IBIS Macromodel Task Group Meeting date: 23 February 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 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: - None. -------------------------- Call for patent disclosure: - None. ------------- Review of ARs: - None. ------------------------- Review of Meeting Minutes: - Arpad: Does anyone have any comments or corrections? [none] - Radek: Motion to approve the minutes. - Arpad: Second. - Arpad: Anyone opposed? [none] ------------- New Discussion: De-embedding C_comp IBIS-ISS Subcircuits: - Walter: [sharing presentation] - Overview: - What is C_comp De-embedding? - It refers to the tool's process of taking the v(t) tables and deriving the K(t) scalers that describe the switching behavior of the pull up and pull down devices. - What if simple C_comp is replaced with a more complicated IBIS-ISS subcircuit model? De-embedding becomes much more difficult. - Proposed new approach for use with IBIS-ISS C_comp models: - Model maker sweeps the value of a simple C_comp to generate K(t) functions. - These K(t) functions are used with the IBIS-ISS C_comp to see which ones produce v(t) waveforms that best match the actual waveforms. - Take the best K(t) waveforms (KT_fit) and generate v(t) waveforms using a simple C_comp=0.0 (VT_fit). - The finished IBIS Model has VT_Fit, C_comp=0.0, and the C_comp IBIS-ISS subcircuit. - The EDA tool takes the v(t) waveform (VT_Fit) and the C_comp (0.0) and generates K(t) in the usual manner. This K(t) is then applied with the C_comp IBIS-ISS subcircuit at simulation time. - Discussion: Walter noted that he wasn't sure if we'd used the term "de-embedding" in this context before. Arpad replied that we had used the term "C_comp compensation" for this topic. Arpad summarized this proposed method as a way for the model maker to find a v(t) curve that is not laden with the C_comp subcircuit but is such that once the C_comp subcircuit is applied the actual v(t) waveform is recovered. Walter considered this an excellent summary of the idea. Radek noted that in fact we would very likely only hope to come "close enough" to recovering the actual waveforms. Walter agreed with this statement, but noted that C_comp's role in shaping the edge of the initial waveform entering the channel was only one aspect of its importance. The shape of the rising edge could be created with virtually any C_comp, but the return loss when a wave were reflected back would be dramatically wrong without a reasonable C_comp model. Bob agreed with the basic idea of Walter's proposal. He suggested that perhaps the model maker, given that they know the contents of the C_comp IBIS-ISS subcircuit, might be able to define a driving waveform at the input of the C_comp subcircuit that would yield the exact v(t) waveform at the output. This could then be folded into the IBIS model, without having to go through the KT_Fit procedure. Bob said he would have to think about potential details more carefully. David asked if the idea was to get around having to modify the IBIS standard to let the model maker specify the K(t) directly. Walter said that this would have that effect, but that the primary goal was to find a way to avoid having to burden the EDA tool with deriving the K(t) given a general C_comp model. Arpad noted that there were more model makers than tool makers, and that it might make sense to have the tools take care of this issue. Walter countered that the EDA tools would have to perform this computation at the start of every simulation, and that while it was "easy" with a regular C_comp, an algorithmic approach or something else that was time consuming would be necessary for a general C_comp model. He felt that it would be best to have the model maker do it once and not to burden tools with this if we introduced a general C_comp. S-parameter port numbering: - [With 25 minutes remaining and no existing topics ready for further discussion, Arpad made a motion that we discuss some s-parameter port numbering topics that had been under discussion in the interconnect task group] - Discussion: Arpad was interested in discussing the actual circuit topology implied by the N ports, N+1 terminals (common reference) shortcut specified in the current interconnect proposal. Arpad, Radek, Walter and David discussed issues including when and/or if it is necessary or useful to go to a 2*N terminal approach, preservation of common mode information, and when vendor provided "signal pin only" sNp models with an implicit use of node zero as the reference terminal were valid. The main discussions on this topic can be found in the interconnect task group meetings. For interconnect task group minutes on this topic: http://www.freelists.org/archive/ibis-interconn/02-2016, refer to the posted minutes from the February 17th and 24th meetings. - Arpad: Thank you all for joining. ------------- Next meeting: 01 March 2016 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives