80m AM Class E 400W Transmitter
Eric Edwards GW8LJJ presents an interesting, albeit advanced project for a class E transmitter.
Eric Edwards GW8LJJ presents an interesting, albeit advanced project for a class E transmitter. Even if you don’t plan to build it, there is a lot of good technical information here.
This is a fully explained project for the serious constructor and although almost a full kit of parts is available it is not a ‘place the components in the board and switch on’ type of project. It is designed for the more ambitious constructor wanting a low-cost solution to an AM class E transmitter. It is designed as a very stable single-frequency transmitter for the 80m band and although the on-board oscillator has been programmed for the popular 3.615MHz part of the band, there are two other outputs on the oscillator board that can have any other frequencies added. If all the major parts are obtained from me as a kit, the oscillator will be programmed for 3.615MHz as a single output. Others can easily be programmed by the user or I can, on request, program the other two for any other frequency. One, for example, could be for topband (160m) and the other for 60m or another part of the 80m band. If other bands are to be used, the output transformer/capacitor has to be changed to suit.
Class E PA
The ‘class’ of a power amplifier (PA) refers to the type of emissions and is related to the bias arrangement. Class A is the most linear but most inefficient because there is always current flowing through the device when there is voltage across the drain/source even with no input signal applied. The amplifier is biased so that the voltage at the drain of an FET, collector of a bipolar transistor or anode of a thermionic valve is at half the supply to it. This allows the input signal to reach its full swing from top to bottom so with a sinewave input, the output is a perfect replica but larger sinewave output. Class B is biased so that top of the waveform is reproduced and another stage in a push pull arrangement takes care of the bottom half of the waveform. There are variations on this with class AB, AB1, AB2 and so on and class C, which relates to the amount or lack of biasing used. This project uses class E, which is similar to class D and F and it is the output tuning that decides the actual class and which differentiates it from class C and class F.
To get the most efficient system, the devices we are using in this design are FETS and the FET must be turned fully on (saturated) and fully off. This is achieved with a square-wave applied to the gate. Turning an FET hard on produces a near maximum current through the device with no voltage across it (from drain to source). When the output stage is correctly tuned there is minimum time overlap between the voltage at the drain and the current flowing through it. There must, of course, be voltage to allow current to flow but the ‘flywheel’ action of the PA tuning produces this. When the FET changes state from fully on to fully off, the voltage (source/drain) is almost zero so no power (or very little) is lost as heat as the FET switches. The PA is tuned to achieve this required phase relationship between voltage and current. The coil, which is a reactive element, and the drain capacitance along with the added capacitors create this flywheel effect by charging and discharging of the coil/capacitor network. You can think of this as the capacitance being used to time the charging and discharging of the coil along with a similar action for the capacitor or, in this case the output transformer on each part of the push-pull stage. The capacitance is varied to achieve the correct timing and rounds off the square wave to produce sinewaves ready for putting to the antenna (via a lowpass filter). It is this output tuning network that makes it a class E PA.
The oscillator is fixed frequency at the frequency used for transmitting. Usually, a class E transmitter has an oscillator or VFO at twice the transmitting frequency and is divided to produce two equal and opposite square waves to connect to the FET drivers because the output FETS are in push-pull. The oscillator I am using is connected to a high-speed comparator with complementary outputs for the two FET drivers so there is no need for twice the frequency and a divider.
The oscillator is a small module (Si5351A) that is programmed and controlled by an Arduino Nano. Once programmed from the free open source software on the PC, the USB lead can be removed and the transmitter 12V (13.8V) can used. The frequency is very stable as would be expected with a programmed device. The use of an Arduino and Si5351A was an idea by Ray G7BHQ and he wrote the simple ‘sketch’ in the Arduino IDE, which is a free open source download.
As mentioned, it is usual to find the oscillator to be twice the frequency and divided to produce the opposite phases for the push-pull stage. The frequency source used in the FAT5 design by Dave GW4GTE uses a DDS oscillator and produces sine and square waves and is used at twice frequency. The reasoning behind this was to obtain accurately timed antiphase waveforms for driving the push-pull circuit. The DDS modules are now becoming difficult or expensive to get hold of. To use an oscillator at the transmitting frequency, creating two square waves exactly in opposition is difficult using conventional integrated circuits because normally Q and bar Q outputs are shifted in time. This causes problems because the push-pull stages will not be working in exactly opposite modes. I have found a high-speed comparator that can easily convert a sine wave (or square wave) to complementary output square waves, Fig. 1. This is an LT1016 comparator, which I use with a single-rail supply.
