Using computer PSU's

[DebboR]

Hi all!

I am at the moment redesigning my workbench, and I thought it would be cool to have a LED strip as desk lighting. So I ordered 5M strip, and it runs fine at about 12V 3A. First, I was planning to buy an adapter for the strip, but then I came up with this.

I only have one lab power supply ATM, and most of the time it is used for my digital projects (+5V). Now was I thinking of using a computer PSU to drive my LED strip at 12V and also break out the different voltages it can deliver (so +3.3V, +5V, +12V and -12V) too so I can use them in my testing.

I know how to start the PSU (putting a switch between the green wire and ground), but I am wondering if a PSU has some kind of short circuit protection? If it doesn't, is it easy to make one myself? Will the fuse pop if I short circuit anything or will it just shut down? And if it shuts down, does it shut down all channels or just the one that is shorted? Can I add some sort of reset button to reset it after a short?

I hope you can answer these questions!

DebboR

[alm]

A decent ATX PSU will have short circuit protection. For a cheap one, all bets are off and it might catch fire and eat your first born if you short its outputs. The catch is that the power supply will be able to deliver tens of amps, so it will happily fry your circuit by continuously pumping 10A into it.

I would expect it to shut down all channels when shorted. Resetting usually involves cutting the power to the power supply for a while. The supply is designed for a relatively balanced load between 12V, 5V, 3.3V and -12V. Don't expect it to be in regulation if you draw no power from the 12V rail but several amps from the 5V rail. The 12V rail will usually go up in this case. In power supply tests, this is usually referred to as a cross-load test.

Also keep in mind that the power it can deliver from the -12 V is very tiny (<< 1A), and that all of the outputs are referenced to earth.

[DebboR]

A decent ATX PSU will have short circuit protection. For a cheap one, all bets are off and it might catch fire and eat your first born if you short its outputs. The catch is that the power supply will be able to deliver tens of amps, so it will happily fry your circuit by continuously pumping 10A into it.

I would expect it to shut down all channels when shorted. Resetting usually involves cutting the power to the power supply for a while. The supply is designed for a relatively balanced load between 12V, 5V, 3.3V and -12V. Don't expect it to be in regulation if you draw no power from the 12V rail but several amps from the 5V rail. The 12V rail will usually go up in this case. In power supply tests, this is usually referred to as a cross-load test.

Also keep in mind that the power it can deliver from the -12 V is very tiny (<< 1A), and that all of the outputs are referenced to earth.

Ok! thanks for this useful information. So you won't suggest using a PSU in this case?
Maybe I might use amp meters at my outputs, and a switch and a pushbutton wired in parallel so I can quickly test the power draw with the pushbutton before switching the output on.

DebboR
 

[alm]

I wouldn't recommend them as lab supply replacement, for the reasons I mentioned (grounded, high current and also relatively noisy). I think they're fine as high-current 12 V supply (most modern ATX supplies are essentially 12 V supply with some buck converters for the lower voltage rails), and probably the cheapest way to get that kind of power. Driving motors or large numbers of LEDs should be fine.

[Strada916]

Most PC power supplies will require some load on the 5v rail in order to get a decent current draw from the rest of the rails. In other words you may need to load the 5v rail with a resistor or a light in the order of about 10W. 5V or even 3.3v rails are the regualted rails. Open the box and see which one has the feed back. that will tell you which rail to load

[alm]

Modern supplies still use the 5 V rail for regulation? I'm surprised, since the only thing using 5 V is low power components like drives. All of the power-hungry stuff (eg. CPU) is powered from 12 V these days.

[mariush]

Most PC power supplies will require some load on the 5v rail in order to get a decent current draw from the rest of the rails. In other words you may need to load the 5v rail with a resistor or a light in the order of about 10W. 5V or even 3.3v rails are the regualted rails. Open the box and see which one has the feed back. that will tell you which rail to load

This only applies nowadays to cheap 10-30$ power supplies, which are based on old designs. Yes, on those power supplies there has to be something connected to the 5v output so that 12v output will be stable. These types of power supplies already have resistors connected inside on the 3.3v and 5v and 12v to provide the smallest amount of load (usually 1-1.5 watts) to keep the power supply from oscillating or go way out of atx specs but in some cases an additional load on the main rail (usually the 5v one) is required.

From 40$ and up, you can get power supplies from brand names that are based on modern designs and manage to regulate 12v and 5v properly, no matter what so extra load is not really required.

Anyway... light strips are made out of leds... and leds are current driven, not by voltage.  LEDs have a forward voltage and once you reach that, they'll light up and the brightness will vary depending on how much current you push through them.
In a led strip, they may be connected usually in two ways...
One way, a few in series to get the 12v divided to 4-6 leds and get the voltage drop on each led around the led's forward voltage without caring about the current,  then all these series are in parallel connected to 12v
Second way, series of sets of leds, where each set has its own constant current regulator which gives a constant current to the leds no matter the voltage.

So in both cases, the led strip won't care that you don't have quite 12v on the output and won't care about the ripple

A linear adjustable power supply will usually have under 10mV ripple on the output, actually should have 1-2mV ripple or somewhere around that value. So the output voltage is much cleaner. In addition, a linear power supply will have a method to limit the current to a certain value, an ATX power supply won't, it can output all its power if there's a short.

An ATX power supply only has to have the ripple under a certain value specified in the standard...for 12v the value is about 120mV, for 5v is about 50mV... most no-name cheapo power supplies  don't respect that and at low loads such as a microcontroller and a lcd screen and some leds the ripple can be much higher.
Also due to design of the power supply, which switches 50-80k times a second, the output fluctuates so many times a second, it's a noisy output.

Some microcontrollers or chips are sensitive to such high ripple values so you can get erratic behavior in a circuit and pull your hair out unable to determine what's wrong.

Now you could still use an atx power supply to power your circuits but I'd recommend at least taking the 12v and moving it through a linear regulator such as a 7805 chip to get 5v with a lot of that noise removed... or use a low drop regulator such as *1117 (lm1117, amc1117, nxp1117, it's jellybean components, lots of manufacturers make it)  to get 3.3-3.6v out of the 5v rail (3.6v is a value often used on microcontrollers)
If you're paranoid you can also make a pi filter on the output to smooth it out even more (an inductor between two capacitors) or even ferrite beads at the input and output of the linear regulator

[mzzj]

Modern supplies still use the 5 V rail for regulation?

Applies only to OLD power supplies. ATX 2.0 spec is already around 10 years old and it defines all the high-power loads to 12 V lines.
ATX power supplies before 2.0 specification usually deliver most of the power from 5 V output and need couple of amps load to 5v line.

[IanB]

As you have heard computer power supplies can make fine utility power supplies for lighting, motors and such like. Just don't use them as a precision supply for anything delicate.

When you talk about an LED strip I presume you mean the ones designed for a regulated 12 V supply, the kind you just wire up to 12 V and forget about? A PC power supply should be fine for that.

If you are bothered about overload or short circuits you can always fit a fuse on the output.

[AndyC_772]

Definitely fit a fuse on the output. A PC power supply is an enormously powerful beast, and although it might shut down safely if actually shorted, if you put 0.25 Ohms across the 12V rail, it'll deliver over 500W into it quite happily.

I saw the effect dramatically demonstrated in a PC based product at a company I used to work for. We built equipment which was based on a PC chassis and a nice, big power supply that could run its high performance motherboard and a bank of hard discs. The product also included some simple relay I/O cards which were mounted separately and were powered from 12V via some flying leads with 0.1" pin headers on the end. They didn't draw much power and so were connected with ordinary 24AWG hook-up wire. Fine, except there was no fuse.

One day, a short on one of those headers resulted in the largest cloud of smoke I've ever seen come from a piece of electronic equipment. The wire was reduced to a glowing orange thread in seconds.

Fuse. Seriously!

[DebboR]

Hi,

Thanks for all these very useful reactions. So I can conclude this:

- The LED strip will be fine on the +12V rail on the PSU (with a 5A fuse just to be safe)
- If I get a more expensive PSU, the 12V rails are often better regulated then the 5V ones (in terms of ripple, ...)
- I don't have to use a dummy load on a (modern) PSU output
- Fuse all channels to be safe
- Get a 5V line from the 12V and a 7805 + caps to power more delicate circuits under 1A.
- Maybe also use the 5V line directly from the PSU to power things above 1A.

Only one question left: are there any quickly resettable fuses so I don't have to replace the glass ones once they've blown (something that will certainly happen pretty often, beleive me )?
I will certainly post pictures once my setup is complete!

DebboR

[AndyC_772]

Yes, you could use a self resetting PTC fuse. Something like one of these might be suitable:

http://www.bourns.com/data/global/pdfs/mfr.pdf

[DebboR]

Yes, you could use a self resetting PTC fuse. Something like one of these might be suitable:

http://www.bourns.com/data/global/pdfs/mfr.pdf

I will use one of these, but actually I meant something like this: http://homemadecircuitsandschematics.blogspot.be/2012/04/add-this-short-circuit-protection-to.html. I don't think this circuit will work with PSU's (because the voltage won't drop as quickly), but maybe with a power resistor before the relay to create a voltage drop.

Or maybe something microcontroller regulated? With a relay and a shunt resistor to measure the current?

I hope to get some good ideas for this here!

DebboR

[LaurenceW]

Hmmm, this is just MY opinion, but -

I wouldn't trust a PC power supply as a bench supply. They vary in quality tremendously - expect noise and poor regulation from cheaper ones. Also, the -12V rail will be pretty puny, while the current capacity of the +12V rail will be MASSIVE. They will be quite capable of outputting tens of amps, which will ensure that any 5V circuit you had erroneously connected to it was long dead, by the time any fuse you had thought to put in had blown - fuses can stop FIRES but will often NOT protect electronics. <<I KNOW THIS>>  This applies equally to one-time traditional fuses as it does to circuit breakers or resettable poly-fuses. Better protection is going to be afforded where you can control voltage and current limits electronically.

Why not look at a dedicated 12V "brick" power supply for your LEDs? They are cheap as chips on eBay.

[metalphreak]

http://www.seeedstudio.com/depot/atx-breakout-board-bench-power-supply-p-1222.html?cPath=155

Get the super cheap acrylic case thing as well. For a pretty low price you can get a few different rails at once, for not much money (on the assumption you already have a PSU spare). Plenty of electronics in a computer runs straight from the 5V rail without additional power conditioning (apart from a few local smoothing caps perhaps). I wouldn't recommend it for something involving opamps or analogue electronics, but for LED lighting/testing and stuff like that, its not a bad option. It's also much safer with that kit above than running off a PSU directly. There are 2A PTC fuses on every output and a handy power switch. Don't bother with the power resistor - you don't need it on most modern PSUs (They include it separately for you to solder on if you need to).


Oh, and there's no binding post included for the second GND and -5V connections. None of my power supplies have -5V anyway. I recommend getting another post for the GND though - it's nice to have two when using both 5V/12V for example.

[gnif]

Heh, interesting topic because I just did this today.. I will have to re-read the thread but I managed to successfully modify a Delta branded ATX psu for variable voltage and constant current (not settable yet).

It is not too hard, seems most modern supplies use 3 optos, one for power-good, one for +12V & +5VDC feedback, and one for 3.3V feedback. There is normally a adjustable precision shunt (in mine: TL431) that feeds the optos via a current limiting resistor. I just cut the trace to the reference of the shunt and put a voltage divider with a 100k pot inline between ground and +12V.

To prevent the circuit shutting down due to over/under voltage or short circuit, I bridged the output of the power-good opto. This worked fine and I was able to dial the voltage in that I wanted, until the protection IC went POP and let out its magic smoke due to over voltage, which I don't need anyway and is now not doing anything so I just removed the part.

The only thing to be careful with when doing this is all the caps in mine were only rated for 16V, so that is your absolute upper limit, possibly less as some caps are close to parts that get warm/hot. If I was to upgrade the caps I would be able to drive up to 36V. The minimum voltage obtainable with the supply I used is 6V. You must also re-wire the fan to use the 5V stand by power otherwise when you are under/over voltage your fan will not run or burn out.

The PWM driver is a constant current driver, just configured for the maximum the PSU can handle (in theory, who knows for the cheap ones), so it is safe to short the output. I have tested this and can nearly weld with the 19A I get out of this unit

[NiHaoMike]

Use a car light bulb as an overload protector. It reacts faster than a PTC and is very common and cheap.

[potatogun96]

I have tried similar projects with 3 supplies: A $15 POS, a 10 year old dell, and a 15 year old dell. Only the 10 year old dell has short circuit protection, the other two will stay on even if there is a dead short. If there is no protection, a mistake could be very bad (ATX supplies can supply >10A on some rails). If there is short circuit protection, and your projects do not need precise voltages, AND you are very careful, it can work.

[Eonir]

There is a ton of projects similar to this one already. One thing to look out for is not trusting the cables too much! A cheap ATX power supply might claim up to 30A on its 12V rail, but these tiny wires wouldn't be able to carry that much current.  Remember that in order to safely conduct, say 20A you need a conductor that's roughly 3mm in diameter!
Why am I saying this? Because I've seen lots of people who simply forget about this stuff and add new loads carelessly! Just because something is low voltage doesn't mean it's safe. An overloaded wire is a fire hazard.
I understand OP won't have this problem, since he's only going to use 3A on his lighting, but don't be tempted to drive too much from a crappy computer PSU.

[gnif]

There is a ton of projects similar to this one already. One thing to look out for is not trusting the cables too much! A cheap ATX power supply might claim up to 30A on its 12V rail, but these tiny wires wouldn't be able to carry that much current.  Remember that in order to safely conduct, say 20A you need a conductor that's roughly 3mm in diameter!
Why am I saying this? Because I've seen lots of people who simply forget about this stuff and add new loads carelessly! Just because something is low voltage doesn't mean it's safe. An overloaded wire is a fire hazard.
I understand OP won't have this problem, since he's only going to use 3A on his lighting, but don't be tempted to drive too much from a crappy computer PSU.

You are quite correct, what I have done to resolve this is take one of the bundles of wires for each rails (I think 5 wires) up to the front to some heavy banana sockets. I was playing a little with the supply today and found that it can supply enough juice to perform small spot-welds, so I may look at building a jig to perform battery tab welding with .

[ptricks]

If you use a PC PSU or really any PSU without current limiting stock up on 1,2,3,5 watt resistors that will provide the current ranges you will be using. Fuses react too slow to prevent that 20A 5V supply from killing your 5V 20ma mcu.  With the correct resistor in series you can even prevent a lot of costly mistakes.