DSL Filters?

Larry Kayser kayser at sympatico.ca
Wed May 7 15:02:39 CDT 2003

>DSL Line conditioners, or filters, are installed between the incoming
>telephone line and each telephone set. The blurb states that "the
>in-line DSL filter eliminates all erratic impedance from telephone
>equipment that interfere with DSL".
>See for instance:
>Does anyone actually know what this is supposed to mean?

Yes, the narrative description, "the in-line DSL filter eliminates all 
erratic impedance from telephone
equipment that interfere with DSL", is really quite accurate.  The problem 
is really to isolate the on and off hook analogue telephone  connection(s) 
from the echo canceller in the front end of the DSL box.  This was always 
the major problem from the very beginning of digital loops, even back in 
the ISDN days and before that, was to get adequate performance out of the 
hybrid echo canceller.

Back so many years ago, in fact back to January 1980 when I returned to the 
telephone industry my first action was to quickly fund a digital loop 
product into Northern Telecom using a Ping Pong technology.  This came out 
in Northern DMS products as Datapath and it was done with Ping Pong 
transmission, time compression multiplexing, because we simply could not 
get enough performance out of the then A/D and D/A devices to do an echo 

Later, before the ISDN standards were settled, we managed to get a decent 
ISDN hybrid (it was a huge thing physically) working with the 2B1Q code, 
just in time to severely test the silicon manufacturers with their ability 
to put A/D's with low enough noise performance into DMS line card chips for 
the ISDN era.

No sooner had the NT manufacturing people solved the silicon yield problems 
that we raised the bar and made the first HDSL boxes using 2B1Q coding at 
about 800 KHz so we had a complete T1 on two copper loops for local 
distribution. This was done in 1987so long ago in todays terms.   AT&T was 
promoting a multi tone concept for HDSL which required a few dB of 
additional A/D performance which was eventually achieved and we all know 
what happened with the ADSL market.

So in fact the issue remains that every pots circuit action, if it is seen 
by the ADSL hybrid process, causes the ADSL up top to have to what we used 
to call it "retrain" - that is balance the echo canceller again.  The 
canceller range in the vastly cost reduced silicon being used now has such 
a narrow training range that the analogue filters are essential to separate 
the impedance changes and the make and break noise burst from being seen by 
the canceller.  The worst impedance problems on pots lines come from the 
random occurrence of ON and OFF connections of multiple POTS telephone 
sets.  There is another effect as well but at least it is a constant, that 
is the effect of random pots sets going ON and OFF and the "end effect" of 
the bridge tap created by wire in the telephone cable plant that keeps on 
going past the subscribers termination point.  If the bridge tap is long 
enough it will kill almost any digital loop transmission system.


More information about the Tacos mailing list