RE: [SI-LIST] : W-Elements

Haller, Robert ([email protected])
Thu, 5 Aug 1999 00:10:45 +0800

Richard,
I haven't looked at roughness (and we presently don't model it), but I have
done detailed correlation between simulation and measurement (your right it
would make a good paper). I designed some test boards, took data, then
cross-sectioned the boards. I used SIMPEST (DEC's Field Solver) to create an
RLGC matrix and also a Pi model with an RC ladder (for skin effect). I
compared the simple Pi with RC ladder model, to Hspice W-lines (98.4) ,
HSPICE U-lines (ugh) and Ideal T-lines. The W-line (for my case) was
surprisingly accurate and ran very fast. The RC ladders correlation was good
but exhibited a small discrepancy at the very top edge of the rising
waveform (dielectric losses ?) and was sensitive to hspice 'method'.
U-line's correlation was poor. T-lines were surprisingly good at predicting
delay, but as expected don't demonstrate the 'rolling over' of the waveform.
I ran HSPICE using these different models varying risetime, timestep,
method, and segment lengths. I was surprised to find significant difference
in output shape depending on method, timestep and the other variables.

I concluded that W-lines were fast and accurate utilizing HSPICE 98.4 with
my edge rates, lay-up and conductor cross-section. Use a timestep of ~ .1
times your risetime for best performance and accurate results. I recommend
W-line users designing high-speed systems perform bench correlation
especially when edge rate is critical and pay particularly close attention
to measurement techniques.

regards,
bob

Robert J. Haller
Compaq Computer Corporation
AlphaServer Product Development
Phone: (978) 493-4112
Fax: (978) 493-0941
[email protected]

-----Original Message-----
From: Mellitz, Richard [mailto:[email protected]]
Sent: Wednesday, August 04, 1999 9:49 AM
To: '[email protected]'
Subject: RE: [SI-LIST] : Proposal: Rs correlation/collaboration for
W-Elem ents

Very good point about roughness. I wonder if some these EM modeling tools
could incorporate this?

Richard Mellitz,
Intel

**** 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 ****