This circuit will double almost any DC input voltage as well as handling plenty of current. 12 to 24V is just an example. With a few changes, it can also supply any desired output voltage.
There are many possible applications. One I can think of would be to get the 28V for the ADC connector on Gigabit Ethernet and later Power Macintosh G4s from the +12V on a standard PC ATX power supply. Basically any time you need more voltage than what you've got. Possibly on camping trips when you need 12-14 volts but only have 6V cells. Also you can use it to boost the cooling power of the fans in your computer if they can handle the extra voltage. Some can't. Keep reading for more on this. What I use it for is to increase the speed and thus the airflow of a standard 80mm DC brushless fan. This is a part of the homemade electric car heater.
The idea for this circuit came from a common type of fluorescent ballast that's in a lot of flatbed scanners. The schematic for that is here.
As power is first applied, due to differences in the resistors and the gains of Q1 and Q2, one of the transistors will turn on first and/or harder. (For instructional purposes, we will say this is Q2, but in reality it could be either Q1 or Q2.) Current will then begin flowing through Q2, the bottom 13 turn winding on the transformer, and the 1m inductor back to the +V source. Through transformer action, voltage will be induced in the 4 turn feedback winding of the transformer and begin applying more current to the base of Q2 and thus turn it on even harder. The path for this current is from +V, through the top resistor, the 4 turn winding of the transformer, the base of Q2 and back to ground or -V. This same voltage keeps Q1 turned off since its emitter-base junction is now reverse biased. At this rate, Q2 saturates very quickly. Eventually the transformer does as well and at this point, the voltage in the 4 turn winding peaks and begins to decrease. Base current for Q2 decreases along with this decreasing voltage and the collector current following this; tracing a downward curve at the transformer's resonance frequency. After crossing 0V, the voltage in the windings changes direction due to the flyback effect and Q1 turns on and takes over while Q2 is forced off. Q1 then functions just as Q2 except it drives the voltage in the transformer in the opposite direction. Q1 and Q2 oscillate back and forth in this manner. A capacitor is also placed across the 13 turn windings to smooth out the waveform and prevent any high voltage spikes. The inductor tunes out some of the high frequency generated by the 13 turn windings and keeps it away from the power source and the bases of the transistors.
The 13 turn windings act as a sort of autotransformer and produce roughly 4x the input voltage peak-to-peak. This is then rectified by a diode bridge and smoothed by Mylar (polyester) and aluminum electrolytic capacitors yielding slightly less than double the input voltage. The purpose of the Mylar capacitor is to counteract the higher ESR of the electrolytic.
For the transformer, a ferrite toroid (doughnut) works quite well and reduces RFI production. I used one roughly 3/4" OD (30mm). An E core with a bobbin works well also and is easier to wind. On the toroid, a turn is counted every time the wire passes through the hole. On the E core, it is counted each time it goes down through one leg of the E and back up through the other. Note how the feedback winding is wound in the opposite direction than the others. It can also be wound in the same direction and have its connections reversed. If the circuit fails to oscillate, try reversing the feedback winding.
One modification that might be obvious would be to add a third transformer winding and attach the diode bridge there. This will give you any desired output voltage. That would be the way to go if you were making this for your ATX PowerMac mod. To get 28V from the +12V supply, you would need around 32 turns on this winding. You will need more or less turns depending on the current drawn. Then you just connect a yellow for + and a black for negative from the power supply, tie the negative output to the negative input and the positive output is your +28V. If you're running an ADC monitor, you will probably need to use higher current transistors such as D965 which are common in camera flashes. If even those aren't enough, you may have to pull out the heavy artillery, like TIP120 or TIP3055 in the extreme and use a heat sink. Also be sure your windings are heavy enough gage to handle the extra current and your diodes are up to the task. Check your ATX power supply to see if the 12V will handle the current. If heavier gage wire won't fit on the transformer, just use a bigger core.
If you're adding a third winding, might I suggest using an E core with a bobbin to make putting those extra windings on a little easier.
One way to reduce the output voltage would be to tap the 13 turn windings closer in to the center tap and connect the diode bridge here. The voltage can be increased in a similar way by having more than 13 turns and connecting the diode bridge to the ends while keeping the collectors of Q1 and Q2 connected to taps at 13 turns. Don't use this autotransformer method or the circuit as drawn if your + or - outputs has to be connected to one of the input wires, such as in a grounding situation like the ATX PowerMac mod. In these cases you must use a third winding as described in the first second paragraph. Otherwise one of the 13 turn windings will be shorted!
I mentioned earlier that not all 12V DC brushless fans will work reliably at 24V. The way you can tell is to connect the fan to 24V and start smelling the air as it comes out. You're smelling for the odor of cooking magnet wire insulation. If you've ever blown a speaker, you know what this smell is. I would describe it as a sort of tangy smell with acrid and chemical undertones. Sometimes it will take a few minutes for the coils in the fan to heat up, so be patient. One fan I've found to work so far is the Jamicon JF0825B1M which is found in some Bestec power supplies. Most fans will not work. Also, one thing to note is that this will more than double the noise produced by the fan. In my electric car heater, this extra noise is welcomed on an amputation-producingly cold day.
This circuit has been found to work at between 5 and 13.8 volts. If you need to operate it at much higher voltage, say 30V or more, you will need to either add more than 13 turns to the coils and connect the collectors here or use larger resistors or both.
As the circuit appears in the schematic, it will handle at least 500mA at the output. Various transistors have been tried for different output currents, such as the ubiquitous 2N3904 (aka. 2N4401, PN2222, MPS2222 C945, NTE123AP) which was found deficient for all but small currents. The best transistors for running one or two suitable 12V, 120mA, 80mm DC brushless cooling fans at 24V will have a current rating of 1A and a gain of around 200-300. In an ideal situation, the transistors will have roughly the same gain. The pair I chose for my electric car heater project differed by 50 IIRC and they still work fabulous, so it doesn't have to be dead on.
One precaution on the diodes: As stated above, make sure they're ultra fast and have a reverse recovery time of 35 nanoseconds or less. DO NOT just grab a bridge rectifier or a bunch of random diodes and expect them to work. All bridge rectifiers (the ones with 4 diodes in the same package) have reverse recovery times greater than 200ns and will basically short this circuit out. The ones I ended up using were SF10?? just because there were 4 lying around. 1N914 and 1N4148 are good though for the circuit as drawn and handle around 4 amps. To give it a bit more oomph, you can use SUF30J or UF510 through UF540 which are fairly common and will boost your current handling to 15-20A. If you need even more, the big Schottkies from power supplies work.
The primary capacitor can be either polyester or polypropylene. Anything with a low ESR works. I ended up using a polypropylene because that is what was in the scanner lamp ballast; though it does get a little warm. Electrolytics are out of the question because their high ESR would overheat them at such high frequencies. The same goes for the secondary capacitor marked 470n in the schematic
For the transformer, a ferrite toroid (doughnut) works quite well and reduces RFI production. I used one roughly 3/4" OD (30mm). An E core with a bobbin works well also and is easier to wind. On the toroid, a turn is counted every time the wire passes through the hole. On the E core, it is counted each time it goes down through one leg of the E and back up through the other.
Andre de Guerin had a few ideas:
> 1) I found that for high current as well as efficiency, the primary winding of > Royer-type circuits needs to be wound with Litz-type wire. > > The best way to do this is to take four strands of 30 or so gauge enamelled > copper wire, tape one end to an electric screwdriver and the other end to the > desk. Turn screwdriver on and let it wind into a tight "strand" but don't let > it overwind. > > 2) If suitable diodes cannot be found, at a pinch the diodes used in the same > camera flashes the D965/D2504/etc are from can be used; i've had these running > at up to 1A/2500V when six are used in series. > > 3) This circuit may well work for driving strings of LED's, two in inverse > parallel should work and only a single resistor is then needed.
Transformer: 3/4" ferrite toroid or similar size E core with bobbin
