Need help fixing EVGA 500W PSU 3.3v rail overvoltage

[Kyuunex]

Too many things happened but the short version is that a brand new psu died on me after 33 days of use. I did not abuse it or anything it was running on my pc normally. I can not rma it because shipping costs and etc. instead I wanna fix it my self.

I did a lot of testing and determined that, 3.6 volts are outputted on a 3.3 volt rail, then the psu over voltage protection kicks in and shuts it off. (trust me, i spent a lot of time figuring this out)

After looking on a disassembly page of the review of this psu on jonnyguru, I think I could possibly fix it by replacing STPS30L45CT Schottky Rectifier that we have for the 3.3v rail or i could try replacing MDP13N50 n channel MOSFET. Or maybe the problem could be the PWM signal. I have no clue. (capacitors are fine btw)

What I wanna know and want help with is figuring out what component could be causing the power supply to spit out 3.6 volts instead of 3.3 volts, on a psu that came out defective out of the factory. I wanna a general idea what could be the problem. Like how do we even get 3.3 volts? Then I'll just grab that component from ebay and replace it by soldering.

[Toasty]

They have a 3 year warranty. Why not take advantage of it...

T

[Kyuunex]

They have a 3 year warranty. Why not take advantage of it...

T

Too late, I already opened it up and made it obvious I opened it up.

EDIT: everyone keeps telling me to rma it. i can't. on evga website it says: "The customer is responsible for the costs of shipping in order to return the product to EVGA for the RMA process. ". it's a psu, it's gonna cost a lot to ship. it's not worth.

[Kyuunex]

Here's some more info.

Here's how I came to this conclusion. The PSU starts, and shuts off after a second. if I use the green wire method to start it, it shuts off. Many different motherboard I tried, it still shuts off. However I managed to find 2 things that start it and make it not shut off.
1. ASUS P4B533-X motherboard.
2. Spamming the green wire. When I connect my multimeter to the 12V rail on my psu and setting it to measure voltage, It obviously shows the voltage coming on that rail. Which is 0 V when I haven't done anything on it yet. When I short the green wire, the voltage spikes to 9v and slowly starts coming down. However, if I unshort and short again fast enough I can make it go up to 10V for a second. If I keep spamming it like that I can get it up to 12V, in that case, the PSU starts.

I have no clue what's going on and why it's like that. I thought I tricked the OVP on the micro-controller or something.

As far as ASUS P4B533-X motherboard goes, it works fine but aida 64 was showing 3.6 volts, so then I took my multimeter and measured 3.3v rail and confirmed it was true. (Even shuted it off and use the green wire spam method to confirm it.) That's when my brain stopped working and decided on that this was the issue.

I'm really confused how a 15 year old ASUS P4B533-X was able to start it though.

Some more info:
Also one guy recommend me to cut the brown wire that is connected on 3.3 v rail on atx 24 pin and it did not do anything.
All traces look fine on the PCB.
All capacitors look fine on the PCB.
Contacting evga with many details and they insisted that I should RMA it.
This PSU also had this weird habbit of buzzing on the filter sometimes. (The one that's on that small board, next to that switch)

[amyk]

I'm really confused how a 15 year old ASUS P4B533-X was able to start it though.

That's a P4 board, and P4s are known for being power-hungry; it could be drawing enough current to keep the voltages (just barely) low enough to stop OVP from shutting it down.

I don't think it's the rectifier nor the FET (you wouldn't get any voltage out if those were damaged.) Instead, look at the components on the feedback path.

If you can link to the exact model that would help, since it seems EVGA has made several PSUs over the years.

