Re: non-monotonicity

There are 3 issues (I am aware of) on monotonicity:

1) The measuring technique used in making models can sometime lead to
non-monotonicity through round off error (though this is usually so far
from the operating region that you don't care much).

2) The non-monotonicity could be extremely small and caused by small
noise oscillations in a spice simulation in an area that was supposed to
be flat.

3) The device is really non-monotonic in the operating range. If this is
true, this implies the device is using some active feedback technique.
If THIS is true then there is some implied delay in the feedback loop.
This feedback delay is not modeled by IBIS. The IBIS committee has
requested IC vendors to come forward with SPICE circuit examples of
devices that exhibit this effect so that we can figure out a behavioral
mechanism to describe it but (at least to my current knowledge) no such
circuit has been delivered. So your choices are: Ignore the
non-monotonicity (that is, remove it from the IV curve) or simulate with
the non-montonicity but with no info about the delay. Depending on how
the circuit works, either of these could the best approach and I believe
different EDA vendors have made different decisions on what is most
appropriate in their simulators.

I think people who have problems with IBIS models showing this effect
should go ask their EDA vendors what is the appropriate solution for a
given model.

regards
jon
Received on Fri Aug 7 08:31:45 1998