In an earlier blog a commenter asked asked a question about how much inductance a Joule Thief should have; is there too much or too little. Well, I assembled and 
experimented with this Joule Thief to find out how little was too little.
I cut of two 16 inch (40 cm) lengths of 30 AWG (.25mm) magnet wire and wound them bifilar around a short length of 1/4 inch (6.35mm) diameter soda straw. Then I put adhesive tape around it to hold the wire in place. I tinned the wire ends with the soldering iron (this enameled wire insulation is the kind that melts with a soldering iron). I measured the inductance at 1.38 microhenry, which is extremely low for a Joule Thief.
I connected it to a conventional Joule Thief with a PN2222A transistor, white LED and a thousand ohm resistor. I knew when I connected the wires that I was connecting them with the correct polarity. But when I applied the 1.5VDC, the LED did not light up. It kind of flickered once in awhile when I touched the alligator clip to the cell’s contact.
I knew that the JT was going to be switching at a very high frequency with a coil with such a low inductance. I figured that the transistor might not have enough gain at this high speed, and that I might help it to oscillate by putting a small value capacitor across the resistor. I soldered a 100 pF ceramic disc capacitor across the resistor.
I powered it up and the LED lit up. But less than a minute later, the LED went out. I disconnected the power, visually examined the wires to see if there might be a loose connection but found nothing wrong. I reconnected it and the LED lit up, but it again went out less than a minute later. Weird. I put my fingers on the transistor and it was warm. Hmm, it seemed to be overheating and quitting soon after power was applied. That’s very interesting.
I disconnected it and reduced the power supply down to 1.25 volts DC. When I reconnected the circuit, it stayed lit, even though the transistor still felt a bit warm. I connected my DMM set to the frequency range but the frequency was too high for it. I got out my Heathkit GDO (actually a FET dip oscillator) and measured the frequency at 2.5 MHz. Wow, that’s really high!
The reason that the circuit wouldn’t oscillate without the 100 pF capacitor is simple. The transistor has an input capacitance of up to 25 pF. At 2.5 MHz, 25 pF has a reactance of about 2540 ohms. So the 1k and this capacitive reactance make a low pass filter that attenuates the signal coming through the resistor. The 100 pF lets enough of the signal bypass the resistor and go to the base, so the transistor can sustain oscillation.
Another factor is the Miller capacitance, or the capacitance between the collector and base. At high frequencies this capacitance causes negative feedback to the base and reduces the gain. Also another factor which becomes significant at high frequencies is the capacitance of the LED. This has to be charged each time it is switched on, so it becomes a significant part of the load on the transistor.
So the transistor was getting really hot inside and overheating when the power was 1.5 volts. Apparently the time that the transistor takes to switch on and off was a large percentage of the time during each very short switching cycle, and the transistor wastes a lot of power and overheats. Each switching cycle is 1 / 2.5 MHz, or 400 ns (nanoseconds). The datasheet for the PN2222A shows a risetime of 25 ns and a fall time of 60 ns. These are a significant part of the switching time, and during these times the transistor is not saturated or cutoff, so the transistor is dissipating power. The storage time, which is the time it takes for the transistor to come out of saturation, is also important. For the PN2222A, this is as much as 225 ns, over half of the switching time, which is 400 ns. What this all adds up to is that the transistor is huffing and puffing as it tries to keep up with the high switching frequency. The answer to the question I asked is this coil has too little inductance. I need to increase the coil windings and inductance so that the circuit will oscillate at a much lower frequency, such as 1/10 of 2.5 MHz or 250 kHz, or even less. The transistor will then waste much less power, and the circuit will be more efficient.
As would be expected, increasing the coil inductance will reduce the frequency, which is clearly too high in this case. To reduce the frequency to 1/10, the inductance will have to be increased by more than ten times, and from previous Joule Thief designs, it looks like increasing it to 100 microhenrys is a good estimate. The thin 30 AWG wire will have too much resistance, so heavier wire will have to be used.
Back to experimenting…