On Fri, 29 Dec 2006 13:37:43 -0500, chuck wrote:

Absolute accuracy is unimportant in this application.

Hi Chuck,

Quite so, hence the last digit is merely resolution, and the
resolution of 20 Ohms of loss against an Rr of 5 Ohms is easily
resolvable, and its reduction is easily seen. Further, automatically
assigning an error of 1 to the last digit presumes the underlying
measurement hasn't been rounded up from a smaller range of error.
Given the product is MFJ, it is hard to credit it with that kind of
craftsmanship. However, even with a 1 digit error (which has been
presumed to be quantization error), it will always be inherent in any
Relative reading. If 20 Ohms reading is actually 19 to 21 Ohms of
loss (irrespective of absolute accuracy), then driving out 1 Ohm of
loss will actually be 18 to 20 Ohms exhibited as 19 Ohms. The only
problem here is that you may have to drive out slightly more than 1
Ohm to kick the reading down to 19.

If you cannot drive out more than 1 Ohm of loss out of 20, then
there's no point in trying, is there?

Dennis,

If you've followed the discussion thus far, you can well expect that
the SWR will climb as you drive out loss. This means you need to
anticipate matching to something like a 5 Ohm load (in all likelihood,
as described here, it will be higher). Drive out the reactance and
use a 9:1 UnUn. Beware that such an UnUn will require 16 Ohm
transmission line (if it was all that easy, everyone would have
stopped posting similar problems like this long ago).

73's
Richard Clark, KB7QHC