============================================================================== IBIS INTERCONNECT TASK GROUP Mailing list: ibis-interconnect@freelists.org ============================================================================== Attendees from July 23, 2025 Meeting (* means attended at least using audio) ANSYS Curtis Clark, Wei-hsing Huang, Juliano Mologni Arista Networks Jim Antonellis Broadcom James Church Chipletz Stephen Newberry Intel Corp. Michael Mirmak*, Xiaoning Ye Keysight Technologies Ming Yan Marvell Steve Parker MathWorks Walter Katz* Micron Technology Justin Butterfield Siemens EDA Weston Beal*, Arpad Muranyi*, Randy Wolff* Simberian Yuriy Shlepnev ST Microelectronics Aurora Sanna Synopsys Ted Mido, Edna Moreno University of Illinois Jose Schutt-Aine Michael Mirmak convened the meeting. No patents were declared. Michael noted that the June 25 minutes are almost finished, and that he will try to send them out for review after this meeting. Michael reviewed the minutes of the July 9 meeting. He noted an extra period followed the word "etc." (as in "etc.."). Walter Katz moved to approve the minutes with an appropriate correction; Randy Wolff seconded. The minutes were approved without objection. Michael noted that, aside from making revisions to Draft 15 on the text explaining Physical and Logical concepts, all other ARs have been closed. Before reviewing Draft 16, Arpad Muranyi noted he was involved in an ongoing conversation with Walter and Benjamin Dannan regarding SPIM. Ben also provided presentation slides from DesignCon 2022 and the Signal Integrity Journal, in which loop vs. partial models were discussed. Arpad is asking Ben detailed questions and will report back any findings relevant to Interconnect. Arpad noted that Ben is attending SPIM discussions in IBIS-ATM on Tuesdays. Walter suggested bringing this up there as well. Michael mentioned referencing as part of SPIM as well. Kinger Cai's proposal uses - and + terminals to connect to GND and POWER. Arpad believes this to be a loop case; there, we combine power and ground into one effect. This could be a reason why there is still controversy, as using a loop removes the details between power and ground separately. Michael summarized the changes he made to the Port Mapping proposal in Draft 16. Arpad asked about the use of refdes vs. file names; we still cross-reference refdes assignments in filenames. Randy noted that page 8 of the proposal is not an exhaustive list of file extensions or file formats; one could also refer to wrappers, netlists, and the like. Weston Beal suggested that we need a generic term, instead of referring to extensions. Michael asked whether we have the right concept when we state that Logical assignment doesn't determine connectivity (only Physical and possibly Net do). Walter suggested that the sentence beginning "Note that connectivity..." isn't needed. Arpad asked whether this rule only sometimes applies to Physical? Randy countered that Physical is *for* connectivity. Walter noted that Physical defines where to connect *to me* (as the connected device), not where to *connect to*. Randy agreed, stating that connections are created at a higher level in the component definitions. Walter added that all the ports I am exposing are what I'm showing for connectivity purposes. Michael asked whether this is like a subcircuit definition or function definition, where position and not name is what matters. Arpad stated that the port declaration simply identifies that it exists, not how to connect it. On the use of "Net" to help define connectivity, Walter agreed with the Draft 16 revisions. In addition, "Side" is a guide to how to simulate this complex structure. Michael asked whether one is allowed to use all name/value pairs in the same Port. Arpad noted that, for Diff_port, the data is not to be in mixed-mode format. For Example 6, Michael suggested it is either inconsistent example OR referring to a DIMM or a similar subsystem. Walter replied that it's definitely for a DIMM. Arpad asked how a Touchstone file maker would know what's outside of it. Walter answered that, if he is the creator of the Touchstone file, or measured it, he would know. Michael also noted that Example 6 has no Data_usage assignment, but 7 and 8 do. Arpad asked what the difference is between Example 1 and Example 6. He added that, in Example 7, the first six ports are die names while the second set are pin names. Walter suggested that, in the referencing EMD file, there's a subcircuit with the appropriate net names. Michael asked whether the connectivity to EMD is by name. Walter replied that the EMD model calls out this Touchstone file, after the Terminals line, and each Terminal will have the appropriate pin number. EMD and this list in the Touchstone file will be redundant, and the simulation tool will verify that no errors are present between them. Arpad: everything in EMD is a cross-reference to connections in Touchstone; it's Data_usage that indicates what we want. Arpad added that IBIS defines differential differently than what is in Draft 16 of Port Mapping; do we still need a differential identification? Walter suggested that, if the port is already connected, we don't need it. A differential port requires that we still have this indication before it is connected. In Touchstone, the Diff_port is only and explicitly for waveform display. Arpad asked whether the Data_usage section should mention the connectivity assumption? Michael to clarify this text in relation to IBIS. [AR] Arpad noted that the Touchstone file connectivity for Ts4file in Example 8 is ambiguous. Walter replied that this data declares how measurement was done; the AMI file declares how it will be used. Michael will check whether the Ts4file assignments in IBIS can be clarified relative to this draft. [AR] Randy asked what A and B notations mean for on-die buffer connections. Arpad showed the port numbering used in IBIS 7.2. Walter noted that we are talking about the pad side and the buffer side (pad high, pad low, buffer high, buffer low); we can give these terminals names in Ts4files. Michael asked whether we need an assumed set of names in IBIS; Walter suggested using simply 1,2,3, and 4 for Ts4files. Arpad stated that we need a rule for 1,2,3,4 assignments. Arpad moved to adjourn; Walter seconded. The meeting adjourned without objection.