IBIS Macromodel Task Group Meeting date: 4 June 2019 Members (asterisk for those attending): ANSYS: Dan Dvorscak Curtis Clark Cadence Design Systems: * Ambrish Varma Ken Willis Intel: Michael Mirmak Keysight Technologies: Fangyi Rao Radek Biernacki Ming Yan Stephen Slater Maziar Farahmand Mentor, A Siemens Business: * Arpad Muranyi Micron Technology: * Randy Wolff * Justin Butterfield SiSoft (Mathworks): * Walter Katz * Mike LaBonte SPISim: * Wei-hsing Huang Teraspeed Labs: * Bob Ross The meeting was led by Arpad Muranyi. Mike LaBonte took the minutes. -------------------------------------------------------------------------------- Opens: - Mike LaBonte offered to take minutes. ------------- Review of ARs: - Randy to update the C Comp Model draft with today's discussed changes and send it to ATM for review. - Done. Mike LaBonte had posted not the May 21 version, but the May 30 version with updates from the meeting. Bob Ross asked if that was the latest version. Mike said it was. -------------------------- Call for patent disclosure: - None. ------------------------- Review of Meeting Minutes: Arpad asked for any comments or corrections to the minutes of the May 28 meeting. Randy Wolff moved to approve the minutes. Walter Katz seconded the motion. There were no objections. ------------- New Discussion: Complex C_comp modeling: Randy Wolff said for a C_comp model the driving/non-driving mode must be set. We need to decide what should be done if there is no [C comp Model] defined for non-driving mode. Arpad proposed falling back to the original C_comp for non-driving mode. Randy proposed Bob said an I/O model could always use to full C_comp if non-driving mode is not defined. He asked how this would affect Submodel. Arpad said Submodel C_comp was not used. Bob said a Submodel for the driver or receiver might select a different capacitance. Walter said C_comp would be used for generating the k-t function, not a subcircuit. Therefore C_comp must be kept for driving mode. A receiving mode C_comp is needed. Arpad asked if this would be in addition to the main C_comp or a substitute. Walter said it would be a substitute. Bob said the receive C_comp is the more useful one, because important reflections come from there. Walter felt they were both important. Arpad asked if it would make sense to fall back to original C_comp, because it is required. Randy said [C Comp Corner] is required with [C Comp Model]. We agreed to falling back to [C Comp Corner] values for missing non-driving C_comp. BIRD197.3_draft_4(DC_Offset): Arpad Muranyi showed a DC_Offset presentation from Fangyi Rao. Arpad said Fangyi was unable to discuss it this week. Walter Katz said in approach 1 shown on slide 1, the latch was essentially centered around 0V. In approach 2 the latch was at some non-zero voltage. The model would be informed which approach was used and what the offset should be in approach 1. Slide 2 showed an arrangement with a comparator, with the reference voltage provided. The VrefDQ would not be exact because of step size, and the model would not know if it was too high or too low relative to the ideal offset. This could be compensated in the model. Arpad said the presentation could be sent the email list. Randy Wolff said comparing DDR5 waveforms with no DFE to IBIS-AMI output could only be done with level adjusted output. Arpad agreed that some work was needed to overlay waveforms. Walter felt tools should have no trouble implementing that. Ambrish Varma asked if the voltage would be the same as with a non-AMI model. Walter said it would not. Ambrish said the tool would shift the voltage level back as needed. Walter agreed. Arpad noted that getting the right amount of shifting was necessary. Walter said the small 3 or 4 mV shift due to resolution would be handled. Arpad noted that the presentation was using 0.4V shift. Walter said controllers would determine the right voltage by sweeping up and down until it finds errors, finding some mid-point. The mapping of register values to voltage may be non-linear, and the DLL does not know about that. Arpad suggested models that report actual offset voltages should be supported. Walter said that could be done. The model could output the voltage value for the DQ register setting it was using, and the tool could use that as the threshold. Bob Ross asked if the this would involve rejecting the BIRD. Arpad noted we only have a BIRD draft for BIRD197.3. Walter said our choice would be only whether to submit it. Ambrish agreed. Digital signatures: Walter said some companies would prefer not not to accept unsigned DLLs. Mike LaBonte said signing should up to model makers, IBIS need not do anything. Walter suggested we could make a simple statement that DLLs could be signed. Arpad wondered if we should offer some kind of certification service. Walter suggested IBISCHK could report simply if a signature is found, no more. Arpad asked who does the signing. Could IBIS do it? Ambrish asked if documents were being signed for this. Mike said the digital signatures were embedded in executable file headers, and that the certificates could be seen on Windows by inspecting the Security properties of an executable or DLL. IBIS would have difficulty obtaining a certificate because we are not an incorporated entity. Also the models would be "from IBIS" in that case, not the vendor. Wei-hsing said it would be necessary to pay for certificates, since self-signed models would be rejected. Yearly renewal payments was required for certificates. - Ambrish Varma: Motion to adjourn. - Walter Katz.: Second. - Arpad: Thank you all for joining. ------------- Next meeting: 11 June 2019 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives