80 Meter Vertical Yagi: Completed

From first models to final design to final antenna took three years. It isn't that I was so slow than it was relatively low priority to putting up towers and other antennas. Construction began almost 2 years ago, then abandoned over the winter, and continued in earnest in 2018. When that was done I had a manually switched 3-element vertical yagi. For me that was an accomplishment, and I enjoyed the fruits of my labour over the winter.

What I didn't do last winter was to complete the array. The missing piece was the switching system. Its components include:

• Switchable L-networks for the yagi and omni-directional modes
• Tuned, parasitic element switching system: off-line; director; reflector
• Switching matrix: diode array to select relays for the selected mode and direction
• Control unit: manual switch in the shack to select mode and direction

All that work is now complete and the antenna is fully operational. Barring any repairs due to construction errors the antenna is complete. It works and it's a lot of fun to use. With this project out of the way I can concentrate on raising yagis, weather permitting.

In this article I'll step through the construction, testing and tuning. After reading this you'll likely wonder why I didn't build a 4-square with a commercial control system. Good question. For now I will only say that I enjoy designing and building antennas -- as should be evident from this blog. This antenna is an interesting and economical way to an 80 meter array with gain. I've made progress since the day it was little more than a giant pile of parts.

Parasitic element switch boxes

When I first tuned the parasitic elements and successfully rigged the antenna as a yagi I gathered sufficient data on the ground and on the air to ensure performance would be in accord with the design. Over the winter I gathered parts and refined the design of the switch boxes. Only then did I proceed to build the switch boxes.

The units are almost perfect clones. That is important to achieve near identical behaviour in all 4 directions. The coils were designed by software and a prototype was built. It was modified until it exhibited the exact required frequency shift -- from 3680 kHz (director) to 3450 kHz (reflector) -- when installed at the monopole base. With one working to my satisfaction I built the other 3 to the same specifications and confirmed that they behaved identically.

Box details are shown in the two above pictures. A few points about the design are worth mentioning:

• Coil Q is not critical since the loss is low. Even at a kilowatt the estimated dissipation is less than 10 watts, and perhaps a little more due to dielectric heating of the PVC form. Insulated solid AWG 14 THHN wire is used. The insulation prevent winding shorts and ensures consistent winding pitch.
• Wire routing and gauge is roughly identical to ensure all conduction path lengths are equal, and therefore frequency shifting is identical. AWG 18 wire is used for RF, which is easy to work with and more than sufficient in these short lengths to handle high power.
• Stainless steel fasteners provide the studs for the monopole wire and radials. A solder lug is placed under the screw head inside the box.
• The relays are sealed SPST-NO 12A in a PCB mount package. They are inexpensive when purchased in quantity. The pins are directly soldered, with the wire providing mechanical support. These relays are perfectly good when placed in a low impedance (hence low voltage) antenna point, just as they are in antenna switching products. When the element is floated the voltage across the contacts of the bottom relay is very low since the element is non-resonant. This was confirmed using EZNEC.
• There is a hole in the bottom for the Cat5 cable and a pair more for the cable tie to hold it in place. These holes double as weep holes for moisture. The coloured wire pairs connect to a small terminal strip. If corrosion is a concern the screws can be coated with dielectric grease. Don't use silicone or other heavy duty gunk or you'll regret it when it comes to maintenance.
• The boxes and coil forms are PVC. They are UV resistant and the box has a rubber gasket. The small 4" × 4" × 2" size is perfect for this application.

The boxes are mounted on a short length of scrap pressure-treated lumber which is turn is clamped to ~1 meter of rebar that is pounded into the ground inside the round PVC pipe that serves as the radial hub. You can study the physical construction in the picture below.

Holes are drilled into the wood to serve as a strain relief for the wire element. This is more reliable than wiring it directly to the box.

A length of wire beyond the connection point allows the element to be lengthened when more radials are added, which will raise the resonant frequency. The wire stub does not affect element resonance or performance if it's short with respect to wavelength; I tested this with up to 40 cm of wire stub with no problem. The end is pushed through another hole in the wood to keep it from coupling with the radials (and lawn mower blades).

Element tuning

While testing an element all the other elements, including the driven element, are floated; that is, disconnected from its radials and, in the case of the driven element, the transmission line.

A 9 volt battery powers the relays. Modern sealed relays with a 12 VDC coil reliably close with as little as 8 volts. The battery is small and portable, ideal for this testing in the field. Check the voltage periodically since it will wear out powering many relays.

Clip lead length does not affect tuning since DC common is isolated from antenna ground. The extra wire needed to insert the antenna analyzer between the radial hub and the box does matter. On 80 meters the frequency shift is ~10 kHz for every 6 cm of wire. The effect of the extra wire is therefore predictable and can be compensated for during tuning. For the best impedance measurements the analyzer should be connected on the radial side of the box.

All the coils were individually tested on one parasitic element to ensure they shifted resonance from 3680 kHz (director, with the coil shorted or out of circuit) to 3450 kHz (reflector). The same coil location and wiring topology in all boxes (see above) ensures identical behaviour of the elements.

With predictable coil performance all I had to do was adjust the monopole wire length to set the resonant frequency. The bottom relay connects the radials to the monopole through the coil to make the element a reflector, and floats the element when not energized. Energizing both relays -- ground the element and short the coil -- the element is a director.

The tarp provides a clean surface for tools, equipment and small parts. It further reduces the risk of ticks which are quite common here from May to July.

Tuning proceeded surprisingly well. All the elements achieved the 230 kHz resonance shift within measurement error. Using the 6 cm per 10 kHz rule mentioned earlier each element took only one or two adjustments. For example, to raise resonance from 3420 to 3450 kHz a knife is used to remove insulation 18 cm up the wire and then reconnected to the box.

The driven element resonated substantially lower than before due to the new longer stinger. Without a matching network the driven element has an SWR below 2 from 3.5 to 3.8 MHz. The resistance ranges from 29 Ω to 34 Ω and the resonant frequency (X = 0 Ω) is 3680 kHz. That this frequency is exactly that of a reflector element is purely coincidental is irrelevant to antenna behaviour.

Switching matrix

Direction is set from a control unit in the shack by placing +12 VDC on one wire and common. The control cable is Cat5 rated for UV and direct burial, in the same trench as the LDF4 Heliax transmission line. The power supply is 13.8 VDC which will be lower at the antenna. Voltage drop is due to the switching diodes, RF chokes and AWG 25 conductor resistance.

In some instances it may be helpful to draw up a voltage budget to ensure that switching systems work as expected. You can often get away with smaller gauge wire -- more economical -- than may be specified for commercial products.

Although I have yet to measure the voltage at the relays they are rated for full closure down to ~8 volts and there is no problem. Testing and tuning with a 9 volt battery in the field worked well, even as the battery aged and sagged down to 8.5 volts. 

The prototype board on which it is built is sitting on the circuit diagram. The 4 wires at the bottom are for direction selection: NE, SW, SE NW. Each side of the board is for a pair of opposite elements, with the 2 wires destined for the parasitic element switch boxes, one for grounding the element (make it active) and the other to short the reflector coil (make it a director). The other 2 elements are floated to be non-resonant and therefore inactive (no induced current).

The small components are 1N4148 switching diodes. The larger ones are RF chokes, one on each line to the shack and one on each line to the parasitic element switch boxes, so that all lines in and out, including the common, are choked; there are more chokes elsewhere.

The chokes reduce the risk of RF on the control lines which could affect the array and control software and EMI due to rectification of RF by the diodes. EMI protection is desirable on the shack end of the control lines although I do not have that as yet since there is no ill effect during use.

The diodes double as back EMF protectors when the relay coil voltage is interrupted. The common lines -- one to the shack and one to each parasitic element -- are also protected by RF chokes but are not on the PCB. The top diodes on the board select the L-network configuration. Antenna impedance is different for yagi and omni-directional modes. More about that in the next section.

The switching matrix is installed in a large PVC box installed at the base of the driven element. Content and wiring of the box is discussed below.

