[Asterisk-Users] Digium cards, so disappointing !

[Steve Underwood]

Kevin P. Fleming wrote:

>Rusty Dekema wrote:
>
>  
>
>>If this works, I don't see why a fax transmission wouldn't work. Is it
>>because the fax protocol doesn't have error correction? Is that even
>>true?
>>    
>>
>
>FAX transmission is massively more complex than modem transmission. At
>higher speeds, it involves 3 or 4 different 'carrier' frequencies and
>signaling rate shifts, and these are done with very critical timing
>requirements.
>  
>
What complete and utter rubbish, except the last bit - all types of fast
modem, FAX included, demand perfect timing.

>Yes, error correction is available, but it just means that sending FAXes
>over a lousy connection will take a very long time, instead of failing
>completely.
>  
>
This is accurate, though.

FAX is generally *less* demanding of channel quality than other fast
modems, because it never operates in duplex mode (some Super G3 faxing 
excepted). It just sends one
way at a time. All modems hate going into a digitized FXS/FXO port and
back out another FXS/FXO port, since every time the signal goes through
an analogue->digital->analogue cycle it gets corrupted a little bit
more. V.90 will never work. V.34 will rarely work. However, the V.17 and
V.29 modems usually used for FAXing tend to survive one cycle of this
OK. More than one cycle (e.g. both FAX machines going through these
analogue->digital->analogue cycles at the customer premises) is usually
a recipe for failure. If the FXS/FXO ports are of poor quality, they can
introduce enough distortion to cause trouble. Applying echo cancellation
can also cause substantial distortion, and must be avoided for modem
calls, including FAX ones. If echo cancellation is needed for a duplex
modem it must be applied end to end by the modems themselves. The real
killer, though, is imperfect timing. Normal telephone calls suffer
clicks, and pops, and hiss, and buzzing, and all sorts of other crap.
However, they never never never dilate time. The one thing you can
guarantee on a properly functioning PSTN path is flawless timing. Any
modem faster than the good old V.21 300bps modems from the 50's demands
perfect timing. Guaranteed perfect timing is never available with VoIP,
and its not always always available within a PC. PCs are designed around
best efforts handling of data. They don't handle continuous streaming of
media well, even if the data rate is fairly low. They handle it
especially badly if latency must be kept low, as is the case with
something as interactive as voice and modem calls.

That said, a well design PC environment can achieve the timing needed 
for FAX calls, as long as you don't load it up too much. It might not be
guaranteed perfect, in the way the PSTN is. However, if a timing hiccup
only occurs a couple of times a day, timing issues are not going to
screw up many calls. The problem for many people using cards like the
TDM400 is hiccups are occuring all the time. They are audible in voice
calls, though most people are undemanding enough not to complain or even
notice. FAX just won't work. If you use spandsp + rxfax there is
currently no FAX error correction. A hiccup kills the call. If you use
spandsp + iaxmodem + hylafax, hylafax provides error correction, but the
call time may become very long, with much retrying. If you use a FAX
machine plugged into an FXS port, results vary.

Regards,
Steve