Computer ATX PowerSupply to Lab Bench power supply

[ag273n]

we had this old computer lying around that we don't use anymore and I needed a Lab bench power supply. i find the output voltages appealing as it matches the common voltages you can get - like from batteries.. so i thought of converting one.

After following some instructables, i got the ATX supply to work, and using a small voltmeter, i got the output voltages, +3.3, +5.4, and +12 volts, I didn't need the negative voltages (who does?...) those positive outputs seems ok - no visible issues... The "dummy load" is on the +5V rail as it has the most current: 20A

Now.. i didn't exactly like the idea of having one terminal for every output voltage, so i came up with the idea of using an SPTT switch(not sure about that name) on the schematic below:
(the switch can toggle around the three voltages)



So, after adding this switch, everything seems stable. I can toggle between 5V and 3V - totally ok, but when i try move up to the 12V something funky happens. The supply stays on while on 12V, but when i move down to the lower voltages, the whole ATX supply suddenly turns off - nothing blew up(whew!, just turned off by itself). So i can move on the 3V and 5V ok, and even move up to 12V, but when i move down from 12V to the lower voltages, thats when the supply suddenly turns off...

my guess is that, the ground recognizes the difference of the voltages it receives and turns off if there's a huge spike....
could this be correct?...

 

[Harald Kapp]

my guess is that, the ground recognizes the difference of the voltages it receives and turns off if there's a huge spike....
could this be correct?...

No, ground cannot 'recognize the difference of the voltages it receives'. One possibility is that the 12 V power supply rail increases in voltage when you disconnect the load by switching to a lower voltage. What happens when you load the 12 V output with a constant oad (e.g. 1 kΩ)?

Note that the output of a computer power supply can be particularly noisy. Depending on the circuit you use the supply for additional filtering may be required to get a reasonably clean output.

 

[ag273n]

What happens when you load the 12 V output with a constant oad (e.g. 1 kΩ)?

i have not added any load to it yet (only the voltmeter connected), i will try that, I only have 1/4 Watt resistors, are those ok for the test?

Note that the output of a computer power supply can be particularly noisy. Depending on the circuit you use the supply for additional filtering may be required to get a reasonably clean output.

I did observe the output voltages fluctuates a bit.

 

[Harald Kapp]

i will try that, I only have 1/4 Watt resistors, are those ok for the test?

P=V²/R -> 144V²/1000Ω = 0.144W -> o.k

 

[ag273n]

ok, thanks, i will try that and will post what i find probably in the next 24 hrs..

 

[AnalogKid]

ATX supplies make such high power at such low cost that using them outside their narrowly-defined application is risky. IO!, them puppies be weird, and no two are alike. Your model might require a minimum load on more than one output to stabilize the magnetics, particularly when handling a step decrease in load. There is "electronic inertia" in the switching regulator control loop bandwidth. That is, there is a minimum time delay between when you remove a load and when the regulator starts to adjust its output. During that time the output will rise, possible enough to trip an overvoltage monitor.

ATX supplies are relatively robust, and somewhere between cheap and free. And there are lots of sites with instructions for turning them into other things. But I cringe at the thought of someone new and young to the world of electronics experimenting with a couple of opamps and a power supply that is poorly regulated, noisy, and can deliver 200 W in a heartbeat to a circuit that needs 0.1 W.

ak

 

[duke37]

If your voltmeter has an input capacitor, then switching from 12V to 5V will put 12V on it for a short time. The PSU might not like this.
A heavy load might stop this.

 

[ag273n]

I have to start somewhere; everyone does. I find batteries to be costly if i have to keep buying replacements, and i didn't like waiting several hours just to get them charged for the tests... wallworts seems ok, but i only have a 5V 0.5A wallwort, and having no other variety is a limitation to the possibilities.



Now, i tested the 1KΩ Resistor as load to the 12v rail and did the toggling around the voltages, it did the same thing. As soon as i move to the lower voltages from 12V, it automatically shuts off. Same effect without the voltmeter and only the load resistor.



I observe this does not occur when I only toggle around the three rails and there's nothing connected to those rails(by nothing i mean nothing- not even the voltmeter). When in closed circuit, its that abrupt change in the voltage that causes it to shut off

 

[tedstruk]

I agree with Analog...a power regulation/ overheat sciruit built in to the PSU, is firing when the draw is switched. A power supply is only a supply, and not an integral working part of the circuitry. There are many different requirements for motherboard power... try looking up the board the psu drove, it might have power specs it required to give you some kind of a block diagram. Personally with power, I would leave out the switches as much as possible, to much chance of a brown out.

 

[ag273n]

I eventually ended up leaving the power rails into individual posts. so i got 4 terminals; 3v, 5v, 12v and GND.
Then yesterday, i found a youtuber who had the same Idea as me - to use a switch to change the voltage. He used a twist knob with three outputs and just two binding posts - his ATX didn't switch off like mine. While I am certain his ATX is probably a different model than what i have and may work a little differently, I'm still open to the idea of using a switch to minimize the terminals as soon as i find a similar switch.
I simply find dedicating one terminal for each voltage a little less elegant than just a consolidated terminals.

 

[ag273n]

I finally found the solution!
this I observed after i dismantled the switch assembly.
It has something to do with how the switch was designed mechanically...
On the three-mode sliding switch i first tried out, the contacts from the 5v and 12v were making contact on a split second every time I toggle it between the 12v and 5v triggering the overvoltage protection - shutting off the ATX unit. this switch isn't modifiable because of its size.





So I got myself a three-mode rotary switch. the inner contacts were bendable in a way that it doesn't affect the operation of the switch


On this small modification, when the switch is toggled between the voltages, it creates an open circuit in a split second until the contacts reach its intended position and prevents the 12v rail from grounding to the 5v rail.

I already installed everything and tested.. works!