NEC-LIST: Cylinder connected to Ground by a Wire

This is a restatement and update of my problem. Sorry for the long lines.

The problem is to calculate radiation at low frequency from a 1-m diam. by 1-m long
cylinder connected with a 0.01-m radius, 1-m long axial wire to a ground plane. The range of
frequencies is from 10 to 100 MHz, i.e., with the 1-m dimension corresponding
 to .03 to 0.3 wavelengths and the wire radius .0003 to .003 wavelengths. The prominent
 response feature expected is LC resonance, where C is the capacitance of the cylinder
 to the ground plane and L is the inductance of the wire. For a wire with radius .01 m
this should occur around 30 MHz (i.e., wire length of ~0.1 wavelengths).

The cylinder is constructed of a vertical strip of patches, and quadrilateral tops and bottoms,
rotated to fill a quandrant and then reflected to complete the cylinder. A square patch
is overlaid on the bottom center to provide attachment for the wire. The calculation is
performed with the 1-m wire separated into between 1 and 11 segments with voltage
sources in each segment adding up to 1 V for the full wire. At the lowest freuencies
(wire length ~.03 wavelength) the current should be determined by C alone, but the
 calculations show a strong decrease in wire current with increasing number of segments.
The position of the resonance peak also increases in frequency with increasing numbers
of segments. Roughly, the response appears correct only if the wire segment lengths are
 at least .03 wavelengths. The NEC warning comes at .001 wavelengths. In the range
 between these values the calculated response appears to be drastically wrong, e.g., low
frequency wire current too small by almost an order of magnitude.

The NEC output indicates that the wire is correctly attached. The square patch in the middle
 is divided into four sub patches, as expected. The anomalous resonant behavior also occurs
 with a voltage source in only one of the wire segments, or with the voltage source removed
and a vertically polarized incident wave. Disconnecting the wire by shortening it to 0.95 m
produces a low-frequency response and resonance peak slightly beyond those observed
in progressing from 2 to 11 segments. The results act as if increasing the number of segments
 is equivalent to gradually disconnecting the wire from the cylinder.

At first I had a similar problem with a 1-m cubic box instead of the cylinder, but it was
overcome by dividing the bottom surface into smaller patches. That trick, including
segmenting the bottom of the cylinder into anuli at 4 radii, did not work with the cylinder.
The results appear to depend only on the number of wire segments, not the size of surface patches.

Perhaps the answer is to restrict wire segment lengths to at least .03 wavelengths for some
problems, but which ones? Any insight into why this happens or how to circumvent it would
be appreciated. Of course, if I'm doing something stupid, appropriate instruction would be invaluable.
I'll gladly furnish a copy of the .nec file.

-- 
The NEC-List mailing list
NEC-List_at_robomod.net
http://www.robomod.net/mailman/listinfo/nec-list