RE: IBIS Open Forum minutes of 12/5

Dear Greg Edlund,

> I agree that a single sub-committee on IBIS model quality would be in
> the best interest of the IBIS community.
>
> As I see it there are two main topics under model quality:
> 1. Syntactical correctness
> 2. Accurate prediction of electrical hardware behavior
>
> While topic 1 may cover a myriad of common modeling errors, it can be
> easily addressed: every IBIS model must pass the golden parser.

If it were only that easy. The Golden IBIS Parser checks the syntax of the
model, but not all of the semantics. Although the latest version of the
parser does do some testing of the data within the model, there are some
things that it simply cannot test for. For instance, the Golden IBIS Parser
has no way of knowing if all of the part's pins have been listed. So a model
could pass the parser and still not be usable with all simulators.

This is why it was suggested that the model quality sub-committee run a model
through several different simulators as well as through the Golden IBIS
Parser.

> Topic 2 is a bit more involved. The user community needs to see proof
> that a model developer measured the model against test hardware and the
> results were favorable. This leads to the questions, "What test
> hardware?" and "What is favorable?" This is where an amendment to the
> IBIS specification that covers model accuracy would lend great
> credibility to the IBIS model database. I think the sub-committee on
> model quality should work toward developing such an amendment, and I
> think the users and semiconductor vendors should work together toward a
> mutually agreeable solution.

Proof of a model's accuracy is an admirable goal. However, you yourself
pointed to some big questions: "What test hardware?" "What is favorable?".
Since we would need an answer for all types of parts and since there will be
many opinions concerning what is the best answer, I would prefer not to burden
the sub-committee with such a weighty problem just yet.

I propose that those questions be left to the full forum or be assigned to a
separate sub-committee. The model quality sub-committee (I don't believe it
has an official name yet) could be charged with "enforcing" the answers to
those questions in the future.

Don't misunderstand me. I am not saying that the answers to those questions
shouldn't be pursued, I just don't think it is appropriate at this time for
the small sub-committee to do so.

Regards,
Matthew Flora
Senior Engineer
HyperLynx
(425) 869-2320 PH
(425) 881-1008 FAX
mbflora@hyperlynx.com
 
Received on Thu Dec 11 14:30:55 1997