Diagnose and Fix an ATX PSU

[balnazzar]

Hi fellas.

I have an ATX Psu that doesn't work properly. It's a rather good product (XPG Core 850W, Gold). No matter the fact that it powered a system that never absorbed more than 400W, after less than a year it started powering itself off unexpectedly.
If one tries to turn the PC on, the relay inside the PSU clicks on, but nothing happens. Only after a while it allows itself to be powered on again.

It's on warranty, but I want to use it for learning purposes. So, I'd like to know:

1. How to diagnose & fix it (I have multimeters, oscilloscope, AWGs, etc...)

2. How not to die due to the huge capacitors that are inside it, and could be still charged.

Thanks.

[TheMG]

It's under warranty, find something else to use to learn on.

If you've never worked on a switchmode power supply before, a modern ATX computer power supply is not a good place to start. They're quite complex, with things like PFC pre-regulation, active rectifiers, supervisory circuits that will shut down the PSU under a variety of conditions. Usually the construction is quite dense and not easily serviced.

I would choose a simpler power supply to learn on. Start simple, and once you've figured out the basic concepts move on to more complex stuff. Without a firm understanding of how these things work, it will be quite challenging to figure out what is causing it to shut down.

[balnazzar]

I would choose a simpler power supply to learn on.

Linear?

Or, I have an old (like early 2000s) 200W ATX power supply. It works, but it's useless. I can easily sacrifice it for learning purposes.

[TheMG]

No not linear, but a more basic SMPS such as the ones that would be found in a USB charger, and other small electronic devices. These are usually just simple flyback SMPS nothing fancy. Do read up on the theory of how a flyback SMPS works that would be a good starting point.

[MathWizard]

My whole post just got lost, anyways here's a schematic of a TX750W PSU, in LTSpice ( a free circuit sim/design program). That sch. doesn't run, it's more for me making a map of the PSU's I worked on.

You can find out a lot just with the PSU unplugged, and going through the circuit. Most modern ATX PSU's are pretty similar, at least 10 years ago anyways. I fixed a few, and I would always do things like un-power it before moving alligator clips in area I was worried about. Or I'd solder on little test wires all over the place. There's lot's of stuff that can go wrong and be fairly easy to track down. If you are lucky, the problem won't be on some control PCB that is full of SMD parts.

The most dangerous caps usually discharge in a few seconds to minutes after unplugging, as the auxiliary and standby sub/mini-SMPS keeps running directly off the main bulk 400V cap (usually). Most all others have some bleeder resistor or load that will drain them pretty quick once unplugged. 



There's lot's of stuff that could require soldering tho too, I hope u have a good soldering iron, there's lot's of stuff on huge heatsinks.


Intermittent problems can be a lot of things. Is it temperature or load related?


PSU's are cheap these days, if time is an issue, just replace it. But if u are like me, I always just started with the PSU's unplugged, and worked my way in from the plug. I learned a pile by mapping out a few PSU's. And taking a few completely apart, and measuring all the inductors, and SMD cap's out of circuit.

Now I'm learning more of the math/EE side.

[balnazzar]

Mh, It's a little more complex than I originally envisioned. Maybe it'd be better to start with usb chargers, indeed.

[MathWizard]

A typical 5V or 12V charger that u plug into a wall, will basically be like the auxiliary and standby circuits in a computer PSU.

The Aux voltage supply is usually like 12-15V, and it derived from the rectified mains voltage. It makes the low voltage to power various control chips and mosfet gates, on the hot side. It also makes the 5V standby on the secondary side. That runs the control chips on the secondary side, and the power button on the PC is tied into that.

When the PC pwr is pressed, basically a 5V standby circuit, uses an opto-coupler to tell a chip on the primary side, to start the circuits for the PFC boost converter (if it has it, most PSUs of any size these days do), to double the rectified mains voltage. Then feed that to the main converter, which on the other side of the big transformer/inductor, make some stepped down voltage, closer to the desired output.

There's different ways they get to the final 12V, 5V, 3.3V, -12V, but sometimes it's just stepped down, rectified, and LC filtered to a smooth level. Some use another stage of switching DC-DC converters, to take 12V and like a buck converter, step down the voltage. That's what the 1 in the schematics does.


Well on your computer, did you try running any benchmarking/stress-testing programs ? Try running a shareware program like CineBench, or OCCT. They can use a lot of power for the CPU and GPU, and see if that upsets the PSU.

And have you really narrowed down that's it's the PSU, and not the mobo ? Do u have another PC to try it in ?

You could probe various output cables, (helps to tape some test-leads or wires into the molex connector. The regulated voltages are usually pretty close 12/5/3.3V . 1 of the common things to go wrong, is the output caps, start acting more like resistors, and like a voltage divider, you start to drop too much voltage across the caps, and instead of getting 12.2V out, you get 11V, and the more current being drawn, the worse it gets. That would show up easy.

There's a few control chips, that monitor different voltages/currents. If stuff gets too high/low, the PSU can shut itself down.

It would be nice to see on an oscilloscope, what happens to the various voltages, when it shuts down, some problems will make certain voltages die before others.

There's clues to look for all over the place.

[DavidKo]

Your symptoms look same to mine PSU - capacitors. All PSU that I had death, failed due to the electrolytic capacitors.

If it is in warranty, it makes no sense to exchange them. Last PSU that I had, have some thin tall capacitors (hard to get, does not make sense from the finance point of view) glued to the PCB.

[balnazzar]

A typical 5V or 12V charger that u plug into a wall, will basically be like the auxiliary and standby circuits in a computer PSU.

The Aux voltage supply is usually like 12-15V, and it derived from the rectified mains voltage. It makes the low voltage to power various control chips and mosfet gates, on the hot side. It also makes the 5V standby on the secondary side. That runs the control chips on the secondary side, and the power button on the PC is tied into that.

When the PC pwr is pressed, basically a 5V standby circuit, uses an opto-coupler to tell a chip on the primary side, to start the circuits for the PFC boost converter (if it has it, most PSUs of any size these days do), to double the rectified mains voltage. Then feed that to the main converter, which on the other side of the big transformer/inductor, make some stepped down voltage, closer to the desired output.

There's different ways they get to the final 12V, 5V, 3.3V, -12V, but sometimes it's just stepped down, rectified, and LC filtered to a smooth level. Some use another stage of switching DC-DC converters, to take 12V and like a buck converter, step down the voltage. That's what the 1 in the schematics does.


Well on your computer, did you try running any benchmarking/stress-testing programs ? Try running a shareware program like CineBench, or OCCT. They can use a lot of power for the CPU and GPU, and see if that upsets the PSU.

And have you really narrowed down that's it's the PSU, and not the mobo ? Do u have another PC to try it in ?

You could probe various output cables, (helps to tape some test-leads or wires into the molex connector. The regulated voltages are usually pretty close 12/5/3.3V . 1 of the common things to go wrong, is the output caps, start acting more like resistors, and like a voltage divider, you start to drop too much voltage across the caps, and instead of getting 12.2V out, you get 11V, and the more current being drawn, the worse it gets. That would show up easy.

There's a few control chips, that monitor different voltages/currents. If stuff gets too high/low, the PSU can shut itself down.

It would be nice to see on an oscilloscope, what happens to the various voltages, when it shuts down, some problems will make certain voltages die before others.

There's clues to look for all over the place.

Yes, I did some testing. It's the PSU, since I swapped in another PSU of the same wattage and the computer works flawlessly. Anyhow, that's what emerged: indeed, the PSU shuts itself down after putting some load on it (cpu+gpu at max, prime95 + furmark). Note that the maximum load I can put on the PSU is not even half its rated wattage. However, it tends to shut down some seconds *after* I cut the load. It seems to go into some protection state, because if I try to turn it back on, the relay clicks, but nothing happens. Only after it cools down I can turn it on.

As for disgnosing with the scope: I'm using my scope to probe somewhat basic circuits for learning purposes. I don't have the faintest idea about where to put the probe(s) in an ATX psu... Suggestions? Thanks!

One more question: your explanation above was interesting and informative. Tell me one thing.. An ATX psu basically turns 230V AC into 3.3, 5, 12V DC. Why is there a 400V capacitor inside???

[TERRA Operative]

I have repaired a number of ATX PSU's, but usually don't bother. The two recent ones had the -5V rail that's missing from modern PSU's for powering the now obsolete ISA bus in older motherboards, so they were worth a repair.

One needed all new capacitors and the other ended up having a faulty control/supervisor/current limiter chip.

They weren't too hard to mentally break down into their logical blocks of operation, but I did draw on years of inverter welder repair experience.

If your unit is under warranty, use the warranty. Go get an old laptop/USB charger or one of those metal framed power supplies that you see cheap on ebay etc that are often used in 3D printers etc and reverse engineer that.
You can go a long way without even needing to power it on, and besides slight topology differences, the basic theory of operation is all the same, from a USB charger, to an ATX PSU all the way to a 2000A welder power source.

Input mains filtering -> PFC -> input rectification -> DC filtering -> High frequency switching -> transformer -> output rectification -> output filtering -> load

Add a control loop that references the output to control the duty cycle of the high frequency switching and there it all is.


Remember to stay safe and isolate then discharge and test before you touch!

[ingalopez]

Hi Balnazzar

On reply as how not to die...

Always be sure you bulk capacitors (those of 400V) are discharged prior to dissasemble, solder or anything inside such a power device. There is enough juice in those to give you a bad day.

Also, is good advice to use an isolation transformer to power de device under investigation. It may prevent  a shock and /or the destruction of your osciloscope.

A series incandescent light bulb (which you can bypass after checking there are no shorts) is a must.
There is a lot of information about this on You tube, if you are a starter.

On your power supply, as you describe the fault, there is probably the power supply supervisor circuit turning off the supply (and blocking it to be turned on again for some time). This includes a series of comparators and other circuitry on a chip, that will generate the "PG" (power good") signal to the mother board, and ultimately turn off the supply is detects an anomaly on the 5, 3.3 or 12 V outputs.

The most frequent cause are filter electrolitycs with high ESR , broken solders , etc. Check for bulked capacitors, and measure their ESR and capacity.

On the other hand, is true that the common ATX PSU  is quite a complex and crowded device to start with.
Good luck

[balnazzar]

There is a lot of information about this on You tube, if you are a starter.

The problem indeed is that there is too much..! Would you link some quality resources to start with? Thanks!

[fzabkar]

A series incandescent light bulb (which you can bypass after checking there are no shorts) is a must.

I also use a light bulb to discharge the 400V/200V bulk capacitors.

[DavidAlfa]

ATX working with low-mid load, suddenly dying with larger loads - primary electrolytic caps, 99% of the time.

[balnazzar]

ATX working with low-mid load, suddenly dying with larger loads - primary electrolytic caps, 99% of the time.

In all my noobness, they were my primary suspects too. If there are electrolytics in something, and that something breaks, it's quite probable they are the culprits.

[DavidAlfa]

I've recapped mine already twice in the last 10 years or so.
The first time it was pretty weird, a sudden smell like fresh grounded coffee filled the room, it took few minutes for my brain to notice I was alone at home, it was 1am, so what the hell was that freaking smell?
Few seconds later the computer went off, I could see all the electrolyte raining down over the internal parts!
Actually had few parts from other broken ATX power supplies, so I just replaced with used ones.
They seem to last 4 years or so, recently it started to die with high GPU loads, so here we go again.
I'm too lazy, I very rarely play anything, it can wait 

[tautech]

Hi fellas.

I have an ATX Psu that doesn't work properly. It's a rather good product (XPG Core 850W, Gold). No matter the fact that it powered a system that never absorbed more than 400W, after less than a year it started powering itself off unexpectedly.
If one tries to turn the PC on, the relay inside the PSU clicks on, but nothing happens. Only after a while it allows itself to be powered on again.

It's on warranty, but I want to use it for learning purposes. So, I'd like to know:

1. How to diagnose & fix it (I have multimeters, oscilloscope, AWGs, etc...)

2. How not to die due to the huge capacitors that are inside it, and could be still charged.

Thanks.

Apply basics and identify the SMPS controller then get datasheets, several if you can.
Carefully study 'Typical Application' circuits for the IC supply circuitry as it's a common failure point. Dropper resistors drift and/or go open and the VCC cap is also a common failure point.
Check VCC is not dropping below UVLO levels.
The bulk caps if the SMPS is working anything like correctly will be depleted after power is removed until the UVLO is triggered.
The controller can be powered with a bench PSU without any further mains supply and the IC output (gate drive) can be safely scoped at these low supply levels. One only needs to ensure there is no ground loop and there shouldn't be with a properly isolated DC bench supply.

[balnazzar]

I could see all the electrolyte raining down over the internal parts!

That's why they learned to put the psu bay at the bottom 

[balnazzar]

The controller can be powered with a bench PSU without any further mains supply and the IC output (gate drive) can be safely scoped at these low supply levels.

That's a good tip, I didn't think about it!

[tautech]

Forgot to mention, the VCC flyback diodes can be another failure point and the whole VCC supply is something I always carefully check....but this is in the case of dead SMPS.
As they start from the tiny currents through the dropper resistors any significant performance failings of the VCC cap will prevent them kicking off and then their supply is via the flyback where again the VCC cap must do its job to keep ripple under control.

I've long lost count of SMPS VCC caps replaced.