Electronics gurus - 'joule thief' circuit...

Joule thief.

Can somebody please explain (in layman’s terms) exactly (or roughly even) how this circuit does what it does?

I remember sketchy bits about inductance from college, but it’s all a bit of a blur.

I’d like to make one of these, but I want to run it off ~3v (in a 2 cell torch) - Is this possible? (it must be, surely)
How would I go about calculating the new component values?

Coil B1 and the resistor together with the base of the transistor form an oscillator. When the base of the transistor has voltage over 0.7 volts applied to it, the transistor conducts current through coil A1. When current flows, it builds up a magentic field in A1. When the oscillator causes the transistor to stop conducting, the field collapses and generates a current in the coil. The voltage builds up from the collapsing field until it gets higher than the forward voltage of the LED (probably about three volts.) At this point, the LED conducts and discharges the coil, generating light as it does so. The collapsing field also generates current and voltage in B1, so that the circuit will continue running even when the battery voltage is below 0.7Volts. The circuit will not start if the battery voltage is already below 0.7Volts, but will keep running when the voltage drops below 0.7.
It should work just fine on two cells without modification.
BTW: This is not a replacement for a normal light bulb. White LEDs don’t put out enough light for that kind of use. This circuit just shows a way to use supposedly “dead” batteries. As such, I don’t know why you would want to run it on two cells.

Thanks for that, Mort.

I thought there were white LEDs now with a reasonably high output (I’ve seen loads of LED torches on the market, although I’m aware that none of them are going to be as powerful as a halogen filament lamp).

A couple of brief comments:

#1 That device should work with a run down battery (high internal resistance) but may fry the components if used with a new battery.

#2 LEDs have progressed to the point where they can be used for the lighting of public streets and commercial signs.

It’s just an oscillator.

Initial conditions

  • Transistor is off
  • LED is off

Then you hook the battery up. What happens?

First notice that there are three, parallel mesh circuits:
Mesh #1: R, Coil B, and the base-emitter of the transistor
Mesh #2: Coil A, the collector-emitter of the transistor
Mesh #3: Coil A, LED

When you first hook the battery up, a slight bit of current (a few micoramps) will flow almost instantaneously through Mesh #3. This is due to an extremely short time constant in this mesh. Note that, even though there is a slight amount current, the LED will not illuminate; this is because the supply voltage is too low. And because the transistor is off, no current flows in Mesh #2.

Now let’s look at Mesh #1. This mesh has a longer time constant vs. Mesh #3 due to the 1K resistor and voltage drop across the transistor’s base-emitter junction. But eventually the current will increase to the point that the base-emitter junction will reach 0.7 V, at which point the transistor turns on.

When the transistor turns on, a huge & abrupt rush of current flows through Mesh #2. (The transistor effectively shorts-out the coil). Coil B “senses” this. As a result, a voltage spike is induced across Coil B. This voltage spike directs current in the opposite direction in Mesh #1, and abruptly shuts off the transistor.

When the transistor shuts off, it interrupts the (relatively large) current that was flowing through Coil A. Because v = L* di/dt, a relatively large voltage spike appears across Coil A. This illuminates the LED.

The LED eventually “dies out,” and the process starts over…

The on time is short enough that it shouldn’t be a problem for the transistor, and the LED current is limited to the energy that can be stored in the coil. Since that isn’t much, I don’t see a problem for the LED either. Since white LEDs have a forward voltage of over three volts, you shouldn’t have any trouble even if you have two batteries in series.

Yes, red and green LEDs. The blue LEDs haven’t been on the market as long and haven’t reached the efficiency of the others. White LEDs are actually blue LEDs set in a phosphor compound, so they are less efficient even than the blue ones.

Max collector current for that transistor is stated as 100 mA and with the circuit supplied at 3V I believe it can easily be exceeded. The characteristics of the ferrite transformer and the entire circuit will have a huge range of variation so it is difficult to say. In any case, I do not believe it is an efficient circuit to be used just as a current limiter when sufficient voltage is available. It only makes sense to me for the express purpose of using batteries which would be otherwise useless.

We had a thread not long ago about ICs used for switching power supplies for circuits just like this but I was never any good at searching here and I am traveling and have limited access (on a broken laptop).

Wouldn’t allowing a cell to run down that low drastically increase its leaking risk? (Not that I wouldn’t be guilty of replacing my flashlight bulbs with smaller ones when they get dark. :))

Dunno= there nust be non-flashlight applications (I’m thinking quartz clocks and smoke alarms etc) that drain cells more or less completely dry before they stop working, although they would do so very gradually, which might make a difference.

Yup, I have some LCD clocks, thermometers, etc, which us button type batteries which are very expensive but I just use AA batteries which are almost dead. They still have enough energy to power these devices for many months, several years in some cases.

Hah!
I’ve got an LCD clock that’s been running for ten years on one AA battery.

Seems that, if I want to make an LED torch, it might be better to use bright red, green and blue LEDs in a cluster (although I realise this will result in incomplete mixing).

I am not sure of what you are trying to accomplish but LED “torches” (I can’t help but giggle) are sold commercially if you want to buy them.

yeah, but the really good ones aren’t widely available here in the UK yet (and it’s going to cost me $20 plus import tax to get one shipped over).

My experience as a tinkerer is that in the end it is not worth the time and the effort if the product is commercially available. At any rate, do a search for high flux leds
http://www.google.com/search?hl=en&lr=&ie=UTF-8&oe=UTF-8&q="high+flux"+led
http://www.gilway.com/pa-superfluxleds.html
http://www.ledtronics.com/datasheets/Pages/spiderled_hi-flux_led_lamps/40.htm

Thanks

The search is made no easier by the vague terms that manufacturers seem to have chosen for these components:
High intensity
Super bright
Ultra bright
Hyper bright
Mega bright

None of these are based on any objective standard.

In the fine print of one of the ‘LED Torches’ I saw, IIRC said that the LED isn’t ‘on’ but flashing at a high enough frequency that the eye perceives it as ‘on,’ and the flashing extends the life of the batteries. Sounds good to me.

that’s my 2 cents anyway

Mangetout, the spec sheets will give you all the data you need: lumens @volts, amps, etc. Check out the data sheet in the last link I posted

I hope nobody minds me resurrecting this thread; I just finished making my first ‘joule thief’ tonight - I built it into the empty shell of a krypton flange(band name?) bulb - my white LEDs are still on order, so this one uses yellow ones - they are superbright (something like 7cd) and I’ve connected three of them in parallel instead of just the one as shown in the link above.

The result is remarkably bright and when installed in a torch (I had to open out the hole in the reflector a little to get it in - a cluster of three 5mm LEDs is just a touch wider than the original glass envelope) it is very bright - easily bright enough to be usable as a torch

The LEDs draw 50mA each, AA batteries are about 1,500 mAh, so allowing as much as 50% or so for inefficiencies, I reckon (and this is a pretty uneducated guess) it may give me 5+ hours of usable light on 2 fresh AAs.

I forgot to mention that I also substituted a BC184 transistor for the BC549 - it’s in the same family and the specs are quite similar - seems to work OK anyway.