[TowerTalk] modeling help?

[Jim Lux]

At 01:04 PM 11/24/2004 -0500, you wrote:
>I am not a modeler, and honestly have no desire to go thru the learning 
>curve necessary to have confidence in the results.  That said, is there 
>one of you out there who would be willing to do some modeling for me?
>
>My plan is a three tower antenna farm.
>
>Tower 1 is a stack of Opti Beam OB-16-3's at 40-80 and 120 feet, and a 40 
>yagi at 130 feet all on a rotating 55 tower.  Hanging from the top from 
>phyllistran  catenary lines will be an 80 meter four square vertical array 
>and I will shunt feed the tower on 160.   (currently it is an 80 footer 
>with 40 beam at top, 80 vertical 4 square hanging off it and a 160L 
>attached to the top.)
>
>Tower 2 is a guyed 80 foot 45 tower 120 feet away from the one described 
>that will be converted to a WARC array. This one will hold a pair of M2 5 
>element 17 beams at 40 and 80 feet, a 4 element M2 12 meter beam at 85 and 
>a 2 element 30 meter yagi at 90 feet.  (currently it is an 80 footer with 
>an OB 16-3 at the top and a C3E on a tic ring at 40 feet)
>
>At some point in the future I would put up tower 3; a 40 footer with a C3E 
>fixed on EU, making a triangle of towers with about 100 feet between it 
>and each of the taller sticks.
>
>All guys are/will be phyllistran.
>
>Question is, what if any interactions can be expected and should perhaps I 
>not be considering such a complex trio of arrays so physically close 
>together?  I have analyzed the take off angles etc. using HFTA and am 
>impressed with its potential from my flat as a pancake location. I am also 
>very impressed with the single Opti Beam OB-16-3 currently up at 78 feet, 
>and the C3E below it at 40 feet is strong into EU.   Anyone wish to offer 
>some help?  We will first discuss compensation.  Thanks and Happy 
>Thanksgiving. 73 bob de w9ge

I think you need to define what you're looking for in a bit more detail. 
Here's some things to think about.

For instance, are you looking for accurate predictions of gain?  (how 
accurate?  1 dB, 5 dB? 0.1 dB?)

Are you concerned more with forward gain or with things like F/B or F/R 
ratios.  The relative uncertainty in a model for small signals is much 
greater than for large signals. That is, if you're computing front/rear 
ratios, the error in the rear facing power is going to dominate the overall 
uncertainty.  Say you radiate 100 watts in the forward direction, and 1 
watt in the backwards direction. Your F/B is 20dB.  But say the uncertainty 
in both those numbers is 0.5 watts.  The F/B could be anything from 
100.5/0.5 (23dB) to 99.5/1.5 (18 dB), but it's mostly the uncertainty in 
that backwards power that creates the overall uncertainty. The forward gain 
only changes by 0.02 dB.  (this is why designing antennas with sidelobes 
down 40dB is a challenge!)

Or, are you looking for a more qualitative planning model to identify 
possible trouble spots, and to evaluate various physical arrangements (for 
which you really only need to look at whether the induced currents are 
"significant").

For instance,HFTA makes some fairly generic assumptions about the antennas 
you're stacking, ignores any vertically polarized components, and ignores 
mutual coupling among the antennas.  But, as you note, it's easy to fool 
with different spacings and get a feeling for how the pattern changes.

On the other hand, if you use something like NEC you'll get a good model of 
the antenna system with mutual coupling among elements(and potentially, any 
feed system effects).  However, none of the terrain effects will be covered 
(NEC assumes "flat earth").

And, as far as fidelity of modeling goes, NEC (even NEC4) doesn't deal with 
tapered elements particularly well, so, if you have tapered 
elements,  while the pattern will be reasonably accurate, the feedpoint 
impedances won't be so accurate (which affects the accuracy of the feed 
network model, if you're running multiple antennas through a BIP/BOP 
stacking box).

And, another important point.. what do you want at the end of it?  A report 
on the results, or a model that you can run and alter?  If the latter, what 
modeling tool will you be using (EZNEC, NEC4WIN, 4NEC2, MultiNEC, NEC4, 
etc.etc.etc.) because the input files aren't all the same.

I suppose what I'm trying to get at is that this could be a trivial project 
or a hideously complex one, depending on what you expect to get out of 
it.  Just the tedious mechanics of entering the physical model is part of 
it; although one that you can do a lot of the "grunt work" on, since it's 
just measuring/researching dimensions, and typing them in.  Especially if 
you're doing a very high fidelity model, the time to enter and validate the 
geometry is non-trivial.  None of the modeling tools have a "Rohn 45" block 
that you can just drop in. (I should qualify that.. none of the tools 
commonly used by hams.  For all I know HFSS has some fancy stuff like that 
available, but for $60K a seat, it should)

Jim, W6RMK