Full Brightness Myth – The typical JT (Joule Thief) uses a single transistor, draws 80 milliamps from the 1.5V battery, and puts out up to 20 milliamps to the 5mm white or blue LED. Let us consider this 20 milliamps “full brightness” or “full brilliance”, which is a term often used but has no quantitative meaning because it uses no unit of light measurement. I see this term used by a JT builder who connects 12 LEDs in parallel to a JT that uses a single pipsqueak 2N3904, and claim the LEDs light to “full brightness”, a totally unsubstantiated claim. A typical JT using a 2N3904 puts out 8 or 9 milliamps to a single LED, and when 12 are connected in parallel, each LED will then get about 3/4 of a milliamp, which, in fact, is less than 4 percent of the current needed for “full brightness”. Such a myth!
You can’t get something for nothing. You will be spending 12 times as much for the LEDs and get the same amount of light from 1 or all 12. “There ain’t no such thing as a free lunch”, you know. Putting bigger tires on a race car won’t make it go faster if you’re using the same underpowered engine (2N3904).
What should we do? First off, get rid of the 2N3904 and replace it with a 2N4401, PN2222A or best, a BC337-25. This will bring the total LED current up to about 15 to 20 mA. Divided 12 ways, that’s still only about 1.5 mA per LED, still way too low.
Let’s Get Serious We need a dozen times more current to each LED. The quickest way, design-wise, is to give each LED its own JT circuit. Just use 12 coils, 12 transistors, and 12 1000 ohm resistors, and duplicate 12 JTs. Each JT will put out roughly 17 mA to its own LED, but at 83 mA battery current per JT, that is a total of 1 AMP battery drain. The poor AA cell’s voltage will drop down to 1.3 volts or even less, and that will reduce the output of each JT circuit.
To help with this problem, I would put only 6 LEDs per battery. A half amp is still a lot for a single AA cell, but the voltage won’t drop so much. Then the next thing is to use a transistor that can handle 6 times as much current as a BC337. The 2SD965 is capable of 5 amps, but it’s hard to find. The Fairchild Semi KSD5041 is the same and is available from Mouser for a reasonable price. Attention! The KSD5041 is a Japanese transistor and has a pinout of E C B.
The resistor will have to be lower, I would guess that a starting point would be 330 ohms. With this much current, a 100 uF capacitor across the battery is a must have. The core for this JT should be heavier, I would use a FT87-75 available from surplussales.com. I would use four windings wound quadrifilar, of 24 AWG solid insulated wire, with 3 of the windings connected together in parallel for the primary winding. A short piece of cat 5 cable should work, long enough for ten turns.
Will this put out enough current for 6 LEDs? I haven’t tried it yet, so I don’t know for sure.
Update Feb 16 I put one together (see picture) using a 27 LED light that I got from OSH for 6 bucks. This yellow and black light has 24 LEDs on top and 3 on the end. It has a 3 AAA cell holder, and draws so much current that the batteries quickly go dead. So I drilled holes through the case for the two wires and soldered them to the holder contacts. Note that if you do this, the pushbutton switch allows you to switch from 24 LEDs to 3 LEDs and then off. You do not, ever, want to disconnect the LEDs from a JT, for the possibility of damaging the transistor.
The coil is wound with 10 turns of Cat5e solid wire, three wires for the primary, and 1 for the feedback winding. The length of each wire was 14 inches or 0.35 meters. The core was the FT87-75 from surplussales.com, and each winding of the coil measured 350 microhenrys. I used a 330 ohm base current limiting resistor for a start.
On a 1.5V power supply it drew 210 milliamps, and the frequency was 8 kHz. Assuming about 50 percent efficiency, that is about 50 milliamps LED current, more than 3 times better than the typical JT, but still far short of the 120 mA we want for 6 LEDs. I think removing a few turns from the coil and reducing the resistor would help increase the LED current.
Update Feb 18 I reduced the resistor to 100 ohms, and the supply current went up to 280 milliamps, an improvement. Then I removed three turns so the primary and feedback windings now have 7 turns. The supply current went up to 350 milliamps, and the frequency was 12.5 kHz. Assuming 50% efficiency, that calculates to about 13 milliamps for each of the 6 LEDs. This is at the low end of what I would expect from a multiple LED JT. Note: of course this ridiculous 24 LED monster light has four times as many LEDs, so each LED is only getting about 3 milliamps, which isn’t all that bright. But it’s doing pretty good, all things considered.
Back to experimenting…