W8WWV - 40 Meter Parasitic Delta Loop Array

Greg Ordy


I have always liked delta loop antennas. I think it goes back to the early 1970's, when a nearby amateur who used a single delta loop had a tremendous signal on 40 meters. He also had a Collins 30-S1 amplifier, and no doubt that helped too.

I never erected a delta loop because I lacked a high enough support that was free from neighboring obstacles that would always get in the way of the loop wire. When I finally installed a tower for an upper HF antenna I got the support and space I needed. I decided to construct a two loop array using a driven element and a parasitic reflector. The array would be designed for 40 meters. Open coax stubs would be used to provide the loading necessary to turn the parasitic loop into a reflector. I had used that technique on other antennas, and I felt comfortable with it. Since I am interested in DX, I chose to feed the loops near a lower corner, in order to obtain vertical polarization. A relay is used to select the driven loop, the other becomes the reflector. The array will have two possible directions.

When fed in the vertical polarization mode, a single loop is conceptually similar to a pair of sloping verticals fed in phase. The verticals are the sloping sides of loop, and the bottom wire acts as an elevated radial. L. B. Cebik, W4RNL has an excellent series of pages that describe a number of these so-called self-contained verticals. A pair of loops becomes an array of 4 verticals, two driven and two as parasitic reflectors.

Following the analysis on the W4RNL pages, I selected a right-angle loop as opposed to an equilateral loop. This choice slightly improved the loop gain, as well as shortened the height of the loop, making the base higher for a constant apex location. For modeling purposes I took the initial dimensions from the W4RNL example, which is to say a 60.8 foot base, and 43 foot sides.

My tower has a convenient mounting point at around 55 feet above the ground. Rope would be used to pull the loops away from the tower. The spacing between the loops could be controlled by selecting the distance from the apex of the loop to the tower, along the diagonal support rope. Since the rope slanted towards the ground, loop spacing would be (inversely) related to loop height. As the loops became further apart, they would also become closer to the ground. In other words, as the loops conceptually slid along the support rope they would move from being right next to each other at the 55 foot level to eventually touching the ground while being quite far apart. A little trigonometry indicated that I could vary the loop separation distance between 20 and 30 feet while keeping the bottom of the loop around 20 feet off of the ground. From the W4RNL pages, a 20 foot bottom height provided acceptable performance, and it's also high enough to make sure that the wire will not get in the way of ground activities.

20 to 30 foot space at 40 meters represents approximately 1/8 to 1/4 wavelength spacing. This is a common range of front to back spacing for multiple element end-fire antennas.

An antenna very similar to what I constructed is shown on page 12-8 of ON4UN's Low-Band DXing, the third edition. His version is cut for 80 meters.

My next step was to use EZNEC to model the antenna. I was interested in performance, and the feed point impedance, in order to see what, if any, matching was required. I was particularly interested in the front to back (F/B) ratio of the antenna at low take-off angles. At this part of the United States, the eastern part of the midwest, the big problem with 40 meters is foreign broadcast stations. The European broadcasters start to roll in before sunset, and they nearly chew up the entire phone segment with their powerful and wide AM signals. I have heard people talk about the watering holes on the phone part of the band. This is a description of the handful of clear frequencies, barely the width of a signal or two. The term watering hole is borrowed from the interstellar search for life, where some deep thinkers believe that certain frequencies will be used for communication and greeting because of their universal characteristics known to all intelligent species. At times, I'm still trying to find intelligent life on this planet.

On 40 meters, my experience is that you need at least 20 dB of F/B ratio in order to be able to turn down European broadcasters enough to be able to hear the west coast stations, especially the west coast stations with dipoles and 100 watts. The band really does take on a new character when the broadcasters are cut down. Any F/B above 20 dB would be wonderful, but if it starts to drop much below 20 dB, Europe is just too strong.


I created an EZNEC model of two loops with a single source and a length of open coax hanging off of the parasitic element. I had to pick some initial parameters. I used the loop dimensions stated above. I used 75 Ohm coax for the parasitic stubs. Since I assume that the impedance of the driven element will be near 100 Ohms, the 75 Ohms section will be a good starting transition towards the final desired impedance of 50 Ohms. I set the bottom of the loop at 20 feet off of the ground. Finally, the spacing between the two elements was set at 25 feet.

The W4RNL web page provided dimensions for a single right-angle loop. The only information on feed point location, however, was that it is about 12% up from the bottom of one leg. When I created my  EZNEC model, I specified that the source was 88% down from the apex, which would be 12% up from the bottom. My model used 24 segments per wire, and that caused the actual location of the source to be placed at 89.58% down from the top, which is 10.42% up from the bottom. Since I obtained good performance with the models, I went with this feedpoint location. Here is a diagram of a single loop.

Initial 40 Meter Delta Loop Dimensions (from W4RNL web page)

The red dot is the feedpoint. This is a so-called full or one wavelength loop. The length formula, shown on page 5-2 of the ARRL Antenna Book (18th Edition), is 1005 / F(MHz). In our case, that would be 1005/7.15 = 140.6'. The loop specified on the W4RNL page has a length of 43 + 43 + 60.8 = 148.8'. Of course this is designed to produce a resonant loop, which I interpret to mean the reactance is zero. The tables on the W4RNL page show that the feedpoint impedance of a single loop varies with height above the ground. So, it's not clear if we are comparing apples to apples. In any case, I initially went with the W4RNL measurements in order to keep down the number of variations. As the story unfolds, the loop size will become an important factor.

In EZNEC, the two loop array will require six wires, three per triangular loop. Here is a picture of the wires in the array, taken from EZNEC.

40 Meter Parasitic Delta Loop Array Wire Diagram

Wires 1, 2, and 3 represent the driven loop element. The red circle on wire 1 is a source, which is where we will connect the transmitter to the loop. This location is selected to be 1/4 wavelength down from the apex. Please consult the W4RNL pages for the different feed point alternatives, and their implications. Wires 4, 5, and 6 form the parasitic loop. The dimensions are identical to the first loop, and the second loop is spaced 25 feet behind the first loop. The blue/green circle on wire 4 is the connection point of a transmission line. This will be the stub that turns the back loop into a reflector. In this picture, the maximum gain is down the positive X axis.

My only variable to experiment with was the length of the coax stub connected to wire 4. Although at this point in the project I was mainly concerned with this antenna in a conceptual sense, it's a good idea to make sure that any transmission line models can also be physically realized. This keeps problems from happening later, when they may require reworking of the design. For example, it does no good to have a transmission line which must be 20 feet long electrically but also physically connected to two points that are 30 feet apart. That just won't work. If that is your only solution, it is often times possible to add an extra half wavelength to the line, and maintain the same results. This is expensive in cable, not always possible, and may add more loss. So, whenever I model transmission lines, I try to make sure that they are at least as long as the physical constraints. In this case, the distance between the two connection points is 25 feet. This suggest that the transmission line really should not be less than 13 feet. If they were any shorter, the two ends could not be brought to a single point, at a central switching box. Even 13 feet is too short, since the common connection point would have to be located exactly between the loops, up at the same height. This might not be convenient. So, it seems as if a longer line length would give us more freedom to locate the switching box. Because the loops are hanging off of a tower, it makes sense to bring the transmission lines back to the tower. If nothing else, this keeps them from falling to the ground, and creating more obstructions under the antenna. I cut that grass, and I don't want to be moving cables every week! The trigonometry of the loops implies a desired line length of around 30 feet, since this would have the connection point very near the tower, at the center of the whole antenna.

Once that approximation is known, it will determine if the line should be terminated in an open circuit, or a short circuit. When a transmission line is less than 1/4 wavelength long, a shorted stub will have inductive reactance, and an open stub will have capacitive reactance. Since we are creating a reflector, we can be fairly certain that we need inductive reactance. But, once the line is longer than 1/4 wavelength, the reactance sense switches for the next 1/4 wavelength. That is, an open stub has inductive reactance, and a shorted stub has capacitive reactance. With an assumed RG-11 75 Ohm cable, with a velocity factor of 0.78, a 30 foot cable is more than 1/4 wavelength long. That indicates that we probably want an open stub, not shorted. An open stub switch can also be implemented with a single DPDT relay, which keeps the switching circuit simpler.

So, with all of this in mind, I started to model versions of this antenna with different length open stubs of RG-11 connected to wire 4. All modeling was done at the target frequency of  7.150 MHz. The range of interest I found was between 27 and 30 feet of length. While one could certainly model the line inch by inch, I decided to model at the one foot increment. In practice, higher accuracy is lost in all of the various implementation flaws and errors. The length of 28 feet had the highest F/B, at 29.46 dB. This was at the take-off angle of maximum gain, which was 18 degrees. In all modeling, I used the real/high accuracy ground model, with above average ground quality, which I believe represents my actual area. Here are the azimuth and elevation plots of that antenna model.

Initial Design Azimuth Response Plot Initial Design Elevation Plot

Several attributes caught my attention. First, I had a greater than 20 dB F/B ratio, which was a goal. The response at high take-off angles was low. That's good for reducing local signals and noise coming in at a high angle. Of course that suggests that this truly is more of a DX antenna rather than local ragchewing antenna. The take-off angle of maximum gain is a low 18 degrees, which is a good value for 40 meters. The 3 dB azimuth beamwidth is 92 degrees. Usually, we want a narrow beamwidth to reduce noise reception. In this case, however, the implication of this beamwidth is that we will always have poor coverage off the sides of the array. The array can only be switched in two directions - to the right, and then in the reverse direction to the left (on these plots). The side response will always be almost 10 dB down from the maximum gain. If you really need coverage of all directions, this may be a fatal flaw. Since I had other 40 meter antennas, and I was in experimental mode, I wasn't bothered too much. In any case, positioning of the antenna array should be carefully considered. You get two directions, and lots of holes.

In addition to these plots, I captured simulation data at stub lengths of 27, 28, 29, and 30 feet. That data is shown in the next table.

Initial Design Properties as a Function of Open Stub Length (7.150 MHz)
Stub Length (ft) Gain (dBi) F/B (dB) RDF (dB) Take-Off Angle (degrees) Beamwidth (degrees) Current Ratio (wire 1 / wire 4) Current Phase (degrees) Source Impedance
27 5.32 23.59 11.63 18 90.8 0.92 121 91.22 + j 62.72
28 5.26 29.46 11.51 18 92.0 0.92 117 95.67 + j 60.01
29 5.20 29.08 11.40 18 93.2 0.93 113 99.65 + j 56.68
30 5.12 23.58 11.28 18 94.4 0.94 109 103.10 + j 52.84

This table gives us a lot more information about what to expect. The gain is rather constant, without regard for the stub length. The F/B ratio, however, is very sensitive to stub length. This is the usual case, where nulls are very sensitive to many aspects of the antenna set up. The RDF is the receiving directivity factor, a metric which I discuss on other pages. 11.5 dB is a very good RDF value, and suggests that this will be a good receiving antenna (so long as you are pointed in the right direction).

The current ratio is calculated from the reported current on the wire right at the source, and right at the transmission line stub. I wanted to capture this value for several reasons. First, it gives us a sense of  how much current is being coupled into the reflector. Only the driven loop is physically connected to the transmitter. The parasitic loop receives all of its power from mutual coupling with the driven loop. According to these results, there is actually more current in the parasitic reflector than in the driven element. This does give me some concern. My experience has been that when you don't want mutual coupling, you get more than you want. When you do want mutual coupling, you get less than you want. The coupling will be influenced by the alignment of the loops. As I look at this high degree of coupling, I just have to wonder if I can get that much out in the yard. The current phase is the phase difference between the driven and parasitic loop right at the two feed points. The second reason why I like to capture the current ratio and phase information is so that if I need to figure out why the antenna doesn't seem to work as expected I can actually measure those values and compare them to the expected. The measurements can be made with RF current probes and an oscilloscope. From these modeling results, it appears as if the stub has far greater impact on current phase as opposed to magnitude.

Finally, the source impedance tells us the impedance right at the source point on wire 1. With the real antenna, we will connect a stub of transmission line to each loop. Those two lines, which are RG-11, will go to a common point where a relay will select one line as the real driven element, and the other line will be left open - as the true stub. This is how we switch directions. This means that the driven element will have at least a stub connected to it before we can connect more line on the end. The important impedance is not so much the feed point impedance on the loop, but the impedance of the driven loop after it is transformed by the stub. That is what we will need to match back to 50 Ohms. Since the line is long enough to reach the ground, there is some hope that I can measure it, if I want to make a good match. This array is a good candidate for a series matching section.


Before I left this model, I also modeled the antenna at two other frequencies. My goal here was to see how well the pattern held up as a function of frequency. The results were not very good. At 7.050 MHz, the F/B dropped to 6.03 dB. At 7.25 MHz, the F/B dropped to 10.36 dB. At best, we will have a narrow sweet spot with this antenna, in terms of rear response.


Don't forget, this is just a model of an antenna. It's a paper design exercise. This is probably as good as it gets. It assumes many things, in particular, two loops which are perfectly aligned with each other. I doubt if I can build it that perfectly. From here, the various compromises and problems will start to creep into the implementation. Still, I found these results encouraging enough to build one.


Designing an antenna is one thing. Building the design is usually something else. There are always so many variables and little details. For example, the W4RNL models used #12 solid wire, whereas I was using #14 stranded wire (for the loops). I would probably never get the real wires in the exact locations specified in the paper design. I knew I would want to put current baluns at the feed points in order to reduce radiation from the feed line. What would that balun do to the length of the coax stubs? What about the impact of the metal tower holding the whole antenna up? What about the ground quality under the antenna? Should there be a radial field under the antenna?

For me, I tend to view the paper design as a suggestion of what I really want to build. In some cases it's a goal, or even an upper bound of performance.

I got out the #14 stranded insulated wire, two baluns (Radio Works B1-2K), and some Dacron UV-resistant rope. After a few hours of work, I had two loops up in the air. It certainly looked like an antenna.

Each loop required three tie down points. Two corners, and the top. The top rope pulls the loop away from the tower, and establishes the loop separation as well as height. I wanted these ropes to be tied off in trees, ideally at least six feet off of the ground. That's motivated by weekly grass cutting. Ropes which come down to the ground would create a problem with cutting the grass in the area around the end. I also needed to pick six points which allowed the two loops to be as parallel and aligned as possible. Given all of these factors, and where Mother Nature chose to plant trees, I ended up orientating the loops with a nearly exact east-west signal orientation (probably 80 degrees). The loops would be run in a nearly north-south plane, and the maximum signal strength would be at 80 degrees and 260 degrees. This also happened to closely align the array with other antennas I had up at the time, so I could perform some antenna to antenna comparisons. Perhaps the only down side would be that my rear null would not be directly pointing at Europe, which can be found at approximately 45 degrees at my location.

After I tied off the loop ends, I started to walk around and size up what I had really erected. My first observation was that the loop bottom was not 20 feet off of the ground. It measured approximately 13 feet off of the ground. It turned out that I could not pull the top ropes far enough out from the tower to elevate the loop to the height of my model. If I would have spent a few more minutes with pencil and paper, this would have been easy to see, given the location of my potential support points. Since height was tied to spacing, I suspected that I would not be at the modeled 25 foot distance. A quick check showed 22 feet of spacing. The good news was that the two loops appeared somewhat parallel. One loop was nearly vertical. The other seemed to tip in about 6 feet towards the tower at the top.

By the way, did I measure the SWR of each loop as I put them up? No, and be careful if you do. First of all, SWR is a comment about the match of the antenna to the line, not a comment about the antenna performance. As is often said, the SWR of a dummy load is perfect, but nobody considers it to be a good antenna. For whatever you measure after you put up the first loop, it will surely change after you put up the second loop. The loops are coupled together, and one influences the other. Don't start adjusting the loops to improve the SWR at this point, that is almost always a serious mistake. This is an antenna system, not two individual antennas.


Now that I had a real antenna, I could take the dimensions back to the modeling program, and see how this antenna configuration might work, and what, if anything I would need to do to the length of the stub, my only variable. The stub length for maximum F/B at the maximum gain take-off angle decreased from 28 to 26 feet. Here is a table of modeled results for several stub lengths.

Actual Antenna Design Properties as a Function of Open Stub Length (7.150 MHz)
Stub Length (ft) Gain (dBi) F/B (dB) RDF (dB) Take-Off Angle (degrees) Beamwidth (degrees) Current Ratio (wire 1 / wire 4) Current Phase (degrees) Source Impedance
25 4.99 27.35 11.30 21 89.0 0.95 128 90.72 + j 82.0
26 4.96 39.95 11.19 21 90.0 0.96 124 95.34 + j 80.54
27 4.92 29.28 11.08 21 90.9 0.96 121 99.69 + j 78.54

While the gain and RDF slightly decreased, the maximum F/B actually went up by 10 dB. The take-off angle increased by 3 degrees. The source resistance remained nearly the same, but the source reactance increased by approximately 20 Ohms. This really impacts the matching system, not the performance of the antenna.

I made a few other simulations at this point, in order to help me understand what I was starting to build, and to set my expectations appropriately. The term front to back (F/B) ratio usually is computed from the gain in two opposite directions, aligned with the direction of maximum gain.  With a fixed array (not on a rotator), very few signals are directly off the front or the back. I like to compute a front to back ratio that  is in directions other than the direction of maximum gain. Perhaps some software packages compute this data, but I'm not aware of them. I don't care about the back half (180 degrees) of the antenna, and because the antenna is symmetric around the center axis, the performance will be the same to each side, at the same angular deviation. Here is the F/B ratio as a function of the azimuth angle away from the front center of the antenna.

F/B Ratio as a Function of Azimuth Angle Deviation
Angle 0 Degrees 15 Degrees 30 Degrees 45 Degrees 60 Degrees 75 Degrees 90 Degrees
Front Gain (dBi) 4.96 4.59 3.55 1.91 -0.17 -2.43 -4.59
Rear Gain (dBi) -34.99 -31.01 -19.1 -12.95 -9.31 -6.81 -4.61
Front - Rear (dB) 39.95 35.60 22.65 14.86 9.14 4.38 -0.02

Since these numbers reflect one side of the total response, the actual beamwidth is twice these amounts. For example, the azimuth rotation around the center point that has a F/B ratio of at least 22.65 dB is 60 degrees, 30 degrees each side of center. When you deviate 45 degrees off of the center line the F/B drops to 15 dB (14.86). This sort of information matters when listening to stations on the air, and trying to figure out if the expected F/B performance is being realized.

Another interesting trend is the F/B ratio and gain relative to the maximum as a function of take-off angle in the primary direction. Here are the numbers for this example.

F/B Ratio and Relative Gain as a Function of Take-Off Angle
Take-Off Angle 21 Degrees 30 Degrees 42 Degrees 54 Degrees 66 Degrees 78 Degrees
F/B 39.95 27.22 17.8 11.94 7.52 3.98
Relative Gain (dB) 0 -0.64 -3.02 -6.81 -12.05 -14.69

At 21 degrees we see the previous data. As the take-off angle increases, the F/B drops, but so does the gain of the antenna (at that angle). At higher angles, typical of night time local NVIS stations, the F/B may only be a few dB. That won't provide much discrimination against unwanted signals. The relative gain drops by over 10 dB with respect to the gain at 21 degrees. The higher angle local signals are rejected not so much by the F/B ratio, but by the overall reduction in response at higher take-off angles. This makes the antenna relatively quiet on reception, and improves the RDF.


Let me now present the antenna schematic which includes the control system. Many of my antennas use the technique of inserting a DC or low frequency AC control voltage onto a coax cable, so that a single line is both signal and control. I use a control box next to the radio for all antennas that required this signal. The control box can produce 4 different voltage levels. 0, +12, -12 volts DC, and +12 volts AC (60 Hz). These levels are sufficient to control two independent relays, assuming the use of diodes and filter capacitors on the relay coils. One relay responds to positive voltage, and the other to negative voltage.  In this case, I need to control a single relay, so the required voltages are 0 and +12 volts DC. The components located within the control direction switching box are enclosed in the red box.

Antenna Array Schematic

I insert the control voltage into the transmission line near the radio shack, where all of my cables and transmission lines leave the house. All of the other components at the antenna were assembled into a common Radio Shack plastic box. I picked a plastic box so that the parasitic loop open stub would truly have no electrical contact with the driven element transmission line. Here is a picture of the box. Click on the picture for a larger view.

Stub Control Box
Stub Control Box

The box has three SO-239 connectors. One is the input from the radio, which includes the control signal. That is the connector at the bottom of the picture. The other two connectors go to the two loops, through the stub lines.

I used a blob of adhesive caulk to glue the relay (which is in a sealed plastic case) to the bottom of the box. With the addition of a terminal strip or two, all other connections are made by standard point to point wiring. The relay is also a Radio Shack part (275-218), made by Potter and Brumfield. It has a 12 volt coil, and DPDT contacts which are rated at 15 amps as 125 VAC.  I have used this relay on many projects, and it works well, up to the 1500 watt level.

When the control voltage is 0 volts, the relay is not energized. The front loop is connected via the DPDT relay to the transmission line going back to the shack. The rear loop transmission line stub is left open. In order to switch the direction of the loop, 12 volts DC is inserted on the transmission line. It turns on the relay, which causes the front loop to become parasitic with an open stub, while the rear loop is connected to the transmission line going back to the radio.


You may think that this story is reaching a conclusion. Actually, it's just starting. My concerns seemed to come true. In the end, I could not achieve a 1 to 1 current ratio between the driven and parasitic loops. The simulations suggested that I would actually have more current in the parasitic loop. I was hopeful, but not optimistic.

This is not to say that I have trouble with the modeling software. On the contrary, I would be lost without it. But, my experience has been that the estimates of parasitic current can be optimistic, especially on low to the ground nearly vertical antennas. Yagis and quads seem to model just fine, and they are usually parasitic. But in the few times I have modeled and built arrays like this one, I just never can get enough current in the parasitic elements. One solution is to move to an all driven array. The complexity goes up, but so does the performance. In any case, let me drag you through the story as it unfolded.


I set up the oscilloscope and the current probes.  The current probes were inserted 6 inches from the balun, on the apex side of wires #1 and #4. The open stub was hanging off of the parasitic loop.  I used my MFJ-269 as a signal source. I dialed in the target frequency of 7.150 MHz. Here's what the scope showed (please click on the picture for a larger view).

Initial Current Measurements
Initial Current Measurements

The taller trace is the driven loop. The smaller trace is the parasitic loop. The phase difference is approximately 100 degrees, not too far from the target shift of 120 degrees. The current magnitude, however, was 3 to 1 in favor of the driven loop, not slightly less than 1, which was the modeled result.

So now what do I do?

I started to adjust the frequency on the MFJ-269, seeing if the ratio improved. The ratio improved to approximately 2 to 1 (driven to parasitic) at 6.766 MHz. Ok, the frequency is low, so I should start shortening the loops - right? Well, not so fast. First of all, that's work. Maybe something else is going on. My first thought was that the support tower in the middle was accepting current. Acting as an unwanted antenna element. I measured the current in the tower near the base of the tower using the procedure described in the ON4UN book. Alas, no detectable current. Maybe it was the beam on top of the tower? Maybe it was something to do with my ground quality? Perhaps, but outside of laying down a radial system, the soil was the soil. What about the current baluns? They were not even in my EZNEC model. My next though was that the loops were not aligned precisely  enough to each other. As I made more and more measurements, I saw that one corner of the loop was at least 2 feet higher than the opposite corner. I adjusted several of the support ropes, and the current magnitudes stayed very much the same.

After running out of alternatives, I decided to shorten the loops. Earlier on this page, I noted the length difference between the W4RNL web page, and the usual loop formula in the ARRL Antenna Handbook. The difference was about 8 feet. On the other hand, the simulations of this antenna showed good performance at the longer length. I would hate to change the wrong parameter.

Where should I shorten the loop? In the bottom leg? The side legs? All of the legs? Laziness prompted me to shorten the bottom leg, since it was easy to access. I removed 2 feet from each loop. The point of maximum parasitic current did indeed move up almost 100 KHz. Well, if 2 feet are good, 4 feet are better. Again, the frequency moved up. If 4 feet are good, 6 feet are better. The frequency moved up, but now the problem was that because I was shortening the bottom loop wire, the whole loop came closer to the ground, since the side legs became more vertical, pushing the bottom down. No matter what, I was done taking wire out of the bottom leg.

Still, the point of maximum parasitic current was too low. I decided to take 1 foot out of each side leg. This meant that each loop was shortened by 8 feet - 6 on the bottom and 2 on the sides. This would bring me close to the ARRL Antenna Book length, and how wrong can that be?

After all of this work, the current ratio near the target frequency of 7.150 MHz was 1.5, still far away from the desired magnitude ratio of approximately 1.

Whenever I made measurements, I also experimented with different lengths of the parasitic loop stub. I had a number of small sections of RG-11, and I was able to easily configure different lengths with barrel connectors. I found that I could indeed achieve a 120 degree phase shift. At least that was not causing trouble. At one point I inserted a variable capacitor at the end of the stub. It was quite obvious that as you changed the value the parasitic loop current changed in both magnitude and phase. One of the reasons why working on antennas can be frustrating  is that most adjustments change more than one aspect of the antenna. It's all a closely linked system.

My last thought was that the tilt in at least one of the loops could be causing the problem. I decided to work hard to get the loops totally vertical, parallel, and aligned. After all, in the EZNEC simulation they were perfectly vertical and aligned.  In order to get the loops vertical, I had to pick support points that also caused the loops to become closer together. With a spacing of approximately 18 feet I had two very parallel and aligned loops. The reduction in spacing suggested that the 120 degree phase difference would need to be increased in order to achieve the largest null off of the back. 140 degrees would improve the pattern.

In the end, the best current magnitude ratio that I could achieve was approximately 1.25. That is, the parasitic loop had 80 percent of the current in the driven loop. I also looked at the high and low ends of the band. As expected, performance tapered off in each direction, but not as bad as the simulations suggested. Of course I was not achieving the best performance of the model. Perhaps this was some small victory. Here is a picture from the scope showing an improved current relationship. This was close to the best I was able to achieve. Remember, the simulation software predicted that the current in the parasitic loop would be larger than in the driven loop. I could never get it higher than 80% of the driven loop current.

Improved Loop Current Relationship
Improved Loop Current Relationship

The final stub length that produced the desired phase shift was approximately 32 feet.