[Rasz]

you didnt specify which "EVGA 500W" supply you are talking about (there are at least 2 models?) nor uploaded any pictures.

buy new power supply
+isnt there some kind of consumer protection in Georgia? in normal countries we can choose between warranty and shop liability, no need for expensive shipping.
evga isnt even good, they lie about this product ('Japanese' capxon/luxon caps )


as to the fault(if there is any), it works on obsolete motherboard because P4 still used 5V as the main power rail, you can test this hypothesis loading 5V rail with incandescent 12V 60-100W bulb (old car light etc), on the other hand this is very old design regulating all rails at the same time, meaning you shouldnt see 3.6V but not 5.5V 13.2V = you are doing something wrong/out of your element, buy a new power supply

[Kyuunex]

EVGA 500W 100-W1-0500

that's the model.

I bought this from amazon.com, and use a cargo forwarding company to to ship it to my country. Paid a lot in shipping since this thing is 2 kilograms. I didn't buy this locally. Nobody sells any psu that's good locally.

Instead, look at the components on the feedback path.

hm... thanks.

you can test this hypothesis loading 5V rail with incandescent 12V 60-100W bulb

I'll definitely test this later.


I'll post pictures tomorrow.

[Kyuunex]

you can test this hypothesis loading 5V rail with incandescent 12V 60-100W bulb

I don't have a 12V 60-100W bulb but I tried 12v 24W bulb and it did not start the PSU.

One thing I noticed is that the motherboard it self can't start the PSU. However, if I do hock up the 3.5 inch IDE hard drive to this PSU, it can start the PSU.

Hm...

Here's are the pictures BTW:

[Mattylad]

I take it you have given it a really good going over for bad joints?

[Kyuunex]

I take it you have given it a really good going over for bad joints?

"really good going" not really, if you mean going through every single one of them with a magnifying glass. However it looks fine.

Can't really believe a bad joint can happen after only 33 days of use or from factory. it wouldn't just melt, right? or it might considering how terrible quality psu this is.

[Kyuunex]

Here's a high resolution picture of the PCB, where resistor values are clearly visible, taken using a scanner with 2400 dpi and slightly downscaled because imgur upload limit.

no idea what those purple markings are for though.

[Kyuunex]

Update: I figured something out.

I happened to be cleaning my dad's old pc that has G41M-VS motherboard with E5300 overclocked to 3.12 GHz. While I was at it, I decided to test out this psu there. And it started on it. I checked voltages and 3.3v rail as showing 3.61 volts constantly, not dipping, as opposed to 3.58 i usually get on that p4 motherboard that sometimes peeks to 3.6.

So, i got thinking and decided to attach large loads onto different rails. as suggested, I tried attaching loads to 5v rail, but it didn't help. But attaching to 3.3v rail actually makes it more likely for psu to start.

I'm saying "more likely" because it starts on 2 tries of shorting the green wire. (second time shorting (after 1 second cooldown from first shorting) is guaranteed to start the psu.)

Loads I attached were, 12v 80mm fan (to 3.3v rail) and a laptop charger that is designed to go in cigarette jack in cars. (attached to 3.3v rail starts psu. attaching to 5v or 12v doesn't)

(Note, on other forum, they told me that overvoltage protection kicks in at 3.8v actually)

My theory is that when PSU start the first time, the voltage is 3.8 or higher and triggers the OVP, but if a large enough load is attached to 3.3v rail, it brings the voltage down to 3.5-3.6. This specially makes more sense considering the fact that old motherboards were utilizing 3.3v rail to some point while modern motherboards barely use it at all. (afaik)

So we're back where we started but with a better idea of what could be the problem.

Now, here's the important thing I discovered

if I manage to jump start the PSU by adding load to 3.3v rail, and then removing the load and then measuring the load, the multimiter shows "3.67" volts.

if I cut the brown (aka 3.3v sense) wire it shows "3.67" volts.

if I then attach that brown wire to 5v rail, it shows "3.48" volts.
if I attach the brown wire to 12v rail, it drops it down to "3.18" volts.

When the brown wire is attached to the 5v rail, the PSU starts without shutting down normally. No need to add load. it just simply works. Though I'm not sure how comfortable I am with this solution.

[EPTech]

Hi there,

Near the opto-couplers, it looks to me like there is a TL431 (U550?). It should be a T-92, transistor-like housing. Looks like the middle pin is not soldered. That is the voltage reference for the feedback path, so a pretty important component.

Happy hunting.

[Kyuunex]

Hi there,

Near the opto-couplers, it looks to me like there is a TL431 (U550?). It should be a T-92, transistor-like housing. Looks like the middle pin is not soldered. That is the voltage reference for the feedback path, so a pretty important component.

Happy hunting.

Any chance you can point out on a picture with a red dot or an arrow what part you mean?

[amyk]

That does sound like a problem with the sense/reference circuitry. According to an article I've found this is a "group regulated" supply where the multiple voltages are summed together to create the final error signal, but I haven't inspected the pictures closely to determine if it's summing all 3 (12V, 5V, 3V3) -- you'll see a separate winding on the main transformer for each of the voltages in that case --- or if it's the other common scheme of regulating 5V and 12V together and using a magamp (saturable reactor) off the 5V to give 3V3. The former case will have only one reference, the latter will have two. If the 12V and 5V are OK and do not vary if you change the 3V3 sense voltage, then you have the latter design; if they also increase/decrease in proportion, then it's the former.

Example of ATX PSU schematics:

http://danyk.cz/s_atx_en.html

http://danyk.cz/s_atx01b.png - sense on 5+12, "floating" 3V3
http://danyk.cz/s_atx01i.png - sense on 3V3+5+12
http://danyk.cz/s_atx02d.png - sense on 5+12, magamp for 3V3

[Kyuunex]

That does sound like a problem with the sense/reference circuitry. According to an article I've found this is a "group regulated" supply where the multiple voltages are summed together to create the final error signal, but I haven't inspected the pictures closely to determine if it's summing all 3 (12V, 5V, 3V3) -- you'll see a separate winding on the main transformer for each of the voltages in that case --- or if it's the other common scheme of regulating 5V and 12V together and using a magamp (saturable reactor) off the 5V to give 3V3. The former case will have only one reference, the latter will have two. If the 12V and 5V are OK and do not vary if you change the 3V3 sense voltage, then you have the latter design; if they also increase/decrease in proportion, then it's the former.

Example of ATX PSU schematics:

http://danyk.cz/s_atx_en.html

http://danyk.cz/s_atx01b.png - sense on 5+12, "floating" 3V3
http://danyk.cz/s_atx01i.png - sense on 3V3+5+12
http://danyk.cz/s_atx02d.png - sense on 5+12, magamp for 3V3

Whether the brown wire is attached to 3.3v or not doesn't make any difference.

If the brown wire isn't attached to anything, here are voltage readouts:
12.20
5.15
3.67

If brown wire is attached to 5v rail here are the voltage readouts:
12.23
5.12
3.48

If brown wire is attached to 12v rail here are the voltage readouts:
12.19
5.17
3.18

while other's change very slightly, 3.3v rail changes more noticeably.

[Kyuunex]

it's time for an update:

I have been using this psu for 2 weeks now on my main pc which has G4560 and GTX 970. I had no problems. It works fine.

only one odd problem i encountered was aida64 showing wrong voltages on my new AsRock H110M-ITX/ac motherboard. (it was fine on my other gigabyte motherboard) I went into bios and voltages were correct. it's just AIDA64 showing wrong ones.



out of this, it was only showing 3.3v rail voltage correct and I actually checked and it was that and it turns out I did a bad job at connecting the brown wire to 5v rail, other than that, everything is fine. (not sure how it booted with high 3.3 v rail voltage though I imagine it lost the contact while it was already booted, or maybe something was pulling enough power)

so yeah, the psu is functional. I kinda feel sad it turned out this way. But I decided to just gift this psu to a friend who doesn't even have a computer and is gonna build one soon.

Thank you everyone who contributed in any way for me to get this hunk of junk to work. next time i buy a psu, I'll just overpay for a tier 1 or tier 2 and get a quality unit.