IBIS Macromodel Task Group Meeting date: 18 August 2009 Members (asterisk for those attending): Adge Hawes, IBM * Ambrish Varma, Cadence Design Systems Anders Ekholm, Ericsson * Arpad Muranyi, Mentor Graphics Corp. * Barry Katz, SiSoft * Bob Ross, Teraspeed Consulting Group Brad Brim, Sigrity Brad Griffin, Cadence Design Systems Chris McGrath, Synopsys David Banas, Xilinx Deepak Ramaswany, Ansoft Donald Telian, consultant Doug White, Cisco Systems * Eckhard Lenski, Nokia-Siemens Networks Essaid Bensoudane, ST Microelectronics * Fangyi Rao, Agilent Ganesh Narayanaswamy, ST Micro Gang Kang, Sigrity Hemant Shah, Cadence Design Systems Ian Dodd, consultant Jerry Chuang, Xilinx Joe Abler, IBM John Angulo, Mentor Graphics John Shields, Mentor Graphics Ken Willis, Cadence Design Systems * Kumar Keshavan, Sigrity Lance Wang, Cadence Design Systems Luis Boluna, Cisco Systems Michael Mirmak, Intel Corp. * Mike LaBonte, Cisco Systems Mike Steinberger, SiSoft Mustansir Fanaswalla, Xilinx Patrick O'Halloran, Tiburon Design Automation Paul Fernando, NCSU Pavani Jella, TI Radek Biernacki, Agilent (EESof) * Randy Wolff, Micron Technology Ray Komow, Cadence Design Systems Richard Mellitz, Intel Richard Ward, Texas Instruments Samuel Mertens, Ansoft Sam Chitwood, Sigrity Sanjeev Gupta, Agilent Shangli Wu, Cadence Design Systems Sid Singh, Extreme Networks Stephen Scearce, Cisco Systems Steve Pytel, Ansoft Syed Huq, Cisco Systems Syed Sadeghi, ST Micro Ted Mido, Synopsys Terry Jernberg, Cadence Design Systems Todd Westerhoff, SiSoft * Vladimir Dmitriev-Zdorov Vikas Gupta, Xilinx Vuk Borich, Agilent * Walter Katz, SiSoft Zhen Mu, Cadence Design Systems ------------------------------------------------------------------------ Opens: Bob: We should discuss AMI specification clarification - The parser developer has raised a number of questions about .ami files - Walter: This should be emailed to the list - Arpad: We need to address this at another time - The agenda for today is full and key participants are here -------------------------- Call for patent disclosure: - No one declared a patent. ------------- Review of ARs: - Walter send updated IBIS-ISS to Mike L - Done - Mike L post IBIS-ISS to website and email open forum - Done - Arpad send updated IBIS-ISS presentation to Walter - TBD - Mike change web page to make items easily linkable - TBD - Arpad write a BIRD to clarify time period accuracy requirements - TBD - Todd: Write IBIS s-param BIRD - No update - Arpad: Write parameter passing syntax proposal (BIRD draft) for *-AMS models in IBIS that is consistent with the parameter passing syntax of the AMI models - TBD - TBD: Propose a parameter passing syntax for the SPICE - [External ...] also? - TBD - Arpad: Review the documentation (annotation) in the macro libraries. - Deferred until a demand arises or we have nothing else to do ------------- New Discussion: Walter presented the SiSoft proposal for Tx_GetWave input clarification: - Slide 1: - IBIS 5.0 does not support non-LTI TX models - Arpad: We are not talking about analog, right? - Walter: right - Slide 2: - Input to Tx_GetWave is stimulus convolved with channel - This is incorrect - Slide 4: - Kumar: This may not be convolved with the bit stream - Slide 5: - Changing the order of evaluation solves non-LTI problems - Slide 6: - Non-LTI examples: duty cycle distortion and n-tap table lookup buffer - Slide 7: - Walter corrected a typo - Slide 8: - Table lookup output has no linear relationship to input - Tx_GetWave is unable to reconstruct the original bit pattern from the analog input waveform - The equalized pattern bears no resemblance - Slide 9: - Duty cycle distortion has pattern shifted 1 bit - Slide 10: - This is similar to IBIS [Driver Schedule], used for two-tap pre-emphasis buffers - Slide 12: - 75% duty cycle dist means each segment is a different number of time steps - Fangyi: Shouldn't the first one be 6 steps, not 7? - Walter explained DCD calculation - Slide 13: - This proves that DCD is non-LTI - IBIS 5.0 With or without DCD, Tx_GetWave gets the same input signal - Another effect is DCD amplification, which is too involved to show here - Slide 14: - Tx_GetWave needs to receive a digital stimulus pattern - The digital signal is either below 0 or above it, swinging -0.5 to +0.5 - Slide 15: - IC vendors currently use digital patterns for their models Kumar: non-LTI is not the right terminology - The bit stream is a continuous waveform - It can have amplitude distortion - This is nit-picking that will cause confusion - Walter showed slide 5 - Walter: Tx_GetWave input would have channel distortion - Kumar: Tx_GetWave just adds equalization - Walter: But that is at the beginning of the channel, not the end - If Tx_GetWave is non-LTI you can't add it there - Kumar: The order of convolution does make a difference with non-LTI - The bit generator should be able to add DCD - There should be no confusion about where the DCD is added - System jitter would appear in the bit stream - Walter: Why can't Tx_GetWave accept just transition times? - Kumar: The time should not be fixed, there can be jitter - Walter: The times can be anything, not constrained by bit time - Fangyi: Why have the simulator add jitter? - Walter: The simulator should have the choice - Kumar: If we go that way, 0/1V would be better than +/-0.5 - Walter: We mostly work with differential signals - Fangyi: There are 2 problems here - Who is responsible for jitter - What is a non-linear TX - Walter: TX_Init stores the analog impulse - Tx_GetWave combines that - Kumar: We should not do that - Walter: Tx_GetWave produces the input to RX_GetWave - Kumar: That is an unnecessary restriction - Fangyi: That should be up to the EDA tool - Walter: This is the BIRD SiSoft will introduce - Fangyi: The TX needs to have a 1 or 0 input? - Walter: Unlike the RX, the TX has a digital input - Simulators have chosen different methods to convey it - You could give just a list of times - Kumar: Agree with Walter that TX_G output is the bit generator - Walter: No it isn't - Kumar: This is just the way SiSoft works today - Walter: IC vendor models work this way - Kumar: We should not be convolving - Walter: Why is TX_Init convolved? - Kumar: I should have the option to use circuit simulation, not convolution - Walter: That would be a new BIRD, this is about making the current setup work - Fangyi: There is no current requirement to convolve in Tx_GetWave - Walter: We can add a switch to say if Tx_GetWave convolves - Kumar: Agree - Walter: Tx_GetWave is only needed for non-LTI - Tx_GetWave only needs to support non-LTI - Kumar: The channel is passive, not non-LTI - Analog channels are a different beast - NMOS drivers are non-LTI - Walter: It would be better to have objections in writing Fangyi: Is this proposal backward compatible? - Walter: No, there will be a version 2 AMI - It will have digital stimulus - Bob: The stimulus is analog - Walter: The BIRD will call for digital - Tx_GetWave only works for LTI - But then it is not needed - Fangyi: So analog stimulus models would not be supported? - Walter: Those would be version 1 AMI models - Bob: Most vendors want backward compatibility - SiSoft is free to submit a BIRD on it's own - Compatibility is a fundamental IBIS requirement Walter: We made a serious mistake with the AMI design: - Bob: Ideally we will have consensus in this subcommittee - Arpad: We can discuss this - Will the BIRD be submitted directly? - Walter: I will explain equalization next week - Arpad: Is that a separate BIRD? - Walter: We don't want to have too many BIRDs - Getting this working has been painful - It may be better to identify all deficiencies and write a new AMI - Bob: it is a dangerous process to abandon what we have done - Not sure how this impacts other vendors - Walter: What vendors do we have? - SiSoft, Agilent, Sigrity, Cadence, Mentor - They should all participate - Arpad: These things should all be written up - Walter: I will not write BIRDs for each item - Arpad: Should we have more presentation style discussions? - Walter: I have slides for other topics Next meeting: 25 Aug 2009 12:00pm PT -------- IBIS Interconnect SPICE Wish List: 1) Simulator directives