Jim Lux wrote:

I doubt that any antenna an amateur is likely to build has some physics
not adequately modeled by the math in NEC. Exactly what might be these
discrepancies in equations, formulae and assumptions? Considering that
NEC has had decades of validation against actual measurements and a lot
of really, really smart people looking at how it works, I'd be kind of
surprised.

However, I can also easily believe that an amateur (or professional)
could build an antenna that has measured performance different than
expected from their NEC model of that antenna. The differences would
lie, most likely, in these areas:
1) Inaccuracies in the model itself. Things like earth properties (NEC
assumes uniform dielectric, it isn't) are an important source of error
for antennas close to the ground. Most amateur models do not include a
very good model of the surroundings (supports, trees, feedlines, etc.)

2) Inaccuracies in the measurements or not measuring the right things. A
good example is using NEC to get feedpoint characteristics, then
measuring at the rig, and not properly accounting for the transmission
line, particularly if the feedpoint Z is reactive.

In the professional antenna world, if someone models an antenna, then
builds it and tests it on the range, and the measurements differ from
what the model predicted, the usual assumption is that what was built
differed from what was modeled, or the measurements were off.

A lot of people are aware that some knowledge and skill is necessary in
order to construct a good model which will give accurate results. What's
not so widely appreciated is how difficult it is to make decent
measurements of even such seemingly simple things as impedance, let
alone gain and pattern.

Some years ago I was at what was then NOSC (Naval Ocean Systems Center)
in San Diego, where they had the very best equipment to do full 3D
pattern measurements of antennas mounted on carefully constructed and
accurate scale models of U.S. Navy ships. They were also modeling the
antennas and ship structures with NEC-4. One of the engineers confided
to me that they'd learned that when the measurement disagreed with the
model result, the model result was probably better. Of course, these
people were very highly skilled in using the modeling software and how
to avoid and detect its limitations. But they were also very highly
skilled in making the best possible measurements.

A claim by an antenna manufacturer or creator that an antenna "can't be
modeled by NEC" very often means that NEC's accurate results don't back
up the manufacturer's or creator's inflated claims. If the claim
includes a statement that some alleged physical rule or phenomenon
("critical coupling" comes to mind) isn't "accounted for" by NEC, you
can be certain that the disparity is due to inflated claims rather than
a shortcoming of NEC. There are, of course, some antennas that truly
can't be modeled with NEC, for example a patch antenna with dielectric
between the patch and ground plane. But those of NEC's limitations that
aren't apparent from the basic program structure have been pretty well
discovered and documented in the 30 years or so it's been in use.

Antennas are being designed every day with NEC and EZNEC which provide
critical communications functions for military and government agencies,
aerospace companies, space agencies, domestic and international
broadcasters, scientific researchers, and many others. Many of those
antennas have been carefully measured and verified before being put into
service, and countless of them are in current use. I personally have
designed a number of antennas with EZNEC, tested physical prototypes at
a professional range, and seen them put into daily service performing
critical functions as designed.

Roy Lewallen, W7EL