IBIS Macromodel Task Group Meeting date: 08 Jan 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: We will not meet Jan 29 since most of us will be at DesignCon - Radek: Would like to discuss the latest dependency table BIRD -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Fangyi, Ambrish, Walter to discuss merging their "Redriver" proposal with BIRD 131. - Fangyi: There has been no progress - Walter believes one BIRD is generic while the other is AMI specific - This makes it too difficult to merge them - Arpad: Should we cancel this AR? - Walter: Yes, we can cancel ------------- New Discussion: Interconnect task group report: - Michael M: There has been no meeting since the last report - We will continue discussion of tying AMI analog with interconnect proposals - The next meeting will be tomorrow - Walter: I sent a presentation in Dec reviewing that topic - It was an email sent Dec 18 Arpad showed a presentation Analog Modeling Discussion: - slide 2: - Arpad: Our discussions have been going in circles - We have too many BIRDs - Some conceptual questions must be answered - slide 3: - Arpad described classic IBIS I/V behavior with regard to voltage - Arpad: It handles only two "on" states - Walter: Our fundamental differences are on three points shown here - Everything here is true for [External Model] in IBIS 4.1 on, but not for anything else - slide 4: - Arpad: Current flows from either of two supply nodes - There is no direct transfer function from logic input to output - The input stimulus is really a selector - High and low states can have different impedance - slide 5: - Arpad: [Driver Schedule] describes events, the EDA tool handles the transition - slide 7: - Arpad: LTI models such as s-params almost could be used directly - But the spec says the impulse response has to be convolved - slide 8: - Arpad: This is the regular flow - slide 9: - Arpad: The alternate flow has the combined channel between the GetWaves - Walter described the process differences - slide 10: - Arpad: The TX GetWave input stimulus is ideal, except maybe adjusted for V=0 timing - The output is semi-digital - Walter: In this case only because the FFE is ideal - slide 11: - Arpad: For this flow the TX GetWave input has to be analog (semi-digital) - The RX output is 1, 0, or X - None of this is stated in IBIS 5.1 - Walter: There are many issues with this - slide 12: - Arpad: BIRD 116 proposes something similar, except a D-to-A is added - slide 14: - Arpad: These possible changes to [Model] may be too complicated - slide 15: - Arpad: Do we really need supply terminals? - slide 16: - Arpad: [Algorithmic Model] is always differential, but [Model] is single ended - It could not be used for single ended simulations - slide 17: - Arpad: There are many issue using [Model] for AMI - We need a new model keyword for AMI slide 4: - Walter: We should focus on only Pullup and PullDown curves - AMI assumes they are linear - Also we can only measure in high and low states - But the K(t) functions simulate intermediate states - With SerDes the devices really just change current flows - A SerDes can have many states, not just two - It can be represented with Pullup and PullDown coefficients - The Pullup/PullDown impedance doesn't matter because they are used differentially - The total impedance remains the same - The control logic was implemented the way push-pull drivers typically worked - But nothing in IBIS forces it to be that way - This is why using the IBIS buffer for the analog model is a problem - Radek: That ignores some items in the IBIS spec - An s-param block can't have DC sources - I agree with Arpad - Walter: The TX and RX models, especially above 6GHz, are s4p models - We can do a better job describing how to generate those - Arpad: What is used for the control input - Walter: Effectively the GetWave output - Radek: IBIS did not specify this well enough - David: It was described in terms of mathematics - We are changing the value of PUref and PDref - The FIR output gives these values - Michael M: Would like to hear how others are interpreting the spec - It doesn't have to be digital vs. analog, it can be seen as real numbers - David: It is the reference that changes, with UI granularity - Michael M: We may need new model keywords to handle this right - David: This also means clock-to-out for IBIS models is meaningless - Walter: Timing is always done with standard loads - IBIS says nothing about clock-to-out - Values change in UI increments only with a "nice" FFE - I have been modifying the Ramp boxes, not the PU reference - Arpad: That relates to the idea that these are just real numbers - slide 12: - Arpad: The TX analog model input supplies power, so PU reference does not - Walter: The diagram here doesn't quite work because it doesn't show correct s-param connection to the differential driver Michael M: We need a BIRD to address this - Arpad: First we must settle our differences, whether a new keyword is needed - David: We probably need to discuss the flows again to test assumptions - Michael M: We need to see if people adhere to flow understanding A, B, or even C - A majority vote would not work well - Arpad: We might need to settle Walter's assertion about pre-4.0 IBIS - David: That may be going too far back - Walter: We just need to be able to describe the analog model as s4p - Arpad: Where is it connected? - Walter: To the output of the logic block - Arpad: We still need to determine the logic voltage level - Walter: A document will be presented on that - David: Each EDA vendor should present on how they handle TX GetWave - Fangyi: We would have to get internal approval for that ------------- Next meeting: 15 Jan 2013 12:00pm PT Next agenda: 1) Task list item discussions ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives