[Asterisk-Users] Alternatives to SpanDSP??

[Colin Anderson]

>Having played around some with zttest (modifying the code to better
>understand the issues), it would appear the TDM card consumes about
>1.02 seconds to obtain one second of data. That would suggest the

I would like to chime in with my experience:

We are trying to use SpanDSP off of a PRI to recieve > 200 faxes a day. It's
gone OK but not perfect. Some gotchas that I have found: 

1. Timing (as others have said) is totally critical. I found a subtle timing
error because our Asterisk box is behind an Adtran channel bank and the
Adtran introduced tiny slips occasionally. Upgrading the firmware in the
adtran and monkeying around with how the Adtran took it's timing from the
PRI took care of it (after consultation with Adtran tech support which is
first-rate BTW)

2. ZTTEST is a critical metric. I was getting disconnects on about 20% of
faxes until I looked at the output of ZTTEST and found that it was dropping
below 99.98% occasionally. Using setpci I changed the latency on the Zaptel
boards (T100P & TDM04) to the max, 254 and cranked down the latency on
everything else as low as I dared. Now, I get 99.9873% across the board as
long as I run the test, and I even get the magic 100% on 1 in 10 test
passes. 

3. Yes, we have the HP problem, and I don't know how I'm going to deal with
it yet. I'll probably set up a "problem fax" line with an analog fax and
give that number to those people that have the problem. It's always the same
guys. 

I'm getting a reject rate of about 2-3% which is ok but the endusers of
course want no rejects. I have to offset that with the convenience of
getting the faxes as PDF's (we would take the paper fax and scan it into our
CRM if you can believe it) and the monetary savings of not printing the
faxes; we have a click rate from our print vendor and he loves it when we
make paper 'cause it's more money for him. No more busy signals on the fax
line is a bonus too, people being people the fax will sit idle all day then
15-20 faxes will try to come in simultaneously.