Let's Make Robots!

Regulated Joule thief, how?

Many of you have seen or read about a Joule thief, and for the rest of you: http://en.wikipedia.org/wiki/Joule_thief. They come in a great varieties, google easily finds them. One thing that I like about the Joule thief is that it is so forgiving about the applied transformer, the circuit simply adapts to that ever you throw after it. No need for calculations or anything.

Anyway, I'm searching a Joule thief circuit that is a little more advanced than just driving a white LED, I'm searching a regulated power supply. Still able to run on a half depleted AA cell, but providing 5V regulated output, a few mA's will be fine. But I do not just want to clamp the output with a zener, that would be total waste of energy.

So what I'm looking for could be a Joule thief circuit with some sort of feedback where the oscillation is dependent on the output voltage.


Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Check this out mogul, I found it surfing today. Looks like it might be a possible solution for you.


So I've been searching google about JT's and saw this topic.  This might be considered dead revival, but it doesn't look like the OP ever found a solution.  I took one of the circuits posted earlier and modified it a bit, and I think this would solve your problem.


I know you said you don't want a zener clamped on the output, but this does not perform the same way.  Rather, the cap reaches the zener voltage and then turns the transistor off.  Might need to play around with the resistor into that 2nd transistor though.  Looks like it turns out to be quite a bit more efficient!



I really wish you could edit posts.  Played around with this JT a bit more.  


Flick the load off and watch it charge up, and you can see the oscillations cut off and the quiescent drain is 680uA max.  Flick the load on and it will sustain 5ish volts at about 2.5mA.  The 2K load resistor is obviously just to model your current draw. This is all powered from a 1V supply.  If that Rload is different, you might need to mess with some other values to see what you can get. I'm sure you know that smaller inductor means faster switching frequency, or you could even go with a larger inductor and larger cap/multiple caps, and when it oscillates it will probably shoot above 5V and have the zener drain into the other caps so they all level out at 5V and you have more energy storage.  Still wouldn't be "constant" load supply technically, but you said you didn't need it to be on all the time didn't you?


Here is the circuit.


You can edit your posts by mousing over the More link. A "window" will drop down and edit is the first item on the list.

Hmm, it won't run here. can you upload the schematic as an image?

Preferable at a high enough resolution to be actual visible :-P

here is the schematic in another simulator, LTSpice.




So I've been tinkering with it a bit, trying to make it more efficient.  Most of the quiescent current is defined by R1 because Q2 is fully on after C1 is charged and has spilled over the zener.  So raising R1 limits that current, but also has an effect on the oscillation properties.  I started off with a 1K, and raised it and found a few things that help fix oscillation problems.


Here is the capacitor voltage:



And here is the current on the battery just as the oscillations cut off.

So quiescent drain is barely above 200uA.




I'm still a little iffy about it though because sometimes it will work the way I want it to in LTSpice and not in falstad, and sometimes it will in falstad and not in LTSpice -__-

Also, the I'm sure you know the sustainable output current is mostly related to the inductor size after what the battery supplies.  In the falstad simulator, I raised the inductors to 5mH and lowered the winding ratio to .1 so the frequency wouldn't drop so much, and it could sustain 5V ~12.5mA or around there.


I was looking at your schematic, and am a wee bit puzzled by it. Where is the output? I see you have used a BZX84C6V2L which is a 6.2 Volt Zener Diode which means the output must go at least that high for the zener to pass current, but it is being fed through a LXHL-BW02 which is a special power LED, which has a voltage drop of its own between 2.79 and 3.99 volts (with 3.42 volts typical) so you must get about 3.42 + 6.2 = 9.62 volts at the collector of the transistor Q1 to make it through the zener. Let me ignore the R2 resistor for a moment as the voltage drop across it directly relates to the current through it and mention also transistor Q2, which has a base-emitter junction voltage drop of about 0.65 volts. This adds to the prior drops to give ~ 10.27 volts.

You mentioned something about 200 µA, so if you had that current, the drop across the R2 resistor will be only 0.02 volts which is negligible.

I am not sure, however, how you have achieved oscillation with the transformer you are showing. I am assuming that the coils L1 and L2 are magnetically linked to form a transformer to give you the necessary feedback to achieve oscillation, but it appears you must have gotten your values reversed. The way these are shown with L2 (going to Q1's base) is .005 (presumeably mH ?) and the output side going to the collector being only .001, thus forming a step-DOWN transformer rather than a step-up which is needed.

Without stepping up the voltage, the voltage at the collector of Q1 will only be 1.5 volts from the battery (plus a tiny amount more if the circuit did actually oscillate). It is nowhere close to the ~10.3 volts needed to act as a regulator and charge C1 to a useable voltage.

If there is something I am not seeing here, please explain where I went wrong.


Here is the same circuit, with R1 moved around and the output load modeled as a resistor.


Here is the voltage across the load, and the current through the load.  So it maintains 5V - 11mA about.  55mW isn't too bad!



Hey there.

The zener thrown in there was just for the sake of an example.  LTSpice does have a 4.7V one, so I'm not sure why I used that one.  But it was just to show the regulation properties.

The same goes for the LED.  I didn't actually know the specs of the LED but I just picked a random one from the library and it seemed to have worked.  (In falstad, it is just a generic LED)

When the JT reaches the zener voltage, it turns on Q2, and the majority of the current passes through R1.  So 1V / 5000 Ohms ~ 200uA.  Now this does drain down over time, and Q2 will eventually turn off and the Q1 turns back on, kicks it up a notch, then it cuts off again.  So it is constantly turning off, waiting until the voltage drops, then turning back on and oscillating once just to kick the cap voltage back up.

Yes the K1 L1 L2 1 line creates the mutual inductance between the two inductors, and they are hooked up opposite to eachtother (phase dots).  

It is a step up transformer.  Coil turns can't be specified in LTSpice, so a higher inductance is analgous to higher turns.  That makes it a step up. 


Here are some more pics for you.


Current through R1.

 You can see how it oscillates at the normal frequency up until the capacitor is charged, then the current draw remains around 200 to 300uA, until it drops far enough and the large spikes are when it steps it back up.


Voltage of L1 and L2

L1 is in green, L2 in blue.  Notice how L1 peak is roughly 17V, L2 peak is about 7.5V



In fact, it is almost identical to your circuit.  Your use of a schottky is probably much better.  It looks like you have the step down transformer, 15 turns on Q1 base and 75 turns on collector.  It should be stepping up from collector to base, like 75 turns on base  --> 15 turns on collector.  That way when the current begins to flow through the collector, the induced voltage change gets stepped up on the base.

Your schematic diagrams all seem to have disappeared.

As to the transformer, no, mine is correct. I have built that circuit. I do not use spice or any of those so-called design help programs because I have not seen one yet that correctly analyzes circuits. Spice will usually not give the correct voltages and current except in the very most simple and standard circuits (and sometimes not even those).

The current needed to drive the base-emitter of a transistor is always smaller than the current needed in the collector-emitter loop. Also, you need a higher voltage out at the collector, not lower.  With my drawing set at 15 to 75, you get a 1:5 step-up in voltage. I show a somewhat discharged battery giving 1 volt. At 1:5, this will step up to about 5 volts. If the battery is closer to full charge (1.5+ volts), you will get 7.5 volts (or higher) hence I used the diode string to limit it so I get about 5.0 volts.

I never recommend using the design programs (such as Spice and others) as I mentioned. Here is another example where the program gave you faulty information. You cannot just place two loops of wire near each other or the coupling will not be strong enough to pass the current you need. (You might get a micro-amp or so, but it depends on the magnetic coupling. To get an acceptable output current, the coils must be wound on the same core and would be labeled as a transformer, not as two separate coils. The only way you would get sufficient magnetic coupling between two adjacent coils would be if you were working in ultra-high frequencies or above.