[Asterisk-Users] Email2fax big problemo

[Lee Howard]

Andrew Nowrot wrote:

> Email with attached .pdf file ---> Asterisk 1.2 box with qmail, 
> connected to PSTN by ISDN HFC-S card, with a_law codec---> PSTN 
> (ISDN)---> Another Asterisk with the same card and codec ---> fax 
> machine connected to Asterisk by Digium card. 


So...

Asterisk -> alaw ISDN HFC-S -> PSTN -> alaw ISDN HFC-S -> Asterisk -> 
TDM -> Fax Machine

Hrmm... well I'm unfamiliar with how the ISDN HFC-S cards work, so let 
me just discuss the general case of "fax machine only receives a partial 
page" problem with you a bit.

A typical non-ECM fax receiver will start receiving modulated page image 
data using a high-speed carrier.  When the carrier is lost/dropped then 
the receiver assumes that to indicate the end-of-page marker.  "Carrier 
loss" is most usually synonymous with ~10-40 ms of silence being 
detected.  (Now, some techies may wish to debate me on the accuracy of 
that statement, but what I'm saying is that in practice carrier loss 
detection is acheived by detecting silence.)

Now txfax/rxfax don't support ECM, but for comparison, ECM fax receivers 
work similarly, but the page image data is encoded and the ECM protocol 
allows for the receiver to request partial page retransmissions and 
tends to be a bit more resiliant to premature carrier loss detection.

So if the receiver is only getting a short page, a page that is 
truncated or cut-off somewhere in the middle then it means that 
"something" happened to the carrier in the middle of the page.  The only 
definitive way to know exactly what the receiver saw would be if you had 
some DSP logging mechanism on the receive-side (probably not), so 
usually you just have to guess and make tests based on the guesses.

Make a recording of the call on the Asterisk receive-side (the monitor 
app should work for this).  Then open that recording up in something 
like Audacity and look at the audio close enough to see every 20 ms.  
Look for a small "glitch" in the audio... maybe a small portion (even 
only 20 ms long) of silence.  If you find it, then you've found the 
problem, and then you need to find out where that corrupted audio is 
coming from (make recordings all along the call path and compare them, 
see where the corruption is occurring).

If the audio is clean ... well, then I guess you should just ignore me :-)

Thanks,

Lee.