This is the final stack switch for my 5/5 stacks on 20, 15 and 10 meters. The 10 meter yagis were built and raised last year, but without the stack switch only the upper yagi was connected. Once I make and install the phasing harness and control cable the system will be complete. Of the three stacks, this one was the lowest priority because of the solar flux. Now that cycle 25 is blossoming, I want the stack fully operational for the fall contests.
The design of the stack switch is almost identical to that for the 20 meter and 15 meter stacks. In this article I will focus on the differences. Please refer to the earlier articles, and this one about L-network stack switches. In addition to construction and testing details, you will find the theory of operation, component selection and a circuit diagram.
The major differences from the other stack switches include:
- More compact L-network to reduce stray inductance
- Compensation for the internal wiring when using the upper or lower yagi
- Components are mounted on the enclosure cover
The "dead bug" construction method used is not ideal for preserving a constant impedance throughout the circuit. A better method is to mount them on a circuit board and place the traces to approximate a transmission line between major components. It can never be perfect since wires travel to and from the board and there are uncompensated paths through the relays.
Impedance deviations are small on the 20 meter stack switch, and imperfect but still acceptable on the 15 meter stack switch. I knew I would not be so lucky on 10 meters. The problem of stray reactance in inverse proportion to the wavelength; the reactance of an inductor is proportional to the frequency. Wire is an inductor, whether in a coil, as component leads or for circuit connections.
Before diving into those challenges, it makes sense to start with the final item: mounting components on the enclosure cover. My motivation to try this method of construction is due to Steve VE6WZ. He has several videos on his YouTube channel describing its advantages. It does indeed make working on a circuit easier. I'll definitely consider it for future projects.
Referring to the earlier articles you'll find that I'm using an identical set of components. They fit neatly on the aluminum cover. All I had to do was position the top connectors around the protrusions for the enclosure's screw holes. The aluminum enclosure is from the Hammond 1590 product line.
I'm using the same layout as for the 15 meter stack switch since it has less stray inductance than the layout used in the 20 meter unit. As we'll see, the wiring is be the same but mounting of the L-network will be quite differentl The terminal strip I used in the other stack switches is not needed. I hadn't yet made that decision when the picture was taken, which is why it's there.
The 100 pf ceramic capacitor I purchased for a few dollars in the Hamvention flea market last month is suitable for a kilowatt in this application. It's worth mentioning that there are no extreme voltages and currents in a L-network with a low transformation ratio (2:1) and a modest SWR at the edges of the yagis' frequency range.
The picture at right shows the stack switch fully wired. Unlike the ones for the lower bands, the L-network is built over its relays for the shortest possible lead lengths. The lead lengths to and through the relays to the yagi ports at the top are about the same as the other units. The compact layout is made possible by the small capacitor and coil.
TLW was used to design the network. While this network works well as a single band stack switch I must empathize that it is not a narrow band device. That is a common belief going by the feedback I've received from readers and others. A (transmission line) broadband transformer certainly has its uses, such as for tri-band yagi stacks, but it is larger, more difficult to build and in most situations it will be less efficient. The L-network single-band design works well and it is easy to design and build.
I did not adjust the default coil Q in the TLW design. The actual coil Q is calculated at over 300 using K6STI's Coil app. Power dissipation is lower than the calculated value in inverse proportion to the Q. For example, the loss would be 6 watts for a Q of 300. The coil is wound from 12 AWG copper. I tried 10 gauge wire but it was so stiff that I was concerned about damaging the circuit during mounting and network adjustment. I made a temporary coil wound with thin wire for the initial bench test since it was easy to work with.
Stranded wire connects the two ends of the coil to the relay pins to relieve mechanical stress on the relays during installation and adjustment. The standoff and aluminum angle bracket on the capacitor's cold side have sufficient mechanical strength to rigidly support both the coil and the capacitor.
The VNA plot of the completed stack switch in BIP mode (both in phase) shows a remarkably good match and low insertion loss. The insertion loss is not perfectly accurate because the impedances to the yagi ports are not exactly equal and the VNA was not recalibrated for this measurement. When I reverse the loads on the yagi ports the measurement is lower. We want the average insertion loss through the two yagis ports as close to -3.01 db as possible to minimize heat dissipation. With properly selected capacitors almost all the loss is from the coil.
There is one extra component in the stack switch that you may have noticed: a capacitor at the top of the unit. This brings us to a discussion about impedance compensation.
Due to the dead bug design there is an impedance"bump" between the input port and the yagi ports . The impedance through the switch with a calibrated 50 Ω load on the upper or lower yagi port is 52 + 6j ± 1j Ω. On a PCB with traces positioned to form a transmission, line the series inductance of the conductors is cancelled by the parallel capacitance between the conductors, leaving a reactance-free characteristic impedance. Even with a well-designed PCB there may still be a need for impedance compensation.
The impedance bump causes an SWR of about 1.15. That is no disaster but it is worthwhile correcting if not too difficult. As we'll see, the SWR can be corrected quite easily.
It is sensible to assume that the unwanted inductive reactance can be cancelled with a shunt capacitor on the input port of the stack switch. Designing a low-pass tuner with TLW confirms that this should indeed work. There series inductor calculated for the L-network can be ignored since it very small and dominated by the circuit's stray inductance. The value of the compensation capacitor is calculated to be in the range of 22 to 25 pf, depending on the yagi port.
The input port is a poor location for the capacitor because it would be in parallel with 100 pf capacitor of the L-network used for BIP mode. That wouldn't be so bad if I had a suitable L-network capacitor somewhat less than 100 pf, but I don't. The capacitor is instead mounted at the far side of the L-network output relay on the NO (normally open) contact that connects to the relays for each yagi port.
The picture also shows why there is a small difference in the inductive reactance to each yagi port. To minimize the total amount of interconnecting wire, the length of wire to the upper yagi port (penned with a "U") is longer. The small difference in reactance did not justify rewiring the relays, which is a delicate job. Of greater consequence was the capacitor's location and its value.
My first choice was a 27 pf ceramic transmitting capacitor. It turned out to have twice the required reactance to compensate for the inductive reactance because it is not at the input port. The location of the capacitor in effect creates a π-network, with two series inductors: one from the input port, through the L-network relay to the capacitor; and the other from the capacitor, through the yagi port relays, to the yagi ports.
Interpolating from the measurements, I needed a capacitor of about half the value. More digging in my junk box turned up a 15 pf ceramic transmitting capacitor. Above is the VNA plot of the through impedance and insertion loss for one yagi port. There is no need to show the nearly identical plot for the other yagi port. The compensation is very effective.
At this point I had a fully functional stack switch for 10 meters. The final step was the smoke test. Since I do not have a suitable dummy load I used the 2 × 8 antenna switch to select the upper 10 meter yagi and the TH6, and connect the coax for each operating position to the stack switch. RF power is supplied by an Acom A1500. To avoid annoying anyone, testing is done when 10 meters is mostly dead, and on a frequency well away from the main activity areas.
A power of about 1000 watts was applied for a 30-seconds during each test. The results exceeded my expectations. The L-network capacitor (BIP) and the compensation capacitor (upper, lower) remained cool to the touch. The L-network coil (BIP) was warm but not at all hot. The coil in the 15 meter stack switch got hotter during its smoke test.
To complete the project I need to design, make and install phasing harnesses for the 10 meter stack. The stack switch will be installed ~140 feet (40 m) up the tower and the control lines spliced into an already present run of Cat5 cable. More on that later this summer when I'm ready to proceed.