This post wraps up a series describing the construction of a 200W Class D AM transmitter for 160m, built, tested and put on air over 6 months from August 2019. The transmitter consists of a digital PLL VFO and driver circuit, an Arduino controller, a microphone amplifier, a Pulse Width Modulator, two H-bridge Class D RF modules each capable of at least 100 watts of carrier, an RF power combiner, Low Pass Filter, a 300W power supply and a switching regulator. At the Australian legal limit of 120W carrier, all parts of the transmitter run cool due the use of switching designs. Previous posts describe the switching regulator, 300 watt DC power supply and dummy loads for this project.
The engine of the VK3SJ low-band AM transmitter is one or more of the Modulator/RF modules, pictured above. These modules, designed by Laurie VK3SJ, consist of a Pulse Width Modulator (PWM) and a Class D H-bridge RF power amplifier, capable of delivering 120 watts of carrier without raising a sweat. The modules can be stacked in parallel, with an off-board RF combiner doing the work to sum each unit’s RF power. Another feature of this module is the combination of PWM and the RF stages on the same compact board, something that defies the conventional wisdom of the Class E crowd. All up, a remarkable piece of audio and RF design.
Laurie’ s design has taken years of development and experimentation time and he has kindly shared his circuit and construction details for experimenters and home construction. Some of his other designs are here courtesy of Ralph VK3ZZC. The complete schematics of this transmitter are below.
PWM build and test
Construction began by hand-copying Laurie’s PCB layout in my usual way, using permanent markers to draw in the copper lands and tracks, followed by a Ferric Chloride bath, a wash and clean, and finally a spray of clear satin enamel. I was careful to reproduce Laurie’s original board layout which is arranged carefully to cope with high DC and AC switching currents.
The module is designed to be mounted vertically, employing a 3mm x 25mm x 40mm angle aluminum bracket, a heatsink bolted to the 40mm face. The space underneath, between the overhanging angle stock and the board, houses the PWM LPF and RF output transformer. The vertical design allows a pair of modules to be mounted together, back to back, sharing a fan. More on this later.
The Pulse Width Modulator part was assembled first, comprised of an 8MHz crystal and a CD4060 CMOS oscillator, 14-stage binary divider and buffer, to generate a stable 250kHz square wave. The LM311 comparator, TL071 audio preamp and IR2110 high side driver and some additional control and squaring discretes were added in. After a few trimpot adjustments, the circuitry worked first time, rewarding me with a stable, modulated 250kHz waveform, and drawing about 35mA. Using only an electret microphone insert, I could fully modulate the pulse train with a loud whistle. So far so good.
PWM Low Pass Filter
Next came the PWM low pass filter (LPF). This is a conventional Butterworth 4th order LPF with -3dB point at 28kHz, designed by VK3SJ, reproduced using SVC Filter Designer from Tonnes Software, the report is here.
The inductors were wound on RM10 cores. RM10 cores are a pair of ferrite formers and cups, designed to encase a bobbin wound with wire. Use of a bobbin is preferable but I didn’t have any so 1mm enameled copper wire was wrapped around a convenient marker pen, then worked by hand into one RM10 half. Two and a bit more layers got it close to 126uH with a 0.5mm gap between the two halves, which I made up with filed down Veroboard super-glued to one piece.
You can buy gapped RM10s — that would be sensible! As I hadn’t, I worked with what I had. Once in the ballpark, adding or removing a half-turn is done to get as close as possible to the desired value.
RF switching PA
The RF part (H-bridge) of this module is minimal. A 1V PP signal at the transmit frequency from the VFO/buffer drives an IXDD614 high-speed gate driver, able to source and sink 14A of peak current, and intended for high-frequency and high-power switching applications (Class-D switching amplifiers are specifically named on the datasheet).
