Many cool things require high voltage DC supplies. With cool things, I mean things that measure radiation—Geiger-Müller tubes—as well as things that can multiply and measure individual photons—Photomultiplier tubes—and finally, things that can measure individual photons and project them onto a visible phosphor screen: Image Intensifier tubes.
In today’s world, you’re more apt to find a nice +3.3V or +5V supply—with a USB cord, naturally, than a good ol’ kilovolt supply to power these marvellous devices. Fear not, however, for this is exactly what we’re going to rectify. Now, we can’t just have a single, fixed output voltage; some devices will want 500V, others 1000V. So we need something with a bit of choice.
The Inimitable Transformer
There are several ways to go about this. One would be to use a simple boost converter. Charge up an inductor (or a coupled inductor), then shut off the current and let it discharge. This works pretty well, but is a rather unpredictable process. The high voltage spike generated (the back EMF) depends on a number of factors, including inductor resistance, inductance, the core material, parasitic capacitance and so on. Completely without regulation, just letting an inductor kick high voltage spikes into a circuit, is rather what bug zapper rackets do.
We want a predictable supply, and we want one where we can have multiple output values. Sure, you could close a feedback loop around the boost converter. But the individual spikes are sharp—so with a reasonable amount of capacitance at these high voltages, we’d get insane amounts of ripple. Not nice.
Instead, let’s go for something a tad more orderly. How about a high-frequency inverter with a high step-up ratio? By “chopping” up our DC, we can pass it through a transformer. And because we don’t have to output 50/60Hz mains, we can keep the transformer pretty small. We could use an H-bridge to switch this transformer in both directions, but I wanted to keep part counts low—and didn’t have any nice, high performance H-bridge ICs lying around—so I instead decided to make the primary winding centre-tapped. This in practice means you create two windings, of equal size, and join them at the middle. The end of one primary winding is connected to the beginning of the other. You must, of course, remember to wind them the same way around the bobbin—otherwise it won’t work well.
So I went and bought an EE35 bobbin, and a set of PC40 ferrite cores. They’re made by TDK EPCOS, but you can get reasonably priced options on Aliexpress, like this one. All you need now is some magnet wire (that is, varnish insulated copper wire)—0.4mm diameter for the two primary windings, and 0.28mm diameter for the large secondary. And then, I wound a nice step-up transformer with a ratio of about 1,1:12. That is, 20 turns for each of the primaries, and 250 turns of wire for the secondary.
The circuit design
Now, I needed a circuit design. Fortunately, since we’re using a simple step-up transformer and an inverter, we don’t really need closed loop control. The cool thing about transformers is, if you increase the load on the secondary side, the primary sides’ effective inductance will decrease, allowing more current in. This is a property of mutual inductance—and the reason that this effective decrease will let more current through, is because an inductor’s reactance is given by
X_L = 2π \cdot f \cdot L, with f in Hz, L in Henries and X_L in ohm.
In other words, the lower inductance at a given frequency, the lower inductive reactance (in ohm) and thus higher current “pulled through” the inductor.
Now, with an input voltage of 12V, a step-up ratio of 12 won’t get us far beyond 150V. I could’ve kept winding until my fingers came off—or the winding wouldn’t fit inside the core *cough*, but this wouldn’t be good in perpetuity. More copper wire means more resistance, and additionally, since the secondary winding is essentially wound around the primaries (with a layer of tape in-between), more turns on the secondary also means more parasitic capacitance between the primary and secondary side, which makes for slightly less predictable performance and generally worse efficiency. Also, it’d still be a bit of a problem to control the output voltage if the transformer just output 1000V(AC).
So I opted to use a Cockcroft-Walton voltage multiplier on the ~150V(AC) secondary side. Why? Well, first of all, it allows us to tap the DC voltage for each multiplication step. So we can tap at ~300, ~500 and so on, using a simple, multiple-throw switch. This gives us exactly the control we want! Additionally, our secondary-side output is already AC and at high frequency, so we can use small capacitors and relatively inexpensive fast-rectifier diodes to get serious output power—up to around 20W—from our simple circuit.
The circuit, as designed, looks like this
The output frequency is adjustable, to allow for some control in terms of output current. Inexpensive MOSFETs are used to drive the circuit, and a cheap, low-cost signal MOSFET used to invert the control signal for the top driving FET. Instead of a BS170, a 2N7000 or BS107A is perfectly applicable here.
You’ll also notice there are two resistors, RD1 and RD2, that both serve to provide a “multimeter-friendly” lower-voltage output tap. The output voltage is divided by 11, so that a measurement of 90V = 990V. Bear in mind the input impedance of your multimeter, however, and don’t forget to factor this in as an additional resistance in the voltage divider. You can typically find values for this in your multimeter’s manual or datasheet. Apart from allowing lower-voltage measurements, these two resistors also serve to discharge the compound output capacitance formed by Co1 and Co2. The circuit discharges to sub-100V within 20 seconds, but I would recommend waiting at least a minute before poking about the circuit after turning it off.
Building the Beast
I chose to build this circuit rapidly, on one of my self-designed Protoface prototyping boards. It came out like this, before adding the final multiplication stage, putting in a box and adding a rotary 1P6T switch to select the voltage taps
The huge red caps are 1µF polypropylene 630V capacitors, connected in series. They really are some big beasts!
On the Scope and Meter
The output at the secondary side, with the voltage multiplication circuit connected, looks like this:
The large ringing artifacts are primarily caused by the capacitance of the voltage multiplier. You can see the top and bottom edges of the ringing is “cut off” by the rectifier diodes. Additionally, some ringing is contributed by the primary-to-secondary parasitic capacitance, which measures in at around 80pF at 100KHz—a not insignificant amount of capacitance at 50KHz. But in this case, the ringing is not detrimental to our circuit.
On the multimeter, I chose the voltage tap designated as 960V, and measured it directly with a multimeter, instead of at the 1:11 designated measurement point, to show that the voltages generated are no joke, even with the 1 megohm load of a multimeter:
Just to show a fun (yet also quite dangerous part), here is the supply eating a good centimeter of silver wire, discharging from the output – to +. Filmed in slow-motion, with the last bit in “real time” so you can hear the normal-speed discharge sound
A final word…
So that’s how you make a low-cost, efficient high voltage supply. One word of advice though: follow the warning in the schematic. High voltage electrocution is no joke. It causes fatalities that could’ve been avoided every year. So don’t treat it like a play-thing. If you are an idiot and don’t heed my warning, I won’t weep at your grave.
With that said, this circuit is an excellent way to power diverse high-voltage DC-requiring devices around your personal lab. Personally, I’ve got a GM tube and some photomultipliers coming this way… I hope you’ll join me, and if you do, be safe!