IBIS Macromodel Task Group

Meeting date: 10 November 2015

Members (asterisk for those attending):
ANSYS:                      * Dan Dvorscak
                            * Curtis Clark
Avago (LSI):                  Xingdong Dai
                            * Bob Miller
Cadence Design Systems:     * Ambrish Varma
                              Brad Brim
                              Kumar Keshavan
                              Ken Willis
Cisco:                      * Seungyong (Brian) Baek
eASIC:                      * David Banas
                              Marc Kowalski
Ericsson:                     Anders Ekholm
GlobalFoundries:            * Steve Parker
Intel:                        Michael Mirmak
Keysight Technologies:      * Fangyi Rao
                            * Radek Biernacki
Maxim Integrated Products:    Hassan Rafat
Mentor Graphics:              John Angulo
                            * Arpad Muranyi
Micron Technology:          * Randy Wolff
                              Justin Butterfield
QLogic Corp.:                 James Zhou
                              Andy Joy
SiSoft:                       Walter Katz
                              Todd Westerhoff
                              Mike LaBonte
Synopsys:                     Rita Horner
Teraspeed Consulting Group:   Scott McMorrow
Teraspeed Labs:             * Bob Ross
TI:                           Alfred Chong


The meeting was led by Arpad Muranyi.

------------------------------------------------------------------------
Opens:

- Arpad mentioned that the meeting in two weeks (November 24th) occurs two days
  before the Thanksgiving holiday in the US.  He asked if people would be likely
  to miss the meeting because of travel plans.  Ambrish stated that today's
  meeting would be the last he could attend this calendar year.  However, since
  he was the only attendee to state that he would be unable to attend, the 
  meeting on November 24th was not cancelled.

--------------------------
Call for patent disclosure:

- None.

-------------
Review of ARs:

- Fangyi to update Walter's Redriver Flow BIRD proposal.
  - In progress via email discussions.

-------------------------
Review of Meeting Minutes:

- Arpad: Does anyone have any comments or corrections? [none]
- Arpad: Motion to approve the minutes.
- Curtis: Second.
- Arpad: Anyone opposed?  [none]

-------------
New Discussion:

New Attendees:
- Steve Parker of GlobalFoundries (the group acquired from IBM) reintroduced
  himself and said that his current interests are in developing NRZ and PAM4
  AMI models.
  
- Seungyong (Brian) Baek of Cisco introduced himself.  He is interested in AMI
  model development and has been studying correlation of AMI simulation results
  and hardware measurements [see presentation below].
  
New Redriver Flow BIRD:
- Discussion:  Arpad asked Fangyi to summarize the state of things after the
  private email discussions that had occurred over the previous week.  Fangyi
  said that details were still being ironed out, but he and Walter still agreed
  on Fangyi's proposal for an additional IR to be added the IR matrix passed to
  the AMI_Init() function of the downstream Rx.  Since Walter was unable to
  join today's meeting, further discussions were deferred.

- Arpad: Since Walter isn't here, I think we can also defer discussion on the
         cleanup of ground language.
  - Brian asked me how he might best introduce his work on correlation of AMI
    model simulation results.  I think the topic might be a bit more related to
    the IBIS Quality Group, but since the ATM meetings are better attended I
    thought we could have it presented here first.
    
IBIS-AMI Hardware Correlation:
  Brian reviewed a presentation on his work with correlating AMI Model
  simulation results and hardware measurements.  He pointed out that model
  makers sometimes show correlation at one point, for example, show the measured
  and simulated values of a given metric like eye-height under one set of
  conditions.  He stated that this is generally insufficient because one of the
  primary purposes of AMI models is to allow the user to determine the optimal
  settings for their channel.  If AMI results and measured results matched under
  one set of conditions, but they trended differently as settings changed, then
  the AMI model would not be useful in determining the optimum setup.  He
  presented examples of a more useful "trend correlation" approach.  A metric,
  such as eye-height or eye-width, is plotted as a function of swept tunable
  parameter settings.  Even if the absolute values of the metric are not
  identical for the measured and simulated cases, as long as the trends are
  similar the model will be useful in determining the optimal settings.  That
  is, as long as the shapes of the measured and simulated curves are consistent,
  the model will be useful for optimization of settings.  Brian had studied
  time domain flow and statistical flow if the models supported it.
  
- Brian: Could we add a trend correlation approach like this to the standard?
- Arpad: IBIS generally doesn't mention correlation at all.  There is some basic
         info on data gathering.
  - Why do you feel this should be in the spec?
  - Do we see a big problem with correlation of hardware to simulation models?
- Brian: Overall, we often see issues if only one correlation point is given.
  - If we added correlation work in the standard, results and trust in the
    models might improve and allow AMI to be further adopted.
- Arpad: IBIS Quality has a document, perhaps this could be added there?
  - Would you consider that, or do you only want to consider the main IBIS spec?
- Bob R: I think it's appropriate for the IBIS Quality spec.
  - Perhaps even an IBIS-AMI Quality spec.
- Brian: Is there anything related to correlation in the main IBIS spec?
- Arpad: Not that I know of.
- Bob R: There is some mention of measurement setup and how to derive data.
  - No information on correlation studies.
  - The IBIS Quality document would be the place for it.
- Ambrish: Generally a correlation report is provided by the model maker.
  - We don't generally see correlation procedures defined for other types of
    models either.
- Bob R: Did you try different simulators?
- Brian: Yes, we used two different simulators.
  - The absolute values differed, but trends were the same.
  - We spent a lot of time debugging tool vs. model issues.
  - That's part of why we like this trend correlation approach.
- Bob M: Both the model and the tool can have problems.
  - The more complex receivers get, the harder it is to get good correlation.
- Arpad: IBIS-AMI now supports PAM4, any thoughts on PAM4?
- Brian: We started correlation work, but as yet haven't found good eye
         monitoring software for PAM4 eyes yet.
  - We can get PAM4 results once tools catch up.
- Arpad: What BER did you use for your plots?
- Brian: As a simulation time compromise, we used a BER of 1e-10.
  - We used a PRBS31 pattern.
- Arpad: How closely did you work with the model vendors?
  - Did you have access to proprietary information?
- Brian: It varied by vendor.
  - For some we had access to proprietary information and settings.
  - For other vendors we only had access to the standard AMI model.
  - All vendors provided AMI models and evaluation boards.
- Bob R: Did you also use eye-width as a metric?
- Brian: Yes, we saw similar trends but larger offsets (relative to eye-height).
- Arpad: Now is a good stopping point.
  - Thank you all for joining.
  - Note: At this point Radek and Fangyi said they would not be able to attend
    next week's meeting (November 17).  Arpad then said he would check with
    others including Walter and might cancel the meeting if too many people were
    unavailable.
  

-------------
Next meeting: 17 November 2015 12:00pm PT
-------------

IBIS Interconnect SPICE Wish List:

1) Simulator directives
