From Gemini:
The following guidelines break down how to tune and adjust each component:
1. The Core & Windings (The Transformer)
* The 1:1 Baseline: For standard builds, use a bifilar winding (twisting two wires together and wrapping them simultaneously). Aim for 10 to 15 turns around a small ferrite toroid core.
* More Turns vs. Fewer Turns:
- Adding more loops increases inductance and reduces the operating frequency, which typically improves efficiency and lowers battery drain.
- If you need a significantly higher voltage boost (e.g., driving an array of LEDs), move to an asymmetric ratio like 1:5 or 1:10 (primary to feedback).
* Wire Thickness: Use thicker magnet wire (e.g., 24–26 AWG) for the primary winding to reduce DC resistance and limit current loss.
* Salvage Tip: Excellent small toroidal cores can be harvested from dead compact fluorescent lamp (CFL) bulbs or old computer power supplies.
2. The Resistor (R₁)
* The 1kΩ Standard: A 1kΩ resistor is the universal starting point for a 1.5V input.
* Increasing for Efficiency: If your transistor is running hot or you want the battery to last as long as possible, increase the resistor up to 2.2kΩ or 6.8kΩ. Higher resistance drops the input current draw, stretching out battery life at the cost of a slightly dimmer LED.
* Lowering for Low Voltages: If you want the circuit to keep stealing energy down to ultra-low thresholds (below 0.5V), you can drop the resistor below 1kΩ (even down to a few ohms), but monitor the transistor so it doesn't overheat.
3. Transistor Selection
* Key Specs: Look for general-purpose NPN BJTs with high DC current gain (h_FE) and a low collector-emitter saturation voltage (V_CE_sat).
* Top Choices: The 2N4401 and BC337 often outperform the standard 2N3904 or 2N2222 in efficiency because they feature a lower saturation voltage, letting them "pump" more cleanly at ultra-low battery levels.
* The Temperature Rule: The transistor should always remain cool to the touch. If it exceeds roughly 60°C, increase your base resistor to safely throttle the current.
4. The LED Load
* Wiring Configurations: Always connect multiple LEDs in series rather than parallel. Parallel setups suffer from current hogging (where one LED steals all the power and burns out early) unless you add individual balancing resistors.
* Voltage Rule: For every 1.5V battery added to the input, you can reliably string another 3V white/blue LED in series on the output.
Sorry I can't help you with calculations here.
I can give you my joule thief circuit that I've used to build dozens of them over the years.
It's a really good one. Will light up brightly with near-dead cells.
I wind the transformers on small (1/8" or so) ferrite rods using 32 gauge magnet wire.
(I based this on designs I found on Colin Mitchell's "Talking Electronics" site; see
this page for Joule Thief specifics.)
YIKES! I have never heard of "Joule Thief" but I can tell you that the original schematic design will not work with any old ferrite core for the "transformer."
The circuit is sort of a self oscillating flyback. For that, you need to have energy storage in the core...that is done with either a discrete or distributed gap in the core. A gap in an inductor core is analagous to the gap between capacitor plates....it is where the energy is stored!!
So for the most part a ferrite core will need some sort of calculated gap for energy storage and the correct inductance. Core materials such as powdered iron, or Permalloy (Molyperm) or Kool Mu have a distributed gap. Powdered iron is the cheapest but least efficient (more power loss) but might be worth experimenting with. For easy work in the lab, I really like ferrites with round center legs (such as PQ shapes) and PQ2020 might be good unless you find and want something smaller. Grinding a gap in the PQ core assembly requires the right equipment, but a spacer gap made with polyester film often works well enough for me.
YIKES! I have never heard of "Joule Thief" but I can tell you that the original schematic design will not work with any old ferrite core for the "transformer."
Well, what can I tell you?
Not only do all mine work, they all work very well, with the kind of transformer you insist won't work.
Couple pics of one of my homemade transformers:
I used to break up salvaged transformers with ferrite cores and use the pieces for my cores.
The circuit is sort of a self oscillating flyback.
Close; technically it's a blocking oscillator.
Here's a closeup of one of my units:
This one is in a loupe. The JT circuit drives 4 high-brightness LEDs in series/parallel. Very bright.
YIKES! I have never heard of "Joule Thief" but I can tell you that the original schematic design will not work with any old ferrite core for the "transformer."
The circuit is sort of a self oscillating flyback. For that, you need to have energy storage in the core...that is done with either a discrete or distributed gap in the core. A gap in an inductor core is analagous to the gap between capacitor plates....it is where the energy is stored!!
So for the most part a ferrite core will need some sort of calculated gap for energy storage and the correct inductance.
Whilst a closed loop ferrite like an E-core or toroidal core requires a gap or distributed gap, a solenoid ferrite core naturally has a massive air gap from one end of the solenoid to the other.
A joule thief does not require a gapped transformer because it operates as a resonant blocking oscillator that naturally prevents core saturation.
A gap in an inductor core is analagous to the gap between capacitor plates....it is where the energy is stored!!
That is somewhere between misleading and complete nonsense. There are plenty of inductors without an airgap. Creating an air gap lowers the induction, and thus the energy storage per Ampere of current. It also changes other things, such as making the saturation limit "softer".
YIKES! I have never heard of "Joule Thief" but I can tell you that the original schematic design will not work with any old ferrite core for the "transformer."
...works ok with a generic Ferrite bead (RS 238-283 - as per Big Clive, which I copied)
A joule thief does not require a gapped transformer because it operates as a resonant blocking oscillator that naturally prevents core saturation.
Where is the capacitor?
It will run at the SRF of the inductor.
Thanks Picuino,
I'll have to try out some of your tips to brighten my MAG-LITE replacement LED; starting with a cap across the 1K and a smaller core. My goal is to place it (mostly) inside an old bulb housing.
The "MAG2CELL HK 1A2" incandescent bulb currently pulls 775mA@3V. The joule thief (attached) starts-up at 13mA@500mV (99KHz), then 56mA
@1V (44KHz), 134mA
@1.5V (35KHz), 237mA@2V (37KHz). I'm not sure where the power is going (transistor seems cool) nor how far to push the LED. It's a NOS Radio Shack 276-0005, 28500MCD 20mA@3V LED. I may need to rewire the 2 D-cells for parallel operation if I can optimize the brightness
@1.5V.
Thanks Picuino,
I'll have to try out some of your tips to brighten my MAG-LITE replacement LED; starting with a cap across the 1K and a smaller core. My goal is to place it (mostly) inside an old bulb housing.
The "MAG2CELL HK 1A2" incandescent bulb currently pulls 775mA@3V. The joule thief (attached) starts-up at 13mA@500mV (99KHz), then 56mA@1V (44KHz), 134mA@1.5V (35KHz), 237mA@2V (37KHz). I'm not sure where the power is going (transistor seems cool) nor how far to push the LED. It's a NOS Radio Shack 276-0005, 28500MCD 20mA@3V LED. I may need to rewire the 2 D-cells for parallel operation if I can optimize the brightness @1.5V.
Be careful here. The traditional "joule thief" was designed to run from a depleted (dead) AA cell, and power a small 20 mA LED.
If you want a proper boost circuit that runs from large fresh batteries and powers a bright LED, then the joule thief is not really the answer. It will not be efficient enough. You will instead want a more sophisticated circuit, potentially one that uses a boost IC rather than a simple transistor.
I want to buy this small ferrite ring:
https://www.tme.eu/en/details/rt-100-60-40/ring-ferrites/richco/Sizes and material:
- Outer diameter: 10 mm
- Inner diameter: 6 mm
- Height: 4 mm
- A = 8 x 10^-6 m2 = 8 mm2
- l_m = 0.025 m = 25 mm
- Material: K5B
With 25 turns:
- Between 135 µH and 225 µH (180 uH +- 25%)
- Between 280 mA and 340 mA of saturation current
The problem is that with that saturation current, won't the LED be getting too much current when the battery is fully charged to 1.5V?
Edit:
According to my calculations, the peak current would be 100 mA with a resistor of 2200 ohms.
In the simulator, with a 4700 Ohm resistor and 1.5V in the battery, the peak current is 26mA, and average current 4mA.
It would be difficult to accurately predict the LED current in this circuit.
You should start with a higher value of base resistor and experiment with LED brightness by comparison with one driven with a relevant DC current.
The circuit can be made to provide regulation by modifying the drive to use feedback, the simple circuits output will vary a lot over the discharge life of a cell, but is a good place to start before adding complications.
By adding a small resistor in series with the battery, the peak current is greatly limited (28mA max). I haven't taken the battery's equivalent series resistance into account, so I'll have to conduct some real experiments to determine the true peak current and whether the LED can withstand occasional current spikes greater than its rated current.
Why do you need so many turns in the base?
You only needs some small current which just closes (or opens) the transistor, imho.
Smaller number turns there in the base may increase the efficiency..
PS: do find the minimal number of L1 turns which would work reliably, for example..
From what I've gathered from all the comments, the best solution is a 1:1 transformer for a single LED, and the more inductance it has, the more efficient it will be. That's why I'm winding it with so many turns.
In any case, I'm going to buy the components, assemble the circuit, and test its efficiency and operating limits.
This is a circuit I want to use with 15-year-old technology students. The technical details won't be too much for them to handle.
The most important thing is that the circuit be robust so that it works with a very wide range of component variations.
..from this sim it seems the largest power into the diode you will get with 10u L1..
It seems that reducing L1—and thus changing the transformation ratio—increases the current flowing through D1. But I don't see how that would lead to greater efficiency.
Reducing the resistance R1 also increases the current through D1, but right now, I don't think it's safe to go above 30 mA for a diode rated at 20 mA.
At this point, the best thing to do would be to build the circuit and make practical efficiency tests, but I'm more interested in having a robust circuit than one optimized for maximum efficiency.