L-networks

Designing L-networks is easy using TLW (comes with the ARRL Antenna Book). All you need is the impedance of the antenna, which is best determined by measurement not the software model. As much as I and others heavily use computer modelling the impedance calculation for verticals and their radials over real ground with its variable composition, the model isn't sufficiently accurate.

Modern antenna analyzers come to the rescue. Use one of suitable quality and accuracy and the job of L-network design and tuning will be easier. My weapon of choice is the RigExpert AA54.

With the parasitic elements tuned and their switch boxes installed and operating the entire array can be configured from the control lines at the base of the driven element. Using a battery and clip leads I manually run through the parasitic element modes. When the array is unpowered it is in omni-directional mode and tuned for the CW end of the band.

Impedance was measured and recorded in 25 kHz steps from 3.5 to 3.65 MHz in all 4 yagi directions and every 50 kHz from 3.5 to 3.8 MHz in omni-directional mode. Recall that the yagi functionality is, at present, only available for the bottom 150 kHz of 80 meters. I wrote all R and X values on paper, which I find is more convenient than pulling the data from the analyzer onto a computer.

The impedance is not the same for all yagi directions despite the care taken in tuning and radial layout. There is a small amount of asymmetry due to construction and (very likely) lack of homogeneity within the mass of soil and rock across the 1 acre of land the antenna encompasses.

The reactance among the 4 directions is very close but the resistance at any one frequency varies a few ohms. That seems small but when the R component of the impedance is 15 Ω a difference of 2 or 3 Ω is proportionately large. It impacts the impedance curve among the 4 directions. In practice the SWR deviations aren't problematic (measurements further below).

With a full set of measurements in hand I played with TLW. My objective was a set of C and L values that would be convenient for switching among modes.

The basic L-network for the yagi modes is as calculated above. The impedance is an average among the 4 directions. The result is a low SWR across the CW segment and acceptable for the digital modes up to 3650 kHz. Yagi performance is degraded but still effective up to 3650 kHz.

As it happens I have a coil already wound, once I trim it down to size. Using K6STI's most recent Coil program my coil with 1" diameter and 6 turns per inch needs to be 1.9" long and will have a Q in line with TLW's estimate for loss. The coil is tapped approximately halfway along for an inductance of 0.75 μH which is needed to improve the SWR in omni-directional mode for SSB.

For the shunt C I use two 1200 pf capacitors in series to give 600 pf for omni-directional mode, one of which is shorted for yagi mode. Each capacitor is a vintage 1000 pf mica transmitting capacitor in parallel with 100 pf high voltage, low RF loss and zero temperature coefficient disk ceramic capacitors. Due to their relative values when in parallel the mica capacitor carries the bulk of the current, for which it is better suited.

The carefully engineered compromise of L-network components results in low SWR in all modes, including omni-directional SSB. Interestingly the SWR in omni-directional mode dropped to well below 2 without an L-network after I rebuilt the stinger. The reason is that the resonant frequency dropped to 3650 kHz due to its longer length. Of course the R value is low enough that impedance transformation remains worthwhile.

During final tuning I found that I could not achieve an SWR below 1.5 for the yagi modes. After making a few measurements and testing alternatives with TLW I added two 100 pf capacitors to the shunt capacitor (the other series capacitor is shorted in the yagi modes). Only one was needed to drop the SWR to 1 at the design frequency of 3550 kHz.

Frequency of minimum SWR differs among the yagi directions due to impedance differences among the elements. Since the SWR is between 1 and 1.2 at 3550 kHz it is inconsequential. The SWR at the edges of the design range -- 3.500 and 3.650 kHz -- is about the same since it is dominated by the large resistance and reactance change. This is typical of antenna matching networks.

The SWR curves are displayed further below. They were measured after the L-network adjustment complete.

A negative consequence of the L-networks is that the antenna no longer works on 30 meters. The topology acts as a low pass filter that has a high SWR above 80 meters. I chose the network topology to reduce inter-station interference during contests. The 80 meter inverted vee through the rig's ATU works well enough on 30 meters to be an interim solution.

