IBIS Macromodel Task Group Meeting date: 08 March 2016 Members (asterisk for those attending): ANSYS: * Dan Dvorscak * Curtis Clark Broadcom (Avago): Xingdong Dai Bob Miller Cadence Design Systems: * Ambrish Varma Brad Brim Kumar Keshavan Ken Willis Cisco: Seungyong (Brian) Baek eASIC: * David Banas Marc Kowalski Ericsson: Anders Ekholm GlobalFoundries: Steve Parker Intel: * Michael Mirmak Keysight Technologies: * Fangyi Rao * Radek Biernacki * Ming Yan 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 Teraspeed Labs: * Bob Ross TI: Alfred Chong The meeting was led by Arpad Muranyi. -------------------------------------------------------------------------------- Opens: - Ming Yan of Keysight Technologies introduced himself to the ATM group. He works closely with Fangyi and Radek, and he has been working on IBIS and AMI modeling for 10 years. He has previously attended Open Forum meetings and summits, and he was co-author of a paper on IBIS overclocking issues. -------------------------- Call for patent disclosure: - None. ------------- Review of ARs: - None ------------------------- Review of Meeting Minutes: - Arpad: Does anyone have any comments or corrections? [none] - Michael M.: Motion to approve the minutes. - David: Second. - Arpad: Anyone opposed? [none] ------------- New Discussion: New Redriver Flow BIRD: - Fangyi: [Sharing an updated version of his presentation "AMI Simulation Flow Round 3"] - Thinking about ways for the tool and model to work together. - Introduce two new Reserved_Parameters: - Simulator_Supports_Augmented_Rx_Init_Impulse_Matrix - Optional, Boolean, Type Input, Default value False - Used by the EDA tool to tell the model if it supports the new flow. - Rx_Init_Supports_Augmented_Impulse_Response - Optional, Boolean, Type Info, Default value False - Tells the EDA tool if the model supports the new flow. - Three scenarios in which the new flow is supported by the model, the tool or both. - 1. Model does not support new flow. - Tool follows existing flow. - 2. Model and tool support the new flow. - Tool tells Rx model it is running the new flow. - Tool provides augmented IR matrix. - Tool follows new flow. - 3. Model supports new flow, tool does not. - Tool does not indicate that it follows the new flow. - Tool passes legacy IR matrix to the Rx model. - If model supports legacy flow, then it follows it. - If model does not support legacy flow, its AMI_Init() function returns an error. - Discussion: Walter noted that he would prefer not to have "Supports" in the names of the new parameters. He also stated that a single parameter could be used for the model to indicate support to the tool and vice versa. Radek agreed that a single parameter should be sufficient. Fangyi reviewed the slides detailing the augmented flow proposal (the slides originally presented in the Feb 16th meeting). Walter noted that if we were going to consider augmenting the IR matrices and fully revamping the flow, he would have some suggestions for augmenting the IR matrix for the Tx AMI_Init() as well. With respect to the "Normal Time Domain Flow: Get Wave Tx" (slide 4), Walter had several questions/proposals: - One of the three things the Rx AMI_Init() returns is the non-DFE portion of the Rx convolved with the analog channel IR. Instead, why not simply return the non-DFE portion of the Rx itself? The tool can do that convolution instead. - The three things returned by the Rx AMI_Init() would then be the DFE portion of Rx, the non-DFE portion of Rx, and the total containing the entire Rx applied to the entire input IR. - The third returned item is just what we have with the existing flow. - We would no longer need to pass the hRx,in partial (analog channel IR itself). - This simplified flow would make it easier for a model to support the legacy and new flows. On first reflection Fangyi agreed with Walter's comments and said the flow could be reduced in this manner. Mike LaBonte then asked if not having the analog channel IR combined with the non-DFE would deprive the model of any special processing needed to align the cursor. Fangyi said that was a good question, and that different models would have different ways to locate the main cursor dependent on their CDRs. For this reason, it might be necessary to pass the analog channel IR to the model. After the review, Walter noted that Fangyi's proposal would solve the original problem of an Init() only terminal Rx() not being passed the complete IR, including upstream effects, it needs for its optimization. However, he said this proposal still fell short if the redriver's Tx wanted to optimize itself. Fangyi agreed that this was true, because the redriver Tx was not presented with the effects of the upstream channel. Walter noted that this was a separate issue. He said it was the reason his proposal contained a new parameter indicating whether the Tx optimizes itself. - Arpad: Thank you all for joining. ------------- Next meeting: 15 March 2016 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives