Chuck wrote in
:

....
Just as an aside, Owen, have you
considered just a single pair of
non-resonant, opposed radials for all
bands? The ATU can just as easily "tune"
a single radial pair + vertical
element as it can the vertical element
alone.

Any additional benefit of separate
radial pairs for different bands may be
slight.

That would seem the case if you just regard the radials as providing a
counterpoise, the "other" connection that provides a return path for
current to the source. I have modelled this scenario where the source is
at the feedpoint (ie no transmission line) and the radials and radiator
are suspended above ground by a non-conducting structure, and you are
right that the radials need not be resonant, residual reactance being
dealt with by the auto-tuner at the feed point.

However, if you connect the radials to ground by some conductor (eg feed
line and / mast) that conductor is not part of the picture, and as I
modelled a conducting mast with a lossy ground connection, the big
picture is very different.

Here is a plot of modelled system losses with the configuration that I
described: http://www.vk1od.net/multibandunload...al/13mEV03.gif .
Not the large losses at 5MHz, this loss is mostly in the 20 ohms
equivalent earth resistance. The high ant+gnd loss at 1.8MHz can be
reduced to less than a dB with a pair of ~40m long radials (but tuner
losses increase to 3+dB).

So it seems that one could do as you suggest and effectively isolate the
radiator and radials from ground, or the radials need to be carefully
adjusted to minimise the mast / feedline current to ground, especially
where the feedpoint resistance is small wrt the equivalent mast to eart
resistance.

Owen