================================================================================ IBIS INTERCONNECT TASK GROUP http://www.ibis.org/interconnect_wip/ Mailing list: ibis-interconnect@freelists.org Archives at http://www.freelists.org/archive/ibis-interconn/ ================================================================================ Attendees from April 17, 2019 Meeting (* means attended at least using audio) ANSYS Curtis Clark Cadence Design Systems Bradley Brim Intel Corp. Michael Mirmak* Keysight Technologies Radek Biernacki Mentor, A Siemens Business Arpad Muranyi* Micron Technology Justin Butterfield*, Randy Wolff* SiSoft Walter Katz*, Mike LaBonte* Teraspeed Labs Bob Ross* 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 April 10, 2019 meeting. Bob Ross noted a spelling error where the word "roll" should be "role". Randy Wolff moved to approve the minutes with this correction. Bob seconded. The minutes were approved without objection. Review of ARs: - Walter Katz to make changes to the EMD draft with tracking turned on [AR]. - Walter noted this is not done, as Bob highlighted some fundamental issues, and he would like to wait to resolve these issues before making changes. Bob noted one of the issues is if we are documenting 2 file formats, one for EMD and one that is analogous to interconnect model sets. One of the things we try to avoid is using the same names to describe different structures. Walter asked if this is simply a keyword naming issue. Bob noted there are some structural changes that are needed. Michael suggested to track the suggested changes to EMD as part of our bin list. Opens: - Bob noted at some point he will ping the parser developer for his thoughts on an IBIS-ISS parser. Michael suggested to track this as an a AR to be done later. - Walter stated he wanted to discuss about the signal_type in EMD. The purpose of signal_type is to document whether the signal is an I/O, Power, Ground, or NC. He noted we could rename the EMD Pin List. Bob asked where signal_type is used. Walter replied we need to know if it is an I/O or a rail. Bob suggested Ground could be GND. Interconnect Example Model Review: Justin showed the Interconnect Model examples prepared my Micron. The examples include numerous package models of various types and formats, including IBIS-ISS, Touchstone 1.0, and Touchstone 2.0. The models are intended to show the different ways Micron may use the new Interconnect Model syntax in IBIS 7.0. Justin noted the models include cases of signal only and signals plus power delivery parasitics. In the cases where power delivery is modeled, the on-die decoupling model is included. The examples also include a case where the package connects to an on-die interconnect model, which includes detailed on-die power delivery. Arpad Muranyi asked whether the full package includes the PDN. Justin replied, for the full package model, all signals, package PDN, and on-die decoupling capacitance are included. Randy commented that ideal power rails were included for on-die, while the model package includes the PDN. Randy stated, for the package and on-die model case, they wanted to have models of parasitics between each of the on-die decoupling capacitors. A workaround was used, where the SPICE model has a shorting resistor for many of the signals, because connections are required from pin-to-pad but also pad-to-buffer. Michael asked if this is the fan-out/fan-in case. Arpad noted this is a limitation that all connections must go from pin to buffer terminal. Bob noted the example includes many SPICE and Touchstone model structures; he asked if they are shipping those structures already, though not in IBIS 7.0 format. Justin replied yes, though the on-die PDN model is not common and under discussion, and they may ship these type of models in the future. Bob asked if there are overlapping Groups with the same signals covered in multiple Groups. Justin replied there are several Groups which cover the same signals. Michael pointed out examples that include the same DQ signals, but one with Quasi-static SPICE models and the other with a full-wave SPICE model. Bob suggested more documentation on the differences between the models. Walter commented a spreadsheet listing the nets in one column could be automatically generated. Arpad suggested we could add a description keyword to describe the Groups. Walter asked if the syntax allowed Micron to wrap the models they needed it to. Justin responded it did, with the exception of having to add the shorts to the signals for the on-die interconnect model. Randy commented that this shows how many options there can be. The question is how will the EDA tools present these options to the user. Bob suggested there could be a standard naming convention for the models. Michael asked which model Group would be the default. Randy noted the default would depend on which signal you select. Bob thought we decided there is no such thing as a default for Groups. Mike LaBonte stated you would need to find a model with the signals of interest and as few other pins as possible. If that results in multiple possibilities, you would have to fall back on some default. Michael asked if it will take significant processing time for the EDA tool to figure out the association. Walter replied it is not a compute time issue, as long as you do not read in the IBIS-ISS and Touchstone files. He suggested the default be the models without coupling. Bob asked what source information is used to extract the on-die decoupling capacitance models. Justin replied they do not want to provide transistor-level spice information, thus the model is extracted to a simplified broadband spice subcircuit. Bob commented these models use node 0. Randy noted that the models for the on-die decoupling and package models have a references to node 0. Bob asked if Micron plans to use the Aggressor_Only designator. Justin replied, for the package models they currently provide, they don't see a need for this. Randy noted the models they typically provide are post layout and include all coupling. Michael asked Bob to prepare talking points on the EMD structure discussion [AR]. Touchstone Industry Mail Draft Review: Michael deferred this discussion until next time. Next Meeting: The next meeting will be April 24. Mike moved to adjourn. Randy seconded. The meeting adjourned without objection. EMD Comments to be Resolved: 1. Change top level keywords to distinguish between EMD and Interconnect Models 2. Add Touchstone to the introduction 3. Clarify the meaning of signal_type 4. File format structure IBIS-ISS Parser: - Bob to contact the parser developer for initial thoughts