300W Class D Amplifier for 630m
This amplifier is based on the ‘300W Class D Transmitter for 136kHz’ published by the RSGB in Radio Communication and reproduced in the LF Experimenters Handbook by Peter Dodd G3LDO.
This circuit, updated and modified for 630m, uses 2 MOSFETs in push pull driven by square-waves from a CMOS D-type flip flop. The power can be varied from around 25 to 300W by increasing the supply voltage over a 12 to 32V range. Circuits are included to protect against excessive supply current and high reflected RF power.
The amplifier is suitable for any non-linear mode. Examples include CW, QRSS, WSPR, JT-9, JT-65 and FSK.
The circuit has been built on a dedicated double-sided PCB measuring 164 x 120mm. For ease of construction 90% of the components are leaded but it has been necessary to use a few surface mount devices as the normal DIP versions of some ICs are increasingly difficult to find.
A signal at twice the desired frequency is applied to the input connector. (1) This is AC coupled to the clock input of a CMOS 4013 D-Type flip flop. Potential divider R1 and R2 hold the clock input just below its switching point of 6V DC. Any signal from C1 exceeding apx 2.5V p-p will clock the circuit correctly. In practice any output from 3.3V to 12V logic, or an input from 5 to 50mW from a 50 ohm source can be used.
The outputs from the Q and not Q pins of the HEF4013 are
square waves at the wanted frequency and are 180 degrees out of phase.
These signals are buffered by a dual FET driver- IC2. The MC1404-EP has
fast rise and fall times which minimise internal power dissipation.
The 8 Pin DIP package is used as it can dissipate more power than other
surface mount variants of this device. Keying the drive is achieved by TR4 a
P-channel MOSFET in the 12V supply to the FET driver IC2.
Looking at the Q output, The buffered output from IC2A is AC coupled via C2 to the gate of PA transistor TR1. Diode D1 ensures the drive is referenced to ground. C2 and R7 ensure that in the event that a drive signal is lost, the gate is safely held at 0V and the PA transistor is off. C3 D2 and R9 provide a similar role for the other half of the amplifier.
The pair of IRFP250-N MOSFETs, TR1 and TR2, are driven in push pull. Each MOSFET has 1500pF in series with a 10 Ohm 20W resistor connected between drain and ground. This reduces voltage spikes on the drain and damps any ringing from unwanted resonances. The DC supply voltage for the MOSFETs is fed to output transformer T1 via RF choke L3. The output transformer is wound on a 42mm ferrite toroid core and has a ratio of 4+4 turns on the primary to 16 turns on the secondary. In the prototype, the secondary winding has been made from Teflon (PTFE) coated multistrand wire, while the primary uses a 16A rated twin flex which was wound on top of the secondary winding. The red and black twin flex used to connect power to most amateur radio transceivers is suitable, or a high temperature version can be found on e-bay for wiring boats or motorcycle engines. This is the same red and black multistrand wire but contained in a heat resistant plastic covering. The 4+4 to 16 ratio is suitable for 300W output from a 32V supply. If you wish to run the amplifier at a lower power, then just reduce the supply voltage. A rule of thumb is that doubling the supply will give 4 times the output power. The core used is a 42mm toroid from Ferroxcube using their 3C90 material. Other ferrite materials suitable for 100kHz to 1MHz will probably work without issue e.g the -77 mix from Fair-rite. However, the popular -43 mix used at HF will not have enough inductance per turn for 630m.
The RF output from the secondary of T1 passes through TX
/ RX relay RL1 to a 5 section low pass filter L1 / L2 and C11 – 14.
The LPF has a nominal cut off around 520kHz
making the design suitable for frequencies up to 510kHz.
The capacitive input of the filter works hard removing the sharp edges
of the square wave output and consequently L1 and C11 run slightly warm to the
touch. If needed, space is provided on the board for C11 and C11A allowing 2
components to be connected in parallel. A
computer simulation of the low pass filter is shown in fig3 below,
while a spectrum analyser image showing measured performance is shown in fig4.
A second relay is used to connect the LPF to an external receiver. It also terminates the receiver in 47 ohms when the amplifier is in transmit mode.
The RF output from the LPF passes through a 14mm diameter toroid core T2 which is used as a directional coupler. The primary is a short length made from the inner of some RG58 coax.
The directional coupler has a 15 turn bifilar secondary feeding a pair of Schottky diodes. These provide DC voltages representing forward and reflected RF power. RV2 and RV3 are used to calibrate the FSD of a suitable milliammeter.
The circuit implements over current and excessive reflected RF power protection by removing the squarewave drive from both PA transistors. In normal operation the set input of the 4013 (pin 8) is held low by R31. However, if the voltage on pin 8 goes high, then the 4013 sets the Q output to 12V and the not Q to 0V. This removes the drive from the PA MOSFETS.
VSWR protection is implemented by taking a sample of the DC voltage from the reflected side of the directional coupler and feeding it to RV1 which sets the trip point. The wiper of RV1 connects to op-amp IC5. When the DC voltage from RV1 exceeds the reference voltage on the non-inverting input, it triggers monostable IC4. This causes the output at pin 6 of IC4 to go from 0 to 12V which is connected to the set input (pin 8) of the 4013 via diode D6. The monostable inhibits the drive from the 4013 for apx, 3 seconds. The timing period is determined by R24 and C23. After the monostable has timed out the drive is restored. If the fault still exists the circuit will remove the drive for a further 3 seconds.
The over-current protection uses similar techniques.
Current sense resistors R14 + R15 produce a voltage across them proportional
to current flow into the amplifier. When this voltage exceeds 650mV, TR3 will
conduct and current flows through R36 illuminating the LED inside optocoupler
IC3. As the output at pin 4
changes state from high to low, it triggers the other half of the monostable
IC4 via pin 11. When
triggered, pin 10 of IC4 goes high causing the set input of the 4013 to remove
the drive signal for 3 seconds. After
the monostable has timed out, the drive to the PA
It’s important to note that both these protection circuits only remove the drive from the PA. Should a PA device fail for some other reason it is possible for the supply to be shorted to ground. Therefore it is important that a 15A quick blow fuse is fitted in the supply line. This provides protection for the power supply.
630m Amplifier Circuit diagram
Figure 1 630m Amplifier – Component overlay with top copper pads and tracks
Here are a few images of the amplifier and its components.
Figure 3 Original breadboard prototype. Cores on right are 42mm and 58mm 3C90 material
Figure 2 Printed Circuit Board. Less inductors for clarity.
Figure 3 Computer simulation of Low Pass Filter design
Figure 4 Measured performance of low pass filter using T157-2 cores. 1R) is 475kHz 1) is 2nd harmonic
Construction and testing:
The PCB (2) can be built and tested in 3 stages. First assemble the drive circuits around IC1, IC2 and the keying circuit TR4 up to the gate components of R6 – R9. When soldering the surface mount devices, carefully check that you have them oriented correctly and that there are no bridges between pins. Any surplus solder can be removed with de-solder braid. Remember to include R31 but not D5 or D6. At this point it’s possible to pause the assembly, apply 12V and an input signal, connect the key input to ground and check that a nice 12V P-P signal is present at the pads for each gate. If all the signals are as expected, the rest of the components with the exception of D5 and D6 can be added.
The PA transistors are mounted with at least 6mm / ¼ inch of their leads soldered to pads on the top of the board. They need an isolating washer between the body of the FET and the heatsink. A little heat sink compound is always a benefit. The transistors can be mounted vertically with their leads bent at 90 degrees or they can be mounted horizontally if that method best suits the heatsink and enclosure you have. Either way, should you ever have to replace an FET then it can be done very quickly without major disassembly.
The suppression components between drain and ground dissipate some power and so the resistor must be mounted to the heatsink. No washers are required as the resistor is electrically isolated from the body of the device. Areas of the groundplane have been left clear of solder resist for the earthed end of the resistor. The 1500pF capacitors are soldered directly to the drain at one end and connected to the resistor at the other. – See pictures for detail.
Output transformer T1 is probably the most difficult part of the assembly. Start with the secondary winding making sure it is in close contact with the core. I found that a cable tie is very useful to hold the ends of the secondary in place and under tension, while you wind the 4 turn primary over the top. The difficulty is that you end up with a nicely wound transformer with 6 wires that need soldering. Strip each wire allowing an extra 10mm or 3/8 inch so that the ends can be bent at 90 degrees once they are passed through the board. This provides a nice mechanical fixing and it increases the area of wire in contact with the PCB tracks giving better current distribution.
The directional coupler is bifilar wound with a pair of loosely twisted wires. I used two different colours of 28SWG enameled copper. Hold one end with a pair of pliers and the other in an electric drill. A quick press of the trigger will give you evenly twisted pair. About 4 turns per inch is about right for this transformer.
The low pass filter inductors should be wound leaving a small gap near the ends of the wires to allow the core to be glued directly to the PCB. This prevents the cores or the enameled wire vibrating against the PCB which can cause short circuits over a period of time.
When selecting a power supply, use one with good regulation. This is especially important if you are running a mode like WSPR, as any mains hum on the supply will produce sidebands every 50 or 60Hz either side of your transmitted signal.
With the circuit complete, except for D5 and D6
, do some quick checks to make sure there are no shorts to ground
around the LPF or T1 and then fit a 5A fuse in the PA supply line. Connect a
50 ohm load to the output and a 12V supply to the drive stages, and TX/RX
Connect a 12V supply to the PA stage and then short the PTT input to ground. Hopefully you will hear the relays switch to TX mode. Apply a signal at twice the wanted frequency to the RF input and then; the moment of truth ! - Short the key input to ground. At this point you should see between 2.5 and 4 amps of supply current and an RF output of 25 – 40W. Assuming all is well it’s time to do some initial checks on the protection circuits.
