Minutes from the January 26, 2022 IBIS Interconnect Task Group meeting are
attached.
Regards,
Justin
================================================================================
IBIS INTERCONNECT TASK GROUP
http://www.ibis.org/interconnect_wip/ ;
Mailing list: ibis-interconnect@xxxxxxxxxxxxx
Archives at //www.freelists.org/archive/ibis-interconn/ ;
================================================================================
Attendees from January 26, 2022 Meeting (* means attended at least using audio)
ANSYS Curtis Clark
Intel Corp. Michael Mirmak*
MathWorks Walter Katz*
Micron Technology Justin Butterfield*
Randy Wolff*
Siemens EDA Arpad Muranyi*
ST Microelectronics Aurora Sanna
Teraspeed Labs Bob Ross*
University of Illinois Jose Schutt-Aine*
Zuken USA Lance Wang
Professor Jose Schutt-Aine from the University of Illinois introduced
himself. His current research is on algorithms for signal integrity
simulation and modeling.
Michael Mirmak convened the meeting. No patents were declared.
Justin Butterfield took minutes.
Review of Minutes:
- Michael called for review of the minutes from the January 19, 2022 meeting.
Michael displayed the minutes. Randy Wolff moved to approve the minutes.
Arpad Muranyi seconded. The minutes were approved without objection.
Review of ARs:
- Arpad to ask Vladimir if it is possible to reduce ports in the pole-zero
format.
- Arpad reported, from his discussion with Vladimir, it is possible to
reduce ports in a pole-zero format.
- Everyone to provide their requirements for the pole-zero format.
- Michael reported this is the topic for today's discussion.
Opens:
- Walter Katz sent out an email about the format of the network data, and he
thought there could be an issue with the specification.
- Walter asked if we have reviewed the survey. Michael replied that we did
briefly discuss it, and he sent out the link to the survey results. We can
add this to bin list.
Pole-Zero Discussion:
Jose shared a presentation on interconnect modeling using model order
reduction. Since the interconnect circuit model is linear, model order
reduction can be done to reduce the complexity of the frequency dependent
system model. The model form can be a rational function that can be used in
SPICE. Jose gave an overview of the process going from S-parameters to SPICE
simulation. He presented the flow of model order reduction with vector
fitting. He showed the flow for passivity enforcement with the Hamiltonian
Matrix. The goal is to generate a circuit netlist which is modeled as an
admittance calculated from the S-parameter. He compared simulation results
from the recursive convolution method and the pole/zero SPICE model, which
showed good agreement between the simulation methods. Jose noted the
algorithm may come up with negative circuit element values. But, his team
developed an approach to add dependent sources to keep the circuit elements
positive.
Arpad asked if causality can be a problem. Jose replied, when you find a
solution for the rational function, it will be causal. The passivity
enforcement is more computationally intensive due to the size of the matrix
and number of eigenvalues.
Michael stated we are considering a pole-zero format as a way to convey the
same information as a Touchstone model. If we were to use the pole-zero
format, would the pole-zero data be the only data required, or would we need
additional information? Jose replied getting to the pole-zero values is not
easy, while the S-parameters are straightforward for field solvers and VNAs to
measure. Arpad stated the motivation is to reduce the size of the model
format, and the idea was to standardize the pole-zero format. Jose agreed the
pole-zero format would be a much smaller format. Michael commented we have
the opportunity to update Touchstone. For Touchstone 3, one of the feedback
items was to reduce the size, and pole-zero format is attractive for that.
Michael asked if there is any additional information needed to ensure a
quality model and the circuit synthesis is correct. Jose replied you need to
take care of the model order reduction and passivity enforcement. Michael
noted IBIS does not describe how the data should be created and extracted.
Jose stated you would only need the constant term, poles, and zeros, as
complex numbers. You would also want to specify the order of the model, and
you may want to state whether the passivity has been enforced.
Bob Ross asked about support for active S-parameters. Jose replied, for
active systems, the stability depends on the load and is not guaranteed.
Bob asked if there are strategies for zeroing out "Sij" cross terms that are
not significant. Jose replied this exercise does not address this, and this
would be better to address at the S-parameter level. This is a judgment if
the term is important or not.
Next Meeting:
The next meeting will be February 2.
================================================================================
Bin List:
1. Touchstone 3
2. Pole-zero support for Touchstone
3. New version of Touchstone to include 3 TSIRDs
4. Clarification BIRD drafts from ATM
5. Review Touchstone survey results
