[802.3ae_Serial] From serial PMD call 17 Apr: need more experimental evidence
Participants to the call feel that the experimental verification of some
aspects is still insubstantial. See for example Raj's comment "Allow time
for at least 3 companies to submit measured data confirming the test
parameters and set up." Many vendors have reported transceivers, how come
so little data?
As I understand it the concern is whether parts will pass a test in one lab
and fail in another, as opposed to a concern that the parts won't work or
won't interoperate with another manufacturer's parts.
The key items, serious first, are stressed eye implementation, TDP
measurement, transversal filter implementation. If these concerns can be
allayed by more experimental evidence, we have a good chance of a clean
balloting round on D4.3. As the next meeting will be on 30 April (or 29)
(in the Bay area), we have this week and next week!
Participants are asked to look at the experimental information they have and
see if sharing some of it will progress the standard and the market without
competitive disadvantage by revealing too much. For transceiver vendors,
aspects of test technique could be revealed. For more product-related
information, Greg LeCheminant again offers to provide a confidential
"information broker" service for creating an anonymous combined report.
On stressed eye: Tom reports that in theory, the filter in the stressed eye
generator may not create enough jitter, and the quantity of interferer
and/or sinusoidal jitter would need to be increased for a generator with an
ideal filter response. Others report that in practice they have more than
enough jitter already and no action is needed. How can we close this
disconnect between theory and reality? Need a way of simulating more
realistic filters.
Petar thinks ISI from filter and ISI from interferer are reasonably
interchangeable within limits, as a stress-test.
There is a small question of whether the jitter metric is from 1st to 99th
percentile, 0.5th to 99.5th percentile, or from mean to each side, excluding
1% or 0.5%.
On transversal filter: the latest information from Picosecond Pulse Labs may
allay the concerns.
No other new issues were raised.
Present
-------
Piers Dawe Agilent SPG
Petar Pepeljugoski IBM
Greg LeCheminant Agilent T&M
Tom Lindsay Stratos
Raj Savara, Mike Stout Network Elements
Peter Öhlén Optillion
Piers