NEC-LIST: How much should the size of patch be?

I cannot say anything more specific about the required size of patches
or wire mesh segments, but the comment about NEC surface patches
together with Tod Hubing's Canonical problem of a thin plate with
attached wires raises a red caution flag:

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
! Surface patches can only be used to model closed, perfectly !
! conducting surfaces. !
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

That is, the patches must form a closed surface so that only their
outer surfaces (normals outward) can be seen from the region of
interest. A closed surface formed by patches on the top, bottom and
edges of the plate would not be likely to work in this case. The
single-point integration used for the current over each patch would be
very inaccurate until the patch size became much less than the
distance between the top and bottom surfaces, and that would require a
hugh number of patches.

Looking at the Magnetic Field Integral Equation, given in the NEC-2
and 4 manuals, it can be seen that for patches on a single flat
surface the integral over current vanishes due to the vector cross
product. What is left looks like physical optics, J=n X H. This may
give reasonable results for a case such as a monopole on a flat plate,
but it is not an integral equation solution. It would certainly not
be valid when currents on both sides of the plate were significant.
On their back sides the surface patches have the boundary condition
that tangential magnetic field is equal to zero, while they carry an
electric current. That is a very strange, non-physical situation.

I have not tried modeling Tod Hubing's plate problem, and the
description in the ACES Journal seems to be missing a dimension to
locate the wire with respect to the plate. I would try a rectangular
mesh plate, possibly refining the mesh in the region under the wire.
It might be necessary to make the mesh under the wire smaller than the
height of the wire, but larger mesh, based on wavelength, should be OK
away from the wire. Tod gets surprisingly good agreement with
measurements with the simple wire plate model using squashed wire
loops to model the thickness of the plate. Since the plate thickness
is only 1.e-3 wavelengths at 100 MHz, it seems like a single mesh
surface should be sufficient. The current on the mesh would then have
to be the sum of the currents that would flow on the top and bottom
surfaces of the plate, but that should be OK for both impedance and
radiation patterns if the solution was accurate. However, there is no
guarantee that that would work as well as Tod's loop model. Guess
that is where the art comes in in the modeling business.

Jerry Burke
LLNL
Received on Fri Jul 10 1998 - 09:37:39 EDT