ATX Power Supplies

[engrguy42]

As I put the finishing touches on my ATX power supply conversion, I'm still trying to sort out what I sense to be a (very) strong bias against ATX power supplies here. I'm trying to figure out to what extent the apparent bias is, well, just personal bias, and to what extent it is justified and based in fact.

Now one of the big concerns seems to be that the ATX supplies don't have current limiting, and therefore your expensive circuit can get damaged. But isn't that a bit like saying "Wrenches make crappy hammers". Well, yeah, it's presumed you know enough to use the right tool for the job, and AFAIK nobody ever claimed that an ATX supply includes current limiting. So don't use it where current limiting is important.

On the other hand, a decent ATX supply has internal protection to shut off overcurrents over their set point. As well as overvoltage and other protective elements, depending on whether it's a piece of junk or name brand that follows (or exceeds) ATX specs. As I've shown before, I've done tests to show that my supply (and presumably other name-brand ATX-compliant supplies) has supervision circuitry to instantly shut off the supply at or below rated current. No spikes (unless they last for 10mSec or less and I missed them...). So we're back to knowing the specs of your power supply and applying it appropriately. Like you do for any other piece of equipment.

So are people actually saying "never, ever, ever use an ATX power supply on your bench or else worlds will collide, and there's absolutely NO possible use case for you to ever justify using a converted ATX supply", or is it more like "yeah, they're fine when appropriate, as long as you consider their limitations, just like you do with every other piece of equipment".

Is there something I'm missing? They're fine for many gazillions of computers around the planet, and assuming you buy one with all of the important certifications and protection (unlike some of the equipment many of us use on our benches  ), isn't it just a matter of making sure you don't do something dumb and apply it in the wrong situation (like on a $10,000 circuit where you need an appropriate power supply with current limiting in case you do something dumb?).   

Personally, I just used mine to test/calibrate a high current sensor, while my main power supply was busy. Worked fine. Was that a mistake?

[tautech]

They have their uses sure but NOT as a precision bench PSU.

[engrguy42]

They have their uses sure but NOT as a precision bench PSU.

I suppose it depends on how you define "precision", doesn't it? Mine was "precision" enough to help me calibrate a current sensor. I assume you're referring to more professional-oriented requirements for high precision to meet industry specifications, etc.?

[Nusa]

Just don't call them bench or lab supplies in forums like this, as that triggers people that know what a real bench supply is, and how to use it.

Sure, it's fine as a general purpose constant voltage supply for higher-current projects. But for many projects, it's like using a sledgehammer on a pushpin and then wondering why there's a hole in the wall.

I wouldn't use it for breadboard projects...12 volts at 50+ amps for 10 ms (not "instantly"!) applied to the wrong pin by mistake is going to be fatal to a lot of components. And that's for a dead short. If there's some small resistance in the circuit, you've got a power supply that'll happily pump 40 amps until the smoke is released from something.

[engrguy42]

Just don't call them bench or lab supplies in forums like this, as that triggers people that know what a real bench supply is, and how to use it.

Sure, it's fine as a general purpose constant voltage supply for higher-current projects. But for many projects, it's like using a sledgehammer on a pushpin and then wondering why there's a hole in the wall.

I wouldn't use it for breadboard projects...12 volts at 50+ amps for 10 ms (not "instantly"!) applied to the wrong pin by mistake is going to be fatal to a lot of components. And that's for a dead short. If there's some small resistance in the circuit, you've got a power supply that'll happily pump 40 amps until the smoke is released from something.

Thanks. I'm curious of anyone has any references of actual test data for decent, name brand ATX supplies that shows what the actual transient overcurrent/overvoltage can be? Because my testing shows it can barely get up to its rating before the whole thing shuts down. No transient overcurrents at all. Again, my tests had 10mSec sample resolution and showed nothing. Hypothetically something could have happening in-between a couple samples, but that seems a big stretch. Something like 16A(?) rating for 12v rail, and it shut down around 12A.

Has anyone actually done tests to show these high currents? Maybe I'm missing something?

