our configuration ?

The FAX answer tone is detected in *, not in the driver.

Regards,
Steve




More information about the asterisk-dev mailing list