Well reasoned.

Think of a three-dimensional curve of cost, uncertainty, and frequency to
measure gain on a range. Think of a second 3D curve involving modeling. My
guess is that below something like 20 MHz (use your own number) modeling is
to be preferred.


On a related topic: I saw with my own eyes NBS in Boulder (c. 1978)
using a different technique to measure gain. It was a near field scheme
where a probe was moved in front of the antenna while its vector voltage and
position was measured. (As I recall, a pair of lasers was used in the
measurement of the probe's x and y position.) The (vast number of)
measurements were then imported into a computer that computed the gain. As
we say: "you could do that!" I never thought to ask what the expected
uncertainties were expected to be.
73 Mac N8TT

--
J. Mc Laughlin; Michigan U.S.A.
Home:
"Richard Clark" wrote in message
...
On Wed, 27 Apr 2005 03:01:20 +0000 (UTC), "Reg Edwards"
wrote:

But if you, as an employee of a
reputable laboratory, were given the job of determining the forward
and reverse gains of fractal or other weird antennas, at 7 MHz and
144 MHz, what uncertainties would you state? I'd believe you.

Hi Reggie,

Measurement Mismatch Correction Error 0.04
Noise Power of Power Sensor 0.00
Zero error of Power Sensor 0.00
Power Meter Linearity 0.04
Space Loss Measurement Error 0.01
Multipath Curve Fitting Random Error 0.04
Proximity Effect Correction Error 0.05
The errors remain across all applications, only the assigned values
change. If I arbitrarily scaled all values by 25, few could challenge
the numbers.

At 7MHz we can all agree that the errors are going to be inversely
proportional to the astronomical cost to determine. No one is going
to perform it at HF when they can only afford 1/100th scale models
that offer the accuracies implied above. What would spending more
money buy them anyway?

73's
Richard Clark, KB7QHC