Thursday, March 24, 2016

3-element, 4-direction Vertical Parasitic Array for 80 Meters CW

Before describing this antenna it may help to declare my thoughts on the 4-square vertical array. This is an antenna that has become the "standard" for competitive stations on 80 and 160 meters. They are occasionally used on 40 meters where big towers and yagis are not feasible..

The attributes of the 4-square that I find unappealing are:
  • Complex and often finicky power division and phasing by use of a hybrid coupler. They can and do fail at the most inconvenient times.
  • Dump load that, ideally, dissipates well under 100 watts due to system imbalances. Imbalances can be due to weather, freeze/thaw cycles, aging, element and feed faults and other component issues. It's difficult to control.
  • Selected directions must be at 90° intervals. This is perhaps a small limitation since the -3 db beam width is greater than 45°.
  • Expansion of the array to more elements requires a total rebuild, and an even more complex power division and phasing system.
  • Performance and power handling requires a precision in element tuning and phasing that can be trying.
  • Typically requires 4 structurally independent 20 meter tall elements made from tubing or lightweight tower, guyed and electrically isolated from ground.
On the other hand the antenna has many positive attributes. These include:
  • Excellent F/B and compass coverage.
  • CW and SSB segments can usually be accommodated without tuning and little loss of gain and F/B.
  • Commercial products are available for the hybrid coupler and direction selection, which removes a significant technical challenge.
  • Lots of contesters and DXers use them so there is a community you can lean on for advice.
A parasitic array has its own set of issues, so the decision comes down to trade-offs and personal choice based on local circumstances and what one is comfortable with. These will become apparent as I proceed to describe the antenna.

My objective is to find an uncomplicated and relatively simple directional array for 80 meters to be used in my next station. The 3-element vertical yagi looks promising. Like the 3-element vs. 2-element wire yagis for 40 meters and 2-element vertical yagis I discussed some time ago, the use of 3 elements simplifies the physical and electrical design by introducing symmetry, and has better performance.

What you think is for you to decide. I suspect that most would choose to go with the tried and true 4-square. In fact I more than suspect it since I see that choice being made all the time by serious DXers and contesters for whom performance is a never-ending pursuit.

Where it comes from

This antenna is not my invention. It is a variation on a 160 meter parasitic array by K3LR, as described in ON4UN's Low-Band DXing on page 13-42 of the 5th edition. My aim is to delve more deeply than the brief description it received in that text. I believe it is worth a closer look. If you haven't read it or don't have the book, get it and read it. Don't stop there: read the entire book. It's a gold mine of information for those serious about high performance on the low bands.

I know one ham who has built the K3LR vertical yagi pretty much as described and I know he gets out very well on 160. But it does require a 40 meter tower with a base insulator! The challenge on 80 meters is not so severe, if one has the land and the motivation.

The driven element (tower) is tuned with an L-network and the parasites are tuned as directors with coils switched in to convert one element into a reflector. Additional coils and L-network adjustment are needed to cover both CW and SSB segments. Radiation resistance is low so it is critical that many radials are used to bring the ground loss as low as possible. Gain and F/B are critically dependent on tuning and radials. SWR bandwidth is not broad but can be made sufficient for most purposes.

Physical design

A light-duty tower can serve as the driven element and as a support for the parasitic elements. The DMX-52 tower I currently use could work well in this application with a couple of added sections to bring to 19 meters height. I might just do that. The height does not have to be tuned with exactitude since we'll be using an L-network at the feed point. A short mast projecting above the tower provides all the adjustment room we'll need.

Isolation from ground will need to be improved with plastic spacers and fibreglass or similar pins rather than screws to secure it to its base. Non-resonant guys are necessary. These should be separate from catenary ropes for the elements. Radials are not connected to the tower.

Each element base requires a ground anchor to support a switch box and to take the tension applied to the wire element. The tension comes from adjustment of the catenary rope, which is anchored beyond the extent of the radial field. The wire elements in this design are full height, which requires a tower extension mast to attach the ropes. Use good insulators such as ceramic eggs at the tops of the wire elements. The ON4UN design uses a T-top incorporated into the catenaries so that a mast extension isn't needed.

The mast, if used, must be non-conducting or metal with a fibreglass coupler to isolate it. It can serve double duty as the top of a 160 meter vertical using the same tower and radial field. But I will set that option aside for coverage in a possible future article. I have drawn in the 160 meter vertical wire in the manner suggested in ON4UN's book. A trap can be used though initial modelling shows that the SWR bandwidth on both bands narrows.

Geometry and trigonometry will help us calculate what length extension mast we need. If the mast extension is 3 meters and the rope anchor is 30 meters from the tower, for an element spacing of 10.5 meters (0.125λ at 3.5 MHz) the mast extension must be 10 meters. That is excessive. Longer rope or use of posts for anchors can bring the mast down to a reasonable length. I would aim for a maximum mast of 6 meters (20'); it'll be guyed by design and, as already said, it can be used for 160 meters.

The radials reach outward from all 5 elements. Where they intersect they can be tied together (as shown in an earlier antenna design) or a bus wire running between them terminates the radials from each side.

The tower driven element requires the same arrangement when intersecting the other radial fields. This is the design presented in ON4UN's book. A diagram showing this is at right. Notice that the overlap of radial fields is more pronounced than in the 4-square due to the closer spacing of elements.

Coax transmission line and a multi-conductor control cable connect to the tower (central) element. Control cables, but not transmission line, connects to the base of each parasitic element. For best results they should be buried and common mode chokes employed.

Total area of the antenna is a square with 55 meter sides, or about 0.75 acre. This is slightly less than the 4-square.


As with any ground-mounted vertical antenna the radial system is critical. It is easy to lose several decibels with an inadequate set of radials, and even more with this antenna due to its lower radiation resistance. Since the radial field of this antenna is too extensive for modelling I have reverted to an alternative technique that is widely used.

The model uses MININEC ground rather than "real" ground in EZNEC. In this treatment wires can be connected to ground (not possible in standard NEC2). That ground is a perfect conductor or, if you like, a zero-loss radial field. Real ground -- EZNEC medium ground in my model -- is only used to construct the far field pattern, including the reflection ground losses and low-angle field cancellation due to phase shift.

A real radial system is emulated by inserting a resistance load in the base of elements connected to ground. The resistor represents the net ground loss for that element. Since the radials are not present in the model the resistance value must be carefully chosen from measured values published in the literature for the number and length of radials to be used.

For the initial run through the model I assume that all radial fields have an equivalent ground loss resistance of 3 Ω. This is approximately the loss for 60 symmetric radials of λ/4 length (20 meters, for a total of 1.2 km of wire) over good farmland. The effects of different radial arrangements and ground quality will be addressed further below.

The parasitic elements (wires) reach 19.5 meters high, and are constructed of 14 AWG copper (2 mm). The driven element is 20 meters, with a model diameter of 250 mm (10"), which is an approximation of a light-duty tapered tower such as DMX. The driven element is adjusted for resonance at 3.55 MHz, however this is not critical provided it is kept well within the band edges if it is also to be used as a omni-directional vertical. The K3LR design includes this option, and I would like to have that as well.

Performance and adjustment

The antenna is a 3-element yagi. That it is comprised of ground-mounted verticals and radials does not alter that. It therefore has a narrow usable bandwidth, with respect to match, gain and F/B. In my opinion this is its primary disadvantage with respect to the 4-square.

In its simplest configuration it does well over 100 kHz. This is sufficient for CW (3.5 to 3.6 MHz) or the main SSB segment for DX and contests (3.7 to 3.8 MHz). To get both, or other segments of 80 meters, requires switching of loads to tune the elements.

The antenna as described here is for the CW segment. Direction switching and band segment selection is in the next section.

The azimuth and elevation patterns are unremarkable for an array of this type. The broad azimuth beam width is typical of vertical arrays since radiation cancel poorly off the sides. Horizontal arrays have sharp nulls to the side since the elements on edge. Instead it is the elevation pattern that is narrow in a vertical array.

The gain is quite good and is comparable or better than a 4-square near its best frequency. The difference is that the gain bandwidth is narrow this vertical yagi. The same assessment applies to the F/B. Even so this antenna does quite well over the entire CW segment.

Note: be careful how you compare this antenna to the 4-square since the efficiency of each is critically dependent on ground and radial system. I will not do so in this article since any comparison can be deceptive. Not only for that reason, but because the antennas behave differently over the same ground and radial system, as we'll see.

The plot goes as high as 3.65 MHz for the interest of those using digital modes. It is still usable though not a great performer above 3.6 MHz. The reason is the radiation resistance. It is ~21 Ω at 3.5 MHz and declines gradually before steeply falling above 3.6 MHz. With a superb radial system and good ground the gain will remain good through to 3.65 MHz. Otherwise the array can be tuned higher but not without affecting CW performance. Gain, SWR and F/B rapidly degrade at lower frequencies, in this case below 3.5 MHz.

Of the four wire elements only two are active at any one time. The other two must be left floating by disconnecting them with a switch or relay, otherwise the pattern will be substantially degraded. The director is grounded (connected to its radial system) and the reflector is grounded through a 2.4 μH coil. Although coil Q is not critical in this application it is still good practice to make the coil diameter at least 40 mm (1.5") and inter-turn spacing approximately the same as the wire diameter. Make the coil inductance greater than required and adjust it by use of a tap or spreading the windings.

If SSB is your primary interest you have only to shorten the elements by the desired scale factor. Another and better alternative is described below in the section on switching.

The L-network consists of a shunt capacitor of 1,000 pf across the transmission line port and a series inductor of 1.2 μH . Again, the coil Q is not critical since losses are small. The capacitor should be a low loss ceramic transmitting type or a variable capacitor of suitable size. If a fixed capacitor is used you should determine its optimum value with a variable capacitor. The L-network coil can be tapped for tuning.

Direction switching and additional modes

The transmission line and an 8-conductor cable connect to a switch box at the base of the driven element (tower). The 8-conductor cable is used for direction switching. Signalling is low-voltage DC to power relays. A 4-conductor cable runs from there to a box at the base of the four wire elements. The transmission line and switching cable should be buried and use common mode chokes at the driven element.

Inside the shack a switch box selects one of 5 directions, one of which is for omni-directional mode. One conductor is for DC ground. The remaining 2 conductors can be used to select a sub-band (e.g. CW or SSB) or a 160 meter vertical utilizing the same radial field and tower, but its own matching network.

The switch box at the driven element uses a diode matrix to tell each element what to do. For example, when the array is to be pointed northeast, the northeast element is grounded (director), the southwest element is grounded through a coil (reflector), the other two wire elements are floated (open) and the L-network at the driven element is switched in. The reflector coil is shorted with an SPST relay when that element is a director. In omni-directional mode the L-network is bypassed (two DPDT relays) and the four wire elements are floated (open).

When SSB and CW are included the directors are tuned (shortened) as a director for the SSB segment. Its modelled length is 18.54 meters. A coil of 1.85 μH is placed in series with the reflector 2.4 μH coil. For SSB operation the 1.85 μH coil is shorted with an SPST relay at all four wire elements. The CW performance of this dual band segment array is almost identical, suffering only a slight decrease in gain and F/B due to the shorter elements.

The L-network, using the same 20 meter tower height for SSB, requires a shunt capacitor of 760 pf and a series coil of 0.12 μH, configured as before. For the CW segment the L-network is slightly different due to the shorter elements. Use a 900 pf shunt capacitor and a 1.0 μH series coil.

When 160 meters is included by means of a separate vertical wire parallel to the tower (as described in ON4UN's book) the transmission line is switched to the matching network for that antenna element and all the wire elements are floated. Since the radial field has an approximate circular diameter of 60 meters the 160 meter ground loss should be quite low.

I have not bothered to design the switching matrix for this antenna since these are adequately described in ON4UN's book, chapter 11. The matrix is placed in the switch box at the base of the tower. If the default (unpowered) selections of float, CW/SSB and direction are different from that shown in the above element switching diagram you should use SPDT relays instead of the SPST relays shown so that you can wire the normally open or normally closed contacts per your choice. My choice would be to default to omni-directional, where are wire elements are floating. In this case the element wiring is as shown in the diagram.

Radials and performance

The efficiency of the elements in a vertical antenna is a function of the ground equivalent series resistance and the current flowing through that resistance. The loss of the array is the sum of the element losses. This efficiency loss is not to be confused with ground reflection loss that often plagues verticals over real ground at low elevation angles.

For our ground-mounted vertical elements the return current flows in the radials and the ground beneath. The more and longer the radials the less current flows through the lossy ground. It should be apparent that the local ground quality affects the antenna efficiency for all but the most extensive (and expensive) radial system.

I will assume reasonably good ground such as that found on or near farmland since that is most likely where this type of antenna is likely to be built. If your ground is rocky or sandy, or even just very dry, the antenna efficiency will be lower. All you can do is put down more radials or elevate the antenna. Elevating an 80 meter vertical array is rarely practical.

The following chart plots the current in each element across the band segment. For comparison the current in the driven element when in omni-directional mode (single vertical element) is 4.7 A for a power of 1,000 watts.

The magnitude of the current does depend total ground resistance (including the radials) but it does not vary abruptly. So the selection of 3 Ω in this case is not material. What I want to draw attention to is the current distribution among the elements and relative to a simple vertical. Recall that P = I²R, so for a doubling of current the loss is 4 times higher.

Since the radiation resistance is lower than in a single vertical the current in the driven element is higher. It increases with frequency as the array's radiation resistance declines. The current in the parasites is lower, which is typical of yagis. At the 3.55 MHz centre frequency the driven element current is 6.84 A. In comparison to a single vertical (omni-directional mode) the loss in the driven element doubles to ~200 watts from ~100 watts. Taken together the loss in the parasites is ~100 watts, for a total loss of ~300 watts, or about -1.7 db.

That's a lot but it gets worse for a typical radial system and for poorer ground quality. Not only that, the radiation peaks at higher elevation angles as ground loss increases. If you think you can escape this by retreating to the 4-square, well, think again. The loss is similar and can be even worse since the radiation resistance is low and the currents are more equal. To increase efficiency we will have to be creative.

One method is to exploit the current imbalance for a more economical radial system. We want the lowest ground resistance for the driven element since it contributes the lion's share of the total loss. Therefore we should install the maximum number of radials we can afford from the driven element base to the bus bar that rings the driven element.

Possibly better is a metal mesh filling that area. For a 6 meter on a side square inside that bus bar that amount to 36 m² (360 ft²) so it is not cheap or easy. Some hams use galvanized chicken wire mesh (1" squares or hexagons) that comes in widths of up to 3' or 4'. Bonding the edges and connecting to the bus bar and feed system is a challenge since soldering is impractical. Further, copper to zinc contact promotes corrosion and zinc (the coating used for galvanizing) rapidly corrodes when in contact with acidic soil. A dense set of copper wire radials is likely the best choice.

The radial fields under the parasites can be more modest since the current is lower. Unfortunately we can't use a sparser radial field for the lower current reflector since every parasite can be a director or reflector. Further, due to overlapping radial fields the return current for the driven element also flows in the radials for the parasites.

We can model the behaviour by adjusting the load resistances in the parasites and driven elements, making the resistance lower for the driven element. Without collecting and presenting lots of data from the models I ran, including assumptions for the resistances due to the complex radial system, I will only say that this appears to be a promising approach to reduce total antenna loss.

Here are a couple of articles references I know of that provide some background on estimating the resistance associated with radial fields over different ground types.
And don't forget to look at ON4UN's book and also the ARRL Antenna Book if you have it. There are many other useful resources you can find by looking around.

More directions and gain

Unlike the 4-square this vertical yagi is not as limited in the directions it can point. Although the director, driven element and reflector must be on a straight line the other two parasitic elements do not have to be at right angles; that is, direction intervals of 90°. Depending on your global location this flexibility can ensure you get the maximum gain or F/B for the directions of greatest interest.

You can even have 6 or 8 directions with more parasitic elements and increasingly complex switching and radial system. You must only ensure that the unused elements are floated. Of course this may be a little excessive since the -3 db beam width is so broad that the extra directions can only gain you 2 to 3 db. That is, unless you add more gain...

If it's more gain you want you can add a second director, either for all four directions or even just one. Although the latter would require L-network selection in tandem with direction. The added director increases SWR, gain and F/B bandwidth and adds about 1.5 db gain for a director spacing of 16 meters. The second director is 20 cm shorter than the other directors, whether the measurement for CW or the one for switchable CW and SSB band segments. The 2:1 SWR 150 kHz. The gain band width stays excellent except at the upper 25 kHz of the 150 kHz band segment.

But this becomes a very large antenna. The point is that you have this option. You don't get that with a 4-square.

The good news is that the element switching is simpler for the second directors because they have no reflector coil. They either float or are directors for CW or SSB. Even the mode coil can be eliminated if you use the second director for only one mode. The radial system can be more modest since the current is relatively low (see discussion above).

Perhaps in future I'll describe the 4-element vertical yagi in more detail, just for the sake of doing so since I believe few would build it. I find it helpful and interesting to discover what is possible if we are willing to put in the effort.

Final thoughts

I hope that the material in this article adds some insight into the antenna as described in ON4UN's book. That was my intention, if only for my own interest. If you have the land perhaps this antenna will interest you as well.

I could grow to like this antenna . The performance is good and it has an uncomplicated electrical and physical design that appeals to me. It is still a lot of work, though less than I'd expect for a 4-square and with some potential advantages. Supplementing it with a directional receive antenna is desirable since the F/B is not exceptional. But then you'd also want one for a 4-square, so the comparison in this respect is not unfavourable.

If for some reason the vertical yagi does not work out it is not very difficult to convert the antenna to 4-square. The tower would become a support for the wire elements and double as a 160 meter vertical. The element anchors and their associated radial system would have to be moved to achieve the λ/4 element spacing of a 4-square, which is 20 to 21 meters, rather than the ~15 meters of the vertical yagi.

To start off I would build the tower vertical and then, over time, install the wire elements and build the switching system. At first it could have only two directions, with the others added later. By comparison, you must build the 4-square all at once. In parallel I could add the 160 meter extension to the tower. It would be best to install the bus bars, wire element anchors and full radial system right from the start. The radial system can be sparse at first and more added when the antenna looks like a keeper.

Although CW is my priority for 80 meters I would build the elements for both modes. Switching the elements for CW and SSB does not require a switching matrix, and can be easily added when I am ready.

Sunday, March 20, 2016

VFO Baggage

After returning to amateur radio after a 2012 hiatus I had to familiarize myself with very different technology used within the shack. When I went QRT in 1992 the typical transceiver had few memories and settings of various features was universal, not associated with a VFO. Indeed there usually was only one VFO.

Several generations of transceiver technology later we see the opposite. A large number of features are associated with each VFO, of which there can be many and even many registers per band. These include frequency, filter, bandwidth, mode, attenuator, pre-amp, noise blanker and noise reduction, VFO tuning rate, CW offset, antenna port, ATU and more.

Our rigs are more software than hardware these days so it is easy to keep this data around. It is intended as an operator convenience. It can be very helpful that when you change modes, bands, memory or VFO that you return to exactly where you were when you last used that selection, with same selection of feature settings.

Can it go too far? Let's look at a few examples from my recent experience.

Other operators

At a multi-op contest operation I brought along my FT-950 for the group to play with during the day (this was a 160 meter contest). Of course everyone has their own ideas about filtering and other feature settings to use. On switching bands or VFOs they'd naturally make adjustments to suite themselves. The next person would do the same.

By the end of the weekend the rig was pretty much unusable. Before putting it back on the air at my home station I had to undo all the changes. With 9 HF bands, 2 VFOs and 3 registers per band that amounts to a lot of changes! It wasn't quite as bad as that since not all combinations were touched.


In the recent ARRL DX SSB contest I changed the AGC setting to fast. Slow is more typical for SSB to make listening more pleasant by covering up band noise and lower-amplitude QRM and QRN. This is detrimental in a contest since many of the stations you need to copy are boxed in by louder stations that dominate the AGC.

In similar fashion I would narrow the SSB filter to as narrow as 1,800 Hz to contend with the crowded bands. Copy is enhanced despite the unnatural sound of the result.

Every band change, VFO change and band register selection required setting the AGC and filter width. After the contest these changes had to be reversed.


When noise appears I make adjustments. I may select a noise blanker or noise reduction, or I may narrow filters, turn off the pre-amp and make other tweaks. Some noise is frequency dependent while other noise affects multiple bands. In other cases it is antenna dependent.

Since these settings are all settable by band, VFO and register I am repeating the same sequence of feature settings every time I select a band, VFO or register. When the noise goes away I gradually undo those changes. That's annoying.

Counter-trend move

I purchased the Elecraft KX3 in late December 2012. It was the rig I chose when returning to amateur radio after 20 years of inactivity. At the time a small, QRP, moderately priced modern rig appealed to me. The research I had to undertake to reach that purchase decision leads me to sympathize with new hams. There is so much out there, and so much that was unfamiliar to me, that getting to a decision was difficult.

One thing that puzzled me about the KX3 was their design choice of having the filter settings associated with mode and VFO, not with the band. I thought that an odd and potentially annoying choice, though nothing more than a minor irritant. My opinion gradually changed.

Someone at Elecraft had clearly given the matter some deep thought. It is a design choice that I grew to appreciate. Filter settings do make sense when associated with the mode and not have change on me when I change bands. On rigs that do make the association it seemed I was often surprised at what filter settings came up. They were whatever seemed best to use the last time I was on that band. But who remembers that or even cares?

What about all the other settings that are remembered with every band, mode, memory and VFO register? Have we overloaded these registers for good reason, or only because it is easy to do with modern technology? That's what I have come to wonder.

Solutions are possible

There is great value in having many feature settings uniquely settable for different bands, modes, VFOs and memories. I believe that "feature creep" has overridden good operating practice. The question is where to draw the line. It isn't even clear that a line can be drawn that most would agree with. What Elecraft has done is one example of walking back the feature creep that makes sense to me.

One kind of solution is to assign personalities to a rig. If you use N1MM Logger or a WinKeyer in contests you will understand the concept. In a multi-op contest with Logger when you sit down you log in with your call and many of the feature settings and memories (include voice messages) switch to your own preference. The next operator does the same. It works well.

For CW contests with the WinKeyer the memories and feature settings are pushed aside while Logger is in use. This allows the full set of contest-friendly feature settings to be programmed in Logger and easily invoked. When you exit Logger the WinKeyer features return to their previous settings.

I can imagine the same being done for our transceivers. This could be directly programmed in rig software or via external software such as Logger. You could set personalities for yourself for when you're DXing, contesting or casual operating. All the settings follow the selected personality. When a visitor operates your station you assign them a personality. When they leave you get your station back to where it was with one simple command.

Most transceiver manufacturers are unlikely to take the initiative. It is more likely to come from the developers of software applications for rig control, including those for SDR. Most of the CAT (computer aided transceiver) controls needed are already present in the majority of late model rigs. It needs only to be done.

Friday, March 11, 2016

Hidden Openings

Those active in last weekend's ARRL DX SSB contest are likely aware of a peculiar opening on 10 meters on Sunday. The band to Europe and even further points from North America was better than on Saturday and arguably better than expected for a solar flux below 100 at this time of year. But that's when it got very strange. Here is an illustrative excerpt from my Cabrillo log:

QSO: 28484 PH 2016-03-06 1827 VE3VN         59  ON     S58N          59  K
QSO: 28507 PH 2016-03-06 1830 VE3VN         59  ON     F8ARK         59  1K
QSO: 28395 PH 2016-03-06 1837 VE3VN         59  ON     EA2LMI        59  100
QSO: 28402 PH 2016-03-06 1838 VE3VN         59  ON     HA1AG         59  KW

Notice the QSOs with stations in central and eastern Europe that took place up to 2 hours after their local sunset. These occurred immediately after I took a lunch break when I was sure 10 meters was finally closed to Europe. My operation was very casual -- 95% S & P and 17.5 hours -- which gave me freedom to tune around to see what I could find.

What happened? It was a common theme in the reports filed on 3830. The answer appears to lie with the sun. A bright visual aurora was reported at that time (early evening) in Scandinavia. That was a hint.

Later in the day I browsed to WM7D's web site to check on the geomagnetic indices. At right is a screen shot taken the next day that shows the entire story. The Kp index spiked to 5 in the period 1500-1800Z and kept climbing. My guess is that the auroral activity created a high MUF auroral-E region in northern Europe, linking European signals to the F-layer propagation over the north Atlantic.

Contests as a propagation tell

Let's face it, if there hadn't been a major contest underway this delightful opening is very likely to have gone unnoticed. Would you be looking toward eastern Europe on 10 meters at that time of day? I know that I would not. Contests drive an intensity of activity rarely seen at other times. The relentless pursuit by thousands of hams for QSOs and multipliers leaves no opening undiscovered.

Why openings are hidden

Hidden openings like these are only hidden because we don't look for them. This happens on HF more than many realize. It would not be typical on bands such as 6 meters where unusual propagation is the norm. There are many beacons and dedicated enthusiasts who find and advertise openings when they happen or appear imminent.

Imagine the following scenario. You might find it familiar. It's late afternoon and you are tuning 15 meters with the beam pointing northwest. The K index is low so you are expecting an opening to Japan and possibly southeast Asia. You hear nothing from that direction. So you call CQ. Again, nothing.

You then run across a Caribbean station, easily copied off the back of your beam. He finishes the QSO and you hear a pile-up respond. They are mostly Japanese stations. Clearly the band is wide open. What is going on?

The obvious answer is that they want to talk to the Caribbean station, not you. But of all the stations in that part of the world you would think a few are willing to work any DX. This is not an isolated example. The same result often occurs when the JA is calling CQ toward North America, and indeed for most hams thoughout the world who call CQ DX.

Tuning across the bands most days is not as inspiring as it was years ago. Casual QSOs, DX or otherwise, seem to be on the wane. My guess is that the casual QSO appeals less as we grow older since we've done it so many times. This shows up in the aggregate since the ham population is getting older here and elsewhere.

It was a surprise to me when I returned to the hobby after being away for 20 years. The prevalence of the "599" QSO, without so much as an exchange of names, seemed to have become the norm. What we now find is the intense pursuit of awards such as DXCC and its endless sub-categories, and contesting. Every day the bands seem emptier yet the participation in CQ WW and many other contests grows. DX competitions such as DX Marathon appeal to the retirement crowd (soon a majority of hams) who have the time to invest in these activities.

I am as guilty as anyone. Sure, I try new things, but DXing and contesting are the majority of my activity. To counter this I regularly make a point of calling most anyone just for the sake of doing it. But that doesn't compensate. Calling CQ on a seemingly dead band is something else I will occasionally do.

A story from my youth

I had the misfortune to become a ham in 1972 during the waning years of solar cycle 20. Like cycle 24 (the current cycle) the peak was mediocre. The decline therefore offered poorer than normal conditions, and the storm activity was high. During the depths of a minimum during the mid-1970s the solar flux would stay below 70 for days at a time. It doesn't get any lower than 66, the value of a quiet sun. Openings on 10 meters were rare and 15 meters was hit or miss.

Being a VE4 tucked just beneath the auroral oval the waning years of a solar cycle were a poor time to pursue DX. I still did of course; I really didn't know better, except for the fantastic stories the old times told of cycle 19. The contesting activity by me and my friends centred on ARRL Sweepstakes. The few DX contests I entered were done multi-op if only to relieve the monotony of depressingly low rates with conversation, especially on the low bands during the long winter nights. Working DX, sometimes any DX, could be quite exciting.

One day while relaxing at the university club station VE4UM a friend (still a friend and a contester) and I had a good-natured argument over who was the better DXer. The argument went something along the lines that I had the smaller home station and so had to hone my skills more than he did with his better station. We decided to put it to the test.

We walked over to the club's operating desk. Each of us would in turn call CQ DX on 20 SSB (this was the middle of the day) and see who caught the better DX. Well, this is hardly much of a contest but it was simple to do and it appealed to our humour.

I went first. I turned the yagi north (to enthusiastic laughter from my friend and a few onlookers). The first CQ came up empty. The second drew a weak signal that required asking for a repeat. It was a VU station. We were all stunned. Who would have thought that there'd be a workable opening over the pole midday in mid-winter in the depths of a solar minimum? Yet there it was.

When my friend sat down for his turn the pressure was on. I would be hard to beat. To more laughter his CQ drew a response from a friend of ours from across town who called in to say hello. The accidental discovery of a hidden opening won the day.

The lesson

How many more hidden openings are there? I'm sure there are many. Unless it's a band with numerous watchers and beacons, such as on VHF, most likely go unnoticed. The reason is that we don't try. That's a shame. The magic of radio is in part about not knowing what might happen when you venture to transmit a CQ into the aether. Clicking on spots from the global spotting networks is convenient and a great time saver, but we should also occasionally venture beyond that.

Every now and then I put out a CQ into an apparently dead band. You just never know what hidden opening might be there for the taking. The anticipation of who might answer harkens back to decades ago when every QSO brought excitement. These foolish attempts to find an opening no one else suspects helps to keep me young. That the majority of these lonely CQs fails doesn't damp my enthusiasm.

Tuesday, March 8, 2016

3-element Coil-loaded Yagi for 40 Meters

The all time top 3 posts on this blog are designs for 40 meter wire yagis. In retrospect this is understandable. On 20 meter and above a rotatable yagi is within the reach of most hams. Below that it is not. Wire yagis are an economical and effective way to get a boost on 40 meters for DXing and contesting. From personal experience I know they can work very well indeed.

But now I am looking to join the minority with a 3-element rotatable yagi for 40 meters. The XM240 I have in storage is only part of my future plans. I also want something with better gain and performance across the band. I have the tower and the rotator to make this possible. Since a full size yagi is so challenging I am exploring 3-element yagis with shortened elements. The quest is to find how close we can come in performance. I am certainly not the first!

Selecting a shortening mechanism

There are a few ways to shorten dipoles and yagi elements. We can use one or more to meet our design objectives.
  • Loading coils
  • Capacity hats
  • Linear loading
  • Traps
  • Bending the elements
When yagi elements are electrically shortened the antenna can be a nightmare to tune. It is no surprise that most hams who want such an antenna will opt to buy rather than design and build. There is something to be said for paying someone else for doing the heavy lifting.

On the other hand there may be a lingering suspicion that the compromises in a commercial design may be greater than advertised. Too often these products focus on low SWR rather than gain. The former is easier for the customer to observe, plus it makes operating more convenient.

This should not be taken as a smear against commercial products, many of which do measure up very well. In my quest for a larger rotatable yagi for 40 meters it is only natural that I would try to design something suitable before making the decision to buy or build, or to choose a full-size yagi. The knowledge gained is worthwhile no matter the outcome.

For example, the Cushcraft XM240 uses coils and capacity hats. Moxon rectangles bend the elements, and may even use loading coils. What they all have in common is that gain must be compromised, whether by inefficiency (loss) or reduced aperture. SWR bandwidth is often reduced as well because shorter elements can have lower radiation resistance.

The capacity hats in the XM240 are small and therefore have only a modest impact on shortening. Their advantage may be more about increasing element coupling to raise feed point impedance and improve F/B. But I haven't studies this closely other than simplified computer models. Those capacity hats are vulnerable when working on the antenna, so there had better be some performance benefit!

Loading coils must be high Q to keep loss low in a yagi, which means large, and large means increased wind and ice loading, and risk of breakage. Linear loading has similar efficiency challenges, and with greater vulnerability to ice. Traps are like coils with respect to loss, with the further disadvantage of not reducing element length all that much. Do it if you feel you must have a huge multi-band yagi (such as for 30 and 40). It will be challenging to design, build and tune. This won't be the first time I've linked to VE6WZ's site for high Q coil construction. Have a look if this style of yagi interests you since it's a critical component.

I chose to keep it simple: coils. With a target Q of 600 and elements 70% of full size it is possible to design an effective 3-element yagi for 40 meters that is reasonably rugged for the VE3 climate. It was for this purpose I earlier modelled elements of this type.

Modelling woes

Tapered elements are not well modelled with NEC2. The Leeson stepped diameter correction (SDC) in EZNEC deals with that nicely, but not when the element contains loads. EZNEC 6 is reported to correctly model such elements under certain constraints. Unfortunately the resulting model's correlation to reality is not always predictable. I've chosen to stick with version 5 for now.

Since NEC4 is not in my immediate future I modelled the element as constant diameter. This is sufficient to closely approach the true antenna pattern and impedance behaviour even though the length will be incorrect. As we'll see there is a way to translate the finished model into a working antenna by doing some work with the antenna in the air. Even with NEC4 there will be enough difference between the model and reality that a similar tuning procedure would be required.

Design procedure

Designing a 2-element yagi is straight-forward, even simple. All you do is make a resonant element for the design frequency and then place a copy of it next to it. You can adjust the element spacing to meet your objectives. If the performance is at the wrong frequency all you do is shift the resonant frequency of the parasitic element (reflector in this example). You then choose a feed system for the driven element, such as a beta match, and adjust for best SWR.

A 3-element yagi is almost as easy. That is for one with full size elements. Add loads and it can become a chore.

Let's walk through the steps.
  1. Model a resonant driven element of the desired length as a percentage of a full-size element. Place a coil at the mid-point. Adjust the coil value, not the element length to bring it to resonance at the selected centre frequency. I chose 7.1 MHz. Make sure the coil in the load table has the correct ESR (equivalent series resistance). Use the formula R = X / Q.
  2. Make two copies of driven element. Place one behind (reflector) and one ahead (director) of the driven element. As a first step place the driven element slightly back from boom centre since this usually improves performance and allows room for a boom-to-mast clamp.
  3. Next, shorten and lengthen the director and reflector, respectively, by about 6%. Do this even if the element contains a load, whether a coil, trap or capacity hat. check the gain and F/B over the desired frequency range. Don't worry yet whether that frequency range matches what you want. However be sure to keep the loads on all element the identical distance from the boom. This will come in handy later. It will help to model each half element as two wires with the inside wire a constant length and the load at the outside edge. The other wire is the one you adjust.
  4. Adjust the length of one of the parasites until the F/B and gain are optimized, or at least as good as you make them. You'll know you're close if the performance is similar to that of full-sized yagis with the same number of elements and boom length that you can find in the ARRL Antenna Book and other texts.
  5. We still have no matching network so the SWR will be quite poor. All you need right now is to get the feed point resistance (the R component of Z = R + jX) as high as possible while keeping the gain and F/B as good as possible. Remember to check this over the frequency range, not just one frequency. You want performance over the whole band! The process is iterative.
  6. Move the resonant frequency of all the elements by the same percentage (that is, multiply length by the same number) to move the performance to the band or band segment of interest. Do it again if necessary.
  7. Design and model the matching network and adjust the network and driven element to optimize the SWR curve. Perfection may not be possible, but if you paid attention in step 5 you ought to come close.
  8. Save the model! Now, delete the driven element and one parasite. Place the source on the remaining element and measure the reactance at the centre frequency. Write down that number, or save this model as well. Open the full model and do the same for the other parasite.
  9. Build the antenna. Use a tubing taper schedule for your robustness requirements. A fibreglass tube is a good choice to bridge the length of the large-diameter coil.
  10. Assuming that the driven element is a split for dipole feed (e.g. beta match), adjust its length as a reflector. Raise it on the tower, preferably at least λ/2 height. Hook up your antenna analyzer and measure the reactance. Adjust element length when the (inductive, or positive) reactance matches that from step 8. Do the same but for the director dimensions.
  11. Assemble the full antenna, with the reflector and director lengths per your measurements in the previous step.
  12. Raise the antenna to its final position. Adjust the feed system to optimize the SWR. Get on the air and enjoy the product of your labour.
Now you know why the vast majority of hams buy rather than build yagis!

Since the antenna has a narrow bandwidth when gain is maximized I developed two alternatives: one with maximum gain and one with broad bandwidth. They make an interesting contrast. Since beauty is in the eye of the beholder you decide which is better.

Design A: maximum gain

My results, based on ~70% of full length 25 mm diameter elements on 40 meters, are as follows. Coils are 7.2 μH with a Q of 600 and ESR of 0.5 Ω, with centres 3.385 meters from boom centre.
  • Reflector: L = 14.20 meters; X = +26.3 Ω
  • Driven element: L = 13.92 meters; X = -11.5 Ω
  • Director: L = 13.64 meters; X = -49.0 Ω
The actual centre frequency as measured by resonance is 7.140 MHz. That is why the parasite reactances are not of equal magnitude. Both parasites are ±38 Ω relative to the driven element.

An L-network feeds the driven element, but a beta match could be used by shortening the driven element (capacitive reactance). The L-network was selected so that it can be more easily tuned, and another reason as we will discover. The 2:1 SWR bandwidth is 100 kHz, from 7.0 to 7.1 MHz.
  • Shunt, across the input port: 575 pf
  • Series: 0.94 μH
The gain of this antenna is above 7 dbi from 7.0 to 7.2 MHz, reaching a peak of 7.5 dbi near 7.050 MHz. Gain sharply falls above 7.2 MHz and below 7.0 MHz. The narrow SWR bandwidth is due to the low radiation resistance, a consequence of using loaded elements. F/B follows a similar frequency profile. The charted performance appears later in this article.

In comparison to a full size 3-element yagi there is approximately 1.5 db less gain. There are several reasons for the lower gain:
  • Shorter elements contribute about -0.3 db
  • Coil loss varies by frequency between -0.3 and -0.6 db for coils with a Q of 600; loss rapidly rises as the Q declines, so don't skimp on the coils
  • Up to -0.5 db due to the shorter boom length of 12 meters (40') versus 14.7 meters (48')
This amount of gain is about the best we can do for elements of this size and loading. You cannot do better with alternative loading schemes, including linear loading.
Design B: maximum bandwidth

In this design the elements use the same construction and spacing. The only difference is the total element length, which is achieved by changing the length beyond the coils.
  • Reflector: L = 14.30 meters; X = +51.5 Ω
  • Driven element: L = 13.92 meters; X = 0 Ω
  • Director: L = 13.64 meters; X = -92.4 Ω
The reactances are with respect to the 7.14 MHz resonant frequency of the driven element. Notice the much wider and unequal tuning spread of the parasitic elements. The L-network to get the SWR curve below is as follows:
  • Shunt, across the input port: 370 pf
  • Series: 0.64 μH

The 2:1 SWR bandwidth is now a respectable 200 kHz, double that of the maximum gain design. It covers all the spectrum of interest for DXing and most contests. In Europe and other regions this is the entire band.

We pay the price with lower gain, averaging around -1 db. However, the gain remains good all the way to 7.3 MHz. The F/B is also quite good across the full band.

Performance comparison

The chart at right shows the gain and F/B of both designs alongside each other to ease comparison. These are for free space. The F/B will differ by several db over real ground. As for any horizontally-polarized antenna the elevation angle of peak gain and position of lobes and nulls is height dependent.

SWR curves of the yagi are virtually unchanged by the presence of ground when 20 meters or higher, and even a little lower. This is typical for yagis.

Notice that the maximum gain design (A) reverses direction above 7.250 MHz (negative F/B). Design B has usable gain and F/B across the band, although its matched SWR bandwidth is 200 kHz. That 200 kHz can be placed higher in the band for SSB enthusiasts by changing the L-network.

Q of the L-network coil is not critical since the power dissipated is very small. A loosely-wound coil of bare solid copper (approximate Q of 150) is sufficient. It must be bare to allow a movable tap for the purpose of tuning the network. The capacitor should be a high current (low ESR) transmit door knob or equivalent, or an air-core or vacuum variable. For an example see my earlier article on the L-network for my 80 meter tower vertical.

Tuning challenges

The driven element is resonant at 7.140 in both designs. If you use a beta match you will have to shorten the driven element to add capacitive reactance. However I am specifying an L-network (see explanation below). The design process discussed earlier includes a step where the driven element is used to find the actual lengths required for the reflector and director based on reactance. But the impedance of a simple dipole like this varies with height, in both the R and X components.

When the driven element is moved from free space to the real world the resonant frequency (where X=0) depends on height. Notice in the chart how the resonant frequency changes. You must account for this during tuning.

You should also use the coils of the parasitic elements on the driven element to find the correct length of the parasitic elements since if, like me, you cannot build a set of exactly matched coils there will be unit to unit variation. After tuning with those coils you can move them back where they belong.

Be sure to keep the leads from the analyzer to the feed point very short or the reactance of the parasites will deviate from what you measure. If you use a length of transmission line between analyzer and antenna it must be an exact multiple of λ/2, or you should use an analyzer that can automatically measure and compensate for the transmission line length.

The L-network does not have to be redesigned for every adjustment of the yagi's tuning. You can make modest changes by adjusting the shunt and series components. Tap the coil and, at least initially, use a variable capacitor.
  • Adjusting the coil shifts the frequency of minimum SWR, thus shifting the 2:1 SWR points 
  • Adjusting the capacitor changes the SWR near its minimum point; however it does little at the band edges where the SWR creeps toward 2
To reiterate, this is not a general rule. They are only valid for small changes to the network.
More bandwidth

To improve the SWR across a larger swath of the band, in particular for design A, there are a few possible solutions:
  • Remote tuning of the L-network
  • Coupled resonator
  • Dual-driven elements
The first is most convenient with a simple switch to move the low SWR region higher in the band. This is easier with design B since the impedance is less variable in the SSB segment. A couple of SPST can suffice to switch a capacitor and coil tap. This works less well for design B because the impedance varies quickly above 7.150 MHz. I was able to get a 2:1 SWR bandwidth that only covered about 70 to 80 kHz below 7.2 MHz.

The coupled resonator works well with full size elements as we've seen before. It does poorly with design A since the radiation resistance is low. Perhaps if I'd kept trying I'd have had more luck. Even so an L-network would still be required to raise the feed point impedance to 50 Ω. When I placed the couple resonator close to the driven element the impedance was difficult to control. At wider spacing it began to act as a director and fouled the performance no matter how I adjusted the real director. Trial and error has its limits.

Dual-driven elements work well in a similarly sized antenna: the M² 40MDDLL. I tried a few variations of this type. I had no more luck than with the coupled resonator and put it aside, at least for now.
Interaction management

Interaction between antennas is a product of nearness, orientation and resonance. It can be a performance destroyer. Loaded antennas often have an advantage in this regard since the loads typically alter antenna resonance on harmonics. This is important since most of the HF bands are harmonically related. On 40 meters the major concern is with 15 meters, the third harmonic.

Both versions of this antenna do not resonate on 15 meters. The raw SWR profile shows one resonance at around 25 MHz. With the L-network in place there is only a weak resonance near 23 MHz. This makes these 40 meter yagis more friendly to stacking arrangements.

However, since the model does not account for element taper the real antenna will be different. For example, my simple model of the Cushcraft XM240 shows a shallow resonance around 20 MHz, with a potential for some interaction on 15 meters. In practice the actual resonance is in the vicinity of 18 to 19 MHz, as reported by those who've measured it. Therefore until this antenna is built and measured there is reason to be cautious.

The other aspect of interaction is of concern to contesters operating SO2R or multi-op. When two or more receivers are simultaneously used a transmitted signal, or its harmonics, can interfere with reception or cause IMD due to overload. There are several filtering techniques to address the problem. The feed point matching network on the antenna can supplement, but not replace, the filters.

The above chart shows the SWR of design A (design B is almost identical) without a matching network. There is a resonance near 25 MHz, far from 15 meters and close to 12 meters. Now look what happens when the L-network is inserted.

The resonance has moved downward and is greatly attenuated. The L-network helps protect the rig connected to the yagi from overload from transmitters on other bands. Harmonics form the 40 meter rig are suppressed and so will help protect rigs on higher bands.

The difference is that the shunt element in a beta match is an inductor while in the L-network I've specified it's a capacitor. The series element is also the opposite. An L-network of this type is a modest low-pass filter. It can be easier to adjust than a beta match when a low loss, high voltage variable capacitor is used in the shunt. In either case you must use a good common mode choke or you'll lose this benefit.


The coil-loaded designs in this article are viable if imperfect alternatives. Are they worth the trouble? Maybe. There is no simple answer. Certainly it is a fantastic learning experience. But that learning comes at a steep price.

Most hams would sooner choose a commercial design such as the M² 40MDDLL. I don't like the linear loading design because of its mechanical complexity and susceptibility to ice loading (more surface for ice to accumulate on). The price is reasonable for an antenna of its size. It has found favour with many hams.

Some would choose to take the step up to a full size 3-element yagi for its simplicity and better gain, or stick with a small 2-element yagi and its performance limitations. A full size 40 meter yagi is more economical to build than buy and is easier to tune that a yagi with loaded elements. But it is awfully large. Cranes are often used to raise and maintain these monsters.

I have many months to consider my options. This design and comparison to alternatives are food for thought.