Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [10GBT-Modeling] RE: [10GBT-Cabling] [10GBASE-T] a channel capacityestim...





George,

I did not mention or emphasize any optimal DFE modeling in my message that
you referred.  I only agreed to work on a program that can help us to
evaluate the situation.  Before that I did not talk about anything of DFE
modeling at all.  The first thing I sent out for people's evaluation is the
program that calculates the channel capacity.   I have emphasized many
times in some following emails that if the channel capacity calculation
tells us that this is not feasible, then the optimal DFE modeling won't do
anything better.  The channel capacity metric can be somehow useless if
most people in this group disagree with me.  I believe that I have tried to
put away my personal opinion in the discussion as much as possible.  The
channel capacity metric is not for use of implementation, but of
feasibility only.  In that sense, I think it is "flawless" if we use the
group-accepted, scaled, selected, measured channel data and group-accepted
achievable noise/interference cancellation levels with the program.

Then the problem becomes that how much noise/interference cancellation can
be achieved in implementation.  If we can cancel all noise/interference
completely, we can do anything we want.   But the truth is that we cannot.
We need to reach a consensus on how much we can achieve.  I think this will
be a topic in the Dallas meeting.

Regards,

Xiaopeng





"George Zimmerman" <gzimmerman@solarflare.com> on 02/26/2003 12:01:35 PM

To:    <xichen@marvell.com>, <stds-802-3-10GBT-Modeling@ieee.org>,
       <stds-802-3-10GBT@ieee.org>
cc:

Subject:    RE: [10GBT-Modeling] RE: [10GBT-Cabling] [10GBASE-T] a channel
       capacity estim...


Xiaopeng -

Perhaps you misunderstood me on using optimal DFE or time-domain
simulations.  As models of implementations, those are worse than
capacity calculations.  HOWEVER, capacity is sensitive to how much noise
is left over after cancellation.  A main issue was raised as to
implementation losses & how much cancellation was feasible.  DFE & time
domain simulations are a proposal to address the concerns over
implementation losses only.


It is unfortunate that you have now taken a "flawless" model for your
metric.  I can only interpret that as a complete design for the system,
including finite word length & circuit noise effects.  This is WELL
beyond what has been done in other standards.


To refresh your memory:
---begin snip ----

-----Original Message-----
From: xichen@marvell.com [mailto:xichen@marvell.com]
Sent: Wednesday, January 15, 2003 9:18 AM
To: William Jones
Cc: stds-802-3-10gbt@ieee.org
Subject: RE: [10GBASE-T] The channel model



Bill,

Sure.  Once the common channel model (I only need the frequency domain
model) is available, I can work on it.

Regards
Xiaopeng

To refresh your memory:

"William Jones" <wjones@solarflare.com> on 01/15/2003 08:56:07 AM

To:    <xichen@marvell.com>, <stds-802-3-10gbt@ieee.org>
cc:

Subject:    RE: [10GBASE-T] The channel model


Xiaopeng

Would you be willing to sign up for technical feasibility based on an
optimal DFE calculation?

regards
Bill

-----Original Message-----
From: xichen@marvell.com [mailto:xichen@marvell.com]
Sent: Tuesday, January 14, 2003 3:01 PM
To: stds-802-3-10gbt@ieee.org
Cc: William Jones
Subject: RE: [10GBASE-T] The channel model



Bill,

I has no problem to use TIA or ISO based channel and noise model once it
is
standardized.  My point is that even using the frequency-domain model we
can still give a quite good estimation of the practically achievable
distance at a throughput of 10Gbps over CAT-5E or CAT-6 cables.

Regards,

Xiaopeng
------ end snip -----




George Zimmerman
gzimmerman@solarflare.com
tel: (949) 581-6830 ext. 2500
cell: (310) 920-3860

-----Original Message-----
From: xichen@marvell.com [mailto:xichen@marvell.com]
Sent: Wednesday, February 26, 2003 11:30 AM
To: stds-802-3-10GBT-Modeling@ieee.org; stds-802-3-10GBT@ieee.org
Subject: RE: [10GBT-Modeling] RE: [10GBT-Cabling] [10GBASE-T] a channel
capacity estim...



Hi, all,

I may be short-memory, so I cannot remember where and when I have
proposed
to this group a so-called optimal DFE modeling.  If somebody can recall
that for me, your help will be highly appreciated.

Any success in either 1000BASE-T, DSL, RF&wireless or any other standard
does not guarantee our success.  To my knowledge (I may be wrong since
there are way too many things that I do not know and I am ready to
learn),
if you give a channel capacity estimation by the way I am using, you
will
find out that all the above systems have quite a lot SNR margin given
what
their specifications state.  And I don't think that the channel capacity
metric is unfortunately meaningless and "obstructive" in our case.

