EVGA NEX650G power supply detecting a fault on startup.

[nikload]

Re: EVGA NEX650G power supply detecting a fault on startup.

good evening
A similar problem
replaced R76 10 om
Worked for 5 days and burned out

[momaka]

Re: EVGA NEX650G power supply detecting a fault on startup.

good evening
A similar problem
replaced R76 10 om
Worked for 5 days and burned out

Hmmm.... that suggests nothing is majorly wrong in there if the PSU can work "OK" for a while. But since the same resistor has burned again, perhaps some component somewhere has drifted in value and is causing an overload on that resistor. Or maybe the design was marginal to begin with and so that resistor just burns out after a while when components start to drift in values from age.

OR OR

Maybe there's an intermittent ceramic cap somewhere shorting out... though I don't see how that could happen with the same cap in two different PSUs.

Another possibility is the output CapXon caps going bad (high capacitance perhaps?) It would be interesting to see someone actually check the ESR and capacitance of the output caps to see what exactly is going on.

And if not, perhaps the easiest solution would be to use a big (1 Watt?), through-hole 10-Ohm resistor for R76 and see what happens then. Will the PSU still burn it or will the PSU burn itself?

Given how majorly over-cooked the trace looks in the picture in post #5, I honestly wouldn't suggest trying a resistor rated for bigger power. IMO, that's the same as using a fuse that's larger than what the application calls for... and typically the results can be lots of smoke and sparks when something fails.

[kuntao]

Re: EVGA NEX650G power supply detecting a fault on startup.

what is the code of Q29

[kuntao]

Re: EVGA NEX650G power supply detecting a fault on startup.

that is a 10 ohm resistor

[socketa]

I just destroyed one of these that had the same symptoms

Re: EVGA NEX650G power supply detecting a fault on startup.
Finally, Q29... it also appears to be either a free-wheeling or forward rectifying device for the 5V rail (thus equivalent in function to either Q3 or Q4 in Figure 1 on the FPS article linked above.) Obviously if Q26 is one, then Q29 will be the other and vice versa. But again, I see a minor error here in the table - pin1 (Gate?) of Q29 shows as reading "1.92" to the 5V rail. I'm pretty darn sure this measurement is not in Ohms. If it really is, that may be a problem. But I really think you're using the continuity/diode test mode here and not resistance as I asked... so I can't actually tell if the reading is problematic or not. Also, when stating the resistances, please state the units too. I don't mean to make this into a high school science class all over again . But clearly "1.92" doesn't really mean much without units. If it's Ohms, that's no good for sure. If it's KiloOhms or MegaOhms, then it's a completely different story.

After a lot of effort getting the large toroid off, i saw that the gate of 031N03L (Q29) goes to one of those 2R0 resistors, and then to that through-hole dot, then through the board and onto the top side of the board to a trace that goes to through-hole dot that is between the two mosfets (G26, Q27), which are connected to both big square pads, which are connected to the 5V rail
So it looks like the gate of Q29 is meant to be connected to the 5V rail via a 2R0 resistor
Gate Q29 ----- 2R0 ----- 5V rail
Seems a bit odd, since in that FSP schematic all of the gates are connected to a controller chip
Underneath that 14-pin 6601 chip, it looks like some components got hot - the smd caps either side of the resistor aren't shorted and the resistor doesn't give a stable resistance
Didn't notice that till after though

[socketa]

So it looks like it's probably the same fault
I've put it back together and will order replacement mosfets, and then try using a 10 ohm resistor (post #24)
The output caps tested good ESR and resistance (couldn't read two of them because i pulled out the leads when removing them)

[socketa]

Working good now
Except the 10K resistor heats up to more than 100 °C
Couple of youtube vids of folk replacing this resistor that had obviously burned out due to to excessive current, but they failed to check the temperature
Over 1M ohm at the resistor to ground, and the only other path that's connected to the resistor (except for the SMD caps and the FSP chip pin) is a diode that has 47K ohm to ground on it's other end.
This diode (D1) is biased for current flow into the resistor and then into the FSP 6601 chip
Maybe there is something wrong with the chip

