LED chain powered by 'Joule Thief'

Rationale

Some years ago a friend of mine asked me whether I could help her power some LED chainlights she bought using batteries, so that she could take the lights with her to M'era Luna Festival, creating nice atmosphere in the group tent. As she was already taking a 12 V car battery to the festival, and the chain light was powered from mains using a 24 V wall plug transformer, I simply suggested to use another 12 V battery in series to the one she already got to obtain the 24 V necessary for lighting the chain. This setup has been working since, but I always thought of it as a rather inelegant solution, as car batteries tend to be quite heavy.

Having played around with a MC34063A switching-mode power supply controller IC, I thought about using a circuit designed around that IC for generating 24 V out of a single 12 V car battery (which my friend will take with her anyways).

As M'era Luna is still some weeks away, I have not built anything yet. But for the Feuertanz Festival I was going to attend I wanted to build some kind of LED throwie to place at my tent pegs so that at night people won't stumble over them and the ropes attached to them.

Yet, LED throwies are quite resource-intensive in my opinion; you need a small lithium cell with 3 V to light a LED, and afterwards it's junk. Even at a cost of only 0.20 € ten of these things already cost 2 €, and I figured they'd only last one night, given a capacity of 170 mAh and a drain of 20 mA for a single LED^[1]. Besides the cost of 6 € for three nights of festival, I'd waste precious material for (more or less) nothing – an ecological nightmare. Using 2 or more AA/AAA cells, possibly rechargeable ones, might work – but I'd need at least 10 cells for just 5 lights.

So I searched the web for ideas to power a LED from a single AA/AAA cell, and I found something called 'joule thief'.

What is a 'joule thief'?

With the circuit being explained over at wikipedia, I just cover the basic idea here. Take a look at the schematics (taken from evilmadscientist.com for its clarity):

That's basically a blocking oscillator using the stray capacitance of the transistor and the inductance of the transformer to swing. With the transistor in ON-state, the battery is connected to the right winding, causing a magnetic field to build up. When the transistor opens, the magnetic field collapses, generating a voltage spike. Feeding that spike to an LED makes it light up, and as that on-off-switching is repeated very fast, the human eye won't notice that the LED is in fact flickering and will perceive a steady light.

I built a quick-and-dirty version of such circuit to verify it's working; I added a 20 nF capacitor in parallel to the 1 kΩ base resistor to force oscillation (even when measurement probes are connected to the circuit):

Then I hooked up my old oscilloscope to have a look at what's going on:

As you can possibly see, that thing is oscillating very fast; in the above picture, there were only 0.5 μs between the DIVs, and it took only 0.6 DIVs from peak to peak. This results in a frequency of about 3 MHz! That's unnecessarily high for practical use, as even 30 Hz (= 0.00003 MHz) are too fast for the human eye to notice, but as that circuit just serves demonstrational purposes I don't care :)

To take a closer look at what's happening, I pushed the 10x-button for the oscilloscope's time scale, adjusted the vertical diversion for the two measurement channels und took a screenshot of the resulting waveforms with my camera; the green one is V_CE (at 2 V/DIV), the purple one V_BE (at 0.2 V/DIV) measured at the transistor:

With the zero reference line for V_CE being the center horizontal axis, V_CE reaches about +4 V (enough to light the LED). The zero line for V_BE is the axis in the lower half of the screen, so that with 0.2 V/DIV the voltage V_BE oscillates around 0.6 V, switching the transistor on and off.

Practical use

Such circuit can be made quite small, so when I found a LED tealight at a local store for 1 € including three AG13 batteries with a built-in switch, I immediately thought "nice case!" and bought two of them to try fitting a joule thief into them. Here's the result, with a regular tealight right next to it (click on the picture for a larger view):

As the modified tealight only needs one battery instead of three, I put in a spring to bridge the gap between the battery and the holder tap:

It was fun building one of these, but I ran out of enameled wire (the one I had was salvaged from a broken radio), so I had to stop here and gave up building 10 of these to place around my tent at the festival. But while looking around the Internet for joule thief applications and stuff, I found a video on youtube showing someone using a joule thief to light a LED chain with 50 LEDs. That reminded me of my friend and her LED chainlight, so I decided to give the blocking oscillator a chance, because output voltage stability is not critical for that application – it's not that important to get exactly 24 V, approximately 24 V is absolutely okay for decorative lighting – and the circuit is far more simplistic than one using above-mentioned MC34063A IC. You just need about four parts, put them together (even wildly, without a PCB), and it should work.

The LED chain light

So after I came back from the festival, I ordered LED christmas lights from Amazon (turns out christmas lights are hard to come by in retail stores in June): 64 LEDs arranged in a 8x8 net spanning 2x2 m². I chose a net that's powered by a mains transformer delivering 24 V, so that when I successfully build a device to power my LED chain light I can easily build another one for my friend to power her chain light, too.

That's the transformer of the christmas lights:

And that's what the lights look like when powered with the provided transformer:

As you can see, they are giving nice warm yellow-ish light ♥ instead of ugly blue-ish light; I kept the net bundled together as it is simpler to handle that way during playing around with it.

Modified joule thief

The next step was to build a circuit based on the original one above to power the lights. With the LEDs needing about 24 V to work I couldn't use the collector-emitter-approach with the transistor I have at hand (BC368), because the 24 V exceed its maximum permissible voltage of 20 V for V_CE – I would fry it using that circuit without modifications. I could use another transistor, but I wanted to try another idea: wind another coil onto the transformer with more turns, thereby increasing the output voltage, like in a video showing a CFL being powered by a joule thief.

So after I acquired more enameled wire, I wound a new transformer (I hate doing that. I'm just not patient enough, so the result usually looks ugly and performs not too well) with two coils having the same number of turns und one coil with many more turns (I guess I ended up with only about twice as many turns; knowing that this is at the edge of being not enough turns I just hoped for the best. I didn't want to wind another coil...)

I hooked up my new transformer to the existing circuit, removing the single white LED as it is not needed any longer. I also put a normal rectifier diode between the base and the emitter of the transistor, with its cathode connected to the base. That's a precaution, because the transistor's datasheet states a maximum of -5 V for V_BE, which might be exceeded, thereby damaging the transistor. The diode keeps V_BE at about -0.7 V at minimum, so the transistor may live long and prosper. This is what the test setup looked like:

I placed one of the oscilloscope's probes at the base of the transistor to see whether the oscillator is really working because now there is no white LED that gives visual feedback, and the other probe at the output of the transformer's third coil (which was also connected to the LED christmas lights) to see how much voltage I could get in case the LED chain wouldn't light up. I connected the battery, and behold!

It worked out-of-the-box :) Seems I was lucky today... Yet, the LEDs weren't too bright, I'd say about one third as bright as when operated with the original wall plug transformer. A quick look at the oscilloscope told me the reason for this: the third coil's output voltage peaked at about 26 V, so for one thing the voltage was a little low, and for another thing most of the time the LEDs were in off state – which wouldn't be that tragic as we know from the simple single LED experiment done at the beginning, but for that bad duty cycle being unnoticeable I'd need higher voltage to run the LED chain.

Optimization

So, the next step was to wind a new transformer (*sigh*), with a higer turns ratio for the third coil compared to the two other coils of the blocking oscillator. This generates higher output voltages for the LED lights, thereby making them brighter.

As I was going to take the final circuit with me to a festival, I needed to think about how to make everything more robust compared to the initial prototype. So I decided to make a PCB to mount the components on. I used the free lite version of EAGLE to draw the schematic:

After that I continued drawing the PCB layout (when clicking the image, you get a PNG file that, when printed at 600 DPI, can be used to create your own PCB):

As you can see in the next picture, most of the space is reserved for the transformer, which is placed in the middle of the PCB. The four vertically placed holes around it can be used to fix the toroid onto the PCB using cable ties or plain wire:

After the usual processes like printing the layout, exposing the base material, etching etc., you get a neat PCB like this:

After that, I drilled the holes, mounted the components and soldered them onto the board:

Measurements with the oscilloscope at the LED output terminals showed swings when connected to the LED chain; this is because with the rectifier built into the cord current is allowed to flow in both directions, basically forming an RLC circuit that can oscillate. Therefore I removed the rectifier from the power cord, and the swings disappeared.

Even at daylight one can see the LEDs shine. Doubling the input voltage to about 3 V makes them light really bright, but then the current increases (the primary coil's ohmic resistance R remains the same, yet the voltage U across it doubles, so the current I doubles, which in turn yields much higher power dissipation P_T = R_CE · I² in the transistor) and the transistor gets warm.

Conclusion

For decorative lighting, the joule thief I built is sufficient while being cheap to build and operate (just use "dead" batteries recovered from recycling bins at supermarkets), but for exhaustive illumination of a tent you either require higher input voltage (and, therefore, probably another type of transistor) or a transformer that's better designed than my "let's use a toroid I salvaged from a dead PC PSU and do some wiring around it". So for my purpose that's OK, but I guess for my friend's LED chains I'll be needing more "serious stuff" like a plain voltage doubler she'll attach to her 12 V car battery.

Nevertheless, it was interesting and fun building such a device, and I hope the result will make a nice addendum to my tent at the festival next weekend :) - at least I think it will look nice:

---

Annotations

[1] In the three weeks I've been writing this article I learned that you can't drain such a small battery with 20 mA, it will only supply far less current, so a LED throwie will last longer than just one night.