[Asterisk-Users] Clarification on Fax capability?

[Rich Adamson]

> Wondering if someone (Steve?) can clarify something form me.  I think the
> recent "soho fax solution?" thread has mixed things up for me.
> 
>   - Is it possible to get reliable fax reception using
>     a Zaptel FXO interface connected to a standard POTS
>     line and a fax machine connected to station interface?
> 
>   - Is it possible to reliably send outboud faxes in
>     the reverse direction?
> 
> I understand the issues with fax over VoIP.  I just want to handle faxes
> here in my small office without dedicating a line to the machine.

The function of faxing over a zaptel interface (eg, x100p, tdm) is 
not consistent from one implementation to another, regardless of
whether your using Stable or Head.

The two basic approaches that are commonly discussed on the list are:
 1) using Steve's spandsp patches to intercept faxes, creating an *.tif
    file that can be emailed and viewed outside of *, and,
 2) simply switching a fax call through * to a tip/ring interface of
    some sort that has an attached traditional fax machine.

Option #2 works in many cases "if" the incoming fax call is handled
by the g711 codec. The fax call is no different then receiving any 
other call, however any analog-based modem tones passed through 
a digital interface (eg, asterisk) are _not_ reproduced reliably. 
The higher the modem speed, the greater chance of distorting the 
analog signal tones, and the greater the chance of fax not working 
at all. (You'll find the older-slower modems will work better then
newer-faster fax machines. Essentially, modems that operate at
9600 baud or slower are more reliable then anything faster.)

Option #1 works in some cases with Digitum fxo cards, however something
close to 50% (or more) implementations fail to work in any form of
reliable way. The issuse seems to be related to funcky things happening
with the zaptel/wctdm drivers that cause missed pcm frames (or slips)
between the digium cards and the asterisk code. Missed or slipped 
frames will negatively impact any modem-based communications, including 
fax machines.

Steve has received recent reports (#1) that suggest that outgoing faxes
are now failing at a significant rate indicating that something has
changed in the zaptel/wctdm drivers negatively impacting such calls.

No one (to the best of my knowledge) has complained about the drivers
impacting voice, however small numbers of missed or slipped frames
are generally not noticed (or are not that objectionable).

If you dig through the archives, you'll probably notice a fair number of
folks having issues with the digium x100p/tdm cards that revolve around
interrupt latency and/or pci bus problems. Those issues tend to be 
associated with echo and many have found that swapping motherboards
corrects the problem. But, no one (to the best of my knowledge) has
ever discovered "why" swapping motherboards corrects the problem; they
just know that it does. Likewise, no one has assembled a list of 
motherboards that work simply because it is very difficult to determine
motherboard models when companies like Dell & HP do not publish who's
board they actually use in their products (not to mention that some
system vendors have different boards in the exact same system models).
We do know that processor speed, number of processors, amount of ram,
etc, has little or no impact on the issues.

Some have noted that moving from a P4 to a slower speed P3 processor
has corrected the issues, but those type changes essentially have
hundreds of other changes along with that switch. Its unlikely the
actual processor switch had anything to do with it; more likely is
that changes that came along in the form of a different pci bus 
structure actually improved it (or something like that).

So, the best guess (today) is that a driver or hardware problem (pci
chip set) exists between the digium cards and the underlying motherboard
that negatively impacts the reliable transfer of data from those cards
to asterisk code, thus impacting voice (echo) and fax (modem signals).

It doesn't help that the drivers and cards are essentially considered
digium property and that all support should come from that source.
Opening a trouble ticket with digium (on this particular issue) tends
to go directly into a black hole. Some of that is likely due to the
sophistication (and/or lack of understanding) of those responsible for
supporting those products since one has to fully understand how to deal
with specific hardware (eg, chip sets), kernel drivers, asterisk code,
etc. If you try to analyze the code in zaptel and wctdm you'll 
understand why that is.

Bottom line is the tdm card & drivers seem to be just okay for voice,
but no where near reliable or even predictable for fax. That's based
on cvs head and spandsp-pre9 code as of this morning.