IBIS Macromodel Task Group Meeting date: 09 July 2013 Members (asterisk for those attending): Agilent: * Fangyi Rao * Radek Biernacki Altera: * David Banas Julia Liu Hazlina Ramly Andrew Joy Consulting: Andy Joy ANSYS: Samuel Mertens * Dan Dvorscak * Curtis Clark Steve Pytel Luis Armenta Arrow Electronics: Ian Dodd Cadence Design Systems: Terry Jernberg * Ambrish Varma Feras Al-Hawari Brad Brim Kumar Keshavan Ken Willis Cavium Networks: Johann Nittmann Celsionix: Kellee Crisafulli Cisco Systems: Ashwin Vasudevan Syed Huq Ericsson: Anders Ekholm IBM: Greg Edlund Intel: * Michael Mirmak Maxim Integrated Products: Mahbubul Bari Hassan Rafat Ron Olisar Mentor Graphics: * John Angulo Zhen Mu * Arpad Muranyi Vladimir Dmitriev-Zdorov Micron Technology: Randy Wolff Justin Butterfield NetLogic Microsystems: Ryan Couts Nokia-Siemens Networks: Eckhard Lenski QLogic Corp. James Zhou SiSoft: * Walter Katz * Todd Westerhoff Doug Burns * Mike LaBonte Snowbush IP: Marcus Van Ierssel ST Micro: Syed Sadeghi Teraspeed Consulting Group: Scott McMorrow * Bob Ross TI: Casey Morrison Alfred Chong Vitesse Semiconductor: Eric Sweetman Xilinx: Mustansir Fanaswalla Ray Anderson The meeting was led by Arpad Muranyi ------------------------------------------------------------------------ Opens: - Arpad: I added a last minute agenda item - Walter: I would like time to discuss Resolve_Dependency -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Fangyi to add Tx_DCD example to BIRD 155 - Radek: There was discussion of not using Usage Out - Fangyi: It is done, but it is an Rx example ------------- New Discussion: Interconnect update: - Walter: Time used for editorial work - Arpad: We can drop this standing agenda item Usage Out for Jitter parameters: - Walter: There was objection to those parameters being Out. - The idea was that there was nothing the tool could do with them. - I think it was the wrong decision. - It is now deprecated. - SiSoft believes Out should be allowed. - Fangyi: Agree, as long as the note about it is sent out. - Radek: Is it too late to change for 6.0? - Michael M: This would be unapproving an approved change? - Radek: Yes. - Walter: This should be legitimate by our rules. - It may delay 6.0 due to voting schedule. - Bob: It would be better in 6.0. - Walter: Approving it at the next open forum would speed up the process. - Walter motioned for the group to approve such a BIRD in advance. - Arpad: Where there other reasons to take that Out? - Walter: No, we might find out more from the minutes. - Mike L: I could not find anything in the minutes. - Todd: It may have been something about changing the course of the simulation. - Arpad: That would apply to Model_Specific, but these are Reserved. - Todd: I think you are right. - Ambrish: In the minutes Walter said only RX_Noise should be Out. - Did that make it to the spec? - Bob: Rx_Noise can be Out. - Rx_Sensitivity is Info or Out, a carryover from 5.1. - Ambrish: The minutes say Ken objected to jitter being Out. - Walter: A Nov 2011 draft allowed Out, a Dec draft did not. - Arpad: Would Out be allowed for GetWave and Init? - Bob: We should allow it for both. - Walter withdrew his motion. AR: Ambrish ask Ken Willis about reasons for jitter not being Usage Out AR: Walter draft BIRD to allow Usage Out for jitter Resolve_Dependency BIRD 155: - Walter: Some times we want parameters to depend on corner or model name. - AMI_Init does not have corner or model name as an input. - We could allow passing input parameters. - Ambrish: They could be Model_Specific. - Walter: Corner can be, but not model name. - Fangyi: Isn't the AMI root the model name? - Walter: No you can't count on that. - Todd: The root is supposed to be the model name. - David: That is not correct. - Walter: There may be other possible solutions. - David: It should be resolved in the final version of this BIRD. - Arpad: Each [Model Selector] entry has a unique name. - David: But they usually all call the same DLL. Eye mask definitions: - Arpad showed a presentation on eye mask definitions. - Arpad: Eyes have both width and height requirements. - We lack timing for setup and hold checks. - Rx_Receiver_Sensitivity only lets us bound top and bottom. - A full eye mask would let us check setup and hold. - These standards usually stay away from things in the core itself. - It is not clear where Rx_Receiver_Sensitivity applies. - The core logic is not part of AMI, but Rx_Receiver_Sensitivity may apply to it. - Some people think it applies at the analog model. - Michael M: Is an eye mask just for a pass/fail check? - Arpad: Yes, but also margins. - David: Some people inject jitter and watch for failure. - This is an indirect measurement. - If you do both this and check CDR at the slicer you are double counting. - Walter: There is a fundamental difference between eye opening and Rx_Receiver_Sensitivity. - Eye opening is slightly related. - It is usually a rule for compliance to a specific standard. - For DDR4 JEDEC has nomenclature for eye width and height. - In SerDes eye masks are measurement masks, built into scopes. - Ambrish: Is there a compliance requirement? - Michael M: It is advantageous to have a compliance check that avoid getting into inner details. - Arpad: This is similar to Vinh/Vinl. - We know the real chip will have smaller hysteresis. - We can put in spec numbers or actual device numbers. - Walter: Model makers will want to put in spec numbers. - Some people will want actual operation though. - Ambrish: AMI models are not made for a specific spec. - Michael M: We could take a kit style approach with different kit types. - Arpad: Do we agree we want this? - David: How else would anyone characterize jitter without double counting? - Ambrish: The jitter parameters handle that. - Walter: Sometimes we get negative RJs. - There probably will be a BIRD for DDR4. - There are no timing constraints and no Vref. - It will be done in legacy IBIS. - Michael M: I would like to see an eye mask option. - This came up at a 2008 Asia summit. - It should ideally be compatible with AMI. - Arpad: There would be voltage, time, and BER in the definition. - Fangyi: That would be for pass/fail, BER would not be needed. - Arpad: A BER would give a minimum to pass. - Ambrish: The tool figures out BER. - Michael M: Vinh/Vinl give a concept of voltages to guarantee latching, not actual performance. - Todd: We agree with David, it is baked into jitter. ------------- Next meeting: 16 July 2013 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives