============================================================================== IBIS INTERCONNECT TASK GROUP Mailing list: ibis-interconnect@freelists.org ============================================================================== Attendees from September 23, 2026 Meeting (* means attended at least using audio) Amphenol Graham Kus* Cadence Kyle Lake* Chipletz Stephen Newberry Intel Corp. Michael Mirmak* MathWorks Walter Katz Siemens EDA Weston Beal, Arpad Muranyi*, Randy Wolff* Synopsys Ted Mido, Edna Moreno ============================================================================== Michael Mirmak called the meeting to order. No patents were declared. Michael reviewed the September 9, 2026 meeting minutes. Arpad Muranyi moved to approve; Kyle Lake seconded. The minutes were approved without objection. Michael noted that he would not be available starting on October 7, for two weeks. Arpad stated his desire to have a "clean" document ready for review by that date. Graham Kus noted that he would be unavailable for the September 30 meeting. Michael showed a highlighted, minimalist example as part of a brief presentation, illustrating basic EMD and proposed Touchstone 3.0 rules. Arpad noted that, if you look at the EMD pin list (either variant supported), the structure is as follows: Pin_name Signal_name Model_type Bus_label Each pin has to have a signal name but Bus_label is optional. The rules we are defining may be summarized for the example as: - for Port 2, "Net VSS" associates a Signal_name with the Pin - Bus_label splits up a group into smaller chunks; when we have Bus_label, Signal_name is still required, and provided by the Net name/value pair. Port 1 already has the Signal_name in the Physical construct; but Arpad would suggest the name associated with Net should match (but still be optional). Randy Wolff noted that Net is required for the statements and structures identified in the draft with blue highlighting; this should be true for the green-highlighted items as well. Arpad noted that the Port 2 instance was associated with EMD.G1; imagine another pin EMD.G2 having an EMD:VSS reference. Randy replied that we don't need to have a Signal_name declaration for EMD:VSS. Arpad noted that we cannot refer on a Port line to a Pin name, then overlap with a reference to the same signal. Michael asked whether the specification should specifically include a redundancy prohibition. Randy replied that it would be better to declare it only once; alternatives can complicate the parser. Michael asked whether the following statement was still true: "Note that the values that appear on a given Signal_name line shall all match". The team agreed this was true. Randy added that the only exception is the use of a wildcard to short all designators. Arpad raised a concern, in that Signal_name: can use a wildcard, but this is a "hammer approach" where everything is shorted. Michael asked whether the next statement was also true: "Note that the values that appear on a given Bus_label line shall all match." The team agreed this is true; Arpad added that, in this case, everything is shorted. Michael asked whether the structure "Bus_label EMD:" is not allowed. The team confirmed that this is allowed, but does not support wildcards. Further, Net is required to identify the association with a Signal_name. Arpad added that there is an error in the IBIS p. 421 example, in that there are two unused Bus_labels. These should be removed, or additional structures are needed to make use of them. Arpad raised concerns about the use of the phrase, "ideal simulator terminal". Michael noted that this was a euphemism for SPICE ideal node zero. Arpad suggested that "global simulator reference" could be used, though references can be different across different tools. "Simulator global reference node" is another possibility, as "terminal" should be avoided. Graham suggested that the term could be "junction", per Kirchoff. The team discussed avoiding "node". Graham asked about ECL support. While this was a past concern in IBIS, questions about ECL have not emerged in some time. Arpad asked whether the example using Symbol_left, etc. should have a fourth terminal. Randy replied that, if the structure uses 2N referencing, you would put the refernece next to the terminal; if Port 1 is on the left, you get 2 terminals on the left. The parameter is purely defining how the image of the symbol is arranged. Arpad noted that the statements around "collections of power/ground Port terminals" can be simplified to "pins" in most, if not all, cases in the draft. Arpad moved to adjourn; Randy seconded. The meeting adjourned without objection.