Meeting date: 18 apr 2006
Members (asterisk for those attending):
*Arpad Muranyi,
*Bob Ross,
*Todd Westerhoff,
*Mike LaBonte,
Paul Fernando,
Barry Katz,
*Walter Katz,
Ken Willis,
*Ian Dodd
*Lance Wang
*Richard Ward
Doug White
*Sanjeev Gupta

-------------
Review of ARs:

none

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

Welcome to Sanjeev Gupta from Agilent:
- ADS has a good system for modeling SerDes channels.
- Todd would like to have a simulator-independent version of it.
- Agilent trying to move away from encryption.

To be or not to be - continued:
- Example of a difficult problem: Self-calibrating buffers
  - Trying to model the maximum calibration error.
  - This can't be done in IBIS at the moment, but it can with Matlab/AMS.
- Richard not sure if macro-model performance will be good enough.
  - Difficult to put non-linearities like compression into AMS.
  - Simulation will be faster if linear.
  - It is important to correlate with silicon.
- End user requirements:
  - Mix and match silicon and behavioral models
  - A non-proprietary solution is needed, not single-vendor.
- Building blocks that relate to silicon structures
- Original Telian-based macro-model vision was for buffers
  - Now we are looking at system behavior
  - Now we want to do design stage what-ifs
- How to extract data from silicon for more complex model structures?
  - no cookbook
- Current AMS function library is immature
  - Example: easy to do an FIR filter in Matlab
  - More difficult to do in AMS
    - Richard: verilog AMS has one
- Again, do we need SPICE compatibility?
- Gary Pratt has 2 people coding AMS models
  - PLL filters tend to have to be recoded moving from one vendor to another
- Standards like 802.3AP: can we get groups to produce reference models in AMS?
  - Like having the industry create macros
  - Better off having building blocks, even if they have to be customized
    for each use.
    - May be able to parameterize them most of the time
- We don't have to discard our work so far
  - We really need to keep looking at simulator compatibility issues

Will IBIS 4.1 offer encryption?
- Proprietary solutions have to be encrypted for each tool
- IBIS committee could get in the encryption business
- Could use something like PGP
- Difficult to keep the private key private among multiple vendors
- This is an important issue, although we would prefer unencrypted models
- Encryption should be restricted to the parts that need it
- We could define IBIS BEGIN/END encryption keywords, but not the algorithms.

Further discussion of future direction in the final minutes of the call:
- We should restate our goals and prioritize in the next teleconference.
- We need to get simulation companies to weigh in here.
- We are going from phase 1 to phase 2.
- Some vendors may be left behind.

-------------
Next meeting: Tuesday 25 apr 2006 12:00pm PT

