Joule Thief Filament Lamp

Intro: Joule Thief Filament Lamp

LED filaments are thin, stick like LEDs. They are used in many Edison bulb look-a-like LED bulbs.Each of those thin sticks contain many - 20 to 30 LEDs connected in series. So they are quite bright and energy efficient, however typically require over 70V to turn on.I wanted to make a small, battery operated lamp using those LED filaments. After trying out many circuits, I discovered a very simple circuit to light a filament LED with only one, 1.5V battery.

STEP 1: Circuit Designs

I knew I should be able to use boost converter to boost the low voltage from a battery, however didn't think I can easily get the 70V. I have tried using specialized boost LED driver ICs with success, but those ICs need 3V or higher to operate.Next I experimented with a simple, two transistor Joule Thief (blocking oscillator) circuit. I did not have a transistor that can withstand 70V, so I used charge pump circuit to double the output of the Joule Thief. This way the transistor is exposed to only half the final output, or 35V.

This circuit worked, and I was happy with the performance for a while, but still wanted to reduce the component count. So I obtained a few transistors that can handle over 70V of voltage, and tried to see if I can light filament LED with just a Joule Thief. After some tweaking of the component values I found the circuit to work just as well as the charge pump assisted Joule Thief circuit!

STEP 2: Final Circuit

So here is the final circuit. It is deceivingly simple, but works as well as the previous version with many more components.

The key is using transistors that can handle high enough voltage. I used KSP06, which has Vceo of 80V, just high enough for this project. Other specs such as hfe and Vbe is still good enough to operate at low supply voltage.

I tuned the components to not draw too much current, since the power source is AAA battery which is small. You can adjust R1, R2, as well as C1 to draw more current and light LED with more power if you want. For example R1: 470 ohm, R2: 47k ohm, and C1: 22pF would produce higher output, but the battery will drain much quicker.

STEP 3: Final Touch

I designed a PCB to fit into a glass test tube.

It uses a single AAA battery (alkaline or NiMH) and draws about 50 mA.

I also added a tilt switch to turn on the LED when the unit is standing upright, and turn off when resting. I inserted the unit into the tube to make it look a bit like a vintage tube.

I put together the PCB and components as an easy to assemble kit - available at my website: https://www.theledart.com/products/jt-filament - if you are interested.

Comments

[DavidM638]

There is a 300V Transistor A42 NPN hv>300V >500MA EBC. I got mine from AliExpress. I opted for an assortment, and A42 was included. A92 is also 300V, but is PNP.

[DavidM638]

I built your circuit with the A42 transistors as close as I could get with available parts, but it still won't go much above 70 Volts, even if the transistors are rated at 300 Volts.

[ledartist]

I actually tested the circuit with MPSA42 (as well as a handful of candidates), and the result was not good. There's more to transistors than just the breakdown voltage - other parameters affect switching performance, I guess.

[squeakbat]

Very cool circuit. (Availability of low-voltage filaments does not make this any less fun.)

Anyway, while playing with a prototype, I noticed that Vb on Q1 drops well below the breakdown voltage (or would except the junction breaks down, in my case, at around -10V).

Have you noticed transistor failures? Mine seems to be fine after 10s of hours of use.

Adding a limit diode to ground doesn't seem to affect performance, so that's probably what I would go with.

Thanks!

[ledartist]

I know that Q1 Vb drops negative, which is caused by the charge from C1 when Q2 turns on. Since C1's capacity is so small it doesn't seem to stress Q1 any bit. I have many units running for many years, yet to see any failures.
Of course you can add a diode as you wish.

[squeakbat]

Thanks for answering! I was basically noticing on the scope that Vb was clipped. Typical Vebo is like 6 volts and C1 is charged to the full output voltage, so it made me a little uncomfortable. But I agree that there's so little charge involved so it's probably fine.

Thanks!

[THX 1138]

Fortunately these filaments are now available in 12 V and 3 V versions (and some more colours and lose so no need to break open glass bulbs to obtain them). I'm struggling to find inspiration for projects though!

[AnJo888]

Nice PCB/project.
What about using a PET test tube (bottle preform)? It has a watertight cap and is unbreakable (great for outdoors activities)... Not sure about the heat, though...

[ledartist]

I don't think the heat would be a problem. The circuit doesn't produce much heat at all.

[AnJo888]

Thanks for the answer.
That's great... The tube format would be different, os course, probably wide enough to use a pair of AAA in parallel (longer duration), but would be, virtually, indestructible.

[craftyjam]

is the inductor value crucial?

[ledartist]

No really. I tried values between 47uH to 100uH and see only a small difference. Values up to 470uH or so should work ok.

[luemmel]

Very cool, is it possible to add a potentiometer to dimming it?

[phil46]

Very cool. I love the addition of a tilt sensor to turn on/off, Why don't they do that simple trick with commercial lights to eliminate the manual on/off switch? Brilliant!

[Victor805]

Those look amazing. The PCBs are also really neat.

I've checked your webpage and your other projects are equally as great.

[ledartist]

Well thank you very much!

[Bongmaster]

oh nice :) have you considered selling this as a kit? :)

[ledartist]

Yes! It is available at my website: https://www.theledart.com/products/jt-filament

Thanks!

[radmans]

that's some very clever engineering and design! will you be sharing the pcb design also?