Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.
The electronics cognoscenti may question the need to add to the 880000 Google hits that a search on 'Joule Thief' yields but I hope that the circuit shown might allow me to be shown some indulgence!

Detractors will point out a whole number of disadvantages of the JT as usually constructed:

No control over the output.
No ability to switch the circuit off in daylight.
Poor efficiency.

It is hoped that the circuit shown below addresses all of the above points.

An inspection of the circuit below reveals a few slight alterations and additions to the JT circuit normally seen. One transformer winding is connected directly to the base of the 2N2222A oscillator transistor and the other end of the winding is connected to the 100KOhm variable resistor VR1 which is in turn joined up to the positive supply via a backstop resistor R1. VR1 is our means of controlling light output and allows us to vary current through the LED from several milliAmps (very bright) down to a few microAmps and we use this control to bring up the light output as the battery decays. The circuit will allow a useable output to the LED until the battery voltage drops to around 0.7 Volt and then light output drops as the battery decays to around half a Volt and oscillation ceases.

There are applications where a low consumption low level light is required such as a backlight for a battery clock which can be powered by the clock's own battery - this circuit enables this.

C1 is very important. Counter-intuitively, addition of this component may ostensibly reduce the light output but measurement of actual circuit performance shows an enormous increase in efficiency. Many constructors have previously reported the improvement in efficiency that a capacitor across the base resistor in a JT circuit brings.

In order to turn off the circuit in daylight some circuit additions are required. Transistor Q2 is connected between ground and the base of Q1 via D1 and the base of Q2 is fed with the output of the solar cell via R2. When illuminated, the solar cell turns on Q2 and this shuts down oscillation of Q1 via D1 resulting in feed current dropping to around 10 microAmps. The solar cell can be one scavenged from an old garden solar light or one purchased new - it can be physically small as very little actual power is needed.

Initial efficiency measurements were carried out with a 15 cm OD ferroxcube toroid of known pedigree wound with 15 plus 15 turns of 28 SWG, this being fairly typical for a JT as usually built:

Feed Voltage V Feed Current mA O/P Current Through LED (mA) Eff%
1.59 1.15 0.54 76
1.21 1.80 0.55 65
0.99 2.09 0.50 66

The running frequency of the circuit was measured at around 390 kHz.

Substituting the toroid with a 15 plus 15 turn bifilar wound item achieved a consistent 5% increase in efficiency at the various feed voltages. This satisfied me that bifilar winding (i.e. when the two wires are tightly twisted together,) is better and it makes the toroid easier to wind.

Feeling that the running frequency was rather high I substituted a 70 plus 70 turn bifilar wound toroid. This is getting towards the maximum number of turns of 28 SWG that this toroid will physically accommodate. With a feed voltage of 1.10 Volts the running frequency was measured at 33 kHz and the efficiency jumped to a satisfying 86%. This circuit is now my preferred version.

We find that several strategically positioned JT's giving half to one mA to a diffused white LED and, allied with night vision, allow for navigation through the house at night e.g. for the various bathroom visits that advanced age brings. These JT's are truly run from 'dud' batteries retired from other duties and thus follow the original JT philosophy of scavenging the last scrap of power from 'dead' batteries.

This has finally confirmed to my satisfaction that bifilar wound toroids are superior in this application and readers might be interested as to why? My own take on this is that most JT's as published have far too few turns on the toroids of various (and dubious) origin that are used. Is it possible that these toroids are lossy if the running frequency is in the hundreds rather than the tens of kiloHertz? The simple method of switching off the circuit in daylight may be of interest. The base of Q1 is very sensitive to capacitance effects and the diode provides low capacitance isolation. I found that simply connecting a 2N7000 FET directly to Q1 base, for instance, gave rise to some very strange effects.

The experience has brought home to me the importance of measuring the actual efficiency of any circuit development. You can wind a transformer on a rusty nail, a horseshoe or anything you like and achieve oscillation but the efficiency might be terrible! I recall being very encouraged by binding together two axial inductors that are cheap and easy to source, and away it went, but the efficiency was very poor!

Note that medical advice may dictate that those of senior years should switch on a light before moving around in the house at night rather than rely on a device such as this.

Joule Thief.png


Crude construction but it works! Squares of PCB are attached to a PCB backplane with epoxy resin and the circuit laid out on these. The toroid, (note the large number of turns,) is mounted on an old scrabble tile with heat setting glue. The solar cell is on the left and the 'dud' battery on the right. Ignore the 'plus' sign written on the lid - this is from a previous use of the case.


Here the circuit has been forced 'On' in daylight by placing a card over most of the solar cell.
Further to my previous reply, I will add 'sorry I should have anticipated this one'. In fact AWG and SWG are very close in this region and this gauge is ideal for our purpose. In truth it is not critical so long as the gauge allows the required number of turns to be wound on the toroid and is not so fine as to be delicate and easily break during the winding process. Also with very thin wire the DC resistance rises.
what is the inductance? 500mH?
Last edited:
- Phase must show dots inverted on secondary.
- Transformer , L DCR, C and SRF must be defined.
- Can you make a simulation match your performance?
- I think the FET versions with Vt <.8 ought to do better.

Cute box.
Unless you have a true RMS meter reading, the trapezoidal area of the Relaxation Oscillator will be accurate.

RMS Voltage * RMS Current = true Average power may be computed in 1 repeating cycle or many.

Measurements that measure average rectified voltage or peak and convert to reading are only calibrated for a sinewave.
Slightly enhanced schematic inserted below.

The DC resistance of one winding is 0.3 Ohm.

I am not equipped to measure inductance but by dint of substituting known fixed axial inductors in a running YX8018 solar light circuit and measuring the current, I estimate it to be midway between 470 uH and 1 mH.

For efficiency I measure input current which is smoothed by C2. The output is measured by disconnecting the LED from ground, inserting a 47uF electrolytic and the measuring the current across this with a mA meter. This may be crude but it serves for comparison.

I have dallied with FET's such as the ubiquitous 2N7000. I think that results are better at higher feed voltages but this application needs the circuit to work down to the low voltages of failing batteries.

I have not tried circuit simulation.

Joule Thief.jpg
Many moons ago, I also became quite interested in JTs and experimented with it significantly to maximize efficiency. I arrived at almost identical conclusions to all of yours, and implemented those efficiency upgrades, Which meant a longer battery life. I am glad that we have both independently arrived at similar conclusions. :)

The difference in my circuit was that the control transistor (Q2 in your circuit) is a PNP and is located on the top of the winding.

I also was fortunate to have access to a large variety of core materials, and found that Micrometal’s T50-8 core provided the best efficiency.
I have a 1 transistor Joule Thief circuit with 2 resistors and 1 capacitor. I bought a package of 10 toroid induction to experiment with. I wound 6 turns of wire on the inductor to get the output voltage I wanted. This was a fun interesting project about 6 years ago.


Latest threads