IBIS Macromodel Task Group Meeting date: 13 October 2015 Members (asterisk for those attending): ANSYS: * Dan Dvorscak * Curtis Clark Avago (LSI) Xingdong Dai * Bob Miller Cadence Design Systems: Ambrish Varma Brad Brim Kumar Keshavan Ken Willis eASIC David Banas Marc Kowalski Ericsson: Anders Ekholm IBM Steve Parker Intel: Michael Mirmak Keysight Technologies: * Fangyi Rao * Radek Biernacki Maxim Integrated Products: Hassan Rafat Mentor Graphics: * John Angulo * Arpad Muranyi Micron Technology: * Randy Wolff Justin Butterfield QLogic Corp. James Zhou Andy Joy SiSoft: * Walter Katz * Todd Westerhoff * Mike LaBonte Synopsys Rita Horner Teraspeed Consulting Group: Scott McMorrow Teraspeed Labs: * Bob Ross TI: Alfred Chong (Note: Agilent has changed to Keysight) The meeting was led by Arpad Muranyi. ------------------------------------------------------------------------ Opens: - Walter mentioned that he had some new thoughts on the Redriver Init Flow BIRD and would be making a motion to untable it, time permitting. -------------------------- Call for patent disclosure: - None. ------------- Review of ARs: - Arpad to clean up draft 14 of the Special_Param_Names BIRD and send it to Mike L. for posting to the ATM website. - Done (Mike posted it, and Arpad sent an announcement to the ATM reflector). ------------------------- Review of Meeting Minutes: - Arpad: Does anyone have any comments or corrections? [none] - Todd: Motion to approve the minutes. - Arpad: Second. - Arpad: Anyone opposed? [none] ------------- New Discussion: Item #6: renamed: New IBIS-AMI Reserved Parameter Special_Param_Names. - Discussion: Arpad reviewed draft 14 and highlighted the changes incorporated based on last week's meeting. Radek expressed concern over the wording of the long sentence describing the Usage Rules. After some discussion, all agreed that the phrase "in which this parameter appears" could be removed from the sentence. Bob M. then reminded the group of a suggestion he had made that the BIRD might state that an EDA tool should throw a warning if it does not recognize any of the special parameters in the table. He felt this was a way to ensure that an end user would know of the issue, which was the primary benefit a model maker would derive from using the new Special_Param_Names. David Banas and others had agreed with Bob M.'s suggestion (first made via email). Attendees understood the value of Bob M.'s proposal, but Arpad and Bob R. pointed out that the specification generally avoids defining "what the tool should do." They both stated that it was really a tool vendor issue, and the benefit in the BIRD was that it provided a path for model makers to declare their special parameters so the tools could do something with that information. Todd said the benefit to the model makers was the fact that they could explicitly declare their special parameters. Bob M. said that he had wanted to raise the issue, but was fine with the decision not to define what a tool should do with the information. - Bob R: I move that we submit this BIRD, as modified here today, to the IBIS Open Forum with a recommendation from the ATM that they approve it. - Walter: second. - Arpad: I think we should take a company by company vote. Results of Company Vote: ANSYS - yes Avago - yes Keysight - yes Mentor - yes Micron - yes SiSoft - yes Teraspeed Labs - yes [unanimous approval] - Bob R: Submit it to Mike L. to be uploaded as BIRD 179. Item #7: Discussion of language corrections regarding "ground". - Discussion: Walter wanted to briefly discuss this topic in anticipation of full discussions in future meetings. He restated that we had all generally agreed that "ground" in a buffer referred to a local ground, and that we needed to iron out the language of a generic statement of that fact. He asked that Radek take an AR to propose such language. Radek stated that he had completed a smaller AR to offer comments on some of Walter's original slides, but that there was some more work to be done to draft a more comprehensive description. Arpad asked if we might start an email thread so we could more easily evaluate and comment on any proposed language. Radek accepted the AR to start the email thread aimed at hashing out the ground language. Item #11: New Redriver Flow BIRD. - Walter: I move that we untable the Redriver Init() Flow BIRD. - Arpad: Second. - Anyone opposed? [none] - Discussion: Walter reviewed draft 2 of his Redriver Init Flow improvements proposal. Walter reminded everyone that the current spec has a known issue with statistical flow for redrivers, and that earlier attempts to alter the flow had opened a Pandora's box of issues that he and Fangyi could not resolve to their mutual satisfaction. His new proposal was more limited in scope and defined a new Tx_Init_EQ_Depends_On_Input_IR Boolean parameter defined for Tx models. If it was false for a redriver's Tx, as would often be the case, that would result in a significantly simplified Init() flow that still provided the correct input IR to the downstream Rx. The proposal detailed four distinct methods of combining the effects of both channels for the final Rx that would all yield identical results. If the parameter was true, however, the flows get quite involved. Todd wondered if it made sense to have a parameter defined for a Tx model when it was the downstream Rx that was the driving requirement. Walter replied that the Rx was almost always initializing its equalization based on the input IR. It was knowing that the Tx did not that allowed for simplifying the flow for this common case. Fangyi objected that cross talk was not handled properly by the proposed flow even if the Tx did not optimize itself. He said that to analyze a network with crosstalk one needed to have the Impulse Responses handled independently for each channel, because crosstalk contributions could be introduced at any channel in the chain. He stated that the flow currently defined in the spec properly dealt with this issue by treating each channel independently, but he acknowledged that this left the problem with the Redriver Init() flow that Walter was attempting to correct. Fangyi suggested that one possible solution that would handle crosstalk was to pass the downstream Rx two different impulse responses. If one were the combined upstream and downstream channel IR, and the other were simply the downstream channel's IR, the Init() could optimize based on the combined IR and apply those optimized settings to the downstream only IR. Walter stated that he understood what Fangyi was proposing and why, but that he felt that his proposal could stand alone and offer a separate simplified flow when the parameter is false. Fangyi said he and Walter could continue to discuss the topic. - Arpad: Now is a good stopping point. - Thank you all for joining. AR: Arpad to submit today's final version of the Special Parameters BIRD to Mike L. to be uploaded as BIRD 179. AR: Radek to start email thread regarding finalizing the language for the cleanup of references to "ground". AR: Walter to email draft 2 of the Redriver Init Flow proposal to Mike L. to be posted to the ATM website. ------------- Next meeting: 20 October 2015 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives