[Asterisk-Users] Help with strategy for echo cancellation.

[Rich Adamson]

> I have just installed * (RH9, P4 3.0GHZ, 1G RAM) in a small office,
> using three TDM400's with 4 FXO's each for incoming calls.  Outgoing
> calls are (for the moment) routed via VoicePulse.  Phone sets are Cisco
> 7940G's using SIP.  I'm getting intermittent echo on outgoing calls, and
> my understanding, based on reviewing the wiki and several posts here, is
> this:
> 
> 	>>>>  The source of the echo is the analog tail circuit at the
> far end of the call.  This is consistent with the facts -- I don't have
> echo on internal calls or on IAX2 calls to another Cisco 7940 on another
> * box.

That's because those paths are basically equivalent to 4-wire full-duplex
links. The only place where echo 'could' be generated is from within the
sip devices themselves, or from the handset (eg, echo tunneled through the
handle). Handset echo 'has been' an issue with some cheap sip phones, and
usually stuffing foam rubber into the handle takes care of it.

> 	>>>>	 There's nothing that I can do about the echo using *
> echo cancellation because (according to Cisco's Echo Analysis paper) my
> echo cancellation only deals with echo originating at my end.  (Am I
> wrong?  Hope so).

Not necessarily true, but could be. If you're hearing your own voice
when talking, you're getting feedback from "something" along the path.
In very general terms, the delay between a spoken word and when the
feedback (echo) occurs should help determine the source location, but
you have to listen very closely. The greater the time between a word
and the returning echo, the further the source of echo is from you.

The Cisco paper assumes a near-perfect world; be careful with assumptions.
Example: it assumes that if you have an echo canceller running on
your end that its doing what it is supposed to be doing (eg, a quality
echo canceller). As you've seen in many many earlier posts, the * echo
canceller is not a high quality piece of software and has a rather 
narrow range of operation. When echo occurs outside that range, the
canceller is not handling it at all. Also, you've probably read some
of the posts relative to differences that motherboards have on the *
echo canceller; if the delay in moving packets from * to the TDM cards
and asterisk reading packets back from the TDM card is lengthy, then
you will hear echo. Depending on how long that delay actually is, you
can easily jump to an incorrect conclusion that its caused by far-end
problems (per the Cisco document), when in fact its not. The motherboard
issue has something to do with interrupt latency and/or pci bus
characteristics, and has absolutely nothing to do with the speed of
the processor, brand name on the front of the box, or far-end echo. 
(I've not heard anyone in six months actually offer up a way to figure 
out what the issue truly is, just lots of opinions thus far.)

> 	>>>>	 I may be able to minimize the problem by tweaking the
> Rx/Tx gain in zapata.conf.  
> 
> So, if my understanding is right, can someone please suggest a strategy
> for adjusting the gain controls?  There are two controls, and each can
> be adjusted up or down.  I'd like to adopt a method other than random
> fiddling.  Where's a good place to start?  

Each site seems to be a little different, so there's no such thing as
a good value to start at. Some machines are very close to Central Offices
(where cable loss is insignificant) while others are some distance from
the Office (where the loss might be 10 db or so). In very general terms
start with 0,0 (rxgain, txgain) and adjust in maybe 2.0 db increments.
In most cases you need to stop/start asterisk (not just a reload). The
smaller the gain settings, the less echo, but will also become difficult
to hear as well. Going to low will also kill DTMF and/or CallerID functions.

Rich