============================================================================== IBIS INTERCONNECT TASK GROUP Mailing list: ibis-interconnect@freelists.org ============================================================================== Attendees from March 25, 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 suggested deferring review of the March 11 minutes until a later meeting with larger attendance. He noted that no ARs were assigned in the March 11 meeting. Michael reviewed the Port Mapping draft 28 document, with a focus on examples. He added that he had updated two areas: the Probe/Physical and A_gnd rules and definitions, plus adding a new, preliminary section of text regarding how referencing works in different contexts. The team focused initially on Example 5. Michael asked why it is problematic. Is not A_gnd is the assumed reference? Arpad Muranyi noted that ports 1 and 2 are for the pin and buffer, respectively; ports 3 and 4 are Vdd at the pin and buffer, respectively. There is no interconnect between Vss and the buffer defined explicitly. Randy Wolff suggested that such an interconnect and reference must exist, as it was measured; we just don't say what it is. Michael noted that the new rules limit A_gnd to Physical and Net. Randy asked, if you have an IBIS buffer model, what do you connect the Vss terminal of the buffer to? Arpad agreed. A_gnd reference is fine for these ports, but it doesn't connect the Vss terminal to Pulldown_reference, for example. Michael noted that tools assume A_gnd connections in IBIS, but there is no explicit specification statement about this. Arpad added that there is also no specification assumption, either. Arpad stated that the example connects Vdd and signal, but not Vss. If we are providing package models, we are connecting pins to buffers. Randy replied that the specification is either leaving it up to the EDA tool or you have to choose one location for reference per rules of IBIS Interconnect. Arpad asked whether one can define Port 4 as the + (pullup_ref for Vdd) with -(Physical Pulldown_ref.7). Michael suggested that this is ambiguous. It's 2N with lots of A_gnd connections. Does this get us into trouble? Randy replied that, unless we update IBIS, there is a problem with 2N syntax. IBIS Interconnect doesn't support a 2N syntax and we should. Example 10 has issues, but at least it doesn't have Data_usage. Michael suggested this might be confusing "law" and "a good idea". No Data_usage means that fewer rules apply. Arpad: what's the timeline for fixing IBIS vs. Touchstone 3; 8.1? Randy: reasons to get TS 3 out soon; good for 8.1 Arpad: Push power integrity to 8.2 Michael: syntax is still good, even without examples 3 options: comment out Data_usage add -(Physical Pulldown_ref.7), keep everything as-is, add -(Physical Pulldown_ref.7) only; need notation regarding IBIS support Arpad: prefers first, as it is a full circuit description Changed Example 5. Randy: "interface" is IBIS Interconnect terminology. Randy suggested copying the same language to Examples 11, 12, 13 et al. for signal connections. He added that Bus_label is the official EMD syntax; we don't have Group outside Touchstone 3. Grouping creates pins, which is what Bus_Label does. It's translatable by machines. Arpad suggested we use Group names for a collection of pins Michael asked whether the team needed to explain Group. Randy replied that we sort of invented Group for the non-IBIS cases. Arpad asked whether the * character is a wildcard or is it just shorthand for "everything". He noted that we have an example with *.2 to represent any RefDes U2.2, U3.2, etc. This is not supported in the EMD specification. Michael replied that we only have Group descriptions for the separate Group lines, not for Group as argument to Physical Type P declarations. Randy replied that Table 51 of IBIS 8.0 shows * only for Pin_Rail, including definitions in Rule b under section 13.6. We don't use Signal Name in Touchstone 3; we only use Bus_label. We are matching *.Bus_label. Arpad noted that Example 11 uses Touchstone 3 wildcards not IBIS wildcards, as IBIS does not support *.2, *.11, etc. Randy replied that we need to think about this problem to support auto-translation. Arpad stated that this is a localized expansion of GND_Ux into all refdes with a given pin number. Michael replied that we don't explain this. Arpad suggested that the rules here are slightly different than EMD in IBIS. Randy suggested that we can write up the automation process. Arpad took the AR to write a statement summarizing that Touchstone and IBIS wildcards behave differently. [AR] Randy further suggested stealing the language from IBIS on what a wildcard does and adding it to Touchstone 3.0. Arpad moved to adjourn the meeting; Randy seconded the motion. The meeting adjourned without objection.