• Please review our updated Terms and Rules here

ATX PSU as a workbench supply? Running into some quirks...

Eudimorphodon

Veteran Member
Joined
May 9, 2011
Messages
6,866
Location
Upper Triassic
Wondering if anyone on these forums who dabbles in electronics has encountered some of the "issues" I'm seeing.

Over a year ago I started using a hotwired (just stuck a loop of wire between the power-on signal and ground on the motherboard plug) ATX power supply to power breadboard experiments running at 5v. This seemed to work fine; I sourced the power to the breadboard by sticking wires in one of the 3.5" drive connectors. Recently I bought some 3.3v microcontrollers (Pi Picos) to fool with, and decided I needed to try to formalize this mess to be less jury rigged. As luck would have it I noticed that breakout boards for converting ATX PSUs into bench supplies were all over Amazon; I'd been thinking of laying out a PCB for this on my own, but these preassembled boards are far cheaper than the parts separately so I bought one. Here's the thing I'm running into with an old "Enhance" branded FlexATX power supply (2003-ish vintage?):

The cheapo boards from Amazon don't include any kind of ballast resistors or other dummy load. I remembered from the "old days" that switching power supplies didn't like running with insufficient load, but I hadn't really noticed a problem when I was just using the 5v feed from the supply, so I guess I assumed that "Modern" PSUs didn't have that problem. But what I discovered with the breakout board is while the other voltages (+/-12, 5v) were in the ballpark at least of being right (5v is 5.2v unloaded) the 3.3v feed was only putting out about 2.5v and, worse, if you hung any load off just it the PSU would shut down. An obvious thing to assume would be that the 3.3v output on this supply is bad, but on a hunch I cut some sockets off a string of dead Xmas lights I keep in my junk drawer (it's a convenient source of resistive loads), and experimentation revealed that if I load the +5v output with about half an amp's worth the 3.3v supply starts regulating properly. (Also of note, with a load on it the +5v also starts regulating much closer to a "proper" 5.0v instead of that 5.2v unloaded figure.)

That's... okay, but fiddling around with it some more it kind of seems like that there may still be a limitation that I have to draw as much or more from the +5v feed as from the +3.3v to *keep* its output from falling; IE, with about half an amp +5v I can pull at least around .35a(*) on the +3.3v feed and it's fine, but if I double the draw on +3.3v, IE, make it higher than the draw on the +5v, it goes down to around 3.1v. (Which I can then kick back up by adding more load on the +5v line.)

I guess the long and short of it would boil down to this: does this ATX power supply just suck and I should get a newer one, or are quirks like this likely to be common even in newer supplies? Should I just pile a couple amps of Xmas tree lights (or, you know, a boring old power resistor) on it and call it a day, or buy a new supply?

(* All figures are ballpark because, yeah, I'm using more light bulbs as the test loads because they're here; they run about 9 ohm or so each.)
 
*Minimal* Loading, Some PSU's AT and ATX have minimal load resistors built in, They only regulate properly when more load is attached ie: Motherboard etc, Your 3.3v is likely down regulated from the 5v rail so loading the 5v rail will result in better regulation on the 5v and 3.3v.
 
How much power (and what outputs) are you looking for?

Near term, probably not more than a couple hundred mA on 3.3v, tops, followed by maybe a couple amps on +5v if my craziest ambitions come true. (A major item on the to-do list is getting acquainted with various logic level shifter solutions so I can use a Pi Pico to solve some ”peripheral issues” for a home brew 6502 system.) So tying on a few light bulbs or investing in a couple beefy resistors to load it is probably a reasonable course of action; it’s only when I’m solely messing with 3.3v testing that the draw on that side would outweigh the other.

*Minimal* Loading, Some PSU's AT and ATX have minimal load resistors built in, They only regulate properly when more load is attached ie: Motherboard etc, Your 3.3v is likely down regulated from the 5v rail so loading the 5v rail will result in better regulation on the 5v and 3.3v.

Yeah. I made the wild guess that the 3.3v was dependent on one of the other voltages like you described, if loading the 5v line didn’t do it I was going to try adding to the 12v lines as well. The way the PSU shut off entirely with a load on the 3.3v line alone was what really initially confused me. I’m guessing there must be some “power good” circuit that crowbars it because of the undervolt before the 3.3v regulator can draw enough load from 5v to make everything happy?
 
Yeah. I made the wild guess that the 3.3v was dependent on one of the other voltages like you described, if loading the 5v line didn’t do it I was going to try adding to the 12v lines as well. The way the PSU shut off entirely with a load on the 3.3v line alone was what really initially confused me. I’m guessing there must be some “power good” circuit that crowbars it because of the undervolt before the 3.3v regulator can draw enough load from 5v to make everything happy?
Yep, Some ATX PSU's have a sense wire with the 3.3v wire at the Motherboard socket end or if not it will be sensed from somewhere else.
 
A lot of older ATX power supplies used group regulation on the various power rails, but there was no standard to how it was done, every manufacturer did something different. Usually the 3.3v and 5v rails were regulated together, and the 12v was regulated by itself, but that wasn't always the case. On really cheap trash supplies, all of the power rails would be group regulated together.

On a group regulated power supply, the power supply is designed to expect specific load ranges across all of the rails. When you start loading the supply out of what it was designed for, say everything on the 3.3v rail, the power supply will have erratic behavior and shut down.

More modern ATX power supplies from the past 5-10 years or so have started using buck regulation instead. They're basically big beefy 12V supplies and use buck regulation to get the 5v and 3.3v rails. On these types of supplies, you can generally load the individual rails by themselves and not run into issues.
 
I'd recommend an inexpensive Chinese import bench supply. They're not expensive and are more suited to the task.
When you get through prototyping, you can choose a standard fixed-output PSU. At a couple hundred mills, a plain old "wall wart" might be the easiest.
 
I'd recommend an inexpensive Chinese import bench supply. They're not expensive and are more suited to the task.

Most of the cheap adjustable ones only output a single voltage at once, which was the main attraction of breaking out an ATX supply.

I’m definitely not above using a wall wart even for prototyping, but the closest to 3.3v I could find in the junk box was 3.6v and kind of… sketchy.
 
Add a cheap linear regulator on a 5V wall wart, if you want 3.3. Even a veneered and generated LM3940 will be good for an amp. Comes in SOT as well as TO-220.
 
I’m definitely not above using a wall wart even for prototyping, but the closest to 3.3v I could find in the junk box was 3.6v and kind of… sketchy.
How about one of those adjustable Universal wall wort type power adapters, I have a couple i bought years ago, They go from 0/3/4.5/6/7.5/9/12v @ 300 mA, The 3v setting on mine gives out a stable 3.285v.
 
Months later...

Decided to get off my rear end and actually work on a stalled project, and for that I needed a bench power supply. So I got creative. For anyone who needs a laugh here's what I came up with; witness the sunk cost fallacy in all its glory. (I'd already paid for the ATX breakout board, might as well use it.):

benchpsu.jpg

To "fix" the balky 3.3v regulation on the power supply I had I cut 16 sockets off a dead xmas light string and wired them into 8 parallel dual-socket serial pairs. (Threaded classily into a hunk of scrap board from the garage.) This assembly is soldered to the back of the breakout board on the 5v and ground pins and sucks significantly more than an amp, which seems to do a fine job stabilizing the 3.3v supply. (5v is rated for 12 amps, so it can spare it.) Used a few more chunks of scrap board to make a couple rakishly slanted stands to hold the PSU breakout, and screwed the mess down to a piece of Lexan with rubber feet in the corners. The PSU is mounted with a set of zip ties, and just for a little extra peace of mind slapped on a $2.50 miniature voltmeter that monitors the 3.3v line. Seems rock solid at around 3.35v.

So far so good, I guess. I anticipate it might be handy to have both the screw breakouts and the drive plugs available. And with all those lights it's definitely no problem telling if it's switched on or not...
 
Every day on the workbench is a party.

It's like Christmas every time I get a chance to tinker. Sucks when it only comes once a year. :p

I was pleased to discover after not touching it for a year my video generator breadboard still worked after blowing the dust off it. Next step is to start pawing through my IC stash and see if I have enough parts on hand to implement the next round of weirdness.
 
I wouldn't worry about 3.6V vs. 3.3V, if that's your only option. I've got a couple of PIC32MX chips here that don't function reliably at 3.3, but work fine at 3.6-3.7.
 
Back
Top