Re: Re[2]: connector models

Hi Gus,IBIS

At 06:44 AM 12/15/97 -0330, apanella@molex.com wrote:
> The connector companies main issue with IBIS is the lack of coupled
> line modeling.
> The connector companies suggested the use of a Maxwell matrix for the
> connector models a couple of years ago. I think a partial option has
> been installed in the V.3.0 IBIS specification.
Close but it is not used and it isn't set up for connectors.

> As far as "first pass".... I beleive that IBIS in V.2.0 can support
> series L, parallel C, and series R. This is the simplest solution
> with the most probablility of inaccurate results.
It doesn't work for connectors. Connectors need an input and an output
defined.
The pin definitions in IBIS 2.0 don't provide that. The Version 3.0 sytax is
much closer with the additions that Steven Peters made.

> Now I think the other significant problem is going to become
> apparent... The fact the IBIS simulators have different methods of
> "circuit solving". When a connector model is provided in matrix
> format, the results will depend on the simulators solution algorithms
> (not an issue with Berkeley SPICE).
Not true.
It is the same issue as for Spice, the problem is a standard. Just
a coupling matrix is not a connector model. It could be; but just as it could
be in Spice the format must be agree upon and then the simulators all work.

> I am concerned that connector models will perform differently on
> different simulators. Is this a valid concern??
Yes, but it should not be a problem if the format is well defined.

> A part of the problem might be... (besides the above concern)
> * Semiconductor manufactures need a way to provide models that do not
> provide insighth to proprietary fabrication methods. Connector
> companies already have this... SPICE.
Perhaps but it is not in any sort of standard format. Each connector company
uses different syntax. Also not all Spice packages support all connector
formats. Also most connector companies do not provide their models for open
download over the web. It requires an NDA or even extra fees.

> Some SPICE simulators can accept some versions of IBIS models some
> can not support IBIS at all. Are there any IBIS simulators that
> support SPICE? In this case, it all depends on were the "energy" is
> assigned.
IBIS is not a simulator. IBIS is a description format, as has been
pointed out by several people on this forum.
Many Spice packages support IBIS. All SI simulators I know
of support IBIS. Spice syntax has evolved in 30 separate directions over
the years. It is now to the point that it is very difficult to exchange Spice
files between two different types of Spice simulators and have it work.
With IBIS we are attempting to maintain a standard syntax that all simulators
can use including Spice.

> Most connector companies do not sell simulation software.
> As such connector companies might not be the best source for driving
> this issue.
This I believe is the real issue, I feel the connector company's
participation is critical. As I work at a simulator company I
would very much like to see all the connector companies participate
in developing a single standard that all companies could use. It
is certainly not required that the connector company have a simulator.
In fact I think it is a big disadvantage because then they become
biased toward their simulator.

Perhaps we need a multi-day focused session on connector modeling to
hammer out a syntax. I would very much like to see all the connector
company's represented at the Jan. Design Con IBIS meeting and
start to hammer out a solution. Perhaps we do a Phase 1, Phase 2, Phase 3.

 
Received on Mon Dec 15 09:48:51 1997