[10GMMF] Notes from Aug 17 meeting on TP3 definition
1. Attendees
Lew Aronson
Chet Babla
Venu Balasubramonian
Piers Dawe
Jens Fiedler
Ali Ghiasi
Jesper Hanberg
Jonathan King
Ryan Latchman
Tom Lindsay
Martin Lobel
Jan Peeters Weem
Petre Popescu
Yu Sun
Norm Swenson
Lars Thon
Kevin Witt
2. Normative tests, what to test for, philosophy
Lew asked what was decided last week. Tom said that rather than pursue the "philosophy" we would work to establish the magnitude of the effects then consider again how to test for them. We have not made a decision about separation of dynamic test, or not. Those on the call were agreeable to this position.
3. Designing the ISI generator for receiver testing
Petre had sent a slide pack to the reflector and we went through this. It was well received and thought to be headed in the right direction. Slide 3 shows a parameter extraction method where the parameters A1,A2,..., deltaT of the synthetic response candidate for a stressed eye generator are parameter-fitted to fibers (or groups of fibers), chosen to have appropriate difficulty, from the 65 fiber model set. Because different response shapes will exercise actual equalisers in different ways, more than one "synthetic" stressor vector (A1,A2,..., deltaT) will likely be needed; then we would group the modelled channel responses into groups such as precursor-heavy, neutral, postcursor-heavy. Two of these would be mirror images of each other. There was discussion about the metric of goodness of fit between synthetic response and modelled responses. Also questions such as would we average the modelled responses then fit, or fit each then average. When we have the right categ!
ories, it may not make a significant difference. We should calculate the several metrics including a measure of pulse length for the synthetic responses as well as the modelled ones; they should agree. The actual level of metrics e.g. PIE-L need to be correlated to the percentile of the installed base, and will evolve as our understanding of the latter evolves. They are lower than Petre's example, which is based on 300 m and we should base the work on the agreed 220 m (for FDDI grade). Individual companies will have their own constrained equaliser metrics and they should check consistency using those. Also, the committee can consider results with "public domain" constrained equaliser(s) as several presentations have done already.
Should there be a filter such as Bessel-Thomson in addition to the transversal filter? Would we reach the same synthetic responses if we did the calculations based on impulse response rather than 100 ps pulse response?
For the simple signal generator (Bessel-Thomson filter only) the relevant bandwidth to start the search is not 500 MHz.km (which is appropriate for overfilled launch) but a higher value appropriate to a controlled launch. The diagram on p4 does not show a second 7.5 GHz B-T filter because it would make no noticeable difference.
4. Electrical-to-optical conversion for stressed eye generator, and its linearity
Nothing to report this week, will be on the agenda for next week.
Please let me know if you have any comments or corrections.
Best Regards
Piers