IBIS Macromodel Task Group

Meeting date: 24 January 2023

Members (asterisk for those attending):
Achronix Semiconductor:       Hansel Dsilva
Amazon:                       John Yan
ANSYS:                      * Curtis Clark
                              Wei-hsing Huang
Aurora Systems:             * Dian Yang
Cadence Design Systems:     * Ambrish Varma
                              Jared James
Google:                       Hanfeng Wang
                              GaWon Kim
Intel:                        Michael Mirmak
                            * Kinger Cai
                              Chi-te Chen
                              Alaeddin Aydiner
Keysight Technologies:        Fangyi Rao
                              Majid Ahadi Dolatsara
                              Ming Yan
                              Radek Biernacki
                              Rui Yang
Luminous Computing            David Banas
Marvell                       Steve Parker
Mathworks (SiSoft):         * Walter Katz
                              Mike LaBonte
Micron Technology:          * Randy Wolff
                            * Justin Butterfield
Missouri S&T                  Chulsoon Hwang
                              Yifan Ding
Rivos                         Yansheng Wang
SAE ITC                       Michael McNair
Siemens EDA (Mentor):       * Arpad Muranyi
Teraspeed Labs:             * Bob Ross
Waymo:                        Zhiping Yang
Zuken USA:                    Lance Wang

The meeting was led by Arpad Muranyi.  Curtis Clark took the minutes.

--------------------------------------------------------------------------------
Opens:  Arpad noted that the meeting scheduled for January 31 was cancelled
because of DesignCon.  He asked whether we should also cancel the meeting on
February 7th, as it's so close to DesignCon.  Kinger suggested we cancel the
meeting on the 7th, and he asked people to use the time to review the SPIM BIRD
draft.  There were no objections.

- Status of regularly scheduled upcoming meetings:
  - January 31, 2023 - cancelled - week of DesignCon
  - February 7, 2023 - cancelled

-------------
Review of ARs:
  
- None.

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

- None.

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

Arpad asked for any comments or corrections to the minutes of the January 17th
meeting.  Kinger moved to approve the minutes.  Ambrish seconded the motion.
There were no objections.

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

Standard Power Integrity Model (SPIM) BIRD draft:
Kinger reported that he had met with Bob.  Bob had updated the keyword hierarchy
tree (section 3.3 of IBIS 7.1) with the new SPIM parameters.  Bob had also
suggested some formatting cleanup in the example for [Device SPIM Group].  Arpad
noted that the Solution Requirements section is a separate section of the BIRD
explaining the end goal, but it does not belong in the Proposed Changes to be
incorporated in the specification.  Kinger said he would send out a draft8
incorporating Bob's changes and addressing Arpad's comment.  Arpad and Kinger
asked everyone to carefully review the draft and provide feedback.

PSIJ Sensitivity BIRD draft:
Kinger reported that he had created a new draft by incorporating relevant
suggestions he had received for the SPIM BIRD draft, per Arpad's suggestion at
the previous meeting.  Kinger said he would send out draft5 after the meeting.

Multi-level analog buffer modeling (PAMn, C-phy, etc.):
Arpad previewed a presentation he will be giving at the DesignCon IBIS Summit on
Friday, February 3, 2023.  He noted that IBIS 6.1 introduced support for PAM4,
and IBIS 7.2 will support PAMn, but only in the AMI context.  He said there are
scenarios for which supporting multi-level signaling with conventional analog
IBIS [Model]s would be beneficial, and he referred to a previous GDDR6X
presentation:
   https://urldefense.proofpoint.com/v2/url?u=https-3A__ibis.org_summits_aug21b_wolff2.pdf&d=DwIGAg&c=nKjWec2b6R0mOyPaz7xtfQ&r=DcQR-qLpQg5lIreuM6-NYECRIAFXt268PRNS5WO043M&m=CVNFfzBskqJ0L9vcmQNqZrw_cV5Kz7_kM6tDxcv9O-3h1WL0CUUDVja8RWZlrz2Z&s=7sKZGGaYP0bQrSuTksQkvNwjD94svUju8KLQ5ucaow8&e= 
   
Arpad said that extending the traditional I-T, V-t, K(t) algorithms to support
multi-level signaling would not be too difficult.  The difficulties might lie
in the logistics of interactions with other keywords.  Arpad mentioned [Driver
Schedule] as an example of such a keyword.

Arpad noted that Randy had suggested that a minimal solution might be to extend
IBIS by adding a multi-level stimulus for use only with multi-lingual models.
Arpad suggested that instead of the currently defined '0', '1', and 'X' states,
we might allow an integer value to represent the various states.  He said one
complication might be the A_D and D_A converters defined for multi-lingual
models, but he suggested a simple linear equation of the form:
   V = k * (stimulus_index) + c
might be sufficient to represent the voltage for each stimulus level.

Bob asked if this was intended to work for the true differential modeling
supported by [External Model].  Arpad said he thought it would.  He said the
existing true differential models still take a single value to represent the
stimulus.  Arpad said the pseudo-differential models might encounter an extra
complication.  Currently, we expect one [Model] to get the stimulus and the
other [Model] to get the inverted stimulus.  We might have to decide how this
would work for a multi-level stimulus, but he said these are details we could
work out later.

Walter said he had concerns about a solution that only supported multi-lingual
models.  He said he wasn't sure anyone would need to model these devices outside
the AMI context, but if we were to enhance conventional analog IBIS [Model]s, he
would prefer something that a SPICE engine could support.

Walter suggested a simple model based on a lab measurement scenario.  He said
you might connect a PAMn device to a 50 Ohm load to ground, put the device
in each of the n states, wait for it to settle, and measure the voltage and
impedance of the device at steady state.  You would end up with a table of the
impedance of the device at each of the levels.  You could add in legacy power
and ground clamp tables, and have the model maker define some effective K(t)
rise/fall time curves for the transitions from state to state, and you'd end
up with a simple and effective model that is easy for EDA tools to implement.
You could define an algebraic way to extend such a basic table model to handle
non-linear regions.

Alternatively, Walter suggested that you could extend a typical [Model] set
of active device I/V curves to 4 curves (for a PAM4 example) and define a set
of K(t)s that describe how you switch from one state to another.

Arpad said these were interesting suggestions.  However, he said that we will
need an expanded stimulus definition regardless of how we choose to implement
the multi-level model.  Arpad said he thought we could consider the expanded
stimulus definition and the multi-level modeling as separate BIRDs.

- Walter: Motion to adjourn.
- Curtis: Second.
- Arpad: Thank you all for joining.

AR: Kinger to send out draft8 of the SPIM BIRD containing changes reviewed and
    discussed in today's meeting.
AR: Kinger to send out draft5 of the PSIJ Sensitivity BIRD containing changes
    incorporated from SPIM BIRD draft feedback.
-------------
Next meeting: 14 February 2023 12:00pm PT
-------------

IBIS Interconnect SPICE Wish List:

1) Simulator directives