----- Original Message -----
Sent: Wednesday, October 20, 2004 6:12
AM
Subject: [10GMMF] TP3 meeting notes Oct
19
Dear TP3'ers,
Here are my notes from yesterday's meeting. Any
comments or corrections please let me know.
Next
Meeting:-
Tuesday October 26th at 9am SJ, 5pm UK, 6pm Germany
Dial
in (650) 599-0374, Meeting ID: 136169
1. List
Attendees
Aronson, Lew
Dawe, Piers
Ewen, John
Hanberg,
Jesper
Latchman, Ryan
Lawton, Mike
Lindsay, Tom
McVey,
Jim
Popescu, Petre
Shanbhag, Abhijit
Telang, Vivek
Van Schyndel,
Andre
Weiner, Nick
Witt, Kevin
2. Review meeting notes from last
week
No comments
3. Review Lew's additional motions (which were not
heard at Ottawa). These were forwarded by Lew last week.
Lew described the
motions he did not get to present in Ottawa
#1
This one presented
more details around the parameters used in the stressed Rx sensitivity test,
namely:-
i) set OMA
max (shown as -6.6dBm but due to change as dynamica adapation penalty
was not voted out)
ii)
Sinusoidal Jitter freq and amplitude (40MHz,
0.1UI)
iii) remove reference to
sinusoidal noise term
iv)
add reference to OSNR figure (TBD) to represent noise powers
i), ii)
and iv) did not get much discussion. I assume that is because there is broad
agreement on these.
TAL - did you mean iii) instead of ii)?
MCL Yes. Apologes my error.
For
ii) there was a discussion about how jitter is represented in our stressed
test. Clearly jitter contributions come from several different sources (Tx
clocks, tx circuitry, channel impairements ...) The group discussed which are
the significant sources for jitter and hence how it should be represented in
our test. Our repesentation of the channel ISI will create some jitter but we
need to understand if this is sufficient or if there are other significant
sources of jitter (with potentially different characteristics) which need to
be represented in the test.
ACTION: Petre kindly agreed to evaluate the
jitter testing required and consider the need for a high frequency sinusoidal
interferer.
Tom has already helped here with an email to the
reflector.
#2
Adopt methodology
in line with 52.9.5 for OMA measurement
In
particular:-
i) for
TP2 use the 4 "1"'s 4 "0"'s pattern defined within
802.3ae
ii) for TP3 use
the same technique with a 10 "1"'s and 10 "0"'s
pattern
iii) Calibrate OSNR with
the same singal as in ii)
Petre asked if there was value in developing
a technique which could be supported more easily in the field. The current
approach is assuming that this is not a requirement.
Piers raised the
accuracy and the usability of the 1's and 0's method vs a histogram approach.
Adding that he sees subtle differences between a 4 "1"s and "0"s vs a 10. He
asked that we allow him to progress this work further.
The group gave
its support to this motion but is happy to listen to recommendations to
improve the accuracy/usability of the OMA measurement.
TAL - this reminds of earlier discussions
regarding the dynamic penalty. Here, we want the best balance of an
accurate test (one that links to the budget and interoperability) and one that
is that is easy and versatile. If we can get both, then we've won. Otherwise,
I tend towards accuracy, especially when the budget is tight - we don't have a
lot of room for slop and sandbagging.
Although I favor Lew's approach, this comment should
not be interpreted as reading against Piers's work on this. Hopefully he can
come up with something that indeed does it all.
I recall that we do all generally agree on the
conceptual definition of OMA that we're using. That is important for
budgeting, analysis, etc. Testing can stray a bit away from that as long as we
have an accurate way to associate back to the budget.
#3
This one is not directed at the
standard directly but more attempts to provide a framework for advancing the
ISI portion of the test.
Lew offered the
following:-
i) we
should pursue using a set of 3 different chanels for testing (pre, post curson
and quasi symmetric)
ii)
we should define a "goodness of fit" metric (PSR) for a 5 impulse peaks model,
each with uniform time spacing
iii) Not use PIE-L but focus on
PIE-D
iv) Assume Tx
rise/fall times of not less than 4 7 ps
i) was not
heavily discussed.
ii) was discussed. A counter view was expressed that
maybe we could describe minimum tap weightings and minimum PIE-D's for an
exact channel and let the implementer decide his exact channel
implementation
iii) the debate has moved on since the motion was presented.
There were some concerns presented around PIE-D as a metric arguing it has
limitations in terms of representing a finite length equaliser. A metric
relating to finite equalisers would be valuable.
TAL - we did not get to it, but this
last thought was added also to my revised TP2-TP3-budget presentation. It
came up again on the TP2 call today, and there seems to be some growing
support for it.
iv) some confusion around rise/fall times. Lew wrote
this to be relax the rise/fall time burden on the equipment used for the test.
That said it may end up being important that we relate this figure to what TP2
says about rise and fall times.
TAL - I will try to respond to this in a
separate email.
4. TP2/TP3 link budget work. Tom Lindsay's
Feedback from TP2 call
Ran out of time before this item
Upcoming
items of focus (for subsequent
metings):-
i)
Jitter requirements
ii) Static
channel
methodology
- noise
loading
- channel types and exact
characteristics
o i.e. do we pick pre, post cursor and quasi symmetric and if so where from?
108 fiber model?
iii)OMA
measurement methodology.
Best
Regards
Mike