I ordered some parts from Mouser, and along with them I ordered some BD433 and BD438 transistors. Belza uses the BD433 in some of his LED projects and as far as I can make out using google translate, Belza says the BD433 is supposed to do a very good job in a Joule Thief. The specs say it’s capable of handling 4 amps.
I built up a conventional JT using the BD433, and I added a third winding to the coil. The core is a Fair-Rite 26723002402. It’s trifilar wound with three lengths, each a foot and a half (45 cm) long, of 30 AWG enameled wire. The third winding is two additional lengths of 30 AWG in parallel. The third winding is the output of this coil and is rectified by a 1N5817 and filtered by a 47 uF ‘lytic, and then goes through the LED and a 1 ohm resistor to allow measuring the current. The LED is a 10mm 1watt single chip white; it can handle a few hundred mA.
The third winding is polarity sensitive; connection gives a brighter LED one way and dimmer the other way. The supply current was 120mA, the LED current was 22.5mA and the frequency was 10.5kHz. The BD433 in this JT is putting out more current than what would be typical for a BC337-25 or similar transistor.
Next JT – I used a core from a dead CFL and wound 2 windings, each 9 turns of 24 AWG, and measured the inductance of each winding at 101 uH. I connected another BD433 to it in the conventional JT circuit, with the usual 100 ohm resistor. I used a red 1 watt Seoul Semi ‘star’ LED. I connected this JT to the power supply set at 1.5V, and the LED would go on at the moment of connection, but it wouldn’t stay lit. I turned the power supply down to 1.2V and it worked okay, lit up brightly. I varied the voltage, and found that the LED got brightest at 1.1V, but got dimmer below that (as would be expected), and dimmer above that voltage (I didn’t expect that). At 1.1V the current was over 120 mA. Above 1.1V the current increased even though the LED got dimmer. I think the toroid core was having a problem handling all that current.
I unsoldered the core and soldered in a different core. This one had 6 uH and 6 turns of 24 AWG phone wire for the primary and 2000 uH and unknown but more than 20 turns of wire for the feedback wire. The LED lit up easily and very brightly at 1.5V, and the current was 250 mA.
I still haven’t opened the bag of BD438 PNP transistors. Back to experimenting…