[masterx81]

I've used for years an atx psu ( 2003 ic) modded with voltage and current limiting (sense resistor, opamp and comparator). Worked really nice. Then i've bought the entry level bench psu korad kp3305... it's another thing. More precise, more stable on transients, almost no ripple... And for who use "true" bench supplies the korad is a toy. I trust it to power sensitive boards (as dual i can do both 3.3v and 5v), while with the atx i was not so confident and i was using always some linear regulation after the switching.
The bigger thing that i miss is the 20a capability, but i'm planning to buy a dps3020 for that.
Would be nice also an rd6006 but is almost redundant with my current setup.

[tautech]

They have their uses sure but NOT as a precision bench PSU.

I suppose it depends on how you define "precision", doesn't it? Mine was "precision" enough to help me calibrate a current sensor. I assume you're referring to more professional-oriented requirements for high precision to meet industry specifications, etc.?

Not at all.

Without the ability to limit current an ATX supply can not only vaporise components but PCB traces too. Not only will there be risk of injury and that could to eyesight or burns to the person. Breadboards have been mentioned too and they could be damaged with high and uncontrolled currents also.

Those with enough experience can mitigate some risk however the novice with limited knowledge and experience can be injured or permanently damage DUT's using uncontrolled current PSU's so as a responsible forum we are best to advise the use of user adjustable current controlled PSU's as we have absolutely no idea of the experience of our readers.

It was not long ago the dangerous advice of how to float a scope was freely given as at one time it was common industry practice whereas today for just a couple of hundred dollars purchasing a differential probe rightly consigns that dangerous process to the foolish and to history....and yes, at one time I did it too !

You are not the first here to extol the virtues of ATX PSU's and certainly not the first to be told forgettaboutit !

The first half decent PSU I built used a LM338 5A regulator capable of 7A peaks but as it was otherwise uncontrolled current I rarely use it now instead opting for a SPD1305X precision PSU that more than suits my quite basic needs.
For a while I used only a MCH K305D, a tiny 30V 5A SMPS however as these are somewhat electrically noisy it's been relegated to mostly charging duties.
At one time I used the 3 channel SPD3303X-E hacked to the X model but customers keep buying them off me rather than wait weeks for more stock to arrive.

[madires]

An ATX PSU is fine for powering your PC or something else which needs lots of power. Imagine you are repairing a laptop with a power problem (doesn't turn on, no LEDs lit). A common cause for that is a shorted component, like a MLCC for example. Without current limiting the PSU will push as much current through the laptop's board as possible. Components in the power circuitry will heat up, might blow up, and PCB traces will get hot or even vaporize. With current limiting you'd limit the current to a reasonable value and look for the shorted MLCC without damaging the board.

[engrguy42]

Thanks. So far it seems, again, the major issue people have with ATX power supplies is that they don't have current limit capabilities. Agreed. They don't.

Just like a crescent wrench "sucks" because it doesn't drive wood screws. 

But other than that, does anyone have any actual data (not just opinion) to show that ATX power supplies are dangerous and worthy of mock and scorn and ridicule?

Personally I don't care whatsoever about ATX power supplies. I don't like them or dislike them. They're just another tool, and I use them where appropriate. And I don't get "triggered" (as someone here mentioned) if someone dares to discuss them. Just looking for facts, to see if there's some limitation/danger I'm missing and need to protect for on my bench.     

[Electro Detective]

So is there a frugal way to get some limiting on ATX PSUs?

lamps in series, ballast trannies, wet feet  etc   

[BravoV]

Personally, I just used mine to test/calibrate a high current sensor, while my main power supply was busy. Worked fine. Was that a mistake?

Its fine if it fits in at your existing tasks, its just do NOT be too over confident ... or should I say be naive that a PC PSU, is all you need for more "serious" electronics tasks, thats all.

As this is like reading a boy who can heat up a big nail using gas lighter, and did some simple soldering by melting some solder at PCB, and yelled proudly that he has the ultimate soldering tool, and denying the need of advanced and expensive soldering tools, as he is so confident that only using that nail only, can properly solder/de-solder SMD components. 

Hint, forget about the owners of the PSUs, just look at the bench PSUs collections in my sig at below right, and think why the hell all those manufacturers built them from the 1st place and reasons for that.

My 2 cents.

[Whales]

Use whatever is appropriate to your tasks.  I'm sure a good 50% or more of this forum wouldn't approve of my lab equipment



I'M HERE TO DONATE SOME BLOOD

[Nusa]

So is there a frugal way to get some limiting on ATX PSUs?

lamps in series, ballast trannies, wet feet  etc 

If you mostly care about the oh-shit scenario, a suitably low value in-line fuse on the PSU output is cheap. Remember to have spare fuses, or you'll be tempted to eliminate it after the first oops.

If you want real limiting, google "dc current limiter" and see what works for your situation.

[jemangedeslolos]

Thanks. So far it seems, again, the major issue people have with ATX power supplies is that they don't have current limit capabilities. Agreed. They don't.

Just like a crescent wrench "sucks" because it doesn't drive wood screws. 

But other than that, does anyone have any actual data (not just opinion) to show that ATX power supplies are dangerous and worthy of mock and scorn and ridicule?

Personally I don't care whatsoever about ATX power supplies. I don't like them or dislike them. They're just another tool, and I use them where appropriate. And I don't get "triggered" (as someone here mentioned) if someone dares to discuss them. Just looking for facts, to see if there's some limitation/danger I'm missing and need to protect for on my bench.   

I don't understand why you are looking for data ?
They are already there in the datasheet.

In case of problems ( HS component, short circuit, handling error, tiredness), there is no current limiting to remain within reasonable orders of magnitude.

Even if the current protection is very fast on ATX power supplies, there will probably be something along the way that will create resistance.
So you will to stay below the max current of your ATX power supply and the OCP is not going to trigger. ( >40A on the +12V rail for mid range 500W ATX PSU ).
You're just going to have a lot of smoke and/or an exploding component.

[engrguy42]

jemangedeslolos, are you saying that an ATX supply's OCP protection will only operate if the TOTAL of all the 12v rails exceeds the total current rating of all those rails?

For example, my ATX supply has 3-12V rails, rated 8A, 16A, and 18A. As I've posted before, I actually tested this and applied current to one of those rails. The power supply shut down at 12 amps (see attached), not >40A.

Maybe my test was in error? Or maybe it depends on the quality of the power supply and whether it fully complies with ATX?

In any case, as I've said many times before, yes, an ATX power supply does NOT have current limiting, nor is it intended to.

What I'm looking for is OTHER issues that cause people to dislike them so I can make a factual list of limitations.

 

[Shock]

Use appropriately rated fuses on the outputs for basic circuit protection or get as elaborate as you want. As the supplies are normally mounted inside a case add some psu mesh to prevent screws and tools from finding their way into the air/fan vent. As always check the ground/earth is properly connected and use eyewear for soldering, cutting and blowing up testing electronics.

Saying you must have this or that to do electronics is a crock of shit. You can get by with very little to start off with. A proper adjustable bench supply is handy for obvious reasons but people make their own from parts or modify or use what they have, hobby electronics in a nutshell. In the US there is also a ton of surplus bench supplies on the market so I'd consider that first before dropping hundreds on a new supply.

[engrguy42]

Use appropriately rated fuses on the outputs for basic circuit protection or get as elaborate as you want.

Careful...when I suggested using fuses for ANY important circuits to allow for the fact that ANY power supply can fail, not just ATX, I got bombarded by judgemental finger-wagging. "Yeah, but what if your components can't handle the transient current before the fuse has a chance to operate and your entire city explodes??".

[Fungus]

Thanks. So far it seems, again, the major issue people have with ATX power supplies is that they don't have current limit capabilities. Agreed. They don't.

Nope.

The problem is people calling them "bench power supplies". They're not, they just "power supplies" (even if you place them on wooden benches).

Avoid using the word "bench" and you'll be fine.

[jemangedeslolos]

jemangedeslolos, are you saying that an ATX supply's OCP protection will only operate if the TOTAL of all the 12v rails exceeds the total current rating of all those rails?

For example, my ATX supply has 3-12V rails, rated 8A, 16A, and 18A. As I've posted before, I actually tested this and applied current to one of those rails. The power supply shut down at 12 amps (see attached), not >40A.

Maybe my test was in error? Or maybe it depends on the quality of the power supply and whether it fully complies with ATX?

In any case, as I've said many times before, yes, an ATX power supply does NOT have current limiting, nor is it intended to.

What I'm looking for is OTHER issues that cause people to dislike them so I can make a factual list of limitations.

All ATX PSU have their own architecture.
I have one Seasonic with just 1 x 12V rail (70A on the 12V) and one Be quiet with 4 x 12V rail.

Even with only 10-12A on 1 rail, it is enough to blow up your logic board, cause cancer, kill seals and spread a global virus.
PC power supplies are made for PCs ( crazy right ? )

There is no need for 0.000000001% load and line regulation, ripple and noise is terrible and you have no UI.
I don't know how it is possible to have precision result with that.

It is not a solution to do serious work.

[engrguy42]

With all due respect guys, I'm not sure how many of you are actually engineers, but I'm trying to imagine an engineer telling a boss:

Well boss, we shouldn't consider this piece of equipment, EVER, for these reasons:

1. It "sucks"
2. It's not "precision"
3. It's not "bench" quality
4. It's not for "serious" work
5. It's analogous to a stupid kid doing stupid stuff
6. It doesn't do stuff it's not designed to do

If I'm the boss, my response is "Oh really? You're fired". I'd then hire a guy who knows how to do a real engineering evaluation.

Engineers look at facts, not emotional, judgemental finger wagging. 

[Shock]

Variable unregulated supply is a bench supply, that is the minimum qualifier. A regulated smps supply is just a power supply.

I'd fire you for being silly. Was just seeing when you signed up, you are also starting to exhibit troll like behavior, calm down a little.

[engrguy42]

Variable unregulated supply is a bench supply, that is the minimum qualifier. A regulated smps supply is just a power supply.

I'd fire you for being silly.

Really? So the difference between "bench" and something like an ATX is the "variable" part?

Thanks. I learn something new every day.

[jemangedeslolos]

With all due respect guys, I'm not sure how many of you are actually engineers, but I'm trying to imagine an engineer telling a boss:

Well boss, we shouldn't consider this piece of equipment, EVER, for these reasons:

1. It "sucks"
2. It's not "precision"
3. It's not "bench" quality
4. It's not for "serious" work
5. It's analogous to a stupid kid doing stupid stuff
6. It doesn't do stuff it's not designed to do

If I'm the boss, my response is "Oh really? You're fired". I'd then hire a guy who knows how to do a real engineering evaluation.

Engineers look at facts, not emotional, judgemental finger wagging.

Is it possible to have a live webcam during your next job interview ?

 

[engrguy42]

With all due respect guys, I'm not sure how many of you are actually engineers, but I'm trying to imagine an engineer telling a boss:

Well boss, we shouldn't consider this piece of equipment, EVER, for these reasons:

1. It "sucks"
2. It's not "precision"
3. It's not "bench" quality
4. It's not for "serious" work
5. It's analogous to a stupid kid doing stupid stuff
6. It doesn't do stuff it's not designed to do

If I'm the boss, my response is "Oh really? You're fired". I'd then hire a guy who knows how to do a real engineering evaluation.

Engineers look at facts, not emotional, judgemental finger wagging.

Is it possible to have a live webcam during your next job interview ?

 

Nah. I've been doing electrical engineering and managing (and hiring) electrical engineers for almost 50 years now. No more job interviews for me. Actually I was hoping to retire soon, but with the market the way it is that might have to wait.

Though I've never been called a "troll".     Wow.

Not sure what it means other than "someone I hate cuz he disagrees with me" maybe?

[masterx81]

You speak in that way because you have never tried a proper bench psu. Atx psu for lower the ripple have LOADS of output capacitance. And despite that, dynamically they can't mantain a precise voltage.
On my atx i've also added the current control. But was not so precise, with dynamic load was HORRIBLE. I never trusted it.
I've had your same opinion before using a bench psu, but now that i have one, the atx is on one side, i use it for something coarse but never anymore attached to a board. With a proper bench supply you can can set with mv/ma precision the voltage/current limits, dynamically they are way better. With a dynamic load i was also able to set the atx psu in some sort of resonance with incredible voltage spikes.
On atx supply i wasn't able to have a precise current control, my supply had 2 ranges, 0-25v/0-10a, 0-10v/0-20a (from 5v rail and 12v rail, 2 ranges for have a bit more precision on the settings), i was able to let it goes to 0v (normally on the 2003 controller the minimum is 2.5v). I've lost a lot of time developing it, and was really far away from being usable on sensitive device. As stated before, i always had the need to use some other more reliable voltage and current control after it.
Maybe with a good development can become more usable, but after i've already spent so much time for that result, i think that simply not worth.