Signaling ad-hoc: September 9 minutes
Attendance
Joe |
Abler |
IBM |
Michael |
Altmann |
Intel |
Stephen |
Anderson |
Xilinx |
Majid |
Barazande-Pour |
Vitesse |
Ali |
Ghiasi |
Broadcom |
Joel |
Goergen |
Force10 |
Adam |
Healey |
Agere Systems |
Glen |
Koziuk |
Vitesse |
Mike |
Lerer |
Rapid Prototypes |
Cathy |
Liu |
LSI Logic |
Jeff |
Lynch |
IBM |
Mary |
Mandich |
Lucent |
Richard |
Mellitz |
Intel |
Charles |
Moore |
Agilent |
Brian |
Seemann |
Xilinx |
Jimmy |
Sheffield |
Tyco |
Jeffrey |
Sinsky |
Lucent |
Fulvio |
Spagna |
Intel |
John |
Stonick |
Synopsys |
Schelto |
van Doorn |
Intel |
Brian |
Von Herzen |
Rapid Prototypes, Inc. |
Nirmal |
Warke |
TI |
- Difficulties in
joining the call (busy number) were reported by several
participants
- Mike Altman will
send out alternate numbers for next conference
- Meeting called to
order
- Charles Moore
reviews his presentation (http://ieee802.org/3/ap/public/signal_adhoc/moore_02_0904.pdf ). The following are
comments generated during the review.
- On the use of
HSpice as a simulator
- Dimitri
was the use of HSpice and S-parameters verified
- Fulvio
the discussion should be about the technique proposed rather than a
specific time-domain simulator
- Mike
A given that a level
of checking on S-parameters usage is performed, how key is
HSpice?
- Charles not
familiar with other simulators. Have been assuming that most have access
to HSpice ...
- Richard
M ...it is really important to verify
self consistency in time and frequency domain when using
S-parameters
- Mike
A It would be helpful to
put some notes together on how to simulate with S-parameters in the
time domain
- Richard
Mellitz volunteers to do so
- On the use of an
interfering sinusoid
- Fulvio is
the sinusoid interference test to be run at a single frequency, a set of
frequencies or what?
- Charles not
clear yet
- On the model
/block diagram
- Ali should
we not been using a 64/66 pattern instead of a PRBS 2^7? What about AC
coupling effects?
- Charles if someone
generates it I can convert it in the appropriate format
- Ali
what about the package model
- ...discussion
follows regarding the package itself, the effects of
ESD protections, terminations etc.
- Mike
A it is impractical for us
to specify the details of Silicon and/or package implementation. The
approach we are following is to specify the system from TP1,
TP4 point of view.
- ... some more
discussion ...
- Charles in
the end I would like to note that the more details are added to
the package model, the more this becomes an impediment to the
generality of the approach.
- Charles With regards to
the channel model these will be provided by the Channel ad hoc
...
- Mike
A ... Channel ad hoc will
provide limits. This group will select the actual models to be
used.
- On the
post-processing of the data
- Fulvio It seems that this
approach is based on an architecture where the receiver equalization is
done in the continuous time domain
- Charles yes, I have not considered sampled
equalizers
- Mike
A. how do we include imperfections/degradations
introduced by timing loop, equalization loop
- Charles some of it is done in
post-processing
- Fulvio but in the case of
a sampled receiver the interaction between the two systems is not taken
into account
- Ali
especially in this kind of system where the eye is closed you need to
consider that
- Fulvio so if the
assumption is to decouple CDR and Receiver, how is this different then the
proposal put forward last week by Bill and John Stonick?
- John S. If we just look at the eye after
some equalizer with ideal timing seems the right thing to do.BER
estimation inclusive of CDR effects are more complicated. Spice can not
simulate worst case patterns.
- Ali
the same can be said about the equalizer.
- John
S. Right. So if I provide a set of coefficients (a
finite number of taps and finite resolution) I can make an estimate of BER
and allow quantification of time and voltage margins and allow fair
comparison.
- Mike
A. We need a detailed proposal on how to do
so.
- John
S. But is this the correct approach to the
problem?
- Brain I feel that we need
to concentrate on what goes on in the channel. What happens in the
receiver is helpful but should not be used to specify constraints for the
receiver design.
- John
S. Absolutely not.
- Brian You mean that the
signaling should be accompanied by a certain amount of disclosure and not
by imposing constraints?
- John
S. Right. We want to determine the right trade
off.
- Ali
What are we really trying to do here? Are we trying to set an
interoperability standard?
- Mike
A. The task force is now concentrating on selecting a
method by which we can compare different signaling methods.
- Ali
What about StatEye ?
- Steve
A We did present an analysis of the synthesized channel
using StatEye.
- Mike
A ... non of which has been presented to the task force.
The schedule and target have been now public for some time.
- Ali
I am concerned about the time to develop a methodology ...
- Mike
A I can ask you to present what it would take to
develop a StatEye based methodology to compare signaling
schemes.
- Adam As a clarification,
the September deadline refers to presentation of new technical content.
The decision for signaling will take place in November.
- Mike
A Ali has the AR to present to this body a methodology
based on StatEye.
- At this point a
straw poll was conducted to get a measure of the ad-hoc level of interest in
Charles proposal.It is not binding and it is meant to assess whether we want
Charles to fine tune the his simulation system
- John
D Can we poll the "no" to see if they have
alternatives:
- one abstained
form doing so, while the others pointed to the proposal put forward last
week.
- At this point some
more discussion concerning StatEye and its viability took place. The
discussion was stopped by the chair who pointed out that any discussion on the
subject should take place in the context of a formal presentation to avoid
duplication of effort.
- Next Mike
introduces John D'Ambrosia presentation of Channel models. Before giving the
floor to John, Adam H made the following statement: "I want to make clear that
John has volunteered to provide test cases. These are not needed to be the
only ones and I am soliciting other participants to bring other models for the
consideration of this group"
- John D'Ambrosia
goes through his presentation (http://ieee802.org/3/ap/public/signal_adhoc/dambrosia_02_0904.pdf )
- John D A number of pending slides
will be updated and added
- John D All channels presented are
real channels (there is no mix and match of data from different
sources)
- Ali
Is the resonance on TC7 due to reflection
- John
D Yes
- Richard It would be interesting to have well
behaved short channels
- Meeting is ended
due to time constraints. The ad hoc chair thanks all presenters and encourage
the use of the reflector to keep the discussion
going.