computer SMPS repair -- Help needed

[rodd]

Dear all,
I am trying to repair a Corsair AX850 gold computer power supply.
Since it is a medium to high quality product and the problem is intermittent, I decided to try to fix it.
The supply works fine for several days, until there is a power glitch and the supply shuts down.
After that, I have to unplug the ac cord and wait 5 to 10 min before it can start again.
I have checked the voltages with a PC check gadget and all the voltages are within spec.
I think that this "dead time" suggests something related to capacitor discharge time.
There are a lot of dangerous voltages around so I would like to narrow my guesses to a minimum.
Any ideas would be greatly appreciated.
Regards,
Roger

[PKTKS]

Not trivial stuff. 

You need several things and assume too much without the schematics.

First you definitely need an ELTRONIC LOAD  capable to support
the proper wattage you are requesting  in multiple rails

Second you NEED A TRANSFORMER ISOLATION bench gig to start

Then you have be aware of the CHIPSET used as controller

From that point you can LOAD THE PSU and put some
minimal readings to see if the OVERLOAD PROTECTION IS ACTING..

Or you have a really bad thermal device bouncing on and off..

That would require some data logging for the period you mention.

USB DMM with sigrok is my top choice.

LOAD the PSU and let the sigrok data logger collect data...

I know.. this is not trivial GIG to put ...
other methods are just a bunch of guessing wo information

Paul

[rodd]

Hi,
I am sure I have understood your reasoning.
The power supply is not tripping by itself.
What is happening is If, for some reason (usually a small glitch in the 127AC supply) the supply shuts down, I have to unplug the 127AC power plug and left it unplugged for 10-15 min.
After that period of time, I can reconnect the AC power and turn the SMPS on normally.
I can not see I should monitor the outputs since the SMPS is NOT tripping under load.
By the way, I found this information after some more internet search    "an ICE2QR4765 Quasi-Resonant PWM Controller handles the standby mode of the PSU. "
Thanks,
Roger

[PKTKS]

This thing is quite confusing due to the fact that ATX PSUs changed  a lot

Fomr early ATX to ATX12V (aka ATX-P4) to modern (new standards) ATX12V[1234...]

What is important to consider is that on the bench YOU CAN TURN THE POWER ON.
in simple terms you can turn the power ON even without load...

**BUT** the thing is .. EACH specification of ATX states that the supervisor is
**REQUIRED ** to sense a MINIMAL and a MAXIMUM load to trip the OFF condition

As the ATX evolved .. so did the required MINIMUM load.

Check your ATX version: ATX 2.03 or 2.x - confirm the MINIMUM load to sustain ON

USE a proper dummy load to that ampearage on EACH RAIL 3.3, 5 and 12V (aka P24)

Then start your ATX and wait if it can sustain **MINIMUM** LOAD ..

That said -- in case POSITIVE - you have no ESR faulty readily seen ...

The HARD PART..

MODERN ATX required  P4 P6  and P8 connectors of ** VERY HIGH AMPERAGE **

You now need to see if it can TURN ON POWERING HIGH AMPERAGE connectors...

Best case scenario you FAIL THE  TURN ON procedure... MINIMUM LOAD..
Great because you probably have
- FAULTY CAPACITORS
- FAULTY MOSFETS

Best lucky shot...  *IF* your MOVs, termistors and PTCs/NTCs.. are still good to go..
Glitches may render them faulty as well..

Some folks start changing all these items in a lucky chance...

whatever.
Paul

[PKTKS]

Another  confusing detail ... is that ATX 2.31 specification

changed the time period to the POWER_ON to POWER_GOOD signals

That means that ... some MOBOS or chips may have trouble with incorrect
time delay issued from the POWER_GOOD line

You should also check (measure) the specific (SINGLE_SHOT) pulse of POWER_GOOD
on your PSU and check the ATX specification  version...

