[Asterisk-Users] strange echo problem

[Rich Adamson]

> We have a strange echo problem.  Maybe echo isn't the correct term.
> When we make a call f/ a SIP phone (we have several 7960's, some 3coms,
> and I've even tried a softphone, all on the same 100BaseTX network) to
> the pstn, if the person I'm calling has a PRI or channelized T1 f/ Bell,
> then the sound is perfect, couldn't be better.
> If I make a call to a person with a plain POTS line, I hear everything I
> say in my earpiece about 1/4 second after I say it.  It's very
> irritating.    We have tried 2 different * boxes, using 2 different
> T1/PRI cards f/ digium.  
> 
> After calling digium about it, we set echotraining to 800 in
> zapata.conf.  It got better but was still there, if I turn the volume
> down on the phone, it does "almost" go away, but it's still detectable.
> No where near as clear as calling a person that has a PRI or channelized
> T1 for phone service.  The POTS persons we call that we do have the echo
> issue with 
> all say the call sounds perfecto to them.
> 
> Am I missing something obvious?

No, your not missing anything specifically. There seems to be a fair
number of people with those same type of echo problems, which are the
most difficult things to diagnose. There's no easy way to determine
whether this is an * or pstn problem, so most of the previous efforts
have been focused on ruling out what is not the problem.

It would help all of us better understand the issues if at least some
additional data were provided, however. Such things as cvs version,
significant portions of 'zap show channel...' with the echo can data,
and anything else that seems to narrow the focus.

In most cases, turning volumes/gains down is a method to bypass the
problem, not solve the root cause. (Certainly not true in all cases.)

Personal opinion (based only on the various words posted to this list)
is there still is an echo can issue floating in * that none of us 
have documented with sufficient data that would guide a developer towards
the root cause. If you easily recreate the problem, that would go a
long ways towards gathering qualifying data.

Rich