# Building a 10kV DC power supply

It’s time for the first post filed under ill advised experiments. Just in case you’d not got the hint from the name, don’t try this unless you know what you’re doing; one small slip and the only silver lining will be the cut-price cremation you’ll get for half doing the job yourself.

I’ve spent the last week building a 10kV DC power supply capable of about 100mA. In the main part it is composed of a microwave oven transformer(MOT), 4 microwave oven diodes and 60 camera flash capacitors. I got the microwaves from freecycle; I’ve put the capacitors and other 3 transformers aside for other projects. The old disposable cameras were the result of a phone call to boots; they get money when they send them back for recycling, but the phrase “I’m a physics student” and an offer to pay 10p a piece for them was enough to convince them to give me some. In addition I used a fan from a microwave to keep the whole thing cool, some spark plug cable to wire it all up and some assorted bits & bobs from the bits box and garage.

The MOT I’m using is out of an older microwave; modern ones tend to have much smaller cores and have the secondary bound to the core at one end. Using less iron in the core saves money at the cost of becoming magnetically saturated much more easily; this limits current and wastes power as heat in the core. Having 2 non-tied terminals allows the whole power supply to float (to a reasonable degree, the winding insulation will breakdown eventually). This one is capable of outputting up to about 500mA at 2100V.

The transformer output is fed into the following circuit. It is a pair of voltage doublers with opposite polarities stacked; each one outputs (in a perfect world, when not loaded) $2 \times \sqrt{2} \times V_{rms}\approx 6kV$. With both we have 12kV DC that can be zeroed either side.

High voltage ceramic capacitors do not have a high enough capacitance for this purpose, while buying high voltage electrolytics is prohibitively expensive; strings of camera flash capacitors in series can take the voltage required while being cheap and high capacitance. Ones salvaged from disposable cameras are usually rated for 80uF at 330V. I’ve soldered these together into strips of 5, which are then assembled into blocks of 20. 2 of these form the final smoothing capacitors, while the other one is centre-tapped in order to act as both of the first stage doubling capacitors. Each capacitor has a 390kOhm bleeder resistor across it, this allows the charge stored in these capacitors to decay when the power supply is turned off. Without these they would hold potentially lethal charge for days after use; with these resistors each capacitor will decay from 300V to 1.25V (for a total of 50V) after $390k\Omega \times 80\mu F \times \ln\left(\frac{300}{1.25}\right)\approx171s$. The power supply output is not safe to touch until 3 minutes after turning off.

In order to test the multiplier/rectifier circuitry I initially attached it to 240Vac with a multimeter connected to the output. As you can see it reads just over 1100Vdc; it is obviously working, output is 4.7x the rms voltage of the input and rectified to dc.

Now to test the output voltage when the MOT is included. Obviously I can’t just hook a multimeter up to the output, so what I’ve done is to string together 5 10MΩ in series to create a measured 51.3MΩ load and measured the current. Under this small load the voltage is $226\mu A \times 51.3M\Omega \approx 11.6kV$. I did manage to get over 13Kv a couple of days ago; I assume this is because the slightly damp air today is causing corona discharge somewhere. This is larger than the theoretical 12kV from earlier, either because my meter hasn’t been calibrated for about 20 years or because the MOT voltage is above the rated output when barely loaded. So far I have been unable to take any measurements under heavier load as I can’t find a suitable load; 100kΩ 1kW resistors are about £100, although it might be possible to build one for much less.

While the power supply is now operational, it could do with a few things doing with it. I’m still waiting on delivery of a 6A circuit breaker to cut power if the output is shorted. Under these conditions the HV output will give a large current surge; I tried using a 1kΩ 10W resistor to limit this, however I had an accidental spark over and blew this resistor up. It didn’t overheat a bit or burn out; it went bang and ended up in several pieces.

I’ve replaced it, this time before the capacitors rather than after; the capacitors will manage fine with sudden discharges (this is what they’re built for). This means that the resistor no longer has to take the 150+ Joules stored in the smoothing capacitors so it might survive if the circuit breaker triggers fast enough. If not, it’s cheaper than needing to replace a properly rated fuse. Giving the box a coat of paint is also on the to-do list.