ATX -> Bench Power Supply Mod

Yay bench power!
Yay bench power!

WARNING WARNING WARNING! Switching power supplies have many BACs (beefy-ass capacitors) that can hold lots of juice long after you’ve unplugged the unit. Do not crack one open like I did without knowing what you’re doing, it could seriously injure you or at least be really uncomfortable when you shock the crud out of yourself. I’m not responsible if you do just that, don’t say I didn’t warn you. I have a cable I made with a brush at the end that grounds everything it touches, I use it to brush the back of all power electronics before I work on them. You should too.

This is a pretty common one, but I like to add flair. I needed a bench power supply. Sadly, bench power supplies are very expensive, and I’m a poor college student. What I do have are computers. Lots of computers. In fact, I just brought to the recyclers about 100lbs of computer waste from machines that are too slow for even my standards. Luckily, one thing computers have going for them is very refined, VERY powerful power supplies. To put it in perspective, a really good bench PSU is adjustable from around 0 to 30V, and is current limited adjustable from around 0-3A. ATX power supplies are fixed, but for almost all purposes, they’re very accurate, they have a good selection of voltages (12, 5, 3.3, -5, -12), and they’re extremely powerful – the one I’m using can supply 6A on the +12 rail, and this is a pretty crappy old power supply. They also have two other REALLY nice features – over-current protection (if you draw too much, they just turn off, no harm no foul. Remove the load, turn it on, everything’s back to normal), and short-circuit protection (this one’s already saved me. If you short circuit anything, same deal. The PSU just turns itself off and waits). So the logical conclusion is to just cut the power plugs off of one and use it as a bench power supply. And you can do that, but I did some other things to make it look a bit prettier on my bench.

Transistor for the TTL all-systems-go indicator.
Transistor for the TTL all-systems-go indicator
  1. The most obvious is that I made a pretty box out of 1/4 inch MDF (medium density fiberboard – it’s a favorite building material of mine since it’s GREAT for speaker enclosures, which I like to make). I originally planned to use screws for the whole thing, but even using pilot holes, that didn’t work out too well – drilling the holes was hard enough, and the screws still tended to split the boards down the middle. I ended up just using clamps and wood glue.
  2. I drilled holes on the sides for air intake, and I used a Dremmel router tool to cut a fan hole in the back to relocate the built-in fan. One thing about MDF is that it’s made up of a lot of wood pulp and a lot of glue. It’s acutally a lot like really carcinogenic (wear a mask or respirator cutting it), really flammable paper. As such, I don’t want things getting too hot.
  3. I made a pretty front panel out of sheet stainless steel.
  4. I added a lot of LEDs. There are some interesting notes here. First, all ATX power supplies are switching DC supplies, and as such, they don’t really work without a load. If you tried to turn it on with nothing plugged in, it’d probably just fail into protect mode. As such, you need to prevent this by using a small ballast on at least one of the rails to essentially burn a little power and complete the circuit. I’ve seen guides online that recommend using a 10W beefy power resistor on the +12 line, but A. I didn’t have one, and B. I thought that might be overkill. My alternate plan was to jut put a single LED+series resistor on each line. You can never know too much about what’s going on, this configuration means that the ballast doubles as an indicator of power. Perfect! (note: in the end, I decided NOT to put an LED on either the -12 or -5 lines [hence the empty LED holes] because this PSU only supplies 300mA on either one. I figured that if I had need for them in the future, I might as well not waste such limited power).
  5. ATX power supplies are smart – they monitor themselves. As such, there’s a 5V TTL line that’s pulled high when the PSU thinks that all of it’s rails are within ATX spec bounds. Similarly, there’s a +5VSB line that supplies limited current when the PSU is off to keep things like the power button working. I figured that I’d use the 5VSB line to light a red LED, indicating mains power presence, and I figured I’d use the indicator line to light a green LED in addition to the others to say that the PSU is on and operating within spec (I’d expect it to enter protect mode if it weren’t, but maybe someday I’ll be surprised to find that the PSU is dying and the all-ok led is off even though the rails are on). There was only one problem. Since the indicator lead is TTL, it supplies barely any current – not enough to light an LED. I’m not the best EE in the world (just finished freshman year, so barely an EE at all), but I put on my EE hat and just put an NPN transistor from my box in the way. I wired the base of this to the TTL line, the emitter to ground, and the collector to the cathode of the LED. The anode is wired through a resistor to the +5 line. Basically, when the TTL line is high, the transistor sinks current for the LED and it’s lit.
  6. The last thing to do is add the switch. I chose to wire my switch between the green control line and ground – when this line is connected to ground (typically by the motherboard when it’s on), the PSU is on. When it’s not, only +5VSB is on. I could also have just shorted this and used the switch on the mains side, but I guess I just thought that was lame.
Exposed posts. Drat!
Exposed posts. Drat!

The only other point of interest is that I made the mistake of selecting the cheapest binding posts from RadioShack. These happen to have exposed posts, and while they come with insulating spacers to screw on on either side of your panel, I was worried that they might shift side to side and short out against my panel, which I’ve chosen to be ground. To combat this, I just put the first spacer on, slipped a ~.5mm long sliver of heatshrink tube on and shrunk it, then made sure that my holes were big enough to take both the post and heatshrink insulation. This has worked fine – I just put the other insulating washer on and screwed everything on, crushing the excess heatshrink into the gap between post and panel. I’m sorry I don’t have any good pictures of that part.

The next project will involve making a current and voltage variable power supply for those times when these selections don’t satisfy, but for now, I’ve got a great PSU that can run my Arduino, case fans I have kicking around to exhaust soldering fumes (I just screwed the +12 half of the molex connectors I cut off into the binding posts), and whatever other projects I’m working on. The best part is that I can test the 12V-based audio project I’m working on up to about half volume, since the thing has so much juice behind it.

I hope I’ve inspired you to go off and make your own el-cheapo PSU now!

8 thoughts on “ATX -> Bench Power Supply Mod

  1. Greetings. Very first I want to say that I truly like your webpage, just discovered it last week but I have been following it ever since then.

    I appear to agree with most of the thinkings and opinions and this post is no different. totally

    Thank you for a good webpage and I hope you maintain up the excellent function. If you do I will continue to look over it.

    Have a wonderful day.

  2. Thank you for your posting of your power supply mod/conversion. I also salvage what I can from any kind of tech but I usually save them for spare or replacement parts, I don’t have any experience in re-purposing equipment. but lately my box of power supply’s has been increasing, and I am trying to find creative ways to make use of them. I have been asking for any kind of assistance from those who have demonstrated a working knowledge of modding these psu’s. what I am wanting is to maybe build them into a type of component cabinet to power all the extra computer fans I also have, and maybe some kind of cool lighting, LED, UV, or EL string. Since I have been researching my idea I have found it exhausting to get any good information. I would greatly appreciate any thoughts, or insights, to what I am wanting to do. Most importantly should I event try it, or is the idea just not suitable for the equipment? Also if you can point me in the right direction as far as reference material, how to guides, text books etc.. I’d appreciate any knowledge you may share.

    Thank you,

    t. 1

  3. Well, DESIGNING power supplies is tricky, and very complicated, but once you have the device it’s pretty dead simple. You’ve got 5 rails to pick from and ground, +12 +5 +3.3 -5 and -12 volts. Linking things between them you can get other voltages, like +12 to -12 = 24v, +5 to +3.3 =1.7 volts, so on. There are only two tricky things to worry about. First, not all the rails can supply lots of current. If it’s an older or a smaller supply, the 12v rail will only be able to push a few amps, maybe 5. If it’s newer, it could push lots of amps – 30, 40, 50, I have a power supply that can do 100. That last bit is kind of important. Every ATX supply will have short circuit protection, but the meaning of that depends on the strength of the rail you short. For example, if you accidentally touch the two wires +12 and ground together, on most supplies that will turn it off immediately. But on my 100 amp supply, for example, the resistance of the wire might be high enough that what you see to be a dead short actually draws less than 100 amps, in which case the supply never turns off. And you set the wire on fire :).

    The other thing to watch out for is basically a more specific note about checking current ratings for the rails you want to use: the -5 and -12v rails aren’t used for much, so they rarely have high current ratings. Or even moderate ratings. Generally they’re on the order of 100mA to 300mA – nowhere near the level of the other rails.

    But generally you can’t really hurt the supply. If you attach too much load or short outputs, it’ll just turn off and you can power cycle it to reset everything. Happy playing!

  4. Alex,

    The last comment you posted saved my night πŸ™‚ I spent almost 5 hrs trying to figure out what happened to my psu coz I shorted +12 and ground by mistake . So i have two psu and i did shorted both the same way πŸ™
    after reading ur comment and waited one of them back to life but the other died πŸ™ which is the one that i spent my whole evening working on it . could you please advice if there is safety fuse or it might fried some components.

    Thank you,

  5. Interesting info here. What do you think about using the + 12 volts outputs from two separate ATX power supplies, and paralleling them together to get more current ???
    I want to run a transceiver with the 12 volts, and need about 20 Amps. One of my ATX power supplies puts out 14 Amps……and the other 17 Amps. Any problem paralleling the two outputs to get more current ??? Thanks, Bill

  6. Don’t do it, it is likely to go very wrong very fast. The problem is regulation and feedback. A switching power supply (like an ATX supply) uses a closed feedback loop to ensure that its output is within specification. The problem is, if you tie two together, the feedback loop opens (rather, supply A doesn’t know if what’s happening on the output is its own fault, or supply Bs, and it will probably fight B to get to the right value). This isn’t necessarily catastrophic, but it could be. Think about carrying a serving tray full of glasses of water. With your eyes open and paying attention, it’s very easy not to spill any – you sight and feel are both feeding back to tell you how to correct your motion to not spill anything. When you close your eyes, it gets a little harder, but still manageable. Now imagine two people carrying one serving tray from either end, with their eyes closed. You can’t see that your partner lifted up their end, so you can’t compensate, and very quickly the water will just spill. On the power supply, the likely result is that the output will oscillate aggressively, possibly causing no problems and possibly causing lots.

  7. Does anyone know if an atx power supply can be fooled to operate after the voltage ic monitoring chip has been ripped from the board?


Leave a Reply

Your email address will not be published. Required fields are marked *