IBIS Macromodel Task Group Meeting date:18 May 2021 Members (asterisk for those attending): Achronix Semiconductor Hansel Dsilva ANSYS: * Curtis Clark * Wei-hsing Huang Cadence Design Systems: * Ambrish Varma Ken Willis Jared James Google: Zhiping Yang Intel: * Michael Mirmak Kinger Cai Alaeddin Aydiner Keysight Technologies: * Fangyi Rao Radek Biernacki Ming Yan Todd Bermensolo * Rui Yang Luminous Computing * David Banas Marvell Steve Parker Micron Technology: * Randy Wolff * Justin Butterfield Missouri S&T Chulsoon Hwang Siemens EDA (Mentor): * Arpad Muranyi SiSoft (Mathworks): * Walter Katz Mike LaBonte Teraspeed Labs: * Bob Ross Zuken USA: Lance Wang The meeting was led by Arpad Muranyi. Curtis Clark took the minutes. -------------------------------------------------------------------------------- Opens: - None. ------------- Review of ARs: - Walter to add a diagram for the "Upstream" flow to BIRD211.2. - Done. - Walter to send out an updated draft of BIRD211.2. - Done. BIRD211.2_draft7 was sent out. -------------------------- Call for patent disclosure: - None. ------------------------- Review of Meeting Minutes: Arpad asked for any comments or corrections to the minutes of the May 11th meeting. Curtis moved to approve the minutes. Randy seconded the motion. There were no objections. ------------- New Discussion: BIRD211.2 draft 7: Arpad noted that the Open Forum had approved a motion to schedule a vote on this BIRD at its next meeting on June 4th. He said he hoped we could finish discussion on any cleanup of the BIRD and get it submitted by Friday, May 21st, which is the deadline if we want to vote on it at the next Open Forum meeting. Walter noted that, based on email discussions on the ATM list since the last meeting, he had added a paragraph explicitly stating that the "Downstream" flow should be the flow for pre-7.1 IBIS models and new models that don't specify Tx_Impulse_Input because the existing 7.0 flow has known problems. Walter said the substantive difference between his position and Bob's was that Walter thought pre-7.1 models and new models that did not specify Tx_Impulse_Input should default to the "Downstream" flow, and Bob thought they should follow the legacy incorrect flow. Bob said the first issue was whether we should document the legacy deprecated flow with a flow chart of the type Walter's BIRD provided for the new flows. Bob said once you document the deprecated flow, then you have the question of how to handle the versioning issue. Bob said he was okay with Walter's new paragraph, but he wanted a block diagram of the legacy deprecated flow. Ambrish said he still thought the BIRD gave the impression that everything was broken in IBIS 7.0. He said we need to make it clear that it's only the Init only statistical flow for Redrivers that has a problem. Fangyi agreed that we might want to preserve the parts that are working in 7.0 and only fix the parts that are broken in the statistical flow. Bob said that was what Walter's BIRD does. Ambrish said we should explicitly state that we are correcting the statistical Redriver flow for Init only models. Fangyi suggested the "Redriver Flow" section be changed to "Redriver Statistical Flow" to make the scope of the change clear. Walter said we could do that, but then we would have large sections of text duplicated in the Statistical and Time Domain flow sections. Walter said he had deliberately combined the Statistical and Time Domain flow descriptions into one since they shared so many steps. Walter said the EDA tool still has to call the Init functions even if it's doing the GetWave flow. Ambrish said an EDA tool doing the GetWave flow may not care about any of the IR processing. Arpad said the EDA tool executes the same Init flow even if it is ultimately doing the GetWave flow and isn't using the modified IRs. Ambrish said Init had become overloaded with initialization and IR processing. He said the models in a channel could have multiple combinations, and if the Tx and Rx both have GetWave then there's no need to use the IRs returned by Init. Fangyi said in IBIS 7.0 the statistical and time domain flows are similar enough to be combined into a single section, but with Walter's new Tx_Impulse_Input parameter and the new flows for statistical the difference becomes too big to combine them into one section with if-else conditionals. Walter reviewed the flow descriptions in IBIS 7.0 (below Figure 41). He said this was the current flow to which models were written. He said he had not changed any of this functionally, with the exception of fixing the problem with the existing Redriver statistical flow (i.e., make sure the Rx2 now gets its entire upstream response). He said he had simply rewritten the existing flows in a different way that allowed him to combine the descriptions. He said anyone who had written an Rx2 model had done it with the assumption that its input would be everything upstream of it, regardless of what the existing incorrect Redriver statistical flow said. He said what Ambrish was suggesting about a tool not being required to follow the Init flow would actually change what's in the specification. Ambrish said he would like it to state that if you're doing the GetWave flow you can disregard the Init flow. He had previously stated that historically the original intent had been to have Init and GetWave flows as independent flows. Walter said the current specification does not say that. Walter suggested that he would send out draft8 after the meeting, and Ambrish could mark it up with suggested changes we could all review. Arpad returned to the original question about the disagreement between Bob and Walter on deprecating the legacy flow. He asked if we all agreed we want to deprecate the existing flow [and have it replaced by the "Downstream" flow]? - Ambrish: Motion to adjourn. - Curtis: Second. - Arpad: Thank you all for joining. AR: Walter to send out BIRD211.2_draft8. ------------- Next meeting: 25 May 2021 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives