IBIS Macromodel Task Group Meeting date: 04 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 create a BIRD211.2_draft2 with today's changes and send it to ATM. - Done. Walter distributed a draft3 prior to this meeting. -------------------------- Call for patent disclosure: - None. ------------------------- Review of Meeting Minutes: Arpad asked for any comments or corrections to the minutes of the April 27th meeting. Walter moved to approve the minutes. Randy seconded the motion. There were no objections. ------------- New Discussion: PAMn BIRD: Arpad asked Walter to comment on the PAMn proposal he had recently sent to the group. Walter said a new USB standard using PAM3 is coming in the not too distant future. Designers are starting to work on these devices, and end users may want to simulate by the end of the year. Walter said he had considered a straightforward PAM3 BIRD, which could be analogous to the earlier PAM4 BIRD and its PAM4 specific keywords. As a more general approach, he had drafted a PAMn proposal that wouldn't require a new BIRD for each PAM5, PAM6... Walter said rather than PAM4 specific keywords, for example, the PAMn BIRD uses parameters of format Table to define the thresholds, sampling offsets, etc., for the particular PAMn being used. Tables are also used to specify the mapping from binary data into PAMn symbols. Walter noted that the mapping from bits to symbols is more complicated for PAM3, PAM5, etc., than it is for PAM4. David asked if PAM3 is another name for duobinary. Fangyi said duobinary is one of many flavors of PAM3, and it is entirely different from what USB will use. David said duobinary relies on bandwidth shaping of the channel to define the pulses shape. Fangyi said there are two styles of duobinary: delay and add logic, or filtering. He said the impulse response of the channel is always available. PAMn is usually referring to the transmitter. Fangyi said the PAMn proposal is quite general, but duobinary would be one of the outliers that can't fit into the proposal's scheme. He said duobinary will need a special case or possibly a separate BIRD. Walter said the goal is to consider this proposal and either get it into 7.1 or at least have a BIRD approved for 7.2 by the end of year so tools can start implementing it. Arpad asked people to read the proposal and start considering it. Arpad noted that the proposal still hadn't been uploaded to the ATM work archives, and he asked Walter to submit it to the Open Forum so it will be posted on the website for people to see. Michael and Arpad noted that the Editorial task group has completed work on all approved BIRDs for 7.1 and is down to editorial tweaks. He said the more we wait for large scale technical changes like BIRD211 or the PAMn proposal, the more we delay 7.1. Michael said the latest draft of 7.1 is available in the wip directory on the IBIS site thanks to Bob. BIRD211.2: Walter reviewed the latest changes in draft3. He said Bob had sent out a list of good comments, and he broadly characterized them as editorial or organizational. In response, Walter had changed two of the values of the new Tx_Impulse_Input: "IBIS7.0" was changed to "Downstream" - this is the legacy input to the Tx. "DoNotCare" was changed to "Upstream" - specifying it this way cleans up the logic a bit. Walter said there were various minor changes to the flow descriptions based on these new values. Arpad again noted that he disliked the use of "upstream" and "downstream" in the descriptions and thought they could lead to confusion in multi-channel redriver flows. Terms like immediate and direct were again discussed, and Radek suggested preceding and following. Ambrish agreed with Radek's suggestion. Radek suggested that if there are any places in the descriptive text that still refer to "combining" responses, then we should probably replace them with a proper mathematical term like "convolving". Bob noted his logistics comments and said that the BIRD starts with changes to 10.2.3, then provides the new parameter definition, which would be in section 10.3, then jumps back to the flow descriptions in 10.2.3. Bob suggested moving the new parameter section to the end of the BIRD to match the order of changes in the specification. Walter agreed with Bob's point, but he thought that was an editorial issue and that the BIRD read more clearly in its current order. Bob said the clarity and readability would be addressed by adding the cross references he suggested. Walter added a sentence to the parameter's Definition. It notes that the parameter affects the AMI_Init's impulse_matrix and the AMI flows, and it adds a cross-reference to section 10.2.3. In response to a question from Ambrish, Walter noted that he'd taken the regular non-redriver flow(s), which had been separate for statistical and time-domain, and combined them into one flow. Steps 1 through 3 applied to both. The group then reviewed a list of suggested edits and corrections from Fangyi: (In the notes below: A ---> B means change A to B). - Definition of the "Combined" value of the new parameter - Add "and channels" because it previously only said all upstream models. - Normal flow step 3d - "which is presented" ---> "and present the result to the Rx..." - Normal flow step 4 - "calculated in step 3" ---> "returned in step 3" - "given [Algorithmic Model]" ---> "the Tx and Rx algorithmic models" - "model filter equalization" ---> "filter that represents the model's equalization, which can be determined using..." - subsection 2.b. - "ignore for the purpose of" ---> "ignore while determining" - Normal flow step 6b. - "Step 4" ---> "Step 5" - One instance of "Model makers that use..." ---> "Models that use..." Fangyi and others had more changes to review. Walter said he would send out draft 4 with the changes made today. He said if people could put their changes in email form he would attempt to make those corrections ahead of the next meeting. If not, we can continue review next time. - Ambrish: Motion to adjourn. - Walter: Second. - Arpad: Thank you all for joining. AR: Walter to send his PAMn proposal to Randy to submit to the Open Forum as an official BIRD. AR: Walter to create a BIRD211.2_draft4 with today's changes and send it to ATM. ------------- Next meeting: 11 May 2021 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives