[Asterisk-Users] Will Echo problems EVER be solved, I'm scared

[Rich Adamson]

> On Friday 26 August 2005 18:14, Rich Adamson wrote:
> > As mentioned in my last private email, all four analog pstn lines attached
> > to my TDM04b have been tested with a professional transmission test set
> > and all four are in excellent condition. Fxotune did not generate any
> > noticable differences to my ear. I'm only mentioning that for those readers
> > that might expect something different to happen and don't notice any
> > changes/improvements on US pstn lines.
> 
> Can you elaborate on the specific issues you're having?  It's nice to be able 
> to talk to someone who's got experience in telco and also computing (and CTI 
> too) to get to the heart of some of these issues.

I'm the one that opened bug 2022 and 2023 some time ago involving
low audio on VM when using either the TDM or X100P card. After doing
extensive testing, I also found that setting the rx/tx gains were no
where near realistic from a "telephony" perspective (eg, balancing
levels and echo), and no where near competing external gateway devices.

> Specifically, have you tried the KB1 echo canceller?  What about enabling MMX 
> (if you're on intel), enabling/uncommenting the CALC_XLAW #define and 
> compiling zaptel with 

Yes the KB1 as of today, and its good. Tried all of those other items
over the last two years, and worked with Mark to implement some of the
stuff that's there today (eg, echotraining=800). You'll also find my
postings related to the zttest and trying to identify some of the pci
(and/or interrupt latency) issues that cause spandsp to fail (95% of
the time) with the TDM card. Think Kevin/digium may have some resources 
poking at the pci stuff.

> KFLAGS+=-march=pentium4 (or whatever) and
> CFLAGS+=-march=pentium4 (or whatever)
> 
> in your zaptel Makefile just underneath the comments regarding everything 
> being done in zconfig.h.
> 
> I'm *really* interested in hearing what your findings are.

Any experience with how those flags might impact spandsp with the TDM?

My interest in TDM/echo postings is to help identify why the TDM card (and
older X100P card) are so unrealistic in the telephony standards world.
Just about everything that one does with those two cards have major
exceptions in telephony, and I'm a rather firm believer those exceptions
can be corrected. It just so happens that spandsp is an excellent tool
to help point out those exceptions. (I don't actually need spandsp
any more since converting to a commercial fax service.)

I've had a pair of digium x100p's for two years and a TDM04b a week or
two before it was annouced. The original TDM has since been replaced
with a rev H card, which is now stable, but stil has those other nagging
issues that we all seen posted in the last 12 months or so.

The TDM is in use on a production server and levels/echo are 'acceptable'
but can use some serious improvements. Hoping to help make the TDM an
even better card (and not rag on digium). Its heading in that direction
and there are some specialists truly focusing on it now. :)

As mentioned in several previous postings, the level/echo issues are far
more negative for those using asterisk a greater distance from the CO
then what other implementations are that are very close to the CO. That's
one of the reasons why there is such a hugh difference in what folks
see with that card. (Its almost like bits 5 & 6 are miswired (reversed)
on the card's digital audio path to (or in) the TigerJet chip.)

I had signed up to teach a basic telephony class (to cover all of this
telephony interfacing stuff) at the first Astricon in Atlanta, but 
cancelled out since there were so many different problems with these 
cards. (It wouldn't do much good to teach a class on how to do this 
only to say 'well, it doesn't work that way in asterisk'.)