All and without me know about the problem of saving electricity. To do this, invent various schemes and use well, and not very well known effects. One of these I suggest checking out. It is based on the principles stated by Tesla, do not require scarce parts and going for one evening.
The basis of circuit solutions is the so-called kacher Brovina, but improved, or rather, redesigned to obtain the desired effect of energy saving. Spent one Watt of electrical power we can obtain several tens of Watts of power lighting.
About kachere there is a lot of information on the Internet, so its operation will not stop. The same features of the scheme presented below is a method of energy extraction and supply circuit voltage. Energy is removed by means of two diodes VD2, VD3 and ferrite winding L3 of the transformer by adding up (or, as gauvreau afirmaci — multiplying) current to power the circuit and the voltage output kachera. Of the supply voltage is slightly lower.

Ferrite transformer (L1, L2, L3) is a three wire cut, same length, twisted together and wound on a ferrite ring. Mark ferrite and its size did not play a special meaning. Many people know that afirmaci do not operate the number of coils, and meters, and instead of a flat schemes transferred to volumetric. Here you need to do also: on the wind of trifilar (three wires twisted together) to fill the space across the surface of the ferrite. For example, I have every piece of wire was about 1.5 m. Its diameter is 0.25 mm or more.
A very important point in the scheme of the LEDs. You need to apply only a powerful series GBZ or the like. Perhaps this is the "scarce" aspect of this device, because all other details are the most common. Of the capacitor C2, C3 — any ceramic, or polyester film; diodes VD2, VD3 — pulse with small recovery time and a reverse voltage of 400 or more volts; the transistor is preferably КТ805, but you can substitute any of the series; the inductor L4 is any 100..200 mcg.
First, it should be selected based on the voltage of the LEDs, or Vice versa — under voltage, to select the number of series-connected LEDs. The point is that the supply voltage must be equal to or slightly less than the voltage threshold of the ignition LEDs. If the scheme is to close the base and the emitter of the transistor, the LEDs or should not glow or glow very weakly.
The second is the AC voltage at 50Hz, rather, only the positive half-cycles. This is achieved simply by including in the power circuit rectifying diode bridge VD1. The input of the same circuit can apply the voltage directly from the secondary winding of the transformer. Please note that no smoothing capacitors should not be is important. The values of the supply voltages (given the above) can be from 5 to 30V. For large values it may be a little warm up transistor and a trimming resistor, so VT1 is preferably set to a small radiator, and R1 to take a more powerful.
If the scheme is to supply power to the constant voltage, the efficiency will decrease to normal. But is possible to do so:)
Some special configuration is not required, except for selection of number of LEDs (described above). Trimming resistor R1 is necessary to set the required brightness of the glow. Also, you can still slightly improve the efficiency of the device by selection of the capacitor C2 to increase/decrease its capacity to achieve the best resonance in L3-C2-circuit.
If you change a little the inclusion of the diode VD3 and connect it to the throttle, under heavy loads the schema can give better results.

As a result of the scheme, some of your colleagues have any questions. I will try to provide additional clarity and answer some of them.
1. Rings or cups? The scheme was tested with ferrite rings, and cups. The last wound on the above principle of maximum fill. The scheme works almost identically, the only difference is the cups, I got very buzzed and, perhaps therefore, the efficiency was here a little less (part of the energy goes into the sound).
2. The load I connected the incandescent lamp, the efficiency was low. The most important point of the map is the type of load. The scheme will have maximum efficiency only with led load; for incandescent lamps it is necessary to use completely different schemes.
3. Scheme I does not start. If done correctly — not to run she can't. From my experience I can say that you need to pay attention during Assembly:
- the correct inclusion of the windings are start and end specified in the scheme. If you turn the wrong winding diagram will not work;
- the circuit between the windings — it should not be;
- the capacitance of the capacitor C3 is better just check the tester. Its capacity should be 0.1..0.15 µf — no more, no less;
- on the throttle L4. Take your standard 100..200 mcg, or from computer power supplies. If too much or too little inductance, the circuit will work, but with lower efficiency.
4. What should be oscilogram? She looks like a "fence" of kacelnik pulses with envelope, having the shape of a half sine wave with a failure inside. The measurements were performed on the collector of transistor VT1 is:

Below shows in more detail one half cycle from the frequency of 50 Hz:

The topic was very relevant and my colleague, Alexei Lukins (http://vk.com/id33500385), sent its version of this device. If from the above schemes to remove VD2, L4, C1, and C2 and C3 to replace the electrolyte, we obtain a new graph based on the same principle: kacher + of trifilar. Its peculiarity, according to the author, is any type of load: from LEDs to incandescent bulbs. The circuit operates from 5 volts and up to the most extreme values still stand details. Trifilar dangle just the same as described above.
The experiments showed that this scheme works better with incandescent lamps, and two above — with LEDs.
This variant of the scheme was investigated by Vladimir Lukasina. Next I quote the original text and photos from the author.

The result of the research revealed that the effect of high efficiency of the device is obtained due to the intermittent series of pulses (see oscillogram). Theoretically, the efficiency of the scheme can be further increased if you get a "fence" without failure inside the sinusoids. Practically to do that yet, in any case, simple circuit solutions. I would be glad if you send us your design options: kacher + of trifilar and also, if you share your thoughts on this.
This theme got an unexpected continuation in the recovery of charges from katscher and partial restoration of battery capacity. Continued reading and leave your comments!