[Asterisk-Users] Current echo status?

[Rich Adamson]

> I've been following the list for months, and I have a working Asterisk 
> setup, but it'd be *really* useful to me at this point if someone could 
> summarize when Asterisk has echo problems and when it doesn't.  For 
> instance, I usually hear a far-end echo when talking on my 7940, but 
> not when using a POTS phone plugged into a TDM400 FXO port.  It doesn't 
> seem to matter if the call goes out over a POTS line or via NuFone; 
> either way there's a fair bit of echo on most (but not all) calls.
> 
> Do different SIP phones have better echo cancellation then the Cisco 
> 7940/60s?  How about the Polycom IP500/600s?
> 
> Does it matter if calls go out via POTS/T-1/PRI/VoIP?
> 
> The general impression that I've received is that "fast" channels 
> (basically traditionally telephony interfaces) don't exhibit noticeable 
> echo, but the slight delay associated with VoIP packetization unmasks 
> existing echo.  Is that a reasonable summary?
> 
> We're starting to plan for a new office build-out at work, and I'd love 
> to use Asterisk and SIP phones in the new office, but I'm not going to 
> try to sell management on a phone system with a horrible echo problem, 
> even if it will get fixed eventually.

For the most part, echo is less of a problem today then what it was
a month or so ago. The source of the echo varys case by case, however
key points that need to be addressed in _any_ installation include:
 - * fxo pstn interface cards (eg, x100p, tdm-fxo) have to be configured
   to match the impedance of the telco standard for whatever country
   you're in. The x100p, by design, appears to be limited to a 600 ohm 
   telco impedance (US standard), and will likely generate echo when
   attached to pstn lines that are not 600 ohms. The tdm cards are
   different and can be configured to match just about any telco standard
   worldwide. (Don't know about the x101p, never seen one.)
 - echo issues with the zap interfaces (eg, x100p, tdm, T1/E1 cards)
   rely on a software echo canceler within asterisk. If the above
   impedance match is correct, most implementations seem to have found
   parameters that minimize echo for them. However, there are still some
   in which echo is a problem and best 'guess' as of this afternoon is
   those cases appear to have something to do with undocumented internal 
   system hardware. Some folks have found swapping a motherboard for
   another with no other changes reduces echo by noticable amounts.
   That would suggest buss speed, PCI version, or something like that
   other then processor speed or RAM. There are working examples of
   300 mhz machines with no echo, 2.2 ghz machines with echo, and dual
   processor systems with echo.
 - Digital interfaces to * (eg, T1/E1, PRI, ISDN) tend to be less prone
   to echo, however there are some implementations that still have it
   and non-technical users of those systems do complain. Technical
   users tend to tune it out.
 - there has been very little list noise associated with echo that 
   could be honestly pinned on any sip phone.
 - there's been a lot of opinions stated about the cause of echo on 
   the list, and at least some have no technical factual basis.
 - there's a rather strong belief that additional echo problems exist
   within asterisk, and a group of non-programmer types are attempting
   to isolate common items in an effort to document the problem for
   those that have the programming skills to address. (That's happening
   on a non-asterisk email list.)

It seems that production systems either don't have any echo issues, or
they have objectionable amounts. There does not seem to be anything
in between. Its probably fair to say the majority (if no all) developers
don't have the problem, making it most difficult to isolate and
troubleshoot the cause.

Best guess (based only on the list noise level since Octomber 2003) is
that probably 80% of the production * systems are near echo free, and
20% still have objectionable echo. Others may argue a different ratio. 
YMMV.

Rich