Another ATX to bench supply conversion

[Imalayan]

Hey,

after messing around for quite some time with some dodgy 9V to 5/3.3V breadboard converter, I thought, I'd get a "real" powersupply, with y'know knobs, plugs and blinking lights.

And having an old ATX PSU lying around I thought I'd try my hand at one of those infamous ATX to bench supply conversions. I had only done ready-to-solder kits and some messing around on breadboards before, but I thought to myself "I litterally just have to pass the rails thorugh to some switches and banana recepticles, how hard could that be."

About 2 months and ~50 hours of work later, I can attest that it can be quite hard (If you don't know what you are doing and have neither the tools nor skills to compensate.)

I had put everything together for a quick test before final assembly and all seemed well. Then I disassembled everything, painted the box, assembled and soldered everything. Now here is where I hit a wall: The thing wouldn't turn on. The standby power would work, but when I "started" the PSU (shorting the ON cable to ground) it would start to get to the desired voltages, but shut down mere moments later. To my laimans eyes it looked like one of the automated safety shutdowns modern PSUs apparently have integrated.

So I checked and rechecked every connection, thinking I surely must have shorted something somewhere. Having no luck with that, I decided to one by one disconnect everything from the PSU, trying to find the faulty section by process of elemination. Only that faulty section never turned up. I now have the bare PSU module with only mains power attached and it still exhibits the same behaviour, wich leads me to suspect that I must have damaged the board somwhere in that process. A visual inspection finds nothing wrong, wich aligns with me never noticing anything out of the ordinary during assembly. I might have dropped some solder or something on some pins though, so idk.

I know remote troubleshooting is inconvenient at best and regularly annoying, so while I'd be happy to try any troubleshooting steps anyone would think of right now, I have basically given up on that PSU module and have already bought another one from ebay for 5 bucks. I'm not very comfortable with messing with that PSU outside from basic probing, with all the big caps holding 300V and stuff, so buying another one seems like a prudent choice. I debated switching to a custom built power supply, but especially with all the panels already drilled, I don't think the monetary investment would be worth that.

I would however like to not blow up another one of those ATX PSU's so, I'd greatly appreciate it, if someone could look over my shematic and tell me if there is something I could/should improve, especially if there is something that might have damaged the PSU. Thanks in advance for any help



I attached a circuit diagram to this post (the png is a bid dodgy if not zoomed in, but .svg's aren't allowed :/)

My (heavily feature creeped) goals were:
- 3 fixed rails directly from the PSU (3.3V, 5V and 12V)
- 1 variable rail with constant current mode
- a 9V rail, since that's a common voltage for powering stuff
- a 12V fan rail for powering fans while soldering
- a earth passthrough for an ESD strap
- all outputs fused (as a crude form of safety current limiting)
- all rails seperatly on/off-switchable and an indicator (LED) for each rail


Some notes:
- The ATX PSU used is a BeQuiet SystemPower 7
- for the variable rail I used a cheap DC to DC variable power supply module from ebay (https://www.ebay.de/itm/CNC-DC-DC-Buck-Boost-Converter-CC-CV-4A-36W-Adjustable-Regulated-Power-Module/133404944413) (German ebay, I apologize)
- this is the buck converter used: https://www.ebay.de/itm/3x-LM2596-DC-Step-Down-Spannungswandler-Arduino-Modul-Regler-LM2596S/252785167788

[gcewing]

Have you tried connecting a load to any of the outputs? Some switchmode supplies don't like running without a load.

[0culus]

FYI, do realize that even if you do get it working, you're going to quickly find that the switching noise makes it a crappy bench supply, especially if you are working with analog circuits. A multi output linear supply with proper voltage and current limiting will far more useful to you as a tool.

[Imalayan]

Have you tried connecting a load to any of the outputs? Some switchmode supplies don't like running without a load.

I had it running without any load before, but I did shorten and remove cables, so maybe the ratsnest had acted as some kind of inductance (picking up something from the transformers or the ripple itself). I'll try adding a load and see if that changes anything. thanks for the tip
EDIT: adding a load sadly didn't help :/

FYI, do realize that even if you do get it working, you're going to quickly find that the switching noise makes it a crappy bench supply, especially if you are working with analog circuits. A multi output linear supply with proper voltage and current limiting will far more useful to you as a tool.

yeah, I am/was aware of that, but ultimately my "bench" has various electronics inducing all kinds of noise anyways (monitors, powercables running directly under the worksurface, several plug pack transformers). I looked into linear supplies, but I didn't find circuits that looked both safe and feature rich enough, but maybe I just don't know where to look (hints ofc appreciated).

[MarkF]

What's the point of all the transistors?  Big waste.
Simple resistor and LED to each rail is all you need.
Using a transistor to limit the current is a bad idea.

And the Standby LED indicator doesn't have a resistor.

[Imalayan]

What's the point of all the transistors?  Big waste.
Simple resistor and LED to each rail is all you need.
Using a transistor to limit the current is a bad idea.

And the Standby LED indicator doesn't have a resistor.

Thanks for having a look. Like it says in the notes, the LEDs are integrated into the switches and have some kind of currentlimitng built in. This is why I have to run them off the 12V rail instead of just using a resistor on the corresponding rail. Believe me, I wasn't thrilled either when I found out... (It's most likely just some high value resistor in series, but I didn't wanna draw in something I wasn't sure about, so I just added it to the notes.)

EDIT: Just to clarify the transistors are just switching, not currentlimiting. They are opened to about 12mA, if I recall correctly, and the LEDs top out at 7mA, so whatever they are using to currentlimit in there is the deciding factor.

[0culus]

FYI, do realize that even if you do get it working, you're going to quickly find that the switching noise makes it a crappy bench supply, especially if you are working with analog circuits. A multi output linear supply with proper voltage and current limiting will far more useful to you as a tool.

I had it running without any load before, but I did shorten and remove cables, so maybe the ratsnest had acted as some kind of inductance (picking up something from the transformers or the ripple itself). I'll try adding a load and see if that changes anything. thanks for the tip
EDIT: adding a load sadly didn't help :/

FYI, do realize that even if you do get it working, you're going to quickly find that the switching noise makes it a crappy bench supply, especially if you are working with analog circuits. A multi output linear supply with proper voltage and current limiting will far more useful to you as a tool.

yeah, I am/was aware of that, but ultimately my "bench" has various electronics inducing all kinds of noise anyways (monitors, powercables running directly under the worksurface, several plug pack transformers). I looked into linear supplies, but I didn't find circuits that looked both safe and feature rich enough, but maybe I just don't know where to look (hints ofc appreciated).

Other noise is certainly a problem, and you may find yourself mitigating it in the future. My point is that the switching noise coming from your modified ATX supply is conducted EMI that will cause more problems in your circuits than just some radiated EMI around the bench. I don't know what you mean by safe enough. I would say building your own linear supply is much more straightforward than building or modifying a switch mode supply. Certainly far fewer "gotchas". Both of course are going to involve mains potentials...

[Imalayan]

Other noise is certainly a problem, and you may find yourself mitigating it in the future. My point is that the switching noise coming from your modified ATX supply is conducted EMI that will cause more problems in your circuits than just some radiated EMI around the bench. I don't know what you mean by safe enough. I would say building your own linear supply is much more straightforward than building or modifying a switch mode supply. Certainly far fewer "gotchas". Both of course are going to involve mains potentials...

By "safe enough" I mean something that not only can take, but is designed for stupid mistakes like shorting the outputs for an extended period of time, reliable current limiting, having a soft start, temperature limiting, etc.

I certainly debated changing gears to a linear supply after reading your message. And it's not that I'm not confident that I could build something like that (following some guide), but rather that I don't trust myself to build something good. With an prebuilt ATX supply, I know that it's designed to be plugged into the wall by a consumer and just work and that it has some features already built in that are deemed at least acceptable enough by some consumer protection standards.

I'm quite frankly not familiar enough with all the concepts to truly understand, much less judge a power supply design. So I have no way of determining if a given design is actually good/safe/decent/horrible and I'm not really confident enough to plug something into the wall, wich I can't judge.

Basically with an ATX conversion, I know what I'm getting: A noisy supply.
But with a self built supply (at least at this stage) I have to take a gamble and while I'm willing to work on mains voltage, I'm not willing to gamble on it.

If, however, you happen to have some good resources on linear power supplies on hand, maybe the very next project will be a linear supply^^


A truly beginner question at the end:
Is there a reason why I couldn't just put some caps on the outputs to bleed the noise to ground?

[0culus]

One resource that may or may not be useful depending on your level of experience is the new Art of Electronics X chapters. In it, the authors do a pretty deep dive into an Agilent branded bench supply and take a look at how the professionals designed one and why.

[bob91343]

Capacitors may or may not help.  Much of the noise is radiated, more is conducted via input and output leads.  The chassis is always a lousy ground, especially when it comes to noise.  And each capacitor has its limitations with regard to ESL and ESR.