Control box

I purchased a 6" × 6" × 4" PVC electrical box to house the switching matrix, L-networks and cable terminations. It sits at the base of the driven element (tower). External connections are for the coax, driven element RF connections, control cable to the shack and control cables to the parasitic elements.

The 5 Cat5 (8 conductor) cables pierce the wall of the box and are routed to the barrier style connector strips. On balance I deemed this approach better than connectors with respect to waterproofing, convenience, cost and labour. It requires some care in the layout to make it work well. The picture is of the completed box before being installed and the cables attached.

Yes, it does look messy! Despite that it works quite well. The relays are very lightweight and can be supported (suspended) by the solid connecting wires for RF. Control wires are separated from the RF section (bottom and lower left) to minimize interaction even though coupling with short wires at 3.5 MHz isn't a serious problem.

The control wires are made as long as necessary and RF wires are kept as short as possible. The photograph makes the depth look shorter than it is; there is more vertical separation than is apparent.

Layout detail:

• Switch matrix is on the right wall. One screw with an insulated spacer supports it.
• Connector strip for the control cable is on the bottom. Every terminal is labelled. The blank line carries +13.8 VDC to allow testing on site without need for a battery.
• Labelled connector strips for the control lines to the 4 parasitic elements are at top centre. Chokes for the common lines and reverse EMF protective diodes for the 3 internal relays are between the strips.
• Holes for the control cables (5 of them) are at the lower right. There are barely visible due to the camera angle. There are additional  holes for tie wraps to hold the cables in place.
• There are 3 stainless studs at the bottom. From left to right they are for the driven element monopole, radials and 160 meters. Again, these are not easily seen due to the camera angle.
• Coax connector (N) is at the lower centre of the left wall.
• L-network series coil with taps for yagi and omni-directional modes. The relay just to its bottom right shorts the lower coil section for omni-directional SSB use.
• The relay above it shorts one of the series L-network capacitors for yagi use. The capacitors are large transmitting mica with a few 100 pf 1 kV parallel high-Q ceramic capacitors. These are partially hidden by the coil and relay.
• The DPDT relay at centre bottom directs output of the L-network to either the driven element or the 160 meter stud. The 160 meter option will be described in a future article after it is added to the antenna.

High value carbon composition resistors will be added later to drain precipitation static to the radials and ground from the driven element. It is a minor concern in our cold winter so it can wait for spring. The lightning protection system will be added at the same time.

The picture shows the completed box attached to the driven element and with all cables connected. Despite being even messier with the cables attached it was straight-forward to wire and test. The bracket that holds the box to the tower needs to be replaced with a piece of lumber for improved mechanical support.

Colour coding is standardized with colours written on labels inside the box and documented in my files. Make sure you do this since you won't remember (trust me on this). There are 24 control lines: 8 to the shack and 4 to each parasitic element. After considering options I opted to install short cables to the internal terminal strips, which could be down indoors in comfort. Crimps on the outside of the box to attach them to the 5 cables.

It took a couple of hours in the field to get it all connected, tested and temporarily weatherproofed. For maintenance that requires box removal it is inexpensive and quick to simply cut the lines and reattach them later. I will use connectors for the external connections if maintenance and antenna improvements occur more often than currently anticipated.

Impedance tuning of the antenna

The antenna has 7 operating modes: 2 omni-directional modes optimized for CW and SSB band segments; 4 yagi directions; and 160 meters. The last is not yet implemented. That leaves 6 modes to be matched to the 50 Ω transmission line.

With the L-network installed the initial tuning could proceed. First to be done were the CW and SSB omni-directional modes. These did not require manipulating the control lines to the parasitic element, which were left disconnected and not yet routed into the control box. As before a 9 volt battery was used. The taps for the full coil (CW) and SSB were adjusted until the SWR curves were optimized.

The coil is at least 50% longer than needed. It was pulled out of my junk box and I didn't bother shortening it to give me more tuning room should I ever need it. The coil values calculated with TLW are approximately 1.5 μH for CW and 0.75 μH for SSB. Based on the coil dimensions the taps ended up not far from these values.

The shunt capacitor should be approximately 725 pf for CW and 500 pf for SSB. Instead I fixed it at about 600 pf for both band segments; that is, with the two 1200 pf capacitors in series. One of them is shorted for yagi modes since its shunt capacitor is calculated to be in the range 1150 pf to 1300 pf. The range is due to the impedance differences among the 4 directions as previously noted.

The component choice allows for a minimum of alterations (by relay) among modes. The design was intentional in this regard to keep the switching as simple as feasible. This is an interesting topic on its own which I may cover in a future article.

The SWR curves for the omni-directional modes are very good. They cannot both be perfect since it is only the value of the series inductor that changes, not the shunt capacitor. Adjusting the capacitor value would help little since although it would bring the minimum SWR to 1 the tails of the SWR curve would be about the same. This is typical since as you move away from the centre frequency the R and X departures from the ideal dominate the impedance.

With that done I connected the battery to the parasitic element control lines to adjust the L-network for the yagi modes. There is variation among the 4 directions due to the aforementioned parasitic element impedance differences. Rather than show all of them here is just the one for northeast. The others are similar or better.

Per the design the SWR soars at the 3650 kHz upper end of the design range. This is unimportant since the gain and directivity decline rapidly above 3625 kHz.

With the L-network adjustment complete the box was brought indoors to solder the coil taps and install the parasitic element cable harness (see below). After reinstalling the box there was a problem. A clip lead to the radial hub was used during L-network adjustment. When I replaced this with a permanent wire the SWR for all modes increased since it was shorter. It was far earsier to increase the wire length than adjust the networks. Wire length matters.

The SWR plots were done in the shack not at the antenna. Out in the field I forgot to save the analyzer plots and I didn't want to brave the cold and redo the weatherproofing just for this article. The SWR curves are more nominal at the other end of the 300' transmission line. The difference is probably due to a deviation from 50 Ω in a section of old RG213. Attenuation is very low at 3.5 MHz and thus contributes less to the impedance difference at the shack end of the transmission line.

Component choices

There are not many types of components in the switching system and networks. Judicious choices are necessary to ensure proper operation and reliability. Some of what I needed comes from my extensive junk box. Other components were purchased. The price is low when you buy in bulk from the major electronic component outlets. Most of these were purchased from Mouser and there are many other sources with good reputations and prices that also make it easy to do business with them in Canada.

• The 1N4148 switching diode is used throughout the switching matrix. They are cheap, small and effective. They double as back EMF protection for the relays they power. While not ideal in this latter application I was willing to take the easy route. Others use them in similar antenna systems with good results.
• For relays not powered by the switching matrix I used 1N4007 rectifier diodes for back EMF protection. They are more robust than the 1N4148 and indeed are overkill with their 1 kV reverse voltage rating. I had them on hand so I used them.
• Old 1000 pf mica transmitting capacitors in combination with parallel 100 pf disc ceramic capacitors make up the two 1200 pf series capacitors in the L-network. These are stable and can handle a kilowatt of power. The 100 pf capacitors are 1 kV, low temperature coefficient and the ceramic material is low loss at RF. They each carry about 10% of the antenna current since they are in parallel with 1000 pf mica capacitors.
• The RF chokes (lower right) are rated for 600 ma and are resonant at 6 MHz. The current capacity is well above the draw of the one to three relay coils they each support. The series resistance of ~1.5 Ω has negligible voltage drop in my control system. It is important that the self-resonance of the chokes be outside of amateur bands to ensure they perform properly.
• The TE Systems 12A SPST-NO relays do the bulk of the switching chores in the L-network and parasitic element switch boxes. Both sides of the Omron G2RL DPDT relay switch between 80 and 160 meters at the output of the L-network to double the contact rating. All the relays are small, sealed and reliable up to a kilowatt. This type of relay should only be used for low impedance switching since with high impedance the RF voltage is too high. For those applications use vacuum relays or relays with a large contact gap.
• PVC pipe as a coil form is acceptable at 80 meters but not for QRO above 20 meters or so. Loss (heat) in most PVC formulations rises with frequency and becomes a problem. I use PVC forms and insulated (THHN) wire in the parasitic element coils to achieve dimensional uniformity. The L-network has an air core bare wire coil from my junk box. Bare wire is handy for tapping the coil to adjust the impedance match.
• Internal RF wiring in the 5 boxes is AWG 14 and 18. Smaller 18 gauge wire is acceptable despite the high RF currents since the lengths are short and ohmic loss is negligible. The dead bug construction method favours small gauge wire connections to the relays.

Control unit

The cutouts for the 80 meter array were included in the manual antenna control unit when it was built. A 6-position rotary switch and toggle switch do all that is required in the present antenna configuration. The only design question is which direction and mode to assign to each position.

When unpowered the antenna is in omni-directional mode with SWR optimized for CW (see L-network discussion above). For switching convenience I placed the CW omni-directional mode between the 160 and yagi position. Since the most common yagi directions are northeast (Europe) and southwest (bulk of the continental US) they are the adjacent positions. The less common directions are in the furthest clockwise positions: southeast (southeast US, Caribbean and South America) and northwest (Japan, east Asia).

During CW contests antenna mode and direction requires one one click most of the time. This is usually between northeast and southwest, with occasional forays to southeast, and rarely to northwest except grey line openings to Asia.

For SSB contests there are only two choices, both omni-directional modes with SWR optimized over different band segments. Most often the SSB switch is the only one needed; set once and forget unless you want to "rest" the L-network relay when 80 meters isn't in use.

As currently configured the SSB switch position disables the rotary switch. This works since the only SSB mode is omni-directional. Should I add SSB to the yagi modes the wiring must change so that the mode selection would select (short) coils in the 4 parasitic elements to move the centre frequency from 3550 kHz to 3700 kHz or thereabouts.

A wiring error among all the cable harnesses reversed the SE and NW directions. I'll have to track this down. I could reverse the wires in the control unit except that would violate my Cat5 colour coding and invite future confusion. There is no rush and there is no great consequence during use. Apart from this one error the control unit works perfectly.

Grounding

With the driven element, parasites, radials and control system isolated from physical ground there are a few challenges implementing lightning protection. Although this region has a lower risk of lightning strikes than many other protection is still required. After all, an 80 meter full size vertical sitting in the middle of a hay field can be a very attractive place for a lightning channel to form.

There are two problems to be addressed: lightning strikes and precipitation static. For a direct or secondary strike I want an easy path to ground for the lightning current despite the antenna's isolation from ground in normal operation. The electronics are unlikely to survive and are easy to replace, but I do want little of the strike current reaching the shack ~100 meters distant. Precipitation static needs to be continuously bled to ground to avoid excess receiver noise when it rains or snows.

For both problems ground rods are needed. To drain static it is enough to place a high value resistor between the antenna elements (including the radials) and ground rods. The typical method to deal with direct and secondary lightning strikes is with two copper balls separated by a spark gap, one on the driven element and one on the ground rod.

I hope to implement both solutions next spring before summer storms arrive.

Performance

Does it work? The short answer is yes. There is close agreement between what the computer model predict and how it performs on the air.

This is not a "perfect" antenna and it was not designed to be. Its performance pros and cons are quite interesting. This article is already long enough and I would like more experience using it before writing it up for the blog. Perhaps in December, after CQ WW CW and when the weather forces an end to tower and antenna projects this year and I spend more time indoors.

Summing up

For someone who enjoys playing with antennas this has been a fascinating project. I've learned a great deal and I've gained an effective directional antenna on 80 meters. I will continue to monitor its performance over the winter in contests and daily DX chasing. The future of the antenna will then be decided.

I definitely plan to add 160 meters to it, per the design and construction. It won't be difficult and will be an interesting project. I may in time want to add SSB to the yagi modes, something that isn't possible now due to the narrow bandwidth (less than 150 kHz) of the yagi. It will require changes to the parasitic elements and control architecture; the control lines are already installed.

What is particularly interesting is how the yagi's performance compares to the popular 4-square that is used in many big gun stations. The major differences I knew before starting this antenna project. Now I'd like to make the comparison quantitative, for my benefit and for readers. It will be enlightening.