=============================================================================== IBIS INTERCONNECT TASK GROUP Mailing list: ibis-interconnect@freelists.org ================================================================================ Attendees from November 6, 2024 Meeting (* means attended at least using audio) ANSYS Curtis Clark*, 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 Zuken USA Lance Wang Michael Mirmak called the meeting to order. No patents were declared. During review of the minutes of the October 30 meeting, Arpad Muranyi noted that the spelling of Yuriy Shlepnev's name was incorrectly spelled. Arpad move to approve the minutes with this change. Curtis Clark seconded. The minutes were approved without objection. The team reviewed referencing issues with IBIS examples using Touchstone. Arpad showed a February 2018 IBIS summit presentation from Vladimir Dmitriev-Zdorov with the phrase "port regularity" used to describe the requirement that current in to a node within a port must equal current out from the other node in the same port. Michael noted that Yuriy said the same thing but not using that name during the October 30 discussion. Arpad noted that the IBIS File_TS reference terminal rules on page 347 of IBIS 7.2 appear to be missing a needed reference. In the examples from the Summit presentation, arrows symbolize ideal node 0. Arpad noted that some examples show an n=16, four-element structure. Walter Katz noted that there is always a buffer terminal for GND; for [Voltage Range], what node is that? The team discussed several examples Arpad assembled, particularly using *_ref keywords. Walter asked, if Pullup Reference is 2 V and Pulldown Reference is 0 V, is this equivalent to Pullup Reference of 2 V only? Arpad replied this was not exactly the same. The GND terminal of buffer may not be node zero. Walter asked how you measure the voltage at that point. Arpad replied that we are assuming ideal node zero is NOT connected to buffer GND; Walter noted that he is assuming ideal node zero IS connected to buffer GND. Arpad explained his concerns with Example 3 on page 360 of IBIS. IBIS only supports a single reference terminal. The board doesn't have a way to connect to an on-die reference terminal. Unless you assume an A_gnd on the board side, you violate the reference concept. Randy Wolff noted that this example should use A_gnd and not Pulldown_ref. The IBIS specification is therefore wrong. Walter agreed, adding that Pulldown_ref does not necessarily have to be A_gnd. The use of the *_ref keywords distinguishes between Power Integrity and Signal Integrity analyses. Arpad suggested several potential fixes to the situation, including allowing per-interface referencing. In this case, more than just one S-parameter interface (set of ports or side) would be supported. You could have two different references, one per each side. He added that IBIS doesn't support direct shorts between pads pins, etc. Michael asked whether several of the examples are equivalent to Arpad's conceptual test case of an ideal transformer with no DC connection to SPICE ground. Walter noted that all EDA vendors for 40 years have been doing things the same way. Arpad advocated that the team needs to fix this in IBIS 8.0 and that this issue will likely result in a BIRD. Michael asked whether the long interconnect problem with different (non-shorted) ground connections could be solved by imagining the limit of a set of infinitesimal slices - like a wave, the signal has a return path that can be assumed identical in properties from one slice to the next. Walter replied that measurement of VCC rails and associated signals includes the effects of the return path, so that's why you can throw away the effects of the return path. These are ground-based models. That's why we do this for signal integrity and not power integrity. Michael asked about the distinction between this concept and "ground-reduced". Walter noted that S-parameter information includes ground return effects in the data. Arpad stated that one needs to have S-parameters containing power and ground if you want to model all these effects. Walter asked whether Arpad's example would be described with an S4P or an S6P file. Arpad suggested one would ignore the top Vcc line and use an S4P, to which Randy agreed. Randy added that the structures being discussed were currently allowed in the IBIS examples. Walter advocated removing these from the specification. He added that any other reference nodes in the examples may be connected by EDA tools to A_gnd. Randy suggested that a special rule for S-parameters may be needed for IBIS, as a bedspring model, for instance, may have more explicit connections. Arpad asked about the proper connections to IBIS-ISS circuits. Walter replied that IBIS allows you to create a subcircuit with Vss terminals on it; you don't know how that subcircuit will be connected. This allowed people to write subcircuits with no reference at all - a silliness condition. If you are doing that, and putting in ports like Vss (a synonym for ground here) they should be connected to node 0. Direct shorts are also not supported, according to Arpad. Walter replied that these shorts are meaningless. Randy noted that an EDA tool cannot leave these floating. Arpad asked whether A_gnd should be the only reference in the examples. Or should per interface referencing be permitted? Michael additionally asked whether the team wanted to lift the shorting prohibition. Randy suggested that, as a solution for IBIS, adding assumption statements regarding referencing and removing the incorrect examples may work. Walter concurred. Arpad took the AR to send out his slides. [AR] Arpad moved to adjourn; Randy seconded. The meeting adjourned without objection. The next meeting will be held November 13, 2024. ================================================================================ Bin List: 1) Complete port naming proposal (Katz et al) 2) Complete/revise Touchstone 3.0 draft outline (Mirmak) 3) Complete ISS-IRD 1 Draft - enable cascading of S-parameters through W-element (Mirmak) - TABLED Tabled ARs: - Arpad to give an example of the physical connectivity needed for EMD automation.