Geoff:
There was never any
doubt about the viability of 1000BaseT over installed copper cabling during
the development of the 1000BaseT standard. The only question was precisely how
to do it (what line code, error correction, noise margins, etc). As you can
tell from the debate that is going on this reflector, several
individuals including myself have expressed serious doubts
regarding 10GBaseT's theoretical feasibility over 100 meters of CAT5/6
regardless of precise line codes and implementation aspects. These doubts are
well-founded since they were backed up with detailed MATLAB analysis
programs. This is exactly why we are considering the Shannon Capacity
limit now. My point was that regardless of the innovations in DSPs and line
coding schemes that may or may not have happened since 1997, this limit is
only a function of the properties of the cabling, background thermal noise and
alien NEXT assuming that all other self impairments are
canceled.
Sreen
-----Original
Message-----
From: Geoff
Thompson [mailto:gthompso@nortelnetworks.com]
Sent: Monday, March 03,
2003 3:17
PM
To: Sreen Raghavan
Cc: GEisler@aol.com;
stds-802-3-10GBT-Modeling@ieee.org; stds-802-3-10gbt@ieee.org
Subject: RE: [10GBASE-T]
Reflections
Sreen-
We did not use or even
consider the Shannon capacity of twisted pair cabling in our
previous work.
We did do a lot of investigation into twisted pair cabling
characteristics when we developed 1BASE5 and 10BASE-T. After that experience
we decided that we didn't want to do cabling and worked very hard
to:
1) See
that useful specs for our applications were developed in TIA and SC
25.
2) Use
those specifications.
When those cabling groups were starting on their
specification of Cat 6 we did ask them to characterize the cabling well beyond
the zero ACR point (their traditional cutoff point). They, in fact, did so.
What all of this discussion is about is that they didn't go out far enough so
support this project.
Geoff
At 10:17 AM 3/3/2003 -0800, Sreen
Raghavan wrote:
While
there may have been innovations in DSP architectures since 1997, Shannon
Capacity of CAT5/6 cables is a theoretical limit, and is unaffected by an
individual s time horizon, or implementation sophistication.
Sreen Raghavan
Vativ Technologies, Inc.
-----Original
Message-----
From:
owner-stds-802-3-10gbt@majordomo.ieee.org [mailto:owner-stds-802-3-10gbt@majordomo.ieee.org]
On Behalf Of
GEisler@aol.com
Sent: Friday, February 28,
2003 8:44
AM
To: stds-802-3-10GBT-Modeling@ieee.org;
stds-802-3-10gbt@ieee.org
Subject: [10GBASE-T]
Reflections
Guys,
Nice to hear from a
lot of the old crew that created the 1000BASE-T standard. I think that you can
all take a bow, judging from what is happening in the market and the
remarkable improvements in implementations achieved in three short years. But
no good deed goes unpunished, as they say, so you guys earned the job (and
created the need for) taking the next step up. Without a doubt, there are big
jumps in the performance demanded and we don't have the luxury that 1000BASE-T
enjoyed, namely the introduction of coding into the design. That has been done
already. This time around, we have to take advantage of the residual capacity
of the cabling, which we barely touched.
On the other hand, you
will agree that the levels of sophistication in analysis and implementation of
DSP based systems has also increased dramatically since 1997, when GigT
started in earnest.
While I enjoyed fond memories of your recollections
of what we knew (or suspected) then and what we added later when we got
smarter, it strikes me that that history is relatively immaterial today. That
was then and our current circumstances are quite different. We now need to
agree in the study group that it is feasible to do 10Gig on some form
of twisted pair cabling at the desired length and go on to form a Task Force
to do the hard work of coming to agreement on the precise details of the
signaling and channel specifications.
It is clear that the channel
specifications as written in 11801 will have to be modified and expanded for
our use and installation qualification test parameters will have to be
defined. This is more or less what happened for 1000BASE-T and seems to me to
be a normal and reasonable step. After all, why should the cabling industry be
expected to have already characterized cabling for the use of a standard for
which we are still in the Study Group phase? As for installed cabling, we will
see whether or how it may be qualified for 10G as part of our effort.
I
look forward to our meeting in Dallas and hope
that we can get to work on a PAR draft there.
George
Eisler