|
Attendees
- Nick Weiner
- Tom Lindsay
- John Jaeger
- Pavel Zivny
- Norm Swenson
- Lew Aronson
- Sudeep Bhoja
- Ali Ghiasi
- Paul Voois
- Abhijit Shanbhag
- Others who did not ID?
Agenda discussed and approved
Previous minutes/notes approved
Technical progress
- Objective - resolve as many details as possible by/before comment deadline
Jan 7th, 05
- Waveform capture method
- Trigger block & timing of pattern key
- Reviewed presentation uploaded by Pavel before previous call.
- Use same basic clock recovery bit-wise trigger method as in 802.3ae.
- Divide output of clock recovery by pattern length (of BnBi or AnAi) to
create pattern rate trigger.
- Easily done in HW. May not be available in all scopes now. One
near-term method would be to run recovered clock into pattern generator
with any pattern of the same length and use it's pattern trigger output.
Other options also discussed. Integrated methods built into more scopes by
2006.
- Use SW to scan input waveform to find key to align trigger with
specific subpattern. Pavel says this is straightforward.
- ACTION - Pavel to write detailed comment to clause 68.6.5.2
and submit by 1/7.
- Noise and jitter control
- Other options proposed, but rely on mask to capture non-Gaussian
events.
- Therefore, use only rms jitter method.
- ACTIONS - Pavel to write detailed comment to clause 68.6.5 and
submit by 1/7. Tom to send Pavel part of earlier comment (now done) and
support Pavel on this.
- Pattern definition
- John Ewen uploaded spreadsheet of patterns mentioned in previous
minutes.
- Explained during call; some discussion on what most important criteria
are.
- ACTIONS
- Choose top 5 patterns with widest variety of 9 bit blocks
and white power spectral density. Keep final disparity values below
~1%.
- Running disparity and max run lengths may be additional
criteria, but no clear conclusions on call.
- ClariPhy (and others?) to run selected patterns (per above,
plus compare to PRBSn) through waveshapers and penalty
simulator.
- EDC simulation tool
- Norm presented summary of basic approach and assumptions. General
agreement on most items.
- Change 6 to 6.5 dB (but subject to final dispersion budget).
- Concern about double filtering with Bessel Thomson + Butterworth. Some
preferences stated to filter only noise with Bessel Thomson (signal
already filtered during acquisition).
- ClariPhy may test both ways to determine difference.
- Some steps need more details (such as steps 3, 6, and 7).
- ACTIONS/PLAN - (Preparation for ballot and next
draft)
- Refine description, seek/develop consensus at conceptual
level.
- Corroborate result against PIE-D with common linear pulse
shapes (including 47.1 psec Gaussian).
- Final form TBD (pseudo-code?), but ClariPhy to extract MATLAB code
from internal model and distribute as soon as capable. May be rough at
first, but group can refine to follow conceptual level
work.
- Submit comment with extent of what is available to submit by
1/7.
- Prepare presentation (and motion?) for Jan meeting to
capture evolution developed in TP2 con-call forum since
comment.
-
Waveforms
- Reviewed Finisar waveforms (and 1 from Intel but w/o eye) run through
ClariPhy penalty tool.
- Note, fiber models were Cambridge 2.1, 3 offsets, w/o
connectors.
- Results show slow edges increase penalties, but overshoot may have a
helpful effect. (This is observed previously).
- Future plots should add 47.1 psec Gaussian for baselining.
- ACTIONS - More of this type of work, with penalty tool
(above) is required to understand TP2 power budget and test limit
requirements.
- Other topics (not discussed this call)
Next call
Good call today. Thanks for the contributions.
All - please plan to report progress and proposals on
actions on the 1/6 call. Please let me know in advance of any agenda
items.
----- Original Message -----
Sent: Wednesday, December 22, 2004 6:05
PM
Subject: [10GMMF] Reminder for TP2 call
12/23/04
All - here are details for 12/23 TP2 call.
|