From: LaFlamme, Peter (Peter.LaFlamme@compaq.com)
Date: Mon Mar 20 2000 - 07:37:54 PST
DC has some great points about screening the IBIS models. The only true
verification of the accuracy of an IBIS model from an IC vendor standpoint
is to ensure that the IV and package data are accurate and the syntax of the
model is consistent with the IBIS rev. In my past experience with modeling
IC's it is very difficult to tell customers that you have an accurate model
when thier system simulations do not exactly match thier lab data taken.
There are many variables to the accuracy of a simulation using an IBIS model
(SPICE also), which have been covered previously and I will not get into
here (beat the dead horse). I think that screening the data in the model and
syntax is a great way to ensure that the model creator has done due
Compaq Computer Corporation
200 Forest Ave (MRO1-1/P5)
Marlboro, MA 01752
(508) 467-4541 Phone
(508) 467-4493 FAX
From: D. C. Sessions [mailto:firstname.lastname@example.org]
Sent: Monday, March 20, 2000 10:15 AM
Subject: Re: [SI-LIST] : Bad IBIS models!
Kim Helliwell wrote:
> Dan (and the list)
> I guess I'm reluctant to do this, as I'm not sure what purpose would
> be served. Some of the vendors are actually aware they have problems
> and are, to their credit, trying to fix them. My comments were aimed
> at vendors who are (or appear to be) clueless about how these models
> are actually used. But I don't think it would do to embarrass them
> publicly. At least not yet!
> In a sense, I don't know that I've found *ANY* vendor who supplies IBIS
> that are consistently good across the board. Vendors are not monolithic;
> they have different product groups, and some groups do a better job than
> others. If my comments can heighten the awareness of these vendors or
> product groups, I will feel it was worth posting my diatribe. And I've
> already had some indications that vendors themselves welcome this sort
> of criticism as it helps them know how to do a better job.
> So the best thing to do about it is to complain to the vendor whenever you
> get a model that is not up to par. Let them know in specific detail what
> is wrong, why it's wrong, and how it should be fix (if you know). Some
> will push back, but the better ones will thank you I believe.
Kim, please consider telling them that incoming QA will screen against their
IBIS models, and make sure that they know what that means WRT ultranarrow
VI curves and so forth. If they try to get cute by making anything-goes
models, tell them you can't qualify them because the part won't work in
Until IBIS data is taken seriously (ie it affects the bottom line) we
have to accept that it's part of the advertising budget and has the
same degree of technical quality.
-- D. C. Sessions email@example.com
This archive was generated by hypermail 2b29 : Thu Apr 20 2000 - 11:35:46 PDT