IBIS Macromodel Task Group Meeting date: 11 February 2014 Members (asterisk for those attending): Agilent: Fangyi Rao Radek Biernacki Altera: * David Banas ANSYS: * Dan Dvorscak * Curtis Clark Luis Armenta Cadence Design Systems: * Ambrish Varma * Brad Brim * Kumar Keshavan Ken Willis Ericsson: Anders Ekholm Intel: * Michael Mirmak Maxim Integrated Products: Hassan Rafat Mentor Graphics: * John Angulo * Arpad Muranyi Micron Technology: Randy Wolff * Justin Butterfield QLogic Corp. James Zhou Andy Joy SiSoft: * Walter Katz Todd Westerhoff * Mike LaBonte Teraspeed Consulting Group: Scott McMorrow * Bob Ross The meeting was led by Arpad Muranyi ------------------------------------------------------------------------ Opens: - Mike: The attendee list has been pruned to retain only those joining since Jan 1, 2013. -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Walter collect user statements regarding their packaging needs. - This needs to wait until a combined solution is defined. - We will consider this AR tabled. - Brad has asked package vendors for this information. ------------- New Discussion: - Arpad: It was suggested that I meet with Walter to resolve issues. - In fairness all co-authors should be involved, maybe other company members. - That might lead to having more people than expected. Back-channel: - Ambrish showed a proposed back-channel BIRD . - Markup colors are relative to the last posted version. - Ambrish: Several sections define bit patterns. - The AMI file has Training and Backchannel_Protocol Reserved_Parameters. - The BCI string resembles the other ones passed to/from AMI DLLs. - It has data for setting TX tap settings. - Walter: Some protocols will have different methods. - For example, percent instead of coefficient. - Ambrish: The TX informs when no more changes can be made, training is done. - Arpad asked to keep questions and answers short. - Michael M: Agree with Walter, for example PCIe does it differently. - They do not increment and decrement. - Walter: We may need a cookbook to show how to handle various protocols. - Michael M: How deep will we get into protocol awareness? - Walter: AMI should be concerned with the channel payload, the TX to RX bits. - Kumar: We should solve this at the physical layer. - Walter: I think we agree. - We should be concerned about adding new types and formats. - Ambrish: This is cleaner than space delimited strings. - Arpad: Will this BIRD be submitted to the open forum? - Ambrish: Yes. - Mike L: Will this work only between a specific TX and RX? - Ambrish: It will work with any TX and RX using the same protocol file. AR: Ambrish send updated backchannel BIRD to Mike for posting. AR: Ambrish send clean backchannel BIRD to Michael M for open forum. Package model discussion: Michael M: We need to pick one of the two proposals that have been made. - Mike Steinberger noted that we have a choice whether to refactor software or change incrementally. - John: Refactor actually means to rewrite existing functionality. - Michael M: One question is if we should rework a keyword that is not commonly used. - Keywords like [External Circuit] raise the expectation of supporting AMS languages. - If it can be another name then we might make both proposals work. - Ambrish: We could create a subparameter to indicate that only ISS can be used. - Walter: We should pursue unifying our proposals offline. - A cookbook written in parallel would help. - John: Mentor has introduced BIRDs, SiSoft has presented a methodology. - The methodology may not be complete and may be difficult to use. - It will have to say how both the users and tools will use it. - Arpad: How will the tool automate this for the user assigning models to buffers? - John: We are getting generalized models from vendors. - They may not say which pins provide worst case coupling. - Will model choices be made automatic? - Michael M: The main concern is post-layout? - John: No, pre-layout is just as important. - Walter: In prelayout the assignments would be more manual. - Arpad: How will the tool know how to do the connections? - John: The IC vendor has not given the information needed. - Walter: One approach is for the vendor to supply wide models. - Another is for the EDA tool to derive them, possibly using swathing. - John: There is much to discuss about how this will work in practice. - Michael M: It seems only a buffer model and some package info is needed, no IBIS. - John: A plug and play solution could be made if we provided victim level information. - Michael M: That is easy for the EDA tool, but not the model maker. - We might be creating a solution that does not require IBIS. - Arpad: We need to focus on the user. - There will be more users than model makers. - Walter: There can be chips with 128 SerDes. - Users need IBIS files that allow them to analyze that for noise. - They may find that a model is insufficient, will need a bigger one. - Touchstone 3 sparse models will help. - John: Vendors should be willing to create s12p models, for example. - Brad: It is good that we have been discussion methodology, not a BIRD. - The problem today is I have models I can't use with IBIS. - Things are done manually first, before automation. ------------- Next meeting: 18 February 2014 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives