All - here are my notes for the 10/21 call. Below is dial-in info for
tomorrow's call at 9 AM.
Attendance:
Albreght Rommel, Acuid
Ali Ghiasi, Broadcom
Vivek Telang, Broadcom
Greg Lecheminant, Agilent T&M
Pavel Zivny, Tektronix
Jan Peeters Weem, Intel
Tom Lindsay, ClariPhy
Norm Swenson, ClariPhy
Paul Kolesar, Systimax
Yu Sun, Optium
Martin Lobel, Intel
Andre Van Schyndel, Bookham
Piers Dawe, Agilent
Others? Please let me know.
Agenda
Approved per below, but Paul asked for a summary of TP3 call
progress.
Discussions
- Report on TP3 work from 10/19. Reviewed proposals (see Aronson email to
reflector 10/05/04) that we did not have time to discuss in Ottawa. General
summary is that there were alternatives discussed and no clear conclusions.
Work will continue. See Lawton minutes from 10/19.
- Paul reported Petar's work on modal noise in task 2 group. Accurate
analysis with realistic fiber and laser parameters. May be able to combine
effects of modal noise, mode partition noise, and RIN into one model.
- Will TP2 penalty test be done at TP2 or at virtual TP3 (TP2')? This method
is similar to 802.3ae's TPD in that is indicates overall transmitter and
dispersion deterministic penalty, but different in that the waveform is
captured at TP2. Virtual TP3 behavior is determined by using SW models.
- Next step for penalty work is to gather numerous waveforms that represent
a range of good to bad transmitters. Waveforms should include effects from
drivers, layout and coupling circuits, lasers, etc. I subsequently posted an
email outlining the initial part of this work (sent (10/26).
- Budgeting & metrics
- General agreement that DFE will have a better chance of success for LRM,
so PIE-D or other DFE-based metrics appear to make the most sense for
budgeting.
- However, PIE-L is still a very useful indicator of channel behavior and
should be used for analysis.
- What goes into the implementation penalty? (refer back to thread end of
June 2004). Concern that PIE-D is too ideal and might require too much
implementation penalty.
- Suggestion to consider fixed length equalizer for budgeting. Ideal
metrics may not accurately capture issues with real equalizers.
- Laser non-idealities - Andre verbally presented a list of laser properties
that may degrade link performance. Purpose of list is to help ensure testing
includes all important effects. Since the call, Andre sent a written list
which I forwarded onto the reflector, with his permission (sent
10/26).
Next call
Any comments or changes, please let me know.
Thanks,
|