V/I table sign validation (from Kellee Crisafulli)

To: Jon Powel, Arpad, Don T.,Bob Ward, IBIS commitee
From: Kellee Crisafulli
Re: method to determine if V/I table is valid

Don:
This should not require a new bird because we are not changing
the IBIS specification. This also has little to do with monoticity.
The problem as I described earlier is simply validating that the V/I
tables have the correct sign. For example the source generator
should source current, and the sink generator should sink current.
The reason for the rush is that I have personally seen 4 separate cases
of incorrect IBIS models being created in the last two weeks. We now
have numerous companies including consultants creating IBIS models.

I believe it is very important to make sure models are correct before
being released to the public. If we cannot agree on a simple method
of validating the V/I table sign than it will require a full fledged
BIRD to achieve this and a much longer time table. My first impression
was that this would be a simple matter. I still beleive there is a
simple method of validating the sign of these tables:

I will repeat my proposal:

For each V/I table:
  1) Find the minimum and maximum voltage points (Vmin, Vmax) in the table.

  2) IF:The TYPICAL current corresponding to Vmax is less than the TYPICAL
        current corresponding to Vmin than the table is assumed to have
        decreasing current.
     ELSE: The table is assumed to have increasing current.

  3) Verify that:
        - Pullup V/I table has decreasing current
        - POWER_clamp V/I table has decreasing current
        - Pulldown V/I table has increasing current
        - GND_clamp V/I table has increasing current

  4) If any table moves in the wrong direction report that
     the table has an ERROR, (wrong sign....).

Note: I am proposing an ERROR, not a warning!

  5) Another test we should consider: If the currents at Vmin and
     Vmax are equal then consider that the data may be a single
     point. The IBIS specification requires at least two data points.
     I propose that we flag this with a WARNING message.

Bob Ward, I have revised number 5 to better address your
inputs.

I would still like to get agreement on a method of validating
the sign of the V/I tables as soon as possible.

We had agreement at the last meeting that Will Hobbs would determine
if we had suffient agreement to continue with a modification to the
IBIS_CHK program.
Received on Wed Mar 16 21:34:28 1994