The following is from a cached version of my late watsonseblog.
2011 Mar 15 Use Darlington for Joule Thief
An email correspondent asked me if I have ever used a Darlington for a Joule thief. The short answer is no. The longer answer involves some lengthy explanations below.
A Darlington pair connects two transistors so that their individual current gains are multiplied. If you have two transistors connected, each with a gain of 200, then the overall gain is 40000. Darlington transistors are available with two transistors in a single package, or you can connect them yourself. The single package Darlingtons may have gains of several tens of thousands. This gives the transistor some advantages. The current needed at the input is just microamps or even less. The input impedance is high. But along with this there are some serious disadvantages.
The most important issue is that the two base to emitter junctions are in series. Each junction has a forward voltage of about 0.6V. The total voltage is over 1 volt, typically 1.2 volts. In order to get the Darlington to conduct current, the base has to be 1.2 volts above the emitter, and in the case of a Joule thief this is more than the voltage across a used battery.
The other important issue is the collector voltage can never go below 0.6V because if it did there would then be too low a voltage across the second transistor’s base to emitter junction and it would not conduct. In the case of a Joule Thief with a battery voltage of 1.5V, this issue means that the Darlington pair would waste more than one third of the power.
The two issues above are not a problem when the supply voltage is more than several volts, such as in an audio amplifier. One minor issue is that due to the high impedance, the Darlington’s frequency response is not that high, but this is not an issue for a Joule Thief. But the two issues above are unacceptable for the low voltage Joule Thief circuit. This Youtube video explains it visually.
Sziklai Pair
One way to deal with one issue is to connect a PNP and an NPN transistor together to make a Sziklai pair, AKA complementary Darlington. This reduces the voltage drop across the input to 0.6V, but the emitter to collector voltage is still 0.6V minimum, making for power waste and low efficiency just like the Darlington. This too makes it unacceptable for a Joule Thief. Remember, we are trying to squeeze every last bit of energy from the battery and turn it into light, not waste it on heat.
Other Choices
The advantage of a Darlington pair’s very high current gain is not needed for the Joule thief. The circuit requires very low Vce(sat) – the collector to emitter saturation voltage -which the above pair configurations do not have. It also requires low Vce(sat) at high collector currents. There are several transistors that meet this requirement, so we do not have to consider using the above compound connected transistors.
If you take apart a disposable camera, you will find a flash unit that operates off a single AA cell. At one time this circuit used a 2SD965 (Japanese) transistor, but nowadays the typical flash unit may use an equivalent surface mount transistor. The 2SD965 is available from Fairchild Semi as the KSD965, but the Fairchild KSD5041 is a newer and just as good substitute. Both of these are capable of handling 5 Amps collector current, which is amazing for such a little transistor. Another similar transistor is the KSC2500(D) from Fairchild. All of the above are capable of 1 watt or less dissipation. If you want higher power, the KSD1273 from Fairchild will handle several watts depending on the heatsink. This is a Superbeta transistor, with a current gain of a thousand or more. It does not necessarily have super low Vce(sat) but for a Joule Thief it will put out quite a bit of power.
Also, the D suffix or other letter at the end of the transistor’s part number is usually (for Japanese transistors) the gain range. In the data sheet, there will be a footnote telling what those mean. The A, B, C, gain ranges may be overlapping but usually centered around three points, for example 120, 240, and 360.
Other Considerations
Back to the Joule Thief. When you run a JT on 1.5V, you are already asking the transistor to do a difficult job. Only 1.5 volts supply voltage, and a hundred milliamps or more means the DC resistance of the circuit has to be very low. But the JT gulps large slugs of current with each pulse, meaning that the transistor has to conduct two or three times that much current for brief periods. This means even lower DC resistance, down to 1 ohm or less. You must minimize losses in the transistor, and in the coil primary winding by using wire that is heavy enough to have low voltage drop – a small fraction of an ohm. Success here shows up as low losses and higher efficiency. But remember that the typical JT has only about 50 percent efficiency. That is why I like my Supercharged Joule Thief (See Fig. 2 here this link no longer works, instead, use this link).
Now, when the battery runs down, or you run the JT from a lower voltage power supply, things are even more difficult. If you use a 0.5 volt solar cell, the voltage is one third that of a AA cell. The lower voltage means that three times as much current is needed to give the same power. And the DC resistance of the circuit must be NINE times lower. This means it must be a small fraction of an ohm. Things become critical, such as the resistance of the wiring on the circuit board.
More on this in another blog. Also go check out the Wikipedia article for a more comprehensive explanation.