While performing some off-air tests with the modified MF Solutions converter it became apparent that any variations in a load impedance were going to result in potential variations in the on-board output filter network. This would likely cause matching issues with the FET and you can probably imagine how it could all spiral downhill from there. It’s very common on 630-meters for the match to vary during the overnight session a slight amount as environmental factors vary due to temperature and moisture changes. I am certain that this behavior will be exponentially worse on 2200-meters. I experimented with a 3db pad that had a 50 ohm input and output, which seemed to cure the instabilities but since I am already running lean and mean at 10-watts TPO on 2200-meters, I was not terribly excited about losing 3db. Instead I decided an outboard low pass filter would make a good buffer and provide some additional filtering and stability.
I chose the W1VD 137 low pass filter which is built like a tank. The 630-meter version that I have has worked flawlessly. But I don’t plan on operating QRO on 2200-meters any time soon and those large T225A-2 ferrite’s are quite expensive. I settled instead for a scaled down version using what I had on-hand. If you are judgemental about engineering, you should probably stop reading now and move on.
I opted for T106-2 cores which meant really small wire. I wouldn’t normally use #26 wire on a power project like this but its what would fit the core on-hand. John Molnar, WA3ETD / WG2XKA, emailed saying that the inductor values, transposed from the T225A-2 core size, were 58 uH for L1 and L3 and 143 uH for L2. Now, L1 and L3 are manageable but that is a LOT of turns on L2! In fact, the number of turns work out to 66, 103 and 66, respectively (NOTE 01/04/2018: Rob, K3RWR, indicates that the value of L2 is incorrect. He reported that his calculations and testing suggest that L2 should be 114 uH instead of 143 uH. I have not verified this value here but keep this change in mind and adjust your turns accordingly. Its always easier to wind the larger inductor and then remove turns than splice in additional wire if you need additional turns)
Winding went OK aside from the long lengths of wire I was pulling though the tiny cores and the cat wanting to play with the wires as the whipped around. Installation was on a scrap piece of copper clad board and was placed into a box that originally had a battery charger that was purchased on a junk table at Hamcom for $1. I like deals like that and more and more I find myself building stuff with scraps of junk. It’s very gratifying.
So how did it play? Over a wide variety of loads on the bench, the waveform was very clean and stable. No oscillations were observed. It appears that I was successful at buffering the PA and providing a little more output filtering.
Why is this stability issue so important? In order for the filters to do their job properly, they need to see the match that they were designed for. If they were built for a 50 ohm output and I present 10 ohms at the load, the filter might not be able to do its jobs of properly filtering the harmonics that are present on the signal. The load will vary through a long beaconing session, for example, and possibly even during the course of a CW QSO depending on component heating at the antenna.
I’m excited to see how this system does on the air. I have made some additional adjustments to the matching network, finding greater stability in the match from day to day when the capacitors from the 630-meter network are jumpered out of the network, leaving only a bit of shunt inductance. It seems good enough to find 50 + j0 ohms. Its my belief that with current environmental conditions, my losses are probably only about 30 ohms. It could be a LOT worse.