In this page, I will describe how to make a class-E amplifier from scratch, that is how to design the circuitry required to produce a desired power at a desired frequency from a desired input voltage on a 50 ohm load (or antenna). This will be done through an example of a 50 W, 12.5 V, 137.777 kHz amplifier based on a single IXYS IXFN N55N50 FET. This FET was given to me by Glen, VK1XX, who challenged me to build a big Class-E transmitter with it. It can probably be used in a transmitter capable of more than 1 kW, however, I would not dare build such a transmitter because of the super high voltage involved - over ten kilovolts RMS in my case, due to a physically small antenna.
The selection of 50 W from 12.5 V is a strategic one; 50 W from a single sealed lead-acid (SLA) battery is a good power source for field operations. Scaling the power in a class-E amplifier is easy without any changes in the component values! When the voltage is doubled (either from a 24 V SMPS or two SLAs in series), the power output should quadruple, making this a 200 W transmitter. A further doubling of the voltage, if the components can take the increased currents, will result in 800 W output. The output impedance does not change with the voltage, so the output transformer does not need to change either.
The following schematic is a typical Class-E amplifier. It is taken from Sokal's QEX article (see references).
In order to calculate the values for the various components, the paper offers a set of equations that the amplifier designer can use.
VK2ZAY (and I!) has done all the hard work and has provided an online calculator (see references) that can provide the component values much faster. The following is the output of the calculator for the 12.5 V, 137.777 kHz, 50 W amplifier:
The saturation voltage is related to the voltage drop across the drain and source of the FET and is linked to its Rds(on) resistance. The FET I used has an Rds(on) of 90 mohm.
Before the soldering iron is switched on, it is a good idea to do a simulation of the amplifier using the free circuit simulator, LT spice. I have used a FET with a similar Rds(on) to the one I have, as I don't have a spice model of the IXYS FET.
In the snapshot above, we see the schematic of the amplifier with the two capacitors made of components that were available at that time. Instead of 176.9 nF there are three 56 nF and an 11 nF in parallel (179 nF) and instead of 222.5 nF there are three 56 nF and a 47 nf (215 nF). The output capacitance of the FET will also need to be accounted for when calculating the capacitor from the drain to the ground, however, in the 137 kHz case, it is insignificant compared to the required value.
The green trace shows the voltage across the load (1.388 ohm resistor) and the red trace shows the drain voltage. Even though the output trance looks like a sine wave, it is not as the following FFT trace of the output voltage reveals.
It is obvious that the second harmonic is well suppressed compared to the fundamental frequency. Despite the low levels of harmonics, a low pass filter (LPF) will be required before this transmitter is connected to a real antenna.
The FET is driven with a 0 V to 9 V, 50% duty cycle square signal. The generation of a square signal to drive the FET can be very easily achieved with a common MOSFET driver such as the TC4420 (6 A peak) or TC4422 (9 A peak) and is not the concern of this write-up. For the record, the TC4420 seems to be enough to drive the 10 nF input capacitance of the FET used on 137 kHz. A crystal oscillator and divider based on a 74HC4060 and a 2.2 MHz crystal was used to drive the TC4420. Instead of that, a DDS, or any other type of oscillator can be used, as long as it can provide the TC4420 with a driving signal of the amplitude required (5 V).
In the simulation above, a load of 1.388 ohm was used instead of a 50 ohm. In reality, it is not necessary to change the output impedance to 50 ohm. It is though convenient. For this, a 1:6 turns (1:36 impedance) transformer will be used, to step it up to 50 ohm.
We now have a complete circuit, so the next step is to gather suitable components.
The two capacitors in the load network are built using a combination of common value capacitors in parallel. The capacitors will need to be able to handle high currents, so the choice is rather limited to silver mica RF capacitors or polypropylene capacitors. The WIMA FKP pulse rated capacitors were found to be suitable. The voltage rating of the WIMA FKP capacitors depends on the frequency, according to the following plots (from the WIMA web site):
The photo below shows an assortment of WIMA FKP capacitors. Note the size of the 47 nF/ 2000VDC one!
Besides the current and voltage rating of the capacitors in the load network, it is very important to select capacitors that have very small temperature coefficient, in other words, their capacitance does not change as they get hot. High voltage ceramic capacitors were initially tested but they failed miserably as their capacitance at a high temperature deviated by more than 50% from the capacitance at room temperature!
There are two inductors in the class-E amplifier and their specifications are slightly different. The inductor between the drain and the power supply is not critical. A 270 uH, low DC resistance, ferrite core inductor was identified in the junk box as a suitable candidate. The inductive reactance at the frequency of operation needs to be high, maybe 50 times the load impedance, otherwise the details are not critical.
The second inductor, L2 in Sokal's article, is part of a tuned circuit and therefore needs to be stable in value regardless of the temperature. A T157-2 core and 2.5 mm^2 wire was used to form this inductor. This has proven to be a reliable implementation in real life. An air core inductor could be used instead.
In this design, a ferrite toroid with a 1:6 turns ratio was used for impedance transformation. This was an unknown ferrite from the junk box. However, an LC impedance matching network can be used instead of the ferrite transformer. The LC network has a disadvantage that it will not be broadband, but that's not a problem on 137 kHz.
The primary (low impedance) winding consists of 5 turns of 2.5 mm^2 bifilar. The doubling of the wire is necessary in order to reduce the ohmic losses due to the high current at that point of the circuit. The secondary consists of 30 turns of 1 mm diameter enamel copper wire. It seems that this is the component that is responsible for most of the loss in this amplifier!
The prototype of the class-E transmitter was built on a double-sided PCB with all track work done with a dremel drill. It doesn't look very pretty, but 136 kHz is a very forgiving band. The result can be seen in the photo below. The FET is on the big heat sink on the left hand side. The space on the upper right corner is reserved for a low pass filter. The output transformer is on the red toroid on the bottom right corner. L1 is the ferrite inductor next to the electrolytic capacitor and is a 270 uH unit salvaged from an old power supply. It has a very low ohmic resistance which is desirable, given the current that flows through it.
The oscilloscope trace below, shows the drain (channel 1) and output on 50 ohm load, after a low pass filter. The Vcc is 18 V and the calculated output power is about 79 W.
The oscillogram below is the same setup with Vcc increased to 24 V. The calculated power output is now about 181 W.
The efficiency measured exceeds 90%.
As usual, the proof of the pudding is in the eating. In this case, eating is reception reports from distant stations. The signals produced by this class-E transmitter have been seen in Tasmania, at a distance of over 800 km. The following capture was produced by Edgar in Moonah, Tasmania. It is a QRSS30 signal and was captured after sunset:
By using WSPR, it was possible to cover the distance even during the day, relying on pure ground wave. The following capture of a WSPR window showing reception of VK1SV and VK1DSH was also produced by Edgar in Moonah, Tasmania:
I am quite confident that if there were listeners in Brisbane or Melbourne, they too would have no trouble receiving my signals. So, to use Lara Bingle's famous words, where the bloody hell are you! :-)