Maglite Joule thief

Joule thief circuit

The Joule thief is a really fascinating circuit, simple yet very intricate. Basically, it’s a step-up converted in its most elementary expression. I will spare you the theory since there is plenty of information on it on the web; is a good place to start.

Joule thieves in all sorts of forms have been featured countless time on DIY websites and I felt it was time I build one. However, I did not want to leave the circuit at the breadboard stage because as it stands, the joule thief has characteristics that make it very attractive for all sorts of low power applications and I figured a flash light would be a very good home for a joule thief, where having the option of using dead batteries is certainly a big plus not to mention using less cells because the circuit steps the voltage up. Why dead batteries? Because a battery is never really dead, its voltage just falls down logarithmically until it hits a point where the device it was powering up stops functioning, which does not mean the battery is totally drained but rather that its voltage has fallen below a usable level. Since joule thieves are step-up converters, they can take that “dead” battery, and give it a new life by stepping up its output voltage to usable levels again.

The build

A 2 AA batteries Maglite

For my flashlight, I opted for a maglite body for its sturdiness and simplicity. I have been using those for years and they have served me well, but with traditional incandescent lamp bulbs (I do know they make LED versions now), they eat through batteries like crazy. So the challenge was to convert a 2 AA battery maglite so it could run off a joule thief circuit and a single AA but could easily get converted back to using a lamps(As I will tell later, the joule thief’s light output is not so strong … sufficient in most cases but not strong).

As I would be using one less battery, the trick was to use that space to hold the circuitry. I proceeded to cut a perfboard the size of an AA and soldered all the components on it with the heads from two nails as connectors. Inductors being already pretty hard to come around, one tailored to this application would be next to impossible to find so I had to hand wind one using 20 or so turns of 40 awg enameled wire (almost hair thin (also hard to find, look for it on ebay)) around a ferrite core to build the joule thief’s coil. Once everything was in place, I soldered the circuit ground wire, which when making contact with the flashlight’s body, would turn it on or off. That wire had to be routed inside the plastic insert that normally holds the lamp and its metal pad in place so that when you turn the head of the maglite, it screws up and presses the pad against the body, thus closing the circuit. In order to allow the maglite to be converted back to using a lamp, I just cut a notch under that pad so that pressure on it would contact the wire and ground it.

For protection and isolation, the circuit was wrapped it with acrylic tubing (some leftovers from the time my computer was watercooled) and inserted in the body. Finally, I installed the LED at the top, with its two pins bent to fit in the holes normally meant for the incandescent bulb. The lens fits perfectly on it; the only way to tell it’s a modified maglite is to look at the bulb.

The circuit being constructed.
The circuit being constructed.
Testing the circuit.
Testing the circuit.
Oscilloscope trace of a correctly working joule thief.
A correctly working joule thief.
The finished circuit in its protectective acrylic tubing.
The finished circuit in its protectective acrylic tubing.
The finished product.
It works!


I will right away admit that I am a bit dissapointed with the light output of the circuit. Though I did expect it to be a whole lot less than the incandescent bulb, it is barely usable. The culprit is certainly the LED. At only 3mm, it can only do so much with that waveform going through it ; it’s rated for 20mA and its getting 12 so the circuit is doing a correct job keeping in mind that joule thieves are quite inefficient ( in the order of 30-40% judging by the duty cycle). Using a larger inductance is out of question because it reduces the frequency without modifying the waveform but using a larger wire gauge (thus lowering the resistance of the coil, see below) would probably help. What also does appear to make a difference is the type of transistor used, I noticed that the current draw of the LED was only 9mA with a 2N3904 while it jumped to 12 with a PN2222A. Both are general purpose NPN so maybe another type of transitor would do better. As a side note, the circuit will not work with FETs, I have found plans to build a joule thief with those but its much more complicated.

With a single battery at 1.435V, I got two days of continuous lighting, not bad. At that voltage, the current draw is about 65mA. I was not able to measure the pull of the standard incadescent bulb for comparison because the inline resistance of my multimeter was too consequent, but one interesting thing I noticed was that below a certain voltage, the light would start to flicker at hertz or so. Its hard to see in the picture, but I added an electrolytic capacitor for bypass; it could have something to do with that.

Even though I said that the light output was nothing to brag about, I did take the flashlights to many trips in the woods,  with some lasting a few days and it has held up perfectly. With your eyesight accustomed to the dark a bit, you can see at a few meters  and whatever task your hands are doing is lit well enough for comfort. With all this serious usage, I have not yet ran out that dead battery. Too bad, I wish I could have gone out asking my hiking buddies for dead batteries.

With LED flashlights being quite effective, I can’t vouch for the potential of this circuit for such applications. Certainly, using dead batteries is a plus as they are relatively easy to come by, but the low light output would certainly be a killer for most of us because when it comes to photons during a moonless night in the woods, more is just better.

Still a very fun build…

On’s post about my joule thief

According to rustybolt (thanks), my coil is mainly to blame for my poor performing Joule thief. The wire used is too thin and has a consequence opposes too great of a resistance ( I should have tought about that) to the current, thereby limiting the LED’s brightness. There is a very clear article on Joule Thief coil selection on his site.   He also points to the transistor being responsible for the loss of efficiency. Next time I get my hand on that maglite (I’m travelling right now so it’s an ocean apart from me), I’ll revisit the circuit.