asterisk users - Jan 2007 - TDM400P, fxotune and ADSL filters - Just a FYI, FWIW

This may be commonly known but I haven't come across it so here goes,
maybe it'll help someone:
 
I have terrible echo with asterisk 1.2, zaptel 1.2.12, and a TDM400P
with 1 FXS and two FXO modules. 
The Mark2 echo canceller with Aggressive turned on was the only setting
that would make it acceptable.  
I found fxotune with this zaptel version to be broken.
 
I pulled the latest fxotune.c and fxotune.h from cvs and recompiled
zaptel.  
fxotune then ran but I got the error:  Could not fill input buffer - got
-1 bytes, expected 4000 bytes Failure!
After two days I installed a splitter to listen in and found out that
fxotune wanted 18 seconds of silence on the line but Bellsouth only
gives 15 seconds.
The -m switch in ./fxotune -i -m 15 -vv -b 3 -e 4 -t 2 allowed the test
to complete successfully.
 
Before tuning the TDM400P with ./fxotune -s, I observed the echo
percentage on the line with ./fxotune -d -b 4 to be .32, a far cry from
the .05 I wanted.
After ./fxotune -s, ./fxotune -d -b 4 revealed an echo percentage of
.075, still not good enough.
 
I remembered that there is a DSL filter between this FXO module and the
PSTN to break out signal for my DSL modem. I removed it and plugged the
FXO straight in to PSTN.
After a rerun of ./fxotune -i -m 15 -vv -b 3 -e 4 -t 2 and a ./fxotune
-s, ./fxotune -d -b 4 now reveals .026 percent echo! 
It appears that the DSL filter circuitry affects the .fxotune impedance
test to the point that it becomes ineffective (~.05 delta in my case)
 
FWIW, I replaced the filter and reran ./fxotune -d -b 4 and observed a
report of .11 percent echo, which I do not trust due to the filter's
effect on the circuit.
 
I eagerly removed the aggressive suppression and restored the original
echo canceller to be disappointed that the echo still exists.  So it is
back to Mark2 with Aggressive.
 
If you hang a FXO module behind a DSL filter and have high echo
percentages or echo, this is a gotcha.
 
I'm now experimenting with zaptel 1.4 with similar results, despite a
new default echo algorithm.  
 
Also, any tips on echo reduction from here would be greatly appreciated,
I'm out of ideas.  My biggest fear is installing a hybrid system in a
client's office and to come across a situation where I can't suppress
echo..

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.digium.com/pipermail/asterisk-users/attachments/20070115/b62d3ea6/attachment.htm

Eureka, echo free at last! ahh....
 
I set the rxgain by running my CO's milliwatt test to 14844 from the
original 6688.
I just looped from FXO 3 to the asterisk milliwatt() test on FXO 4, I
just found the txgain was 6686, instead of 14844. 
(http://lists.digium.com/pipermail/asterisk-users/2004-November/064312.h
tml)
After bumping the txgain to 6 (!), I got it to 13500 and that was all I
could get.  However, The echo has disappeared.
 
Sorry to answer a question that hasn't been asked, but maybe this will
save someone some serious frustration! 

  _____  

From: John French 
Sent: Monday, January 15, 2007 11:40 AM
To: asterisk-users@lists.digium.com
Subject: [asterisk-users] TDM400P, fxotune and ADSL filters - Just a
FYI, FWIW



This may be commonly known but I haven't come across it so here goes,
maybe it'll help someone: 
I have terrible echo with asterisk 1.2, zaptel 1.2.12, and a TDM400P
with 1 FXS and two FXO modules. 
The Mark2 echo canceller with Aggressive turned on was the only setting
that would make it acceptable.  
I found fxotune with this zaptel version to be broken. 
I pulled the latest fxotune.c and fxotune.h from cvs and recompiled
zaptel.  
fxotune then ran but I got the error:  Could not fill input buffer - got
-1 bytes, expected 4000 bytes Failure! 
After two days I installed a splitter to listen in and found out that
fxotune wanted 18 seconds of silence on the line but Bellsouth only
gives 15 seconds. 
The -m switch in ./fxotune -i -m 15 -vv -b 3 -e 4 -t 2 allowed the test
to complete successfully. 
Before tuning the TDM400P with ./fxotune -s, I observed the echo
percentage on the line with ./fxotune -d -b 4 to be .32, a far cry from
the .05 I wanted. 
After ./fxotune -s, ./fxotune -d -b 4 revealed an echo percentage of
.075, still not good enough. 
I remembered that there is a DSL filter between this FXO module and the
PSTN to break out signal for my DSL modem. I removed it and plugged the
FXO straight in to PSTN. 
After a rerun of ./fxotune -i -m 15 -vv -b 3 -e 4 -t 2 and a ./fxotune
-s, ./fxotune -d -b 4 now reveals .026 percent echo! 
It appears that the DSL filter circuitry affects the .fxotune impedance
test to the point that it becomes ineffective (~.05 delta in my case) 
FWIW, I replaced the filter and reran ./fxotune -d -b 4 and observed a
report of .11 percent echo, which I do not trust due to the filter's
effect on the circuit. 
I eagerly removed the aggressive suppression and restored the original
echo canceller to be disappointed that the echo still exists.  So it is
back to Mark2 with Aggressive. 
If you hang a FXO module behind a DSL filter and have high echo
percentages or echo, this is a gotcha. 
I'm now experimenting with zaptel 1.4 with similar results, despite a
new default echo algorithm.  
Also, any tips on echo reduction from here would be greatly appreciated,
I'm out of ideas.  My biggest fear is installing a hybrid system in a
client's office and to come across a situation where I can't suppress
echo.. 


-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.digium.com/pipermail/asterisk-users/attachments/20070115/0c2e2a4c/attachment-0001.htm