I tried various means of making an oscillator for 3.615MHz along with a VFO to cover that frequency but I couldn’t get one stable enough. It varied only a few Hertz but that was enough to cause asymmetric sidebands (one sideband output higher than the other). A crystal oscillator would be the ideal solution but as with most components these days, they are not readily available and more than one will be needed for other frequencies. There is a UK company that can make these but they will not be cheap because they are not mass produced. I tried ‘pulling’ a ceramic resonator to frequency and although I could cover the required frequency, it was varying by a few Hertz. This is not detected on air because it’s only a tiny fluctuation but on SDR it can be seen as asymmetric sidebands on the AM transmissions. This shows on average a 6dB difference in levels in the sidebands at low frequencies. It’s not quite so noticeable above 1kHz but it’s there and because it’s there it will be noticed and will commented upon. Fig. 2 shows perfectly symmetrical sidebands at 300Hz. Before SDR no one would know or even care but with many operators using SDR they look for any imperfections! So, a stable oscillator had to be found. I tried a GPS and, as expected, that was very stable but an inconvenience to use with a transmitter. On saying that, I doubt if anyone would say that you were not on frequency! The next best thing was the Si module and an accompanying Arduino Nano. This certainly works and because it has three clock outputs, three frequencies can be programmed into the device.
The full circuit can be seen at Fig. 3 and all that is needed to make it a very efficient AM transmitter is a modulator and a power supply. The frequency generator is on the left-hand side of the diagram and consists of an Arduino NANO Ver. 3. This controls, by means of I²C control lines, the next item, which is a clock oscillator type Si5351A, a chip manufactured by SiLab and. It can provide three simultaneous (independent 50Ω impedance) square wave outputs and can be anywhere between 8kHz and 160MHz. Clock 1 has been programmed to provide a frequency of 3.615MHz. This can easily be changed and the two other clock outputs can be used to provide any other frequency.
The output from the Si5351 is connected via a short coaxial lead that has an SMA plug to connect to the Si chip that I have soldered onto the pad. The other end is to be soldered to the pins on the PCB. I could have designed the PCB for the Si module to be permanently connected to clock one but having a ‘flying lead’ allows the choice of a second frequency to be used if required. The screen of the lead is connected to 0V and the inner conductor to a 100nF capacitor that is the signal to the input (pin 3) of the LT1016 IC. This is a comparator with complementary outputs. The comparator outputs are taken to the FET driver TC422A, which is a fairly conventional circuit, along with the pair of FETS (IRF640s) in parallel on both sides of the push-pull arrangement. The outputs to the transformer are taken via tabs connected to one each of the IRF640 FETs as shown in the diagram Fig. 4. The nut and bolt holding the FET in place on the aluminium strip must be isolated from the aluminium and hence the ground potential. For the two FETS that need to have the tabs connected to the drains and the nut and bolt, the insulated nylon bush is mounted from the underside of the PCB so that the nut and bolt is not in contact with the aluminium angle. See Fig. 5 for the method of mounting the FETs and drivers to the PCB using the aluminium angle.
The Output Transformers
The transformer arrangement and winding details can be seen at Fig. 6 and as a photo at Fig. 7. It may be better to look at this diagram along with reading the text here. Both transformers are the same and the former is a T200-2 toroid. This is an iron dust mix type suitable for the frequency we are using for this project. It is the same one used for 160, 80, 60 and 40m. The wire used is mains grey covered cable as used for house wiring and is 2.5mm², the thicker the better. Take about one metre of the live and neutral leads of the cable and discard the bare earth wire. Place the two wires together and place in a vice or pair of pliers. The other end pair is fitted into a power drill. Switch on the drill gun setting to a slow speed and the wires will twist, forming a bifilar wind. Switch off and remove from the drill gun when the wire is twisted but not so tight that it cannot be separated! Repeat the operation with another pair of the same length.
Take one of the bifilar wires and feed through the large centre hole of the toroid. Make nine turns remembering that the first pass of the wire through the toroid is one turn. This is about 0.9µH. Repeat the process with the other bifilar wire and toroid. Take the brown wire and untwist from the blue one so that you have a free end at the toroid and repeat this for the second toroid. These we will call the start windings. Do the same at the other end (end winding) of both toroids so that you have two toroids with brown and blue open ends at both ends. Take the end blue wire of the first toroid and the start of the blue wire of the second toroid, connect together and insulate. The brown start wire on the first toroid and the end brown wire on the second toroid is connected to the drain tabs on two of the FETs as per the photo. The end brown wire on the first toroid is joined to the start brown wire on the second toroid and connected to the PA voltage. The start blue wire on the first toroid is connected to the variable capacitor and fixed parallel capacitors at the antenna connection.
What we have now are two transformers with the primary windings in parallel and the secondary windings in series to provide a transformation of about 5Ω primary to 50Ω secondary. The photo, Fig. 7, shows the blue wire connecting to the antenna passing through a smaller toroid. This is called an ‘L’ match coil and is used to reduce the drain current if the power supply is near its limit. That will also mean increasing the drain voltage to get the same power output. This extra toroid, if used, is a T130-2 or similar size but must be a -2 type (red with grey/neutral underside). The variable and fixed capacitors are also connected at this point. I used a three-gang variable as found in older broadcast radios but any single- or twin-gang can be used or a large ‘postage-stamp’ type trimmer as I used in the PA shown in Fig. 7. There are fixed capacitors fitted in parallel with the variable and I use silver mica types made up to the value of approximately 2.5nF. It is prudent to use a lowpass filter (LPF) for the band in use to remove or at least greatly attenuate the odd harmonics. The even harmonics are taken care of in the push-pull arrangement.
Choice of FETs
The IRF640 FETs are intended for switch-mode power supplies but work quite happily at frequencies up to 7MHz. However, it must be emphasised here that not all IRF640s will work at RF. It would be prudent to buy branded types and of the fifth generation. I use and supply International Rectifier types. NEC types have also been found to be useable at HF. These FETS have low Rds (Resistance Drain to Source), which means a low voltage across the device when at maximum current through it. These were also chosen for their low cost, gate and drain capacitance and are readily available from UK sources. To fully turn on an FET it needs a square wave of sufficient amplitude and this is supplied by a TC4422A FET driver.
The PCB is available from me either on its own or with a selected kit of parts. It is double-sided, Fig. 8 (not to scale), and I will have fitted copper rivets where shown and soldered to connect both sides of the boards. Where the component leads are fed through the top to the bottom ground plane, these need to be soldered both sides and the source pin of the FETs are one case where this has to be done. The transorbs can be omitted or added as a protection against a high voltage spike at the drains.
There are several types of modulator that can be used, connected in series with the PA voltage. There are some ideal modulators found on the website of Dave GW4GTE (URL below). There you will find two analogue types and one digital modulator for which all parts are readily available. The two analogue modulators are for either medium- or high-power transmission. The digital modulator is a pulse width type (called PUWMA on Dave’s site). This is a very efficient modulator and recommended for use with this transmitter to provide an all-round efficient AM transmitter.
The completed inside view of the transmitter with a pulse width modulator fitted is shown at Fig. 9. This shows the toroids with red wire for the primary and secondary because that is what was to hand at the time. It would benefit from the twin brown/blue 2.5mm² wire as shown on the transformer of Fig. 7.
This transmitter can be tested without a modulator fitted. Apply the voltage for the oscillator, 12V (or 13.8V) and connect a pair of oscilloscope probes from an oscilloscope with a ‘Y’ response better than 20MHz, to the outputs of the comparator at pins 6 and 7. These waveforms should also be seen at the inputs (pin 1) of the TC442A FET drivers. The waveforms should look as in Fig. 1. Next is to connect the antenna socket to a suitable dummy load and power meter along with an analogue voltmeter (AVO) at the PA voltage input with an analogue ammeter in series with the PA voltage supply. Set the voltage, preferably from a variable voltage power supply, and slowly increase from zero to 6V. The current seen on the ammeter will be quite low, less than 500mA. The wattmeter may read a few hundred milliwatts.
Now connect the oscilloscope probes to each of the drain connectors (use the tabs). With the oscillator and FET driver voltages already on, switch on the PA voltage and adjust the variable capacitor until the waveforms appear as shown in Fig. 10. If the waveform looks about right, you can gradually increase the PA voltage while watching the PA current and observing the waveform. You may need to adjust the variable capacitor again to get this waveform. If the waveform is not as in Fig. 10, you may have to increase or decrease the fixed capacitor values across the variable.
Observing the waveform as you adjust the variable capacitor will indicate whether you need to add or remove fixed capacitance. It must be borne in mind that the method of tuning the PA is not ‘tune for maximum smoke’ as in other types of PA but is a careful tuning of the charging rate of the coil/capacitance flywheel effect. If all looks well with the waveforms, you can increase the PA voltages and by observing the current and power output until, say, 150W is seen on the power meter. The PA voltage will be about 24V with 7.5A. If you divide this input power (24 × 7.5 = 180W) into the output power as seen on the power meter (150 ÷ 180) × 100 = 83.3%. This is a guide and I have achieved 95.7% on mine. The higher the output power, the better the efficiency. For a full explanation of Class E visit Dave GW4GTE’s website as before and also look up the site of Steve Cloutier at:
Is There a Kit?
I can supply the PCB and all the parts that are to be fitted on it. Send me an e-mail for the details of a picking list and you can decide what you need if not wanting all the parts from me. The parts off-board are not included in the picking list unless you have difficulty in obtaining any of them.
This article was featured in the September 2018 issue of Practical Wireless