Laurie VK3SJ is an accomplished RF designer and homebrewer in Melbourne’s 160m scene. You can hear his excellent AM signal regularly on the 160m AM Coffee Break net most weekdays. Laurie has spent a great deal of time experimenting with Class D, E and Pulse Width Modulators. He has an interesting Class D AM transmitter design with a pulse width modulator. Laurie’s design has been built by Wayne VK3ALK, who I met online in the Class E Forum on AMFone. After lots of emails and a few calls with these homebrew AM experts, I decided to proceed with my build of a VK3SJ AM transmitter for 160m.
Laurie’s transmitter consists of a pair of 150 watt class D RF power amp and PWM modules feeding an RF power combiner. My construction of these universal RF and Modulator boards will be outlined in a future post. To complete my AM transmitter I will add an Arduino and si5351, microphone preamp, Low Pass Filter (LPF), T/R switching, 12v and 150v DC power supplies. Laurie’s design includes a switching regulator for the HT supply, the subject of this post.
The module is a switching buck mode regulator with over-current protection. It regulates 150v DC from a rectifier/filter bank using a TL598C PWM IC, driving a IRF260FP switch, with the usual shunt diode, choke and filter capacitor. It includes a current limiting circuit and other protections. Output voltage is varied by controlling the PWM duty cycle. This post describes the regulator’s construction and testing.
The first challenge was to make a 120uH switchmode power supply choke. The recommended former is a PC40 EI33 core. Although freely available on eBay, Wayne VK3ALK advised that useable formers can be ‘liberated’ from old PC power supplies. I tore down two (both sourced from one of my favourite electronics suppliers, neighborhood ‘hard garbage’ piles). This yielded one without and one with breakage.
These transformers are quite sophisticated things, with many layered windings, some made from 3 or more pieces of enameled copper wire in parallel. This forms a litz wire, a technique designed to minimise ‘skin effect’.
Removing them from a road kill computer power supply takes about half an hour. Grab the large low-ESR electrolytic capacitors, fan, output toroid for its enameled copper wire, and any other individual parts you like the look of while you’re there. The large switching FETs and diode look tempting, but better ones can be had from the online retailers for a few $ each, so these were left.
Next, it took a good hour of careful unpicking the tapes, insulating pieces and enameled copper wire (ECW), working with eXacto modeller’s knife, jewellers’ screwdrivers and small pliers. If you go in too hard you’ll break things. The two E shaped ferrites slot into the former or bobbin, which is made of some kind of heat resistant synthetic. It appears strong but is brittle, as I learned when one half of the base snapped off under gentle pressure from my tools.
After all the windings and tape have been cut and scraped away, the assembly needs to come apart. The ferrite Es are glued to the bobbin. ‘Use a heat gun’, Wayne advised. I gingerly held the assembly over a low gas jet with plyers, turned it rotisserie style, and quite quickly, the ferrites slipped apart. Easy when you know how. The bobbin was not visibly damaged by the heat.
Next, I roughly wound on 20 turns and measured the inductance. With ferrites inserted and a 2mm gap between the two E pieces, 20uH. With both ferrites firmly pressed together, 130uH. Amazing! Not only do the ferrite cores amplify the inductance by a factor of many hundreds, a couple of mm of physical gap between them accounts for 6 times the inductance.
The term ‘Litz wire’ is derived from the German word ‘Litzendraht’ meaning braided or woven wire. It is constructed of individually insulated wires either twisted or braided into a uniform pattern. When I read ‘litz’ wire I think of wire composed of a large number of very fine wires all in a bundle, something like this beautiful stuff. However, apparently even 3 or 4 conductors wound in parallel is beneficial, reducing the skin effect in low radio frequency alternating currents up to 1MHz, typical of switch mode power supplies. Skin effect describes the tendency for AC to flow through only the outer portions of the cross-section of a conductor, thereby increasing resistance.
I made up a kind of ‘litz’ wire from 6 pieces of 0.5mm ECW held between a vice and a power drill chuck for some quick intertwining. The resultant litz wire made about 15 turns, giving 70uH with a 1mm gap. Not enough. Another six lengths were cut and the process repeated, adding another 5 turns. Now to trim the gap for 120uH. After a fair bit of experimenting I got close to 120uH, but only after filing down two Veroboard spacers to about half hormal thickness.
The choke was completed for testing by applying more epoxy to the remaining E to keep it in place. Wayne advised to fill the windings with glue from a hot glue gun to stabilize them, and to stop them ‘singing’. I deferred this step until after testing. If it was going to sing, I wanted to hear its tune, at least once.
PCB and Assembly
I hand-made a PCB for the regulator in my usual way, with marker pens on the copper side, a FeCl3 bath, then a clean and spray coat of satin enamel. The board was designed to be mounted vertically next to the toroidal mains transformer, sized to sit against a piece of 80 x 25 x 3mm aluminum angle as a physical support, RF screen and heatsink. A cutaway on one side of the PCB allows the FET and power diode to heatsink to the angle, via mica insulating washers.
First, the TL598C PWM IC was checked out. It ran at 100kHz after a small adjustment, and the power potentiometer adjusted the duty cycle from 10 to 90%. The enable pin turned the pulse train on and off.
Next, I put HT on the IRFP260 switch FET. Powering up these kinds of circuits gradually helps to avoid smoke! I started with 30v HT from my variable power supply and connected it to a 6 ohm 80 watt dummy load I had used to test an earlier Pulse Width Modulator. The choke sang a high pitched whistle faintly as it delivered 90 watts into the load.
Wayne advised replacing the 0.22 ohm 5 watt current sense (shunt) resistor with a 1 ohm one so that the current sensing circuit could be tested at a lower power level. It checked this circuit out in an isolated test and the control circuit worked fine. The ultimate test involves shorting the supply at full power. I shall wait until everything is set up in its final configuration before I try this party trick.
Regulating a 150v HT had to wait until the mains power supply was built. This uses a partially rewound 500VA toroidal mains transformer, rectifier bridge and filter bank. With all the usual protections and fuses of course. A 20 watt 12v linear regulated supply for the low power circuitry (microphone amp, TL598C PWM) was added to the same mounting bracket.
With 150VDC HT the regulator performed perfectly without load. When a 50 ohm RF dummy load was connected, a few bugs surfaced. Firstly, the output voltage increased from 0 to 125VDC smoothly, at which point, advancing the power potentiometer caused the pulse train to drop sharply from 90% to about 10%. This is behaviour is still unresolved.
Secondly, the choke started to ‘sing’, in an unnerving shrill and raspy voice, starting at around 50V (1 amp), peaking at 70V (1.5A), then oddly, settling down around 100 to 125V (peaking at 2.5A, that’s 300 watts).
Looking at the AC ripple on the regulator output with the oscilloscope, I could see it change from a few 100mV to 5 volts at the peak of raspiness, around 70V, decreasing to mV above 100V. This problem took quite a lot of diagnosis, but was traced to my decision to mount a trimpot in the power regulation circuitry on the top of the board, for easy access. I had allowed the long leads that were necessary to wire it in to the PWM circuit to run directly next to the FT50-43 transformer on the input to the IRFP260 gate. Once the trimpot was relocated, the gate waveform cleaned up significantly. Lesson: a switching regulator can suffer from feedback just like any amplifier, so keep the sensitive PWM control circuitry switching paths away from the high power switching parts.
The HT power supply delivers 145V and 139V under load into the regulator, which then reliably delivers 120V at 1.5 amps (180 watts continuous) into a single 500 watt halogen globe. This is an efficiency of around 86%. This power level should be about right for a class D PWM AM transmitter with an output of 100 to 120 watts of carrier, the VK limit for AM transmissions for an Advanced Class licensee like myself.
Lower resistance loads will consume more power, of course. The maximum I saw was a shade under 300 watts. At around 200 watts everything remains cool or runs just warm (other than the current shunt resistor). If a higher power level is called for, the circulating air from the RF PA fan(s) in the same cabinet should provide sufficient cooling.
Thanks to Laurie VK3SJ and Wayne VK3ALK for coming up with the design for this regulator and transmitter (Laurie) and for helpful conversations and email assistance (both).