This fascinating circuit¹⁾ steps up tiny DC voltages to more useful levels using only a few parts easily scavenged.

                        //        
                     +-->|-------+
                     |           |
  +------------------+    _+-----+
  |                 C \   / E    |
  |    1:1      BC547 -----      |
  |  = = = =            |B       |
  +--UUU+UUU--+--/\/\/--+        |
  A2    |    B1    1k            |
        |                   |    |
        +-------------------||---+
                         +1V|     

• Selecting Joule Thief Components
• Assembly Instructions
• Joule Thief Variations
  • The Sun Thief
• Joule Thief Notes
• Joule Thief Testing
• Archive of Jewel Thief Stuff

The circuit is essentially a blocking oscillator^{2) 3) 4)} somehow working without a capacitor, perhaps relying on parasitic transistor (Miller) capacitance to operate. ⁵⁾

A brief test of this bare-bones circuit thrown together from scavenged components, powered by a freshly-charged but old NiCd battery, and driving a white LED indicated 30% efficiency,⁶⁾ though it isn't hard to improve it. ⁷⁾ And if you try, you might get up past 75% efficiency using only a few common, cheap parts.

Adding a diode and capacitor as shown below⁸⁾ can more than double efficiency.^{9) 10)} Although this change introduces another loss due to the forward voltage drop of the diode, the capacitor captures the current during the oscillation cycle between V_f,diode and V_f,LED that would otherwise be lost in the transistor or shunted to ground. This configuration also provides a blocking diode for charging battery cells from one or two solar cells. Once started, the oscillator averages the supply voltage as it swings between the forward voltage of the diode, and a value slightly above the sum of the forward voltages of the diode and the load. For example, a 1.2V NiCad cell driving a 3.0V LED through a standard 0.7V silicon diode will average 1.2V as it oscillates between 0.7V and some value greater than 3.7V.

                                     //   
                                  +-->|--+
                                  |      |
  +------------------------+-->|--+--||--+
  |                        |             |
  | = = = = =            |/ C            |
  +-UUUU+UUUU---/\/\/-+--|               |
  A2    |    B1         B|\|E            |
        |                  +-------------+
        |     |                          |
        +-----||-------------------------+
           +1V|                           

                                     //   
                                  +-->|--+
                                  |      |
                              D1  |  C1  |
  +------------------------+-->S--+--||--+
  |                        |      |      |
  | = = = = =          Q1|/       +--Z<--+
  +-UUUU+UUUU--/\/\/--+--|           Z1  |
      T1|        R1   |  |\|             |
        |             |    +-------------+
        |             |              D2  |
        |             +--------------|<--+
        |             |                  |
        |             +-----------/\/\/--+
        |    |                       R2  |
        +----||--------------------------+
          +1V|                           

Although the first circuit works fine and is easy to make from salvaged components, a few more improvements and protection circuitry complete the design.

The lower voltage drop of a Schottky rectifier or germanium diode is more important than its reverse current leakage.

The pulldown resistor does seem to stabilize the circuit.

It surprised me to learn it is entirely possible to blow a white LED with a single NiCd cell. Apparently if the LED disconnects from the running circuit, as often happens when messing with components on a breadboard, the capacitor can charge up to a fairly high voltage and blow the LED when it reconnects. You can solve this problem by placing a 5V (or so) zener diode across the LED to shunt excess voltage to ground. (As an alternative, you may place the zener between ground and the transistor's collector to protect the transistor and Schottky diode from voltage spikes as well.) In either case you now have a zener-regulated power supply. Though an LED load draws so much current it drops the voltage across it to the LED's forward voltage, a higher-impedance microamp load will get a fairly stable supply voltage. (See below for improved power-supply characteristics.)

It is also a very good idea to protect the base of the transistor from reverse voltage spikes with a diode between it and ground. A normal silicon diode is fine for this. Note the diode's reverse current leakage makes the bias resistor redundant.

                                            //   
                                         +-->|--+
                                         |      |
                                     D1  |  C1  |
  +-------------------------------+-->S--+--||--+
  |                               |      |      |
  | = = = = =                 Q1|/       +--Z<--+
  +-UUUU+UUUU--/\/\/-+-/\/\/-+--|        |  Z1  |
      T1|      R1    | R2    |  |\|      |      |
        |            |       |    |      |      |
        |            |       +-|<-+------|------+
        |            |         D2        |      |
        |             \|Q2               |      |
        |              |---------+-/\/\/-+      |
        |            |/|         | R4           |
        |            |           |              |
        |            +-/\/\/-+   +-/\/\/--------+
        |    |         R3    |     R5           |
        +----||--------------+------------------+
          +1V|                               

The third circuit may provide decent voltage and current regulation as well as component protection, though this circuit may be more appropriate for loads other than the LED shown. Note the zener diode Z1 is redundant in this circuit. Also, a more non-linear cutoff response can be provided by putting a different zener diode (perhaps 3.0V to regulate at about 3.6 V) in series with Q2 base resistor R4. It may also be a good idea to pull the base of Q2 down to ground with a fairly large resistor. Another possibility is to put a capacitor between the collector and emitter of Q2 to try to minimize the amount of time Q1 is in the linear response region.

If the supply voltage is 1V, the maximum (saturation) voltage at the base of germanium transistor Q1 is 0.3V, and R1+R2 should be 200Ω during normal operation (Vout=3V), consider R1=100Ω and R2=100Ω. The maximum voltage at the point between R1 and R2 is 0.65V. At this point, Q2 needs to be just below cutoff.

Hmm… Choose R1 and R2 so the voltage between them is at the C-E forward voltage drop of the transistor when the supply voltage is high enough to provide maximum output current ignoring Q2?

Suppose Q1 is a silicon transistor with hFE=100 and provides 20mA at 3.1V while conducting 100mA at 1.2V with a base resistance of R1+R2=1000Ω. (This is similar to test results using a 2N5192). When on, the base of Q1 is at a potential of about 0.7V and conducts 1mA, so the end of the coil must be at 0.7+0.001×1000 = 1.7V. Q2 (also Si, hFE=100) is just barely at cutoff, <0.1 mA at 1.5V. (0.5mA at 1.2V needed for saturation)

First of all, you can throw this circuit together with what you have on hand and it will probably work as long as you power it with one standard rechargeable battery cell or at least two solar cells in series.

A separate document details component selection.

• [Coil/transformer winding]
• Choose a transistor rated at least 50 mA for lightweight applications, 100 mA to drive one LED at 20 mA, 200 mA for more than 20 mA charging current or for driving 2 LEDs, and 300 mA for 2 PV cell use. Make sure it can handle V_CEO spikes at least twenty times the supply voltage. Transistors designed for switching applications usually work better than those intended for amplification.
  1. A germanium transistor should self-start if the supply voltage is above 0.15V, and can provide efficient power conversion.
  2. A silicon bipolar transistor will self-start and can work well with supply voltages greater than 0.7V.
  3. A low-threshold FET is best for use with a supply voltage greater than 1-2V, and should self-start from supplies of at least 0.75V. These can provide the most efficient power conversion.
  4. Two or more similar transistors in parallel often improves performance.
• [Feedback resistor]
• [Bias resistor]
• Protect the base of the transistor from reverse-bias with a diode. A small silicon diode is fine for this.
• Choose an output rectifier with the lowest forward voltage drop, the fastest response time, and the lowest reverse current leakage for the current needed by the load; larger is usually better. 2-5 amp Schottky rectifiers work especially well here.
• The capacitor should be at least 2µF, though a larger value is better for use with very low supply voltages or larger coils. A 10µF capacitor seems to be sufficient for most circuits. Use a capacitor rated above but fairly close to the load voltage. (To be safe, it can be rated about double the output voltage, and should be rated higher than the optional zener diode.)
• If using a zener diode to protect the load, it should have a breakdown voltage a bit less than twice the load voltage. The common 5.6V zener works well to protect a single white (3V) LED.

It is unlikely more than 150mA will pass through the transistor in this circuit, and power dissipation should not be a problem if everything is working correctly.

There are zillions of applications for variants of this circuit.

• See this solar charger, for example.
• Here is a similar patent 4274044.

This document focuses on (very) low-voltage applications.

Using a germanium transistor in the step-up stage can produce the desired supply voltage from as little as 0.2V input. This isn't the best voltage regulation¹¹⁾, but should be sufficient for low-power microncontrollers and such. The above circuit with feedback may work better.

  +------------------------+--->S----+------------+
  |                        |         |            |
  |  = = = =             |/ C        |          |/
  +--UUU+UUU--+-/\/\/-+--|      +-||-+-/\/\/-+--|
        |             | B|\|E   |            |  |\|
        |             |    +----+----+-->Z---+    +
        |             |              |       |    |
        |             +--|<----------+---||--+    +--o
        |                            |             >+2.5V
        |     |                      |
        +-----||---------------------+---------------o
       >+0.2V |                           

Using a silicon diode for the emitter-follower regulator will drop the output voltage about 0.65V, so the common (and temperature-constant) 5.6V zener will supply about 5V. R = V/I = 5.6/(0.006 to 0.011) = 500-900Ω¹²⁾ though higher may be better if the load is not too great.

Moved to The Sun Thief

Moved to The Sun Thief

Combine the Joule Thief with this circuit to make a self-adjustable micrcocontroller power supply.

                                    .----------.  
                                    |   [uC]   |  
                                    |          |  
                                    |       PWM+--+
                                    |          |  |
                                    |GND    Vcc|  |
                                    .-+------+-.  |
                                      |      |    |
                                      +--||--+    |
                                      |      |    |
 +-------------------------+---+--S<--+-->Z--+    |
 |                         |   |_            |    |
 | = = = =               |/      |           |    |
 +-UUU+UUU-+-/\/\/-+-----|_    +<||          |    |
      |            |     |\    +-||----+-----|----+
      |            |    Q1 |   | Q2    |     |    |
      |            |       |   |       |     |    |
      |            +-/\/\/-+-+-+-/\/\/-+     |    |
      |            |         |               |    |
      |            |         +---------------+    |
      |            |                         |    |
      |            +----|<--+--|<--+--/\/\/--|----+
      |                     |                |
      |                     +----------||----+
      |                                      |
      -1V                                   ---
                                             -

Once the micro powers up, the (inverted) PWM stream charges a capacitor to disable the PNP transistor Q1 by pulling its base high, then takes over switching as the PWM signal begins switching the low-saturation PNP MOSFET Q2 (which maybe should be NPN). Using a germanium transistor for Q1 will allow powering the uC from a single solar cell (0.45V). A diode or resistor may be necessary to protect the RUN pin.

If an LED power-on indicator is desired, it may replace the zener diode if the LED's forward voltage is higher than the minimum voltage required by both the microcontroller and the MOSFET. Two or more LEDs in series may be used to achieve this. (Note the LED should be not be reverse-biased as the zener diode is.)

Note some kind of feedback should be used to make this a regulated power supply. For example a second zener diode with a lower value can be attached to a microcontroller input pin to signal when the voltage goes over this diode's breakdown voltage.

This can probably be done better a little differently, perhaps by using a second RUN signal from the uC, or a third transistor so RUN shorts the base of Q1 to +1V (or ground for an NPN Q1), or by putting Q1 and Q2 in series.

A galvanized and a copper nail can be stuck into a lemon to supply the power for a flashing red LED.

                                            //   
                                         +-->|--+
                                         |      |
                                         |  C1  |
  +-------------------------------+-->S--+--||--+
  |                               |  D1  |      |
  | = = = = =                 Q1|/       +--Z<--+
  +-UUUU+UUUU--/\/\/-+-/\/\/-+--|           Z1  |
    T1  |      R1    |       |  |\|             |
        |            |       |    |             |
        |        Q2|/        +-|<-+-------------+
        +-/\/\/-+--|           D2               |
        | R3    |  |\|                          |
        |       |    +-----/\/\/----------------+
        |       |          R2                   |
        |       +-/\/\/-+                       |
        |   |           |                       |
        +---||----------+-----------------------+
      +0.2V |          
   to +2.5V          

This improved joule thief circuit adds protection to prevent blowing the LED if the lemon vampire is attached to higher voltages. By adjusting the base bias on Q1, this circuit should also provide better performance over lower voltages within its operating range. This will probably work best if Q1 is a germanium switching transistor and Q2 is a standard silicon transistor intended for amplification. Note both transistors must be of the same bias, both NPN or both PNP.

To estimate the value for R3, the following formula should give a reasonable resistor value to try first¹³⁾ with a silicon feedback diode and R2 bypassed.

     (Input Voltage - Transistor VBE) * Transistor hFE 
R = ---------------------------------------------------
                    1.3 * Load Current

The 1.3 is the standard fudge factor to be certain there will be enough current to saturate the transistor. The load current is the input voltage divided by the sum of R1 and R2.

R3 = (3-0.6)*100/(1.3*(3/1000)) = 61538 ≈ 68kΩ

Calculate voltage at which a silicon transistor begins to conduct by substituting 0.3V for VBE in the equation above.

R1 and R3 should be replaced with pairs of resistors to make voltage dividers.

Need to make some supercap/Luxeon flashers to chase the squirrels from the attic.

A typical CFL drive circuit has enough parts to make an automatic light. This will not make the most efficient solar light, but it should work well enough.

• 1 Busted CFL Bulb
• 2-3 Solar Cells
• 1 NiCd or NiMH Battery Cell
• 1 Phototransistor
• 1-3 White LEDs

Everything else you need is soldered to the CFL circuit. You can even mount your new circuit in the CFL base and reuse the diffuser (if it had one).

If the base resistor available has a small value (<770Ω), either add a small capacitor (:0.1μF?) between the base and collector of the transistor, or reduce the number of windings on the secondary coil driving the transistor's base. Adding this capacitor is probably the better choice; it's difficult to determine the transformer coil ratio needed.

: Expand on the following.

• Try using a germanium transistor to decrease operating voltage, perhaps as low as 0.15V. ¹⁴⁾, AS127?¹⁵⁾, ¹⁶⁾
• Also, try using a Schottky or germanium diode to increase efficiency.¹⁷⁾ Even a tunnel_diode would be worth investigating.
• In fact, replacing the charge pump diode with a zener or tunnel diode should obviate the need for a separate zener diode to shunt excess voltage.
• This circuit will also drive at least two LEDs in series.
• What about limiting the LED current? (Red LEDs typically have an internal resistance of 22 ohm.)
• Use a shake-magnet coil and rectifier in parallel with a very low voltage power supply (such as a manure fuel cell) to provide a jump-start voltage.
• Use the shake-magnet coil and rectifier to recharge the battery.
• Solar LED Throwie: AAAA NiMH batteries typically 200-400 mAhr. .(Find these inside rechargeable 9V batteries sometimes.) 200mAhr: Flat top, 1.626” tall, .276” diameter, .19 oz. ( 5.2 gms).

Some questions

• How to reduce the minimum power supply voltage?
• How to increase efficiency?
• What effects the oscillation frequency?
• Can I make it flash?
• 19 to 56 W/m² @1V is 1.9-5.6 mA/cm² total area of two cells in series. 4-6 cm² should be safe for trickle charging a NiCd/NiMH cell. Adjust area larger to account for efficiency: 6-9 cm²?

— David Wagner 2007/07/10 10:25

• Make a Joule Thief. by bigclive1@gmail.com
• White LED Drive Circuit
• Charging Methods
• AA Battery Solar Charger
• Cell Chemistries