[802.3ae_Serial] From serial PMD call 19 Feb
Present
-------
Piers Dawe Agilent SPG
Paul Kolesar OFS
Jan Peeters Weem Intel, Optical Platform Division
John Ewen JDSU
Raj Savara Network Elements
Benny Christensen Intel (Giga)
Stretch Camnitz Agilent T&M
Adam Healey Agere
Tom Lindsay Stratos
Petar Pepeljugoski IBM
Greg LeCheminant Agilent T&M
Report from Test Fest
----------------------
The test fest was popular and successful and all participants wished they
had more time in the lab. As the conclusions were quite subtle I won't try
to muddy the waters here.
Accuracy of TDP measurement
---------------------------
Are TDP measurements with really shallow BER vs. power gradients
reproducible?
Does TDP correlate at all with anything in Tx jitter bathtub, as expected?
Does reference receiver really need sensitivity similar or better than
"product" receivers?
Does receiver sensitivity affect TDP results? I believe that receiver noise
does not; it cancels out.
Calibration of stressed eye
---------------------------
Tweak what?
We expect that the stressed eye will not be specified so rigidly that the
four parameters signal amplitude, clock jitter, interfering amplitude and
filter bandwidth must be each accurately met, but say specify three of them
(or metrics relating to what they do) and allow the fourth to be tweaked to
hit the targets.
Should a trade between VECP and signal amplitude be allowed? Some thought
not.
Trade filter bandwidth and one of the sine waves? Note the interfering
amplitude can easily be adjusted to tweak the VECP.
Specify what?
Moving towards specifying something on the scope: traces or histograms.
What filter bandwidth?
Petar had used 8 or 9 GHz. Tom said 7.5 GHz caused about 1 dB VECP; Stretch
said such a filter caused similar "sigma" jitter to real transmitters (we
have to decide if that is desirable: I think we are aiming to keep the
"sigma" components low to obtain better calibration), and avoid having to
correct for scope noise.
Specify bandwidth or risetime?
These aren't linked on a 1:1 basis, the relation depends on filter order.
Laser overshoot
---------------
It is possible to create a transmitted eye which passes a mask test when
filtered but is poorly shaped when not, and then if received with a very
high bandwidth receiver, can cause a problem. Draft standard aims to
address this, or the similar issue with bifurcated pulses from multimode
fiber, with a receiver upper bandwidth limit. Fibre Channel added a
receiver -10 dB bandwidth spec (presumably an upper limit again); for this
reason? Another way would be a reduced mask for a non-filtered eye, or one
filtered with a higher bandwidth. Do we need to do anything?
I was in favour of staying with the present scheme where we specify the
receiver, as receivers are often more consistent, cheaper and vary less with
temperature than transmitters.
But there is one "hole in the spec" which we have created in the last
revision; we now specify an eye mask at a virtual TP3. In the case of
10GBASE-S, this involves a 45 ps transversal filter as well as the standard
7.5 GHz filter in the receiver. Thus we are measuring the transmitted eye
through the low extreme of bandwidth, and there could be a problem with a
very ringy transmitter, a reasonable bandwidth receiver, and no transversal
filter (e.g. short or good MMF link). This can be "easily" fixed (easy for
the editor) by requiring that the 10GBASE-S mask be met at TP2 as well as
the virtual TP3, which is what we do for 10GBASE-E, although we call it the
zero dispersion extreme of virtual TP3 rather than TP2.
Next phone meeting
------------------
The next PMD teleconference will be on Tuesday 26 February, at the usual
time and coordinates:
4:15 pm GMT = 17:15 CET = 11:15 am EST = 8:15 am PST
+1(816)650-0631 Access code 39209
Piers