[TowerTalk] the value (or not) of modeling

[Jim Lux]

Periodically on TT (and in other forums) the issue of modeling, accuracy of
modeling, and the value of modeling comes up.  I thought I'd throw out my
thoughts for comment.

First, I'd like to start with a quote from the late R.W. Hamming (of the
eponymous error-correcting codes): "The purpose of computation is insight,
not numbers".

Modeling is wonderful!  It lets you experiment in nice cozy warm comfort
while it's raining outside, or while the sunspots are few, or while you're
just too darn lazy to go out and build it to try it.  Especially now, with
2+GHz computers available for $350 from Dell, running half a dozen
speculative cases is so quick that there's no excuse not to.

On the other hand, there's no substitute for going down to Home Depot and
buying 500 ft of AWG 12 THHN wire and stringing up some antenna, and just
giving it a shot!

So where does modeling fit in the overall scheme of things?  I think the
first thing to realize is that the "professional" modeling area (which I do
at work) is very different from the "amateur radio" area.  We have different
goals, different budgets, different objectives. If you're working on a
design for an antenna for a spacecraft that costs $100 Million to launch,
then you can afford to spend a bit of cash to make sure it's going to work
before you build it.  On the other hand, if you're scrounging parts at
hamfests and want to make a good showing in the SS, maybe your money is
better spent on a good P.E. to wet stamp the plans for that 100 foot tower
than on a copy of NEC4 from LLNL.

There's also the issue of using models to understand what's going on, in
general, rather than predict to the gnat's eyelash what your forward/back
ratio is going to be. Given that construction and environmental tolerances
for most antennas will be in the few percent range, expecting gain numbers
to be better than a few dB is unrealistic.  For instance, I've been hunting
for a good NEC model of a tract house for years, and haven't found one yet
(for that matter, I haven't even found a bad model). Since the vast majority
of us have houses near our antennas, this would be a useful thing.

So, given that modeling isn't going to give us exact answers, what good is
it?

Going back to Hamming, it gives us insight.

I don't have a model of my house, or of the rain gutters, in any sort of
precision.  However, I can make a decent assessment of the impact of my
gutters fairly easily, using models.  I model my antennas (dipoles laying on
the tile roof).  Then, I add in wires where the gutters are (approximately..
within 10-20 cm).  I run the model and look at the current in the "gutter
wires". If the current is low, I figure, hey, the gutters don't matter.  I
compare the pattern with gutters and without.  Hmm, 0.02 dB difference.
Yep, the gutters don't matter.

Or, maybe I want to know how critical the dimensions are (given that I'm
lame with a tape measure).  Build the model with the ideal dimensions.  Run
it.  Now, change the dimensions by 10cm.  Run it again. Sure enough, the
pattern looks pretty similar, but the feed point impedance has changed a
fair amount.  Hey, I've got an autotuner at the feed point: what do I care
about feed point impedance.  What I care is that the pattern isn't horrible.
Whether the pattern matches, exactly, what I've modeled is immaterial.

The big gotcha in modeling is losses.  Ground losses, losses in surrounding
conductors, losses in the antenna conductors themselves.

There was a great article by Brian Beezly (K6STI) in one of the ARRL Antenna
Compendiums describing his experience simulating a (he thought) new, nifty
antenna design.  At first it looked great., until he started putting in
lossy components.  All of a sudden, that great performance turned not so
great.  The W8JK type antennas are notorious for this.  Great directivity,
getting ever better as you move the two elements closer.  Ooops, if you
factor in element resistance, it doesn't look so hot, because the element
currents get real high, so the IR losses get huge, not to mention the
problems in feeding an antenna with a feedpoint impedance of 0.1 ohm.  Small
resonant loops have the same problem.

Again, though, even though losses are tough to model accurately, you can get
a feel for the impact of the loss.  Change the resistivity of the elements
by a factor of 2. See what happens.  Does the efficiency drop like a stone?
Does the F/B ratio go away? Or, does it work fairly well still. A bit more
loss perhaps, but still a decent pattern? You've got a winner.  Build it,
and 10 years from now, when all the joints are corroded, it will still work
pretty much like how it worked when you built it.  On the other hand, if
changing resitivity by 2 causes the pattern to die, and the efficiency to go
to heck, watch out.  You'll be posting comments to TowerTalk asking about
what sort of climbing gear you should have, and how to keep your tramlines
in order, because you'll be bringing that antenna down every year for
maintenance.


So, use that model for insight, not to predict your run rates in the next
contest!

73,
Jim, W6RMK