IBIS Macromodel Task Group Meeting date: 11 Sep 2012 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 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 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 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 Ray Anderson Note: The name "Abhishek Munjal" appeared in NetMeeting but no corresponding voice was heard. The meeting was led by Arpad Muranyi ------------------------------------------------------------------------ Opens: - None -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Michael M provide examples to show what IC vendors need - Done with part 1 - Need questions answered to do part 2 ------------- New Discussion: Michael M showed "Package Model Format Needs": - Michael M: This is part 1 - slide 3: - Michael M: We address two types of analysis and three user segments - We try not to put a large burden on any single segment - slide 4: - Michael M: For post-layout user may have specific nets to look at - They may not have been identified by name - May want to look at signals with crosstalk over some threshold - Besides format, there is a question if the same data is useful pre and post layout - slide 5: - Michael M: Three ways to look at the problem: - Full model - Small swath model - Many small models - Need to look at how each user sees each type - Swathed models will be hard for EDA vendors to develop - slide 6: - Michael M: The model maker controls the destiny of the other two - The ideal case is where the end user controls what is extracted - That will be hard to do - Arpad: Why is that difficult? - Michael M: The person controlling distribution limits what it will cover - Small models are more work to create, easy to use - Large models are easy to make, hard to use - back to slide 5: - Walter: On large models the through paths must be accurate - There are limited coupling paths that must be examined - Sparse formats will reduce file size - The EDA tool can choose what to use - S20p models can be quickly extracted from s1000p models - Michael M: That is entirely in the context of s-parameters - Am I automatically limited? - Walter: ??? - Michael M: Then the extraction tool should be able to use coupling info for that - James: For most buses the highest pin count is about a thousand - Probably need 1 to 5 aggressors max - The standard should allow arbitrarily large models - Tools should be able to make it easy for end users - Michael M: We don't want the standard to eliminate possibilities, agree - But some customers create "unusual" designs where crosstalk is more of a problem - Time domain models other than s-parameters are an issue too - Walter: Some simulators have trouble with long chains of short s4ps - Package models should not have many of those - Michael M: For part 2 a swathed s12p example might be good - Walter: Some vendors deliver s4ps and expect them to be combined - This might be an alternative to swathing - Brad: It's easy to violate passivity with those - James: It is easy to measure return losses and combine into an s16p - Radek: When we neglect far coupling we sacrifice accuracy - This has to be well understood - Usually we only have full matrices - Michael M: Has anyone tried creating an sZILLIONp model from lab data? - Equipment requires tricks to do that - Walter: It has been done up to 150 - James: We have seen around 130 - Arpad: Have we seen this done? - Michael M: I have not seen a case where sparse data is used for full simulation - Arpad: Walter may have an example - Walter: I emailed one, but it only has about 20 pins - Scott has also proposed creating s1000p models - Michael M: We should err on the side of giving as much data as possible - Reducing data for specific applications may not help - Walter: My format has no inherent size limitation - Model makers will be aware of tool limitations - Arpad: BIRD 125 can be modified to meet these needs - We need to answer some questions before proceeding - Michael M: The hybrid solution may be best - The existing spec has interactions that are hard to understanding - A simpler IBIS spec that connects pins/pads to other things would help - A pad keyword would be better - Walter: I would like to show the email I sent - Arpad: Maybe we should reorder our priorities first to get other BIRDs out of the way - We need to address the AMI analog model problem - Could packaging be handled by a subcommittee - Michael M: The interconnect task group could come back - Arpad and Walter liked the idea AR: Michael M restart interconnect task group meetings Wednesdays 8am or 9am PT Walter showed an example IBIS file and example package tree syntax: - Walter described the tree syntax - Walter: Corners might be limited to three, although more might be needed by some - This might be called a mixed model - Walter showed a full package model example tree syntax - In this one each pin is a victim - It shows the power and ground pins - This is a simple, incremental change to IBIS - A separate file might also describe the on-die models - James: Are these competing against each other? - Walter: One might use one where there is crosstalk and another elsewhere - In pre-layout there may not be pin names - They are three views of the interconnect - James: The semantics are different - Walter: Yes Arpad: Would interconnect meetings start tomorrow? - Walter: I think we have to assume next week - James: We have to consider the model maker burden of these syntaxes - For the user it may be hard to choose which model to use - End users should choose how many aggressors, etc. - Walter: A full model would be the only source data needed - This single syntax allows any use we want - Brad Brim might want to add XY coordinates ------------- Next meeting: 18 Sep 2012 12:00pm PT Next agenda: 1) Task list item discussions ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives