IBIS Macromodel Task Group Meeting date: 13 Dec 2011 Members (asterisk for those attending): Agilent: Fangyi Rao Radek Biernacki Altera: * David Banas Ansys: Samuel Mertens * Dan Dvorscak * Curtis Clark Arrow Electronics: Ian Dodd Cadence Design Systems: Terry Jernberg * Ambrish Varma * Feras Al-Hawari Celsionix: Kellee Crisafulli Cisco Systems: Ashwin Vasudevan Syed Huq Ericsson: Anders Ekholm IBM: Greg Edlund Intel: Michael Mirmak LSI Logic: Wenyi Jin Maxim Integrated Products: * Mahbubul Bari Mentor Graphics: * John Angulo Zhen Mu * Arpad Muranyi * Vladimir Dmitriev-Zdorov Micron Technology: Randy Wolff NetLogic Microsystems: Ryan Couts Nokia-Siemens Networks: * Eckhard Lenski QLogic Corp. * James Zhou Sigrity: Brad Brim Kumar Keshavan * Ken Willis 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 The meeting was lead by Arpad Muranyi ------------------------------------------------------------------------ Opens: - Arpad: The next meetings will be Dec 20 and Jan 3 2012 -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Ken propose reduced set of jitter parameters - Nothing formal yet - Have been discussions - May be some discussion today - Mostly on the RX side - David write mathematical description of combined Dj proposal - On hold, pending further discussion ------------- New Discussion: David motioned to discuss BIRD 123: - Walter seconded - No one opposed Discussion of BIRD 123: - Walter showed an email discussing BIRD 123 - Walter: It should be clear once we go through the derivation methods - Model defines parameters for Rj, Sj, DCD not modeled in clock_times - Also defines *_Clock_Recovery_* for what is modeled - These describe when the EDA tool needs to handle jitter separately - Using only Init EDA tool must determine center of eye - This is the only part not well defined - It should be based on the median of the PDF - David: It might be the center between the two medians - Vladimir: Median is not the mean - It is difficult to find the median - Walter: It should not be difficult in time domain - David: It would be the peak of a fairly wide averaging window - Walter: If we use the mean that would have to be carefully defined - Arpad: One definition uses the mode, the most prevalent value - Walter: That would not be the best choice - Ken: These parameters are not needed if all jitter is modeled - Walter: True - Ken: We should have one parameter set regardless of whether Init or Getwave is used - Walter: Every one of these parameters has been required by a vendor - It's hard to put Rj into clock_times - Ken: Rj is usually handled by the tool - Walter: What if the model does include Rj in clock_times? - Ken: We should be able to do it with less parameters - Ambrish: This allows jitter to be double counted - Walter: The model explicitly say what jitter to add - Ambrish: The user may want to add more jitter - Walter: They still can do that - Arpad: These are Reserved Usage In parameters? - Walter: They are Info, not In - There is a separate discussion whether to allow Out - No objection to not allowing Out - In IBIS 5.0 TX jitter was allowed to be Out - Arpad: Was that an oversight? - Walter: It was thought to be a need at the time - It is not needed - David: I proposed making them Out to cut the number of parameters - Walter: How would it do that? - David: We have two sets of parameters in Walter's proposal - Walter: One is for when you have clock_times, one for not - With a forwarded clock the jitter is not known by the model - That has to be added - It helps to keep the parameters separate - One example implementation is a rotator - Each rotator generates a base frequency - The CDR switches from one rotator to another - The frequency is different from the TX - Another is forwarded clocks - The frequency is the same as the TX - Ken: This is driven by the need to model external clocks? - Walter: It does not depend on whether you have a forwarded clock - Arpad: How does it get into the RX DLL? - Walter: It does not, it is in the AMI parameters - Arpad: The RX never needs to get it from outside? - Walter: We decided that would be in a separate BIRD - There will be additional mechanisms to model spectral density jitter - Arpad: How does this relate to the other drafts? - Walter: This explains the BIRD I posted - That one should agree with this - The only exception is adding the eye center definition - I will review the BIRD and email to the reflector - Arpad: Hopefully people will review and we can vote in January - Bob: We should post the derivation explanation somewhere AR: Walter update BIRD 123 AR: Walter create jitter derivation document - Ken: Did we agree they are Info? - Walter: The only one that should be Out is RX Noise - RX Getwave changes the gain and can affect Noise - Ken: Will we deprecate TX Jitter? - Walter: That can be converted to these parameters - Existing models us it - Deprecation begins with a warning announcement - David: If the DFE adapts maybe these should be Out - Mike: A model might not jitter clock_times depending on the controls - Walter: A dependency table can describe that - Arpad: The BIRD does out describe what the tool does if it is an Out - The model can specify jitter but the tool might ignore that - Walter: The tool should use the model value as default - Ken: Is there a reason to make it Out? - David: It is risky to assume updates will never be needed - Walter: With RX Noise there is a known reason - David: Our model might want to tell the tool how much jitter to add - Walter: Ambrish objected to TX jitter being an Out - That has to be injected back into the stimulus - It gets complicated and should be avoided - We have no request from an IC vendor yet - Arpad: It is usually safer to have less features and add later - We agreed it would be Info - Vladimir: Why not add noise to the waveform? - Walter: You could - This would avoid using Out mode - Walter: Except Gaussian noise has tails - There are ways to handle it after the AGC converges - Some models do this and are not standard - IC vendor will like the option to have it as an Out - Vladimir: We can use the last value as long as it is Stable - Ambrish: There are different kinds of jitter for statistical and TD - There could be an overlap in the kinds of jitter - Walter: It is up to the EDA tool Meeting ended. ------------- Next meeting: 20 Dec 2011 12:00pm PT Next agenda: 1) Task list item discussions ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives