Joule theif

[evillow]

Ok so I've always played with electronics on the minor repair like changing the bad capacitors in a bad LCD tv but my son wanted to learn more and for me to help him I needed to learn more so I found this thing called the "Joule Thief" figured it would be a fun thing to make with him and I had misc stuff laying around that i could recycle some parts so I found some pnp transistors a couple adjustable resistors got some led's, a battery holder and breadboard from radio shack so i find some inductors in an old sony tv and i tried making my own with some torroids i found in an old computer and ive tried making this thing as ive seen in videos and my led's lights up.
Now here's my problem my battery has 1.01 volts and my led that's voltage is 3.6 lights up but when I measure the voltage at the led I'm still getting 1.01v  instead of 3v like I've seen in the videos the transistor I am using is c3311 and I'm also using a 100k ohm variable pot.

[ataradov]

Joule thief generates short pulses of current, your multimeter may not be able to accurately measure them, especially in DC mode.

[evillow]

So that means that when I lit up the blue led and the pink led both being 3.6v I was successful my multimeter just isn't reading the short burst I did but a cheap cen-tech p37772 multi meter

[ataradov]

None of the multimeters will provide a totally correct reading, since there is no good definition of correct here. They are short pulses, which value should it show? Maximum? Minimum? Averaged in some way?

[Zero999]

You really need an oscilloscope to measure the pulses. It will also give you other information such as the frequency, peak voltage and duty cycle. The trouble is, oscilloscopes aren't cheap and even the cheapest ones are a fair amount of money.

[evillow]

Well I'm going to have to do some saving so I can get one for our electronic adventures. I probably should get better multimeter also

[T3sl4co1l]

This is the basic circuit you're looking for:



(Or one of the special-purpose all-in-one ICs that does the same basic thing.)

The average circuit you find on the 'net has all the resistance in Rb and no Rbias and Cbb.  A gentle reminder that, just because a thousand parrots squawk the same tune, doesn't mean it's "right"...

Best efficiency is usually found with zero Rb, Rbias sufficient to get the current flow you need, and Cbb sized so the Rbias * Cbb time constant is around the inductor time constant.

What's the inductor time constant?  Well, here's how it works.

As Cbb charges, the transistor begins to turn on, then positive feedback through the transformer makes it slam on.
 Switched on, current charges up in the inductor (the fundamental inductor equation is V = L * dI/dt, so when we apply a fixed voltage V, current rises by dI every dt seconds).  Meanwhile, base current is initially large, but tapering off (because the base winding is drawing current from Cbb, discharging it).  Eventually, there's not enough base current, and too much collector current, and the voltage swings up.  At that instant, base voltage drops sharply, turning the transistor off.

After turn-off, the collector voltage swings up until it hits something.  In this diagram, you'd probably see a spike of 30V or more, and a very hot transistor.

Normally, you put an LED from collector to GND, so the voltage jumps up to a well defined ~3.6V and the current discharges into the LED, lighting it.

As mentioned, it would be difficult to measure the voltage and current in an LED driver.  We can address this by adding extra steps:



On the left, Rbias has been replaced with a PNP transistor (this, and the other PNP transistor, are optional for demonstration purposes).  Recognize Cbb, the NPN, transformer and Rb are in the usual connection.  A schottky diode takes the inductor's discharge current and directs it into a 10uF capacitor, which smooths the pulses into DC.  (the 1k and Vref diode are also part of the control in this circuit, and are optional.)

To measure the DC performance of this circuit, measure the voltage and current at Vin (note: you need a 10uF capacitor from Vin to GND as well).  Measure the voltage and current out of Vout, when loaded with a suitable value resistor, or some LEDs.

The schottky diode costs a little efficiency, but not too badly: a 20V low-loss schottky (I'd recommend a PMEGxxxx) will drop less than 0.5V, not a bad price for getting a stable reading.

Tim

[alsetalokin4017]