variable voltage from converted PC power supply

Hello all, I'm hoping someone could help me with the DC power supply I'm trying to make. You probably know the sort: a PC power supply with some posts and LEDs stuck on the front\top.

Well, mine's not an ATX one and I'm having an of issue with variable DV voltages on the +12V line. From the annotation on the circuit board I've worked out the following colour-wire code:

black: ground
red: +5V 18A
yellow: +12V 7A
blue: -12V 0.3A
white: -5V 0.3A
brown: +3.3V 6A & one sense wire
green: +5VSB 1A
orange: power good
purple: DC on/off

I've connected the green & purple wires to ground and the brown sense wire to another brown wire. When I'm testing it on the bench with 12V 21W automotive bulbs on both +5V and +12V lines and a CPU fan on the +12V too, the voltage on the 12V line varies quite a lot. Like this:

Its been reported that these PSUs need a load on the +5V line to operate and most 'how-to's recommend the use of a 10ohm, 10W power resistor. I'm not sure that the bulb I'm using on the +5V line produces enough load to keep the power supply running. The bulb only has about 1 ohm resistance, measured by DMM, but at 21W I thought it would be adequate. I've tried running three of these in parallel on the +5V line to up the load but the voltage fell to ~3V and the output on the unloaded 12V line went up to +17V! Can anyone tell me if I've done something wrong that results in these variable voltages or is the PSU duff?

(I should confess to not knowing if this is entirely normal behaviour for PC PSUs. If so, doh.)

It may shock you but usually 5V digital electronics work perfectly well within specs with voltages between 4.5V and 5.5V (up to +/-10% variation on 5V).

A 10 Ohm resistor on the 5V line causes a current of 0.5A.

Its been reported that these PSUs need a load on the +5V line to operate

I have to admit that I haven't tested the 'smoothness' of the output voltage of the PSU I'm currently using, but the 300W PSU I'm using doesn't need to be loaded up to work. I only keep the PS-ON line GNDed so it runs, and it runs fine, regardless of load or no-load state.

That the voltage varies seems strange. I've converted a personal computer 500W power supply to be a bench unit. By coincidence, I've also used it on automotive lights of about the same wattage. It works fine.

Yours dropping to 9 Volts with a less than 2 Amp load sounds like you have a problem with your wiring or something. The power supply should, in principle, be able to handle it fine.

I also read several "how-to"s about a load being needed. But I did not put a load on mine, with no problem seen. It does have a couple of LED fans for cooling, though, so maybe that suffices. Frankly, though, I think the people who recommend a big resistor may be steering people down the wrong path.

Thanks for all the replies. I've just been replacing four bulging/leaking electrolytic capacitors in the PSU and resoldering a large resistor that appeared to be loose. Now when I turn the supply on the +5SB line is good but when I connect the purple 'power on' to ground the fan twitches or turns for a second and then stops. I'm going to open it up again and check for shorts.

Another unusual feature of this PSU is the dedicated (-) sense wire that I'm assuming the +3.3V sense wire should connect to? This is confusing me as I've read elsewhere that the (+) sense wire should be connected to another +3.3V line. Should I bundle all three; +3.3V, (+) sense & (-) sense together?

The - sense wire should go on ground (at the device being powered).