IBIS Macromodel Task Group

Meeting date: 09 January 2024

Members (asterisk for those attending):
Achronix Semiconductor:       Hansel Dsilva
Amazon:                       John Yan
ANSYS:                      * Curtis Clark
                            * Wei-hsing Huang
Aurora System:              * Dian Yang
                              Raj Raghuram
Cadence Design Systems:     * Ambrish Varma
                              Jared James
Dassault Systemes:            Longfei Bai                             
Google:                       Hanfeng Wang
                              GaWon Kim
Intel:                      * Michael Mirmak
                              Kinger Cai
                              Chi-te Chen
                              Liwei Zhao
                              Alaeddin Aydiner
Keysight Technologies:        Fangyi Rao
                              Majid Ahadi Dolatsara
                              Stephen Slater
                              Ming Yan
                              Rui Yang
Marvell:                      Steve Parker
Mathworks (SiSoft):           Walter Katz
                              Graham Kus
Micron Technology:            Justin Butterfield
Missouri S&T:                 Chulsoon Hwang
                              Yifan Ding
                              Zhiping Yang
Rivos:                        Yansheng Wang
SAE ITC:                      Michael McNair
Siemens EDA (Mentor):       * Arpad Muranyi
                            * Randy Wolff
Teraspeed Labs:             * Bob Ross
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 30th will be cancelled.
  January 30, 2024 (cancelled - DesignCon week)
  
-------------
Review of ARs:

Michael: Submit the AMI Test Data proposal to the Open Forum for introduction
         as an official BIRD.         
         - Done.  Submitted as BIRD229.
         
Michael: Develop a full syntactically complete example demonstrating the
         AMI Test Data proposal.
         - In progress.

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

- None.

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

Arpad asked for any comments or corrections to the minutes of the December 19th
meeting.  Michael moved to approve the minutes.  Ambrish seconded the motion.
There were no objections.

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

Cleanup of ATM email agenda items:
Arpad initiated a review of the Topic bin list and Tabled topics.

Topic bin list:
Items 1 and 2 - Removing single ended-requirements for differential buffers:
Michael said he would eventually handle these with a single BIRD.  The group
decided to leave them on the list.

Item 3 - Digital signing of DLL/SO files:
Michael said this was a question for other ATM attendees.  Should we mention
this topic in the specification or require signatures?  Should the parser
implement a test for this?  Arpad, Bob, and Michael agreed that this topic could
be taken up by the Quality task group instead.  The group agreed to delete it
from the ATM agenda.

Tabled topics:
10) Modify C_comp for T-coil modeling
Arpad suggested that we wait for more input from Walter.

11) Jitter HF/LF components and Jitter Amplification
Michael said this topic was still relevant, but he had found it difficult to get
technical consensus on how to proceed.  Michael moved that we untable this topic
and remove it from the list.  Curtis seconded.  There were no objections.

12) Other DDR5 related topics
Arpad said this was a general topic for non-linear issues with DDR simulations.
He said that EDA vendors may be using their own special techniques to deal with
some of them.  Michael observed that the specification never says anything about
what should happen if you have a full set of non-linear I-V tables and V-t
curves in an AMI model.  So, it's already left to vendors to decide.

Arpad said that in AMI we only have one set of channel impulse responses
(including cross talk columns) that gets passed to AMI_Init.  That could be
enhanced by adding a second set, so we have rising and falling edge information.
He said this wouldn't be a difficult change to make to the specification, and he
wondered whether it would be worthwhile to add these multiple-edge techniques to
the specification.  Ambrish said he thought it was possible for this information
to be captured in a single response.  He said that prescribing the use of two
different impulse responses was really prescribing one solution to the problem.
Michael said that unless someone is actively asking for a solution, we should
remove this item.  Michael moved to untable the item and remove it.  Bob
seconded.  There were no objections.

13) New BIRDs from Editorial Task Group?
Bob suggested that we remove all 5 sub-headings of this topic.  He noted that
one attempt at addressing item e), BIRD181.1, had been rejected by the Open
Forum.  Bob said that the whole effort had gotten bogged down in an attempt to
change the entire specification to use SPICE type V(node1, node2) voltage
syntax.  While this was technically the correct thing to do, it was burdensome
to rewrite all of IBIS to address it everywhere.  Bob moved to untable item 13
and remove it.  Michael seconded.  There were no objections.

14) How to handle missing min/max data?
Arpad said he thought there was still an issue with the specification.  Some
parameters specify the use of typ if the min or max data is missing.  Some
parameters don't specify what to do, and in some cases it would not make sense
to use typ data in lieu of min or max.  However, he said this issue had been
around forever, and he had never written a BIRD to address it.

Michael asked whether there was a specific keyword that was particularly
problematic.  Arpad said he first ran into the issue when using I-V and V-t
curves.  The I-V and V-t usages would need to match.  Their rules say to just
use typ if min or max aren't available.  However, what would we do if the I-V
contained typ/min/max, but the V-t only contained typ?  We would have a
mismatch.  There are loose ends like this that we should resolve.  Arpad, Bob,
and Ambrish agreed that we should leave this item on the list.  Ambrish  said
he had recently had to refer to the specification to answer a customer question
related to this issue.

15) AMI_GetWave block size with continually adapting models
Arpad suggested that we keep this item on the list for now.

16) Multi-level analog buffer modeling (PAMn, C-phy, etc.)
Arpad said this was not an urgent issue but a possible enhancement.  He said he
had run into this issue when trying to model multi-level signaling in a non-AMI
simulation.  He had implemented an [External Model] using Verilog-A, and he had
wished a multi-level stimulus were available.  Randy said that one of the
possible solutions, adding support for a multi-level stimulus to multi-lingual
IBIS models, might be a simple change.  The group decided to leave this item on
the list.

17) Updates on various PI related topics
Arpad took an AR to ask Zhiping whether we need to keep this topic.


AMI Test Data proposal:
Michael reviewed several issues that he thought had caused of confusion or
risked causing confusion with respect to BIRD229 and the existing IBIS 7.2
specification.

Michael said that there is no formal definition of "block" in IBIS 7.2, though
informally it refers to the block of data passed to each AMI_GetWave call, i.e.,
a block of waveform data whose number of samples is given by wave_size.  Ambrish
confirmed that one "block" of data contains the number of samples specified in
wave_size.

Michael said that when he had created his syntactically complete example by 
by modifying the original 2009 SiSoft AMI example, he had reduced the total
amount of data processed by AMI_GetWave to minimize the size of the example
files.  He said that coincidentally, the total amount of waveform data processed
in bits (1024 bits - 8192 samples in this 8 samples per bit case) happened to be
the same as the number of samples in the impulse response passed to AMI_Init
(1024 samples).  Michael suggested that he should change the example, perhaps to
2048 bits worth of data processed by AMI_GetWave, to avoid any chance of
confusing readers into thinking these quantities were related.

Michael said he might propose a minor BIRD to clarify language related to block
and block size.

- Ambrish: Motion to adjourn.
- Michael: Second.
- Arpad: Thank you all for joining.

New ARs:

Arpad: Email Zhiping to ask whether he still plans to work on item 17 in the
       Tabled topics list, or whether we can remove it.

-------------
Next meeting: 16 January 2024 12:00pm PT
-------------

IBIS Interconnect SPICE Wish List:

1) Simulator directives