The RF power stage consists of four IPP350N15 N-channel MOSFETs in an H-bridge configuration. These Infineon devices tolerate 150V between drain and source, when conducting exhibit a tiny 53 milli-ohms between drain and source, and can pass 21A. Pretty impressive for a TO220 device.
Using Lauries PCB artwork made this assembly easy, although as my board was hand drawn, I carefully checked the vicinity of each component for any errors. The phasing of the four gate transformers is essential to get right.
VFO and controller
The transmitter VFO is my usual Arduino/LCD/si5351 combination, which for this project generates a 1.8MHz square wave signal when the PTT line is grounded. I use the Raduino circuit with a few modifications. Only a single clock is required to drive this RF deck. The si5351 clock output (around 1.5v PP) at signal frequency is amplified by a discrete MMBT3904 to around 22dbm, padded down by 3dB, then squared using a 74HC04 inverter. A trim-pot across 5V provides a DC offset on the inverter input, allowing some control over the duty cycle. See the circuit diagram for details.
The Arduino Nano handles all transmit-receive switching and sequencing, including enabling the PWM, controlling the T/R relay via a 2N7000 switch, DC switching and enabling/disabling the VFO output. It also provides a CW keyer and metering, not used in this project. My code is here.
RF PA Testing
Bringing up the RF PA was straightforward, after carefully checking for any wiring or transformer phasing errors. The drive is a 5 volt peak square wave with 50% duty cycle. With 12 volts on the IPP530N15 drains and a 50 ohm dummy load connected, the oscilloscope was used to check the waveform and magnitude on the driver input, output, each of the FET gates, drains, and finally, across the load. If the LPF is available it can be connected as well, so that a perfect sine wave can be observed on its output.
At this point, the duty cycle was adjusted to get as close to 50% on all four gates as possible. Turning up the drain voltage smoothly increased power. The audio gain was set full, anticipating that modulation level would be controlled via the microphone amp stage. Thirty volts of HT delivers about 10 to 12 watts of RF carrier into the load. I connected my 160m inverted L and witnessed a nice stable S7 carrier on VK3KHZ s WebSDR about 20km away. If all four FETs are just equally warm and none are hot (or hotter than the others) and the ratio of power delivered to the load is above 80% of DC power in, then the PA is probably working as designed.
The modulator required a line-level audio input, so a simple microphone amplifier module was needed. I chose one from a Drew Diamond VK3XU design using a 2N5484 FET sand a TL071 low noise op amp. This little board was done in one evening and supplies up to 2 volts PP of audio, with some basic frequency low pass filtering.
Building this transmitter was a 6 month project, so construction of the transmitter’s 300W DC power supply and switching regulator were described in earlier posts. Even making up the dummy loads was enough of a story to warrant a post. It remained then to mount the remaining boards and heatsinks and wire it all up to the panel switches, meters, audio and RF sockets, etc. in the 2U rack box from Jaycar.
Depending on the HT available to the switching regulator, this Class D transmitter offers continuously variable RF carrier power from zero to over 100 watts via the duty cycle control in the switching regulator. The PWM fully modulates the carrier across this full RF power range. The AM modulation level is full and does not need additional audio processing.
One PWM/RF module draws around 1A at 80V HT (unmodulated). The scope shows 62 watts of carrier. In other rests odulation peaks easily hit 250W and as high as 300V PP (400W PEP).
This transmitter can now be heard from time to time on the daily Melbourne Coffee Break AM Net on 1825kHz, 1100 local time, or 0000UTC southern hemisphere summertime, 0100UTC otherwise. You can listen to this Net wherever you are via the VK3KHZ WebSDR.
Second RF PA module
The idea with this transmitter was to build two identical RF power modules and combine them with an RF power combiner to deliver VK AM legal limit carrier power (120 watts carrier, 480 watts PEP). Wayne VK3ALK advised that one modulator would comfortably drive two H-bridge RF power modules, a total of 8 FETs with around 70V HT.
