[Asterisk-Users] tuning an x100p in Australia for echocancellation

[James Harper]

> 
> That would be called a milliwatt generator. It likely exists in their
> central office, but its typically used by their technicians to ensure
> new installations meet specs and sometimes in troubleshooting. Call
your
> telco repair number and see if they will give you the telephone number
> for it.

I'll try that. 1st line technical support at Telstra are not known for
their ability to provide useful information that isn't in their script.

> If they won't, you can basically do the same thing by dialing out from
> asterisk on one pstn line coming back in through a second pstn line,
and
> using the asterisk milliwatt generator. Or, if you have another
asterisk
> system available somewhere, call out through a pstn line to that
> asterisk's
> milliwatt generator. (Obviously its not as good as using a CO
milliwatt
> as now you have to take into consideration the loss from the second
> pstn line, but it is a way to get a handle on the transmission loss
> values, etc.)

Would the txgain on the 2nd line also come into play? I guess if you set
it to zero then you only have to deal with an estimation of the line
losses.

> > Next, while the TDM400 card has control over the line impedence
> > circuitry, the x100p doesn't.
> 
> Are the AU telephone standards the same as US standards (eg, 600 ohm
> impedence)?

This is a question I've been trying to answer too. I had a look at the
standard phone that Telstra would provide to customers about 5 years
ago, and it has an impedence switch on the bottom to toggle between
'NORM' and '600', which suggests that 600 ohms isn't the normal
impedence. 

On an au configuration example for the pap2 I have seen on the web, the
impedence is set to '220+820||120nF', which suggests that our standard
here isn't 600.

> > Does anyone know of an addon device which
> > can do impedence matching on the line, or of a modification to the
card
> > (eg component swapping) which could allow some manual adjustment?
> 
> Twenty years ago, the telphone companies in the US had several types
of
> hardware devices available for impedence matching, line balnacing,
etc.
> The devices were used to compensate for several different problems
that
> would be too costly to fix through conventional means. I don't have
any
> clue where one might find those boxes today since those types of
issues
> have essentially disappeared due to the heavy use of fiber, remote
line
> modules/units, higher quality cables, and other technology
advancements.
> Some older telephony jocks may still have some of these in their junk
> boxes.
> 
> Since I don't have a clue what the AU standards are, I really can't
> guess at what might be needed in your particular case.
> 
> One such example that was fairly common back then was a simple
transformer
> that had two primary windings and two secondary windings. One could
buy
> them as 1:1.5 (600 ohm to 900 ohm), 1:1 (600 ohm to 600 ohm), and
other
> commonly used impedances. The transformer along with two 2.0 ufd
> capacitors allowed DC to pass through, but changed the impedence from
> one value to another.

I've thought of using a transformer by itself, but obviously that
wouldn't let the DC signals required for looping through. Could you give
me an ascii schematic?

> 
> As far as substituting components on a x100p card, I don't believe
that's
> realistic. If you can read the part numbers on the chips used on your
> x100p compatible card, its not that difficult to check the chip specs
> to see what impedance value it supports. (For example, Intel and
Silicon
> Labs made some of the chips used to interface the winmodem cards to
the
> pstn lines. Those cards manufacturered for US sales used a 600 ohm
chip
> while those manufactured for other specific countries used a different
> chip to match those country-specific telephony specs.)

I think I've got a spare x100p so I'll check that one out. It's an ebay
purchase so the chances are it's an import...

> > Finally, my echo is really really awful early in the call but then
gets
> > markedly better, which I assume is a result of the echo training
done
> > during the call.
> 
> That's probably a valid assumption. Whether its the result of poor
> impedence
> matching or something else is questionable.

I have since done some checking of calling between the two extensions on
the pap2, and without the pap2's echo cancellation being active, I get
echo there too... and that's with either impedence setting.

> > Is it possible to grab the echo coefficients after
> > training and use them as a starting point for next time? Or would
this
> > vary too much between calls?
> 
> It's my opinion your thought process relative to preloading the EC is
> very reasonable, but in the past two/three years, those with the
ability
> to code such functions "insist" every call is different in terms of
those
> values. I disagree but don't have the programming skills to prove it.
> There is likely to be some middle ground where "some implementations"
> would benefit from it, and a lot more would not.

I modified zaptel.c to dump out the coefficients on ztdiag (just a
#define, removing a few comments, and updating some out-of-date code). I
then started a call and dumped them out every few seconds. They change
wildly, but a large change of the figures may have very little change in
the actual system behaviour of the ec, so I guess that doesn't tell me
anything!

I'll try and write some code to be able to retrieve and load the ec
values today, and see what happens.

After about 20 seconds, the echo is reduced to a quiet and very muffled
echo, which you probably wouldn't notice unless you were listening for
it, and is certainly better than the quality of most mobile phone calls.

Thanks

James