Or is it don’t reinvent the wheel?
Well, actually it is ‘broke’…
I have been troubleshooting a low output power problem on my QCX transceiver this week. I was fairly confident that I had the problem identified as a bad BS170 MOSFET, one of three that QRP-Labs uses in its Class E power amplifier to get 5 watts out. Anyway, it turns out I was wrong so I will look deeper into that problem later. For now, I can live with a watt and a half for 40M CW.
I’ve always been a fan of these MOSFET amplifiers, especially given I enjoy CW so much, as they make for very efficient use of battery power. The downside to Class E amplifiers, unfortunately, is they are both easier and at the same time, more difficult to design. They are, essentially, just high speed switches. Indeed, it might be better to call them switches that are being operated at very high speeds. What they are actually are is neither RF devices nor linear, but they do at times mimic such operations, within limits.
I build a lot of QRP stuff for fun, so I wanted to delve into a MOSFET PA project. My goal is to use a DDS (si5351) as the exciter and link it to a MOSFET amplifier for the power. For many of MOSFET amplifiers, getting them to operate at an output impedance of 50 ohms is an exercise in mathematics. In truth, it isn’t really that difficult, but it’s not for the uninspired.
Whilst digging into the schematic for the QCX, I got to wondering how QRP-Labs designed the finals amplifiers in some of their other offerings. Specifically the U3S WSPR transmitter. It actually does more than WSPR, but that seems to be its primary application.
Unlike the QCX, which uses a TTL logic device to buffer the DDS output into something that will directly drive the gate of the BS170(s) without biasing said gate, the U3S PA does bias the the BS170 gate. I found this both interesting and appealing.
However, whereas the QCX is basically a CW rig, the U3S supports multiple protocols. Given that, I could see why they might need to bias the gate, providing the ability to set the MOSFET in a semi-on state, permitting a linear like operation, further permitting a lower-level/not quite a square-wave input signal. Basically, mimicking (and I use that term loosely) something akin to Class A/AB operation in an analog device.
Reading through the manual it became clear to me that this particular amplifier could easily suit my desire for a very simple QRP transmitter, with possibilities beyond CW.
Figure 1. My ultra ugly, one hour, slapped together MOSFET ‘test amplifier’.
Figure 2. Snippet from the QRP-Labs U3S schematic. Remove two of the BS170 devices and that is what you see in Figure1.
I constructed the amplifier with a single BS170, as described in the manual, and powered it off a 5V supply along with the gate bias, the latter via an adjustable level potentiometer. Indeed, I followed the U3S amplifier description to the letter in order to set a benchmark.
The amp was connected to a si5351/Arduino test setup I use frequently, and each clock output set to a discrete frequency from one of the amateur HF bands and, the amplifier terminated with a 50 ohm load (2x 100 ohm in parallel). In my si5351 breadboard, I can quickly switch clock outputs to see the amplifier’s performance on different frequencies,
As with many MOSFET amplifiers, I started with the bias turned off, because doing it the other way around usually results in a loud pop. Flipping on the 5V supply, I monitored the output on an oscilloscope. With the bias off, there was no visible output. The si5351 alone provides around 3V peak to peak (~22.5 milliwatts), which typically isn’t enough to really turn on even a lowly BS170. Here again is a likely contributor to why the QCX buffers the si5351 clock with a TTL driver, and why this design (and many others) partially turns on a MOSFET with some bias on the gate. As I SLOWLY increase the bias, a perfect replica of the DDS output starts to appear and increase in level. I continued to increase the bias until the amplitude of the signal peaks, at which point I then backed off the setting by just a hair.
A quick count of the scope divisions and the final number was 302 milliwatts. I have to say that I was impressed. Indeed, the U3S manual states the expectation should be ~250 milliwatts, so not bad for a junk box version of that amplifier.
BTW, the math here is peak to peak voltage times 0.3535, then square that number and divide by the load resistor (i.e. 50 ohms).
I measured 11 volts peak to peak, so…
11 times 0.3535 = 3.8885
The square of 3.8885 = 15.12043225
15.12043225 divided by 50 (ohm) = 0.302408645 Watts
Back to the amplifier circuit… keep in mind, this measurement was made with only 5Vdc feeding the drain of a single BS170. The U3S PCB, like the QCX, has provisions to stack/parallel a total of three BS170 devices (you can do that with MOSFETS), to increase the output power by spreading the load. If you want to know the deep details on how that works, read the numerous Class E amplifier design guides.
The QCX is rated at 5W output using three BS170 devices fed with 12Vdc on the drains and the gates driven hard with a 5V Peak to Peak TTL device (74ACT00) fed by/buffering a clock output from the si5351.
Given the differences in the two amplifier designs, I don’t expect to see 5W out of this one. However, my feeling is that adding/stacking two more BS170 devices and, increasing the drain voltage to 12 Vdc will hopefully provide an easy three watts output. That of course, will be tempered by the, absolutely required, 7-element Chebyshev output filter.
One of the biggest obstacles of using a DDS and a Class E amplifier, IMHO, is harmonic content! However, most of the time it can be brought into compliance with a decent low pass filter. At least that’s the thinking.
Only further real world testing will determine that, but I am optimistic.