============================================================================== IBIS INTERCONNECT TASK GROUP Mailing list: ibis-interconnect@xxxxxxxxxxxxx ================================================================================ Attendees from May 28, 2025 Meeting (* means attended at least using audio) ANSYS Curtis Clark, Wei-hsing Huang, Juliano Mologni Arista Networks Jim Antonellis Broadcom James Church 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. The team briefly and informally discussed the state of IBIS and the relatively low attendance at the recent IBIS Summit, and the need for additional IBIS participants and volunteers. Michael reviewed the minutes of the May 7 meeting. Walter Katz moved to approve the minutes; Randy Wolff seconded. The minutes were approved without objection. Michael reviewed updates to the latest draft of the Port Mapping proposal, including a brief list of the coverage, in the provided examples, of the parameters and subparameter in the draft. He noted some typographical errors, and took an AR to create a ninth example to cover the parameters not M - Example 9 [AR]; reviewed examples Arpad Muranyi asked a technical question regarding Examples 1 and 2. Both data sets describe a 4-pin (4-conductor) cable with 8 single-ended ports. Where is the reference? Michael asked, in response, whether a reference is always explicitly required. Arpad wondered how this particular data set would be measured in the lab. Michael reviewed the three node referencing options used in SPICE for Touchstone, adding that the team agreed some months ago to use the SPICE approach, where an unstated reference is ideal node zero. Walter advised being careful about terminology. Every port for measurement, and for documentation of simulation return currents, has a reference node. If the reference is not stated, then it's local to the node being measured. For simulation, that's ideal node zero. Arpad expressed agreement with this, but noted that his recent combined presentation with Weston Beal encountered opposition due to the addition of a physical, floating metal plate to act as a local reference. In the examples, not every connection is shown - where does one connect the shield, for example? Michael replied that the system description in the file data is not implied to be "all there is" (no interaction with any other connections, points or structures not explicitly documented). Walter added that there are always interactions with a reference. Michael asked whether the specification needs to say this. It would be worth adding 2-3 sentences spelling out the referencing assumptions. Arpad asked whether, for this example, the 4-wire system is with respect to a metal bench for measurement. Walter elaborated that one needs to have a return path for TEM wave operation assumed for S-parameters. Arpad asked what local ground means in this context. Walter replied that is means each port has its own local reference. Previous examples of connections between Earth and the moon were cited as showing that the reference could not be universal for some scales of system. Walter replied that, in all cases, currents return in the same path as the signal. All of the local grounds are connected together. As soon as you start talking about AC, you lose the relationship. Ports that mate share the same location for the grounds. Randy suggested that, at least for Example 1, we could add in a statement noting that the reference point is 1 mm away from the measured port. There would be zero difference in the output data. Michael pointed out that loop analysis might be the most helpful way to address this, particularly as the manual method of extracting the information before push-button simulation highlighted how terminations were required as part of data collection. He will find and distribute appropriate references on this. Randy suggested that the proposal could add two twisted pair Ethernet cable instances to Example 1. The team engaged in more discussion of referencing, now using transistor cases (e.g., current flowing from a Vcc rail or signal, and inserting the current into the ground return connection). Walter noted that a mA of current on one side should result in a mA of current on the other side. One would conduct measurements between the pin and a nearby point point. Randy suggested some starting sample language which could be inserted into the draft: "Reference information may be useful for interfacing/mating/cascading S-parameter data blocks to each other or to a circuit described using another method (e.g., SPICE). You may make assumptions..." Michael took an AR to revise this statement for use with Reference. [AR] He also took an AR to invite Pete Pupalaikis to deliver his S-parameter and SPICE interfacing presentation to the Interconnect Task Group. [AR] Walter noted that he has been studying power delivery in much more detail, and now understands the requirements for PD package models and load models (the two parts of SPIM). He suggests the best place to review this is in IBIS-ATM. However, he also advocates integrating PI support into IBIS Interconnect and EMD models, which has an Interconnect Task Group connection. Michael suggested Interconnect could take this up after ATM finalizes the PI approach. Michael asked whether an Example 10 for PI Data_usage should be created. We may need to use EMD or IBIS Interconnect in place of Data_usage "PI". Randy noted that, for Example 4, IBIS Interconnect only allows a single reference per circuit. This makes Example 4 illegal in IBIS. Walter replied that the reference here is only informative. Randy noted that we can make this acceptable in Touchstone, but users will notice an issue when attempting to use the data in IBIS. Walter responded that this issue doesn't affect the result; A_gnd is assumed. You never need a reference for A_gnd or for pin. We should be saying that IBIS is incorrect. Randy suggested that an additional statement would be enough to address this issue. Arpad moved to adjourn the meeting; Randy seconded. The meeting adjourned without objection.