With the amplifier running you need to briefly touch a 12V supply onto one of the pads for D5-7 that connects to pin 8 of the 4013. I usually do this using a wire with a crocodile clip attached and a series resistor (1k to 4k7) to limit the current if you accidentally touch ground by mistake. As you touch the pad, the drive to the PA should be inhibited and the output power will reduce to zero. If all is OK, turn everything off and fit diodes D5 and D6. Adjust the preset resistors as follows. RV2 and RV3 to mid range, RV1 and RV4 fully clockwise.
This sets the current trip to somewhere between 7 and 9 amps and the other controls to a nominal setting. Set the amplifier up as before to run low power from a 12V supply. Use a DVM to test the voltage at the trigger inputs (Pins 5 and 11) of the HEF4538. The voltages at should be 12V at pin 11 and between 8 and 12V at pin 5.
You now need to test the correct operation of the monostable timers by taking these trigger inputs to ground. Starting with the current limit circuit, take a length of wire grounded at one end, briefly touch it on the end of resistor R28 which connects to pin 11. As the pin is grounded the monostable should be triggered. This should remove the drive from the PA for about 3 seconds. Use the same technique to test the reflected power protection by briefly grounding the end of R25 where it connects to Pin5. Again the drive should be removed from the PA, returning to normal after a few seconds. If you have LEDs D8 and D9 fitted these will illuminate for the timing period giving a visual indication of the cause of the shutdown.
Next, connect a voltmeter at the top end of RV3. The low power output should produce a few hundred millivolts from the directional coupler. Move the DVM to the top end of RV4, it should be at zero volts as there is no reflected power.
If you find that the readings from the coupler are reversed you will need to swap the two connections between the coupler and D3 and D4. Switch off for a moment and disconnect the load. Switch on for a few seconds and check that a voltage appears at RV4 then key up. Move the voltmeter to R25 and with the key up note the voltage, it should be between 8 and 12V. Key down again and check that with no load, the voltage reduces slightly, heading towards the switching point of 6V. Now switch everything off.
Before running the PA up to full output it’s necessary to solder some wire to a few tracks on the underside of the PCB. This reduces track resistance and IxR losses. I used 18 SWG tinned copper and soldered it to any track that carries high current. That includes the tracks from the FET supply input pin, past R14, C19 and 20 Also from L3 to the feed point of T1 and the 2 tracks from L1 to the FET drain connections. Less critical are the tracks from the TX/RX relay through the low pass filter to the output. With the wires added you’re ready for some high power testing. Swap the supply line fuse for a 15A quickblow type. If you have a variable supply set it to around 20V and with an appropriately rated dummy load use the key input to activate the amplifier. This should produce between 100 and 150 Watts. Monitor the supply voltage / current and RF output then key up. Do a check on efficiency which should be >80%. Gradually repeat this process increasing voltage as you go. At some point the supply current will cause the protection circuit to trip. When this happens adjust R4 slightly anticlockwise until the amplifier runs normally. Adjust the RV4 so the amplifier trips at 10 amps, hopefully this will coincide with 32V and an output power of 300W. If you have a meter or meters fitted to measure forward and reflected power, you can set the full scale deflection for forward power by adjusting RV3. For reflected power, using the same meter, adjust RV2 to the same percentage as RV3.
The final adjustment is setting VR1 for maximum acceptable reflected power. This depends a little on your operating conditions. Too sensitive and the amplifier will trip every time to try to match an antenna. I have set mine for around 50 Watts reflected power, but in my case I use a switched mode PSU that can be changed to a lower voltage for tuning. That leaves the SWR trip to catch any ‘incidents’ due to antenna detuning due to weather or mechanical issues.
Finally, cooling with a fan is a good idea, but it can be fairly modest. I use an 80mm 12V fan running in transmit only with a series resistor in the supply. This gives 8V at the terminals of the fan providing adequate circulation but with low noise. In normal use the heatsink will only get hot if the SWR is poor. The current sense resistor R14 will run hot in high duty cycle modes while the core of T1 and C11 and L1 may also run warm to the touch.
G0MRF August 2016
PCB and 3C90 toroids www.g0mrf.com/kits
Resistors capacitors transistors relays. www.digikey.com www.mouser.com http://uk.rs-online.com www.farnell.co.ukwww.rapidonline.com
1) A VFO kit that provides an output at twice the required frequency is available from QRP-Labs. http://www.qrp-labs.com/vfo.html
1a) An Analog Devices DDS design has been published by Johan SM6LKM
1b) A development version of WSJT capable of producing tones at twice the normal spacing has been written by K1JT. – This can be used with amateur radio transceivers that have been modified for non-amateur band operation or a dedicated driver at 975kHz.
1c) This amplifier is directly compatible with the QRP Labs U3S WSPR / beacon generator in its x2 mode
2) A PCB for this project with the SMD components IC1 IC4
and IC5 ready soldered is available from the author. The 3C90 cores and a
resistor pack are also available. - See kits
page for details.
March 17: Following a number of requests, a full kit of parts for the 300W amplifier is now available.