[10GBASE-T] FW: emailing the reflector
From Stephen Bates...
-----Original Message-----
From: Stephen Bates [mailto:stephen.bates@massana.com]
Sent: Thursday, January 16, 2003 11:24 AM
To: Booth, Bradley
Subject: emailing the reflector
Hi All
I was wondering if people were willing to give an opinion on the coding
scheme for 10GBASE-T. As we mentioned in our presentation at Vancouver
we do
feel that the balance between SNR required at the slicer and complexity
of
the convolutional code may not be optimal.
Even if the issues regarding alien NEXT are addressed, the requirements
on
cancellation and jitter are very severe. This implies a very strict set
of
design criteria on the mixed signal components of the transceiver. We
could
relax these somewhat by moving to a coding scheme that has the potential
to
provide more gain.
On a similar note we also mentioned pre-coding. We did this for a couple
of
reasons:
1) By performing the DFE operations at the transmitter it is possible to
simplify the convolutional decoder.
2) We have some concerns about code collapse and recovery from error due
to
the very band-limited nature of the channel.
3) Just because the pre-coding is included in the standard it need not
be
used in a specific transceiver architecture. The receiver has the option
of
expressing the desire not to use precoding. It can do this either in the
auto-negotiation step or in the step required to pass the pre-coding
coefficients back to the transmitter.
I'd appreciate any input people have on these issues and how they feel
we
might go about getting some answers. Until we have a figure for the
expected
coding gain of the system we cannot determine the required slicer SNR
target
for a given line code (However we can perform comparisons between
alternative line codes assuming a constant coding gain).
Stephen Bates