Experimenting with the Joule Thief circuit

Out of boredom I got interested into the Joule Thief circuit.

Yesterday I built two of them using a 2N3904 transistor, 1k resistor, 10mm warm white LED and two different handwound coils.

One of the coils is a larger toroid with 10 turns on it, the other one is a smaller one with six turns. So far both of them work just fine. The current draw from an eneloop cell is 40 mA with the larger and 44 mA with the smaller coil. I will make a third one tomorrow with an even smaller toroid with only four turns and see if that works. I would also like to try if they can power two LEDs in series.

Will provide pictures and more details tomorrow.

Joule Thief - Jewel Thief.

Named as such because as the battery runs down the circuit alows the LED to draw the current that is needs (stealing energy (joules) from an otherwise running down battery.

I’ve been playing when them too, and found that if you have more turns on the coil going to the BJT’s base, you can run the battery to a lower voltage, I’ve been able to push it down to a bit more then 0.2v before the LED stopped glowing (~20 turns on the LED, 100+ turns to the base, 2N2222 BJT). I’ve tried increasing the current by decreasing the resistor to the base, but my BJT got pretty hot so I turned up the resistance again (was using a potentiometer). I recently got some 32 gauge magnet wire and I’ve been using those, also used 34 gauge too but I ran out of the little that I had so I’ve got to get more. I tried ~4 turns but it didn’t work well, but that also depends on the toroid, since some have higher micro/nano Henrys per winding then others.

I bought a few Pulse common mode toroid inductors and those worked pretty well too (but I could have fit more windings on the same coil using thinner wire). Also tried some small common mode inductors that had ~7-8 turns on them and they worked well.

Haven’t tried LEDs in series yet but I heard it should work with 2 LEDs in series. Which wires do you use? Also, does anyone know how to increase or decrease the current to the LED? I’d like to make one to power a high power LED at ~50-75 mA average from a 1.5v cell and another to drive a SMD LED at really low currents, maybe 500-50µA.

A warning about using Joule Thiefs. Draining down alkalines all the way with a JT circuit increases the risk of leakage. I found this out the hard way. I had a large stash of half drained alkies and built JT circuits specifically to make use of those cells. I used them for bathroom night lights and thought they were great until I noticed that about one out of three cells began to leak after being drained down. These cells were all fine until I drained them in the JT. I'm not sure of the exact cause and effect involved but it seems like alkalines are not built to be drained down to the levels that JT circuits are capable of.

yup, that’s why mine sits on my breadboard. The battery sits in a battery holder instead of in a tube and makes clean up easier. The only jould thief light I have has a removable head and tailcap, so I can hammer out the battery if necessary.

I think the leaking is has something to do with gas produced as a battery discharges.

The wire I used for the coils is a 0,14 mm² stranded wire.

With the small four-turn coil the LED doesn't light up at all. Apart from too few windings there is also the possibility that the toroid I used this time is made from a different material that is unsuitable for this purpose. I got it from an old mainboard and it's greenish on one side. The other ones I bought at a local electronics store. Unfortunately I have no DMM that can measure inductivity.

These are my toroids. The one that doesn't work is on the right.

Here are the two curcuits that work:

I just did some testing with 1, 2 and 3 LEDs and 1 and 2 NiMH cells. Here are the results:

Number of LEDs Number of cells Input current (mA) Output current (mA)
1 1 44.0 8.7
2 1 47.0 3.5
3 1 48.0 1.9
1 2 113.0 1.2
2 2 83.6 13.6
3 2 87.5 7.5

I'm not sure what happens with one LED and two cells. Maybe the circuit just becomes extremely inefficient if the input voltage is too high compared to the output load?

Those are large toroids compared to what I use (~5mm diameter), but you do use thicker wire. 32 gauge wire = 0.2 mm diameter. 34 gauge = 0.16 mm.

I’m didn’t realize how inefficient the joule thief is, 53 mw input, and 28 mw output. But a DMM might not be as accurate for these measurements, since the output is not DC. Might have to break out the oscilloscope.

I’ll see if I can get some Mosfet to try, maybe even some Jfets and see if I can eliminate that resistor and/or get it to work down to a lower voltage. Do you have a potentiometer? You could try different resistances and see how that affects power output and efficiency.

I just used the thinnest wire I had lying around, maybe I'll try some enamel copper wire.

Could be that my DMM gives false readings to a certain degree. An oscilloscope would indeed be a nice thing to have

By the way, I also tried to measure output voltage across the LED, but my DMM only returned the input voltage at that point.

Another thing I tried, someone said that a 10nF capacitor would improve the efficiency by 300%:
(found on http://www.electro-tech-online.com/electronic-projects-design-ideas-reviews/104554-joule-thief-not-working-correctly.html )

That circuit is slightly different than mine (I've got the resistor between the coil and the transistor base). I tried the 10nF capacitor between + and - and the input current went down by only 1.5 mA without affecting LED brightness.

Yup, I saw that as well, but it seems to make the LED dimmer at lower voltages. Could be my circuit, but with the battery voltage at ~0.55v, the “stock” joule thief drew ~3-4 mA, but when I added the 10 nF cap, it went down to ~1.5 mA, and the LED got dimmer.

I keep that battery around to find the shutdown voltage of the various Joule thieves that I made. The transistor does affect how low the circuit runs to.

i would like to build a kit like this ,will this : http://www.ebay.com/itm/30-x-2N3904-NPN-General-Propose-Transistor-/250853369809?pt=LH_DefaultDomain_0&hash=item3a6806a7d1 do the job?

These should work fine, nofearek9.

There's lots of information on this site, including a more elaborate version with constant brightness regulation:

niec info thanks,that PR4401 chips looks very easy solution.

Ive had good success with the zxld381 chip. It is a bit on the small side (2x3x1mm) but is readily available.

I have used them to modify a couple of Mag solitaires with xpg’s, not the brightest of lights, but very usable with night adjusted eyes.


Yup, I’ve used something similar to that too (zxld380). They work well and can be really small if used with a small inductor. Should also works fine with 2 cells.

Has anyone used joule thief circuitry with a lithium primary?

Steve_the_Chief did some testing with 2 NiMh cells above in post 6, but the circuit seems inefficient with only 1 LED.

Edit: End of quote here. Quote function acting weird.

Last weekend I built another Joule Thief, using the same parts but a different toroid. A very large one actually. Yes, it looks a little bit ridiculous.

I repeated my series of measurements using the same DMM and the same LEDs:

Number of LEDs Number of cells Input current (mA) Output current (mA)
1 1 28.7 7.5
2 1 29.5 4.5
3 1 35.3 2.6
1 2 115.2 23.6
2 2 58.0 14.5
3 2 61.6 9.7

The large toroid improved the circuit's efficiency considerably!

BUT the circuit emits an audible whine now, which is kind of annoying. However, it is quiet in the 1 LED / 2 cells configuration... strange.

I am lucky to have a few 2SD965 lowVce(sat), high hFE transistors. I built this solar rechargeable light using the Joule Thief Circuit.

I really like the look of that. I thought I was doing good following this thread till you said that. After 20 mins of googling, am I right to say that this transistor would act like a booster?