[802.3ae_Serial] From serial PMD call 2 Apr: measurement, preview of a few comments
Collecting test results; measurement
------------------------------------
Not very surprisingly, Stretch had not received any test results yet but was
hoping for them very soon: please provide by Thursday 4 April if possible,
format not critical. A couple of groups had not done as much experimental
stressed eye generation as they had hoped.
I explained that I believe the effective OMA with a sinusoidal interferer
present is the total excursion of the waveform; if everything were linear,
this would be OMA with interferer off, plus pk-pk amplitude of interferer.
In other words, the interferer raises the OMA and also raises the VECP by
more than D4.1 says, giving the same inner eye opening as is really there.
There was a discussion of the effect of the interferer. Petar has found
that ISI from interferer and ISI from filter can be substituted for each
other in the range 50-90% from filter. Meanwhile my colleague has found
unexpectedly large effects from the interferer. I thought that this would
apply with typical receiver bandwidths.
Stretch found that a filter which generated 1.3 dB VECP with scope filter
on, measured 0.6 dB with it off - with PRBS7. Others thought low frequency
effects with longer patterns could affect these figures.
Tom would like 52.9.11.3 to specify measurement of the reference sensitivity
S in the same way as the DUT sensitivity e.g. +/- 0.05 UI. He also would
like the definition of "eye center" there clarified - it does not mean
"optimum" or "wherever suits you". I think he was going to propose a time
midway between the 10^-3 points, which is current good practice.
Specs and other comments
-----------------------
Mike thought the TDP for 10GBASE-S should be 4.3 dB. I wondered if it had
to be 3.6 dB because that is our limit for ISI and the TDP measurement
doesn't define the cause of the penalty. We went away to think about it.
I pointed out that in our RIN test (and in our definition of reflection for
TDP) we had forgotten about reflection noise (10GBASE-L only - we sorted
this out maybe a year ago). Petar mentioned a recent paper on the subject
in January Photonics Technology Letters.
I would be making some other comments as follows:
Muddling signal detect and receive fault is unhelpful for diagnostics;
Need to have minimum average power limits;
Can now have minimum OMA for 10GBASE-E as we have for 10GBASE-L;
Very wide variability of sinusoidal jitter in Rx test is bad.
With a TDP measurement do we need and SMSR spec?
There was a short discussion of the practical difficulties of implementing
the transversal filter.
In summary: no significant bad news, some old issues remain, in particular,
still need to sort out the stressed receiver conformance test. It's still
worth trying to wrap this up next week in Vancouver.
Present
-------
Piers Dawe Agilent SPG
Petar Pepeljugoski IBM
Stretch Camnitz Agilent T&M
Mike Dudek Cielo
Tom Lindsay Stratos
David Kabal Picolight
Adam Healey Agere
Mike Stout Network Elements
Piers