I built up a second board with just the RF H-bridge (no PWM or LPF) and tested it. This module was mounted on an identical heatsink, to be placed back to back with the original module.
RF power combiner
The RF combiner is this one from W8JI. I built it up on its own board, mounted flush against a 2.5mm angle aluminium bracket for physical support and heatsinking. The W8JI Magic T power combiner takes two RF power sources and delivers the vector sum of the two inputs to the output.
The main transformer (5 turns bifilar) does the combining, and presents a 50 ohm load to each PA module. Any difference in the phase or power levels between the two ports is dissipated in the two series 50R ceramic power resistors. The impedance at the transformer’s center tap is half that of either input port (25 ohms) so a 1:1.4 auto-transformer follows it for a 1:2 impedance step-up (7 turns tapped at 5 turns above ground). For the ferrites I used Amidon FB-43-1020 beads.
Powering this combination up was happily uneventful, and the transmitter now delivered 120 watts of carrier into a load on an 70V HT, drawing around 2 amps. As predicted, the Pulse Width Modulator modulated both PA modules and only got mildly warm.
This is not a beginner’s project! As well as many hours of sourcing parts and making PCBs, a bag of parts (mostly semiconductors including four IRFP260 FETs and three IPP530N15s) paid the ultimate sacrifice during development. The IRFP260 is a 50 amp device, but I learned that as rugged as they are, excessive drain to source voltage takes them out in microseconds. An important resource to bring to a project like this is a dose of stubborn determination. At least for an QRP guy like me!
Power supply ‘ghost’ ripple
Final chassis mounting always takes longer than expected, and it was not without a few problems. Despite following normal RF construction techniques, including a 2.5mm thick aluminum chassis base, directly earthing all boards to an aluminium angle bracket at several points, and shielding every signal line, I found RF intrusion or earthing problems were de-stabilising the Nano at power levels above 20 watts, causing the TR relay to chatter.
The scope showed 1 to 1.5v ripple on the 12V DC supply line, increasing with RF output power. No amount of lifting 12v supply wires or decoupling made any difference. Finally (after at least three unsuccessful debug sessions) I called Laurie VK3SJ after the Coffee Break Net one morning, and he diagnosed an un-shielded scope probe immediately. The 1V ripple on the 12V DC internal supply was a fiction – I was seeing scope probe pickup.
Beware! Most ‘cheap’ digital scopes are sold with un-shielded probes. This kind of problem doesn’t occur when you build QRP rigs! The final diagnosis of the chattering TRrelay was a software bug (thanks VK3KR).
Switching regulator issues
The switching regulator takes up to 140V DC and delivers a continuously variable output voltage from 0 to about 90% of the DC input voltage at 1.5 to 2A continuous with peaks of up to 4 times (on AM modulation peaks). It has current protection and can be tied into interlocks such as VSWR sensing control circuitry to reduce power in an exception. A conventional linear regulator with these specifications is a non-trivial thing to design and operationalise and must be significantly over-specified (‘built like a battleship’) to survive AM service, as well as having to sink a lot of heat. For tvese reasons, a switching regulator is the only way to go. But this being my first, it certainly taught me a few lessons!
A key component in any switching converter is the choke (120uH). It is in series with the HT and must pass the peak currents. My first attempt at this, wound with home-made Litz wire rather haphazardly on an old former pulled from a PC power supply, buzzed, hummed and rippled wildly around 70V under a load of 1.5A. In use on my first regulator board, I had trouble maintaining a consistent PWM drive over the range. A number of the components including the FET, its gate transformer, most of the series metalised polyester capacitors, and the filter capacitor were swapped in and out at various times.
My module was prone to a failure mode in which the IRFP260 FET would short source to drain, placing the entire DC input on the output. This happened whenever there was a sudden variation in load (current). On one occasion the halogen globes being used as a dummy load blew, taking out the FET. On another, the FET blew for unknown reasons and rammed 150V DC onto the PA and modulator, taking out half of the modulator and PA FETs. On another, a replacement FET was not in tight contact with the heatsink angle piece and blew in an instant. I got used to replacing the FET and testing the remaining components.
