[TowerTalk] Stack Diversity Pattern

[Jim Lux]

At 08:14 AM 3/16/2005, Keith Dutson wrote:
>Assume a stack of four tribanders, each rotatable.  The top two are phased
>using a StackMatch II.  Same for the lower two.  The upper and lower pairs
>are phased using a third StackMatch II.  Thus, you have a myriad of possible
>antenna selections available (1, 2, 3 ,4, 1&2, 1&3, 1&4, etc. to 1&2&3&4).
>
>Is there a way to model the pattern when using all 4 and each antenna is
>pointed in a different direction?  For example, say the top is pointed NE,
>#2 pointed SE, #3 pointed SW and #4 pointed NW.  In your opinion, would this
>work well enough to select one antenna while receiving and then drop the
>others and use it for a QSO?  Thanks in advance for your comments.
>
>Keith NM5G

Yes.. this is actually pretty easy to model, even in a pencil and paper 
sort of way, as long as you aren't looking for gnat's eyelash accuracy, 
AND, you're not concerned about mutual impedances between the beam antennas.

Assuming that the antennas have a fairly narrow beamwidth (say, 8 dB 
antennas.. something around 60 degree HPBW), you could just draw the four 
patterns on top of each other, oriented as you described.

Since you've got 4 antennas on one feed, the power gets divided 4 ways 
(i.e. you get a 6 dB hit in absolute gain).

On the main lobe for any one of the antennas, the signal from the other 
antennas will typically be 15-20 dB down, so regardless of the phase, it 
won't change the response in that direction.

Where it gets dicey is in the areas between the main lobes (using your 
example, say you wanted to know what the gain is looking directly East 
(between antenna #1 and #2).  Both antennas will be about 6 dB down from 
their boresight gain.  If they happen to be in phase, the two will sum.  If 
they happen to be out of phase, then the two will cancel, and you'll get a 
null.  On most antennas, the phase is pretty smooth across the main beam 
(and across big sidelobes), but gets kind of weird in the nulls.

I did a quick calculation using a 3 element 20m yagi that comes with 4nec2 
(about 13 dBi gain. 6-deg HPBW), looking at the magnitude and phase of the 
horizontally polarized component.  If you hooked them all up in parallel, 
oriented as you described, the polar plot of the gain is almost flat. The 
gain ranges from 7.25 to 6.45 dB, with 4 "lobes" and 4 nulls. (Actually, 
there's a little bump right halfway in between the main lobes)

What's kind of interesting is that if you introduce random phase shifts (as 
if you were casual about the coax lengths, or if you had the antennas 
mounted on different towers), you don't see a lot of difference in the 
overall pattern.  What you do get is fairly sharp nulls, where things 
happen to line up just wrong.

If anyone is interested, I have an Excel spreadsheet that lets you play 
with this fairly easily.