First of all let me say that I do not recommend that anyone builds this charger. Manfred Mornhinweg has published a much better design on his website and also explains some of the considerations when designing a power supply like this. Having said that some of the feature in my design may be of interest so I'm going to put it here anyway.
All images on this page are low resolution versions. Click them for bigger versions (may take some time to load) - actually at the moment there are very few pictures, I need to get my camera out and take some more.
The transformer used in this power supply was made from a rewound microwave oven transformer (as used in the plasma cutter). This one has several important differences but is not as nice and if I were to make it again I would use a similar method to those in the plasma cutter. Also, the trike charger uses a similar transformer and rectifier but is completely unregulated.
One of the problems associated with microwave oven transformers is that, due to core saturation, they take very high magnetising currents and get hot even at low loads. To improve the efficiency of this power supply additional turns were added to the transformer primary and bring the core out of saturation. The number of turns needed was determined simply by adding turns and measuring the current drawn until it reached a sensible level, from memory this was about an extra 50 turns. The wire used was from the primary of another similar transformer and was actually made of aluminium with coloured enamel to make it look like copper. I joined this to the existing primary and to the copper power wires with crimp connectors. The transformer had two secondary windings of 1.5 mm2 pvc insulated wire with an output of around 18 Volts, and an extra winding of small pvc insulated equipment wire which supplys a few milliamps to some of the control circuits. This final winding wasn't really needed but since I was winding my own transfomer I included it anyway. I could have instead connected a voltage doubler to one of the other windings.
The output from the two transformer windings each go through a 16 A fuse to two 35 A metal-clad bridge rectifiers. These are mounted on a homemade aluminium heatsink and two are used since the 35 A rating of these rectifiers is rather optamistic. Having entirely separate transformer windings like this ensures that the current is shared equally without needing any extra series resistors. The reason for the fuses is not to protect against overcurrent but rather against a dead short somewhere between the transformers and the regulator section, for example if some piece of metal found its way inside the case. 50 mF of smooting capacitors are included which is just about enough to remove the ripple at high output currents but given the nature of the connected equipment (lights, motors, a few bits of fairly robust electronics) and the fact that a 12 V lead-acid battery will be connected across the output most of the time when in use more didn't seem necessary.
A 5.6 V zener diode with a 10 V prestabiliser forms the voltage reference. The prestabiliser minimises the effect of changes in current through the diode by giving it an almost contant vltage to the 330 ohm resistor. Zener diodes in the region 5-7 V tend to work by both the tunnelling and avalanche effects which have opposite temperature coeficients and hence the temperature coeficient of a 5.6 or 6.2 Volt zener can be made very small. This is probably still not as good as an IC regulator.
An op-amp compares this reference voltage with a divided down version of the output voltage, the voltage being adjustable over a range of several volts by the 10 kilohm pot. The output of this op amp drives two power darlingtons via a BFY50 transistor. Since the output transistors are configured as emitter followers the op-amp must be able to output at least three base-emitter voltage drops higher than the desired output voltage To do this it is supplied with a regulated 27 volts via the additional transformer winding mentioned earlier and a small rectifier. The power transistors have 50 miliohm emitter resistors to ensure the load current is shared equally between them and are mounted on a substantial aluminium heatsink. A current limit is provided by the 25 miliohm sense resistor and the BC548 transistor. When the current reaches slightly over 25 Amps the transistor turns on conducting some of the base current for the BFY50 transistor to ground and reducing the output voltage whilst also turning on a warning LED.
This point may also be pulled to ground by the thermal protection circuit which uses an op-amp as a comparator to detect when the resistance of a thermistor mounted in the heatsink drops below a speficied level. The diodes in series with the output of the op-amp ensures that the transistor only switches on when the voltage at its output rises significantly above ground and the 100 k ohm pot provides an adjustable amount of hysteresis. An identcal circuit was later added to switch on a fan at a somewhat lower temperature and the circuit was rearranged as shown in the diagram below so that the thermistor was on the negative side of the divider allowing positive feedback to be applied to the reference divider instead. This was needed to allow hysteresis on both fan and overheat circuits without them interacting with each other.
Multiple 0.1 ohm resistors were used in parallel for the current sensing and load sharing resistors, the power supply was then mounted in a wooden case with the heatsink replacing one wall. The fan is a small shaded-pole unit from a microwave oven operated by a relay and sucks air into the case much of which leaves through the output transistor heatsink and the remainder through holes drilled near the other components which need cooling. The heatsink consitst of a thick slab of aluminium to which the transistors and the thermistor are attached (the thermistor is mounted in a hole between the transistors by means of epoxy putty) attached to several sections of finned heatsink from an old cpu cooler and a homemade twisted-vane heatsink from a piece of aluminium angle. The transistors are not electrically isolated from the heatsink which is not a particular problem in this application.
The finished power supply was set to produce an output of 14V and found to drift by a fewtens of millivolts as it warmed up, since it will mostly be on for long periods the voltage was adjusted to be correct after the supply had reached operation temperature although the smal variations were not particularly important. The output was then loaded with an assortment of 12 V loads and the fan-on and overtemperature trip points set by the crude but effective method of “if I can't put my hand on it the fan should probably be on” and “if water boils it's probably hot enough it should shutdown.” The fan probably comes on at about 60 °C and the overheat sensor at about 95 °C. With these set up it was safe to test the current limit (which results in the output transistors dissipating the most amount of power).
The fan normally starts to come on with a load of around 6 to 8 Amps, pulsing on for a few seconds and then off for a minute. Even with the output heavily overloaded and the output dropping below 8 V the fan does not need to come on continuously to keep the transistor cases below the shutdown temperature although I have never actually tried into a dead short. Disabling the fan causes the shutdown circuit to trigger and turn off the power supply until the outpur transistors cool. One minor flaw is that when the supply shuts down it removes power to the fan relay and the fan stops making the cool-down take longer than it otherwise would. This is probably best remedied by switching the fan with an opto-triac instead of the relay but with a battery connected accross the output it's not usually an issue as that supplys the 100 mA or so the relay needs.
Return to main page