[Asterisk-Users] Bizarre Echo Problem

[Florian Overkamp]

Hi Mark,

Citeren Mark Edwards <mark.p.edwards at gmail.com>:

> to add some fuel to the fire, I was monitoring one of the agents last night.
> He made a call to a target and then had to call them straight back to
> confirm some information.
>
> The first call was as echoey as the inside of a cathedral.
> The second (next) call was as clear as a bell.

Although I did not follow your thread from the start, I can confirm there are
significant differences in zap-channel inbound versus outbound
echocancellation, i.e. the direction of initiation of the call is important.

As far as our tracing went, the echo-cancellation itself was properly activated
on the channel, but the exact moment of canceller-training may not be perfect.
Think of when training can occur _before_ the entire call-path is up and
active. This would mean the training is infact useless and may even be
increasing problems.

We have been managing the issue via implementation of a small loss-plan and some
mods to chan_zap, although I need to see with our developers on how that was
done precisely.

Best regards,
Florian