Re: [10GBT] Discussions on precoder decisions
We reduced the symbol rate on the PAM8 proposal from 1Gs/s to 952.381Ms/s,
recognizing that there is no need to adapt the TH precoder coefficients
during normal operations (please see
http://grouper.ieee.org/groups/802/3/10GBT/email/msg00947.html ). Therefore,
there is not much difference between 2 and 3 as far as I am concerned.
In any case, I don't believe that we can make the decision on the TH
precoder structure independent of the proposal that it is attached to. For
instance, it is my understanding that the ARMA precoder relies on a spectral
null at fs/2 for its correct operation. This is achieved (approximately) in
powell_1_0704.pdf using a 5th order modified Butterworth low-pass transmit
filter with a 3dB point at fs/4 . There are practical implications of
specifying such a transmit filter in the 10GBASE-T standard that need to be
considered before we decide to use this ARMA precoder in the standard.
The question is how should we make such technical decisions in this task
force? Personally, I believe that the task force should first choose a
complete proposal, and then optionally make perturbations on certain aspects
of that proposal while ensuring that these perturbations don't cause any
technical damage to the rest of the 10GBASE-T system in that proposal.
>From: Hugh Barrass <hbarrass@CISCO.COM>
>Reply-To: "IEEE P802.3an" <STDS-802-3-10GBT@listserv.ieee.org>
>Subject: [10GBT] Discussions on precoder decisions
>Date: Tue, 14 Sep 2004 17:58:07 -0700
>In amongst the various topics that have been discussed on the reflector,
>I haven't seen any traffic on the subject of the precoder decisions. Is
>this because we have general agreement; because the differences of
>opinion are irreconcilable or because no one has given it much thought
>There are a number of questions that we need to answer:
>1. Do we need a fixed or variable (or multiple fixed) precoder?
>Gottfried has said that we need only one or possibly a small number of
>2. If we need more than one or a variable precoder then how do we define
>that the receiver should choose the parameters to send (or how should it
>choose the precoder to select)? How will that data be passed to the
>transmitter. How will the parameters be defined?
>Jose/Seki said that a 20 coefficient, 12 bit precision precoder is
>required. They suggested that the HDSL2 approach could be used. The
>coefficients would be determined at startup & then fixed.
>Scott has said that we could use an ARMA(3,3) definition to fix 3 poles
>& 3 zeros.
>3. If we have a variable precoder (or even distinct fixed precoders), do
>we need to adapt or change the precoder during operation?
>Sailesh has said that we need a 32 tap adaptive precoder.
>Have we got any further to resolving these issues? Is anyone planning on
>preparing a more detailed and widely supported presentation for Ottawa?
>I would be willing to host a discussion before the presentation deadline
>if anyone thinks that would help.
Express yourself instantly with MSN Messenger! Download today - it's FREE!