Re: Re[6]: Rising Waveform Loading Effects: smart simulator

Hello Arpad, Dileep, others:

     I think Dileep's basic point has been missed, and it really needs
to be responed to.

     Yes, as in Dileeps example of a V/T curve into a
complex reactive load, it *is* possible to create a perfectly
legal yet totally useless (to him, anyway) IBIS model. It
is also possible to make a model that produces different
results on different simulators. IBIS models are, to some
extent, simulator dependent. However, I'd like to point out
that in both these cases the model would also be useless to the
model provider (i.e. silicon vendor). The whole point of this exersise
it to produce simulation models others can use. It is incumbent
upon the model provider to both produce a model that works with
industy simultors and to validate/spec whatever model they provide.
(That, by the way, goes for both transistor level as well as IBIS model).
It is also incumbent upon the simulator vendors to stay current with
the spec and the technology.

     I might also note that the IBIS committee has always relied on the simulator
manufactures to tell us what kind of data is required for acurate
behavioral simulation. If a simulator vendor belives that the spec
is inadiqute(sp), or doesn't provide the right kind of data, they should
speak up. By the same token, the IBIS committee should be listening....

                    Regards,
                    Stephen Peters
                    Intel Corp.

Arpad,

>IBIS allows you to define up to a maximum of 100 waveform tables (V-t curves) in
>each buffer model. Each of these V-t tables have their own subparameters, such
>as Rfixture, so you can make a model which provides a series of V-t curves for
>the various loading conditions. Is that not enough for what you are trying to
>achieve?

At the risk of repeating myself, the point I was trying to make was IBIS does
ALLOW the model developer to specify 100 waveforms, but does not REQUIRE the
waveform tables to follow certain restrictions or conventions.
Therefore, is it then acceptable for each simulator to enforce its own
REQUIREMENTS making IBIS models simulator dependent?
Earlier, it was pointed out that if ONE waveform into ONE RESISTIVE load is
specified, it is possible to predict the waveforms into any load. Then that
particular simulator HAS to make sure that the waveform is specified
into a RESISTIVE
load, whereas, the IBIS specification ALLOWS the waveform specification
into ANY load. What is the simulator expected to do if the model developer
specifies the waveform into ANY load since that is ALLOWED by IBIS and hence
it is a legal IBIS model?
------------------------------------------------------------
Dileep Divekar
Applied Simulation Technology, Inc.
2188 Bering Drive
San Jose, CA 95131

Phone - (408)-434-0967 x 100
Fax - (408)-434-1003
Email - dileep@apsimtech.com
------------------------------------------------------------
Received on Fri Nov 1 18:51:55 1996