[802.3ae_Serial] From serial PMD call 29 Jan: test fest, DDJ and DCD, simulate stressed eye
Present
-------
Piers Dawe Agilent SPG
Raj Savara Network Elements
Juergen Rahn Lucent
Greg LeCheminant Agilent T&M
Stretch Camnitz Agilent T&M
Peter Ohlen Optillion
Petar Pepeljugoski IBM
Gair Brown US navy
John Ewen JDSU
February Test Fest
------------------
See two emails from Greg to serial reflector sent 24-26 January; one about
the 802.3ae interim, including recommended hotel, the other about the test
fest. Our hosts can offer private slots of maybe 2-3 hours each in the lab
where there will be basically one set of equipment, i.e. it is not likely to
support two experiments at the same time. The apparatus will include:
A transmitter with adjustable stress
A receiver made from off-the-shelf parts
A receiving instrument?
Prospective participants are requested to contact Greg quickly; he will sort
out a timetable. Three of those on the call expected to participate, with
300 pin, tentative serial and tentative XENPAK format DUTs between them; all
at 1310 nm. Test apparatus should be available for 1310 and 1550 nm.
DDJ and DCD
-----------
Juergen reported that some 1550 nm EML transmitters have "DCD" (that is, the
average rising and falling edge timing differ) which is most pronounced at
high extinction ratio. I think he saw this as an acceptable feature not a
problem. Raj pointed out that to drive costs down, CMOS ICs and non-clocked
drivers will be under consideration, and (other) pulse length variation will
be seen (in commercially acceptable grades of transmitter).
Stretch reported: a filter around 3.5 GHz gave 8-10 ps eye closure "on
average" but the average lone bit was 5 ps shorter than that, and the
shortest 1 or 0 in the pattern (PRBS7? 15?) was 23 ps short (72 and 67 ps).
Would transmitters show the same behaviour? If so, this distribution of
pulse lengths would likely be quite different to that from a pattern with an
interfering sine wave.
There was a discussion between Juergen and Raj about using the evidence of
~1 dB penalty found for 0.2 UI SJ to help dimension the jitter in the
stressed eye.
Raj suggested a clock modulation scheme to create pulse length variation.
We didn't fully understand it so he was asked to clarify the scheme to
Stretch.
Petar wanted an agreed definition of DCD. As he has requested before, we
should see if our terminology is in alignment with the dictionary
definition. For more on this see my notes from the meeting of 11 December.
We should consult the official IEEE dictionary (IEEE 100, The Authoritative
Dictionary of IEEE Standards Terms) which is not on line
(http://standards.ieee.org/faqs/Std100.html ). Dear reader, if you have
this
book, is Duty Cycle Distortion defined there?
For reference I have pasted in the two comments in the database which relate
to this. We made no further progress on the call.
# 294 Cl 52 SC P L
The terminology for DCD, as used in Rx testing, is not clear and appears to
contradict traditionally accepted definitions. I believe the intent is to
use a term that specifies the horizontal eye closure due to ISI-induced DDJ.
However, this is different (less harsh) than DCD caused or developed by
threshold offsets, mismatched rise/fall times, etc.
SuggestedRemedy
Either use DCD as traditionally defined, or eliminate its usage from this
standard. We obviously still need to define what the Rx test comprises, but
it appears we are leaning towards using ISI-induced DDJ. We need to define
the appropriate test setup to be sure the ISI-induced DDJ is of the desired
form.
This task is and should be part of the ongoing effort of the serial PMD ad
hoc on testing.
Lindsay, Tom Stratos Lightwave
# 49 Cl 52 SC 52.9.11 P 4629 L
We have significantly improved our understanding of what we mean by DCD. We
mean, the amount by which an isolated bit can be shorter (or longer) than
its nominal period. Meanwhile, Fibre Channel(?) and oscilloscope
manufacturers have defined it with regard to the time difference between the
average rising edge and average falling edge. We believe that in our
situation, the former effect dominates. For want of a better phrase, we
could call this "pulse length variation".
On this analysis, the dominant component of high probability jitter ("W") is
likely to be pulse length variation: the vertically innermost trace in
figure 52-14 is likely to be innermost time-wise as drawn.
SuggestedRemedy
Change all references to "DCD" or "duty cycle distortion" to "pulse length
variation" or similar. Revisit the value: maybe it is more than 6 ps (it's
hard to know, experimental problems...).
Dawe, Piers Agilent
Stressed eye: frequency of interferer
-------------------------------------
Frequency of proposed interfering sine wave in stressed eye. The upper
limit seems to be, a smallish fraction of the Baud rate, and the lower
limit, a moderate multiple of the highest acceptable CDR bandwidth.
Something like 100 MHz could be OK. Obviously we need to think through the
interaction between this amplitude sine wave and the one driving the SJ:
locking them together would probably be the wrong thing to do.
Proposed simulations of stressed eye generator
----------------------------------------------
We asked Petar to try a few stressed eye generation simulations if he could
find time.
Suggested: start without the SJ. Create vertical eye closure:
1 predominantly (wholly)? with a filter
2 predominantly (wholly)? with a interfering sine wave
How do they differ in terms of jitter, shortest pulse, frequency of shrunken
pulses ... Does pattern length affect this (try PRBS7, PRBS10, longer if
feasible in a simulator)
Try with receiver bandwidths of 5, 7.5, 10 GHz to see if slow or fast
receivers react differently.
If time, add SJ ...
Thank you Petar!
Next phone meeting
------------------
The next PMD teleconference will be on Tuesday 5 February, at the usual time
and coordinates:
4:15 pm GMT = 17:15 CET = 11:15 am EST = 8:15 am PST, Tuesday 29
January 2002
+1(816)650-0631 Access code 39209
Piers