====================================================================== IBIS FUTURES/COOKBOOK SUBCOMMITTEE MEETING Date: March 25, 2004 Attendees: Cadence Design Systems - Lance Wang Cisco Systems - Syed Huq Green Streak Programs - Lynne Green Intel Corp. - Michael Mirmak Mentor Graphics - John Angulo, Ian Dodd Teraspeed Consulting - Bob Ross ====================================================================== Next Meeting: Thursday, April 8 1 PM - 3 PM US Pacific Time Telephone Bridge Passcode 916-356-4261 3 673-8669 Agenda: Cookbook Discussion - Proper methods of data Arpad Muranyi/Lance Wang extraction Futures Discussion - Reported ICM issue and Michael Mirmak IIRD proposal - Circuit Call References Ian Dodd COOKBOOK MINUTES The meeting opened with a brief discussion of DesignCon East issues. Lance Wang stated that his and Arpad's update on data extraction methods would be provided next week. Michael reviewed a detailed list of keywords added since the 2.1 1997 Cookbook from Stephen Peters was created. While most [Model] keywords and subparameters are described in some detail, most of the non-header keywords in IBIS 3.0 and follwing are not mentioned. Bob Ross suggested limiting the scope of the document update. No keywords past IBIS 3.2 should be considered. Further, a keyword-by- keyword description may only end up repeating material in the specification text. John Angulo inquired whether EBD descriptions should be included, as well as package modeling data. Lynne Green concurred, saying that an entire contextual discussion of package modeling options should be included. This should involve describing the types of package models available, TDR experiments to obtain data and how the data is entered in the model. Lists of lab references should be provided. Overall, the new document should be organized by topic (four), not by keyword. The document will assume that users will rely on the golden parser for specific syntax questions. At a minimum, the document should contain a chapter on data extraction, a single-ended and a differential example from start to finish and a checklist and summary of model creation flow. ARs: Bob Ross - provide outline of updated document (DONE) John Angulo - provide overview of differential model extraction details ====================================================================== FUTURES MINUTES Ian Dodd reviewed a list of concerns regarding data types under IBIS 4.1 AMS-based models: - allowed digital port data types (example, '1', X, '0') - defined port types or natures (analog and digital) For example, VHDL-AMS defines "natures" (electrical, mechanical, etc.) in the style of data types in other languages. EDA tools will need to know the data type to expect being passed into and out of the tool interface. Similarly, digital port states being provided to the tool, if in-line with the IEEE requirements, should be limited to '0' and '1' for inputs and outputs, with X additionally being preferred for outputs. The use of X assists in delay calculations. Michael summarized his findings on ICM integration with IBIS. If IBIS were to add ICM support under the current package model format, several issues are raised: - die pads are assumed in IBIS but must be explictly named in ICM - a name-matching requirement would have to be enforced between the ICM and IBIS names - the treatment of [Pin Mapping] is highly ambiguous for packages - should ICM be used as an [External Circuit] data source? The treatment of missing or mismatched names under the current IBIS parser is not clear, as is the treatment of package models for pins under [Pin Mapping]. Finally, splits between single pins and multiple pads would require explicit naming of die pads in IBIS. Bob Ross commented that the intent was one-to-one mapping from die pad to named pin. Michael will check the current parser's behavior for package models with unmatched pin names and quantity of pins. For reference, the seven original BIRD proposals are listed below: 1) Define accepted digital states for input and output ports on IBIS 4.1 AMS models 2) Provide explicit links between ICM and IBIS 3) Expand [Circuit Call] to accept native IBIS [Model] and [External Model] references 4) Define series links between pins, pads and/or nodes under the multi-lingual extensions 5) Create user-definable buffer delay and input threshold measurements 6) Permit re-use of port names under [External Model] 7) Permit native IBIS data to be linked through external file names, similar to the multi-lingual extensions Attendees were assigned to complete formal proposals or definitions for the above, as follows: 1) Ian Dodd 2) Michael Mirmak 3) John Angulo 4) Ian Dodd & Bob Ross 5) Lance Wang, Ian Dodd and Arpad Muranyi 6) Arpad Muranyi 7) Lynne Green The formal statements should include a problem statement and a general outline of the solution. I-t and core currents will be discussed by Syed at a future meeting. ARs: Lance Wang - check with Cadence experts on Verilog-AMS type definitions and restrictions Michael Mirmak - check the golden IBIS parser's treatment of package models with pin naming and number variations (partially DONE, [Pin Mapping] underway)