15 Meter Stack Switch

The 15 meter stack switch that I recently completed is almost a duplicate of the one for 20 meters. Both are single band, utilizing an L-network. There are a few differences that I'll describe in this article. For background on the design, design objectives, construction and testing you must read that article.

The L-network is very similar, only the L and C value are different. As before it is a low pass design. For 15 and 10 meters a high pass design may be preferable in a multi-operator or SO2R station, however I took this route for convenience in the choice of components. In any case, adjacent band attenuation is modest (no better than -10 db), so the benefit is not critical.

The major change to the 15 meter version is the layout of the input port and relays surrounding the L-network. There was stray inductance in the 20 meter stack switch that required a significant reduction of the coil inductance. The effectively longer lead length -- wire distance from port to port via the relays -- was responsible. At a higher frequency and lower inductance a better layout is required.

To arrive at the optimum layout I spent an hour with pen and paper to see how different component positions affected the lengths and paths of interconnecting wires. For the layout I settled on the two ports and relays for the yagis are unchanged. The input port was moved to the side of the box with the L-network relays beside it, between the yagi port relays and the barrier connector for the control wires. As before the L-network is situated above the other components, supported by conductive spacers and a terminal strip. This style of layering keeps wires short.

For comparison the 20 meter stack switch is shown on the left. On the right is the complete 15 meter stack and in the centre it is shown without the L-network so that the layout is clearly visible. The L-network connects between the NC (normally closed) pin of the input port SPDT relay and the common pin of the other. 

The (unpowered) default position is BIP (both in phase) so the L-network is in the circuit and connected to both yagi ports. For upper and lower yagi selections the L-network is bypassed and the input port is directly connected to the upper or lower port.

The L-network has a 150 pf shunt on the 50 Ω (input) port and a series 0.19 μH coil to the 25 Ω port. The output port connects the two nominal 50 Ω antennas in parallel. After testing with a thin wire coil I wound the permanent coil from AWG 12 wire. Its inductance is calculated as approximately 0.14 μH, which implies 0.05 μH of stray inductance. The stray inductance is equivalent to ~2" (5 cm) leads, and that visually corresponds well with the layout. The revised layout proved its effectiveness by halving the stray inductance.

Testing was done with my VNWA3 by DG8SAQ. With this instrument I could accurately measure insertion loss, port imbalance and input SWR. The final smoke test with a kilowatt is done only after the insertion loss is determined to be acceptably low at no more than -0.05 db. This is achieved with high Q coils and capacitors to minimize ESR (equivalent series resistance). 

The VNA was recalibrated before measurements were done since I discovered that insertion loss (S21) was at least 0.04 db better than expected. The drift from the original calibration, tiny as it is, is enough to make the stack switch appear to have unity or better gain, and that's impossible. Accuracy is worth the few minutes it takes to re-calibrate.

Let's start with the default BIP position. There was little difficulty adjusting the coil to centre the pass band. The excellent match is indicative of the capacitance value being correct. As for the 20 meter unit one yagi port was connected to the VNA to measure the insertion loss (S21) and the other had a precise 50 Ω load.

There is a small imbalance between the ports. The insertion loss with the yagi ports reversed is -3.08 db. Since the exact value per port should be -3.01 db for a perfect and lossless power division the net loss is very good. As tuned the BIP position is still not too bad on 20 meters despite no attempt to make this a multi-band unit. When selecting either the upper or lower yagi the L-network is bypassed so that the unit can be used on any band.

The insertion loss is effectively nil when one yagi is selected. It shows -0.01 db but that is variable with the measurement trial and port and other times shows as 0 db. If it is -0.01 db that is only 2 watts of loss when transmitting at 1000 watts. As the frequency climbs conductor loss increases due to skin effect. This is evident in the slight upward slope of the blue trace. The stack switch uses a mix of AWG 12 and 18 copper wire, and the relay internal wires are smaller. However, as we'll see later in the "smoke test" none of these parts exhibited measurable heating.

Another frequency effect is the input port SWR. As the frequency climbs the internal wire lengths are a larger fraction of a wavelength and increasingly exhibit transmission line effects. The "dead bug" wiring method I use is beginning to show its limits at 21 MHz. A similar unit for 10 meters would be worse though probably still acceptable. Different techniques are required at VHF and above.

An SWR of 1.07 is quite good and is typically dominated by the yagi impedance which is higher over most of the 15 meter band. To alleviate the problem it is necessary to use coax for the internal connection or PCB with parallel conductors (one at ground potential) to maintain the internal connection at 50 Ω and thereby lower the SWR. This is typical of the construction method in commercial stack switches.

Of course the final step is the smoke test, just as it was for the 20 meter unit. For this test a legal limit signal is passed through the stack switch in BIP mode to determine how much power is dissipated, which is the true indication of the insertion loss. This is done in the shack with both sides of the 2 8 switch connected to the yagi ports and each connected to different 15 meter antennas.

Since the new stacks are not yet connected I use the TH6 and TH7, the same as I did for the 20 meter test. I chose a time when the band was dead so there is no inconvenience to others. I picked a frequency on 15 meters where both have a low and near equal SWR.

Driving the two yagis at full power the coil got warm. Everything else stayed cool to the touch. With one minute at a kilowatt the coil was uncomfortable to touch for more than a second but there was no danger of a burn. Coil Q is not as high as I'd prefer: it is calculated by K6STI's Coil to be ~300. This isn't great but neither is it bad. A higher Q coil would be wider and difficult to fit in the space without causing other proble.

With the smoke test completed I am confident that the 15 and 20 meter stack switches are ready to be installed on the tower. That could happen as soon as a week from now. The weather is improving so there is time to catch up on tower work. 

I am looking forward to having 15 and 20 meter stacks fully operational later this month. All 4 yagis are on the tower, and that is a major step forward. More on that in a future article (or two).