Enermax ATX PSU repair

[ceoxrad]

Hello everyone. I'm a long time reader but I posted very little on here.

First of I apologize in advance for my English, I'm not a native speaker and I hope that my post is at least readable.

I have a fairly old but still capable Enermax branded EGA1000EWL atx power supply. This came from my main workstation and was replaced some years ago with a more modern one after developing a fault. That's a short summary of what happened and what I did back then:

The PC was hooked to refused to start, and the indicator led on the back of the unit was red. At this point I replaced the PSU with another one and the PC was fine.

With the PSU out I checked for obvious signs of damage and upon inspection I found some bulging capacitors (as expected given the age...). I replaced every filter capacitor in the unit with brand new Rubycon low esr parts and tried to power the thing back on.

The PSU would start fine (5VSB is there and healthy, back led orange-standby) but when I tried to fully turn it on it refused to start. I can hear a relay click and for a short moment I can see the 5V and the 12V rails turning on but they immediately shut off and the back led turns red again. This is a smart PSU so it has a buzzer for error codes, according to the manual there is something wrong with the protection circutry.

The protection circuit in this PSU is on a daughterboard, on it I can see no obvious signs of damage. The main control IC is a PS232S from SITI, on board there is also a simple opamp.

Some time ago I dug out this PSU from storage and it behaves exactly as I remebered, I would really like to repair it, if not just to learn something new.

I have a fairly good ammount of test equipment and I'm familiar with working with high voltage and electronics (Soon to graduate in physics) but I don't have any expertise on actual power supply design so I don't know how to approach the problem.

I attach here the link for the datasheet of the IC and a review/teardown of the PSU in question.
http://www.siti.com.tw/product/spec/Power/SP-PS232S-A.003.pdf
https://www.hardwaresecrets.com/enermax-galaxy-1000-w-power-supply/
https://www.jonnyguru.com/blog/2006/08/15/enermax-galaxy-ega1000ewl-1000w-power-supply/

I hope you can provide some feedback on how to solve this fault

Thank you in advance

[TheMG]

Couple of things: did you have a load connected to the PSU (3.3V, 5V, and 12V rails) when trying to turn it on? Some of these older ATX PSU designs won't turn on or will trip off on protection when there isn't some minimum amount of load on it or if only one rail is loaded.

Also, you mentioned filter capacitors specifically. Did you check ALL other electrolytic capacitors in the power supply for ESR and capacitance? I've seen cases when small electrolytic caps in the feedback and driver circuits have gone bad and caused PSU to be unstable.

[ceoxrad]

Thank you very much for the kind answer!
Today i did some further testing: it seems this PSU does not turn on if it does't detects the internal fans. before trying to test with a load attached I reconnected those and the PSU starts fine even without load.

I measured the rails with a multimeter and they're all whitin reason. For some reason the beep code I was recieving before was not a fan warning but a protection warning (at least according to the manual).

Anyway after this succesful attempt I tried to load the thing with some halogen 12V bulbs I had laying around. Everything fine till I tried to load the 5V rail. This immediately causes the unit to shut off. It doesn't matter on which order and which power I try to load the rails, everything works except the 5V one (under no load it measures fine BTW).

I don't have an in-circuit ESR meter but I have one of those 20$ LCR ESR testers, I don't know how reliable it is to check ESR but if that's the only option left I can desolder every electrolytic capacitor and check with that thing.
At this point can it be the 5V ocp?

[magic]

Perhaps a cross-regulation issue? 12V/3.3V rail going up when you put load on 5V while the others are completely unloaded?

Are all rails in regulation when you load 5V with something lighter that doesn't trip the protection yet?

The fan problem may also have been a matter of simply adding sufficient 12V load to enable regulation.

[ceoxrad]

Ok further testing done. I tried to do the measurements you suggested.
All the other rails are rock solid in every condition.

 It seems the maximum current I could pull from the 5V rail is approx 500mA, It's not really repeatable but the average trip current is somehere over this value, less than 1A.

 Under this maximum load condition I can hear a little coil whine from the power supply (it disappears as soon I disconnect the load) and the rail sags to approx 4.8V.

[magic]

What if you put some load on 12V at the same time?

[ceoxrad]

Tried now, with a little load (~15W each) on 2 of the 5 12V rails the 5V output returned in spec and I was able to pull almost 2.5A.

