I have three 65-70W power supplies at roughly 20V. I want to combine them into a single 200W source that need only be at 12V, though higher is fine. There are two ways I know of to go about this:
Use some nice 5A 25V min rated diodes (the math works out to max 3.5A of current on each channel), accept the 0.7V drop and live with it. If I’m remembering correctly diodes use the normal I * Vdrop = W heat law and in this configuration might be dissipating as much as 14W of heat a piece. This is far too much.
Use an LTC4412 “low-loss” controller, and a suitable 5A 25V min rated MOSFETs and try to build an ideal diode circuit.
Either way is doable, the first being an order of magnitude simpler. My only questions are will the second dissipate less heat or does anyone with more EE know-how have a method which will dissipate less? I’m not as concerned with losing a few watts of power or having a voltage drop, etc. as I am about dissipating extra heat.
The reason why, is that the power supllies themselves do not have any active cooling, and the device this will go power will not have any active cooling either. In case this helps, the power usage of the device will be roughly <65W most of the time, but occasionally will shoot up to just under 190W, possibly for days on end. It hopefully will be running 24/7/365.
A prebuilt fanless PSU which can output 200W at somewhere between 12-26V, fits in a 2"x7"x7" box, and costs less than $100 would work also.
Hmmm… 65W at 20V is 3.25 amps. BTW, are those supplies rated for an output of 65W, or an input of 65W?. If it’s the latter, then their output is going to be more like 2 amps or less.]
Times 3 is 9.75 amps.
200W at 20V is 10 amps.
If you use normal diodes to combine the supplies, there is a drop of about 0.7V, and a current of say 3.5A, for a power dissipation of 2.45W each (not 14W). You should use diodes that can be mounted on a heat sink.
But if I were you, I’d use high-current Schottky diodes, to reduce the drop to the 0.4V range (or less). Then only a minimal heat sink would be needed.
The ‘ideal diode’ approach will not dissipate any significant heat in the diode itself. But the burden shifts to the supplies - they must be set very close to the same output voltage, so that they will balance their outputs based on the inherent impedance of the supplies. You’re better off using simple (Schottky) diodes, and let the diodes help with the balancing problem…
Keep in mind that your supplies will dissipate non-trivial heat, based on their efficiency. At 190W of output, even an efficient switching supply will waste perhaps 20-40W of heat – that’s a LOT OF HEAT in a 2 X 7 X 7 box! I think you need a much bigger box, and/or you need a fan.
Also, the supplies need to have built-in current limiting. Otherwise, the inevitable out-of-balance situation will cause one supply to overheat and self-destruct, and the others will soon follow.
a small resistor, say 0.22 ohms, in series with each power supply output should help with the out of balance thing. Between the resistor and a shottkey diode (such as a 1N5822) in series with each output you will loose upwards of a volt off of each output at peak draw of 3A. And Pete is right on the money about all 3 supplies needing to be closely matched for output voltage or one will deliver until its current limit kicks in, unequally stressing it much harder than the other two.
Thanks for the help. I think I used 20V * 0.7V for my original calculation which is very wrong. Redoing it correctly makes it seem the heat dissipated from the diodes is not that bad. I remember trying to find a 0.4V drop high current diode and having trouble, but I’ll look harder. The diode you cited seems good (low loss), but only rated for 3A.
As far as balancing and current limiting. I know the power supplies are limited and those wattages calculated are output. I honestly don’t know how close the power supplies are, but will do some testing at various loads. I also can see exactly which of the three are providing current and how much and such when I actually assemble this. Besides a small resistor any general tips about how to avoid an out of balance situation?
Also as far as the heat output, the max room for the power supply assembly is 2x7x7, the whole device is much, much bigger. Unfortunately a fan is not really possible. However, large custom heatsinks are.
For diodes, a starting point is the MBR3xxx series and MBR4xxx series (those are Motorola part numbers, but they’re made by various outfits). There are lots to choose from. Some are ‘stud mount’ (looks like a bolt), others look like flat-mount power transistors. Some are dual in a single package.
If you have any old junk PC power supplies, you can probably find some in there.
I took apart a cheap ATX PSU that died, but couldn’t find any MBR3xxx or MBR4xxx series type diodes. I found a whole lot of FR103s (only rated for 1A) and two unknown larger diodes. It really looked like a cheap PSU, they were using 10V caps on the 12V rail, running a half-wave rectifier circuit, had heat sinks so soft I could bend them with two fingers, etc… Anyway I was able to find three BY704 diodes. These guys are old school power diodes and I found them next to some original (i.e. white ceramic gold plated) 7447 and 7448 ICs down in the basement. They’re rated 6A at 300V max, so they should definitely work. The only question is what the nominal drop is (specs say 1.2V at max and didn’t have a graph), since I can’t find out if they’re Schottky or not. I put them in a simple testing circuit which draws 500 mA from a 5V supply to measure the drop and I seems to be around 0.25 V which would put it on the voltage curve of a nice Schottky as most regular diodes start dropping at least 0.5V after a few mA.
Sorry, I misled you a bit. I didn’t mean to say that you’d find those particular part #s in a PC supply, but rather that you might find Schottky diodes in general.
Note that some come as duals, and are in 3-lead packages that look like power transistors.
Nope, my fault for being an idiot. When you said power transistors I immediately thought the SMD flat-pact form factor since I had been looking at both some FETs for OpenServo and some diodes on digikey which used the form factor. I didn’t think of TO-220 or TO-247. I should have noted the big diode symbols on the TO-247 in pretty much any ATX PSU.
Going though the remains a second time, I Found some even better S30D4s which have less than 0.4V forward drop even at 4 Amps and are rated at an insane 150V, 80A. Even the TO-220 sized ones are rated at far beyond what I need. I like the form factor better than the bolt diodes mainly since I can more easily attach heat sinks and mount them.
Thanks again for all the help. I’ll keep everyone posted on my progress once the actual power supplies start showing up.
Well… after finding a second dead (originally more expensive) PSU in the lab I actually have an MBR4045 along with another 80A rated 100V diode. I now have more than enough diodes for my project. I’m a bit pissed, since at one time I had a pile of 10 or so dead PSUs hanging around the lab (one of my jobs here is to keep all the hundreds of cluster machines running and my boss didn’t quite understand the idea that its better to spend a little more on PSUs than to waste time having to hunt down supply quirks which cause crashes) but I only removed the voltage regulators (nearly all have a 7805 and 7905), capacitors, and some zener diodes. I could have had a big pile of these very nice power diodes.
Anyway… One quick question. I can’t see it myself from the circuit board, what are these guys used for in the supply? Do the major rails in a normal ATX supply have real reverse protection? If so, could I do something as stupid as just put two PSUs together to create a redundant PSU? This of course doesn’t pertain to my project, but I had watched a debate on another forum about a guy who took three 400W ATX supplies and tried to tie them together with an array of FETs and such instead of buying a commercial redundant supply for his server.
The diodes are a fundamental piece of a switching power supply. In the circuit, there is probably a good-sized inductor nearby.
I think they are not used for reverse protection.
And I would not attempt to parallel these supplies without adding ‘balancing’ circuitry. PC supplies tend to be high-current, and paralleling them will cause them to ‘fight’ when the voltages are not identical (which they never will be).
