============================================================================== IBIS INTERCONNECT TASK GROUP Mailing list: ibis-interconnect@freelists.org ================================================================================ Attendees from December 4, 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 convened the meeting. No patents were declared. During the review of the November 20 minutes, Michael noted a spelling error in the line, "regarding use of a local ground as a refernece". Randy Wolff moved to approve the minutes with this error corrected; Arpad Muranyi seconded. The minutes were approved with this correction without objection. Michael noted that there were no new ARs from the last meeting. Randy added that he is still working on generating EMD/Interconnect package examples. Michael noted that several issues need resolution: ⦁ Complete review of “floating plane” reference examples ⦁ Key issues: do we need a Touchstone port, node, terminal definitions section? ⦁ Key issues: should we include the “floating plane” reference example image in Touchstone?  IBIS? Arpad stated that he had developed a presentation on several issues related to IBIS and Touchstone updates as well as referencing. The presentation described how extraction of RLC SPICE models is handled, then Touchstone files from those, then by contrast how full-wave solvers handle Touchstone extraction. This may be a summit presentation later. Previous material referenced includes Chen and Chitwood IBIS Summit presentations. Complaints focus on power/signal coupling and loop inductance where Touchstone files can be incorrectly connected. The presentation discussed a set of examples generated for illustrative purposes, starting with a 2.5D quasi-static solution to generate a SPICE subcircuit, at 100 MHz with 4 die pads and 4 BGA terminals. The SPICE subcircuit contains small capacitances that connect to node zero, but there is no node zero in the subcircuit call. Most capacitances are mutual. Weston Beal noted that these are also residual and small, to capture all parasitic effecs. Walter Katz asked where these capacitances are physically located. Weston replied that a quasi-static solver calculates capacitances between all structures (nodes). These are residuals to node zero. Walter suggested this was an unnatural approach. Randy replied that all quasi-static solvers do this. Michael suggested that the structure suggests there are no DC paths to node zero; if the capacitances of all substances present are only mutuals, then this may be physical but not simulable under SPICE. Arpad added that the U-element using node zero in earlier examples stems from this structure. Arpad noted that simulation with this model used a simplified IBIS driver/receiver pair. The model was not power-aware but *_ref keywords are present. Two ideal sources and node zero were used for terminations on the right side of the SPICE RLC in full circuit. Waveform plots using node 0 as a reference do not show starvation of driver potential, which did appear in the non-node zero cases by about 19 mV. Arpad suggested Touchstone extraction may be the problem - should the file use an explicit reference node (or port, as in s7p)? The port extraction in the tool assumes ideal node zero for each node (each port is a pair with one node connected to node 0). He added that there are minor resistor adjustments for Vss and Vcc plane terminations in the resulting circuit. Using ideal 0 terminations causes waveforms to converge to identical voltages. His conclusion is that a local Vss reference node is needed for correctness when SPICE is used for simulation. Touchstone is only correct when the number of ports is equal to the number of SPICE terminals. Therefore one must use compatible referencing when extracting and using Touchstone files. A six-port Touchstone file would be extracted and used in a similar way as Walter's expressed preference. Analyzing Vdd shows a reduction in voltage similar to that seen earlier (the buffer is now showing Vss-plane effects). The conclusion is that how a model is extracted does not imply how to use it in simulation. The experiements were repeated with an s8p Touchstone file resulting from full-wave extraction (from Weston's work). Power droop becomes correct when a Vss-referenced node is used. The Vss node is not the same as node 0 in this example. Weston noted he explained this referencing approach during Yuriy's previous meeting appearance, and he appeared to agree with this approach. Weston also created an IBIS s6p with local referencing using a wrapper (EMD and IBIS Interconnect do not support multiple references today). The s6p and s8p cases match closely; the SPICE netlist has bandwidth issue. Michael asked whether the specification needs a "how extracted" keyword. Some distinction may be needed between usage and extraction methods. Arpad replied that performing extraction correctly will result in correct simulation usage. Michael took the AR to forward the presentation and discussion to Yuriy Shlepnev and Scott McMorrow for their comments. [AR] Arpad moved to adjourn; Randy seconded. The meeting adjourned without objection. The next meeting will take place on December 11, 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.