Important IBIS question response

Arpad

We plan to use a method equivalent to the summing method (a) and we have
been assuming that the IBIS files are partitioned consistent with (a).

We even assumed that the Output types [Model]s would be partitioned into
the clamping and pullup/down portions. In general, I accept your argument
last Friday that (within the framework of the IBIS model), such a partitioning
is not necessary for the Output Model_type. I would still prefer to see the
data partitioned (even if there are some reasonable, arbitrary assumptions
about how the data is partitioned) because there may be cases where this may
really be necessary. Partitioning should be done in a consistent manner
throughout in IBIS models. Then simulator specific issues can be dealt
with from a common reference.

One case where partitioning assumptions have to be carefully understood
is the Open_Drain configuration with an internal pullup resistor (transistor).
My approach would be to model the resistor using the POWER_clamp table since
I would assume the pullup table does not exist. (Usually these currents
are small enough to be ignored, so this is not a big problem in practice.)

Another hypothetical case is ECL_Output with a clamp to VEE. Since both
the pullup and pulldown tables are connected to VCC, the current path
routing differs based on the partitioning assumption with the ECL device in
the low state. In the (unlikely) event of a large, negative undershoot, the
partitioned model would pull current from VCC and VEE whereas the non-
partitioned model would pull all the current from VCC.

Thank you for raising a fundamental consideration.

Bob Ross,
Interconnectix, Inc.
Received on Tue Apr 5 17:34:05 1994