Since I will not be repeating myself you should use the link above to refer back to that article. For this one I have some objectives in mind when I use the word optimize or, if you prefer, improve the design:
- Simplicity: The more common 4-square used on 80 meters in many high-performance stations is in the judgment of many hams a good return on investment of time and money. But at its heart it is not simple, dependent as it is on some intricate phasing and power splitting electronics. My preference is for an antenna with similar performance that is less vulnerable to weather or component failure, and is far easier to tune and keep in tune.
- Gain and F/B: Gain is paramount to me though I'll take all the F/B I can get on 80 meters. With noise so prominent on the lowest bands gain has an outsize effect compared to higher bands since signals are often so near the noise level. Odds of working the rare DX improve and many more QSOs can be logged during contests. The same applies to F/B but only on receive, which can alternatively be addressed with a separate high-directivity receiving antenna. Thus for me gain trumps F/B.
- DXing and contesting from my QTH: For me the most productive directions are Europe and central USA, which happen to be in exactly opposite directions. The 4-quadrant switchable directivity of a 4-square is nice, in general, but when aligned to those two directions the other two are not so useful: namely the north Pacific and south Atlantic Oceans. One of those will at least garner some JA QSOs while the other is mostly useless, except perhaps for instances of skew-path propagation. I am willing to sacrifice performance in those directions to do better towards Europe and central USA. Other, simpler antennas can fill the pattern holes.
With that introduction out of the way let's proceed to improving (or optimizing) that 2-element parasitic vertical antenna.
Step 1: Optimize the element spacing
The 0.25λ element spacing I used in the earlier article was not selected for its superiority. Rather I chose it as a reasonable value that would allow radial coupling for overlapping radial systems, good though not necessarily optimal coupling between the monopoles, and lastly as an initial basis of comparison with 4-squares (typically arranged in a square with 0.25λ sides). Don't confuse this parasitic antenna with a dual-fed 2-element end-fire array which has a different optimal element spacing.
I ran the EZNEC model by adjusting element spacing in steps of 0.05λ while changing nothing else about the antenna. As before each standalone element is resonant at 3.6 MHz. I successively changed spacing in the increasing and decreasing directions until it was clear that the performance possibilities were exhausted. The model (which has close to 500 segments) requires some work at each step in order to properly maintain radial topology, so it is not a trivial process.
- Move one element to the required position. I nominally set λ = 84 meters (3.57 MHz), and I rounded the spacing to the nearest 0.5 meters to simplify radial calculations without significantly affecting the results.
- With the element moved it is necessary to extend or contract the radials that must be connected. These are radials that would otherwise cross at their nominal length of 20 meters. The included angle between radials must be preserved. I used Cebik's method here, just as I did in the earlier article. All connected radials connect at a line orthogonal to and that bisects the line between element centres.
- Segment counts for the connected radials must be adjusted so that segment length in all radials is as close to equal as possible. This assures best accuracy from NEC2. I normalized segment length as 1.25 meters. I was able to get within a few percent of this for all of the truncated (connected) radials.
From studying these charts I believe that 0.25λ is the optimum spacing. You can get more gain at 0.20λ but the gain bandwidth is narrower. Better F/B can be had at smaller spacing, but again with narrower bandwidth and with higher SWR. SWR increases at smaller spacing due to decreasing feed point resistance (~25 Ω at 0.15λ spacing) and higher Q (more rapidly changing resistance and reactance with frequency).
In all cases the gain and F/B curves can be moved up or down the band by adjusting monopole length (standalone element resonance). SWR curves can be shifted with a matching network as simple as a series capacitor. Impedance, and thus SWR, will change with different ground, wire type and gauge, and radial count, though probably not by a lot.
Gain relative to a single vertical of the same construction is greater by 0.7 db than shown in the charts since its gain is -0.7 dbi. All gain and F/B values are for an elevation angle of 15°, a good median value for typical DX paths on 80 meters. Over medium ground and the specified radial system the gain peaks at an elevation angle of about 25°.
Step 2: Double the gain
Now that the 2-element array is optimized we can stack them. That is, put an identical array beside it and try for 3 db of additive gain. The idea is to exceed the gain of a 4-square with the same quantity of verticals, and to accomplish it with the same simplicity as each 2-element antenna.
Use of term stack is deliberate and accurate. It is the same as we saw in my article on the basics of stacking yagis. The differences are that the two arrays are ground-mounted, 2-element vertical parasitic antennas and that they are stacked horizontally rather than vertically. In other respects the stacking arrangement is the same.
For this to work we must space the two antennas far enough apart that mutual coupling does not grossly interfere with the additive nature of the array. This is complicated by the radials which would touch, and therefore need to be connected for predictable behaviour, at a spacing of 40 meters or less. Thus the minimum spacing is 42 meters, or 0.50λ. There is no theoretical maximum spacing, though land use and transmission line loss are constraints. We ideally want the minimum effective spacing. So let's try 0.50λ and see what we get.
Gain peaks at 6.33 dbi at 3.65 MHz, and again standardizing on 15° elevation. The gain remains in a narrow range over the band segment of interest. In comparison to the single 2-element vertical array the gain increase is proportional to frequency, from 3 db at 3.5 MHz to 4.5 db at 3.8 MHz. The mutual coupling between the close-spaced antennas is having an effect, and that effect is in our favour in that the gain is for the most part higher than the 3 db expected with zero mutual coupling.
Mutual coupling is also having an effect on F/B, and again it is mostly beneficial. We are now seeing some reasonably good F/B, although only higher in the band (SSB segment). SWR has actually improved, staying below 2 from 3.5 to 3.8 MHz.
The forward lobe in the azimuth pattern has become somewhat narrow, spanning just less than 60° at the -3 db points. This is what happens in stacks: the gain comes from narrowing of the forward lobe in the plane of the stacking direction. Vertically stacked yagis narrow the elevation beam width and horizontally stacked antennas narrow the azimuth beam width.
The feed system is included in the model, composed of several length of coax. Performance will not be accurate if two in-phase sources are instead used in the model. Power splitting is accomplished by tying together (T connector) two λ/4 70 Ω transformers, and placing the single source there. I specified 14 meters of solid dielectric coax (0.66 VF), for which RG-11 is an example. The far ends of the transformer sections must be equal lengths of 50 Ω coax. The transformer lengths alone will not reach from the power splitter to the antennas.
I made the transmission lines loss-less in the model. Real (lossy) coax of the builder's choice can be substituted when selected. In any case the loss in the cabling is not high and would not substantially affect the modelled performance.
|0.75λ spacing between the 2-element antennas|
The pattern at right is for a spacing of 0.75λ (63 meters). This is not a pattern you are likely to favour. In an array of this type the spacing between antennas is an important parameter.
Step 3: More directions
While 0.75λ is not a particularly effective way to get more than two directions from this array it is possible to steer the forward lobe of the optimum 0.50λ-spaced array. We do this by switching in (or out) a length of 50 Ω coax between the 70 Ω transformer and just one antenna's feed point. However, exact phasing of feed points is not possible since the unequal lines alter the mutual coupling resulting in what might be unexpected results.
In the skewed pattern shown I have "switched" out 10 meters of 50 Ω coax going to the left antenna. While heavily distorted you can see how the main lobe has turned about 30° to the left and partially filled the side nulls. The same can be done in the right direction by the same change to the right leg of the feed system.
While far from the directional performance of a 4-square it is one way to compensate for the fairly narrow main lobe. Yet it is no substitute for getting good performance in the missing 2 quadrants. SWR changes little for the range of cable lengths differences I experimented with in the model. The 10 meters of coax I cut to skew the pattern was arrived at by trial and error; other lengths, I tried gave poorer results. Keeping the 50 Ω feed sections as short as possible (before switching a section out of one leg) gives better patterns.
Step 4: Direction switching and cable management
Direction switching and cable management in the elevated 40 ground plane array requires some care to avoid the coax and relay lines from degrading performance. The same is true here though with some welcome reduction in complexity. In fact the methods used have much in common with typical construction of a 4-square.
For an individual 2-element ground-mounted antenna the presence of the coax (transmission line to shack and between elements) can be put to use rather than cut out of the picture using common-mode chokes. Relays for direction switching are also simpler. This can be done since these conductors are in the same plane as the radial systems, and several of the radials are connected.
First, let's look at the coax and switching line between elements. Unlike the ground plane array we can permanently connect the ends of the coax outer conductor to the radial origins at both elements. It is only the centre conductor that needs to be switched. We are in effect turning these cables into radials. If the cables are buried deep (for weather and rodent protection) they can be complemented with a separate surface radial above the cables and running directly between the elements per the previous article. No chokes are needed. An SPDT relay at the far element switches the monopole between the radial system (reflector element) and the coax centre conductor (driven element).
The element where the transmission line connects still requires two relays, though these can now be SPDT rather than DPDT. One switches the monopole between the second relay (driven element) and the radial system (reflector element). The other switches the transmission line centre conductor between the first relay (driven element) and coax to the far element (reflector element). The transmission line outer conductor connects to the radial system and to the outer conductor of the coax going to the far element.
The transmission line can also become a radial if run where a radial would normally run and it is (optionally) terminated out 20 meters with a common mode chock. In the 2-antenna array these runs would terminate at some convenient point between the 2-element antennas where power splitting takes place. It is likely that the coax-radial will include a transition from 50 Ω coax to the 70 Ω transformer so there may be a need for superior sealing of that joint. If DC relay lines are run the same way (central DC/RF splitter rather than one for each antenna) similar common-mode precautions are needed for the DC lines.
All transmission lines in each half of the array must be identical for the required symmetry. This includes the 50 Ω lines from the 70 Ω transformers to the feed points and the 50 Ω line between elements.
As with stacking yagis there is the alternative of running 50 Ω coax from the power splitter to the individual antennas by using a 2:1 transformer rather than λ/4 transformers. This has a few benefits worth considering:
- There are fewer joints to protect from weather and burial, especially the ones at the far end of the λ/4 transformers.
- One antenna can be easily disconnected from the array. This is a simple way to broaden the forward lobe (135° beam width versus 60°) and thus reduce the need for array steering. Although we give up 3 to 4.5 db gain when doing this it is one way, for example, to cover almost all of the continental USA from my QTH. In contests this can be highly desirable.
- Pattern skewing as described earlier can be done at the power splitter by switching in (or out) a length of 50 Ω coax feeding either of the antennas.
As I stated up front I am willing to sacrifice ultimate performance in two of the directions provided by a 4-square if I can get better performance in the two directions of greatest importance to my operating. This array can accomplish that and do so with electrical simplicity and a little more land to accommodate the greater separation required between the 2-element antennas.
The cost is one of time and effort to roll-your-own rather than buying a commercial 4-square splitter/phasing box. However the tuning effort in my opinion is simpler and quicker with this array than with a 4-square. You also won't find one day that the "dump" resistor of a 4-square warns of failure or burns up hundreds of precious watts due to weather or other adverse event that will occasionally upset the sensitive tuning of the array.
To me this is good food for thought. Since I am not yet in a position to build either antenna there is time to consider my options. Your situation and objectives may be different and so should your choices.