HV Joule Thief

High Voltage Joule Thief to light a Neon Glow Lamp

Hackaday had a cool post the other day with suedbunker hacking an "I can solder" LED joule thief PCB by replacing the LED with a neon glow lamp and something about it really piqued my interest.


I love joule thieves. I just do. What the hell's a joule thief? It's a basic circuit design called more formally a blocking oscillator -- oscillator because it makes electronic waves, blocking because it uses a transistor to block current flow through its collector-emitter path, release the current, block it again, on and on...

Well-known hardware hacker and tinkerer BigClive gave the circuit the clever name "Joule Thief" years ago because it can be powered by a "dead" or dying 1.5V battery and steal the last bits of energy from it after other more finicky electronics have refused to deal with such a run-down, no-good battery that refuses to even muster 1.5 volts.

Joule Thief circuit image from Wikipedia entry, Rowland.

By using an inductor and through the magic of Maxwell's equations of electromagnetics, the transistor's constant teasing can transform the less-than-1.5V level of our cast-aside battery into 10, 20 or even 30 volts in fits and spurts.

But nothing's free of course, so we're trading a steady flow of power at a 1.2V level for an oscillating flow of power at a higher voltage, but lower current. But this is still enough to get a white led that demands at least 3.5V to light in pulses, at an eye-deceiving rate of several tens if kilohertz.

So the basic joule thief will light a white led at a few volts, but can we boost that voltage even higher? Say, up into the range of 70 to 90V, to light a vintage neon gas-filled glow lamp?

That's what suedbunker addressed in the project, and the answer is yes. Yes, we can boost the voltage that high.

The way he does it is to modify the typical joule thief inductor construction from, say a 20 turn primary coil, 20 turn feedback coil into one designed to boost voltage much higher, 80 turns for the primary coil feeding current down the collector-emitter path, and 10 turns for the feedback coil into the transistor's base to block and allow collector-emitter flow.

He then used a diode and capacitor to rectify the collector voltage from the AC spikes into a smooth DC voltage high enough to ignite a neon lamp.

It's a cool project and a clever hack for the "I can solder" badges, which in his case became "I can hack".

One minor drawback of DC, however, is that it will only ignite one electrode inside the neon lamp -- interestingly the negative one -- since the electrons are actually blasting from negative through the ionized neon gas to the positive terminal.

He notes in his project that it was inspired by an Alan's Lab post about a high voltage joule thief from back in 2003, that actually uses 3 coils to make a sort of joule thief/flyback transformer mashup. Alan used a xenon flash tube trigger transformer that already has a high voltage coil on it made from very fine enameled wire. He removed the other coil laid over top of that and replaced it with his own 2 coil primary/feedback pair of inductors wrapped over the top of the existing high voltage coil. This implements the typical joule thief single-transistor blocking oscillator but will have the nice additional effect of inducing a high voltage (but lower current) signal in the HV coil. That third coil can be connected to the neon glow lamp and light it with an AC current so that both posts inside the bulb ignite and glow in that beautiful red-orange of excited neon gas.

Alan mentions that it needed two batteries (3V) to start and boost it up to the voltage required to ignite both electrodes in the lamp.

I wanted to see if I could pull it off with a single "dead" 1.5V AAA battery in the conventional art of joule thievery.

I also wanted to use a toroid that I would wind myself. Because toroids are cool, let's face it. What's cooler than a magical donut with wire coiled around it, looking like some Ironman arc reactor or something!?


toroid used (14mm OD 8mm ID 7mm height)

So start with a little green toroid from eBay. No specs, but after some experimentation seems to have an AL value of roughly 4000 nanohenries per turns squared. That means a single turn coil would have an inductance of 4000 nanohenries, or 4 microhenries. 2 turns would be 2^2=4*4000 or 16000 nanohenries, i.e. 16 microhenries. etc.

After an initial, sloppy attempt and hours of winding while watching TV (some people knit, some people wind toroids) I started a new toroid, carefully winding 34 gauge enameled wire along the donut, with no overlapping. I coated it with polyurethane (aka gorilla glue). After drying, wind another layer, more glue, and finally a third layer, more glue. In the end, I estimate I got about 220 turns in those 3 layers.

Then comes the traditional joule thief primary & feedback coils. Here I wasn't sure how many windings to go with. It's a trade-off. More turns gets better magnetic flux and sharper working of the circuit, maybe. But more turns means a lower ratio of HV coil to primary, and a lower voltage boost. If we can reduce the number of primary and feedback turns, we'll get a higher HV to primary turns ratio and a higher voltage boost. But a lower number of primary turns may lead to higher collector current in our transistor. So we have to make sure the transistor can handle it.

And which transistor should we use? Good question. With this 3-coil system, we won't get quite the high voltage in the primary that we will in the "secondary" (the high voltage coil itself), but it can still get quite high spikes, perhaps in the tens of volts.

In the end (or for now) I chose the trusty 2n2222 transistor, which can handle up to 40V and repeated currents of up to 600mA. The 2n2222 however needs some care with the emitter to base voltage -- the datasheet states it shouldn't go over 6V. So I've added a 4.7 voltage zener diode from ground to the transistor base. This will block any voltage spikes when the base tries to go below -0.6V with respect to ground, and above about 4.7V in normal positive base operation.

For the feedback path through the transistor base, resistor ranges go from a few K ohms to tens of kiloohms. In my case, since I require some pretty high currents to boost the HV coil output high enough to ignite the neon lamp, I chose a small resistor of 10 ohms, but in series with a 10K variable resistor so that the joule thief can be adjusted.


First I breadboarded the circuit and found after much experimentation that the lamp would indeed light both lamp electrodes with AC voltage when the primary and feedback coils were brought down to 2 turns each. That's a low turn count, but it works and perhaps the high permeability of the toroid ferrite material combined with the very low flux leakage of toroids in general, meant I still get a fair amount of oomph with only 2 turns each.


I built the circuit onto an Altoids tin because they're just cool. Only the battery holder is inside, because the circuit itself is kind of neat looking with everything exposed on the lid. It also sports a handmade knife switch for extra Boris Karloff effect.

Here's the final schematic.

And here's a short video clip tour.