Joule Thief

[jadaro2600]

I recently took apart a borken scanner and removed a bunch of interesting gizmos;  I'm not sure why the scanner failed, everything seemed to be owrking fine and then it just started in with the 'USB, short circuit, power drain detected' type message, on both windows and linux computers.

So, I removed the scanner lamp assembly - it's a very narrow florescent bublb with a driver board similar to the Joule Theif.

I attached a 9v battery to the source chords and it only lit half the tube dimly, then I connected a varaible voltage hobby source and found these results.

The bulb seems to light from a 12 volt source, consuming between 50 and 70 milliamps.  At 18 volts, it consumes 105ma and only gets minimally brighter.

I'll have to measure the voltage output to see what that is like:  ...there's a capacitor leading off of one of the wires to the lamp, ..just touching he lead with one of the DMM probes causes my nearby touchlamp to change settings and I can see sparks- I'll have to setup a series capacitor bank to measure it ( Later ).

[conradelektro]

My attempts to build a high frequency Joule Thief resulted so far in an air coil coil that can drive 10 or 20 LEDs (one or two AAA or AA batteries) very brightly with about 80 to 150 KHz.

I know one can drive many more LEDS easily with a toroid, but at a much lower frequency. It may not be of much use, just an experiment.

L3 = 108 mm diameter, 200 turns, 0.3 mm enamelled wire, height of coil is 72 mm, wound on PVC tube, 3.8 mH

L1 + L2 = 15 turns bifilar, wound over one end of L3, 0.3 mm enamelled wire:

Base resistor 530 Ohm:   10 LEDs on L3, supply Voltag 0,8 to 1,5 Volt (one AAA or AA cell), 80 to 90 KHz, 60 to 130 mA

Base resistor 1.2 KOhm:   10 LEDS on L3, supply Voltage 1 to 3 Volt (two AAA or AA cells), 120 to 150 KHz, 40 to 70 mA

Base resistor 820 Ohm:    20 LEDs on L3, supply Voltage 1 to 3 Volt (two AAA or AA cells), 80 to 100 KHz, 70 to 200 mA

The circuit starts to swing with a supply voltage of about 0.6 Volt and starts to overheat with a supply voltage above 4 Volt (one would need a heat sink for the transistor and a 10 Watt base resistor). All works fine up to 3 Volt (two fresh AA or AAA cells) because the power consumption stays below 200 mA (0.6 Watt) with a base resistor of at least 100 Ohm.

The 10 or 20 LEDs shine very brightly. In case one has a steady power supply with 1,5 or 3 Volts, one can use more LEDs, up to 20 or 40 LEDs. I suggest 10 LEDS with one battery (AAA or AA) and 20 LEDs with two batteries (AAA or AA) in order to have the LEDs shine brightly even when the Voltage drops to 0.8 or 1.6 Volt.

I found that a full bridge rectifier between L3 and the LEDs uses up too much power, the net effect are dimmer LEDs. The Joule Thief circuit (also with air core coils) produces a "one sided signal" (on L3) that is much stronger on the positive side than on the negative side. Therefore a full bridge rectifier has more losses than gain. When capturing the energy from L3 in a capacitor, a full bridge rectifier might still introduce some gain, but not for LEDs.

Greetings, Conrad

[conradelektro]

About lighting a neon lamp from both sides:

The Joule Thief circuit produces a "one sided signal" on L3 (which one may call a "spike"), also with air core coils.

The scope shots below show this fact. Although the signal on L3 without much load looks almost symmetrical,  it is never the less "top-heavy".

Once a small neon lamp is connected, the negative part of the signal almost colapses (and can not light the other side of the neon lamp).

This is the reason why one can light a neon from both sides, but only one side will glow, because the other side of the signal colapses. On a FL or CFL one end will be less bright.

Using a toroid (instead of an air core) the "one sidedness" of the signal is much more pronounced.

To produce a truly symmetrical signal on L3 one would need a more complicated "fly back circuit". See the Original "Andrinerii Revised Mazilli Driver" (far down on this page, last circuit):

http://wiki.4hv.org/index.php/Flyback#High_Power_Drivers.

On this page one can also read about a Joule Thief type circuit (and its high pitched squeal), see "Simple 2n3055 driver" (first circuit on the page).

Greetings, Conrad

[gyulasun]

Hi Conrad,

Would like to suggest (if you have not tried) to load your L3 with another series string of LEDs, connected also in parallel with L3 but in the opposite diode direction than your present 21 LEDs are, to load the negative 100V peak output too. This way both of your original L3 peak to peak voltage (+240V and -100V) could be loaded, use 13-15 LEDs in series for loading the negative 100V and leave the series 21 LEDs as it is now shown.

Another suggestion would be: what if you use a fast diode for doing a half wave rectification for the positive 240V peak and you use another fast diode for the negative 100V rectification, both with one-one puffer capacitor, and you get a summed DC voltage between one positive cap end and one negative cap end? 
Putting the above otherwise: this way in fact you build a voltage doubler which utilizes both the pos and neg peaks of the output waveform, see this link: http://www.play-hookey.com/ac_theory/ps_v_multipliers.html
This way you would have a single diode loss in every half wave versus the diode bridge double loss and also you may gain a little to avoid using the LEDs as rectifiers (because LEDs may have a diminishing rectifier efficiency if frequency increases.) 

The third suggestion would be to use other transistor instead of the 2N3055, I mean bipolar types designed for switching, with low saturation voltage, high beta etc. I know the 2N3055 is cheap and robust and for the time being is almost "perfect" for this job, but later the type should be also a consideration. (Higher beta can reduce power dissipation in the base resistor.)

rgds,  Gyula

[gyulasun]

Conrad,

Oops, I have not noticed you addressed the negative peaks loading while I prepared the previous answer, so my first suggestion is answered with it?

Gyula