[802.3ae_Serial] From serial PMD call 18 Dec: Tx and Rx conformance tests, other
1. Calibrating stressed eye
2. Alternative transmitter conformance test
3. Receiver conformance in two tests or one?
4. Receiver bandwidth test
5. Receiver overload test
6. Terminology: ISI
7. TIA Symposium on Reference Receiver Calibration
8. 10GFC and slightly negative margins
9. Next phone meeting
1. Calibrating stressed eye
------------------------------
I reported very encouraging results with a lone bit pattern containing runs
of five like bits (four seemed too few) with lone bits, e.g. like
11111010000010... . I got clear scope plots from which reasonably precise
readings could be taken using the markers. Tom mentioned a K.??.? pattern
which he has found useful. We don't know if the worst bit would be the same
bit in a pattern for all marginal receivers. [After the meeting - but I
think we could find out that the worst bit will be the same one for all
stressed eyes created with coax cable.] Because the lone bit pattern is
neither transmitted nor received by compliant parts, it doesn't need to be
hard specified in our standard.
2. Alternative transmitter conformance test
----------------------------------------------
If we believe that the BERT bathtub method cannot be calibrated for "W"
jitter, and giving on the order of 10 ps to the test equipment means
over-engineering the device under test (cost!), then here is an alternative.
Use the scope to screen for high probability jitter, mostly "W",
using the left and right most points of the eye mask.
Use a transmitter penalty test similar to the ER/EW one to assure
low BER at eye centre even with stress from receiver noise.
The combination forces "sigma" jitter below a bound.
To do this the eye mask would be lengthened for the two "50% up" points.
But this gives a mask which is tougher than OC-192, partly because of this
but also because our eye is fixed in amplitude and time while the OC-192 eye
is allowed to float. This is in contradiction with the overall aim of
cost-effectiveness.
To compensate for this we could make the mask less high. The area under the
mask would be similar to what it is now, so an integrate-and-dump type of
decision circuit would achieve similar performance. However a truly fast
decision circuit for which vertical opening was paramount would have to
deliver a little more performance (as I expect such a fast receiver could).
A variation on this was proposed, where the mask was lengthened at the 50%
points, the mask height remained, but the four other corners were chamfered,
giving an eight or ten sided mask which would be a better fit to real
transmit eyes but still provide an acceptable "contract of performance" to
the receiver.
Now, can this be applied to all PMDs? There was a discussion of whether the
mask could be measured at the virtual TP3, which is where the transmit
bathtub is measured.
For SR/SW, a receiver, transversal filter and electrical scope head
could be used.
For LR/LW, typical optical scope heads appear not to have quite
enough sensitivity to avoid adding noise to the eye mask test after 10 km of
real fibre. While in principle this could be calibrated out, we were
reluctant to go there. Achieving the same dispersion with less attenuation
probably could be achieved using non-standard fibre for shorter wavelength
lasers but not so readily for longer wavelength lasers, though more exotic
techniques are possible in principle (Peter did I get that the right way
round?). But does this really matter? Transmit wavelengths are likely to
be either well centred (in which case arguably we don't need the dispersion
in the test), or at the short end of the window.
For ER/EW, an EDFA and regular optical scope head could be used, we
think.
3. Receiver conformance in two tests or one?
-----------------------------------------------
Tom prefers two tests because it seems more practical (calibratable) to him.
The two tests are basically, vertically closed eye and jittered eye. Gair
Brown has written supporting one combined test, as a more direct criterion
and to allow trade offs between individual components of impairment. I
thought that it might be easier to generate an eye with defined amounts of
vertically closed eye and jitter simultaneously, than two eyes, one with low
vertically closure and another with low jitter.
4. Receiver bandwidth test
-----------------------------
Tom had expressed surprise that this test was so unfamiliar. I thought
perhaps it had been carried out a few times only when Gigabit products were
first introduced but may not be not done regularly. Because the analog
front end is so much faster this time (and the decision circuit may not have
kept pace), we don't know what will happen this time. No-one on the call
volunteered any experience with this test; has anyone reading this email
tried it?
5. Receiver overload test
----------------------------
Should the overload test be carried out with a nominal eye or a stressed
eye? It seems likely that a jittered eye would be more likely to cause a
fail than a clean one.
6. Terminology: ISI
----------------------
We see "ISI" and vertical eye closure penalty used almost interchangeably.
We think that ISI (inter symbol interference) is the physical effect, and
vertical eye closure is one measure of its results: data dependent jitter
may be another. Are we right? Clarification welcome. A few editorial
changes to the draft may flow from this clarification.
7. TIA Symposium on Reference Receiver Calibration
-----------------------------------------------------
Tom reminds us of this half day(?) meeting, hosted by TIA 6.1.10 and IEC
TC-86/WG4 on January 23, 2002 and advertised by Mike Hackert on the
'stds-802-3@ieee.org' reflector. This meeting will certainly be addressing
issues that concern us right now, we should cover it.
8. 10GFC and slightly negative margins
-----------------------------------------
See serial reflector: 10GFC's line rate is 2% higher, so for similar links
what changes? The PMDs, reach, margins, loss ...
9. Next phone meeting
------------------------
The next PMD teleconference will be in the New Year at the usual time and
coordinates:
4:15 pm GMT = 17:15 CET = 11:15 am EST = 8:15 am PST, Tuesday 8
January 2002
+1(816)650-0631 Access code 39209
The main topic may be optical testing again. Also note, this will be the
last teleconference before the ballot closes and the next face-to-face
meeting.
If you get to this email quickly, Happy Holidays! If not; hope you had a
good one,
Piers
----- End Included Message -----