IBIS Macromodel Task Group Meeting date: 17 January 2017 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 IBM Luis Armenta Trevor Timpane 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 Kevin Li Teraspeed Consulting Group: Scott McMorrow Teraspeed Labs: * Bob Ross TI: Alfred Chong The meeting was led by Arpad Muranyi. -------------------------------------------------------------------------------- Opens: - Arpad noted that we will meet on Jan 24th, but we will cancel the meeting on Jan 31st because it conflicts with DesignCon. ------------- Review of ARs: - Bob Miller to send an update for BIRD 147.5. - Done. - Michael Mirmak to send an update for 187.1. - Done. -------------------------- Call for patent disclosure: - None. ------------------------- Review of Meeting Minutes: - Arpad: The minutes were just sent out a few minutes ago. - Michael M.: Motion to delay review of the minutes until next week. - Mike L.: Second. - Arpad: Anyone opposed? [none] ------------- New Discussion: New Redriver flow BIRD: - Fangyi: [Sharing AMI Simulation Reference Flow Enhancement BIRD draft 0] - Some members of the Broadcom modeling group now included in the discussion. - This is a BIRD draft version of ideas that had previously been given in ATM presentations. - It is a lengthy BIRD. - Requirements: - Current problem: The terminal Rx only sees the IR information for its nearest section of the channel. It doesn't see the effects of the upstream sections of the channel. - Current problem: 6.1 Flow, simple Tx to Rx, right now we rely on deconvolution if the Tx has GetWave() and the Rx doesn't. As we've discussed before, related to the redriver issue, we conclude that deconvolution is not the best way. - Mike L.: In your BIRD's requirements section, could you note that deconvolution is the current solution? - Fangyi: Yes. - [reviewing table of changes] - Flow is modified for the repeater case and the non-repeater case. - Impulse matrix is extended. - Two new parameters are introduced. - Current non-repeater flow has two sections, statistical and time domain. - This BIRD modifies both. - In 6.1 we also describe the repeater flow. - In 6.1, redriver and retimer flows were specified as identical up until the last step. - In this BIRD, redriver and retimer flow discussions are decoupled. - Redriver (statistical and time domain flows) - Retimer (statistical and time domain flows) - A paragraph describes the Impulse matrix (two new columns are added). - Definition of Init_Supports_Extended_Impulse_Matrix. - Info parameter. - Tells the EDA tool if the Model Supports the optional extended IR matrix. - From previous discussion, we require the model to handle the original non-extended IR matrix as well. - Definition of Impulse_Matrix_Is_Extended. - Input parameter. - The tool uses it to tell the model if the extended IR matrix is being provided. - Added new rows for these parameters to the three parameter tables. - Mike L./Radek: We could show the columns headings for those 3 tables to make it clear what the rows represent. - Walter: I responded privately to Fangyi regarding what goes into h1, h2, h3. - Also, when we originally wrote these flow descriptions the IBIS spec was still a plain text file. Now that IBIS is a .docx file that supports real graphics, etc., I would like to see the flows rewritten using the syntax that Todd Westerhoff developed for describing the nine combinations of Tx and Rx. - I think that would make it much clearer, and it would be much easier to see how this proposal affects those flows. - Arpad: In fact, perhaps we could even use flow chart type diagrams to depict the flows. The text descriptions in each step might be large though. - Walter: I think the equation syntax would be nice and easy to follow. - Fangyi: I characterized the four combination scenarios (GetWave_Exists True or False, and Init_Returns_Impulse True or False). - The nine cases Todd outlined before are folded into these four combinations. - Walter: I'm having real difficulty following the text to find out what the flows are. - I'll be interested to see what others think. I found it hard to follow the flow and make sure all the combinations work. - Bob R.: Other than the two new parameters there is no syntax change? - Fangyi: Yes. - Walter: What changes is the structure of the IR matrix data passed to the AMI_Init(). - Arpad: Everyone please review the BIRD draft once Fangyi sends it out. BIRD 147.5: - Bob Miller: [sharing BIRD 147.5 draft 3] - This draft has changes annotated relative to the last submission to the Open Forum. - I've made changes since the last meeting to some of the parameter discussions. - GetWave_Block_UI has been renamed BCI_Message_Interval_UI. - Modified the descriptions, Usage Rules and Other Notes. - Rewrote the example slightly. - Discussion: Ambrish asked if the text should include an example of the performance implications if the EDA tool did not choose its block size wisely with respect to the BCI_Message_Interval_UI. For example, if BCI_Message_Interval_UI were 2000, and the EDA tool used a block size of 1990, it would almost double the time between adaptations. Bob M. said he could add some text to that effect. Radek noted that we didn't actually state that the adaptation would take place after the processing of the block in which the BCI_Message_Interval_UI was exceeded. Bob M. said he thought that was addressed in the last paragraph of the BCI_Training_UI section. Based on some concerns Radek expressed about the relationship between BCI_Training_UI and BCI_Message_Interval_UI, Bob M. and Ambrish noted that BCI_Training_UI could be considered a time-out value. Bob M. added a sentence to clarify this. - Bob R.: Motion to have Bob M. submit this to the Open Forum with the agreed upon changes after Ambrish reviews the final text. - Bob M.: Second. - Arpad: Anyone opposed? [none] - Arpad: Thank you all for joining. AR: Fangyi to email his Flow BIRD draft to Mike LaBonte for posting. AR: Bob Miller to update BIRD 147.5 to draft 4 and submit it to the Open Forum. ------------- Next meeting: 24 January 2017 12:00pm PT ------------- IBIS Interconnect SPICE Wish List: 1) Simulator directives