Signaling ad-hoc: minutes (September 16, 2004)
- Mike Altman begins
with an agenda review from (http://ieee802.org/3/ap/public/signal_adhoc/altmann_02_0904.pdf)
- Ali Ghiasi and
Steve Anderson present the StatEye based proposal (http://ieee802.org/3/ap/public/signal_adhoc/ghiasi_01_0904.pdf ).
- Ali starts with
an introduction to StatEye.
- No Duobinary
support is currently supported but Ali feels that it should not be a big
deal to add it in.
- A question was
raised regarding S-parameters and ABC parameters? What are
they?
- Mike This is a translation step for
matrix manipulations used for cascading s-parameters
sets.
- Mike Does StatEye allow mismatching of DC impedances
?
- Ali Yes.
- Mike
And can these be changed independently?
- Steve
A. Not in the version I am familiar with.
- Ali
yes by modifying the script
- Steve continues
the presentation
- ISI
- Fulvio What do you mean when you say that MSJ
does not include ISI effects?
- Steve I meant that MSJ was developed foe channels
where there was not much attenuation.
- ADS
- Mike Is the ADS support
definite?
- John Yes, Agilent seems interested in doing
so.
- BER
- Mike Does the BER include RJ?
- Steve Yes
- Mike Can you comment on how RJ is added? Is it
just added to the contours extracted from impulse
response?
- Steve RJ is added by computing the composite
PDF. Slides later in the presentation will discuss
it.
- CORRELATION
- Mike Is there any data correlating
StatEye to measured channels
- Joe It is generally agreed
that StatEye is conservative.
- Ali That is always true
when dealing with statistics vs. measurements.
- Charles If we are using the
tool for comparing signaling methods, the only important point is
whether any pessimism may or may not be independent of signaling
techniques.
- ?? but there is still
need for absolute performance indicators
- Mike Yes, I would like to encourage an
approach that shows correlation to measurements.
- RJ
- Mike It is not clear that the probability
distribution (slide 21) includes random effects. It seems pretty
deterministic. Am I interpreting correctly?
- Steve Bit patterns randomize the
results.
- Mike But that is not really
random
- Steve Slide 21 illustrates the conditional
probability function due to ISI.
- ??? Are you sure that this does not
include voltage noise.
- Ali ... whatever shortcomings
there may be we need a starting point ...
- MULTILEVEL
SUPPORT
- Mike Things get even more complicated as we
increase the number of levels. Have the StatEye folks validated other
line codes.
- Steve They have added PAM and RLL. The only thing
missing is Duobinary.
- Ali They have 8B10B,
64/66 etc.
- Mike How do they include the effects of
coding ? 8B10B and 64/66 are probably not the best for
PAM4?
- Ali It should not be
difficult, given a certain encoding scheme. We could add a PAM4 encoder,
modify one or ask them for modifications.
- Adam Actually StatEye is based on random
equi-probable bit sequences. For example, selecting 8B10B has the only
consequence of changing the symbol rate. No other effect is
included.
- Ali You can always
import a bit sequence appropriately encoded
- S-Parameters
- John I just want to verify that
StatEye is only concerned about Sdd
- Steve Yes, StatEye does not use CM to DIFF and
vice versa.
- Charles
Hspice does, but in practice it depends on the information embedded in
the Tx and Rx
- ADAPTATION
- Mike What kind of adaptation? MMSE?
LMS?
- Ali They (StatEye) so
some global optimization
- Steve Originally StatEye tried to determine the
coefficients on a local basis but now it's looking at it globally
scanning both coefficients and sampling point.
- Mike So the optimization is not really
representative of a particular algorithm or inclusive of any
correlation/interaction between different loops. So it is, basically, a
stationary process.
- Fulvio what about the effects of coefficients
quantization/granularity
- Steve It was discussed some time ago but I do
not know if it has been pursued.
- MULTILEVEL
- Mike Do you have an example of multilevel
coding. It would be very good to see examples of how this is done. Are
any of the multilevel guys in this group using StatEye?
...
- Mike It sounds like that is not the case.
Can I ask an experienced user to demonstrate StatEye on multilevel
code?
- John This should include
Duobinary.
- Mike Yes, absolutely. I am not going to
care who does it as long as it gets done. I am asking for a volunteer.
If none is found, we will have to find a volunteer some other
way.
- ?
StatEye for PAM4 was demonstrated at OIF
- Adam I am familiar with this presentation. What
did not come out of it is the suitability of StatEye for PAM4
simulation. We need to have someone do so. We should do the same for
Duobinary (or other signaling schemes that could be proposed in
September, e.g. PR4)
- Mike Soliciting volunteers
...
- Mike thanks the
presenters.
- Mike resumes his
presentation, illustrating the areas that need work:
- channel
simulation model (slide 3)
- viability of ABC
matrix cascading (slide 4)
- type of Data.
Encoding. Typical packaging effects. (slide 5)
- BER or
sensitivity of BER to input noise? Timing Recovery sensitivity. Interferers
(slide 6)
- Mike
these are the topics I would like to have addressed so we can be clear on how
to proceed. Anything else to add ..
- Fulvio Specify the number of
aggressors
- Mike
... anything else? No!
- Mike (to
Adam) Could we have specific ad-hoc signaling meetings in conjunction
with the Interim mtg in Ottawa?
- Adam I would
hesitate to do so. But there are discretionary alternatives (evenings etc). I
can talk to the meeting organizers.
- Mike
Fair enough. We will play it by ear.
- Adam Since this is
not a plenary, evenings are our own and we can certainly set up things for one
night.
- Adam We need to
make sure to have presentations filed for the general
meetings.
- Meeting
adjourned.