[802.3ae_Serial] From Serial PMD 21 Aug: calibrating jitter measurement, stressed eye definition
Link model treatment of stressed sensitivity and any margin
-----------------------------------------------------------
Looks like we have a consensus, but haven't yet got around to documenting it
at any length.
Definition and measurement of stressed eye
------------------------------------------
The draft standard says that the stressed eye should be prepared without the
sinusoidal jitter, its vertical eye closure penalty measured, then the SJ be
added. Even though fibre channel does it this way, it seems to make more
sense to measure the final thing which is used, so suggested that we change
the standard to say that stressed eye should be prepared, and its vertical
eye closure penalty measured, with the sinusoidal jitter. The value of VECP
would change, for the same final eye.
It still isn't clear to the group whether this measurement can be carried
out with a near pseudo-random pattern or whether a very short pattern is
required.
If the principal objective of the SJ is to force a lower bound on PLL
bandwidth with the amplitude ~ 1/f part in Table 52-21, then we could
consider leaving out the amplitude = constant part to simplify the test.
Calibrating jitter measurement
------------------------------
Some experimental findings showing the significance of the equipment.
Comparing BERT jitter bathtub with scope 10,000 waveforms:
2^23-1 PRBS Pattern gen -> modulator E/O -> DCA (scope) 54 ps pk-pk
Sine wave generator -> Scope 12 ps pk-pk.
Implies 44 ps (?) for DUT. My notes aren't too clear but I think this was
to be compared with an uncorrected BERT measurement:
Pattern gen -> modulator E/O -> pin -> TIA -> BERT 37 ps.
Not sure exactly, but:
Pattern gen -> BERT on the order of 20 ps pk-pk
Pattern gen -> scope ~22 ps pk-pk with various trigger methods.
Infers that pattern generator and scope each contribute 10-12 ps. The test
receiver could contribute more than that.
The conference was cut short by a dropped line. Even had it continued, I
think we don't have a way of separating E->O jitter from O->E jitter.
Next phone conference
---------------------
Was on 6 Sept. The one after that will be on 11 Sept, back to the usual
Tuesday slot:
15:15 GMT = 4:15 p.m. BST = 17:15 CET = 11:15 a.m. EDT = 8:15 a.m.
PDT, Tuesday
+1(816)650-0631 Access code 39209
Piers