Things stabilised when Laurie VK3SJ passed on a choke he had made, a fine piece of work, wound with 24 fine enameled copper wires, and tightly packed onto its former. Another change was replacing the IRFP260 with a higher voltage (Vds) rated replacement.
The regulator has built-in current protection, but this was not always straightforward to adjust correctly, and so it was not used for much of the testing. Normally in the presence of a current surge, fuses will go first, but with these high voltage, high current switching designs, the voltage and current spikes are too fast for conventional fuses, and the three-legged silicon fuses blow first.
The final change was to correct sagging gate drive under load. Laurie suspected a dubious gate toroid, but I tried various ferrite with inconsistent results. In the end I ripped out the gate drive transformer and replaced it with the high side MOSFET driver in an IR2110. This fixed the problem with style, delivering a rock solid 12v PP square wave to the regulator’s switching FET under heavy load conditions. This mod is included in the transmitter schematic, above.
All of these failures were largely due to my ignorance rather than any shortcoming in the design. With these changes and the final selection of components in place, the two regulators I’ve built both work perfectly.
Power transformer re-winding
To get 140V DC I wound a new 100V AC secondary onto a surplus toroidal mains transformer. This worked fine, but the various switching regulator failures made me nervous about running the regulator at this power level. Lacking a Variac, the solution was simple. I removed half of the secondary winding, laboriously un-threading it to avoid cutting it, and wound it back on, stopping to create a tap every 10 volts. This gave me a 100V AC secondary with taps at 50, 60, 70, 80 and 90V.
Each tap was brought out to a terminal in a block. Now, moving the input wire to the rectifier up and down these taps gives maximum DC voltages between 70 and 140V DC, allowing any changes to be tested at lower regulated HT voltages.
Higher power AM necessitates components that cannot be found in the junk box or in the local electronics shop. Even from the global e-tailers, supply must never be assumed, as parts that have been easy to source in the past can become unobtainable. Part sourcing consumes many hours of project time.
The FETs, drivers and high current diode cannot be substituted and must be original manufacturer stock — eBay suppliers are NOT to be trusted. Other passives that required effort to source included the ferites, the IE33 choke assembly, RM10 cores, metalized polyester capacitors for the PWM LPF, 250v electrolytics, silver mica capacitors for the RF LPF, and low ESR electrolytics for use around the switching circuitry in the power end of the PWM.
In the end, all of these parts were sourced from Digikey, RS Components and locals Minikits, PKLoops, Jaycar, Altronics, Rockby and one or two pieces from eBay. The big ferrite beads for the power combiner came from local Amidon importer TTS Systems. I also bought some DIL TL598C ICs from Futurlec. Aluminium is from Action Aluminium, Thomastown (Melbourne).
It is prudent to land the parts before designing PCBs, as you may not have the IC package option that you prefer, even from the huge stocks of these global suppliers. I always buy more parts than I expect to use, even allowing for failures, to avoid a 2 to 3 week wait and the Post & Pack costs for a small unexpected replenishment order.
Thanks again to Laurie VK3SJ for the design and for ongoing email and telephone help, to Wayne VK3ALK for extensive email and phone assistance, and to the 160m Coffee Break operators for encouragement. David VK3KR helped by explaining a number of concepts and was a sounding board for ideas and options.
All up, this has been an engaging and entertaining project and the results have exceeded my expectations. Using switching (digital) technology to generate Amplitude Modulation at 100 watt power levels and above is a fascinating world that has extended by QRP sensibilities in some slightly mind-bending ways. There has been much releasing of magic smoke and a lot of head-scratching but each of these explosions brought a little or a lot of insight. More digital AM projects are planned! See you on 160m AM.