Joule Thief With Ultra Simple Control of Light Output




About: I am a retired analytical chemist living with my wife Cynthia in Cornwall, south west England. I have held the UK radio amateur call sign G3PPT since 1961. I have been interested in computing since the day...

Intro: Joule Thief With Ultra Simple Control of Light Output

The Joule Thief circuit is an excellent entrée for the novice electronic experimenter and has been reproduced countless times, indeed a Google search yields 245000 hits! By far the most frequently encountered circuit is that shown in Step 1 below which is incredibly simple consisting of four basic components but there is a price to be paid for this simplicity. When powered with a fresh battery of 1.5 Volts light output is high with commensurate power consumption, but with lower battery voltage the light and power consumption drop away until at around half a Volt light output ceases.

The circuit is crying out for some form of control. The author has achieved this in the past using a third winding on the transformer to provide a control voltage, see:

Whatever control is used it should have the basic property whereby turning down the light output also turns down power consumption so that a low light setting results in low battery consumption and longer battery life. The circuit developed in this article achieves this and is much simpler in that the extra winding is not needed and yields a form of control that could be retro-fitted to many existing circuits. At the end of the article we show how to automatically switch off the circuit in daylight when deployed as a night light.

You will need:

Two general purpose NPN transistors. Non critical but I used 2N3904.

One silicon diode. Totally non critical and a rectifier diode or signal diode will be fine.

A ferrite toroid. See later in the text for more information.

One 0.1 uF capacitor. I used a 35V Tantalum component but you could use a 1 uF ordinary electrolytic. Keep the voltage rating up--35 or 50 Volts rating is not excessive as during development, and before your control loop is closed, high voltage can be applied to this component.

One 100uF electrolytic capacitor. 12 Volt working is fine here.

One 10 K Ohm resistor.

One 100 K Ohm resistor

One 220 K Ohm potentiometer. Non critical and anything in the range 100 K to 470 K should work.

PVC single cored hook up wire which I obtain by stripping down telephone cable

To demonstrate the circuit in the early stages I used a Model AD-12 Solderless Breadboard which I obtained from Maplin.

To produce a permanent version of the circuit you will to be equipped for elementary electronic construction including soldering. The circuit can then be constructed on Veroboard or similar material and another method of construction using blank printed circuit board is also shown.

Step 1: Our Basic Joule Thief Circuit

Shown above is the circuit diagram and a breadboard layout of a working circuit.

The transformer here consists of 2 lots of 15 turns of single core PVC wire salvaged from a length of telephone cable twisted together and wound on a ferrite toroid--not critical but I used a Ferroxcube item by RS Components 174-1263 size 14.6 X 8.2 X 5.5 mm. There is enormous latitude in the choice of this component and I measured identical performance with a Maplin component four times the size. There is a tendency for constructors to use very small ferrite beads but this is as small as I would like to go--with very small items the oscillator frequency will get higher and there may be capacitive losses in the final circuit.

The transistor used is the 2N3904 general purpose NPN but almost any NPN transistor will run. The base resistor is 10K where you might more frequently see 1K used but this may help when we come to apply control to the circuit later.

C1 is a decoupling capacitor to smooth out switching transients generated by circuit operation and thus keep the power supply rail clean, it is good electronic housekeeping but this component is often left out which can result in unpredictability and erratic circuit performance.

Step 2: Performance of the Basic Circuit

Some knowledge of the performance of the basic circuit may be instructive. To this end the circuit was powered with various supply voltages and the respective current consumption measured. The results are shown in the picture above.

The LED starts to emit light with a supply voltage of 0.435 and consumes 0.82 mA current. At 1.5 Volt, (the value for a new battery,) the LED is very bright but the current is above 12 mA. This illustrates the need for control; we need to be able to set the light output to a reasonable level and thus greatly prolong battery life.

Step 3: Adding Control

The circuit diagram of the extra controlling circuitry is shown the first picture above.

A second 2N3904 (Q2) transistor has been added with the collector connected to the oscillator transistor base, (Q1.) When turned off this second transistor has no effect on oscillator function but when turned on it shunts the base of the oscillator transistor to earth thus reducing oscillator output. A silicon diode connected to the oscillator transistor collector provides a rectified voltage to charge up C2, a 0.1 uF capacitor. Across C2 there is a 220kOhm potentiometer (VR1,) and the wiper is connected back to the control transistor base (Q2,) via a 100 kOhm resistor completing the loop. The setting of the potentiometer now controls the light output and in this case the current consumption. With the potentiometer set to minimum the current consumption is 110 micro Amps, when set for the LED just starting to light up it is still 110 micro Amps and at full LED brightness the consumption is 8.2 mA--we have control. The circuit is being powered in this example with a single Ni/Mh cell at 1.24 Volts.

The extra components are non critical. At 220 kOhm for the potentiometer and 100 kOhm for Q2 base resistor the control circuit functions well but places very little load on the oscillator. At 0.1 uF C2 provides a smooth rectified signal without adding a large time constant and the circuit responds rapidly to changes to VR1. I used a tantalum electrolytic here but a ceramic or polyester component would work just as well. If you make this component too high in capacitance then response to changes in the potentiometer will be sluggish.

The last three pictures above are oscilloscope screen grabs from the circuit whilst operational and show the voltage on the collector of the oscillator transistor. The first shows the pattern at minimum LED brightness and the circuit is operating with small bursts of energy widely spaced. The second picture shows the pattern with increased LED output and the bursts of energy are now more frequent. The last is at full output and the circuit has gone into steady oscillation.

Such a simple method of control is not completely without issues; there is a DC path from the positive supply rail through the transformer winding to the transistor collector and through D1. This means that C2 charges up to the level of the supply rail minus the forward voltage drop of the diode and then the voltage produced by Joule Thief action is added to this. This is not of significance during normal Joule Thief operation with a single cell of 1.5 Volt or less but if you do try to run the circuit at higher voltages beyond about 2 Volts then the LED output cannot be controlled down to zero. This is not an issue with the vast majority of Joule Thief applications normally seen but such is the potential for further developments that it could become significant and then resort may have to be made to the derivation of the control voltage from a third winding on the transformer which provides total isolation.

Step 4: Application of the Circuit 1

With effective control the Joule Thief can be much more widely applied and real applications such as torches and night lights with controlled light output are possible. Additionally with low light settings and commensurate low power consumption then extremely economic applications are possible.

The pictures above shows all of the ideas in this article so far brought together on a small prototype board and with the output set to low and high respectively with an on board pre-set potentiometer. The copper windings on the toroid are of the more usual enamelled copper wire.

It has to be said that this form of construction is fiddly and the method used in the next step is far easier.

Step 5: Application of the Circuit--2

Shown in the composite picture above is another realisation of the circuit this time built on a piece of single sided printed circuit board copper side up with small pads of single side printed circuit board stuck on with MS polymer glue. This form of construction is very easy and intuitive as you can lay the circuit out to replicate the circuit diagram. The pads make a robust anchorage for the components and connections to ground are made by soldering on to the copper substrate below.

The picture shows the LED fully illuminated on the left and barely illuminated on the right this being achieved with simple adjustment of the on board trimmer potentiometer.

Step 6: Application of the Circuit--3

The circuit diagram in the first picture above shows a 470k Ohm resistor in series with a 2 Volt solar cell and connected into the Joule Thief control circuit effectively in parallel with the on board trimmer potentiometer. The second picture shows the 2 Volt solar cell (salvaged from a defunct garden solar light,) wired in to the assembly shown in the previous step. The cell is in daylight and hence providing a voltage that turns the circuit off and the LED is extinguished. The circuit current was measured at 110 micro Amps. The third picture shows a cap placed over the solar cell thus simulating darkness and the LED is now illuminated and the circuit current measured at 9.6 mA. The on/off transition is not sharp and the light comes on gradually at dusk. Note that the solar cell is being used merely as a cheap control component to a battery circuit does not itself supply any power.

The circuit at this stage is potentially very useful. With a solar cell mounted discreetly in a window or on a window sill charging a super capacitor or nickel metal hydride rechargeable cell, a highly effective permanent night light becomes a possible future project. When used with an AA cell the ability to turn down the light output and then turn off the light during daylight means that the circuit will operate for a long period before the battery voltage falls to around 0.6 Volt. What a superb bespoke present for grandparents to present to grandchildren! Other ideas include an illuminated doll's house or a night light for the bathroom to allow standards of hygiene to be maintained without loss of night vision--the possibilities are enormous.



    • Fix It! Contest

      Fix It! Contest
    • Tiny Home Contest

      Tiny Home Contest
    • Metalworking Contest

      Metalworking Contest

    11 Discussions


    10 months ago

    like what joule thief it goes better than the flash camera
    1,5v > 3,6v 3w led :)


    1 year ago

    i suggest 27 to more turns for coils also apx. 1:3 for power coil : driver coil ratio

    the minimum of 3 (ampere-)turns for power (Collector or Emitter) coil

    the minimum of 14 turns for driver (Base) coil

    for the oscillations expected to be to start/occur/continue

    keywords "magnetic coupling"

    3 replies
    Lionel SearReneJ2

    Reply 1 year ago

    Thank you for your interest and sorry for the delay in my reply.

    I have an ongoing interest in applying the Joule Thief circuit and I have built a number of them over the last year or so. My feeling is the the single most important factor in the construction of the transformer is the use of bifilar winding i.e. the wires must be tightly twisted together before winding on to the toroid. If you are adding a third winding for some reason the use trifilar. I general I think it is a good idea to keep the number of windings up to bring the oscillator frequency down to reduce capacitive losses and maybe even radio frequency interference.

    It is tempting to do the winding in two stages bringing the centre tap out for connection to the positive rail but my experience is that this can cause difficulties in the starting of the circuit even if the windings are very close on the toroid. (A Joule Thief that does not start reliably is a disaster as there is a near short circuit from the positive rail through the transformer to the transistor collector and then from the transistor emitter to ground.)

    Considering the use of different turns ratios for the transformer I cannot comment as I have always successfully used 1:1 but I will have a go on the breadboard at varying the ratio to see if there is any improvement in efficiency to be had--the LED does present a very strange load to the circuit.

    ReneJ2Lionel Sear

    Reply 1 year ago

    boosting the master switch closing seems to improve efficiency, requires 3-ple winding for best timing+driving -- as the LC values define AC profile -- it must be obvious that achieving the circuit tuning (as in simulator also) in real is much of a fairy tale - though not completely unlikely ... a waste of time stuff -- that's why don't build'em anymore

    ReneJ2Lionel Sear

    Reply 1 year ago

    yes, Hi! - i also was exited built these some while ago - what i learned was - powered from single 1.2 - 1.6 V cell the blocking osc. has a best performance efficiency at about-below 30mA operation e.g. (it's likely not a rule but difficult to avoid) , attempting the high output power v.-s results (speculating) the magnetic saturation - as the first 20s my led was ultra-bright - and then i couldn't repeat it even by changing the transitors , cite "

    Another “core loss” is that of magnetic hysteresis. All
    ferromagnetic materials tend to retain some degree of magnetization
    after exposure to an external magnetic field. This tendency to stay
    magnetized is called “hysteresis,” and it takes a certain investment in
    energy to overcome this opposition to change every time the magnetic
    field produced by the primary winding changes polarity (twice per AC
    cycle). This type of loss can be mitigated through good core material
    selection (choosing a core alloy with low hysteresis, as evidenced by a
    “thin” B/H hysteresis curve), and designing the core for minimum flux
    density (large cross-sectional area).

    " src: Practical Considerations - Transformers - so i assume i permanently saturated/spoiled the small ferrite core with pre-magnetization - must be studied

    . . . those things proof their value during blackouts where you can use 1 (efiicient or not) per room and it works about 24h from little used battery before starting to fade - ! a great invention !


    2 years ago

    is it possible to use some type of light sensor to turn the LED on at night and off during the day? BTW I am obviously an extreme rookie to electronics not even having a grasp of the terms used for those devices I am asking about. But I would love to give it a go.

    1 reply
    Lionel SearNewoldbuilder

    Reply 2 years ago

    Hello and thanks for the interest.

    Yes indeed. The method of control relies on the output being rectified to produce a positive voltage which is fed back to the transistor which is connected to the oscillator transistor base turning it on and turning down the oscillation. So to answer your question, the positive voltage can also come from an external source and a 2 Volt solar cell salvaged from a defunct garden light is perfect this being shown in the final stages of the article.

    The Joule Thief is perfect for a newbie. It will work on a solderless breadboard and you can learn much putting one together and it actually does something i.e. lights up. Be prepared to write off the odd LED or transistor but that is all part of the fun!


    2 years ago

    Great Joule Thief Circuit. This is great for sucking the last juice out of batteries.


    2 years ago

    I have a half dozen of these, well attempts at these laying around on my bench. Never have had one working though. Perhaps this one will be the ONE! Have to give it a try when I am feeling a bet better. Not today though, it's one of those days when nothing gets done but a deeper hole in my recliner where my ass sits.


    2 years ago

    Two comments. The LED connection is not shown in the schematics. In #3, where is the connection that charges the battery from the solar cell?

    1 reply
    Lionel SearQuantumdust

    Reply 2 years ago

    Thanks enormously for that and I will now do some editing. On the first point I will add the LED in to the schematics.

    Regarding the second point I admit to being a little unclear. The circuit is correct but the solar cell is being used as a very cheap (scavenged) component to control a circuit rather than be a source of solar power. Inside a house there is very little energy left in the daylight once it gets through the windows and not enough in general to yield very much from a solar cell--you might source some energy by positioning the cell on a window sill, up against a window or perhaps catching a transient shaft of sunlight. However there is more than enough to provide the few microwatts of power required to switch off the Joule Thief when not required in daylight and thus prevent waste of battery power. I will go back and emphasise this point in the text.

    This does raise the interesting point regarding the possibility of a solar powered night light. I think that it is a viable project but such a device must work 24/7, winter/summer and ideally in as many parts of the world as possible and it certainly is not as simple as it sounds. It will need more effort than modifying a solar garden light!

    Thank you again for your help.