[Asterisk-Users] Still unacceptable echo on X101P

[Rich Adamson]

> I am still having unacceptable echo on my X101P and twidling with the
> rx/tx gain levels and echo settings appears to have no discernable
> effect.
> 
> Some questions for those who may have more significant electrical
> engineering background than I.
> 
> 1. This impedance match thing ... will it affect this solution having
> other phones in parallel with the X101P? This is done so that I can
> test while not having the system pickup/handle all the calls in the
> house until I'm ready to launch it.

In most cases, a bridged analog phone will have no impact on this. I've
tested with multiple phones and no phones, and have not found any
differences. There are some phones (not necessarily cheap ones) that
do create a significant line imbalance and have been proven to impact
echo. Its easy enough to simply disconnect the phones (or connect *
directly to the pstn line without the phones) and prove to yourself
the phones aren't the issue. If an analog phone goes off-hook at the
same time a call is in progress via asterisk, echo 'can' happen.

The impendence mismatch issue ususally pops up when an x100p is used
to attach to a pstn line (in some non-US countries) where the x100p 
represents a 600 ohm impedence, and the telco standards are 135 ohms
(or something like that). The x100p chipset was designed for 600 ohms
only. Its my understanding the Canadian and US standards are the same
(600 ohms), so that shouldn't be your problem.

> 2. What about the effects of it being downstream from a DSL line filter?

No problem as long as the filter is in place. One of four pstn lines
in my office is a dsl line with filter; no problem whatsoever. In all
liklihood, if you had an issue with the dsl filter asterisk interfaces
would kill/impact the dsl before the dsl would impact asterisk.
 
> 3. If impendance mismatch is the (or a major contributing) factor, can
> we not devise some interface circuit which will allow a variable rate
> on the impedance so we can "dial out" the echo based on individual line
> conditions?

That would be called the TDM card. It has a chipset that was designed
to match telco standards in many (if not all) countries. In Canada,
if you have x100p echo problems, you'll have them with the tdm card
as well.

You'll find plenty of discussion in the archives relative to x100p/tdm
echo problems unrelated to impedence mismatches. Seems there is an issue
with interrupt latency, pci controller, or something like that associated
with some motherboards that "is" impacting echo (as well as other things).

Multiple individuals have found that replacing their motherboards fixed
the echo problem with these cards, but no one (as yet) as put their 
finger on why with any degree of accuracy. Lots of opinions, but no
real facts to date.

I believe its fair to say:
 - processor speed has nothing to do with echo
 - amount of RAM has nothing to do with it
 - high end commercial systems can have it; some low end systems do not
 - interrupt sharing has nothing to do with echo (although it can be
   the source of other issues)
 - other apps running on the system has nothing to do with it (but
   can generate other issues)
 - the specific OS distribution has nothing to do with it
 - moving x100p/tdm cards around on a system has nothing to do with it
 - replacing an x100p with a tdm card won't fix it (in countries with
   600 ohm telco standards)
 - the brand name on the system (eg, Compaq, HP, IBM, Dell) has nothing
   to do with it.

There is a high probability (but this is a wild guess) the echo issue
has something to do with the specific chipset used by the motherboard
manufacturer, and if that guess is correct, probably something to do
with how interrupts are handled or pci controller issues, etc.
Something on certain motherboards seem to be delaying the transfer of
data to/from x100p/tdm cards, and that delay is sufficient enough to
fall outside the limits of the echo cancellation software within *.

If you have the capabilities and/or experience to narrow the issues
down further, lots of folks would like to hear more facts.

Rich