The following is from my late Watsonseblog dated 2010 Dec 26 (no pics or links yet).
I made this a few years ago when I tried to see how well these two transistor JT type circuits work, and how efficient they are. The coil is a 180 microhenry RF choke with a single winding, and a DC resistance of about 2 ohms, which is not good enough for efficiency. It would be better if it was well under 1 ohm. The inductance value could be lower, 100 uH would have lower DC resistance. The circuit is one that is commonly found at websites and in projects. Performance was about the same as a regular single transistor JT.
As I said in the pic, it saves copper by eliminating the feedback winding and replacing it with a 3 cent transistor. Silicon is cheaper than copper, but it takes very little wire to wind the feedback winding on a coil. So I’m not sure if this saves money. One thing about coils is they have to be either wound by hand or by machine, and it takes longer to assemble than a capacitor or transistor. Since wire is subject to breaking and/or shorts, the failure rate is most likely greater than for other components. And throwing silicon at a circuit has been the norm for the last fifty years – it’s nothing new. And one advantage is that cheap RF chokes can be bought for about 30 cents apiece (Mouser 580-11R104C). The toroid with two windings is no longer needed, but can still be used.
For the experimenter (or the the newbie or experimenter who’s lazy or wants to throw it together quickly and have it work), the plus side is that there is no concern about how many turns of wire and of what gauge are needed for the feedback winding, or what’s the ratio between the primary and feedback winding. However, the circuit is a bit more complex than the conventional JT. If you’re winding a toroid, the lack of the feedback winding gives you more room to wind more turns of the primary (or only) winding, or even better yet, you can use heavier wire and still get the same number of turns on the core.
One minus is that the gain of the two transistors is multiplied, and transistors typically have a large gain spread, so if you have selected two high gain transistors or two low gain transistors, the value of Q1 base bias resistor R1 may be too low or too high. That’s why I show two resistors in series. The single resistor was on the side of being too low, so I added the second resistor to optimize it a bit.
It would be a good idea to put a potentiometer in there temporarily and vary the resistance over a range to see what seems to be best. Then remove the pot, measure it, and put in a regular resistor close to that value. This is done so the performance of the circuit can be observed while you vary the supply voltage, for example. If the resistance is too low or high, you might have the circuit quit when the voltage gets out of range. We all know that the battery’s voltage will drop as it gets depleted, so you want the circuit to still work when the voltage drops so you can suck the last joule out of it, just like a vampire.*
Performance
This circuit is essentially the same circuit that was used in a flasher, with the capacitor reduced from a few microfarads to a half a nanofarad. The light is replaced by the coil, and the LED is added across the output transistor. Here is a schematic of a similar flasher circuit.
The performance was better than I expected since I figured the extra transistor would use some of the current and waste it as heat. But the circuit takes about the same amount of supply current and puts out about the same current to the LED as a single transistor Joule Thief. That’s about 80 milliamps supply current, and 17 milliamps LED current, which is measured across the 1 ohm resistor; every millivolt is equal to a milliamp.
Silicon Saver Joule Thief
This Joule Tief is a conventional one: the transistor, the resistor, and the coil. But it’s unique in that it has no silicon at all. Instead, it uses germanium, quite uncommon in transistors today, but the only thing that was used in the early days, the 1950s and early 1960s. The coil is an EMI/RFI suppressor sleeve from a keyboard cord or whatever. A dozen turns of 24 AWG solid telephone wire make up the windings. I used a 390 ohm low value resistor so that it would be bright at low battery voltages. I think that if I were to put it on a fresh battery, it would draw excessive current and overheat. The pic shows it operating easily at 0.274 volts from the battery.
I received an email with some good advice from Jack, of muzique.com. With his permission, I’ll quote below.
Just a word about germanium transistors. You cannot usually measure the HFE of a Ge device with a DMM because of leakage that is very common with these vintage transistors.
For example, take one of your vintage PNP germaniums, and connect the emitter to +9V and the collector to a 2.2k resistor then to battery negative. Let the base float free. Since the base is not being biased by a current, there should be no voltage flowing through the transistor, but if you connect your DMM across the 2k2 resistor, you will probably find a small voltage and this is caused by the leakage.
The current leak will cause the DMM to read a falsely high value for the HFE.
I would guess that the leakage is the source of most of your problems with getting Ge-JTs to work. You could select one with a very low leakage and see if that helps.
I have a Peak Electronics transistor tster that can measure germanium leakage and derive a correct hfe value. This is a fast way to sort old transistors.
Also, the hfe and leakage will vary with temperature, as you have probably noted. Germaniums are much more sensitive to heat than silicon devices, and this might reduce their potential for JT use.
Best regards, Jack
Thank you, Jack.