Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[802.3ae_Serial] From serial PMD call 26 Mar: CDR tolerance and stressed eye, feasibility, clock ppm, other




Geoff Garner and Juergen Rahn raised the subject of rate tolerance (20 vs.
100 ppm - see the unsatisfied TRs).  As the optics group does not really
have the expertise or the strong opinions in this matter, this would be
better raised as e.g. a comment to "clause 00" for discussion in the larger
group, or on the main HSSG reflector.  See emails there, written after the
meeting.

Peter Ohlen had grouped the outstanding issues into five categories, now
reduced to two by closed comments:

How to measure things
Rate tolerance

Juergen said he had withdrawn three TRs: 99032, 99035, 99037 I think.

Repeatability of 1% histogram point as in new stressed eye recipe said to be
equivalent to 0.1 dB - found experimentally.

I was concerned that the wide variations allowed in the stressed eye
generator would lead to "false" positives and negatives according to the
implementation of the stressed eye generator.  There are two areas where
variations are allowed:
1.	Relative proportions of filtering and (vertical) interferer - these
are not equivalent in frequency terms.
2.	Variable amounts of sinusoidal jitter in the range x3 wide, are
allowed.  This is effectively "raising the bar" for jitter tolerance,
depending on the stressed eye generator.  We think the resultant "raised
bar" is similar to OC-192 jitter tolerance, but harder, in terms of jitter
corner frequency, than D4.1.  Mike Dudek pointed out that any measurement
standard must have tolerances, e.g. for scope filter bandwidth.

Notice that in D4.2, Table 52-19 and Figure 52-4 do not agree.  The argument
for keeping the corner frequency fixed at 4 MHz (as in the table), rather
than keeping the low frequency sinusoidal jitter fixed (as in the figure) is
that it is related to the fixed-bandwidth clock recovery test equipment used
in several tests.  The argument against is lack of consistency in the test.

I pointed out that we have eliminated the spectral width requirement for
10GBASE-L.  I was assured that we did that deliberately; with a TDP
measurement it is redundant.

Proposed round robin of measurement filters or similar has been abandoned
for logistical reasons.

A disappointingly small number have notified their intent to contribute to
Stretch's anonymising data collection exercise.  Stretch promises not to
divulge confidential information to any transceiver vendor, inside his
company or not, with or without an non-disclosure agreement.  But NDAs can
be accommodated if it helps.
  Without participants to this exercise I don't see how we can show
technical feasibility of the test methods and progress the standard.

Piers

Present
-------
Piers Dawe		Agilent SPG
Peter Ohlen		Optillion
Petar Pepeljugoski	IBM
Stretch Camnitz	Agilent T&M
Mike Dudek		Cielo
Geoff Garner	Lucent
Juergen Rahn	Lucent
Tom Lindsay		Stratos
more...

Next phone meeting
------------------
Tuesday 2 April, 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

Last chance before close of ballot!

Piers