[Asterisk-Users] RFD: With echo and other distortion, can ulaw/alaw line quality ever be good enough for faxing?

Wed May 5 12:35:25 MST 2004

Folks,

The silence was deafening ... I had a few private replies but overall I'd
have to conclude that most people on this list aren't interested in faxing
thru Asterisk. You're all probably jazzed about VoIP and fax is forgotten
for now ;-)

We have learned the answer to _one_ of the questions I asked below - the
Adit channel bank's gain controls can be usefully applied to adjust each of
the incoming telephone lines independently, and by combining the Adit's gain
setting with Asterisk's, fun and good things can be made to happen ;-)

Another interesting observation is that echo can be seen visually when using
ztmonitor. Muting one side of the call (the RX side), and speaking from my
handset I can see my voice appear stongly on the TX side, and then I can see
the echo appearing on the RX side at an attenuated level, but it follows my
voice perfectly ;-)  I spend half a day talking to myself, and I have to
say, for the first time in ages I was running out of things to say! :-)

In other interesting news, updating zapel from CVS seems to have made things
significantly better. I had CVS from about 4 weeks ago previously, so
perhaps something was amiss. We can actually get faxes transmitting now, and
although they fail about 30% of the time, this is definitely progress.

Another data point, is that MARK3 echo cancellation is audibly worse under
our particular conditions. We're using MARK2 with the aggressive option
enabled, and while occasional flares of distorted (but attenuated) echo can
be heard, it's much easier to have conversations with customers! I'm
disappointed nobody commented on the echo cancelation options in the
source... I can only assume nobody actually knows the origins of any of them
any more, and the other ones are mostly anachronisms that are no longer
useful, and MARK2 is the clear winner.

Unfortunately, the news from Digium's technical support is not good. Here's
what they had to say in response to our inquiry about faxing:

"Faxing in asterisk is only at varied levels of support and is
definitely not at a production level.  However, if you wish to work on
this we wouldn't mind at all :-).  The source code is of course free
to your disposal."

I'm not sure I understand that this support rep is saying to me ... we've
had much success faxing from one PRI to another with them both connected to
Asterisk via a TE400P. Asterisk doesn't even break a sweat. We're rarely
getting the theoretical maximum of 33,600 that we can get with a T1 loopback
though, so something's slightly amiss, but it's only when we try to go out
to the PSTN that things start to fall apart badly.

-Darren

--
Darren Nickerson
Senior Sales & Support Engineer
iFax Solutions, Inc. www.ifax.com
darren.nickerson at ifax.com
+1.215.438.4638
+1.215.243.8335 (fax)

----- Original Message ----- 
From: "Darren Nickerson" <darren.nickerson at ifax.com>
To: <asterisk-users at lists.digium.com>
Sent: Tuesday, May 04, 2004 12:35 AM
Subject: [Asterisk-Users] RFD: With echo and other distortion, can ulaw/alaw
line quality ever be good enough for faxing?

> Folks,
>
> I've been following recent discussions regarding echo and echo training
with
> much interest, since it's a problem we've never been able to eliminate
here.
>
> We're facing two challenges presently, and they may be related (or not):
>
> a) Cisco 7960s in the office here echo back to our staff, but the customer
> hears decent sound.
> b) We've yet to be able to pass faxes through asterisk, despite trying a
> number of approaches
>
> A bit about our setup:
>
> Two analog lines come into the office, into the FXO card of an CAC ADIT
600.
> These are 'connected' to the TDM interface of the ADIT, and go out to
> Asterisk via 2 channels of a a 24-channel fxs_ks line into a port of a
> TE405P. We have two other ports on the TE405P occupied by digital fax
> boards, connected to asterisk via AT&T 5ESS PRI. And of course there's the
> 7960 handsets, connected directly to a switch at 100MBps (fdx).
>
> Challenge a)
> =========
> In terms of looking into a), we started with:
>
> http://asterisk.sohoskyway.net/Asterisk_Doc/current/docs-html/x939.html
>
> Clearly some of the details are outdated, since the echo canceller options
> live in zconfig.h now, not a Makefile, but it's a decent starting point.
> Which gives rise to my first question - which echo canceler should we be
> using? We have Steve, Steve 2, and Mark 1-3, with an aggressive option
with
> 3 that could make calls scratchy ;-)
>
> Is there any conventional wisdom here?
>
> Next question ... we have control over gain at the zaptel level, and also
at
> the ADIT level apparently. Which should we use when trying to adjust
things
> with ztmonitor?
>
> Next question ... it's not clear to me what the target is with ztmonitor.
> Are we shooting for tx and rx levels that are balanced. and about 50% of
the
> scale at their max energy? I found that when I called a voicemail service
> and listened to the auto-attendant that the RX meter of ztmonitor was
about
> 50%, but I found the TX meter fluctuated wildly when I spoke. I could
reduce
> my voice level slightly and it would be very low energy, but with a slight
> increase in volume it was pegged. The settings I made seemed to do nothing
> to change this (I stop asterisk, unload and reload all zaptel modules, and
> restart astertisk between each test).
>
> Is this really supposed to work? I've managed to get the connection
> extremely distorted with some settings, but have yet to make a change that
> improved the quality or removed the echo.
>
> Challenge b)
> ========
>
> In some ways, this is the most worrying problem for us, since we would
love
> to be able to pass faxes through asterisk reliably, and from the traffic
on
> this list and the fine efforts of Mr. Underwood, it seems many others
would
> too. People have spoken here of the nearly immediate echo cancelation on
> pure TDM circuits, and so I would have thought we'd escape the echo
problems
> above in either of the following setups:
>
> i) Plain ole' HP fax machine plugged into TDM400 FXS card
> ii) Brooktrout/Eicon fax board connected to TE405P
>
> In practice we're rarely able to train with the remote fax machine when
> dialing to an outside line through asterisk and the ADIT 600. If we do
> manage to get connected and the connection supports ECM error correction,
we
> can see lots of retransmitted frames which again points to very poor line
> quality. In contrast, by connecting directly to the analog lines we can
send
> faxes all day with the trusty little HP ... ie: the lines aren't
inherently
> bad.
>
> I guess what I'm looking for here is a sanity check ... are we trying to
> push the limits here, or should this stuff be working much better than
this?
> We're willing to invest some time getting faxing to work, but I'd hate to
> ask people here to dedicate their time and energy to something that's
never
> really going to work well due to limitations in Asterisk/zaptel
technology.
>
> -Darren
>
> --
> Darren Nickerson
> Senior Sales & Support Engineer
> iFax Solutions, Inc. www.ifax.com
> darren.nickerson at ifax.com
> +1.215.438.4638
> +1.215.243.8335 (fax
>
> _______________________________________________
> Asterisk-Users mailing list
> Asterisk-Users at lists.digium.com
> http://lists.digium.com/mailman/listinfo/asterisk-users
> To UNSUBSCRIBE or update options visit:
>    http://lists.digium.com/mailman/listinfo/asterisk-users
>