12V at the positive side of the diode
There is -485mV across the diode, and there is +15mV across the resistor with conventional current going into the FSP chip (both components are in series, and test leads orientated the same way along the trace)
I=V/R, so current through the resistor = 0.015/10 = 0.0015 amps
Something's not right there, as i don't think that that amount of current will cause the resistor to heat up to over 100 °C

[socketa]

Checked the voltages at either side of the diode:
12.15V at the cathode, and about 11.6V at the anode

[socketa]

Checked the temperature of the chip, the pin of the chip that's connected to the 10 ohm resistor, and the PCB near the resistor; but the resistor is the hottest.
How can a resistor heat up like that with such a small current?

[momaka]

Maybe there's a large AC current across it? Tried measuring the AC voltage? Probably going to be a bit tricky too with just a regular multimeter, as any high frequency may not be measured correctly, depending on the meter.
Me thinks this needs to be looked over with an oscilloscope (if you have one.)

[socketa]

Thanks for you suggestion,
Yes, that did cross my mind.
I have a scope
... and thought that that might be a good option, but wasn't sure about it, so decided not to, since i don't want a high voltage to damage the scope
Also i may have tried connecting a multimeter on AC, and didn't see significant voltage; so that might have discouraged me?
Now the PSU as gone back to the fan briefly spinning... but now with an additional short, mouse-type, squeak that's coming from the main transformer 😠 (used a small rubber hose to pinpoint the area)
I checked the rectifying diodes that are on the secondary heatsink, and one that isn't, and didn't find any shorts
The SMD diode, that's next to the 10K ohm resistor, is in series with it, and it's not shorted
I connected the scope to the side of the 10K ohm resistor (that is also connected to pin 13 of the 14-pin chip) and it has a steady 4V DC on it when the PSU is on standby
When the PSU is powered up, the scope briefly displays a well-defined, 1ms (1KHz), triangular waveform /| /| /| ... that has troughs at 4V, and peaks at 10V
I guess that i will have to take it out of the case again and check the other side of the resistor (between the resistor and the SMD diode), so as to find the voltage that's being dropped across it.

[socketa]

Was checking the other side of the 10K ohm resistor AC with a multimeter and was just about going to connect the scope, but before that i powered on the PSU few times and one of those mosfets blew up with a bang and a spark.
Talk about Murphy's law...Twice!
I took a chance, and replaced the mosfet
(The mosfet that's next to it looks damaged, but that was just some surface damage that happened when i was removing it earlier on)
So now after checking with a scope on both sides of the resistor...
the side that is connected to the chip displays an /| /| /| waveform, of which the peak is about 2V higher than when the probe is connected to the diode side
I think that if there was high AC current that was going through the 10 Kohm resistor it's no longer present because the PSU is now shutting down straight away, whereas before it wasn't, so it's not enough time.
If the main transformer squeaks, does that possibly indicate excessive current?

[socketa]

3.3V and 5V rails are shorted to ground,
so that's probably what's causing the transformer to squeak
also, testing across all four mosfets, in circuit, that are on the back of the PCB, the drain and source are all shorted to ground

[socketa]

Q27 and Q28 were shorted D-S,
which placed shorts across the other two non-shorted mosfets Q26 and Q29

Additionally, after removing the solder and disconnecting a leg, it looks like one of the rectifying diodes is open, as i get no response from it on multimeter diode test mode - this connects to the 5V rail via a 2 Ohm resistor, which, as mentioned before, is unexpectedly connected to the gate of Q29.
Neither he gate of Q29, nor it's series 2 Ohm resistor, are connected to any of the pins of the 6601 FSP controller chip

[socketa]

Ripped off a trace and pads when removing the transformer to get to what i thought were two heatsink-mounted diodes, but they were mosfets that had their third leg soldered to the heat sink
They both tested good as mosfets
Board's very bent
Over it...
It's now a parts PSU now
Maybe someone can add to this topic if they make any further progress with one of these