[magic]

It probably has nothing to do with your repair and just a quirk of the PSU. They can do weird things under unusually light loads.

A few years back there was that whole issue with "Haswell ready" PSUs: Intel dropped the idle power consumption of their CPUs so low that some PSUs were going out of regulation on the 12V rail and shutting down when the CPU entered power saving state.

BTW, the multiple 12V rails oftentimes aren't really independent in any serious sense and are shorted to one power supply internally.

[Emo]

Hi Ceoxrad,

You are not accidently trying to load the separate 5V standby rail?? this rail can usually handle a maximum 500 mA.

Eric

[ceoxrad]

Hi Ceoxrad,

You are not accidently trying to load the separate 5V standby rail?? this rail can usually handle a maximum 500 mA.

Eric

Nope, I'm loading the standard rail, not the sb one

[TheMG]

That's just the nature of the beast. One power supply feedback loop producing multiple output voltages. If they are not relatively evenly loaded, voltage may go out of spec on the lightly loaded rails.

That is why a majority of the more modern ATX PSUs have a single 12V output, and have independent buck regulators at the output providing the 5V and 3.3V outputs, that way there is no issue with voltage regulation even if only one rail is heavily loaded.

Also, I find halogen lamps make a poor power supply test load especially for something as finicky as a PC power supply, since the protection circuit in typical ATX power supplies has very tight tolerances and will easily trip off. The reason is that when the filament of a halogen lamp is cold, its resistance is considerably lower than its operating resistance, so there is a very high initial inrush current when turning on the lamp. This inrush current is often around 10x the lamp's rating, so a 50W 12V lamp for example (4.2A operating current) will draw about 46A for a brief moment at turn-on!!!

[TheMG]

BTW, the multiple 12V rails oftentimes aren't really independent in any serious sense and are shorted to one power supply internally.

That is true, however each rail often has its own current sensing shunt and over-current protection that will shut down the power supply if any rail exceeds its rated current, even if the total load on the power supply is well under the rated maximum.

[magic]

Aren't independent as far as voltage regulation is concerned, anyway

AFAIK the whole point of that is to limit the short circuit current available into any single output cable for some safety regulations.

[ceoxrad]

I think that this PSU has separated 12V rails, at least partially, probably I loaded the 12V rail on the same section of the 5V one. According to this there are 2 transformers https://www.hardwaresecrets.com/enermax-galaxy-1000-w-power-supply/5/

Also, just to test it out I tried to power on an old P4-era PC. The PC would not start and there's significant coil whine when in standby. There's definetly something fishy going on. The 5Vsb sags to approx 3V in these conditions (on another PSU the PC runs perfect)
 

[TheMG]

If you haven't yet checked the remaining electrolytic capacitors, I would do so. Even if you have to remove them from circuit. Capacitors in the switching control circuits (primary side) would be the first ones I'd check.

The way ATX power supplies are designed, the +5VSB output is crucial to the operation of the main power supply. If the +5VSB is unstable or not the proper voltage, then the entire power supply can behave erratically.

Also check that the PFC pre-converter is operating properly: voltage on the primary filter caps should be around 400V. If it's not functioning, the voltage on the caps will be roughly equal to the peak of the mains voltage (170V for 120VAC, 340V for 240VAC) minus two diode drops. Weird things can happen if the main converter isn't getting the right DC primary voltage and certainly the power supply wouldn't be able to deliver anywhere near proper output current in such a situation as the duty cycle would be way high to compensate.

[Miti]

A halogen bulb may not be an appropriate load for a PC power supply. It may take a lot of current when cold and trip the over current protection.

[magic]

Make sure you haven't installed 5Vsb replacement capacitor (if any) backwards

[Dacke]

Whining/buzzing when loaded down with an accompanying voltage drop out of spec on 5VSB sounds like some sharp pulses driving the transformer,  which sounds like the circuit may not be in closed loop.   The first thing I would check is the feedback circuit (and more specifically,  the shunt regulator and the components around it,  possibly a TL431) on 5VSB,  as this tends to be a separate, smaller power supply in itself.   This is just my guess,  as the all the other rails seem to be behaving normally - on an older ATX power supply they can be erratic until they are loaded.