[TowerTalk] Shunt fed tower for 160m

[Lux, Jim]

On 4/8/22 8:58 AM, Grant Saviers wrote:
> EZNEC has a feature to make a XYZ grid of wires, the manual advises 
> less than 1/10wl between them, although I use a finer mesh.  So some 
> insight can be gained with the roof insulated from ground (no walls 
> connecting it to ground).
>
> The hard part is how the metal roof is grounded.  With NEC4.2 a buried 
> ground can be set up, not so in NEC2.  Connecting the roof via 
> something (mesh, separate wires) can take a lot of time to set up, 
> then the high accuracy compute time can be significant.  I've also run 
> some models that run into numerical computation problems, I think 
> because of the large number of segments and junctions. The current 
> distributions made no sense and some tweaking of the model yielded 
> sensible patterns.
>
> From my modeling, I think (and one experience) the grounding of the 
> roof may have a significant effect.  Elevated radials should not be 
> grounded. 


the rule of thumb for "mesh to simulate solid surface" is that the 
circumference of the wire should be approximately the spacing. And 
around 1/10th wavelength is good, both for spacing and segment length.  
If you're doing it in raw NEC, the GM card and clever wire numbering is 
your friend.

The tricky parts are joining the meshes at the edges, so that segments 
line up.

These days, it runs pretty fast: I was running some models of two 
crossed fan dipoles supported by a steel pipe over a steel mesh laying 
on the ground, with the full SOMNEC soil model, and fill times for 2100 
segments were around 30 seconds and factor time about 2 seconds.  Fill 
time goes as the square of number of segments, factor goes as the cube.  
(This was on a 2 year old Macbook pro, running NEC4.2)