Weird behavior on 3.3V rail of a CORSAIR CX600

[bimole]

Hi,

A colleague gave me an apparently ill CORSAIR CX600.
He told me that he heard a banging noise, and smelt the magic smoke.
After tearing down the PSU, no burnt component, everything seems OK, but the PSU often refuses to stay ON. When shorting PS_ON to GND, the fan spins a bit a finally stops. No output voltages.

I tried to treat some apparently bad solder joins and finally it seems to work OK... on the 12V and 5V rails!
It's another story on the 3.3V rail...

I have some devices to test deeply PSUs (scope, AC source, electronic load) so I can diagnose a bit.

The schematic is very close to this one https://sector.biz.ua/docs/power_sup...T-500A-12S.jpg

This is a basic structure : 2 switchs forward converter, one regulation point on 12V and the 3.3V derived from the 5V with a mag amp.

Here is the problem :
I tried to power up an old motherboard (without CPU nor RAM nor GPU) and 1sec after powering it up, it shuts down. Some measurements showed that the 3.3V is falling when loaded, the power good signal trips and it shuts down the PSU.
I have tested almost every single components of the mag amp (resistors, capacitors, the PNP transistor, the TL431 and even the saturable reactor with an impedance meter). Everything looks fine.
But if I load more than 200mA, the 3.3V rail falls, sign of bad regulation. There is no load on other voltage outputs.
In these conditions, the duty cycle of the forward converter is very low and erratic.

I noticed that if I load the 12V rail with a 22R power resistor, the switching signal is now very clean and the duty cycle of the converter is greater than without load.
In these conditions, I'm now able to load the 3.3V with several amps without regulation problem. It's a solid 3.3V
I think when there is no load on 12V rail, the duty cycle is not great enough to give the sufficient room to the mag amp to regulate efficiently, so the voltage falls quickly when the current rises on the 3.3V rail.

My question is : is this normal or is there still something wrong with this power supply ?

Many thanks for your comments in advance!

JB

[stj]

if it went bang and smoked then you missed something.
look for burned or cracked parts again.

[momaka]

Hi,
I have tested almost every single components of the mag amp (resistors, capacitors, the PNP transistor, the TL431 and even the saturable reactor with an impedance meter). Everything looks fine.
But if I load more than 200mA, the 3.3V rail falls, sign of bad regulation. There is no load on other voltage outputs.
In these conditions, the duty cycle of the forward converter is very low and erratic.

I noticed that if I load the 12V rail with a 22R power resistor, the switching signal is now very clean and the duty cycle of the converter is greater than without load.
In these conditions, I'm now able to load the 3.3V with several amps without regulation problem. It's a solid 3.3V
I think when there is no load on 12V rail, the duty cycle is not great enough to give the sufficient room to the mag amp to regulate efficiently, so the voltage falls quickly when the current rises on the 3.3V rail.

My question is : is this normal or is there still something wrong with this power supply ?

Yes, this can be pretty normal behavior for some PSUs, especially forward-converter topologies (resonant LLC, maybe not so much.)
By the looks of it, your PSU is working normally, at least from what I can tell from these tests.
Also, I don't suggest to anyone to use a motherboard without CPU and RAM as a "load" to test PSUs. Part of the problem with doing this is that when there's no CPU and/or RAM, the motherboard disabled most of the chips and power rails, so many will typically draw very little current... especially from the 12V rail, which is what's actually needed the most when it comes to testing more modern / newer PSUs that expect a bigger chunk of their load from the 12V rail.

If nothing looks burned on the PSU and it's powering on normally... then it's probably OK.
If so, maybe buy your colleague a beer for giving you a free working PSU.

if it went bang and smoked then you missed something.

Agreed.
More so, I don't see how something could have burned and banged, then have the PSU work fine again without issues. I suspect something else went wrong in your colleague's PC, like a video card going short-circuit or similar, causing the bang and smoke... and the PSU simply shut down. As to how he concluded it's the PSU and not something else, that you may have to ask him/her. I personally always try to verify the information being given when I receive something broken from someone for repair.

[bimole]

Thank you so much for your answers.
I didn't fiddled more on this PSU since last week.

I was wondering, maybe this could be a bad snubber/damper on the secondary winding...
With no load at all, the duty cycle is at his minimum value. If the snubber resistor/capacitor is broken, some overshoots on pulses could appear and "mislead" the regulation, reducing again the duty cycle to maintain 12V at the output.
Maybe I will check that...

Otherwise, nothing wrong with an IR camera, no hot spots.

