From: Fred Balistreri ([email protected])
Date: Wed Dec 08 1999 - 16:15:53 PST
Mike Mayer wrote:
> >>>>> "Bradley" == Bradley S Henson <[email protected]> writes:
> > Let me further propose we discuss the overall reliability of all
> > of our tools and techniques for gigabit digital.
> On my favorite topic, which is bad IBIS models, here is an (edited to
> protect the innocent) email I received from someone at a large
> corporation we buy complex VLSI chips from:
> Here's the [Part Number] model you requested. I had quite a few
> problems creating it since it was designed using a program other
> than SPICE. Here at [Large Corp], we do not have any software that
> allows us to simulate any IBIS models, so we cannot correlate with
> the original simulation models. The only thing we can do is make
> sure there are no typographical errors. This model was created
> using S2IBIS2 and an [Proprietary Language]. I pieced together from the
> original simulation model ([Proprietary Language]). let me know if you see
> any inconsistencies or if something fails to meet your
> expectations. I am really trying to push for some software that
> allows us to run IBIS and your input would be appreciated!
> Needless to say, I was shocked to find out that I am getting an IBIS
> model from someone who could not test it, even if they wanted to. Any
> vendors out there care to donate software?
> Mike Mayer Artesyn Communication Products, Inc
> Senior Hardware Design Engineer Madison, WI
> [email protected] http://www.artesyn.com/cp
> **** To unsubscribe from si-list: send e-mail to [email protected] In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP.
> si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list
There are tools avaliable that will verify the IBIS information
graphically. Some are free some are not. However verifing the IBIS
information is NOT the same as using the IBIS data, creating a model
and then doing simulation. And this is where people have IBIS wrong.
IBIS is NOT a model. It is a standard, which contains information that
allows EDA vendors to build an SI model. Most SI simulators will NOT
accept the IBIS data as a model in their simulator. There are two
EDA vendor that do this. However internally they parse the IBIS data
and use a topology, equations or model that fits their simulator.
All others or virtually all others have a translator of sort. This
can be as a standalone tool or incorporated in the SI software.
Keeping that in mind its simply not possible to expect the SAME answer
from different simulators. The best one could do is get the same answer
from different simulators for the SPECIFIED IBIS load under which the
data was tested. Under DC conditions all simulators that are respectable
and understand IBIS will get the same answer. However once transmission
lines and loads are added your at the mercy of the EDA SI software.
IBIS information that contains 4 vt curves has a much better chance.
We have been enhancing the IBIS spec to include more and more
information so that EDA companies can make a more consistant and
perhaps accurate model. Lots of Birds come from either the IC vendor
trying to incorporate some new technology or from the EDA vendors in
an effort to correlate better with known results from hardware or
SPICE. There is an IBIS validation group at work for this sort of
As for the semiconductor vendors I have no pity. If SI is REALLY
important to the systems guys then it should REALLY be important to
the semiconductor vendors. If tools are not avaliable for free (some
are) I think the semi guys can invest a bit to put together a
methodology to support the EIA standard.
-- Fred Balistreri [email protected]
**** To unsubscribe from si-list: send e-mail to [email protected] In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****
This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:39:05 PST