"Owen Duffy" wrote in message
...

I have been playing with some NEC-2 models of a multiband vertical with
radials.

The vertical is an unloaded vertical of 13m height, and it is mounted on
a 6m high grounded metal mast, and an ATU installed at the feedpoint
(base of the radiator). I have fitted a pair of opposed nominal quarter
wave radials for each of the 80, 40, 30 and 20m bands.

To simulate ground loss, I have modelled a 20 ohm resistance in the
bottom of the mast, and used a MININEC ground.

An interesting observation is the sensitivity of this model to length of
the radials. Properly adjusted, each pair of opposed radials near
eliminate current on the mast (more than 20dB below the current into the
radiator). The exception to this is the 30m radials which seem to suffer
some interaction with the 80m radials (near third harmonic). Without the
appropriate radials, current in the mast to ground is large, and losses
can be 10dB or more.

The ideas I take away from the modelling excercise is that:
- inadequate decoupling exacerbates ground loss;
- decoupling is very dependent on the length of the radials;
- one pair of opposed radials is enough for a narrow band;
- the radials for different bands have some interaction; and
- the optimum length may be quite a deal longer than the expected length
of legs of a half wave dipole in the same place.

I am grappling with some other way to optimise such a system, other than
measuring the mast current (which often isn't easy).

These effects probably also apply to a trapped vertical with similar
counterpoise, and the traditional wisdom of tuning either the length of
the vertical or radial length to achieve low VSWR is probably less than
optimal, there is an optimal length for each of them.

The traditional wisdom that elevated radials are generally significantly
lower loss than buried radials probably depends on careful "tuning" or
isolation of feed point ground to minimise current flowing to the real
ground.

Comments, thoughts?

Owen

Owen, It may not be too critical, but would not the Sommerfeld/Norton
method improve accuracy?

Frank