The "Reverse Joule Thief" Battery Charger





Introduction: The "Reverse Joule Thief" Battery Charger

Here is a totally different take on the Joule Thief (JT) circuit commonly found in garden lights. Instead of charging a 1.2v battery directly from the solar cell and converting the power to run a 3-volt LED, we'll be using the JT to convert the output from the solar cell and charging a Lithium battery first. Then when night falls, the battery is used to drive the LED directly.

 This method has some advantages: (1) the Lithium cell that was chosen here (and avialable for $2 here) has an output of 3-volts, which can drive a White LED directly; it also has a huge capacity (800mAH) and very low leakage. (2) The solar cell normally can only charge the NiCd battery in full, direct sunlight, but, with the JT circuit, it is able to deliver power to the Lithium cell even on overcast days.

Step 1: The 'Reversed' Layout.

A look at the circuit will tell you this is not a run of the mill JT configuration. Most obvious will be the fact that there is only a single coil involved (the 220uH) - we are using a second transistor (Q2) and C1 to take over the timing requirements. This allows us to use a wider range of coil values, as well as operate over a larger voltage range.

Besides reversing the charge / discharge order, this circuit also reverses the location of the driver transistor and the coil, but wait, that's not all! The transistors all have reversed polarities, and even the output voltage is reversed!

No, it isn't an error! Diode D1, the LEDs and the charged battery all have their polarities  reversed! That's because this Joule Thief is configured as a voltage inverter. This arrangement was chosen due to its advantages for this kind of application.

To improve efficiency, the traditional JT relies on a fairly constant battery supply (over a millisecond or so) to give it a boost when it is delivering power. With the limited output from a Solar Cell, we have to store all its power in C2 and feed it into the Lithium in one big pulse, meaning the capacitor will be "empty" for the few critical millisecond, cancelling the 'kick' the normal JT requires to work well.

Our 'Reversed' JT circuit will work as a regular JT - without the 3v Lithium load, an input of 1.2v will light up the LEDs quite nicely. Not strictly necessary, the LEDs are there so you can SEE the system working, and also to prevent the battery overcharging.

Step 2: The Light

Because the complex stuff is done already, the light itself is quite simple. A BC327 transistor is used as the switch to turn on the LED. The 'inverter' circuit allows us to combine a 'dark' detector, as well as a over-charge limiter in one. Output from the Solar Cell and the charge on the Lithium is monitored through R3 and R4. If the Solar Cell sees no light, or if the Lithium goes over 3.6-volt (the safe charge limit for this cell), LED3, a 100mA unit turns on.
  Switch S2 is optional and allows you to run the LED at full power, otherwise R2 will only allow 20mA to the LED.

The second picture shows the two halves of the circuit together.

Step 3: Parts List

Solar Cell. 2-volts with 100-ohm load

Q1,Q3 BC327 PNP. Can be any low-signal amp of sufficient current rating (>100mA)
Q2 BC337 NPN. Most will work but if you change Q1, Q2 or L1, you may need to adjust R1 for best performance (Try 3.3k to 15k)

D1 1N4148 or 1N914 or similar
LED1 Blue or White LED
LED3 100mA (1/2W) White LED

C1 220pF. Can be 150-500pF
C2 50-200uF

R1 10k-ohm
R2 330-ohm (use 470-ohm for longer run time)
R3 3.3k-ohm
R4 6.8k-ohm (use these values instead of the one on the schematic)
R5 100-ohm. Go as high as 220-ohm for lower brightness.

L1 100-500uH. Many home-made ones will work.


The second image shows the waveform measured at the top of the coil. The portion above the tag (2) is the charge stored in C2 fed into the coil. The sharp negative going pulseis the battery being charged.

Step 4: Testing the Solar Cell

If you, like me, have collected Solar Cells from busted Garden Lights, it is important to choose one which has good performance.

A simple test is to take them out on a sunny day and measure their output with a Voltmeter that has 100-ohm resitor across the + and - leads. This puts a load on the Solar Cell and will give you some idea of how 'powerful' it is.

  In this picture, the meter shows almost 2 volts in a slight shade. Using Ohm's law, we know that 20mA (2-volt / 100-ohms) is also flowing from the cell, and this is the values we will use to design our circuit.

Step 5: Some Calculations

To get an idea of how well our system can perform, we need to figure out how much power we can get, which dictates how much power we can use up.

Using the numbers of 2-volt and 20mA from before, we now know that we can get 40milli-watts by multiplying Volt with Amp. Over the course of an 8-hour sunny day, we should be able to get 320mWH of power from our system.

With a 100mA LED running full steam, it will draw (3.3-volt x 100mA), or 330 milli-Watts! Divide this number into the 320mWH we get from sun-power, it means we'll use it up in less than an hour! Of course the battery will continue to supply current from its reserve, but that is power that would require extra days to replenish.

Step 6: Variations

This circuit performs the same function, but switches the input's polarity so that the output is more 'normal' looking -- positive at the top and negative as ground. It also shows the use of 2N390x transistors and 3.7v Lithium-ion cells, which can also be 3 NiMH batteries in series.

The values of R4 and R5 are important: not only is Q3 tasked with switching on the light after dark, it is also responsible for switching the LED off when the battery drops below 2.9v, as well as to prevent over-charging. In fact, it will delay switching on the light if the battery has not received a full charge during the day.



    • Oil Contest

      Oil Contest
    • Creative Misuse Contest

      Creative Misuse Contest
    • Clocks Contest

      Clocks Contest

    35 Discussions


    For murky cells I use auto clear coat rubbing compound. it brightens them right up. wash the dry (I set mine in the sun). I then gave the cells a quick coat of clear spray paint to keep the shine, Im not sure if that has an effect on the cell efficiency.

    Hope that helps. Bryan

    to anyone with "murky" solar panels, clean them and put a piece of clear packing tape over it. it turned it crystal clear for me and i saw a .5v increase in its output!

    4 replies

    If you wind your own inductors/transformers using parts and methods described in other JT instructables, and using larger wire, I think you can significantly reduce the resistance. I think this may be one key to my desired 0.5V 0.5A cell JT-esque device. And perhaps FETs instead of bipolar transistors, too. Maybe. - Chris whY

    One major hurdle with using a single cell is, as you suggested, the (in-)efficiency.

    Power is proportional to I2 * R so, with any given inductor, the power loss is four times higher for every doubling of the current. Thus to get 3v from 0.5 v, the current draw will be at least 6x, making the losses 36 times higher. Also, getting the proper bias for any semiconductor unit at these voltages will be problematic, to say the least.

    I have a project I need some help with. I am new to this forum and can get lost for hours ready all this cool stuff. I need a PCB board made with a subsiquent schematic that ressembles the joule thief, Would you be interested ? I will pay !!

    Go to
    Here you will find a "Joule Thief" circuit that is new and so simple that you may not have to worry about a PCB board...

    Oh, and it might be worth considering swapping out the overload-protect LEDs with a zener and FET charge-regulator, at little additional time/expense. Can probably grab 'em off a scrounged board. Older boards often even have the schematic symbol for each part, making it easier to ID zeners, transistors, etc.But I still give kudos on this build just the way it is. Haven't built it yet, but it looks sound to me. -Chris whY

    2 replies

    In constructing this circuit, I was trying to specify parts which most interested experimenters would have on-hand. Hence the LEDs rather than anything more fancy.

    The idea of a zener is a good one. In fact that is what's being done in an adaptation of this circuit used in an MIT project in rural Africa, you can read more here.

    Good evening, have just finished building your circuit, have been looking for a more powerfull led light set up for a long time, can only get the main light to come on dimly but does come on gradually when covering the solar panel, have probably made a mistake with the transistors, will be picking my friends brains as to the position of the pins, using a mobile phone battery of 3.6 volts, do you think this would be alright as I know you can have a certain amount of pecentage error in most circuits, also will this circuit charge up in dull winter weather, thanks for a very interesting article, regards Doc Cox.

    Oscillating output for a charger... is this thing functioning as a pulse charger? Now, if I could get this in a size for lead-acid batteries...

    So, is this safe for the battery? What if it were a 2800mah 3.7V li ion battery?

    3 replies

    Once the battery reaches about 3.3v, the LEDs will turn on to make sure the battery doesn't overcharge. A 3.6v Li-Ion at full charge is 4.2v, so the circuit as shown will charge it to about 80%. By replacing LED1 and LED 2 with a 4.8v zener, you will charge the Li-Ion to 100%.

    Actually, though the link you posted earlier isn't explicit, that looks like a LiFEPO (Lithium Iron Phosphate) battery rather than a generic "Li-IoN" battery (which can refer to a variety of exact chemistries, but not LiFEPO).

    A LiFEPO battery has a nominal cell voltage of 3.0V, and an open-cell, fully-charged voltage of around 3.6, as compared to Li-Ion which is typically 3.7 and 4.2 for those conditions. It also has about half the capacity of the Li-Ion batteries, but on the plus side it's a lot cheaper, more rugged and has much longer life (it may last 1000 cycles under circumstances where Li-Ion would only last 200).

    So your 3.3V maximum charge voltage is a good maximum, since at that voltage the LiFEPO battery is really nearly fully charged. Don't increase the maximum voltage any more on that battery; although LiFEPO is less likely to behave catastrophically than Li-Ion, you really just don't want to abuse lithium batteries.

    Can I charge 2 li ion batteries with this? Or do I need a better power source?

    from what you have said using 100ohms to do the maths that gives you 0.02amps
    that would be 2ma

    2 replies

    No, there are 1000-milliAmps in an amp, so 0.02amp is 20mA.

    Great instructable!

    Just the other day I bought a solar lamp at the hardware store for $3.50 bucks. I bought it just so I could rip it apart. Blows me away - only $3.50.

    I'm guessing there is no reason I couldn't use a pair of 1.2V NiMH instead of the Lithium cell. I've got a bunch of AA cells.

    I'm hoping to keep a Zigbee radio running instead of the LED. The radio uses 200uA while sleeping. When transmitting / receiving the current jump up to 35mA. However, the active time is only for a few seconds. The radio then goes to sleep for a programmed amount of time (15 minutes).

    OK, my question, is there anything special about the inductor? Will any old 200uH inductor work? I found a couple at digikey - here is an example:

    I had a hard time seeing your inductor in your pics. Could you put a lasso around the inductor to call it out?

    Thanks again for a great instructable.

    1 reply

    Since the joule thief is nothing more than a boost converter, the value of the inductor doesnt. The value of L just determines the frequency and amount of ripple in the output. So you can probably get away with using a home-made inductor using a ferrite core from a old power supply with a few turns of wire (say like..10+ is a good start)...or if you shack sells a assortment of can pick and chose one that works from there..its like 2.50 or something for the pack