The short answer is efficiency. A true linear amplifier is limited in its efficiency. A realistic number is 60%, meaning if your output is 300 watts, you’ve got 200 watts of heat to get rid of. A class F amplifier is NOT a linear amplifier. The transistor is driven with closer to a square wave than a sine wave, so it acts as a switch, not an amplifier. This means there is no physical limit to its efficiency, but a realistic number might be 90%. So for that same 300 watt output, there is now only 33 watts of waste heat. The amplifier is now much smaller, and when portable, battery life is significantly improved.
In the class F family of architectures, the base or gate of the amplifying transistor is driven with close to a square wave. But doesn’t that create harmonics, you ask? Yes, but you reflect them, conserving all of that energy. Rather than harmonics distorting your output, they distort the voltage and current waveforms on the switching transistor. So a transistor with a much higher voltage and current rating must be used. Fortunately, with GaN technology being used in switching power supplies, they are becoming easier and easier to find!
The GaN FET transistor I selected can withstand pulses of 120 amps and 750 volts, all in a 9 mm x 8 mm x 0.5 mm package! The biggest issue is heat dissipation. Here is the data sheet if you’re interested: https://gansystems.com/wp-content/uploads/2021/10/GS66516T-DS-Rev-210727.pdf
A good primer on Class F amplifiers is here: https://web.ece.ucsb.edu/Faculty/rodwell/Classes/ece218c/notes/Lecture9_SwitchingPAs.pdf
A class F amplifier is generally a fixed output amplitude, great for constant amplitude modulations like FM. Since it’s a square wave driving the gate, it can’t natively be amplitude modulated. But, the output voltage swing *is* directly proportional to the drain voltage. So there is a way to amplitude modulate! Instead of modulating the input, you ultimately modulate the power supply instead. This is the idea behind a Class H amplifier.
In my amplifier this power supply modulation is accomplished with an envelope detector circuit on the input, fed to an op amp to control the power supply. This is the most complicated part of the design. That power supply has to be able to swing pretty fast, or the bandwidth of your output will be limited and distorted. Fortunately with SSB and all digital and CW modes used in amateur radio, amplitude modulation never exceeds 3 kHz, so a target of being able to ramp the power supply up *or down* at 1 volt per microsecond is sufficient. Meaning the power supply must be able to source significant power, but also to sink power efficiently, without creating significant waste heat.
This is not as easy as it sounds. Power supplies designed to source a kilowatt are generally meant to output a fixed voltage. But find the right part, and it can be tricked into doing your bidding. In this case, I chose the LTC7891, and its output can be modulated from about 60 volts, down to about 1.5 volts, with a maximum of 20 amps output. It also requires 2 external GAN FETs to operate, bringing the total number of GAN FETs in the design to three. Fortunately, the power supply FETs are a bit cheaper!
Ultimately, for small signals, there *should* be a linear response. so I added the ability, for small signals, to transition from class F into class A or AB. I use Schottky diodes to square up large signals into the gate at around 6 volts peak-to-peak for maximum efficiency. But there is also a small DC bias on the gate, gently into conduction. This allows the amplifier for very small signals, to operate in a more traditional linear fashion.
Doesn’t that crossover create nonlinearities? Yes, the transition isn’t perfect. And I made the gate bias adjustable so it can be minimized. However, there is no compression (until hard clipping) at the top end, and the signal is already down over 40 dB before the transition from class F to class AB occurs. Meaning it has much less distortion where it counts the most, at the peaks of the envelope, making it more spectrally pure than a typical linear amp.
To be continued…
Leave a Reply