IBIS Macromodel Task Group Meeting date: 29 April 2014 Members (asterisk for those attending): Agilent: * Fangyi Rao * Radek Biernacki Altera: * David Banas ANSYS: * Dan Dvorscak * Curtis Clark Cadence Design Systems: * Ambrish Varma Brad Brim * Kumar Keshavan * Ken Willis Scott Huss Ericsson: Anders Ekholm Intel: * Michael Mirmak LSI Amaresh Malipatil Dai Xingdong Maxim Integrated Products: Hassan Rafat Mentor Graphics: * John Angulo * Arpad Muranyi Andrey Matvienko Vladimir Dmitriev-Zdorov 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: - Ambrish: I had requested time to present today. - Walter: I sent example models to everyone, would like to ask questions about that. - Walter: The redriver flow BIRD should be on our agenda in a future meeting. -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Walter send backchannel example files, Mike post these. - Done ------------- New Discussion: Questions about Walter's presentation last week: - Bob: This requires no BCI file? - You introduce a tap index? - Walter: The TX is independent. - Having a tap increment implies there is a tap index. - I added a parameter to identify it. - Kumar: Increment only means increase or decrease equalization, it does not imply an increment. - Walter: I can send pointers to standards adreesing this. - Are there any questions about the models I sent? - My proposal satisfies the 3 requirements I described. - Ambrish: What about future architectures? - Walter: RXs are complicated, but every standard requires the same things for TXs. - But a TX doing something else could be invented. - Ambrish: The models can go above and beyond what the protocol requires. - Walter: In your files only the established TX features are used. - Kumar: The TX can implement something that does not follow an equation. - The RX will tell it only to increase or decrease, nothing else. - The TX implements it however it wants. - Walter: The TX can accept increments or coefficients, it works both ways. - Kumar: The RX should make no assumptions about how the TX operates. - Walter: There are generic TXs. - Kumar: This is trying to legislate an architecture. - Todd: Walter only suggests that taps and intervals are common. - Ken: A premise of AMI is the algorithmic model is a black box. - Todd: Walter is saying the BCI file adds complexity to keep it out of the AMI file. - Ken: The BCI file was to avoid new Reserved_Parameters. - We would have faster resolution, no need for new BIRDs. - Todd: We would need a committee to approve BCI files. - Ken: It should be this committee. - Todd: We will have to establish a new regime for controlling the BCI specs. - Ken: That is one aspect of the proposal. - Walter: To approve a BCI standard: - Training patterns, that is simple. - New keywords are more involved. - Ambrish: Those need to be only in the string passed to the TX and RX. - Walter: The only content in those strings will be tap increments and coefficients. - That will be true for 5 years. - Even PCIe3 and 802.3bj have nothing newer. - This proposal is to support Cadence IP. - Why should other EDA companies support it? - Ambrish: We have complex models that support multiple protocols. - The protocol definitions are the best way to handle that. - Todd: I don't think we captured those requirements. - Walter: Right now there are only two protocols. - I motion to vote on whether models should support multiple protocols. - This should be asked separately of TX and RX. - SiSoft will accept BIRD 147 if so. - Kumar: It is common to have the same architecture for different protocols. - Why is the BCI file a contentious issue? - Walter: 100G is four 28G copper lines. - 400G Ethernet will be 16x28G electrical. - Training will never change for 802.3. - There is no change from PCIe3 to PCIe4, and there will be no PCIe5. - Where will new protocols come from? - Todd: We need proof that new requirements are needed. - Ambrish: We have documented the need. - Todd: If you can't talk about it. - Michael M.: Are we really saying a proprietary interface would not be supported? - Companies might agree to make something work a certain way, independent of a formal protocol. - Arpad: With proprietary you can't talk about it, how can you make a standard? - Walter: You could agree to a private BCI file if the EDA tool knows nothing about it. - My proposal would not work for proprietary protocols, BIRD 147 would. - It would be straightforward to add that capability to my proposal. - A message string can be added. - Kumar: We are not anticipating that. - The communication between TX and RX should be as simple as possible. - Abstracting it that way allows for applications we were not aware of. - Todd: So we need to add: - multi-protocol. - a facility for private messaging. - Kumar: When abstracted this way the EDA tool only needs to enable handshake. - Walter: We would like to have it work with a TX that was not designed for a protocol. - This helps with design optimization. - That market is important to us. - Ambrish: That could be encapsulated in a BCI file. - Walter: It must work with any TX. - Kumar: BIRD 147 does not propose anything proprietary. - How does the EDA tool come into the picture? - Walter's proposal expects the TX to work on some equation. - Arpad: Is this a question whether to use AMI for the same thing as BCI? - Walter: This is about a solution for TXs that all work about the same way. - Some expose indexes, some coefficients, and some both. - They will work with any RX that understands the same parameter. - We should have thought about this 5 years ago. - Arpad: It should be done through AMI parameter? - Walter: Yes. - Ambrish: It should be in a file that both the TX and RX support. - Arpad: How do we make a decision? - Walter: The other EDA vendors should say what they want. - Todd: Has Cadence presented a BCI file? - Ambrish: No, just an abstracted example. - Walter: There is a BCI for PCIe3? - Ambrish: Yes. - Todd: I would like to see BCI files for the KR and PCIe3 standards. - This should be the first topic discussed next week. ------------- Next meeting: 06 May 2014 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives