IBIS Macromodel Task Group Meeting date: 31 oct 2006 Members (asterisk for those attending): * Arpad Muranyi, Intel Corp. Barry Katz, SiSoft Bob Ross, Teraspeed Consulting Group * Doug White, Cisco Systems * Hemant Shah, Cadence Design Systems Ian Dodd, Mentor Graphics Joe Abler, IBM * John Angulo John Shields, Mentor Graphics Ken Willis, Cadence Design Systems * Kumar, Cadence Design Systems Lance Wang, Cadence Design Systems Michael Mirmak, Intel Corp. * Mike LaBonte, Cisco Systems Paul Fernando, NCSU * Randy Wolff, Micron Technology * Richard Ward, Texas Instruments Sanjeev Gupta, Agilent Shangli Wu, Cadence * Todd Westerhoff, Cisco Systems * Walter Katz, SiSoft Vuk Borich, Agilent Vikas Gupta, Xilinx ------------- Review of ARs: - Michael will try to get access to the IEEE document they are voting on. He will check into access restrictions first. If we get access to it, we will need a couple of volunteers to read the document and see if it is worth our while. - TBD - Macro Library Documentation - Mike will get to this in Nov - Mike posted recent presentations regarding API on the website. - Mike LaBonte looked into the phone bridge service since several Intel calls have been dropped. The MeetMe service used has been flagged for phaseout, so Mike did not pursue support. We can switch over to using the MeetingPlace service, which combines audio and web sharing, using only the audio portion. No immediate change planned. If there are more drops, then we can switch over at that time. ------------- New Discussion: Continued discussion of Kumar's presentation Slide 7: - In the top portion the driver and receiver have to operate in the same simulator - analog/electrical/circuit model. - Bottom portion shows the algorithm model, implemented as executable code. - Walter: what are the inputs and outputs of the Rx model? Kumar: both Tx and Rx contain only the "front end". inseparable from the channel. Init() takes impulse respone V/T, char string parameters - Questions from Walter to Kumar: Q: what does the Rx Init() function return? A: It may pass back it's dataspace, or just pass/fail. Q: Can it return peaking filters? A: Not included in this proposal but yes, it can return anything. Q: Do not have to write software to use an IBIS model. This proposal would need new software for each model. A: ??? Q: What are input and output of GetWave()? A: Output can be CDR output waveform and list of clock ticks. Bit time contained in clock ticks. - Walter will find out if this system is sufficient for his simulator, and how difficult it will be to build models. - Arpad: What generates the "actual waveform" passed to getWave()? Pulse response convolved with stimulus to produce it. - Arrows from top to bottom should be down only - Arrow from pulse to orange Tx should go left only - Arrow from waveform to green Rx should be right only - Key point about this architecture: Algorithmic modeling is completely separate from the circuit-level modeling...there is no feedback from one to the other. Slide 15 shows sequence of operations. - Waveforms 2 and 3 are identical - Waveform 1 is not Init wave, it is returned from Rx (unusual case) Bit pattern may be pure 1/0, or have have been convolved already. Mike: Simulator has to know how to handle everything given it. It has to be plug and play, or it's not IBIS. Arpad: will one model contain both the top and bottom parts (e.g. AMS model)? - Kumar: Lower-left box not needed if upper-left contains both algorithm and electrical. - Walter: 10,000 bits will simulate slowly in this case. - Arpad: For SerDes analysis bottom-left and bottom-right boxes have to be very fast. - Kumar: alternative is to characterize impulse response for all possible tap settings. - Arpad: How will driver know what coefficients to use? Arpad: It is likely that AMS models will be created, but the fact that AMS models will include both algorithmic and circuit portions makes it difficult to understand how the API will handle this. Kumar: If the circuit-level behavior can be clearly delineated from algorithmic behavior, then API should handle. Walter: need clear definition of Tx/Rx ins and outs to Init() and GetWave() functions, including how to pass parameters. Arpad: will help us figure out what IBIS needs to specify Mike: will allow EDA vendors to develop parameter GUI Walter can check compatibility with his system once he knows I/O conventions. Also will begin invesigating how difficult it may be to build these types of models. Arpad: we need to be clear about types of waveforms and not just say "waveform". Need specific labels. Arpad: Does the driver wiggle it's outputs in order to generate the impulse response? -Kumar seemed to imply that the COULD be a way to generate the impulse response. AR: Kumar will go through a tutorial example of this whole process. Arpad asks that it be very clear with respect to terminology, such as exactly what the "waveforms" are, how they are generated, what is being passed. ------------- Next meeting: Tuesday 7 Nov 2006 12:00pm PT