IBIS Macromodel Task Group Meeting date: 10 January 2017 Members (asterisk for those attending): ANSYS: * Dan Dvorscak Curtis Clark Broadcom (Avago): Xingdong Dai * Bob Miller Cadence Design Systems: * Ambrish Varma Brad Brim Kumar Keshavan Ken Willis Cisco: Seungyong (Brian) Baek eASIC: David Banas Marc Kowalski Ericsson: Anders Ekholm GlobalFoundries: * Steve Parker IBM Luis Armenta Trevor Timpane Intel: * Michael Mirmak Keysight Technologies: Fangyi Rao * Radek Biernacki Ming Yan 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 * Kevin Li Teraspeed Consulting Group: Scott McMorrow Teraspeed Labs: * Bob Ross TI: Alfred Chong The meeting was led by Arpad Muranyi. -------------------------------------------------------------------------------- Opens: - Arpad: Mike LaBonte is taking minutes. - Arpad: We will cancel the Jan 31 meeting which conflicts with DesignCon. We will meet Jan 17 and 24. ------------- Review of ARs: - AR: Walter to email his example and BIRD 158.4 to the ATM list. - Done - AR: Mike LaBonte to update the "IBIS 6.2 BIRD Candidates" to rev 7 and send it to the ATM list. - Done -------------------------- Call for patent disclosure: - None. ------------------------- Review of Previous Meeting Minutes: - Arpad: Does anyone have any comments or corrections? [none] - Michael M.: Motion to approve the minutes. - Bob Ross: Second. - Arpad: Anyone opposed? [none] ------------- New Discussion: BIRD147.5: - Arpad noted that BIRD 147.5 had been recently emailed, and that it had been scheduled for a vote in the previous Open Forum meeting, but that had been deferred. Bob Miller showed the BIRD. He said there were issues with BCI_GetWave_Block_UI related to rounding. It needed to be an integer and BIRD 147.5 has been tweaked for that. The parameter name was now BCI_Message_Interval_UI and the definition had been changed. It informs the EDA tool what the interval wants to be. It is wide open as to what the optimization should be. Bob Ross said the Type should be Integer. Bob Miller changed Type to Integer. - Bob Miller proposed that models should be forgiving if AMI block size does not exactly match the requested frequency, and that BCI_Training_UI should be extended by tools to synchronize with the message interval. The expected number of adaptation steps should be achieved. Radek Biernacki said the definition probably should be more precise. When training time ends one more GetWave should be called before actually ending training. Ambrish Varma said that that was in the notes section. Radek said we would need to be careful about interactions with BCI_State. Bob Miller believed there would not be any interactions with that. Ambrish said if the model does not end training the EDA tool will take over. Bob Miller said that to have bit by bit accuracy the model would have to guarantee that it exactly matches device behavior on any EDA platform. Mike LaBonte asked if there should be a rule that Message_Interval_UI must be less than Training_UI, so ibischk could check it. Bob Miller said models would not necessarily message every time they were called, and that performance might degrade about 15% if the message window is long. Ambrish Varma suggested it might be better to add examples to the specification to illustrate the best practice. - AR: Bob Miller to update BIRD147.5 BIRD 187.1: - Michael Mirmak showed the BIRD. He noted that Curtis Clark had suggested removing "Type Value" because there is no such thing. The Format for Usage Out is not used by EDA tools except for Table. The four rules are all expressed in one paragraph. Michael said BIRD would be revised, trying to get down to one simple statement in one location, maybe two. Radek Biernacki said the sentence with "except" twice was confusing. It was not clear how to display Out data. Bob Ross said we had a list of rules on IBIS 6.1 page 194. He asked if the new rules would be moved there. Michael Mirmak said the question was if the rule should be stated in each place where it applies, or if it should be stated only once. Some rules can be far away in the document. Arpad Muranyi said there could be references made to a single location. Mike LaBonte added that including a reference to the original definition helps us to know all of the places where a duplicated rule would need to be changed. Bob Ross said we might want to explain Usage Out. Michael Mirmak agreed that very little had been stated about Usage Out. AR: Michael Mirmak to send an update to BIRD 187.1. BIRD158: - Radek Biernacki said the example needed to be discussed, and that once we agree, it should to be ported to the new BIRD template. Bob Ross asked if we really wanted another solution, since we already had a solution. He said we had multi-lingual, asking if we need a parameter. Ambrish Varma noted the parameter did not have to be in an AMI file. Bob Ross said we could link it through IBIS-ISS. Radek said a simpler solution would be to just give the Touchstone file name. He said if it remained Model_Specific it would always be experimental. Ambrish said that that is provided in the specification if called from an [External Model] parameter. Arpad Muranyi said tools wouldn't know what to do with that. Ambrish said it should be regarded as an analog model. Walter Katz asked what bad thing would happen if it became a Reserved_Parameter. Ambrish said doing so expected the EDA tool to construct a canned circuit, and that we already had a right way to do it. Walter said there was no right way for frequency domain simulation. Radek agreed there was no right way, but it was just an approximation of device behavior. It was "a" way. Arpad asked if everything was IBIS file-centric or AMI-centric. - Arpad: Thank you all for joining. ------------- Next meeting: 17 January 2017 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives