[Asterisk-Users] tdm (and x100p?) echo - fix is coming!

[Rich Adamson]

> > > I am curious did you play with the echotraining flag
> > > echotraining=yes or use the delay values for echotraining=some ms ?
> >
> > Per the doc in the configs samples, you have to implement
> >  echotraining=800 (instead of "yes") to take advantage of the new code
> from >yesterday.
> I'd suggest that one may need to play with the delay to see what the
> tolerance is on your line, I am also interested as to why different line
> require
> different delay's b4 kicking in the echo can, is this due to the different
> switch that the line is connected to ?
> Rich, it would appear on your config the longer delay was the key to
> removing echo

Just guessing here, I'd have to bet some money on the echo can being
related to the CO switch. The reason for that bet is that historically
(at least for last seven months or so), people either had echo or they
didn't (like major 'yes' or no), and none of the research tracked against
config parameters, motherboard model, pstn lines from a common source, 
type of lines, etc.

Changing to 800 had such a significant impact that it's almost like
someone purchased and installed an external echo can box.

I don't think that I'd play around with other values though. There were
three lines of code changed in chan_zap.c and two of those lines were
dependent on values from each other. Example: changing from 400 (default)
to 800 required another statement change from "w" to "ww". I'd have to
assume the "w" is a character/tone code to be used within the cancel
training (or something like that). If you changed to echotraining=600,
then how is the code going to insert a half a "w"? (I didn't look at
the final cvs code, but would guess the choices are "yes" or "800",
and everything else is silently ignored.)

If you have echo, then try echotraining=800. If it disappears, don't
fix something that's not broken.

Rich