|
All -
Here are 2 thoughts that may be of interest for TP3.
1. For jitter, the current proposal is a single frequency well above the
cutoff (supposedly in the 4 MHz range) above the tested CDR's tracking window.
The single frequency has been proposed to be somewhere in the 40-80 MHz
range.
My only real concern is that a frequency in this range does essentially
nothing to ensure tracking of low frequencies. Recall that all outputs are to be
tested with a 4 MHz high-pass jitter filter. This requirement runs throughout
802.3ae and XFI. This allows, and indeed was originally motivated by, outputs
with significant low frequency jitter.
So, I propose another frequency that assures tracking. I propose 5 UI at 40
kHz. These values are per 802.3ae. I would support a note that suggests such a
test could be done at the module or even IC level.
2. For RIN and modal noise, it is not completely clear to (at least) me,
what the spectrum and probability density functions are, but we can expect they
are not single frequency sinusoidal.
Single frequency sinusoidal has been proposed for its simplicity and
history with 802.3ae. These are good reasons. Another fairly simple option to
consider would be to replace the sinusoidal term with a PRBS pattern (rate and
length TBD, if there is any interest). The PRBS pattern would have a wide(r)
spectrum. Unfiltered, this would produce a bi-modal density function, much like
the sinusoidal signal. Filtered, either at the low or high end can affect not
only the spectrum, but also the density function. For example, filtering the
upper end to 1/20 of its rate results a rather "normal" (but truncated)
appearing pdf.
Is there any interest in pursuing this further?
Tom Lindsay ClariPhy Communications tlindsay@ieee.orgphone: (425)
775-7013 cell: (206) 790-3240
----- Original Message -----
Sent: Friday, September 03, 2004 8:55
AM
Subject: [10GMMF] Agenda TP3 Meeting
Tuesday 7th September
Hi Folks,
Here is the Agenda for Tuesday's
Meeting
I am travelling, so Piers has kindly agreed to step in and
manage the agenda.
Best Regards
Mike
Dial in
Details
===============
Meeting Time: Tuesday Sept 7th. 9am CA, 12 NY,
5pm UK, 6pm Germany
International Direct Dial (650) 599-0374
Meeting
ID:
801803
Agenda
======
1. Attendee list
2. Review Notes from
Last Meeting (Aug 31)
3. Update on Experimental Results Regarding
Non-Idealities in E-O-E (Abhijit)
4. Discussion on Parameter Fitting for
the ISI Generator (Petre)
Petre's on-going work is to do the
following:-
o show PIE-L for the
Cambridge vs the approximated channels (validate that the channel metric is
not significanlty altered)
o
present thinking on how many impulses are requiered and which should be used
for
testing
- early thinking from Petre is suggesting that 3 different responses make
sense:-
a. quasi
symmetric
b.
post-cursor
c. pre-cursor
o present thinking
on the relative merits of BT filters, 2 and 3 peak impulse response
models
5. Review work list of addtional items
Review progress on the
the following stressors:-
a.
Fiber loss/dispersion
b.
Connector loss/dispersion
c.
Time varying factors
d. Modal
Noise
e.
RIN
f.
Jitter
g. Optical signal
conditioning
a., b. and c. could potentially be covered through a
combination of the adhoc and Petre's work - is this true?
For d. and e.
Lew is going to talk with Infineon folks regarding some analysis on noise
powers to represent modal noise and RIN. Abhijit is going to do some thinking
on an EDC chips performance with respect to different noise sources and their
nature. In particular is it acceptable to emulate the noise source with a
single sinusoid?
For f. Ali Ghiasi and Tom Lindsay are both going to
review Lew's proposal of testing with jitter applied at a single frequency and
propose if this should be changed and what frequencies should be
tested.
No discussion on g.
6. Discuss goals for what we can
achieve by the September Interim meeting.
|