Meeting date: 14 mar 2006
Members (asterisk for those attending):
Arpad Murayni,
*Bob Ross,
*Todd Westerhoff,
*Mike LaBonte,
Paul Fernando,
Barry Katz,
*Walter Katz,
Ken Willis,
Ian Dodd
Lance Wang

-------------
Review of ARs:
(we did not review ARs, but here is the previous review)

AR: Mike make the website more outsider-friendly
- TBD

AR: Todd contact Cadence about a new representative
- no report

AR: Mike finish documentation examples.
- in progress

AR: Todd contact Ian
- no report

AR: Arpad make buffers scalable

-------------

Second Review of IBIS-Macro Review.doc from Walter Katz
- We can measure an I/V curve for every tap on a driver, but
  they really all differ only in scale factor.
  - So one buffer instance with variable scaling will suffice
    to model all taps.
  - In Intel's AMS model an event-driven PWL moves from one tap
    to another by changing scale factor for one I/V curve.
  - It actually has 2 I/V curves, one to power, one to ground.
- Ibis-macro has been putting logic around B elements
  - Maybe we need macros from which a B element can be built.
  - For drivers with more than 2 taps we at least need V/T curves
    for all possible transitions.
- The Intel approach is different from the 2-buffer current summation
  Cadence approach, a.k.a "driver-schedule style".
  - Arpad found that the driver-schedule approach is not accurate enough.

Macro-model designs beyond pre-emphasis driver:
  - Constant impedance buffer (self-calibrating).
  - BIRD95.6 driver
    - need event-triggered PWL current source
  - Single buffer that automatically implements [Model Selector]
    - Not very compelling, just a convenience
- We could implement the B element as a template
  - More flexible than "black box" for advanced buffer designs
  - Would need to implement an I/V controller macro

Compelling arguments for ibis-macro:
- Can do pre-emphasis, BIRD95, etc.
- Easily translatable to every tool

Is IBIS truly portable from simulator to simulator?
- Simulators implement k(t) switching function differently.
- This is not visible at signal pins, only at power pins.
- A BIRD95 macro-model buffer would help us test this.
- We need data for a real buffer, including composite current.

AR: Walter contact Gary Pratt about model data

-------------
Next meeting: Tuesday 21 mar 2006 12:00pm PT