Damn confusing things..  and amperages surges of 30A/50A/100A
 are not trivial stuff

Note that recent CPUs Vcore are mostly 1.x volts ranging above 100Watt ...
That implies a 100A typical  surge...

Meaning that .. if the POWER_GOOD  starts too soon...
The surge of 100A or more of a modern CPU will TRIP THE OVERLOAD protection ...
while the PSU is not yet ready to put that power out...

Paul

[soldar]

Could it be the mobo and not the PSU causing the problem? Are you 100% sure it is the PSU?

[rodd]

Hi folks,
Thank you for your replies.
Some additional information.
1- My PC was assembled 4 to 5 years ago (everything new at the time).
The system performed without any flaw since July 2018, where the symptom started to develop (the need to unplug ac main and wait 5-10min to achieve a successful power on).
2 - I bought a new PS and the system is working as expected (instant power on after a power glitch), so I think the problem lies in the SMPS not the Mobo.
3 - The power supply test gadget seems to fragile to load the PS with high currents (also it does not heat up during the test), but it shows that all the voltages are on spec and the Power Good signal delay (I guess ) is 300 ms.
regards,
Roger

[PKTKS]

Some years ago..  may be old stock.

What matters is the ATX specification - good brands put that on the stick
outside the case.

**BUT** cheap/budget  models may ditch all the specification requirements
and put things "as is".  CHEAPO is CHEAPO.

MOVs and PTCs/NTCs   are SAFEGUARD only
you can remove them to check if they are bad...
Mains glitches may toast MOVs and some input components

Last thing to note - IF THE MODEL CONTAINS A PFC chip and components.
in this case you will have a long list of things to check...
Not present in budget models... fortunately

There is no simple way - besides that lucky chance guy saying: change CAPS...
You need to check a LOADED secondary side

Paul

[station240]

Attach a multimeter to the 5VSB line and a 0V line, ATX is a standard so
5V standby is purple.
0V is any black.

Now plug it into the mains, see how that rail behaves.
If you can narrow it down to the standby supply, it's a lot less parts/capacitors involved.
Without the standby supply, the main PSU will not start.

[rodd]

Hi,
thank you for your reply.
The PC power supply tester show this voltage and the value  is OK.

[rodd]

Here is a photo of the gadget testing the PS.

[wraper]

It sounds like PSU goes into protection for some reason.

[PKTKS]

It sounds like PSU goes into protection for some reason.

I think too.
By the narrative I would guess a thermal faulty component..

That little tester DOES NOT PUT neither the MINIMUM nor the MAXIMUM load..

It just puts POWER_GOOD and starts POWER_ON..  Benefit of this particular
model is reading the the time delay.. others cheaper won't do that...

You can remove MOVs and protective input devices to test the PSU...
And you should put a MINIMUM LOAD for that..

Paul

[Mechatrommer]

What matters is the ATX specification - good brands put that on the stick
outside the case.

**BUT** cheap/budget  models may ditch all the specification requirements
and put things "as is".  CHEAPO is CHEAPO.

now this is where it went upside down for my case. during the life of my more than 10yrs old Core2Quad PC, i bought 2 gigabyte branded PSU after the original no name brand PSU went dead after years of operation, both worked only for less than a year, and then sudden death no warning whatsoever. i said what the heck i bought less than half the price PSU (unknown TronMonxter brand), this is the 2nd TronMonxter working more than 2 years now. recently it went dead frequently but alive again after few seconds of shutdown, it turned out one of capacitor on the output, i guess 12V rail because its rated 16V, got damaged or shorted. replacing that everything back to normal. in the mean time i've collected 2 cheapo PSU already as backup throughout the time in case this PSU KO beyond repair, but it hasnt dead yet.

[Jfcarrer]

this unit has a 10 year warranty, i'd contact the manufacturer before attempting any kind of repair that could void it.

[james_s]

Most of the time when I see problems like this it is bad electrolytic capacitors. If it's under warranty though it's worth looking into that first, unless it costs more to ship it in for exchange than it's worth.