IBIS Macromodel Task Group Meeting date: 27 November 2018 Members (asterisk for those attending): ANSYS: Dan Dvorscak * Curtis Clark Cadence Design Systems: * Ambrish Varma Brad Brim Kumar Keshavan Ken Willis eASIC: David Banas GlobalFoundries: Steve Parker IBM Luis Armenta Trevor Timpane Intel: * Michael Mirmak Keysight Technologies: * Fangyi Rao Radek Biernacki Ming Yan Stephen Slater Maziar Farahmand Mentor, A Siemens Business: John Angulo * Arpad Muranyi Micron Technology: Randy Wolff Justin Butterfield SiSoft: * Walter Katz * Mike LaBonte SPISim: * Wei-hsing Huang Synopsys: Rita Horner Kevin Li Teraspeed Consulting Group: Scott McMorrow Teraspeed Labs: * Bob Ross The meeting was led by Arpad Muranyi. Curtis Clark took the minutes. -------------------------------------------------------------------------------- Opens: - Arpad noted that the only ATM meetings we expect to cancel in the upcoming weeks are those on December 25th and January 1st. - Arpad noted that he still had received no replies to his emails to ATM and si-list regarding AMI redriver flow. He noted that we could discuss whether to remove the topic from the tabled list. ------------- Review of ARs: - Randy to investigate if/why/how a clock waveform input might be used. - In progress. Arpad noted that Randy had said he was unable to attend today, but that he expected to have some slides on this topic when he returns. - Michael M. to investigate if/why/how a clock waveform input might be used. - In progress. Michael M. noted that he also expected to have some feedback in the coming weeks. - Ambrish to check with IP experts on whether DC_Offset is useful for Tx. - Done. Ambrish had an update for today's meeting (below). - Michael M. to check with IP experts on whether DC_Offset is useful for Tx. - In progress. Michael noted that he expected to have presentations on both of his AR topics for the December 11th meeting. -------------------------- Call for patent disclosure: - None. ------------------------- Review of Meeting Minutes: Arpad asked for any comments or corrections to the minutes of the November 20 meeting. Bob moved to approve the minutes. Michael M. seconded the motion. There were no objections. ------------- New Discussion: DC_Offset BIRD Draft: Ambrish reported that his IP experts said they don't use any DC offset info at the Tx. Any equalization is done to the non-transition bits, so there's no messing with the amplitude. Based on this information and the previous week's discussion, the group concluded that a Tx doesn't need this parameter and doesn't need to return a modified value. Therefore, the proposal can revert to Walter's original plan for DC_Offset to be Rx only and Usage In. Walter took the AR to make these changes. Other DDR5 related topics: Michael M. noted that the clock-forwarding topic would be addressed in a couple of weeks when he and Randy completed their ARs. Review of Topic bin list: Michael M. asked if we were assuming DDR5 and improved C_comp modeling were the next most important topics to address. He suggested we review the Topic bin list to see if they are all worth keeping. Arpad noted that the first two items on the list related to differential buffers and the fact that the spec requires single-ended parameters to be provided even though they don't make sense in context. Mike L. asked if we could add the names of the people who had raised the topics. Michael M. said he might have been the one who raised these two, and noted they had been annoyances for some time. Arpad noted the third item was a technical question about power-aware vs. AMI models. He noted AMI requires LTI-ness, and power fluctuations are a time- varying effect. So, perhaps power-aware constructs shouldn't be used directly with AMI. Should the spec. prohibit this or offer any guidance? Ambrish said he didn't think the spec. had to address it at all. It's syntactically allowed to use power-aware keywords in AMI models, and tools can figure it out. Mike L. noted that this topic had come up in multiple presentations at the IBIS Summits in Asia. Mike noted that he could review the 2 or 3 relevant summit presentations at a future ATM meeting. Mike noted one by Ted Mido, "Study on potential feature additions for bit-by-bit simulation technique to address the DDR5 requirements." Mike briefly summarized the method Ted had tried and reported on: Analyze SSO using SPICE to get an "SSO IR", get jitter from that, and apply it to the algorithmic signal. Mike noted that it might be nice to have Ted discuss it in the ATM meeting, but the timing of the ATM meeting was difficult since Ted is in Japan. Mike noted that he would have the summit presentations posted to the IBIS website soon. Arpad noted that the next item was addressing the referencing ("ground") issues. He noted that a lot of cleanup had been done already for 7.0, but some issues remained. Mike L. noted that we had BIRD181.1 (on the tabled list) regarding cleanup of language surrounding I/V tables, and Bob had noted that other parts of the spec. could use similar cleanup. Bob noted that BIRD181.1 contained good things like differentiating ECL vs CMOS referencing, but he noted that it also contained a change to a nodal reference syntax and this could be hard to apply everywhere. Arpad noted the last item on FEC, which Mike L. had raised. Mike noted that that presentations on this topic appeared at previous summits, and that ZTE and Huawei had raised it repeatedly. One proposal was a more generic reporting function (e.g. an AMI_Report() function that generates a set of .html files). AMI models could then use info the model maker embeds in them, for example a complex algorithm that returns an FEC achieved BER gain based on the simulation set-up parameters. Arpad noted that algorithms in the AMI model are already free to do what they want. Mike agreed that models could create their own files, for example, but what was missing is a protocol that allows the model to tell the EDA tool "here's an .html file for you to show the user." Michael M. said model makers want a more flexible, but also standard, way to pass additional output reports to tools. Arpad said that if it were as simple as displaying an .html file then we could address it with a reserved parameter that returns the name of an .html file to display. Walter asked why we needed to do this at all. He noted that the DLL_ID parameter already allows the model to produce unique filenames. Model makers could document the files their models will create, and the users could view them. Why would the EDA tools have to get involved? Arpad noted that this discussion was just to see if all the topics were still valid, and they apparently are. Arpad noted that he would group the three spec-cleanup topics at the top of the list. Tabled topics: Arpad asked if we could remove BIRD166.4 and BIRD190 from the list (items 11 and 12) given the lack of response to his emails regarding redriver flow. Walter suggested that 11, 12, and 13 (Fangyi's alternate proposal) could be removed. Arpad said he preferred to keep 11 and 12 (actual BIRDs) separate from 13 (a draft proposal). Ambrish said he felt strongly about BIRD190, but was willing to remove it if BIRD166.4 were removed. Bob said the BIRDs should simply remain tabled and be kept on this list until they were voted up or down. Fangyi noted that in the current single-ended (DDR5) discussions some of the same ideas in his alternate proposal would be relevant. He said item 13 could be revisited later after the DDR5 discussions conclude. Ultimately, the group decided none of the items should be removed from the tabled topics list. - Mike L.: Motion to adjourn. - Walter: Second. - Arpad: Thank you all for joining. AR: Walter to update the DC_Offset BIRD draft. ------------- Next meeting: 04 December 2018 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives