ATX to Lab Bench Power Supply Conversion

Step 6 Finished Project

Here you can see the finished project, both with and without the cover. If you have any questions, please leave a comment and I will try to check back often to answer them. Thanks for looking, and good luck!

Keep in mind that while I built my power supply many years ago with only the ground line fused, you should put fuses on all your signal lines and leave the ground line directly connected.

Comments

[milkywayman]

Hi all, just wanted to give you my 2 cents :)

I used a 350W ATX power supply to power a CD car radio and an automated aerial over 12V and here's the results:
- the 3.3V line had to be loaded with almost 2 Ohms of resistance!
The reason is this: the 12V are supplied from the same transformer as the 3.3V and are not regulated as opposed to the 3.3V line. So when I cranked up the volume the PS would shut down since the load on the 12V caused the 3.3V to dip. But since the 3.3V line had no current going out the PS decided to shut down. So in order to get any good power out of the 12V line the 3.3V line should be loaded with at least 1 Amp.
Now I can draw almost 3 Amps on the 12V line without a problem.

PS: your mileage might wary with the load resistor :)

[matthewbeckler]

Another good data point, thank you very much for sharing your experiences with us, I'm sure it will help someone who's seeing a similar problem.!

[khr1stian]

Does anyone know of a way to disable the short circuit protection? I am using mine to power a jacob's ladder / plasma speaker, and it really is quite annoying having my psu randomly shut off on me.

[ChrisInVT]

You ever find a resolution to disabling the short circuit protection? I've been googling for a while now, and this is all I've been able to find.

[shadowdude77]

Ugh, that's a shame, I was gonna use mine to power a flyback for a plasma speaker as well. How often does it shut off? There must be a way to disable short circuit protection, and I can't imagine it's a complicated procedure.

[matthewbeckler]

Perhaps you could get a large capacitor, like the big ones they use for car audio, and install that between the power rail and ground to help provide current when it's needed, without the power supply shutting down.

[hitachi8]

yeah , you can put something like a fake load , on the PSU ... like a Resistor , a little light ...

[Church of the Holy Electron]

Just finished my ATX "bench supply" and wanted to leave some notes on my experience.

Mine was an Astec unit from an old IBM PC (ca. 1998). Contrary to most of what I've read about this conversion on the web, including here, no load resistor was necessary to get this working without a computer. I plugged it in, jumpered the PWR_ON signal to ground, the fan started turning, and I got good voltages all across the plug. Also, on my unit, a lot of the wire colors had no correspondence to what you and others have described, apart from the +5V and +12V lines which are invariably read & yellow. I found a reference online (http://pinouts.ru/Power/atxpower_pinout.shtml) that was very helpful in sorting this out. I ended up going through the output plug connections one by one with a meter, and they all corresponded to the specified voltages. I suggest this method, rather than looking for certain-colored wires, as this may not be correct.

As it was, the only connection I had to make internally was the PWR_ON to ground one to activate the supply. I also connected some wires internally that were wired together at the output plug. I suggest doing this if you find there's more than one wire going to an output pin, as I did. I also found a small capacitor (0.01 uF) across the PWR_GOOD and ground pins, but left this out, as it appears to be a last-minute "fix" probably ordered by IBM for a glitch on their MB. Supply works fine without it.

Thanks for making this information available to all us DIYers!

[crf250rider14]

Hello Matthew, I am doing the same conversion. But what is the reason why you used the fuses. I have read many other guides and none of them mentioned anything about fuses. I will be using mine for a cb radio and it already has a fuse in line before it gets to the radio. So does that mean i do not need to put fuses inside the psu? Also, i pretty much have the supply complete but it is not turning on. All my wires are connected to their binding posts and the power resistor seems to be hooked up correctly. The "live" LED comes on and tells me that their is power going into the unit when i plug it in. But when I flip the switch nothing happens. Also, the fan does not kick on at all. I checked for any bare wire and did not see any. One last thing to remember, I have no green wire because this supply came from a dell with a Pentium 3 processer which did not follow the color coding exactly. I really wish it did though. So my gray wire is ps_on and my orange wire is power ok.
Any help you give will be greatly appreaciated.
Thanks.

[matthewbeckler]

Hello, thanks for the message!

I added a fuse simply to protect my usually delicate, small circuits from the (relatively) massive current capacity of the power supply. Normally I work with little microcontroller projects, designed to be run from batteries, and the 20 amps my supply can provide at 5v is more than enough to smoke traces off my PCBs, if I screw up something. So, the fuse is added just to protect my circuits against my own stupidity. :- ) You shouldn't need any additional fuses, especially since your radio already has a fuse.

Sounds like you did everything right in your conversion, good work. Does the PS follow the standard pinout, even if the wires have the wrong colors? Sounds like you've found the right wires for everything else aside from the power up switch wire. Are there any other wires connected to the big motherboard plug that you haven't used, and don't know what they might be used for? You might try grounding each of these 'mystery wires' in turn to see if any will turn it on?

What size and power rating does your power resistor have, just out of curiosity?

Good luck,
Matthew

[crf250rider14]

Thanks for the reply Matthew. I am using a 10 ohm >10watt "sandbar" resistor. I have also read other guides online besides yours and found that is the one they are using. I was thinking of trying the gray (my ps_on wire) to black without a switch. Maybe something was wrong with that. Im just running out of options...
Thanks again.

[timbont]

Hi, I have built the same, but what stops the supply tripping or resetting itself when turning on from power user? eg: I run a 12v dc fridge 4.5amp and after fridge has reached temp it turns compressor off, when restarting the compressor it trips the power supply and I have to manualy start it again.
I understand the earler AT box did not do this but am unable to get the full 12v.

Tim

[tank1357]

The reason the supply shuts off when the motor kicks on is because the supply has short circuit detection. What this means is that if it senses a large current spike, it will turn off in an effort to protect itself and the circuitry it supplies. Unfortunately, there are not many ways to get around this, as most large motors sometimes need a moderately large amount of current to get them started. This is because sometimes you have to clear a tiny spark gap inside the motor, because the motor may have stopped with its windings in poor contact. One of the only serviceable ways around this is to attach a large-ish capacitor across the terminals of the motor. It doesn't have to be done in a sexy way, just solder it right on to the two terminals. This should help with some of the issues. NOTE: motors and other coiled devices (such as relays) also throw around a lot of nasty EMF when discharged that will easily damage or destroy sensitive devices, like transistors. For instance, I once had a relay controlled by a BJT, and it worked fine for awhile, but after only maybe 50 or so on/off transitions, the transistor no longer worked properly. A way to protect against this is by using a diode arranged across the coiled device so that it is normally not conducting, but if the discharge of the coil is larger than the power supply, it'll back-conduct the spike through the coil again, treating it like a resistor, dissipating the spike until it is harmless. Of course, its hard to find diodes beefy enough to preform this duty in larger power-applications (such as your compressor motor), so in place of the diode, two capacitors in parallel (one big and one small) will suffice.

[timbit1985]

compressor motors on fridges and air conditioners have a very high startup current because of the compression ratio's between the highside and low side of the pump. Some compressors have a capacitor connected to one of the poles to create a slight delay between the electromagnets. This slight delay in effect reduces the compression between the two poles of the motor. You might consider placing a capacitor on one of the lines of the motor to lower the current requirements of startup.

[matthewbeckler]

Hi Tim! Thanks for the comment.

I think that computer power supplies have a short-circuit-prevention where any large spikes in the load after they have turned on are treated as a short circuit and the PSU shuts down to avoid damage. I'm not sure how you could get around this issue, hopefully someone else has run into this problem and has a solution.

[limbokiller]

this wasn't a "be not nice comment :P"

[matthewbeckler]

:- ) No worries, it's been a long time since I've seen a "not nice" comment on Instructables. I've actually never run into a power supply that wouldn't keep itself running without a load, with just a jumper wire between the green wire and ground. Good to know that others have seen the same. Thanks for your feedback and tips, Matthew

[limbokiller]

that could be true but I've done it enough times when i needed to test a dvd rom player or something like that so and i've done it with all kinds of psu's. and then the only thing i do is put a wire in the green socket and in a ground socket and the thing goes on and stays on

[limbokiller]

but you don't need that resistor just connecting them is enough

[matthewbeckler]

Sometimes that will work, but I've read that most ATX switch-mode power supplies (SMPS) need at least a small load on the +5v line, sometimes also the +12v line. In a computer, there's always a large enough load on all the lines, but with this sort of power supply, you might be using only the 12v line, so the resistor is there to provide a minimal load to keep the power supply in operation. If you know that you will always have a load on the 5v line, then you wouldn't need the resistor.

[limbokiller]

just connect the green wire from the 20 pins or 24 pins connector to any ground wire and it will also stay on

[matthewbeckler]

Exactly. Check out the diagram in Step 2, where I show the power switch connecting the ON/OFF green wire with the ground wire.

[jb33]

Nice project. Do you need additional filter capacitors on the outputs to remove high frequency ripple? I have read that computer power supply outputs can be noisy.

[verence]

Hi, jb33!

Just in case, you still worry:
Yes, some capacitors might be fine. But... the matter is not that easy. (Check TI, Intel, Maxim, SE and other chip producers websites if you have the time and will to dive deep into the matter..)
From my experience:
- normally only the 5V rail of a PC-PSU is well regulated (if you draw a high current in 12V, but 5V is 'idle', the 12V voltage might drop)
- the voltage is fine for lower frequencies, but might carry some high frequency ripple
- do not put a capacitor close to the power supply (a very, very high capacitor might even distort the regulation)
- you might put a capacitor (3..300µF) close to a cluster of loads
- you should put an small capacitor (about 100nf, ceramics) as close as possible to every load (IC)
(The as close as possible is no kidding: every fracture of an inch of lead acts as an inductor and degrades the efficiency of the capacitor. Ideally, the capacitor should be buried underneath the load inside the PCB (and it is in high end stuff). Sure, for a home-brew solution this is not possible, but keep the wires to the capacitors as short as possible.
- if you design some high end analogue stuff (12bit DACs etc) , you should forget about the PC-PSU. You should use a linear regulated supply. You should separate analogue and digital ground. - And you will be annoyed about this stuff as you do know about it anyway ;-) .

[jb33]

Thanks for the advice, Verence. Would there be any benefit (or harm) in connecting some capacitors close to the output jacks? I left the power supply in the old computer case and the output jacks are on the front of the computer case. I would still connect a 100nF close to each IC on the circuit. I'm also building a linear regulated variable power supply.

[verence]

When decoupling an electronic circuit, you always have to think not only about capacitors, but also about inductors. In this case any piece of wire (as cable or on PCB) acts as an inductance. While a capacitor tries to even out the voltage by modulating the current, an indictor tries to keep the flowing current constant by sacrificing the voltage. (Well this is not very technical speak, but should give an idea, what's happening) So the wires ('inductors') should be as close as possibly and the capacitors as near as possible to the load. This is, why on a PC main board the CPU is surrounded by capacitors and even has some SMD capacitors right on the CPU. Would it do harm to put an extra capacitor inside the case on the output jacks? Well, I don't know ;-) The relative low internal resistance of an empty capacitor creates a high current flow when the PSU is switched on - that might trigger some over-current protection or 'distort' the regulation (I'm talking about big capacitors here, think >> 100,000uF ) What will happen depends on the design of the PSU, the capacity, the ESR of the capacitor, the connected load etc. etc. Would it do any good? I don't know as well. If you are in doubt about the quality of the voltage regulation, check it on an oscilloscope. You probably always will see some higher frequency noise from the switching characteristics of the PSU. So a switching PSU is almost always (unless you have a lot of time any really know what you are doing) a bad choice for high precision measurement or high end audio stuff. If you want to know more in-depth stuff about decoupling, you will find a lot of information on the web sites of chip producers. They also might have good application notes and design notes for your linear regulated PSU project (and you might be able to order some free samples)

[jb33]

I finally got the PS hooked up. I wound up installing a 1000uF and a 0.1uF near the output terminals for each voltage. I checked the outputs with my small hand-held scope which reads out on a PC. The capacitors appear to have removed the ripple I observed previously without the caps. The voltages appear to be stable under small loads. I will still install appropriate capacitors on the circuits. My PSU is an older unit, not an ATX. I installed a 3ohm, 25W resistor for the load resistor on the 5V line. With no other loads I get the following voltages on each line: Red: 5.1V; White: -5.0V; Yellow: 12.1V; Blue: -11.4 With a 10ohm resistor on the 5V line I got: Red: 5.2V; White: -5.1V; Yellow: 11.7V; Blue: -10.5V

[matthewbeckler]

I've not had any troubles with any analog or digital circuits using this power supply. Nowadays most of my digital electronics seem to be using 3.3v, so I'm using a linear regulator to generate 3.3v, which does a good job anyway of cutting down on noise. In general, regardless of your supply, it's a good idea to throw some capacitors across your power rails to help deal with large spikes of current draw, such as on startup.

[jb33]

Thanks for the info. I left the power supply in an old computer case, so I'll put some capacitors across the different outputs inside the case.

[ghoru]

i was thinking of converting an old psu into a similar test bench for multiple led lighting, would this be a good way to go about it?

[ghoru]

also would removing the metal case be a wise idea? i may also end up using as a perma PSU for said lighting so i was looking to save as much space as possible (

[omnibot]

I wouldn't remove the case since these tend to heat up but it's prolly the best money can't buy. I've made several for all sorts of uses from LED to batterychargers.

[verence]

Totally agreed.
Removing the case might help from the thermal point of view (better air flow - as long as the fan is working),
BUT: The metal case is also something like the last barrier for the dangerous voltages inside the power supplies. By design, switching power supplies generate about 1.4 times the mains voltage (at least..). So, in the US, you would get 1.4*110V (potentially lethal) and in areas like Europe, you might get 1.4*230V (Yieks! - NOT NICE!).
If you keep the metal case (strongly recommended!), make sure, it is connected with the grounding wire on your power cable - disconnect and use a multi meter to check the resistance between the case and the grounding pin on the power plug on the cable. It should be _far_ less than 1 Ohm, ideally a short circuit. So, if something inside the PSU goes astray (loose wire, broken part) and the metal case should be accidentally connected to a lethal voltage, your in-house circuit breaker should be triggered and switch the whole thing off.
Okay, there might be some inconvenience - dark rooms, blinking clocks, lost computer data - but hey, you, your children, wife, husband, significant other (am I the only one who thinks PC-speak is ridiculous?), pets, burglar, whoever will be alive.
If you really want to / have to remove the metal case (which I strongly not recommend), make sure that it is impossible to touch any part of the circuit.
Sorry, if I sound like a fun-stopper (or party brake as we say in Germany). I know, it might - or most probably even will - work without a case, without grounding, even without good insulation. You'll maybe never notice any bad effect of neglecting all good advice.
But ... just eventually ..., just in case, ... if it should happen ... well, the effect might be fatal.
In the truest sense of the word: finally, lethal - 'nuff said.

Have fun, but take care.
PS: Nice I'ble :-)

PPS: If there is a permanent load (like a string of LEDs) connected to the PSU, you can go without the load resistor (the 10Ohm, 10Watts in the I'ble) and save some energy. Just make sure that the PSU doesn't have to work without the minimum load. It might be damaged otherwise.

[matthewbeckler]

Agreed with omnibot, you probably want to leave the case on. There are some large heatsinks inside that need to vent their heat to the outside world, and the case and fan provide good airflow to help get rid of the heat. Also, there are large and dangerous voltages inside, so you probably want the case there to help protect you from accidentally touching anything hot or "hot".

[freelancemissions]

Great work. A couple of things, the 10 resistor could be placed on a small plate of aluminium and mounted near the power supply on rubber feet. This could keep your coffee warm. If one is thinking that the 10 ohm resistor is wasting heat, you could install a 6Volt lamp as a power on indicator as well. Also your warning is correct on the power capacitators in the high voltage side. These caps store a charge over 300 volts at times, so if you do not want to do the "look at me jumping like a fool because I touched something I should not have", please be careful.