Another small victory was that the final configuration had an SWR of 1.1 at 7.160 MHz. It was unnecessary to add any further matching network. I certainly would have preferred the extra parasitic current, and a feedpoint impedance that required further matching.

Although I was unable to achieve the performance of the modeled antenna, the performance was still quite good. Taking my final array dimensions and  putting the information into EZNEC, here are the azimuth and elevation response plots.

Final Azimuth Response Plot Final Elevation Response Plot

The gain dropped by a fraction of a dB, and the RDF was 10.24 dB. The primary performance reduction was in the F/B ratio. This is usually the case when there are phasing errors.

So why was I unable to obtain more current in the parasitic reflector? There are many factors, and possible explanations. One is certainly the orientation of the loops. Were they parallel and aligned? This was a factor that I could adjust,  and I tried to get the loops aligned. It is also possible that there is loss in the transmission line stub. Another factor, the one that I suspect was most important, was the effect of ground and local objects. If power could be coupled into the parasitic loop, it could certainly couple into other surrounding objects.  One test I never conducted was to flip the direction of the antenna while monitoring the antenna current. This simple test might have revealed clues about the antenna environment. I suspect that this antenna would have benefited from either a ground radial system, or, being located higher off of the ground.


I find that I can get a handle on antenna performance during the daylight hours, when signal levels tend to be much more stable, and propagation tends to be by low angle ground wave, since D region ionization attenuates most higher angle signals. Local stations, relatively close to me, can be good reference signals.

I never ask stations for signal reports, that's just too prone to error and misunderstanding.

I use two approaches to performance measurement. The first is my S Meter Lite program, which provides an S Meter with improved calibration. The second technique is comparison with any other antennas I might have up at the particular moment.

Using S Meter Lite, I was able to measure 15 dB of F/B ratio on some broadcast stations. As I mentioned earlier, much of Europe is at a compass heading of 45 degrees from my location. With the modeled  response shown in the last section, the highest F/B ratios occur at approximately 45 degrees off of the back of the antenna. This happens to work well with my east-west orientation. In general, the model suggests a 15 dB F/B ratio across the rear  90 degrees of the antenna (at the lower elevation angles). At one point I had a version of this antenna up with a measured F/B of over 20 dB on some signals. Those extra dBs are significant, and I wish that I could have attained the modeled 25 or 30 dB of F/B ratio. Although the poor current ratio prevented that performance, that performance would have been over a relatively narrow frequency range. Still, if adjusted to the 7.150 MHz to 7.200 MHz span, it would really knock down the broadcast stations in that popular range.

My bi-directional Flag antenna was up while the delta loops were up.  I have found this antenna to be a good reference antenna, and it operates from 40 meters down through the AM broadcasting band. It's performance will be a function of frequency, but so long as you adjust your antenna model for the frequency, you can obtain results which I believe are quite reasonable.

In this case, the two antennas were aligned in (nearly) the identical orientation, and each antenna had two selectable directions. The modeled Flag antenna predicts an average gain of -16.24 dBi, and the modeled delta loop array has an average gain of -6.35 dBi. The difference is approximately 10 dB. I view the average gain as being very close to what I would call the background noise gain, or the response of the antenna to noise in the absence of signal. This will not be exactly true since noise is not uniformly distributed in all directions. Still, several hours after sunset, noise will be arriving from largely all direction, perhaps with a non uniform intensity. I tuned my radio to a clear frequency, and compared the background noise level between the two antennas using S Meter Lite. Across many tests I measured 9 to 10 dB of difference. The delta loop background noise level was 9 to 10 dB higher than on the Flag.

It is more difficult to assess forward gain, since you have to be sure that the received signal is coming from the direction of maximum gain.  The modeled maximum gain of the Flag is -9.47 dBi (at 7.150 MHz), and the maximum gain of the loops is 4.69 dBi. The difference is approximately 15 dB. I did measure a 15 dB difference on the Voice of Russia station on 7.180 MHz. That station seems a little off to the side of the front of the antenna for that difference, but the results are indicative of the expected difference between the two antennas.

My 160 meter vertical is resonant on the 40 meter band. I suspect that my large center loading coil acts as a trap on 40 meters, and shortens the antenna to approximately the size of a 40 meter 1/4 wavelength vertical. As such, the gain would be several dBi. The gain of the delta loop array should be, in the preferred direction, 2 to 3 dB above the vertical. Of course that assumes that the vertical is acting as a 1/4 wavelength vertical, which it certainly might not be.

In any case, the voice of Russia station on 7.180 MHz is approximately 4 dB stronger on the delta loops than on the vertical. Radio Tirana, in Albania, on 7.160 MHz is approximately 2 dB stronger on the loop array. It's quite risky to trust those sort of measurements, given QSB and other instability in the atmosphere. Still, it seemed the case that the loop array was stronger than the vertical in the preferred direction.

All of these measurements suggest that the loops are working reasonably well, and not too far from modeled results.

Would I Do It Again?

The short answer is no. While this array has many strong points, there are just about as many weaknesses. I don't believe that it should be the only 40 meter antenna at a general purpose station. It might be a secondary antenna, designed for DX operation in two very specific directions. If the antenna supported more than two forward directions (see below), I would be much more excited about it.  Here are some attributes, in no particular order.

40 Meter Parasitic Delta Loop Array Attributes
Attribute Ranking Comments
Simplicity Good  There is not much to this antenna. It's easy to model, and easy to build.
Beamwidth 90 Degrees  The beamwidth, at approximately 90 degrees, is similar to a 4-Square. Unlike a 4-Square which has four directions, this antenna has only two. The sides really are at a disadvantage. The problem is that there are only two directions but a 90 degree pattern.
Take-Off Angle 21 Degrees (Good)  DX take-off angles on 40 meters are often quoted as being between 10 and 30 degrees.
Cost Low/Medium  This antenna is low cost. Wire, rope, baluns, and some transmission line. Other alternatives, such as a 4-Square of verticals, or a Yagi on a tower, would cost substantially more.
RDF Very Good  The receiving directivity factor is very high with this antenna. It tends to be a quiet antenna. The response at higher angles is quite low. By the way, I have used this antenna from time to time to listen on 80 meters, and it often times does a very good job.
Local/DX DX  This antenna is more of a DX antenna due to the low take-off angle and narrow beamwidth. It is not as effective in local operation, which usually requires high angle omni directional coverage.
Bandwidth Poor  The rear null largely disappeared at 100 KHz off of each side of the target frequency. This does not even cover the entire 40 meter band.
Size Large  The footprint directly under this antenna is 60 X 22 feet. The 6 tie point ropes push out the overall span of the antenna another 40 to 50 feet. That's a lot of real estate for a one band antenna, especially one with just two directions.

Over time, I have tended to think of this antenna as my California/Africa antenna. It's two main directions point at those destinations, and I find that it works very well with stations in those areas. Whenever African DX shows up on 40 meters, I can hear them well, and I tend to bust through the east coast stations, who are 600 miles closer to Africa than I am. I do run 1500 watts, and that certainly helps too. It also works well into Australia. Japan, however, is too far off the main lobe of the pattern. All of South America is off of the side, so that area is at a 10 dB disadvantage. Asia is off of the other side, and also at a disadvantage.


The work on this loop actually took place over about 3 years. Some permutations existed for many months. It remains a good DX antenna. The inability to achieve the optimum current ratio made little impact on the gain and RDF. The rear null suffered the most. If I were to do it all again, I would drive both elements, and obtain the correct current relationship, or go QRT while trying. I would not construct this sort of array unless I was willing to measure the element currents and have a lot of patience to adjust the array for best performance. Given the variations in local conditions, this is one antenna that must be adjusted for each situation.

If you are going to drive both elements, in order to improve performance, then you should also take some time to look for other useful phasing relationships. Based upon some discussions I have seen on the Internet, it is probably possible to produce Figure-8 patterns as well as omnidirectional patterns, with different phasing. Reference has been made to a December 1993 RadCom article by Tony, G3LNP, which develops these permutations. Since one of my concerns was the limited direction choices of the antenna I built, having some additional choices would be a good thing.


For my tastes, I believe I would prefer using a 1/4 wavelength vertical with a good (32+, 1/4 wavelength long) ground mounted radial system, and a companion receiving antenna, such as a K9AY or Flag. That sort of receiving antenna would have good performance across the whole band, and operate on other amateur bands as well. The gain on transmit might be down a few dB, but often times that difference is not too important. When the atmospheric noise is low, the vertical could also be used for receiving, which would provide more flexibility and alternatives. You can never have too many antennas.

My EZNEC file, 40mdlactual.ez, can be downloaded by clicking on the link.


While searching the Internet one day, I came across another 40 meter delta loop array. This array, by K4TX, has inverted delta loops as compared to mine - they point downward. His comments about the F/B ratio and the lack of additional match my experiences.

Back to my Experimentation Page

Back to my Amateur Radio Page

Last update: Monday, July 12, 1999 06:33 PM
Back to my Home Page