Re: [SI-LIST] : W-element Simulation

Dmitri Kuznetsov ([email protected])
Mon, 29 Sep 1997 11:22:54 -0700

Doug,

I had an error like this reported by one of the W-element users once
before. That time it was caused by a wrong off-diagonal value in
conductance matrix G.

The Maxwellian admittance matrices (such C, Go and Gd matrices in
W-element's RLGC file) have to be diagonally dominant, because their
diagonal elements are the sum of all the physical self/mutial
admittances in each row. In Maxwellian C matrix, an off-diagonal
element being greater in absolute value than the corresponding diagonal
element, will create propagation modes with negative delays. In Go or
Gd matrix, it will cause the energy being generated rather than lost in
the dielectric.

I added a check for diagonal dominance into the W element, so the next
release of Hspice will detect these problems automatically.

Please send me the simplest netlist that creates the error, I will try
to figure out what the problem is.

Regards,
-- Dmitri

=======================================================
Dmitri Kuznetsov, Ph.D.
The W-element Engineer

Avant! Corporation e-mail: [email protected]
46871 Bayside Pkwy. Tel: (510)413-8631
Fremont, CA 94538 Fax: (510)413-7706
=======================================================

Doug Day wrote:
>
> I am currently trying to simulate a package with a W-element model.
> I have borken the model into 4 sections for convenience (it is a 4
> row BGA) This yields very small delays for each segment (4 - 6 ps).
> When I simulate I get the following error on my console:
>
> hspice ttlio3.sp > ! t
> sqrt: DOMAIN error
> sqrt: DOMAIN error
> sqrt: DOMAIN error
> time: command terminated abnormally.
>
> THe number of times "sqrt: DOMAIN error" is repeated seems to have a direct
> relationship to the data in my .rlc file. I can change some numbers and get it
> to occur more times but can not get it go away.
>
> I believe that it is either related the short delay I am using and possibly
> a time step error but I am not sure.
>
> Has anyone else seen errors similar to this?
>
> Thanks
>
> **************************************************
> Douglas Day
> Applications Engineering Manager
> High Speed Interface
> Communications Products Division
> VLSI Technology Inc.
> TEL: 408-434-7660
> FAX: 408-434-7866
> *************************************************