JB

[momaka]

Well, a lot of PSU manuals and labels don't note this anymore (and I'd be surprised if someone even keeps these things for reference), but most ATX PSUs do require a minimum load to work properly. For the old-days ATX PSUs that were 5V-heavy, typically 0.5 to 1 Amp was the minimum required from the 5V rail, and often times between 0.1-0.5 Amps from the 12V rail - as stated by the manufacturer in order for the PSU to output voltages in spec. When 12V-heavy(er) PCs became the norm, this minimum load was more so shifted towards the 12V rail.

In any case, running the PSU without some load and performing measurement tests will likely lead you onto a ghost problem chase.
So put a load and then perform these tests.

Though I doubt any secondary-side snubbers would be bad here, as these typically comprise of low-value resistor(s) in series with a ceramic or film cap. There's not enough high voltage to make these go bad so quickly as to make a popping noise... and a lot of times when they go, either the cap turns dark (overheat) or the resistors does... or in rare cases, both. But you would have spotted that with your visual inspection.

Anyways, keep up posted what you find, as I'd be interested to see if there was actually any problem or not. To me, it makes no sense that something can go wrong so dramatically... and then the PSU can continue to work OK afterwards. But I could be wrong, so let's see just in case.

[bimole]

Hi!

I finally tested the PSU on the same motherboard as before BUT now with CPU, GPU and RAM. Works like a charm !
Seemingly, all ATX PSU doesn't behave the same way depending on the output conditions.
I compared with another PSU, a Seasonic FOCUS+ Gold which is LLC based and has several buck post converters after the main one. On this one, the naked motherboard (without CPU, GPU, RAM) could be powered without any problem.

Anyway, I consider this PSU as OK and it will be used for testing various old motherboards. Output caps are still in good shape and voltage is very clean. So, that's it!

Thank you again for your usefull pieces of advice 😉

Cheers,
JB

[momaka]

I finally tested the PSU on the same motherboard as before BUT now with CPU, GPU and RAM. Works like a charm !
Seemingly, all ATX PSU doesn't behave the same way depending on the output conditions.
I compared with another PSU, a Seasonic FOCUS+ Gold which is LLC based and has several buck post converters after the main one. On this one, the naked motherboard (without CPU, GPU, RAM) could be powered without any problem.

Yes, different topologies and designs have different behavior when it comes to how they run at no or low load conditions.
Generally, most LLC topologies tend to be a lot more stable with low loads, since they don't have an output inductor. What they do instead is, the primary side utilizes pulse-skipping / "hick up mode" when underloaded, which allows the PSU to stay operating even at nearly zero load. Meanwhile, the more "classic" PWM continuous topology designs have an output inductor (toroid), and these cannot run without any load whatsoever, since the output inductor always needs to have some minimum current flowing through it for the regulation to work. This minimum current will also vary by the design particulars (i.e. size of the output inductor, size of the main transformer, minimum allowable pulse width, size of output capacitors, and etc.)

Anyways, glad to hear you got yourself a working PSU.

[TheElectronicus]

From my electronics knowledge i can suggest testing the following things:

If the PSU has TVS suppressor diodes somewhere on the output, they can become conductive because of an overvoltage event, and also emitt the classic burning smell, but don't go bang. So you can not visually see if they are damaged. The often somehow survive such events, tough significantly degrading their standoff voltage, so they will conduct even at or far below the rated voltage.
So i would suggest testing all the diodes on the board with the diode test-function of your multimeter. A caviat on the output Diodes of the switching transformer. They could be parallel to the windings, so you will measure them as a short, even when they are good. If you want to be sure you need to desolder them and test them outside the circuit.

Most switch mode PSUs have a small auxilary power supply, for the standby supply, and suppliying the voltages for the operation of the main power supply itelf.
To kickstart this power-supply, there is often a small resistor in serial configuration feeding from the main bulk main capacitors (carful, they have 400VDC). Also in this circuit there is a zener diode or something like that, which makes a crude 12V rail, to kickstart the auxilary PSUs PWM IC. As soon as this IC is switching, it will get it's voltage from an auxilary winding of the auxilary transformer, which is ofthen diode-or'ed with this zener startup supply. In this area check the series resistor, the zener-diode and the or-ing diode. Most of the time the series resistor gets overloaded and then has a much higher resistance, which can be just sufficient to start the aux supply sometimes, but sometimes it is just not enough. Also the Zener could be degraded.

Otherwise check the board when running with a thermal imaging camera, if you have one available. You should see a degraded TVS or zener diode conducting at far lower voltages light up on the board.

The usual other canditates are testing the electrolytic capacitors and if you have a microscope also check the ceramics for obvious sings of cracks.

But such repairs with non clear faults, are a huge rabbit hole. It's upon you, how deep you want to go down ;-)

Maybe that helps.

Kind regards,
Gregor