Think for a moment about a wire carrying RF current. RF current in any
good conductor stays very close to the surface, so it's valid to imagine
it flowing right on the surface. On a wire, the current spreads
uniformly around the wire (unless it's very close to some other
conductor), and each little part radiates. But at any distant point, the
fields from the currents at various places around the wire (at a single
location along the wire) are virtually the same, so it acts like a
single current filament flowing on an infinitesimally thin wire. This
is, in fact, how NEC and similar programs model conductor currents.

The same even spreading happens as the wire gets fatter and fatter, but
only up to a point. The model of a single current filament begins
breaking down when the fields from different places around the wire are
noticeably different at a distant point. This happens when the wire
diameter becomes an appreciable fraction of a wavelength. Other things
happen, then, too -- circumferential currents -- ones flowing around the
wire -- develop, resulting in (or being caused by, depending on your
point of view) non-uniform current distribution around the wire. And the
wire itself affects the field. That is, the current on the side away
from a distant point can't directly radiate to the distant point because
the wire is in the way. Because the various current contributions around
the wire won't all add together at a distant point any longer, the
pattern changes.

I can't give any more specific answer to the question, really, than that
the pattern will become more complex. In the case of the example I gave
earlier with the cylindrical "car", if you raise the frequency, you'll
reach a point where these effects happen. One result will be that the
horizontal pattern will no longer be omnidirectional, but develop lobes.
The height of the cylinder or car might affect the way current is
distributed around it -- I haven't thought about it enough to hazard a
guess.

The good news is that today's modeling programs do a good job of showing
these effects. The general technique is to represent flat surfaces such
as a car top or side as a wire grid, to stay within the program's
requirements that wire diameter must be no more than a very small
fraction of a wavelength. As long as the holes in the grid are kept to a
tenth of a wavelength or less, results are quite good.

Roy Lewallen, W7EL

Bill Turner wrote:
Roy, your analogy of the car body as a tin can really got me to
thinking.

With the whip mounted dead center on the top of the car, I can see how
the roof acts like a ground plane (a very short one) but I'm puzzled
about the radiation from the lower part of the car body. If one
visualizes RF flowing through the sides, hood and trunk of the car, the
currents will all be in phase with each other (roughly, of course) but
the currents are displaced in space by several feet.

How does this affect the net radiation from the car body as a whole? Is
there some addition or subtraction due to having the same current, same
phase but at a different location in space, and arranged in a more or
less 360 degree pattern?

An interesting thought.

73, Bill W6WRT