Meanwell PSU blew up!

[XXL]

Had a rig on last week that had a meanwell psu powering it. Lasted 2 days and went off. I Took the rig down and found the psu had shorted inside. So I took the lid off and found 3 burnt resistors. It says it can provide 7.3a at 48v but I was only drawing around 3a. I thought these supply’s were meant to be good ? Iv never had a Chinese one blow up before, even when shorting it out to check the max current out.

What would cause those 3 resistors to burn out ? Also the fuse is still intact somehow. So that was useless.

[RF-Head]

RF got in the PSU
something wrong with the ampifier design or antenna problem

[Albert H]

Yes - that's almost certainly the problem. I've had power supplies become unstable with lots of RF about. Sometimes it can be cured with filter capacitors or in-line inductors to stop RF getting into the SMPSU. It can also be due to poor layout, or if the aerial's too close to the rig.
I never truly liked SMPSUs (though they're much better than they used to be). I used to find that they'd put switching artifacts on the DC output (often at inconvenient frequencies!). I much prefer a transformer, bridge rectifier and some smoothing capacitors - heavier and often more expensive - but they're difficult to blow up!

[XXL]

Yeh usually if iv got RF problems getting into the supply it just buzzes and I get lower voltage. Usually because the psu is super cheap crap, probably not even good enough to power some fairy lights. Iv never had it physically blow up before.

[thewisepranker]

This is mostly me thinking out loud.

R18 and R19 are current shunts, they are there to give the PSU some way of inferring the output current by measuring the primary current. The only way they can get hot is for a high current to flow across them. But they don't look to have failed, the outer finish doesn't look to have suffered heat damage. If you look at R60, it's damaged all the way around the circumference. R19 and R19 aren't.

R60 is typically 1k. It connects one side of the two current shunts, which are R18 and R19 and are in parallel, to pin 3 of the NCP1252 PWM controller. The datasheet shows that internal to the NCP1252 is a ~26k resistor connected to pin 3, but other than that they don't specify whether there are clamp diodes for protection. The fact that there's a ~26k resistor tells us information about the order of magnitude of how much power we'd expect R60 to be dissipating. That would be a very low number given that the shunts are a low value - there's two in parallel which halves their effective value. They are deliberately chosen to be low (probably ~0.5 Ohms each) for efficiency reasons and to limit their power dissipation (and thus operating temperature) to sensible values.

If RF injection was the problem, a likely cause of failure is that the primary winding drive FETs became uncontrollable and started conducting randomly. This directly is unlikely given that the gates have got 5k1 pull-downs, the tracks aren't excessively long and the series gate resistors are 51 Ohms for the turn on and 5.1 Ohms for turn off.
The primary drive FET gates are transformer driven so that an N-channel device can be used for both the high and low side. If the gate transformer driver, or the NCP1252 driving the driver, got into trouble or damaged, then random conduction could happen. Worst case is that both the high and low side drivers conduct continuously for a significant amount of time, which puts a large portion of rectified mains voltage (say 100-200 V?) across the current shunt resistors. This puts the same good portion of rectified mains on the current sense pin of the NCP1252, pin 3, which it probably can't handle, so it will blow up.

In this case the lower thermal mass of R60 and any input protection within the NCP1252 would be much lower than the thermal mass of the current shuts so would almost certainly blow up before R18 and R19. Given that the gates of the primary drivers are transformer coupled, no DC can be given and they can't be held on, so once the NCP1252 has blown up, that's it. Hence, the fuse didn't blow.

The primary winding would also be prone to damage but the fact that the gates are transformer driven stops this from becoming a real problem unless the drive FETs are damaged to the point of being D-S shorted - the thermal mass of the NCP1252 is much lower so will fail first.

Can you take the board out and have a look underneath to see if the NCP1252 is also blown up? It's an 8-pin surface mount chip, near to R60 but on the other side of the board.

[sinus trouble]

Oh Dear?? Sorry, I dont mean to laugh! I do really "MeanWell"

The failure could have been caused by many things? Are there any signs of moisture or weather related issues on the PCB?

Did you load test your setup? Or did you assume that it was 3A?

I would suggest that you check that your rig is functioning correctly before replacing the PSU

I am inclined to agree with Albert! A linear PSU is my choice! SMPSUs are great in some applications? But not rugged enough for me!

[jvok]

@thewisepranker do you have a schematic for those meanwells or did you trace that out from the board?

[thewisepranker]

I found this: https://www.kirich.blog/obzory/bloki-pi ... 50-24.html
For some reason R60 is labelled R25 on the schematic but its value and position in the circuit correlate from looking at the pictures, unless I've made a mistake.

[sinus trouble]

I found this: https://www.kirich.blog/obzory/bloki-pi ... 50-24.html
For some reason R60 is labelled R25 on the schematic but its value and position in the circuit correlate from looking at the pictures, unless I've made a mistake.

Nice find WP!

An interesting detailed article which i enjoyed reading!

Looking back at XXLs pic, It does indicate a FET blowout which gave the false impression of the Shunt resistors failing?

The resistor below the shunt has clearly blown but was possibly a consequence of the FET failure?

[thewisepranker]

Where did you buy it from?

Put the serial number into here https://www.meanwell.com/serviceReport.aspx
If you get nothing, it's not a valid serial number and it's a counterfeit.

The MW logo on the transformer looks suspect but that may just be how they are on a certain batch. If you look at images of other MeanWell internals the logos are correct, including the Kirich link I posted earlier.