Joule Thief

[TinselKoala]

The little black things are inductors, probably fraction of a uH so hard to read on a meter. The other things are, I think, gas-filled spark gaps, sort of like modem protectors. I'm not sure about this but that is what they look like to me. Presumably the "470V" might be their firing threshold. Generally used like lightning protectors to shunt HV spikes to ground, hence the heavy terminations.

[Farmhand]

Thanks Tinsel, So two 470 volt GDT's on an inkjet printer circuit board wow curious. Might help make a neat little spark gap coil with
low input frequency and low duty of use. Maybe a switch in a stunner umm.. light. 

That UTsource ebay seller has a lot of parts and cheapish as well, he'll get my business.

I was just thinking one of these NCP1402 parts with a voltage of 2 to 2.5 volts would be good for charging a supercap from a
feeble or a good source of power. And higher output voltage parts also good for simple LED torch circuits, but no adjustment of
power output is not the best for making the torch last longest in use.

..

[Farmhand]

As we can see by my pictures on the previous page where the LED's are not even fully driven, and yet the LED's themselves cannot
be seen, the apparent efficiency of the Joule thief circuit is kinda obsolete. If you can count the LED's you are looking at then they
are not fully driven.

In my opinion according to my experiments if we simply give the LED's a truly pulsed supply then we do not necessarily reduce the
brightness but we do reduce the time the light is emitted and therefore we also reduce the total amount of light emitted. So any
efficiency in that respect is only "apparent".

For dimming LED's, Pulsing them is a fantastic idea, but I would bet that for the maximum practical amount of emitted light we must
drive the LED at it's rated voltage for 100% duty.

The amount of light being analogous to Energy and and the brightness being analogous to Power in a way.

The real efficiencies are determined by the output power as compared to the input power and in the efficiency of the conversion
of stored energy to light.

Now I'm not saying a Joule thief cannot be efficient, but I think it's biggest draw card is being able to use down a single cell to almost
complete depletion of energy.

And saying that I see no reason not to use a capacitor across the output or LED's so that the output power can be measured,
at least in cases where people might claim they are OU or want to claim other peoples setups are OU.

As well people might be pleased to see that the LED's may well emit a lot more light and be more useful than they are with low duty higher voltage pulses.

Even if a capacitor is used and it still allows some ripple the LED's will see the ripple.

Anyway I cannot get anywhere near the same light out of pulsed LED's (without burning them out) as I can from LED's supplied
with "smoothed" DC, and the ones supplied with smoothed DC I think will last longer as long as the ratings are respected.

The LED's on the previous page pictures are now outputting as much light as the really bright ones driven in the regular devices
I have.

Basically if the circuit does not dissipate much power in itself then you have efficiency in that the circuit itself does not waste input.
LED's are inherently efficient so as long as the circuit is efficient and does it's job we have efficient LED lighting.

..

[Farmhand]

OK with a 50 Farad capacitor charged to 2.1 volts and run down to 0.65 volts about 100 Joules is consumed, This energy was
sufficient to run the 3 x 5 mm LED's in parallel for some minutes over a half an hour, ( didn't keep perfect track). But going on only
30 minutes the input would be 100 / 30 minutes = 3.33 joules per minute then divided by 60 seconds = 0.055 Joules per second
or 0.055 Watts average input power.

During the run the LEDs had 2.988 volts DC across them until the input went under about 0.85 volts when the LED's were visibly
dimmer than with the 2.988 volts across them but not much, i failed to record the voltages across the LED's as the input voltage
dropped below 0.9 volts, I should have. With almost three volts DC across them the LED's are driven well enough for practical use.

Anyway I intend to now put together a LED board with a Current Sense Resistor so I can try to measure the output power, and
get a rough efficiency rating for the circuit while driving three 5 mm bright white LED's.

I have on the way some 5 volt NCP1402's and some Si7478DP mosfets so I can switch a coil powered by the 5 volts with the help
of a micro controller and try some fun stuff.

[Farmhand]

Alright in order to make the calculation more accurate I used the 50 Farad capacitor from 2.1 volts down to 1 volt which is about 85 Joules and I got only 15 minutes run time but the output remained consistent during the run at 2.98 volts @ 23 mA = 0.0685 Watts.
So if I take 85 Joules and divide by 15 minutes then again by 60 seconds I get 85 / 15 = 5.66 / 60 = 0.094 Watts input.

So the efficiency for the setup during that run was about 0.0685 / 0.094 = 72.8% efficient. Not bad considering that I am using
regular electrolytic capacitors and the output capacitor is a single cap not a paralleled group for lower equivalent series
resistance. And there is also a 3 mm indicator LED across the output with a 1 K resistor in series with it and the power dissipated
by the CSR which would not be much I don't think. Both should also be considered as output.

The LED's were too bright to look at directly for the entire run. And they were under driven with only 23 mA between them.
I'm thinking they can take that much current each but I'm not sure I have no specs for the LED's I have.

I think my picaxe controlled 12 volt input boost converter is more efficient though when operating at low power and boosting to
only double the voltage, it has paralleled output capacitors, dual coils and mosfets, operates at 40 or 60 kHz (can't recall exactly)
and uses Schottky 1N5822 diodes.

..

Now I'll try to fit a resistor to the input to measure the input current and determine input power at the same time as the output.
But I'm going to say the setup can be between 70 and 85% efficient depending on how it is used and built.

If folks want a circuit to suck the life out of a single cell then these booster chips make a good one.

..