IBIS Macromodel Task Group Meeting date:25 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 send out BIRD211.2_draft8. - Done. -------------------------- Call for patent disclosure: - None. ------------------------- Review of Meeting Minutes: Arpad asked for any comments or corrections to the minutes of the May 18th meeting. Walter moved to approve the minutes. Randy seconded the motion. There were no objections. ------------- New Discussion: BIRD211.2 draft 8: Arpad noted that several emails had been exchanged on the ATM meeting list since the last meeting. Walter said that Bob had proposed a substantive change to the BIRD. With Bob's proposed language, the new Tx_Impulse_Input parameter would only apply to Redriver Txs. The parameter would not apply to all Tx models. Arpad asked why Walter wanted to apply it to all Tx models. Radek said it had been in response to a comment he had made. He said Walter's initial proposal introduced the new parameter for Redriver Txs only. Radek had stated that this would create two different types of Tx model, so Walter had made it applicable to all Tx models instead. The new parameter is needed for Redrivers, but the Redriver Tx model need not be different from other Tx models. With regard to Bob's proposal, Walter asked, "Why not allows this for all Txs?" Bob said that he had kept it open to apply to Retimers too (i.e., all repeaters) because the BIRD modifies the Retimer flow too, even though the new flow figures (block diagrams) are only in the Redriver section. Bob said he thought the four value choices for the new parameter were moot if the Tx isn't part of a repeater. Bob said his proposal is that for version 7.1 and above, if Tx_Impulse_Input is declared then the Repeater_Type is also required. He did not want to carry over an implied legacy flow (the corrected flow corresponding to the "Downstream" setting of the new parameter), and he wanted the model maker to explicitly declare what the setting/flow is for models beyond 7.1. Curtis noted that Bob's proposed language also included coupling to the [Repeater Pin] IBIS keyword. He said that this language had also been used in the IBIS 7.0 specification, but it wasn't correct to require an AMI model parameter, in this case Tx_Impulse_Input, if the [Component] level keyword [Repeater Pin] were used. There could be AMI models that had nothing to do with any of the [Repeater Pin] entries. Arpad said his understand was that we are just trying to fix the flow for Redrivers. For Retimers the channels are independent, so why are we touching the Retimer flow? Walter said he had re-written the normal (single-channel), Redriver, and Retimer flows. He had rewritten them to allow the application of this new parameter to any Tx in any of the flows. Walter said to keep it general, all Txs could specify any of the new parameter's four values to specify their input. Arpad asked if we would now be adding requirements on terminal models that don't need Redriver related changes. Walter said we are not. He said the parameter is optional, and the default behavior is the normal legacy behavior. Fangyi summarized: for a terminal (initial) Tx, the "Downstream" and "Combined" options would be the same (legacy behavior), "Separate" would be a new option, and "Upstream" would mean the unit impulse response is the input to the Tx. Bob said he thought the current BIRD's changes were confusing. Ambrish said that if we think it's confusing, how's it going to be for everyone else? Ambrish said he thought we should just fix the Redriver statistical flow, which is the only thing that needs fixing. He said this proposal is changing a lot of text, and he asked if we could restrict it to just fixing the Redriver statistical flow. Fangyi said he thought this was possible. Walter said that if someone wants to make a motion that says that Tx_Impulse_Input is only allowed when the Tx is part of a repeater, then they should do it so we can vote on it. Fangyi said if we agree on Ambrish's proposal to restrict changes to the repeater flows, then Bob's proposal was one way to accomplish this. Ambrish agreed. Ambrish moved to restrict the changes to the repeater flows. Bob seconded. Curtis commented that he was reluctant to give up on Radek/Walter's idea of keeping the parameter generally applicable to all Txs, which meant it would affect other flows. Walter objected to the motion and asked for a roll call vote. The roll call vote tally was: Ansys - No Cadence - Yes Keysight - Abstain Micron - Abstain Siemens - Abstain SiSoft - No Teraspeed - Yes The roll call vote concluded with a vote tally of Yes – 2, No – 2, Abstain – 3. The vote did not pass. Discussion continued. Walter said he had rewritten the text for the normal reference flow as well as the repeater flows. However, IBIS 7.0 only has diagrams of the repeater flows, so he only created diagrams for the repeater flows given the values of the new parameter. He said it would be helpful if he added flow diagrams for the normal reference flow. Randy asked Walter to also add a statement referencing "Downstream" to the normal flow. - Randy: Motion to adjourn. - Curtis: Second. - Arpad: Thank you all for joining. AR: Walter to add diagrams for the normal reference flow and send out BIRD211.2_draft9. ------------- Next meeting: 01 June 2021 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives