A few years ago, 2007, I believe, I blogged a few Joule Thief circuits on which I experimented. I used the 2N7000 low power MOSFET and the TN0702 (I’ll get to that one in a later blog).
Choosing a MOSFET The 2N7000 is a small signal, low power MOSFET in the standard TO-92 plastic package just like the other transistors commonly used for a JT. Its pinout is also the same as the PN2222A or 2N4401: S G D which correspond to the emitter, base and collector of a regular BJT. It has a Vgs(th) of 0.8V min, 2.1V typ, 3.0V max according to the specs. This is the threshold voltage at which the drain current becomes 1 milliamp. We can see that this is above the 1.5V or less battery voltage of the conventional JT. Something has to be done to get the 1.5V up to the 2 or more volts needed to fully turn on the 2N7000. In the schematic, Fig.1 is the original circuit I devised to make the voltage high enough to drive the 2N7000’s gate.
How it works The Q1, a puny 2N3904, is used as a conventional JT except it uses a much higher 10k resistor, which limits the current that Q1 can supply to the LED. But this is enough to get the 3.3 or so volts across the LED, so the 3.3V goes back through the feedback winding and to the gate of Q2. Notice that the current to the LED is rectified by the Schottly diode SD1, and filtered by C2. Some ‘experts’ claim that doing this makes the conventional JT more efficient, but I can’t see that because the rectifier drops a half volt, which is wasted power. But in this case it’s needed to boost the voltage to 3.3V. In Fig.4 I use the diode to rectify only the bias current going to Q1; the LED is driven directly by the drain of Q2 without rectified and filtered DC. The wasted power across the Schottky diode SD1 is eliminated.
In both circuits the smaller current from Q1 is increased greatly by Q2, which supplies most of the current to the LED. This circuit works very well, but it is not as simple as the conventional JT. If we just remove the 2N3904 and put a more powerful transistor in its place, we don’t need Q2 and other added parts. What we really want to do is eliminate Q1 altogether. I thought about it, and all I had to do was come up with a way to get a few volts to bias the gate of the MOSFET close to its turn-on point.
A Simpler Circuit At first I tried adding another AA cell to the existing one just to use for the gate. It worked okay, so I checked the current of this second cell by temporarily putting a 100k resistor in series with the positive lead, but there was no voltage drop across the resistor. Of course, because the gate of a MOSFET is just about an open circuit, it draws no current. Well, there is a small amount of current when the cell is connected, which is caused by the charging of the capacitance between the gate and source. But after that, there is no measurable current flow. The effective resistance of a MOSFET’s gate is a gazillion ohms. Actually, it’s something like thousands of megohms or more, but that’s so close to a gazillion ohms that it doesn’t matter (what is a gazillion anyway?). So I removed the AA cell and put a 1.5V alkaline button cell in there. Same result: the circuit worked fine. The current drain is so low that the cell should last as long as its shelf life, which today seems to be five or more years for a fresh battery. All I needed to do was order some button cell battery holders so I won’t have to solder to the button cells. But some inventive experimenters I have seen made a holder out of a PC board and a piece of a paper clip.
Measuring LED Current During these experiments I temporarily inserted a 1 ohm resistor in series with the LED as can be seen in the schematic. I can monitor the current with this by putting the DMM set to its lowest voltage range, 200 millivolts, across the 1Ω resistor. (Hey, that’s cool! I unhid the “Kitchen Sink”, which allows me to put special characters like the ohms “Ω” symbol into the text) If I measure 20 millivolts across the 1Ω resistor, then I’m getting 20 milliamps of current through the LED.
Other components
The 470pF capacitor C1 across R1 is there for a reason, I just don’t remember why. It could have been to make it start better with the much higher 10k resistor. The optional Cbyp is a bypass capacitor is an accepted industry practice, to prevent the cell’s internal resistance from making the circuit change frequency or whatever. It should work without it, most of the time.
The Toroid In the schematic in a note I gave the windings for the toroid core. These are typical for a Fair-Rite 2673002402 core, wound with enameled wire. These cores are called RFI/EMI suppressor sleeves, but they are the same size as a regular toroid core, and are available from Mouser for about a dozen cents apiece in small quantities. They work really well for conventional JTs and my Supercharged Joule Thief (see note below).
The LED I’ve used both white and blue LEDs for most of my JTs. I also bought a bunch of aqua colored LEDs which are the ones used for the LEDs in the green light of a traffic signal. These are very bright and have a forward voltage of about 2.8 volts. The size of the typical LED for a JT is the 5mm or T1-3/4 package. The reason is that a conventional JT on a fresh AA cell will put out about 60 to 70 milliwatts, which is just right for full brightness of a 5mm LED. I have used several other size LED, including the 10mm 1/2 watt 5 chip white LEDs, the 10mm 1W white LEDs, and the superflux or “spider” white LEDs, named for their four legged insect look. These are more powerful and require more current from the JT, in the case of 1 watt, that’s up to 300 milliamps.
Note: I use four 18 inch (50cm) lengths of 30 AWG (.25mm) enameled wire wound onto the core at the same time, which is called quadrifilar or 4filar wound. The number of turns is not critical and is usually around 25. I use the DMM on low ohms or tone to isolate one of the windings, which becomes the feedback winding. The other three windings are soldered together and become the primary winding. For the Fair-Rite 2673002402 core this gives about 400 uH per winding. The 2643002402 core gives about 140 uH per winding. Both of these cores are inexpensive, small size, and make very good Joule Thief coils.
What the heck is this “þ” character? Looks like a b and a p glued together! Back to experimenting…