Making a bench PSU from a PC PSU

[Jamesbeat]

I have an old PC power supply, and I'm making it into a bench power supply.
I'm not sure what type of psu it is, I think it's a propietory one (HP branded).
That doesn't really matter, because I have worked out what all of the wires are and it works great.

One thing that is puzzling me though, is the comcept of pre loading the PSU.

I have read a few tutorials, and all of them mention that it is necessary to add a power resistor between +5v and ground, because most PSU's won't power up without a load.

Mine does power up fine, and all voltages are correct, so I was wondering if I still need the resistor?
Does the fact that it powers up without a resistor mean that it doesn't need one, or will the power supply slowly kill itself if it is powered up without a load?

 

[Brevor]

Does the fact that it powers up without a resistor mean that it doesn't need one, or will the power supply slowly kill itself if it is powered up

It should be fine many newer supplies dont need the resistor.

 

[dr peppers]

The regulator circuits in older supplies were designed with a minimum load in mind, as theres always a load on a pc supply, for various reasons including cost if theres no load then the voltage exceeds the protection voltage and the supply shuts down, quite often they will cycle on and off.
The load resistor keeps the output in the regulated region.
You might find some noise on the output with a very small load.

 

[Jamesbeat]

That's odd then, from what I can make out the supply is from 1998, so it would certainly class as 'older'.

One more question: In the tutorials I have read, most of them say to bundle all of the wires of the same color together and attach them to the binding posts.
I presume that this is because of the relatively large currents that these supplies are capable of delivering.
Other tutorials suggest only one or two wires be connected, and the others clipped away.
Obviously, it would be a lot easier to attach fewer wires, as they number between 8 and 10 each, which would be difficult to physically attach to the binding posts.

The supply can deliver 20A on the +5v rail and 4.4A on the +12v rail.
I believe that the wires are 18awg.
What would be a sensible number of wires to attach to each binding post?

 

[dr peppers]

I'm sure everyone has diffo ideas on that, I'd go for 4 wires on the 5v, and a couple on the 12v, the only thing thats gonna cause is a litte volt drop I dont think your gonna be toasting 18 guage with that supply.

Take a look inside the supply at the controller chip, is it a tl494?

 

[Jamesbeat]

No, the chip is an AS2350C, which appears to be a specific Astec chip (the supply is made by Astec).

I ended up desoldering the +5v and COM wires and replacing them with nice chunky 14 gauge wires. From the specs I have seen, this should be enough even in the unlikely event that I manage to draw the full 20A that the +5v rail is capable of supplying.
I cut all but one of the +12v wires, because the supply can only deliver 4.4A on the +12v rail, so I figured 18 gauge should be fine.
I did this because of the the way the casing of my supply goes together.
I have to put the binding posts on the lid part rather than the chassis, and those thick bundles of wire would have been a nightmare to work with.

If these wire gauges are a mistake, then please by all means correct me...

 

[Sceadwian]

James, the resistor could be internal. You can't tell if it's 'okay' or not without detailed requirements of the noise that's acceptable and what the supply puts out when it's not under load, and a way to measure them.

What would be a sensible number of wires to attach to each binding post?

Anything that fits, the more the lower the voltage drop will be. You'll more than likely have more wires than you need. The mechanical bonding of the contacts at high current are very important to keep in mind.

I would definitely use more than one wire for 4.4A just to be on the safe side, the more you can attach the better as it reduces resistance heating of the wires drastically.

 

[Jamesbeat]

I guess I'll add a resistor just to be on the safe side then, it's not like they are expensive or difficult to install...

It's too late to use multiple wires on the +12v 4.4A rail because I cut them off, but I'll desolder it and use a thicker wire instead.

On the other rails, there were two solder pads next to each other because the original wires were split into two groups wired in parallel.
I threaded the wire in through one and out the other and made a huge solder blob over both pads, so mechanically, that connection is definitely over engineered.

Should I use bigger than 14 gauge wire on the +5v and COM? The +5v puts out 20A.

 

[Sceadwian]

I guess I'll add a resistor just to be on the safe side then, it's not like they are expensive or difficult to install...

Exactly. Alternately you could add an additional fan that keeps the power supply under load, it'll keep the supply cooler and reduce heating where as a resistor just generates heat and nothing else.

The ampacity for 14 gauge wire is 20 amps, you can use that fine as long as the runs aren't too long (which they aren't) Anything longer than a couple feet and you'll probably want to figure it what the voltage drop will be from the wires resistance as it might affect devices that expect stiffer voltages. Personally for this kind of stuff I'd simple use 3AWG higher or run two wires instead of one, the extra overhead will reduce heat lost in the wire, keep the voltage stiff and if it's properly fused/over current protected the wiring is basically bullet proof, you can't use too much wiring, but you can use not enough. The only time this is a consideration is if you don't have the materials or cost is a high concern.

 

[Jamesbeat]

Thanks so much for your help everyone, as you can probably tell, I'm rather the enthusiastic amateur when it comes to power supplies!
I remembered that one of my friends had an old pc in his garage, so I bribed him with beer and got a new power supply.
This is a big deal, because this supply has got 3v, which the other one hadn't, and that will be really useful to me.

One thing about this one that I don't understand is that it has two +12v rails.
One is 10A and one is 15A.
The 15A rail has only two wires connected to it (they are yellow with a black stripe as opposed to solid yellow).

This board is different, and doesn't have nice big holes to put chunky wire through like the other one did, so I'm stuck with bundling the 18 gauge wires together.
It seems odd that the 15A +12v rail only has two wires, whereas the 10A one has 7 or eight.

I only need one +12v output, so should I go with the more powerful (and thus more useful) 15A one (which only has two wires) or the 10A one with several wires?

 

[dr peppers]

I think some supplies have mutliple 12v rails to split the load, 2 smaller supplies use common cheaper parts rather than one beefy 12 v supply.
Another reason is short circuit protection, if you only have 1 supply at twice the current if one of the outputs gets shorted the fault current would be greater, also with the 12v rail theres a risk that with standard 1.0mm2 output cables that the short circuit protection wouldnt even detect a short and the supply would just dump current into the shorted cable till it fried, copper is expensive so fat output cables wouldnt be popular with manufacturers.

Trace thetracks on the board, if one of the 12v supplies has its own power tranny thats the one to use.

 

[Sceadwian]

also with the 12v rail theres a risk that with standard 1.0mm2 output cables that the short circuit protection wouldnt even detect a short and the supply would just dump current into the shorted cable till it fried,

Can you substantiate that with some details? Under what conditions can an ATX supply be shorted where the over current protection would be unable to trigger? I've never heard of this before.

 

[Jamesbeat]

I traced the tracks as best I could, and it appears that the less powerful output is the one with its own power transistor.
The 15A one had some very narrow tracks too, so narrow in fact that I felt a little foolish and read the label again, expecting that I had misread it and it was in fact 1.5A.
But no, I was right the first time!

Needless to say, I clipped the wires on that one and went with the lower powered but more heavily built output. I can't see myself needing 10A, let alone 15A anyway.

Interestingly, this PSU also runs fine without a resistor, but I have one on order so I'll pop it in when it arrives just to be on the safe side.

 

[dr peppers]

R.E. current limit.
I read it.
I think it was to do with the specifications of pc supplies, around when the p4 connector became popular pc supplies were able to supply loads of current on the 12v line, some so much so that the over current protection wouldnt have been good enough with just one 12 v output and the specification was changed for multi outputs.
I spose the actual point where the short occured is going to have an effect, if it were on the supplies shortest 12v leads that would be different than inside a peripheral on the longest lead, a 0.5 ohm short would pull 24a and not flip the overcurrent circuit on a supply rated at this or higher.

I pulled this from a pc website:

Short circuit protection only works if there's minimal to no resistance in the short (like two wires touching or a hot lead touching a ground like the chassis wall, etc.) If the short occurs on a PCB, in a motor, etc. the resistance in this circuit will typically NOT trip short circuit protection.

 

[Sceadwian]

if it were on the supplies shortest 12v leads that would be different than inside a peripheral on the longest lead, a 0.5 ohm short would pull 24a and not flip the overcurrent circuit on a supply rated at this or higher.

The reasons for this could be that the over current protection on an ATX supply (outside of the fuse on the mains side) isn't technically over current protection, it's DI/DT protection (current rise over time integrated from a monitoring circuit), the supply could go from no load to full load (and a bit beyond) relatively quickly as long as the DI/DT limit isn't breached. Shorts have incredibly high DI/DT curves, other faults with line resistance or a slow enough loading like a motor may smooth the DI/DT curve and possibly not allow it to trip, but I've never actually seen this occur.

A decent quality modern supply shouldn't have too much of a problem, when used as a bench supply the output should be fused, the type of load expected will determine the type of fuse, for general use I'd think a fast blow rated for 10% over the loads max rating would be a decent one. Or just a straight up standard blow fuse. Fuses are thermally triggered so they're more a power trip than a current trip device. The fuse will protect the load from true over current, the DI/DT will protect the fuse from dead shorts.

If you want a nice power supply a cool feature to add (that would be relatively simple) is an additional sense resistor on each output line and an opamp comparator using a POT to allow adjustable current tripping from a Mosfet output. Far more reliable than a fuse (as long as the FET is over rated for the current it will pass) You lose a little bit of voltage, but depending on how much you want to modify the power supply you can adjust the internal control loop sense resistors to tap outside of the additional circuitry.

Depends on how far you want to take it. Once you get to a certain point you'll just want to buy a decent one more than likley.

 

[dr peppers]

Nice explanation.