Adventures in hobbytronics: bench power from ATX PS

[Analog Kid]

Latest project completed here at Analog Kid Laboratories, GmbH:
A replacement bench power supply from a salvaged ATX unit:







I made one from an older, larger unit (mid-1990s) but it recently up and died. Far too complex for me to diagnose and repair.
But I had this li'l PS someone had tossed out. Nice small form factor.
Unfortunately there wasn't enough room behind the "front panel" for binding posts. Sat and scratched my head, then figured I could just put on an extend-o-matic front panel for them.
The metal for the panel came from a cut-up Macbook, nice thin but strong aluminum. Mounted it on 4 long screws and nuts.

Had to cut away a bit of the sheet metal to clear the top 2 screws.
The last snag was that I hadn't figured on the fan sitting where the power switch on the left was. But that was easily solved by carving away one corner of the fan housing.

So now have a nice, powerful, stable (but fixed) bench power supply to play with.

One question, though:
Do these supplies not like being run without a load? Should I put in some small load resistors?
The last supply lasted a long time without any such, and it was an older design.

[xrunner]

About how many amps can you pull on +12V?

[Analog Kid]

About how many amps can you pull on +12V?

According to the label:

+12V @17A
-12V @0.5A
+5.08V (????) @ 14A
+3.3V @ 8A

(Weirdly precise number for 5V, that.)

[mariush]

Voltages will change as you put load on it .. looks like group regulated design

As you put load on 12v (ex try a few 12v incandecent car/motorbike bulbs) the voltage on 5v will decrease, potentially even below 5v

Also note that 5v and 3.3v usually to "share" input, they're produced from same transformer winding, so they have a total budget of something like 100w, may not be able to max out both of them at same time.

[Analog Kid]

Well, you can stop worrying about that flaw:
I don't plan on putting anything near that load on this little power supply.

Nobody has answered my question yet about operating this with no load.

[Andy Chee]

One question, though:
Do these supplies not like being run without a load? Should I put in some small load resistors?
The last supply lasted a long time without any such, and it was an older design.

Depends on the circuit topology.

If it's a flyback, no load needed.

If it's a forward converter or half-bridge, yes, need a load.

"Older design" doesn't really tell us anything.

Did you take any pictures when you opened it up?  Should be able to guess the topology by the inductors (or lack thereof).

[Analog Kid]

Meant to take pics of the innards but didn't.
Don't really want to open it up again.

Isn't the topology of PC power supplies pretty standard and well-known?

I did see a bunch of inductors in there.

[Andy Chee]

The circuit topology does vary, some use switching transistors, others use MOSFETs.  Forward converters were robust against blowing up, due to simultaneous transistor switching, as opposed to alternate switching in half-bridges (if both turn on, it's a direct short across the +340V rail).  But forward converters also are less efficient with transformer utilisation for the same power output, hence the move towards half-bridges.

Some "really" old designs were fundamentally AT designs, converted to ATX by hanging a 3.3V regulated rail from the existing 5V rail.  And newer designs are resonant LLC topology to increase switching efficiency, more power less heat.

Flybacks notably will lack a large output inductor (there might be a small one for filtering purposes).

[negativ3]

A photo of the label or make/model would prevent advice heading into guesswork.

[Analog Kid]

[negativ3]

Minimum loads are not specified but combined/individual loads are.
I'd profile (observe) the outputs at different loads to see if any significant ripple appears at low load, then you can decide if/what you need regarding clean supplies.

Motherboards generally have quite complex regulation stages to further refine raw PSU output so further conditioning may be a good idea, depending on your use case of course.

External current limiting may be a good option, fuses at minimum, but you could tie back to the enable pin as well.

[SteveThackery]

If it's a flyback, no load needed.
If it's a forward converter or half-bridge, yes, need a load.

@Andy Chee, I think AK was asking whether operating with no load is actually harmful to the PSU. I assume you were talking about whether it can regulate the voltages with no load?

[Analog Kid]

@Andy Chee, I think AK was asking whether operating with no load is actually harmful to the PSU.

Yes, that's eggs-ackly what I'm asking.

I'm really not too worried about it: as I wrote, my previous ATX PS worked just fine for years running without a load (on most of the outputs).

If anyone knows differently, then hopefully they'll speak up here.

[Andy Chee]

@Andy Chee, I think AK was asking whether operating with no load is actually harmful to the PSU.

Yes, that's eggs-ackly what I'm asking.

I'm really not too worried about it: as I wrote, my previous ATX PS worked just fine for years running without a load (on most of the outputs).

If anyone knows differently, then hopefully they'll speak up here.

Yes.

For the aforementioned forward and half-bridge topologies, running without sufficient load may cause the output voltage to rise well above the nominal nameplate rating.

Whilst the obvious concern is with damaging the load, the high output voltage may also overstress the PSU's output capacitors (particularly cheap PSU's designed with little margin), perhaps causing them to explode.

[bolek_bolek]

The PSU mentioned is not the old garden variety HB ATX but a much newer proprietary FSP custom form factor for Lenovo, with probably CM6800 controller and I would say you can run it unloaded without any issues, the modern chips can handle it well.

[MariuszD]

I also sometimes use such power supplies when I need high power. However, I do not recommend such a power supply for daily use with low-power loads; its high current output and lack of adjustable current limiting will only increase the risk of fires and cause more damage.

For example, if you connect a device to a 3.3V output using a 5m long AWG24 wire, during a short circuit the current will be limited to 8A by the wire itself. The power supply won't shut down, but the current will be a high enough that the wire will catch fire.

The use of fuses is not an ideal solution because fuses need to be matched to the powered device.

[SteveThackery]

I also sometimes use such power supplies when I need high power. However, I do not recommend such a power supply for daily use with low-power loads; its high current output and lack of adjustable current limiting will only increase the risk of fires and cause more damage.

I think that's a very good point. The ATX PSU will deliver some serious currents - easily enough to violently destroy wires, PCB tracks and components if you make a mistake. When I'm working with my bench PSU I routinely use its adjustable current limit to keep things safe.

[NiHaoMike]

Do these supplies not like being run without a load? Should I put in some small load resistors?

If it came out of a PC that's Haswell or newer, there would basically be no minimum load requirement. That became part of "Haswell ready" after the exceedingly low idle current caused some PSUs to go unstable.

[coromonadalix]

i have modded some computer psu, never had any glitch,  i did add some 5w resistors load on outputs,  just to load them slightly,  dont remember the values / but for some 100-200 ma loads

yes some derive voltages from other ouputs EX: 12vdc  gives a 5 vdc ...  you have to check

[0100010]

Nice.  I built assembled an ATX PSU many moons ago.  Had to pull it off the shelf to grab a pic.  Sadly it is ATX, and not AT, so no -5V rail.