IBIS Macromodel Task Group Meeting date: 24 March 2015 Members (asterisk for those attending): Altera: David Banas ANSYS: Dan Dvorscak * Curtis Clark Avago (LSI) Xingdong Dai Cadence Design Systems: * Ambrish Varma Brad Brim Kumar Keshavan Ken Willis eASIC David Banas Ericsson: Anders Ekholm IBM Steve Parker Intel: Michael Mirmak Keysight Technologies: * Fangyi Rao * Radek Biernacki * Nicholas Tzou Maxim Integrated Products: Hassan Rafat Mentor Graphics: * John Angulo * Arpad Muranyi Micron Technology: * Randy Wolff Justin Butterfield QLogic Corp. James Zhou Andy Joy eASIC Marc Kowalski SiSoft: * Walter Katz * Todd Westerhoff * Mike LaBonte Synopsys Rita Horner Teraspeed Consulting Group: Scott McMorrow Teraspeed Labs: * Bob Ross (Note: Agilent has changed to Keysight) The meeting was led by Arpad Muranyi. ------------------------------------------------------------------------ Opens: - Arpad: I have a spec clarification issue. -------------------------- Call for patent disclosure: - None ------------- Review of ARs: - Michael M update AMI Directionality BIRD - Arpad: He reported this is in progress. - Walter and Randy produce C_comp BIRD. - Randy: Some progress, but we need to resolve issues about parameters. - Arpad to review IBIS specification for min max issues. - In progress. ------------- New Discussion: PAM4 BIRD: - Walter: Thanks to all who submitted comments. - A list will be allowed to select NRZ or PAM4. - There will be upper and lower timing offsets. - The tool returns an impulse or waveforms, not an eye. - Duobinary is a question. - The keywords are named to not conflict later. - Arpad: Should we number the thresholds? - Walter: We called them PAM4_Upper, PAM4_Lower, PAM4_Center. - Each scheme could have its own names. - Mike: If NRZ and PAM4 are selectable, numbered thresholds might conflict. - Walter: We should not have Duobinary until it is in use. - Fangyi: Agree with Walter. - An IC vendor suggested we use Rx_Clock_Recovery_Mean for the middle eye. - Walter: In GetWave we get center eye clock_ticks. - Offset would be better for that. - For statistical would be offset from the mean. - Fangyi: In TD offset is from the mean. - Walter: SO we have upper and lower eye mean? - Fangyi: In statistical it would be offset from the recovered clock. - Walter: If the mean is 10ps from center, by default I use - Normally there is only one clock for all 3 latches. - There is no offset in time domain, you use clock ticks. - A PAM4 has only one clock recovery. - Upper and lower may have different phases. - One offset for each segment should apply to TD and stat. - Fangyi: So the offsets are added to the center mean? - Walter: Yes. - Fangyi: So clock_times is for the center eye. - A numbered scheme would complicate that. - Todd: I like the idea of numbering the eyes. - Arpad: We could have zero, positive, and negative numbers. - Todd: We have to know which eye the clock_times pertain to. - Walter: A lot of places use upper and lower. - If PAM5, PAM6, PAM7 were coming things would change. - Mike:New modulations will have their own prefixes. - Fangyi: We need to correct "bit time" - Walter: Agree, it should be "symbol time". - Fangyi: ?? should be an integer. - Walter: Agree. Co-optimization: - Walter: No update for today. - We should avoid dynamic modulation switching. - We simulate at PHY level, not MAC. C_comp: - Randy: Curtis Clark sent comments. - I added nodes for a differential receiver. - Curtis had questions about how different simulators would handle it. - Curtis: Walter said these models complicate extracting K(t). - Bob's presentation shows how to move the measurement point. - Maybe the K(t) should be in the IBIS file. - Walter: The format would be simple. - There would be four curves. - Simulation would be easy. - This may be harder for model makers. - Arpad: Without V-T tables verifying would be harder. - With the V-T load you should get the same waveform. - You could have both. - Radek: They might not agree. - Arpad: Tools derive K(t) already , why have model makers do it? - Walter: C_comp makes it hard to get it right. - Curtis: Bob created a complicate system to get K(t). - Walter: Tools reinvent K(t) every time they simulate. - Mike: Wouldn't K(t) tables eliminate current differences among tools? - Walter: Yes. - Arpad: We should finalize the C_comp proposal before this. - Arpad showed [Package] in the IBIS spec. - Arpad: We have a series of overrides. - In [Define Package Model] [Pin Numbers] it does not say they all have to be in [Pin Names]. - The [Package] might be a partial model. - The remaining pins might use defaults. - For example Power and GND pins could be omitted from [Pin Numbers]. - The spec doesn't say what to do with them. - Tools handle this differently. - Some create a short, some leave it open. - Mike: Can [Define Package Model] have both [Pin Numbers] and matrix formats? - Bob: No they are mutually exclusive. - A missing [Define Package Model] pin would fall back to [Pin] RLC, then to [Package]. - Arpad: If no [Pin Mapping] there would be straightforward pin to pad paths. - Why describe 10 power pads when I can describe one and leave the rest disconnected. - That would rely on [Pin Mapping] to connect them. - It assumes shorts would not be created. - Some tools will create shorts though, eliminating the model. - Bob: A rail might be shorted to nine other rails through buffers. - Randy: I get this with my models. - With a power plane 10 balls are connected to the package. - We shorted the powers together, same for grounds. - In [Pin Mapping] all 10 pins are there but the package model has only one. ------------- Next meeting: 24 Mar 2015 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives