From serial PMD call, 3 April
We built a new open issue list; in no particular order:
Trade off jitter and risetime
-----------------------------
Proposal to rely more on eye mask, allowing trade off between risetime and
deterministic jitter. This would simplify testing at 1310 serial and maybe
1550 serial, avoiding an unsatisfactory rise time test involving four
histograms. It may allow a good old fashioned performance improvement at
1550 where faster rise times can be associated with greater dispersion
penalty. There may be technical issues between this and the
wavelength/spectral width/power trade-off, and a highly multi-dimensional
trade-off would become unmanageable.
Specification of optical transmitter spectral width
---------------------------------------------------
Relation between full width, -20 dB criterion and link model.
Double counting of receiver eye opening penalty
-----------------------------------------------
Spreadsheet allows an unretimed receiver sensitivity and then an eye opening
penalty (~0.4 dB). Measured sensitivity of a retimed receiver would include
its own eye opening penalty. Need to at least note the issue in the
standard.
Other issues raised by the shift to a box level spec
----------------------------------------------------
Perhaps without noticing it, we have moved from a component level standard
(Gigabit Ethernet) to a box level standard (10 Gigabit Ethernet). We need
to check that we have made the transition correctly. One known issue is
that of some components which try to appropriate the whole box's allowance
to themselves (e.g. jitter).
Interworking between 1310 and 1550 equipment
--------------------------------------------
It may be that some 1310 and 1550 equipment could interwork, within certain
loss and dispersion parameters, "by coincidence". Should this get mentioned
or encouraged in the standard? Does it have "broad market potential"?
Bit ordering in clause 51
-------------------------
Alignment with SFI-4 vs. Ethernet customs (see reflector traffic). Needs
joint session of clauses 51, 52. Also involves clauses 49, 50.
Meaning of signal detect
------------------------
Needs joint session of clauses 51, 52 and more.
Measurement filter for RIN
--------------------------
Serial ad hoc to find if the filter for RIN should/must be the G.691 (7.5
GHz) one, i.e. what is built into test equipment?
Delay through PMA/PMD/medium
----------------------------
This delay maximum is defined in clause 44. We need to discuss it in the
PMD tracks and add
cross-references in clauses 51, 52 and 54.
Ordering of text in new jitter section
--------------------------------------
Other tests have test method and required values in separate sections:
shouldn't this one too?
Do we prefer dB(OMA) or dB(OMA/2)?
----------------------------------
Argument for dB(OMA): More straightforward relation between OMA
and dB(OMA)
Argument for dB(OMA/2): Allows traditional thinking with dB(OMA/2) ~
"optical power"
We just need to take a poll on this: there weren't enough people on the call
this week.
(My own view is that if Fibre Channel had said "Modulated Optical Power =
(P(1)-P(0))/2" then we wouldn't be having any discussion.)
Topics that don't need to be in weekly call
-------------------------------------------
Adding interferometric noise to link model
Updating link model to reflect [what change?] at 1550 nm
Possible effect of sinusoidal jitter on link model
Test patterns - separate call for these
NEXT MEETING
------------
Agenda
------
I thought we might put on the agenda some items which don't need so much
research:
Delay maximum
RIN measurement filter
Interworking 1310 <>1550
Other ramifications of shift to a box level spec?
Transmitter spectral width
What values of RMS width and full width, -20 dB down, are
anticipated? What dispersion penalties are anticipated in the 1310 band?
Any other issues to add?
Time and place
--------------
At the usual phone number and new usual time:
15:15 GMT = 4:15 pm BST = 8:15 am PDT, Tuesday 10 April 2001, for an
hour and a half -ish, to finish by 17:00 GMT
+1(816)650-0631 Access code 39209
People are having difficulty attaching to these calls; apparently the issue
is congestion not number of ports, so persistence is required. I shall try
to open the call five minutes before time to give people a better chance.
Piers