Re: Standard circuit, validation of EDA "IBIS-capability"

Jon, et al,

This is a good topic. Arpad's circuit is a suggested implementation, but as Jon
points out, is not the only one possible. If, as in Jon's case, an EDA vendor
feels they can achieve better results with a different circuit that employs the
same data, then the vendor is free to do so. Which brings up the question, how
does anyone know if the results are trustworthy?

As someone pointed out already, Intel has seen the results of a standard PCI
driver simulation into the three load topologies, as outlined in the Overview
Document. We have seen extraordinary agreement in all but one case. This work
was done by Don Telian's group last summer and others have joined the forum
since then. I personally cannot volunteer my group to be the validator of each
EDA tool, and I doubt if Don would like to volunteer his group to be independent
validator. On the other hand, the data we were using for the initial test is
still valid and a good sanity check. Maybe we can get some neutral third party
to act as quality checker(?).

Does anyone have any ideas on how we can approach this problem? The alternative
is that Intel has its list of vendors we know is "IBIS-capable" and TI, NSC,
IBM, DEC, etc., generate their own list independently. That seems wasteful.

Will

in response to:

Q1) Is there a "Standard Driver"?
A1) Yes! It was created by Arpad at Intel. And it addresses most of your
    concerns quite nicely. It has been brought up more than once that this
    spice model should be available as part of the IBIS specification.
    It sounds like a good BIRD. Are you or Arpad interested in putting up
    the BIRD? Arpad has done alot of work to create a great start on
-a universal
    simulation model. It is a continuously evolving tool and most of
-the simulation
    vendors have their own methods. We went through a comparision phase when
    IBIS was first started to insure all the vendors simulations compared
-accurately.
    Intel was unbiased the judge for the first validation suite. Note:
-The original
    version of the IBIS spice model created by Arpad did not completely
-address
    open-side issues.

JNP: I disagree, I recall no discussion about a "Standard Driver" nor
-public agreement to
abide by one. I have always considered comparison to physical measurement
-to be the only
appropriate guide.

jon powell
quad design
Received on Mon Feb 7 13:38:18 1994