2012-02-09 1/4 Volt Joule Thief

I found something that drew my attention, a Joule Thief that operates on less than a half volt and uses a silicon transistor.  Here is a short Youtube video of a circuit operating at less than 1/4 volt, and it says it is using a BC107, which is silicon.   I said to myself at first that’s not possible, but then I stopped the video and checked things out closer.  First off, I haven’t tried it yet, I need to wind a coil and try it (I have now, see update below).  What I saw was the LED lit, dimly, and the cell voltage at 0.136 volts.  The transistor looked like a BC107 which is in a metal package.  I saw the schematic of the circuit but he left the LED out of the schematic.  Also it did not use a resistor, so the cell or power supply must stay well below a half volt; if it goes above a half volt, the current would become excessive.

Then I saw the wiring, I saw that the LED was connected between the collector and base.  But I couldn’t tell the polarity of the LED.  My guess is the LED is connected with the cathode to the base, so that negative pulses on the base cause the LED to light when they reach negative 3 volts.  But if I do make the circuit, I will just connect two LEDs in parallel with the cathode of one connected to the anode of the other.  If one LED lights and the other is not lit, then I will know which direction the LED should be connected.

When I thought about it, the high 10 to 100 ratio of the collector and base windings could cause the 1/4 volt change on the collector to be 2.5 volts change on the base, enough to light the LED and turn on the base of the transistor.

I did a similar experiment, and I think I blogged it in my old blog.  I took a conventional JT and removed the LED from across the emitter to collector.  I connected the LED cathode to the base, and the anode to the negative.   Leaving the collector unloaded (connected only to the primary winding) allowed the modified JT to start at 0.45V, and I think I was using a BC550C, which is a low noise, high gain version of the BC547.

But the coil I was using was 1:1, so there was no V increase from collector to base.  By putting 10 times as many turns on the feedback winding, the author may have been able to get it to start at a much lower voltage.

I’ve read that if you plot the V-I curve for a silicon junction below 0.6V on a log graph, it is close to a straight line.  In other words, below 0.6V, the current doesn’t drop rapidly to zero, it behaves square law, or inverse square law, I forget  which.  Anyway, there could be a few microamps at a few tenths of a volt, and with enough turns on the core the base drive voltage may be enough to get it started.

Update Feb 12  I built the circuit, but I cheated.  I didn’t want to wind a hundred turns on to a toroid, so I picked a 420 microhenry choke, and wound 20 turns of 30 AWG on the outside and taped it in place with black electrical tape.  I don’t know what the actual turns ratio is, but I would guesstimate it’s more than 5 to 1. The added winding is the primary, and the choke’s winding is the feedback winding.  I didn’t know what the winding polarities were when I first hooked them up.  I had to reverse one winding to get it to work.  I used no resistor, as he did not use one in the video.

The JT starts up at 0.4V, but it draws excessive current; I had to set the power supply’s current limit at 200 mA.  Once I got it started, I turned down the voltage until it was running at 240 millivolts, which is slightly less than 1/4 volt. The supply current dropped down to about 60 mA, and the LED still stayed lit.

A clarification of the LED connection polarity: it is connected with the cathode (flat spot) to the base, and the anode to the collector.  In the picture the LED is lit up, but the flash overwhelms it so it doesn’t look very bright.  The frequency is about 70kHz.

In conclusion, the circuit is a conventional Joule Thief with three changes: The coil has many more turns on the feedback winding, the resistor is zero ohms, and the LED is connected between base and collector (see above paragraph).   A germanium transistor* will easily work at these low voltages, so  it’s not something to write home about.  But it is interesting to squeeze those last few tenths of a volt out of a dead cell with this circuit.

I am also going to have to make a  simple power supply that will supply less than 1/2 volt at a few hundred milliamps current limited.  BTW, this circuit would be a very good match for converting  the half volt output of a solar photovoltaic cell into a higher voltage.  The single solar cell has the inherent current limiting so that when this circuit takes off, the cell’s voltage will drop, limiting the current to a safe value, as long as the solar cell supplies the correct maximum current – it´s not too big for the circuit.  And that size should be determined in bright sunlight, because there is no resistor in the circuit and it will destroy itself if the cell voltage becomes as high as the base to emitter voltage.  But a solar cell is silicon and cannot put out more than 0.55 volts, which limits the voltage, so the main concern is the circuit´s power handling capability.

* Germanium transistors are no longer made in mass production quantities.  The prices for germaniums that were once very common are now very high – several dollars apiece.  Germaniums cannot be purchased at the local electronics store. so they are not really useful  for any circuit that might be built nowadays.  Also, very low voltage Joule Thiefs can be built using the MOSFETs and JFETs.  See my blogs here and here for more on this.

Back to experimenting…