============================================================================== IBIS INTERCONNECT TASK GROUP Mailing list: ibis-interconnect@freelists.org ============================================================================== Attendees from June 25, 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 reviewed the June 11 minutes. Arpad Muranyi moved to approve the minutes; Randy Wolff seconded. The minutes were approved without objection. Michael reviewed the June 18 minutes. Arpad had noted that the wrong date was listed. He also highlighted two other issues: Change the last “t” to “d” in: “which port connects to which pair of notes” Remove the last “d” in: “how does this related” Arpad moved to approve the minutes with the described changes; Randy seconded. The minutes were approved without objection. Michael noted that all outstanding ARs have been closed. Michael summarized the current usage and definitions for units in Touchstone, including the permitted notation. He found that no distance units were defined anywhere in the existing Touchstone specification. In addition, scientific notation with an arbitrary number of decimal places is permitted. Arpad reiterated that his concern was over the inclusion of the angstrom (0.1 nm) as a unit. Unlike other units, the exponent of 10 in angstrom does not use 3, 6 or 9. Walter Katz suggested that nanometer is used commonly today; is this sufficient? Microns, mils are the only additional units needed. Arpad reviewed his examples, including an s6p with pin G14 on port 1's CA0 signal Port. This arrangement was also true for his original HL example S169P with R and L called out, using ground reference grouping, with power identification. U1 is the die name, which is appropriate for multi-chip memory devices. Michael asked about the usage (package); can't do this with IBIS Interconnect, so one would do this usually with EMD. He added that there is one and only one die and one and only one ball map to which this is associated. Arpad and Walter disagreed, noting that this is a multi-die case. Michael replied everyone knows in these cases how things are going to be connected, physically and logically. But connecting "in the blind" is the other case not yet covered, where connector functions and connectivity are unknown to the model maker. Walter defined the s6p connectivity for the example: Physical: BGA pin G14, DIE1 pin 31, DIE3 pin 31 Logical: not clear what these would be; CAD nets are CA0 and CA1 Randy stated that maybe something general should be used, such as "command-address"; that's a logical, generic name. Arpad noted some inconsistency or vagueness in the definition of "Logical". Randy agreed that "Logical" needs to be clarified. Arpad added that we say more about Physical in the text, but there are also a few issues there. Walter replied that Physical is well-defined; Logical is what shows up on the schematic symbol. For an S3P, the Logical names would be "controller", "memory 1", and "memory 3". Michael stated that "Logical" is very useful, but should it be used for connectivity? Walter answered that the Logical declaration is meaningful for connections. Arpad asked whether the team changed the transistor example to 2-port. Michael asked whether Logical assignments of emitter, emitter, and emitter could be used for all three terminals. Walter answered that one "can't make the horse drink", in that the format does not force the model-maker to use it correctly. Arpad noted that the document also has problems with Physical definitions; isn't A.1 a logical identifier? Walter replied that A.1 is a physical hole in the component. Walter added that Data_usage declares the usage. Arpad suggested splitting up Physical into two different names. Walter replied that Data_usage rules for specific cases are already defined, so such a split is not needed. Arpad answered that we say how a component is measured in the Physical text, but not how it is to be used. Michael asked what would happen if no Data_usage assignment were present. Michael took the AR to clean up the text for Physical and Logical definitions. [AR] Arpad asked whether we should also use negative terminal language, as in the original HL proposal. Walter suggested alternatives, including Reference_Logical or (Reference (Logical Emitter)). For the Example 3 case, the team discussed how many SnP ports should exist; 2 was agreed. In terms of how many terminals should be present, there were several answers, ranging from 2 to 4. Walter suggested that S1 would be the emitter for the BJT case. Michael noted that Example 2 needs a and b variants to cover different referencing. Arpad asked whether all models would be forced to use ideal node zero. Walter answered no, absolutely not but, by contrast, how many transistors are seen in memory DQ decks? Randy forwarded a video from YouTube summarizing S-parameter extraction for BJTs. Arpad moved to adjourn; Randy seconded. The meeting adjourned without objection.