[802.3ae_Serial] SJTP: Minutes from today's call
All,
Attached are the minutes from today's phone call. The
discussion was heavily focused on a presentation by
John Ewen that is too large to send to the reflector.
It has been sent to David Law to be loaded onto the
web site. I'll send out a note with it's URL when it
gets put somewhere.
As for the minutes, 90% of this call was over my head
so I'll apologize in advance for their content. Feel
free to respond to this reflector for any clarification
you might want. I'll ask for others in attendance to
provide answers since it is unlikely that I'll be able
to do so.
The bottom line is that we'll have some LAN patterns by
next week that people can start playing with.
Thanks,
Ben
--
-----------------------------------------
Benjamin Brown
AMCC
2 Commerce Park West
Suite 104
Bedford NH 03110
603-641-9837 - Work
603-491-0296 - Cell
603-626-7455 - Fax
603-798-4115 - Home Office
bbrown@amcc.com
-----------------------------------------
Conference call of the Serial Jitter Test Pattern ad-hoc
19-June-2001, 1:00pm EDT
Attendees:
Ben Brown
Piers Dawe
Tom Lindsay
John Ewen
Peter Pepeljugoski
Proposed Agenda:
Discuss the WAN pattern in regards to whether to require
a SONET framing BERT or a bit-based BERT. Then select some
of the options we outlined.
Continue putting in place the TBDs that will allow John Ewen
(and others?) to start looking for seeds to generate the
patterns.
Discussion:
Put the LAN discussion first.
John Ewen has a presentation he's emailing directly as it is
too large to go via the reflector. I've asked David Law to put
it on the web somewhere. I'll let people know when it gets
there.
Walk thgrough the presentaiton:
What is right kind of stress for once per day event.
1) Disparity (window of certain # of bits and calculate disparity
over it)
2) Transition density (transitions/length)
3) Baseline wander (same as Rick Walker - convolved it with
an impulse response)
1) What is probability I'll get a given disparity for a particular
length series? Equations given for review.
Plot shows the disparity for given pattern lengths based on
multiple probabilities.
2) The same info is provided for transition density.
May be interested in patterns of length ~2500 bits (PLL bandwidth
is baud rate / 2500 bits). This would be a subset of the ~8000
bits in each pattern.
3) Assuming some high pass filter generating the baseline wander.
This high pass filter is doing a running average over the
bits. Corner frequency is baud rate / 5000. This is similar to
a moving average over 2500 bits.
Seed selection:
Use maximum run length (58), run the scarmbler backwards to
put the run in the middle.
Next use run length of 57 and so on.
Get down to 55 1s, put 0s on each end then put a 0 then a 1
in the spare bit.
Get down to 54 1s, put 0s on each end then have 2 spare bits
to get 4 samples from.
Do the same with runs of 0s.
Also test with different data inputs (0s and LF).
See plots of these patterns.
Typical selection point might be (see slide 6):
0.2 for base line wander
0.37 for minimum transition density
Groupings seem very explicit (runs of 1s and 0s, data input
of LF and 0s).
Could check this with inverting the data input and the seed
and seeing that the blue and red marks change places.
More info on slide 6.
0.2 peak baseline wander comes from 8 x 2.5% (from the RMS
base line wander calculation just above).
The results of this analysis are not well understood.
Next step:
Pick a few of these patterns and distribute these so we can
get them through the BERTs and into the hardware.
Request for:
1) more contours, to run these simulations using higher proability
values.
2) Also, use a shorter PRBS (2^15) to look at the entire pattern
and use a sliding window to plot out these values for all windows
across this full pattern.
3) Compare these results against the existing plots.
The point is to make the selection of seeds more random rather
than hand picking them.
Work for next week:
John has been requested to provide some more plots - he may
not get to these after spending time providing the patterns.
Choose some seeds, provide ascii files with seed, data input
and resulting pattern (all 128 blocks) to make available for
testing to all interested parties.
Intersection of 0.35/0.1, 0.35/0.2, 0.4/0.2 & 0.45/0.1 for
basline wander/transition density.
Is choosing a point that is 10^-15 probability for baseline
wander and 10^-15 probability for transition density making
this a 10^-30 probability? Or are these things really not
correlated?
Quick WAN question:
Is using a SONET framing BERT
BOX vendors probably have SONET based BERTs
Transceiver vendors probably have bit based BERTs
All attendees preferred the bit based variety.
I'll enter this information into the WAN discussion on the
reflector.
Thanks to all for attending!
Call ended at 2:40 EDT