If I am not wrong, the channel capacity estimation is relatively simple
AND
can give us a quite clear view of what the BEST thing we can do with the
channel if the complexity is not a concern.  If Shannon tells us that
the
channels that we are discussing cannot support the data/distance that we
are targeting now, I don't think any time-domain simulation or optimal
DFE
receiver or any other remedy can help us at all.  That is the only thing
that I want to tell the group by using the channel capacity metric.
Let's
not forget that Shannon told us that there is something that we cannot
do
(given we well understand the conditions, e.g., noises, interferences,
etc...) no matter how smart or whoever we are.

If anyone can put forward a proposal that is flawless and can be
accepted
by most of the group, I have not problem to follow.

If somebody tell me something that I may not agree, I will not simply
call
it "obstructive".  I will first study what I heard, express my opinion,
then discuss the problem constructively.  I think that is the way I am
trying to work here in this group.

Regards,

Xiaopeng





"George Zimmerman" <gzimmerman@solarflare.com>@majordomo.ieee.org on
02/26/2003 10:09:05 AM

Sent by:    owner-stds-802-3-10gbt-modeling@majordomo.ieee.org


To:    <stds-802-3-10GBT-Modeling@ieee.org>
cc:

Subject:    RE: [10GBT-Modeling] RE: [10GBT-Cabling] [10GBASE-T] a
channel
       ca   pacity estim...



We have put forward more detailed models.  It is very important that the
group agree on a level of modeling necessary to go forward.  Initially
Xiaopeng proposed optimal DFE modeling, which has been used successfully
in standards in the past.  Then he put forward a simpler metric -
capacity modeling.  Now you are asking for more detail.  We've only been
going from what you have put forward. In November we provided detailed
time-domain simulations, based on implementation-dependent
architectures.  That is fairly implementation specific and goes beyond
what I believe is required.
We must identify which items are near state of the art, and then work
them individually.  Today you will find 1000BASE-T systems routinely
working at 140 meters, although lower bandwidth, and half the PAM
levels, that is still 40% more dB loss, which translates directly to
linearity & cancellation requirements.  DSL systems operate with MUCH
higher cancellation requirements today.  RF & wireless systems, as well
as read-channel devices also ship at higher bandwidth, and similar
levels of noise performance.  These problems are hard, but not
impossible.  All are technically feasible.

Let us agree on a methodology for modeling to get to technical
feasibility.  Clearly following Xiaopeng's lead on the capacity
calculation has lead to a dead end.  It sounds like  the second level,
optimal DFE calculations, may be an appropriate point, and then
enumerate the limited number of effects that are first-level models.
Remember - technical feasibility is not a complete design in the
committee.
If you have an alternate proposal that you would consider acceptable to
move forward it would be appreciated.  Otherwise I can only conclude you
are simply being obstructive.

-george

George Zimmerman
gzimmerman@solarflare.com
tel: (949) 581-6830 ext. 2500
cell: (310) 920-3860

-----Original Message-----
From: xichen@marvell.com [mailto:xichen@marvell.com]
Sent: Wednesday, February 26, 2003 9:51 AM
To: Fakterman, Boris
Cc: stds-802-3-10GBT@ieee.org; stds-802-3-10GBT-Modeling@ieee.org
Subject: RE: [10GBT-Modeling] RE: [10GBT-Cabling] [10GBASE-T] a channel
ca pacity estim...



Boris,

I completely agree with you.  The undergoing discussion is to let
everyone
in this group understand what kind of problems we are facing in setting
up
the new standard.  We are dealing with more than a academic problem, we
are
dealing with a standard that will be tested in the field.  Therefore
carefully counting all the factors in the real world is definitely
necessary for our success.  Before doing something, we must know what we
can do and cannot.  If we simply set up a goal and we actually don't
know
whether we can do it or not.  Then a lot of the efforts in this group
can
be wasted without any return.

So let's wait until the cable data accepted by everybody in this group
are
available.  Then the discussion can restart on a solid ground.

Regards,

Xiaopeng






"Fakterman, Boris" <boris.fakterman@intel.com>@majordomo.ieee.org on
02/26/2003 07:38:54 AM

Sent by:    owner-stds-802-3-10gbt-modeling@majordomo.ieee.org


To:    George Zimmerman <gzimmerman@solarflare.com>,
       stds-802-3-10GBT@ieee.org, stds-802-3-10GBT-Modeling@ieee.org
cc:

Subject:    RE: [10GBT-Modeling] RE: [10GBT-Cabling] [10GBASE-T] a
channel
       ca   pacity estim...



George,

I agree that we need more data to collect. We should be based on the
cables
data accepted by everybody and the cables group does a great work to
prepare
it. I agree that it is a high performance system and the cancellation
ratio
you proposed is the state of the art. This is exactly the reason we
should
carefully consider the impact of the system definitions to the
analog/digital blocks definitions, just not to cross the close border to
impossible. For example the 1000BASE-T standard requires 10mV
transmitter
distortion on PAM5/125MHz system. Do you estimate the 10Gb standard will
require much better performance on PAM10/833MHz system?

Do I clearly understand you statement? Do you say that remaining after
the
DFE three dB margin will be enough for all cancellers?
In any case  it is required to remain at least 3 dB margin after all
cancelled noises were counted. I can't see this from the discussed DFE
analysis.

Boris Fakterman

-----Original Message-----
From: George Zimmerman [mailto:gzimmerman@solarflare.com]
Sent: Wednesday, February 26, 2003 5:08 PM
To: Fakterman, Boris; stds-802-3-10GBT@ieee.org;
stds-802-3-10GBT-Modeling@ieee.org
Subject: RE: [10GBT-Modeling] RE: [10GBT-Cabling] [10GBASE-T] a channel
capacity estim...


Boris -
I agree that as we go forward, more data will come out.  The cabling
group was working this.  I believe that we have now countered the
assertions made based on using limit lines that the problem was
substantially beyond capacity, not that it is easy, but hopefully you
can begin to see that it is possible.  Make no mistake - this is a
high-performance system, not everyone will agree they can make it.
Fortunately technical feasibility means not everyone has to make it.

It is important that we begin to come to a common understanding of the
data required.  This was part of the frustration I had with the
capacity-based calculation, in that I was pretty sure the next step
started to get to the time-domain simulations.  It is fairly common
(e.g., in DSL systems) to build systems working close to performance
limits.  We have presented time domain simulations in November which
address this.  Bill's results that you reference use a higher noise
floor, but, you correctly point out that the cancellation ratios
required push most of the impairments significantly below the noise
floor.  If you look at the November tutorial, you will see that this is
in fact where the design sits.

One thing to point out is that we realized that we left Xiaopeng's
cancellation parameters (and his RL estimate, hence the cancellation
will be his) in the simulation. Go ahead and replace them with the
numbers from the tutorial, you'll get substantially the same results.

I'm afraid that somewhere we're going to have to come to a common
ground.  While we need to go further, the purpose of a technical
feasibility is not to do a complete design, it is also not to make it
look easier than it is, and it is not that everyone agrees they can do
it.  It is only to show that multiple vendors think they can do the job.
As you undoubtedly know, this solution will have to minimize noise from
a number of areas.  At the time, so did 1000 BASE-T, HDSL2, and early
ADSL systems.  Yet, all exist as commercial, successful, multi-vendor
parts today.


George Zimmerman
gzimmerman@solarflare.com
tel: (949) 581-6830 ext. 2500
cell: (310) 920-3860

-----Original Message-----
From: Fakterman, Boris [mailto:boris.fakterman@intel.com]
Sent: Wednesday, February 26, 2003 3:48 AM
To: George Zimmerman; stds-802-3-10GBT@ieee.org;
stds-802-3-10GBT-Modeling@ieee.org
Subject: RE: [10GBT-Modeling] RE: [10GBT-Cabling] [10GBASE-T] a channel
capacity estim...

George, All

The primary purpose of the ongoing discussion was to decide if the PAM10
design could exist on 100m CAT5 cable. Following the discussion I don't
feel
on the solid ground as tens dBs in SNR grow and fall mostly in
cancellation
ratio parameters. The Alien NEXT level and possible cancellation are not
based with enough data.
The implications of the Echo, NEXT, FEXT cancellation ratio presented by
Solarflare also are not clear to me. The cancellation ratio will be
impacted
by  coefficients resolution in digital domain, by jitter and other
impairments in analog domain. Does the proposed cancellation ratio
demand
reasonably achievable analog and digital parameters?

Meanwhile to promote the primary purpose I would like to refer to the
document distributed by William Jones few weeks ago (attached).
If I understand correctly it describes the SNR after the equalizer on
100m
CAT5 with ground noise only. The SNR for -140dBm/Hz ground noise (no
Echo,
NEXT, FEXT) is roughly 28dB. Assuming coded signal SNR for BER 10^-10 as
25dB, it remains only 3 dB margin for Echo,NEXT,FEXT, Alien NEXT and
implementation impairments.
Again if I understand correctly the graph, it seems that there will be
negative margin considering all noises exist.

Regards,

Boris Fakterman - Intel Communications Group, Israel
Tel: 972-4-865-6470, Fax: 972-4-865-5999
 mailto:boris.fakterman@intel.com