IBIS Macromodel Task Group Meeting date: 05 April 2011 Members (asterisk for those attending): Agilent: Fangyi Rao * Radek Biernacki Ansoft: Chris Herrick Danil Kirsanov Ansys: Samuel Mertens Dan Dvorscak Deepak Ramaswamy Jianhua Gu * Curtis Clark Cadence Design Systems: Terry Jernberg * Ambrish Varma Celsionix: Kellee Crisafulli Cisco Systems: * Mike LaBonte Stephen Scearce Ashwin Vasudevan Ericsson: Anders Ekholm IBM: Greg Edlund Intel: Michael Mirmak LSI Logic: Wenyi Jin Mentor Graphics: John Angulo Vladimir Dmitriev-Zdorov Zhen Mu * Arpad Muranyi Micron Technology: * Randy Wolff Nokia-Siemens Networks: Eckhard Lenski Sigrity: Brad Brim Kumar Keshavan * Ken Willis SiSoft: Walter Katz Mike Steinberger Todd Westerhoff 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 The meeting was lead by Arpad Muranyi ------------------------------------------------------------------------ Opens: - None -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Arpad post Crosstalk BIRD draft - Done - All review draft - Ambrish start a BIRD on task list row 25 - In progress - Bob write a BIRD on correcting Table 1-3 in the spec. (Row 23). - In progress ------------- New Discussion: Arpad showed the Crosstalk BIRD draft: - Arpad: Radek suggested changing the wording of impulse units - Is it v/s or v-s? - Radek: It is different in time domain - Only in digital do we drop the "per volt" - We should say it right or remove it - This is really voltage impulse response - Arpad showed an email from Vladimir Dmitriev-Zdorov - This explains the v/s derivation - Radek: This paragraph should be in the document - Or just don't use units - This goes beyond what impulse response is - Arpad showed an email from Radek - Walter: Is this about channel response or TX response? - Radek: Nowhere do we go beyond the classical definition of IR. - Walter: IR should be unit-less - Since a 1V input is used, technically volts are involved - But then voltage is divided back out - Radek: Yes it is v/v - Ken: Kumar said it should be v/s - Bob: Has the serdes industry defined this incorrectly? - Scott: It should be v/v, transfer function 1/s - Arpad: Currently the BIRD says v/s - Who wrote this? - Walter: Either me, Ambrish, Kumar ... - Ken: We could call it "voltage impulse response" as Radek suggested - Can we vote on it? - Arpad: This has enough changes to put off a vote - Radek: I had suggested other changes - Need a change to row_size - The number of rows is the number of aggressors - Arpad: We should take this offline Arpad showed the Usage OUT and InOut BIRD: - Arpad: A few paragraphs have been added on page 1 - Ambrish: We should say if simulation results should change as a result - Walter: We have to be careful - That would apply to jitter parameters - Ambrish: This is about model specific parameters - Arpad: That may be OK - Walter: There are backchan parameters that will affect simulation - These are kind of reserved - Arpad: This is about model specific parameters - Walter: They may become reserved - Radek: We should decide how to classify these parameters - Ken: This side data should just be written to a text file - Arpad: The BIRD does not specify how the tool should do it - Some do not like the idea of models writing files - Standard formats would be needed - Walter: Until a format BIRD exists it is non-standard - Ken: This is model specific - Walter: BIRD 121 has that - DLLId gives the file name - Ken: If files are used is this Out paremeter provision needed? - Walter: IBIS 5.0 models would need it - Ambrish: The output does not have to be in any particular format - Walter: It does, the BNF specifies it - Arpad: The paragraph I'm highlighting explains what tool vendors should do - We should add another statement to this - Scott: It should also say that current or future results must not be modified - Anything that does that should be a reserved parameter - Bob: EDA tools should not modify model specific parameters - Radek: They should not act upon them - Walter: Model specific parameters should all become reserved eventually - That may take long enough that some would not want to wait - Scott: The intent is to shorten the cycle - While the parameters are Model_Specific there should be a debate about it - Not everything will go in the specification - Walter: The process for getting into IBIS has to be accelerated - Scott: No, the problem is getting info about these ideas early enough Arpad: The form of our spec is rigid, keyword based - A language would be better - *-AMS languages are good - When we started the macromodel group I thought we would go that way - We made a big mistake with these IBIS keywords - Walter: AMS can add keywords that are transparent to IBIS - The AMI clarification effort began two years ago - But it was put off - IBIS triangle: - Vendors, minimally represented here - EDA tools - Users - Scott: Languages have to do with meaning, not algorithm - This is important when a EDA tool is used - Arpad: What is the meaning of a resistor? - It is well defined in ISS - Radek: If you have a subckt you have to know how it is to be used - Arpad: With a language the spec would not have to be updated so much - Scott: For example, we have to agree on the meaning of each jitter parameter - It is a painful process - It has been better since Arpad took over Ambrish: We should say these output parameter strings are optional - Walter: Meaning a model with no output does not need the string? - Ambrish: Just that the model may output nothing - Radek: If the Parameter is there it should produce a string - Walter: The parameter is ** so it has to point to something - It should not be a null pointer - But it could point to a null string - Ambrish: It should be able to do nothing - Walter: It's not hard to return a null string AR: Arpad add null output language Arpad showed an email from Mike Steinberger: - Arpad: We need to settle on language about arrays - Ambrish: Existing models should continue to work - I wrote language that should work - All multi-row tables require row names - Bob: My alternative does not require that - Ambrish: I haven't read that yet - Arpad: We need to focus on one proposal at a time - Bob: I added an extension for single row tables with no parentheses - Walter: Maybe Ambrish and Bob could work to agree on this - I won't object to the result Arpad: Walter should rewrite BIRD 123.1 for OUT/InOut for the next meeting - Walter: OK - Scott: Could we have block diagrams for TX and RX - This would show where jitter parameters go - Some HSSCDR slides provide good templates AR: Walter rewrite BIRD 123.1 Walter: BIRD 122 should be a good solution now ------------- Next meeting: 12 Apr 2011 12:00pm PT Next agenda: 1) Task list item discussions ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives