----- Original Message -----
Sent: Tuesday, May 11,
2004 4:06 PM
Subject: Re: [10GBT] Symbol
rate
You are right, Jose. I used the 55m
cat 6 case because that was what the presentations were pitching as the best
case for PAM4 – the situation IS even worse for longer lines, or worse
for high frequency ANEXT.
By the way, Figure 7 is the one you want
to look at to see the plots with the crossover of ANEXT & IL.
Jose, I think you meant the 100m models
for Class F IL & Class E IL respectively, models 1 and 3 (2 is the 55m model).
To see the results for Model 1 (Class F
IL, Class E other params, 100m, 60 dB ANEXT at 100MHz), run:
solarsep_varlen7a(-2.5,650,4,100,6,1,7,2)
To see the results for Model 3 (Class E
IL, Class E other params, 100m, 62 dB ANEXT at 100MHz), run:
solarsep_varlen7a(-4.5,650,4,100,6,1,6,2)
-george
-----Original Message-----
From: stds-802-3-10gbt@IEEE.ORG
[mailto:stds-802-3-10gbt@IEEE.ORG] On Behalf
Of Jose Tellado
Sent: Tuesday, May 11, 2004 3:45
PM
To:
STDS-802-3-10GBT@LISTSERV.IEEE.ORG
Subject: Re: [10GBT] Symbol rate
I think the case you considered below
is for 55m of cat6. The problem is even worse for 100m of channel model 1 or
model 2 (see http://www.ieee802.org/3/an/public/mar04/kasturia_2_0304.pdf),
where the SNR "pinching" happens around 500MHz,
Jose
From: stds-802-3-10gbt@IEEE.ORG
[mailto:stds-802-3-10gbt@IEEE.ORG] On Behalf
Of George Zimmerman
Sent: Tuesday, May 11, 2004 3:19
PM
To:
STDS-802-3-10GBT@LISTSERV.IEEE.ORG
Subject: Re: [10GBT] Symbol rate
At the March meeting, some of the
discussion on leaving the door open to the higher baud rates was aksing to have
time to run simulations with the higher baud rates to validate the optimal DFE
results presented.
The MATLAB code has been available on the
802.3an web site for a while now – I assume others have now seen that the
ANEXT model we agreed to (49.5-15log10(f/100)) is within a couple dB of the
insertion loss at 650 MHz – which means that the penalty for signaling at
rates over 500 MHz, where the SNR gets pinched off, is extreme. (to run this
case, use: parameters: solarsep_varlen7a(-10.5,650,4,55,6,1,6,2) ).
What these curves tell you is simply that
at more than 1Gbaud, you are simply signaling faster than the channel can
support, and thus starting out in an SNR hole in building a 10GBASE-T system.
Hopefully at the meeting we can agree to
constrain the baud rates and begin to focus our analysis.
-george
-----Original Message-----
From: stds-802-3-10gbt@IEEE.ORG
[mailto:stds-802-3-10gbt@IEEE.ORG] On Behalf
Of Booth, Bradley
Sent: Tuesday, May 11, 2004 10:36
AM
To:
STDS-802-3-10GBT@LISTSERV.IEEE.ORG
Subject: Re: [10GBT] Symbol rate
The objectives do require that we provide
one PHY that meets at least 100 meters on Class F and at least 55 to 100
meters on Class E. A proposed PHY must meet both these distance
requirements. To take this a step further, a PHY or port type has
traditionally referred to one PCS, one PMA and one PMD. So this could be
seen as one PHY which has only one PCS, one PMA and one PMD is required to meet
both the distance objectives.
In the past, the Working Group has asked
the Task Forces to make the tough decisions and to choose only one PCS, PMA and
PMD to meet the objective. If the Task Force chooses PAM5, then the
decision is made. If the Task Force chooses PAM10, then the decision is
made. If the Task Force chooses PAM5 and PAM10, then the Working Group
will likely send the specification back to us to make a decision.
To quote the movie Highlander, "There
can be only one."
-----Original Message-----
From: stds-802-3-10gbt@IEEE.ORG
[mailto:stds-802-3-10gbt@IEEE.ORG] On Behalf
Of Sanjay Kasturia
Sent: Tuesday, May 11, 2004 12:12
PM
To:
STDS-802-3-10GBT@LISTSERV.IEEE.ORG
Subject: [10GBT] Symbol rate
At the March meeting, there was a motion to bound the range
of symbol rates. The motion, moved by George Zimmerman, suggested a
symbol rate that would range from 714Msym/sec per pair to 1000Msym/sec per
pair.
This motion failed to get the requisite 75% yes vote. Some of
the people who voted against this proposal were in favor of schemes
that would require higher symbol rates - e.g. 1250Msym/sec per pair but
were probably not very familiar with 802.3 operation. With 802.3 voters in
the room, the motion would have passed. See the vote tally appended below from
the meeting minutes.
As I understand it, the PAM 4 type schemes that would use the
much higher symbol rate would NOT meet our distance objectives but offered
some value in that they could enable much lower power transceivers for shorter
distances than called for in our objectives.
Should these schemes, which do not meet our distance
objective, but could still be valuable for customers who want shorter reach and
lower power be considered in separate class - possibly in a different forum
than 802.3an?
Can our chair, Brad Booth, give us his opinion on this?
Vote count from minutes of March meeting
TF Voters Y: 24 N: 15 A: 19
802.3 Voters Y: 21 N: 5 A: 9
Sanjay Kasturia
Editor-in-chief
sanjay@teranetics.com
cell (650) 704-7686
office (408) 653-2235
Teranetics Inc.
2953 Bunker Hill Lane, Suite 204
Santa Clara, CA 95054