IBIS Macromodel Task Group Meeting date: 12 August 2014 Members (asterisk for those attending): Agilent: Fangyi Rao * Radek Biernacki Altera: David Banas ANSYS: * Dan Dvorscak * Curtis Clark Avago (LSI) Xingdong Dai 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 Synopsys * Rita Horner Teraspeed Consulting Group: Scott McMorrow * Bob Ross The meeting was led by Arpad Muranyi. ------------------------------------------------------------------------ Opens: - Michael M: The interconnect meeting tomorrow is canceled. -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Arpad to review IBIS spec for min max issues. - In progress - Brad check within Cadence about overclocking approaches - No report - Arpad - Work with Randy to update BIRD 173. - In progress, maybe next week - Bob and Ambrish - Update BIRD 147. - Can present results today ------------- New Discussion: MinMax BIRD: - Arpad showed a BIRD draft. - Arpad: All spec sentences with min/max are copied here - "NA" means typical is substituted. - Certain places require NA to be used. - If min or max is missing for anything the entire file should have NAs - Mike L: It should be the entire model, not file. - Bob: It's too hard to change a lot of models - Radek: This would be for new models. - Walter: This makes sense for I-V curves but not other parameters. - You might have dv/dt for all corners and one I-V curve. - Arpad: Can a min measurement be used with a typical simulation? - Bob: Threshold values? - Some are fixed, some vary with supply. - Walter: An across the board rule is scary. - Arpad: It seemed last week we wanted an across the board simple rule. - One question is if there should be one statement in the spec or in many places. - Bob: There are both OK and questionable cases, depending on the parameter combinations. - We should have required 3 columns from the beginning. - Arpad: We explicitly called for typical to be substituted for NA. - Walter: Some things like current could be the same for all corners. - Arpad: Using typical I-V will not match V-T min or max. BIRD 128: - Ambrish: The update has been forwarded to Michael M for posting as 128.1. BIRD 147.1 and co-optimization: - Ambrish showed draft 11 of the BIRD 147.1 update. - Ambrish: File prefixes are now lowercase, and "ibis_" is reserved.. - Description changes have been made. - "BCI" root name is reserved, private parameters must use another root. - Walter: What tells us that AMI_Init will do optimization? - Ambrish: The string returned will have (BCI ...) - Walter: I would like a Boolean to say that it is supported. - Ambrish: The Training parameter conveys that. - Walter: Can it do both stat and TD? - Ambrish: No - The AMI files have 0, 1 or 2. - Both TX and RX have to have the same value to do training. - Walter: What is suggested here is not the only way to write a protocol. - Ambrish: Other protocols would not be IBIS approved. - Todd: What makes it IBIS approved? - Ambrish: THE BCI file is created by the IBIS committee. - Walter: The strings produced by the models would have to be approved. - This mandates how statistical analysis must be done. - There may be other ways. - Ambrish: The BCI only gives the parameters. - Todd: There is no checking of the strings produced by the model. - We have no enforcement today of how DLLs use the strings passed. - There is an assumption the models do what BCI says. - Walter: 1.7.1 should say "Here is one way ...", but not be explicit. - Ambrish: This gives the IBIS way of doing it. - Walter: I reject that. - Todd: Ambrish is saying the only comm method is to change taps iteratively. - Ambrish: That will be the only IBIS approved way to do it. - Todd: If the RX tells the TX to change 3 units it needs three calls. - Ambrish: Send (-1 3) - Walter: Where does it say I can do that? - How does the RX know how far it can go? - Todd: It can request a change, but that not not be implemented. - How does the RX find out how much room there is to adjust? - Walter: Does the protocol give the limits? - Ambrish: No - In time domain we only have increment and decrement - Walter: PCIeG3 says it can send tap values. - A BIRD would be needed for that. - This BIRD could simply allow for other ways. - Ambrish: That is valid. - Or we could include both methodologies. - Walter: There can be more ways. - We should not need a BIRD for each. - It should be part of the BCI approval process. - This BIRD is forcing a methodology - Todd: The BIRD proposes no enforcement so it's unenforceable. - Walter: The approval process should not be constrained by these rules. - Ambrish: Private protocols are allowed. - Walter: Can a jitter knob be IBIS approved? - Ambrish: No protocol talks about a jitter knob. - Todd: Only BCI files get approved. - We are pretending this is a standard but it's not. - Ambrish: We have to specify this so the other side knows what we are sending. - Todd: If a DLL doesn't respond correctly it won't work. - This will be more hidden. - Todd: There are protocols that communicate coefficients, not integers. - Ambrish: That can be added. - Walter: Or we could simply allow for other ways to communicate. - Ambrish: Model makers would then have to read 2 specifications. - Walter: But the EDA tool would not have to. - Todd: This should be considered as a change to the BIRD. - Ambrish: We will discuss it. ------------- Next meeting: 19 August 2014 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives