Lop's Blog

Dec 28, 2016 - My beastly power supply project

This is a 12 volt power supply I built shortly out of high school. It can provide 100+ amps easily, however it's unregulated. The voltage can range from 16-18 volts idle, down to 10-12 volts or less under heavy load. It's now more than 10 years old, since I built it, and has gotten very dirty. I recently started taking it apart to clean it, when I decided.. it's time to rebuild this beast, and make it regulated! This will be a challenge.

A quick bit of history, this came from some kind of test equipment used in a vo-tech school. It was labeled for use with low voltage lighting, though I never saw a bulb. The heat sink you see on top originally was mounted with 8 triac's, assuming to turn on/off or possibly dim a set of 8 bulbs. I had replaced the triac's with a set of 4 rectifier diodes rated for 40 amps each, giving me a rectifier bridge rated at 80 amps. Since I'm using an oversized heat sink, it can easily handle more than I've ever thrown at it, and can take a good beating before it gets warm. I used 4x 10 gauge solid copper wires soldered together. This was difficult to solder, even with my biggest iron. The copper is very good at pulling heat away from the source, so I'd have to pre-heat the whole group of wires before I could get solder to melt on the end. I plan to replace all these cables and connections with something better.

Here you can see most of its parts and the scale of the project. The smaller transformer mounted on the upper right side powers the control circuit. The main transformer, on the left, is much to big to simply turn on with a switch. The power surge when turning on the main transformer can damage switch contacts, so the control circuit turns on the transformer with multiple large relays. First, with a 100 ohm resistor in series with the transformer, then after a few seconds the resistor is bypassed with a second relay and goes full power.

Here is one of the original soft-start relay boards which I had modified by adding a second relay. The original delay circuit eventually died, and I never got around to building a new one. It's been working for a long time by simply turning on with no soft-start, which has worn out the relay contacts. I'm replacing this with a new relay board and all new parts. Rather than an analog delay circuit to time the second relay, I plan to use an Arduino Pro Mini to dynamically time the soft-start based on the charging voltage of the capacitors, switching to full power sooner if possible, or allowing a bit more time to charge if needed. Also, rather than the 10 watt resistor hiding at the top of the picture, I'm using a new 50 watt resistor.

Here you can see the basic parts of an unregulated power supply. The large toroidal transformer on the right steps down the voltage from 120 volts ac down to a pair of 2x 12 volt ac secondaries. This has always been wired in parallel, giving me double the current at 12 volts, but my new design will connect these in series giving me 24 volts. It will result in the max current being cut in half, but the buck converter I have planned will use the extra voltage to regulate down to a stable 12 volts and the extra voltage turned back into added current. From there, it connects into the rectifier bridge in the center, which consists of 4 diodes to convert the power from ac into dc. And finally into a capacitor on the left to smooth the ripples into a clean dc current.

Here again, you can see the scale of the project. On the right, a normal power supply, rated for no more than 1 amp at 12 volts. On the left, a transformer rated for 100 times that current, but only one part of what's needed for a dc power supply. And, in the top right, the old capacitor bank I had used. This is undersized, 4x 33,000uF 16v, but still worked well for a long time. I suspect these capacitors have been damaged slightly, due to overvoltage and the abuse I gave them, but I'll test them to see just how much they leak.

I've had these capacitors around for a while but never used them for anything. They're 161,000uF 10v each, and as a set (as it turns out) the bank is 161,000uF 20v. I've cleaned them up and replaced the messy wires with some copper pipe I hammered flat. I found the capacitors are well balanced with only one 100 ohm resistor per pair, rather than a resistor on each capacitor. As it is now, the balancing resistors waste 0.72 watts as heat. I could possibly go with higher value resistors to lower the waste heat. The only problem is trying to fit these big cans into the power supply case.

The top few are the old electrolytic capacitors I removed from the power supply. These are 16 volt, but I need to replace these with 25v capacitors if I plan to make it adjustable up to 20 volts. It's possible these have already been damaged by over-voltage over the years. Across the bottom are 3F and 100F 2.7v super capacitors. My current plan is to use 10x 3F capacitors in series resulting in 0.3F at 27v max, and 8x 100F capacitors in series resulting in 12.5F at 21.6v max. The smaller bank will be connected between the rectifier bridge and the buck converter, and the bigger bank will be after the buck converter at the output of the power supply. I'll also have banks of much smaller capacitors to help smooth the high frequency noise caused by the buck converter.

This is a schematic of the powertrain. The soft-start resistors and relays on the top left, feeds into the large transformer, series outputs to get 22 volts AC, converted to DC by the rectifier bridge, into the first capacitor banks, switched on and off many thousands of times per second by many powerful mosfets in parallel, then into large inductor coils, then finally to the last capacitor banks and the output. Not pictured here are all the extra parts needed to drive the relays, sense the voltage and current in different places, the mosfet driver chips and related parts, and the arduino microcontroller to make it all work.

Jan 12, 2017 - Planning the wiring

The transformer is made of a pair of 6awg solid copper secondaries, which connect into a large terminal block mounted next to it on the bottom of the case. The other side of this terminal block has 4x connections for each polarity, accepting up to 4awg wire. Before I recently took it apart, I had used 4x 10awg solid copper wires bundled into a cable, which is roughly rated for 240 amps max. This time, I'm using 4x 6awg bare solid copper wire, which I can solder together into one solid cable, rated in excess of 400 amps. Since the transformer secondaries will be wired in series, I'll have an estimated working current of around 50 amps, so plenty of head room, over-engineered to the best of my ability and project budget.

This cable made from 6awg wires routes over to and bolts onto the bottoms of the rectifiers mounted to the heat sink, over the transformer. The top of the outer 2 rectifiers will be bolted together with more 4x 6awg wires to form the negative output of the rectifier bridge, and the top of the inner 2 rectifiers bolt together and feed into the power board on the other side of the heat sink.

The power board contains banks of electrolytic and super capacitors, groups of large schottky diodes, and 6 IRF3205 mosfets in 3 groups. I'm using double sided 4oz copper circuit boards. With a quick google search for an online calculator, I'll need trace widths of at least 12mm, mirrored on both sides for the high current circuits, and slathered with extra solder. This should give me power losses across the board itself of less than 1 watt at max capacity. From this board will be a set of 3 outputs to the 3 large inductor coils. (I'm currently still searching for suitable inductors or toroidal cores to make my own)

The other side of the 3 large inductors then merge back together into the psu output, along with large banks of electrolytic and super capacitors. That's all there is to the high current chain of components. The weakest link will cause the greatest losses, so every connection is important.

I still tend to like drawing things on paper for some reason.

Feb 03, 2017 - Assembled the first wires

I've soldered together the first 2 cables. They're sets of 4x 6awg solid copper wire, resulting in a cable about the size of 0 or 2/0 awg. I may shrink tube these. I've emptied my solder roll on these, so will need to get more. From my measurements, I'll be no more than a foot short on wire.. so I've ordered another 10 foot coil.

I was thinking today.. I'm planning to wire the transformer secondaries in series, giving me around 50 amps at 20 volts. This cable I've just made could very easily handle 100+ amps if I connected the transformer how I used to, in parallel. The problem is the secondaries are 10 volts each. A buck boost converter would allow for a high efficiency regulated output of 14.5 volts at high current. But, this would result in very poor performance at lower output voltages. This type of circuit is also very dependent on a precise toroidal inductor. My best bet is to stick with a step-down buck converter, in 2 or 3 phase. Simple, efficient, flexible output voltage range.

More coming soon