IBIS Macromodel Task Group

Meeting date: 21 June 2022

Members (asterisk for those attending):
Achronix Semiconductor:       Hansel Dsilva
Amazon:                       John Yan
ANSYS:                      * Curtis Clark
                              Wei-hsing Huang
Cadence Design Systems:       Ambrish Varma
                              Jared James
Google:                     * Zhiping Yang
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
SAE ITC                       Michael McNair
Siemens EDA (Mentor):       * Arpad Muranyi
Teraspeed Labs:             * Bob Ross
Zuken USA:                  * Lance Wang

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

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

- None.

-------------
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 June 14th
meeting.  Randy moved to approve the minutes.  Bob seconded the motion.
There were no objections.

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

Supporting PI simulation in IBIS:

Arpad recalled Bob's comment, from the previous meeting, that the .spi
extension itself might cause confusion, as the name collides with some industry
standard extensions for power aware simulators' file names.  He suggested we
might use .pim (power integrity model).  Walter suggested that we could also
feel free to move beyond 3-character extensions.  Kinger was amenable to choosing
a different extension than .spi.

Kinger said he had met with Walter to discuss the proposal's requirements and
the best ways to leverage existing EMD and IMS keywords.  Kinger said in these
types of problems we typically want to have a local reference for each port, and
one GND ball might be used in the reference for multiple port groupings.  Kinger
thought an existing limitation in IBIS is that we don't have a fully defined
Port keyword, and the port definitions in IBIS refer only to the positive side
of the port.  There's one provision for a global reference, and no provision
for different local references.

Kinger again reviewed his simplified example from the previous meeting.  He
agreed that existing bus_labels can be used to define grouping, but he suggested
that we could enhance the terminal lines to further define the reference
terminals for the ports and specify which pins were included in each reference.
He asked why we currently have to spread this information across multiple files
when we could include it all in one place.

Arpad noted two ports in Kinger's example that shared two of the GND pins in
their respective reference pin groups.  He noted that since those reference pins
are shorted with the rest of the pins in their respective groups, the sharing
will ultimately short all the reference pins of both groups.  He asked if this
was the intent, and he observed that these local reference groups could become
large sets of shorted pins if multiple ports share pins.  Kinger said this was
the intent, and that in practice this sharing would not daisy-chain into large
groups of shorted pins.

Bob asked if this proposal was only for power and ground.  Kinger said they are
trying to leverage and enhance existing syntax, and it could potentially be
useful for power and ground.  Their primary intent, however, was to update the
bus_label syntax so that one pin can serve in multiple bus_labels, and this is
primarily for GND pins.  Chi-te observed that the extraction tool they often use
will not allow a power pin to be associated with more than one port.  Kinger
said the same situation could occur with SI.  He said a DQ bus might have 8 DQs,
and perhaps two of the DQs only have one nearby VSS pin that they share.

Walter said the existing bus_label construct is really a way of saying that a
subset of pins associated with the same signal_name is shorted.  He said a pin
can belong to only one bus_label.  He said the existing bus_label concept may
already confuse people, and he would be very reluctant to overload the bus_label
concept.  He said he would prefer the addition of a new "group name" keyword.
It could a specify signal_name and the list of grouped pins.  The same pin could
appear in multiple groups, and it would provide what Kinger's scenario required.
He said the group name would have to be unique and not collide with any
signal_names or bus_labels.

Kinger recalled that his original proposal contained some port grouping keywords
similar to what Walter was suggesting.  He said they had arrived at this new
proposal for enhancing bus_label after trying to go back and leverage existing
keywords and syntax.  He said there is legitimate need for the functionality, as
we see more and more power pins surrounded by fewer ground pins.  He said that
if not for the need to be able to use the same ground pin in different reference
groups, they would be able to re-use existing syntax entirely.

Walter asked why this local ground capability was needed by the end users of
these models.  He said board-level PI designers would typically be working at
frequencies well below 300M, and at these frequencies the local reference
concept doesn't make any difference.  Kinger said, for example, we might be
talking about a 100A power rail with hundreds of pins and thousands of bumps,
and at the bump level you certainly need local references.  He said there are
OEMs using Intel CPUs who are doing board level PI designs according to the
scenarios he had described.

Kinger said if Walter wasn't comfortable with the idea of enhancing bus_labels,
then he was willing to go back in the direction of his original proposal with
new keywords.  Zhiping disagreed with Walter's preference for introducing a new
keyword.  Zhiping said we might be better off to enhance bus_label and provide
the same local reference capability for SI designers too.  Arpad and Randy noted
that bus_labels are only for POWER and GND pins.  Kinger agreed and said he and
Zhiping were referring to the fact that SI modeling could also benefit from
providing local ground reference grouping.

- Curtis: Motion to adjourn.
- Randy: Second.
- Arpad: Thank you all for joining.
    
-------------
Next meeting: 28 June 2022 12:00pm PT
-------------

IBIS Interconnect SPICE Wish List:

1) Simulator directives
