IBIS Macromodel Task Group Meeting date:29 June 2021 Members (asterisk for those attending): Achronix Semiconductor: Hansel Dsilva Amazon: John Yan 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: - Bob to work with the Quality task group on a BIRD to address the AMI_Impulse existence requirement and the Init_Returns_Impulse set to True requirement for statistical BCI mode. - Done. Bob reported that the Quality task group had decided they would prefer to pursue a simple editorial statement instead of a BIRD. They plan to discuss the issue at the upcoming Editorial task group meeting, and Bob expected they would incorporate suggestions captured in the previous week's ATM minutes. - Fangyi to create a new draft of BIRD211.3 with the changes discussed in the last meeting and some additional comments from Arpad via email. - Done. Fangyi said he had not sent anything out after the last meeting, but he had preserved a version with the changes discussed during the meeting's review of Ambrish's comments. Fangyi noted that some of Arpad's comments were orthogonal to Ambrish's. So, he preferred to go back to BIRD211.3 draft 1 and begin reviewing Arpad's comments. Fangyi said he preferred to discuss Arpad's comments live during the meeting and get everyone's feedback. We can later merge the versions containing Ambrish's comments and Arpad's comments as necessary. -------------------------- Call for patent disclosure: - None. ------------------------- Review of Meeting Minutes: Arpad asked for any comments or corrections to the minutes of the June 22nd meeting. Walter moved to approve the minutes. Randy seconded the motion. There were no objections. ------------- New Discussion: BIRD211.3 draft: Fangyi shared BIRD211.3 draft 1, and the group reviewed the email containing Arpad's comments. 1. In the 4 flow diagrams shown in the Definition of the Issue section, Arpad noted that first one ("Downstream" case) uses a two-input box representing the convolution step. He said for consistency the second and fourth flows should also use this convention. The second flow ("Combined" case) should show a two-input convolution box with the Output of Rx1 Init and the CH-2 IR as the two inputs. Similarly, the fourth flow ("Upstream" case) should show a two-input convolution box with the output of Tx2 Init and the CH-2 IR as the two inputs. Arpad noted that this section was just the intro and not what would go into the specification, but the issue also existed in parts of the BIRD that would go into the specification. Fangyi added a note to address all the relevant locations. 2. On page 6, Arpad suggested that the block diagram illustrating the physical system should probably remove "IR" from the "Channel IR" box. The box in this case represents the physical channel as opposed to the channel's IR in a flow diagram. He said we need to be careful to avoid confusion about the channel boxes. This is another reason to add the explicit convolution boxes, because currently channel boxes sometimes include an implicit convolution. Arpad also suggested grammatical changes from singular to plural because the section defines multiple flows not a single "reference simulation flow". Fangyi made the grammatical changes and added a note for the "IR" change. Arpad said he was neutral on whether the "Latch" location should stay in the physical system diagram. 3. In the section detailing the single-channel (non-repeater) flow, Arpad said the sentence about the "Downstream" setting indicating that the Tx AMI_Init function does not have the ability to adapt based on the upstream channel was unnecessary. The terminal Tx doesn't really have an upstream channel. Fangyi and Arpad simplified the language and removed the "Downstream" sentence from this section. The "Upstream" sentence will now simply state that the value "Upstream" means the Tx AMI_Init does not adapt based on its input IR. 4. Arpad had concerns about the sentence: "After completing all steps of executing Tx and Rx AMI_Init functions, the EDA tool may use results generated in these steps to perform statistical or time domain simulation." Arpad said he thought executing Tx and Rx AMI_Init functions was the "simulation", and post processing of the outputs was the analysis or post processing. Fangyi said we might have to unify the terminology. The group discussed various thoughts about what constitutes simulation, analysis, or post processing. Curtis noted that the text "simulation/analysis" is used in several places. To address Arpad's issue with the sentence, Fangyi added "the rest of" after perform. Fangyi changed "simulation/analysis" in Step 4 on page 7 to just "simulation." (Note: There may be other instances of this to change). 5. On page 8, in step b. describing what the EDA tool should do if it adds a fictitious unit impulse response crosstalk column, the text "value of argument aggressor" was changed to "value of the aggressor argument". 6. Arpad thought that the step c. sentence was confusing: "Models that use the crosstalk columns of the impulse_matrix to determine the equalization should ignore when determining the model's equalization any column that contains a "unit impulse response" He said when he first read it, he had been unsure whether it was really the model or the EDA tool because of the phrase "determine the equalization", which sounds like what the EDA tool may do by passing in a unit IR as a crosstalk column. Fangyi and Arpad rewrote the sentence. They used the term "optimize" instead and moved "when determining the model's equalization" to the end of the sentence. Further review to continue in next week's meeting. - Curtis: Motion to adjourn. - Randy: Second. - Arpad: Thank you all for joining. ------------- Next meeting: 06 July 2021 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives