Re: NEC-LIST: PCB transmission line modeling

Hi there,

I have been using our own MoM code for such kind of modelling. At that
time we had
no Greens function for layered media and modelling features via
filamentary current
(triangular base pointmatching) and triangular patches (Rao Wilton
Glisson).

Patches are typically more involved and less suited for greater
structures but may
be used to get insight into individual behaviours.
The modelling of the tranmission lines was done by matching radius of
the wires
and dielectric constant of the MoM model to the PUL parameters of the
transmission line.
The recipe is something like:

        - compute PUL parameters of TL (layered media) via 2D solver
          & convert to char. impedance Z0 and phase velocity v0
        - match v0 of MoM wire model to v0 of TL via (effective) dielectric
constant
        - use analytic formula for single wire pair char impedance to compute
radius
          of thin wire model
        - model entire closed(!) loop signal path and return with image theory
        - terminations are to be imaged as well
        - for computing radiation field in free space don't forget to reset
dielectric

I have done some comparisons with voltage at terminations obtained by a
time domain
transmission line solver and was quite pleased. Problems of the MoM
model are due
to discretization of geometry and localization of components. Depending
on the implementation
of the MoM code it is more or less sensitive to this subject and
requires some numerical
experiments (the tiny interconencts to close the loops are sometimes
very hairy).

Also, problems occur if the transmission line width is large (leads to
overlapping radius),
and to coupling computations if the loop area enclosed is different (TL
uses rectangular X-section, MoM circular), hence magnetic coupling is
different.

In order to discretise a ground plane one has to watch out for
artificial current loops
due to the discretization scheme.
Adjusting the width for a microstrip-like structure to the transmission
properties
in a similar manner and using 4 or 5 patches per width gave a good
microscopic view
on the current distribution for our purposes.

For satifactory results one has to use different models for different
purposes
(signal transfer, coupling, emission) and play with the parameters.

regards,

-- 
  Uwe
___________________________________________________________________________________
|                                             
|                                  |
| Uwe Keller				       | tel. : +49 5251-5402-120         |
| Projectgroup FhG/IZM - Paderborn / 	       | fax  : +49
5251-5402-105         |
|  		University - GH Paderborn      | email: keller_at_pb.izm.fhg.de      |
| Technologiepark 34, 33100 Paderborn, Germany |
http://www.pb.izm.fhg.de/~keller |
|=================================================================================|
-- 
The NEC-List mailing list <nec-list_at_gweep.ca>
http://www.gweep.ca/mailman/listinfo.cgi/nec-list