Skip to content

High voltage supply using PC power supply transformers

Project description

Commercial desktop PC power supplies of ATX form factor most commonly operate at power levels around 400W, producing supply voltages required by the computer. In these switched mode power supplies, line isolation as well as most of the voltage conversion is provided by a ferrite power transformer, which will be called simply “ATX transformer” in the continuation.

PC power supplies are a frequent victim of hobbyst ‘recycling’ after their terminal failure. After useful components such as the cooling fan, capacitors, rectifiers, transistors and the TL494 are taken out, the transformers are usually among the components left behind. ATX transformers are usually rather difficult to take apart for rewinding, and rather uninteresting in their original form for the most people. Having several of them laying around for years, an idea sprang up to me. I became curious whether I could get some high voltage from them if I run them in reverse, and perhaps even put them to use as a vacuum tube power supply!

The idea is based on running the ATX transformer in reverse, from a mains powered half bridge of MOSFET’s or IGBT’s. The typical pinout of these transformers goes as follows:

ATX transformer pinout. Pins 1 and 2 are the mains side primary. Pin 5 is ground (the thick wire going out of the top of the transformer); pins 4 and 6 supply the 5V rectifier, while 3 and 7 supply the 12V rectifier

PC power supplies use half bridge forward converter, providing voltage of around 160V to the primary. The peak voltage that can be measured between pins 3 and 7 in that case is around 50V, meaning that the transformer has a minimum step-down ratio of around 3:1 (this can’t be exactly determined as it varies from case to case). The rectifier used is two-diode half wave type which results in rectified voltage of 25V, which is still more than twice as high as required 12V! The supply then regulates the voltage by reducing the duty cycle to required value. One may question this choice as being inefficient usage of the transformer core material (since energy is transferred only 50% of the time). Still, it appears that the designer has decided this is in some way optimal, because it allows for decrease in input filter capacitor size (which are generally more expensive than ferrite) as well as improvement in power factor. It is also interesting to note that most transformers don’t have a winding dedicated for 3.3V on them. 3.3V is produced from 5V windings using a special regulator based on a saturable reactor.

The power transformer is the large transformer in the center. The two smaller ones are auxiliary power transformer and base drive transformer for the power switches. Courtesy: Howstuffworks

The idea was as follows: Two large ATX supply power transformers were taken and their primaries (pins 1 and 2) connected in series. The outermost pins (3 and 7) of the secondaries were wired in parallel (remember that those can be identified because they go to 12V rail rectifier in the SMPS). The paralleled secondaries are now used as primaries and fed with my SG3525 flyback driver. The power switches used were HGTG30N60A4D IGBT’s – I replaced the mosfet’s with these because I expected this contraption to draw significantly more power than flybacks, and hence wanted to be sure that the switches can handle it. The transformers would saturate if they were driven at 3x their rated volts/turn at their usual frequency (50kHz), so a three times higher frequency – 150kHz, was chosen for operation.

With a configuration like this, it still wasn’t possible to draw arcs without tripping the overcurrent protection circuit, because the very low leakage inductance of the transformers wasn’t enough to limit the current to acceptable value. Hence I added some extra inductance into the primary circuit, in form of a ferrite cored inductor with a variable air gap for inductance adjustment.

The first light was quite disappointing – I was only able to draw small arcs, 2cm long at most. But it was actually impressive that I got even that much with less than 1kV which is not even officially high voltage for most jurisdictions! Still, I believed additional improvement could be made. So I built in a full wave voltage doubler onto the former output, consisting of 27nF, 1.6kV capacitors and strings of 3parallel x 3series UF4007 diodes for each doubler diode.

A voltage doubler is known as unsuitable for drawing arcs – because there is very little impedance in the discharge path, it tends to discharge in high current pulses that tend to damage the diodes. Some current limiting is required in order to form stable arcs – the simplest example is a large wattage resistor, which is costly and dissipates a lot of power as heat. The second possibility is to utilize a high voltage choke to smooth the output current. Fortunately, such chokes are actually quite available and cheap, in form of fluorescent tube and mercury vapor lamp ballast chokes! These chokes are iron cored, but seemed to work very well at 150kHz due to fact that they have been conducting mostly DC. In overall, I was actually surprised how well the choke ballasting worked to stabilize the output.

Schematic portraying the final configuration

Upon switching to doubler arrangement, the improvement was immediately obvious, allowing drawing of 6-7cm long, extremely bright arcs. After a while of playing with arcs, the original ferrite ballast inductor demonstrated it’s not up to this task, and melted. Since I wanted more power anyway, I replaced the inductor with a simple air cored inductor consisting of about 10 turns of thick PVC insulated copper wire around 50mm form. I estimated it’s inductance to be somewhere between 5 and 10 uH; the goal was to tune the short circuit current to stay just beyond the lower bound of my overcurrent protection circuit. I attained the fine tuning by inserting a ferrite rod into the coil.

The final results were quite spectacular, with arcs reaching over 12cm in length, and drawing in excess of 2kVA peak input power. I attempted to get an even further increase by using a voltage quadrupler instead of a doubler: this only helped a little, increasing the arc length to about 15cm max. When arcs are stretched this long, they are quite unstable and need to be drawn carefully. They are best drawn vertically upwards because this way heat convection helps keeping the arc hot and ionized, and allows it to stretch.

The arcs start only at some 1-2mm electrode separation, indicating the voltage is still rather low for HV enthusiast standards. This demonstrated that extreme HV is not a requirement for producing large arcs – in fact, as little as 50V may sometimes prove enough (arc welding!). Here is a video demonstrating the drawing of some of the longest arcs from this setup:

The transformers surprisingly stayed cool despite being ran at 3x rated volts/turn and no insulation failures were observed. If the center point between two transformer secondaries is grounded, the transformer insulation will only ever see 450V peak which is too low to produce any potentially insulation-damaging corona, and is actually within rating of most quality magnet wire and transformer insulation standards – yet, the arcs produced surpass those of any flyback transformer I’ve ever tested to destruction.

If a DC power supply for applications such as powering vacuum tube circuits is desired, the filter inductor could be left out and larger value electrolytic capacitors could be used for the voltage multiplier.


It has been shown that high voltage transformers with prohibitive insulation demands are not necessary to produce fairly large arcs, as well as almost arbitrarily high voltages provided an adequate voltage multiplier is used. Of course, the cost may not always be optimal compared to specifically designed HV transformer, but this is offset by parts usually being available to hobbysts at very little cost. The robustness of the transformers even under heavy loads is remarkable – they’ve seen little heating even with prolonged operation.

Links and references

  1. Did you consider using only a portion of the secondary winding as the input? Seems that might give you a higher turns ratio and save you from needing two transformers? Say: use the 12V winding as input and get 170/12 = 14:1 step up

    • mbakula permalink


      The things are actually a bit more complicated here – measurements have shown that these transformers usually run at about 50% duty cycle at rated PSU output voltage to allow good regulation, reduce the size of filter caps and improve power factor. Hence at 100% duty cycle the voltage is around 24V on one end of 12V side and double that for both ends. I generally round up the voltages to 50V sec and 150V pri so the ratio rounds up to about 3:1.

      The transformers are already being ran at 3x their voltage ratings (I had to increase the frequency about 3 times accordingly) which is quite hard on them, as well as on switching devices and multiplier diodes. I could have used even more insane frequency and only one 12V or 5V winding, but I doubt that the transformer would survive that since it is not designed for HV, and I’ve already had my multiplier diodes blow up even at lower frequencies. If you have some good ultrafast high voltage diodes rated for 10A or so you could perhaps try this yourself, though I’d expect a transformer fire there.


  2. Kenneth permalink

    That looks sweet!

    I’m actually having this idea in my mind to use an ATX transformer for the heater supply in my next vacuum tube build; it needs to deliver up to 100 Watts of heater power.
    My current design is some sort of semi switch-mode supply, that takes rectified mains through an inverter (100-150 kHz, far from the audible spectrum), and uses the ATX transformer to step it down to 12.6 VAC for the tubes. It might even incorporate some sort of regulation by means of PWM feedback.

    What do you think, any chance of this working? Has anyone ever tried this before?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: