A LED torch is one of the simplest projects you can build and it's very interesting as it uses a super-bright white LED.
In the history of LED production, red LEDs were the first to be invented and their output was so dim you could barely see if they were illuminated. You needed a darkened room to see them at all.
Then came green, yellow and orange LEDs.
As time went by, the brightness improved and it came to a point where the output would shine into the surrounding air. These were called Super-bright LEDs.
Then came the blue LED. At first it was dull, but gradually the output increased to a dazzling glare.
With the combination of red, green and blue, manufacturers had the potential of producing a white LED.
This was the dream of all LED manufacturers.
Since the illumination produced by a LED comes from a crystal, it is not possible to produce white light from a single crystal or "chip." The only way was to combine red, green and blue. As soon as the output of blue came up to the quality of the other colours, a white LED was a marketable product.
The more-recent way to produce while light is to illuminate a blue LED and surround it with a yellow phosphor coating. The yellow and blue combine to make white light. This is called additive mixing of colours.  
White LEDs are now with us and their output makes them a viable alternative to the globe.
There is an enormous array of LED torches on the market, from $2.00 "give-aways" to $200 "rip-offs."
Although a LED torch is passable for illuminating an area, it certainly does not have the illuminating capability of a $10 lantern, using a 6v battery.
A LED torch is more of a "fun-thing" to see how far LEDs have come in the past few years and see what can be done with a single cell and an handful of components.
When we first decided to produce a LED torch project, we wanted to fit the circuit into a 2-cell torch but a white LED requires about 3.4v to operate, and two cells produce only 3v. So we had to think of a number of ways around the problem. That's why we have produced a number of circuits.
As you know, a LED will not operate on a voltage below its characteristic voltage. It simply will not operate AT ALL.
This characteristic voltage depends on the type of LED and is about 1.7v for a normal red LED, while a super-bright LED is about 3.1v - 4v.
The exact characteristic voltage varies with the colour, the intensity of the LED, the current flowing and the way it is manufactured. This feature cannot be altered after it is manufactured and the EXACT voltage must be delivered, otherwise the LED will be not work or if the voltage is higher, it will be destroyed. This is the cold, hard fact. The supply voltage must exactly match the characteristic voltage.
This sounds a difficult thing to do, but a simple solution is to add a resistor in series and the voltage across the LED will sit at the exact value required by the LED, while the extra voltage will appear across the resistor. According to Ohm's Law, a current will flow though the resistor and this will also flow though the LED. This applies when the circuit is supplied with a DC voltage.
All we have to do is create a voltage higher than 3.4v and we can drive one of the latest SUPER-HIGH-BRIGHT white LEDs with a single cell, using a step-up-voltage circuit.
This will produce a series of pulses to the LED and the brightness will be slightly higher than if a steady DC voltage is applied. These are the things we will be covering in this project. 
 

JOULE THIEF
The web is filled with circuits similar to "CIRCUIT A" below.
Here are 3 circuits:

Although they work, the performance and efficiency can be increased an amazing 300% by simply adding a capacitor.
We will look at the Joule Thief circuit and show the improved design.

CIRCUIT A
The first circuit in this discussion is the simplest design.
It consists of a transistor, resistor and transformer, with almost any type of LED. The circuit will drive a red LED, HIGH BRIGHT LED, or white LED.
The circuit produces high voltage pulses of about 40v p-p at a frequency of 200kHz.
Normally you cannot supply a LED with a voltage higher than its characteristic voltage, but if the pulses are very short, the LED will absorb the energy and convert it to light. This is the case with this circuit. The characteristic voltage of the LED we used was very nearly 4v and this means the voltage across it for a very short period of time was 4v. The details of the transformer are shown in the photo. The core was a 2.6mm diameter "slug" 6mm long and the wire was 0.95mm diam. In fact any core could be used and the diameter of the wire is not important. The number of turns are not important however if the secondary winding does not have enough turns, the circuit will not start-up.

HOW THE CIRCUIT WORKS
The transformer is configured as a BLOCKING OSCILLATOR and the cycle starts by the transistor turning on via the 2k7 base resistor.
This causes current to flow in the 60-turn main winding. The other winding is called the feedback winding and is connected so that it produces a voltage to turn the transistor on MORE during this part of the cycle.
This winding should really be called a "feed-forward" winding as the signal it supplies to the transistor is a positive signal to increase the operation of the circuit. This is discussed in more detail in Circuit Tricks.
This voltage allows a higher current to flow in the transistor and it keeps turning on until it is saturated.
At this point the magnetic flux produced by the main winding is a maximum but it is not expanding flux and thus it ceases to produce a voltage in the feedback winding. This causes less current to flow into the base of the transistor and the transistor turns off slightly.
The flux produced by the main winding is now called collapsing flux and it produces a voltage in the feedback winding of opposite polarity. This causes the transistor to turn off and this action occurs until it is completely off.
The magnetic flux continues to collapse and cuts the turns of the main winding to produce a very high voltage of opposite polarity.
However this voltage is prevented from rising to a high value by the presence of the LED and thus the energy produced by the collapsing magnetic flux is converted to light by the LED. 
The circuit operates at approx 200kHz, depending on the value of the base resistor and physical dimensions of the transformer.
The circuit draws 85mA from the 1.5v cell and the brightness of the LED was equivalent to it being powered from a DC supply delivering 10 - 15mA.

Before we go any further, there are a number of interesting circuits on the web.
The following two circuits need explaining. The first circuit is identical to our "Circuit A" except the design engineer did not do his homework. He only added 8 turns to the 100uH inductor and found the circuit did not start-up. His solution was to add another transistor and tie the base to the collector. What a waste of a transistor!
The second circuit is a very inefficient design. The second transistor is being turned on via a 1k resistor on the collector of the first transistor and when this "turn-on" current is not required, it is being shunted to "deck."
Our circuit uses the "oomph" of the secondary winding to saturate the transistor and this produces the highest efficiency.

Here is a circuit from one of the major chip manufacturers:

Apart from the circuit being enormously complex and expensive, 62mA is too high for many white LEDs. The maximum current must be kept to 20 - 25mA.

The first "poor design" got me thinking. Maybe the signal at the transformer end of the 220R needs to be stabilised to improve the performance of the circuit. I tried a transistor and it did not work.
But I actually thought of placing a small capacitor at the join and taking the other end to the 0v rail. This will allow rail voltage to enter the feedback winding of the transformer but prevent the signal generated by the winding being lost through the 2k7 resistor.
The following circuit is the result:

The brightness of the LED did not alter but the current changed from 85mA to 28mA.
The circuit instantly became 300% more efficient.
I could not believe it.
When I put the CRO across the LED, I realised why. The frequency of the circuit changed from 200kHz to 500kHz. The LED was getting more than twice the number of pulses per second.
That's why you cannot trust anything or anyone. This improvement has never been presented in any circuit on the web. Obviously no-one has done any experimenting at all. 
If the brightness of the LED is equal to a DC voltage of 4v and a current of 10mA, the circuit we have produced is slightly more efficient than delivering a DC voltage to the LED, even though there are some losses in the transformer and transistor.
This proves the fact that LEDs driven with a pulse, are more efficient than being driven by a DC supply.

Here is a photo of Circuit B constructed by a reader. He used a toroid (a circular magnetic circuit - or ring) and this has lower losses because the magnetic flux does not emerge (come out of) the end of the core. The magnetic flux keeps circulating. However if the flux is not very high, it does not saturate the core and there are no losses and the slug performs almost the same as a toroid.

You can clearly see the number of turns on the toroid.
The circuit is not very efficient because it does
not have a capacitor to improve the efficiency.

BIKE FLASHER - Amazing!
This bike flasher uses a single transistor to flash two white LEDs from a single cell. And it has no core for the transformer - just AIR!
All Joule Thief circuits you have seen, use a ferrite rod or toroid (doughnut) core and the turns are wound on the ferrite material. But this circuit proves the collapsing magnetic flux produces an increased voltage, even when the core is AIR. The fact is this: When a magnetic filed collapses quickly, it produces a higher voltage in the opposite direction and in this case the magnetic field surrounding the coil is sufficient to produce the energy we need.
Wind 30 turns on 10mm (1/2" dia) pen or screwdriver and then another 30 turns on top. Build the first circuit and connect the wires. You can use 1 or two LEDs. If the circuit does not work, swap the wires going to the base.
Now add the 10u electrolytic and 100k resistor (remove the 1k5). The circuit will now flash. You must use 2 LEDs for the flashing circuit.

BIKE FLASHER - AMAZING!
THE IMPROVED BIKE FLASHER CIRCUIT

The original 30 turns + 30 turns coil is shown on the right. The circuit took 20mA to illuminate two LEDs.
The secret to getting the maximum energy from the coil (to flash the LEDs) is the maximum amount of air in the centre of the coil. Air cannot transfer a high magnetic flux so we provide a large area (volume) of low flux to provide the energy. The larger (20mm) coil reduced the current from 20mA to 11mA for the same brightness. This could be improved further but the coil gets too big. The two 30-turn windings must be kept together because the flux from the main winding must cut the feedback winding to turn ON the transistor HARD.
When the transistor starts to turn on via the 100k, it creates magnetic flux in the main winding that cuts the feedback winding and a positive voltage comes out the end connected to the base and a negative voltage comes out the end connected to the 100k and 10u. This turns the transistor ON more and it continues to turn ON until fully turned ON. At this point the magnetic flux is not expanding and the voltage does not appear in the feedback winding.
During this time the 10u has charged and the voltage on the negative lead has dropped to a lower voltage than before. This effectively turns off the transistor and the current in the main winding ceases abruptly. The magnetic flux collapses and produces a voltage in the opposite direction that is higher than the supply and this is why the two LEDs illuminate. This also puts a voltage through the feedback winding that keeps the transistor OFF. When the magnetic flux has collapsed, the voltage on the negative lead of the 10u is so low that the transistor does not turn on. The 100k discharges the 10u and the voltage on the base rises to start the next cycle.     
You can see the 100k and 1k5 resistors and all the other parts in a "birds nest" to allow easy experimenting. 
This is the first circuit you should build to flash a white LED from a single cell.
It covers many features and shows how the efficiency of a LED increases when it is pulsed very briefly with a high current. 
The two coils form a TRANSFORMER and show how a collapsing magnetic filed produces a high voltage (we use 6v of this high voltage).
The 10u and 100k form a delay circuit to produce the flashing effect.
You can now go to all the other Joule Thief circuits and see how they "missed the boat" by not experimenting fully to simply their circuits. That's why a "birds nest" arrangement is essential to encourage experimenting. 
Note: Changing the turns to 40t for the main winding and 20t for the feedback (keeping the turns tightly wound together by winding wire around them) reduced the current to 8-9mA.

The circuit can be made small by using a ferrite slug 2.6mm diam x 7.6mm long.
The inductance of this transformer is quite critical and the voltage across the LEDs must be over 6v for the circuit to work. It will not work with one or two LEDs.

2-TRANSISTOR CIRCUITS
The circuits we have presented above use a single transistor and a transformer to provide feedback. This feedback is a form of REGENERATION to turn the transistor on HARDER and HARDER to produce the maximum efficiency.
An oscillator can be produced with two transistors and an inductor, but there are some design-features that need to be applied to produce an efficient circuit.
The first circuit is a POOR DESIGN.

Poor Design - see text

In the circuit above, the base current is constant and will be very small through a 10k resistor.
Base-current is effectively wasted current or "wasted energy" and should be kept to a minimum.
The circuit consumes 10mA and the LED will see less than 4mA.
By reducing the 10k base bias resistor to 470R the circuit-current increases to 25mA but the LED is still not at full brightness.
Secondly, the base-current is shorted to the 0v rail via the first transistor and is completely wasted during part of the cycle.
But the main problem with the circuit is the fact that the driver transistor is not driven into full conduction at any part of the cycle and the circuit has very little efficiency. 

To solve this problem, the two transistors are connects so the "turning-ON" is provided by a transistor and it effectively reduces in resistance to a small value to turn ON the driver transistor.
Theoretically a current-limiting resistor should be added in the base of the driver transistor (about 47R) but this made no difference to the current taken by the circuit.

2-Transistor Joule Thief Circuit

The circuit turns ON via the 220k resistor and the voltage on the collector of the NPN transistor drops to nearly 0v. This action causes current to flow through the inductor and at the same time the 1n capacitor is brought towards the 0v rail and this turns ON the first transistor slightly harder. This action continues until the driver transistor cannot be turned on any more.
The 1n charges a little more and the current through the base lead reduces slightly. This action turns OFF the first transistor slightly and the driver transistor is turned OFF a slight amount.
The voltage on the 1n rises and very soon both transistors are fully turned OFF.
The magnetic flux in the core of the 1mH inductor collapses and produces a voltage in the opposite direction.
This voltage is added to the 1.5v rail voltage and the final voltage is high enough to illuminate the white LED.  
This keeps both transistors OFF and when all the magnetic flux has been converted to energy to illuminate the LED, the voltage on the collector drops. This lowers the top plate of the capacitor and since the capacitor is slightly charged, the bottom plate drops to a voltage less than rail voltage. This action turns ON the first transistor to start the next cycle.  


Another 2-Transistor Circuit:

This circuit is designed for mass production as the choke is a standard 33uH.
It can be made by winding 50 turns of 0.1mm enamelled wire on a 1.6mm diam ferrite slug.
 

With this we turn to a surface-mount chip that has been designed to carry out the exact same task as circuit B. The chip is called PR4401. The following is the promotion advert for the chip:

I could not find any sales literature on the internet, but the manufacturer requires 9,000 pieces to be bought at a cost of 36 cents per piece. This comes to $3,240 if you want to incorporate it into your project.
I have described the pro's and con's of this chip in another article "Circuit Tricks" and you should read it and work out what they really mean.

LED TORCH - JOULE THIEF - INDUCTOR TESTER
Kit of components $3.00  plus $4.00 postage

An equivalent IC (chip) has come on the market for 10 cents and it is a better chip.
Here is the circuit for QX5252F:

Using 220uH, the circuit takes 13mA an illuminates 2 white LEDs very brightly.
Using 100uH the circuit takes 30mA and the LEDs are really the same brightness.
Using 33uH the circuit takes 80mA and the LEDs are just about the same brightness.
Obviously the 220uH creates the most efficient circuit.

Here is the prototype:

The kit comes with a PCB, all parts: QX5252 Chip, tactile switch, 1.5v button cell, tinned copper wire for cell and heatshrink for cell-cover, 100uH inductor, very fine wire and 1M to make your own 100uH, 2 machine pins and length of fine solder, but only 2 LEDs and not the change-over switch.  All for $7.00 posted.
Email Colin Mitchell for details on buying the kit.

The prototype has been built on Matrix Board and shows the change-over switch used to test different LEDs.  You will get the Printed Circuit Board in the kit that has been generated from the layout. This is the easiest and simplest way to make a PCB and avoid any mistakes.

The inductor has been fitted via machine pins and it can be removed and different inductors fitted to see the results. A machine pin is hollow and allows to poke the ends of a conductor into the pin and it will make contact.
The current taken by the circuit changes according to the inductance and this will enable you to compare inductors and even find the value of an unknown inductor.
You need to use one, two or three known inductors and make a table of the inductance and current taken by the circuit.
The current may or may not be linear but we measured inductors from 33uH 1,000uH and recorded currents from 80mA to 2.5mA.
This enabled us to measure unmarked coils.

HOME-MADE INDUCTOR
One more feature of this project is to wind your own inductor and see if it is effective as the 100uH supplied in the kit. 
The kit comes with fine wire and a 1M resistor.
Wind 250 turns on the resistor very carefully and as you come to the end of the winding, you can criss-cross the wire over the other turns to keep them in place.
Leave at least 4cm of wire at the beginning and end.
Now heat the wire very close to the body of the resistor with a hot clean soldering iron that has been fully tinned. The wire will get tinned very close to where it comes from the winding. Now wind 5 turns around the wire coming from the resistor and solder it in place. Break off the wire.
Do the same with the other end.
Fit your home-made inductor and the LED should be as bright as the 100uH.
Measure the current across the switch. It should be about 32mA.
The voltage is being converted from 1.5v to 3.5v and each LED will get slightly less than 6mA by the time you take the efficiency of the circuit into account.

Here is the waveform:

 What is happening?
The first part of the cycle shows the inductor being pulled down to the 0v rail.
This means the 1.5v for the battery is directly across the inductor and current starts to flow in the winding.
This produces magnetic flux (called EXPANDING FLUX) that cuts all the turns and produces a voltage in the turns that is opposite to the incoming voltage.
This means the effective voltage entering the inductor is very small and thus a small current flows. However enough current flows and enough time is allowed so the inductor produces magnetic flux.
This is shown in diagrams A and B.
The circuit then immediately turns OFF and the magnetic flux collapses very quickly.
This is shown in diagram C. This voltage is actually in the opposite direction to the original voltage and this is one of the most important things to understand.
The voltage produced by the inductor will be very high (possibly 10v or more) (and is added to the voltage of the battery. But as soon as the voltage reaches 3.2v, the white LED starts to turn ON and produce illumination.
The magnetic flux keeps collapsing and supplying energy to the LED for about 2uS and when the voltage falls to less than 3v, the LED turns OFF.
The QX5252 IC turns ON again and pulls the inductor to the 0v rail to start the next cycle. The IC operates at about 130kHz and if the inductor has not lost all its magnetic flux, it will add to the flux on the next cycle. 
You can see the inductor only has to produce about 1.7v as the voltage is "produced" on top of the 1.5v from the battery. The circuit will work down to about 0.9v
Understanding and interpreting a waveform is very important because the LED is only turned ON for a very small portion of the time but it is turned ON very brightly.
Our eyes detect this brightness and hold this brightness while the LED is completely turned OFF during the rest of the cycle. This effect is called PERSISTENCE OF VISION.
This is why a very small current will produce high brightness and create a highly effective circuit.
That's how the circuit got its name: Joule Thief. It appears to get energy from nowhere. But we know a LED can operate in pulse-mode and product a very high brightness while consuming a small OVERALL current.
This project will teach you 3 things:
The number of turns to create 100uH inductor, testing different inductors, testing different LEDs and determining the value of inductance by measuring the current.


 oooooooooooooooooo000000000000000000000000000oooooooooooooooooo

When you build circuit "B," you will realise the specifications given in the .pdf for the chip, could be improved. We have achieved a supply current of 18mA for an equivalent brightness of 10mA. The chip requires 25mA. So, all the technology in the world has not surpassed a hand-made circuit.
The advantage of our design is the ready availability of components and you can change them to suit your own application.
If you want to increase the brightness, the 2k7 can be reduced to 1k5.
If you want to drive 2 LEDs, they can be added in series: