============================================================================== IBIS INTERCONNECT TASK GROUP Mailing list: ibis-interconnect@freelists.org ============================================================================== Attendees from March 4, 2026 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 called the meeting to order. No patents were declared. Michael noted that the minutes of the February 18 minutes were still being processed and would be distributed for review at the next meeting. Both ARs from the previous meeting are still incomplete: At the bottom of page 8 of the draft, some names don't match what's actually in IBIS Interconnect in 8.0 (specfically for buffer_IO). See table 49 in IBIS 8.0 for details. Michael took the AR to reconcile these. [AR] Arpad Muranyi took an AR to provide a corrected version of the IBIS-ISS connections shown in Port Mapping Example 5. [AR] Walter Katz raised an open: MathWorks is preparing for Touchstone 3.0. They would like to perform initial sanity checks on Touchstone data, particularly in light of IEEE 370 features. What checks should be implemented for pole-residue? Should a parser run IEEE 370 metrics there as well? Should passivity be checked? What kinds of quality reports can we generate? What gets included in TSCHK? Michael asked about quality checks and parser message spreadsheet; is there a column or category for "quality"? Walter noted that non-monotonicity checks are already included. He added that active vs. passive behavior may be a useful indication. Michael mentioned Pete Pupalaikis and his presentation material from DesignCon. Walter suggested inviting him to IBIS ATM to maximize attention; Randy Wolff agreed, adding that the IBIS Open Forum teleconference would also be suitable. Walter advised the team "make a big deal announcement" and specify the time. Michael reviewed the latest Port Mapping draft, specifically Example 5. He noted a concern about the use of A_gnd. Walter proposed a thought experiment: an EDA tool that imports the named S4P; for each of these ports, there is a specific point (pin, buffer, etc.) where we can put a physical probe or a simulation probe. A_gnd is the reference for these ports. Arpad replied that there is a distinction between how an S4P would be measured and how it would be used. Walter replied that there is a third option: assume there is a nearby ground. Randy asked how we simulate this? N, 2N, 2N+1 connection schemes are supported in SPICE. Arpad noted that, if there is no connection point on a "Port" for reference, then the tool can choose in a way other than intended by the model-maker. Michael asked whether we can change the statement to point to measurement vs. simulation that would make it true. Arpad replied that he thought most everything in the document is intended for connectivity, not for measurement. Is this right? If so, we can have an initial blanket statement about this early in the document. Walter disagreed; the specification documents how the measurement was made. Arpad replied that the reason the document was started was for connectivity; Michael and Randy agreed. Walter agreed that this was one reason, true, but the document is really about measurements or how data was made, and this determines how it is used. Michael replied that connectivity and measurement are separate. See IBIS-ISS 2n terminal statements, etc. Walter added that the Probe parameter notes how the measurement was made to that reference designator pin number. Randy noted that, in the X:Y:Z notation, you can infer connectivity but you don't want to use it for that. For Example 5, Arpad noted that no explicit reference parameter is included. Randy stated that all information about referencing is optional; it can be useful for connecting two circuits together, but may not be absolutely necessary. Michael replied that, even with VSS being used in the examples, we still have an issue with A_gnd. Arpad clarified that we are no longer talking about the Example 5 red comment text. We are instead asking about the "ground" (reference) connection. There is no connection between Vss and A_gnd there. Walter replied that, by definition, all ports of an S-parameter are to its local ground; everything has its local ground and measurements between two local ground terminals or nodes are meaningless. The only reason for ideal node 0 to be used is to account for excess electrons through a return path. Michael stated that there could be no progress on the Port Mapping document without resolution of this problem. He noted that, on p. 10, text apppears starting with "How referencing is handled when the Touchstone data is used in simulation", noting that the referencing is handled by the EDA tool. Arpad replied that our A_gnd rule conflicts the with first sentence of this section. Walter replied that, if you don't have - symbol, the user must assume a nearby point is used as the reference and the tool will behave accordingly. Randy stated that there are many examples in the draft, starting with Example 1, which will need scrubbing if we eliminate the existing A_gnd rule. Michael asked whether, if we remove A_gnd entirely, we will have connectivity issues? Tool consistency issues? Will this cause later problems with PIM? Walter answered yes. But both approaches allow precise connectivity (if you care). Is there any definition of local ground or local reference in the document? Michael replied no, not by that name. Arpad added that ideal node 0 in SPICE is not local ground. Walter stated that all measurements and waveforms are going to be relative to local ground. Michael took an AR to send to the reflector a short, updated statement regarding replacing A_gnd. [AR] Walter summarized CoPilot findings regarding local reference definitions. Arpad moved to adjourn the meeting; Walter seconded. The meeting adjourned without objection.