Switchmode power supply repair, voltage too high, resistors getting hot

[NeedMoreCoffee]

Intro

Greetings! I'm a total noob with magic electronics, so allow me to apologize in advance for the inevitable ineptitude you're going to be exposed to. I'll try and structure the post as best I can.

I'm trying to fix one of those cheap wall wart PSUs. I know it makes no sense economically, but I thought it would be a good learning opportunity - even if I screw up and make everything worse, it's not a big financial loss. So it's not really an important problem, feel free to skip this one.


Description of the device

It's supposed to be a 12V 2A supply with your standard barrel jack. Nothing fancy about it. I think it originally came with an external HDD enclosure. Of course, no schematics. At least none that I could find.


Symptoms

I was using it to power one of those breadboard supplies that can give you either 5V or 3V. While fiddling on the board, I noticed a crackling noise inside the switch every time I turned it off or on. Eventually the breadboard module wouldn't turn on at all anymore and the switch was dead. I checked the 12V PSU with my multimeter and it read about 24V.

I unplugged it from the outlet and let it just sit there for a while. After plugging it back in, it was at 18V, slowly rising at a rate of about 10 mV/s. Unplugging it again, waiting, plugging it back in, and now it read 16,5V. Again, slowly climbing.


Inside

I opened it up and did a visual inspection. First thing I noticed was that one of the secondary side electrolytic caps had a bulged top. But I think this is actually another symptom, not the cause. Why would a smoothing cap cause the voltage to rise? And the caps were rated at 16V, so if the secondary actually did go up to over 24V, I can understand why those caps wouldn't be happy. I replaced both of them with 25V rated caps anyway, just to be safe.

The only other thing I noticed was a blackened resistor at the bottom of the PCB, surrounded by soot. It hadn't turned into charcoal yet and measured at 460Ω. But I have no idea what it was supposed to be because the writing was gone. The resistor connects on one end to the positive output and on the other end to the optocoupler. I presume it's a current limiter, so 300-500Ω seem to be in the ballpark with a 12V rail.

After the gears had turned for an embarrassingly long time, I thought: Ok, if the resistor changes value as power is going through it (i.e. heating up), and it's this resistor that is providing feedback for the voltage regulation, that would explain why the output voltage was off in the first place and also slowly changing over time.

Since I had no idea about the correct value, I put in a 470Ω resistor. And the output was about 16V. Then I switched to 330Ω, and the voltage went to 12.5V.

So all is well? Not really. This time, the voltage seems to slightly decrease over time (not as fast as before). And the resistor got really hot. When it became too painful to touch, I turned off the power.


Conclusions and questions

Here's my thinking:

• The problem is probably on the secondary side. After all, that's where the damage was. Also, the supply turns on all right and I get output.
• The caps are just an unfortunate victim of the actual problem.
• The resistor likely also wasn't the cause and it yet another victim. Something drew too much current through it, so it got grilled.

But I'm not sure where to go next with my diagnosis.

• An obvious suspect would be the optocoupler. It is on the same circuit as the resistor, after all. But how likely is that? Shouldn't those things be pretty reliable? Anyway, in diode mode, I measured a 0.989V drop in one direction and open line in the other. And yes, that's out-of-circuit. Not sure if it's normal though. I expected some kind of LED inside, and they usually have a higher drop. But I don't have any experience with optocouplers. It says "B1646" on it, but a Google search didn't yield anything of interest. I looked at some data sheets for other optocouplers, and they all had forward voltages in the expected range (1.2V or higher). And I think I can see how a lower voltage drop might result in a higher current draw by the coupler, creating heat in the resistor. I hear those LED thingies are not linear with their voltage vs. current.
  
  Even if the optocoupler is faulty, what to replace it with? Is this one of those things where you're probably fine with any replacement that meets some basics (e.g. voltage rating)? Or is it some finely tuned, hand-picked part where the universe will explode if any one of 50 different specs is off by a pico-percent? What should I be looking out for?
  
  
• Another potential issue is the schottky diode. It's an SB5100, and according to all the data sheets I could find, it should have a voltage drop of about 0.6V. But I measure 0.27V (and open line in the other direction, so it's not shorted). The only thing I have on hand as a replacement would be an SR5150. But they have a similarly low voltage drop, so would that even work? Can the lower drop create too much strain on the other components?
  
  
• Or is there anything else that could be at fault here? I probed around a little for shorts but couldn't find any. And there's not much stuff left on the secondary side. A couple of ceramic caps (no shorts), an inductor, and a voltage reference (TL431A). On the latter, I get a 0.6V drop between cathode and anode (and OL backwards), which seems reasonable to me, but with my limited understanding, I'm not exactly sure what I should measure from the ref pin to the two others. The data sheet did not enlighten me in that regard.

I'm now at the point where I might have to order parts or start rummaging for salvage, so I figure I'd double-check with you guys first. Thanks to everyone who made it this far!

[tunk]

The burnt resistor, other resistors and the TL431 could be part of a voltage divider circuit regulating the output voltage.

Edit: Maybe you could reverse engineer the part with the TL431, five resistors (R16+17+19+20+27) and optocoupler?

[LaryPant]

First of all, what a descriptive and well explained post, nicely done. Although I fear that I wont be of much help.

• Another potential issue is the schottky diode. It's an SB5100, and according to all the data sheets I could find, it should have a voltage drop of about 0.6V. But I measure 0.27V (and open line in the other direction, so it's not shorted). The only thing I have on hand as a replacement would be an SR5150. But they have a similarly low voltage drop, so would that even work? Can the lower drop create too much strain on the other components?
  

The 0,27 V drop across the Schottky diode seems very reasonable. I checked the datasheet (https://www.diodes.com/assets/Datasheets/ds30135.pdf) and I think that you may have misinterpreted the graph in figure 2. At room temperature the diode has a forward voltage drop of 0,6 V when the current is 2 A, but a typical multimeter uses a test current of about 1 mA. Figure 2 only shows the forward drop down to 100 mA, where the voltage is 0,34 V. A forward voltage drop of 0,27 V @ 1 mA therefore seems very reasonable and I doubt that the Schottky diode is the problem.

The resistor R16 definitely looks burned and I agree with you that it could be either a symptom or the actual cause. If it is the cause, then you/we might be able to figure out the correct value if you try to reverse engineer the circuit around the resistor, which should be doable considering that this is only a single layer PCB and it is always a great learning opportunity . Just focus on the secondary side around the burned resistor.

Lary

[NeedMoreCoffee]

First, thanks to everyone who responded.

I think that you may have misinterpreted the graph in figure 2. At room temperature the diode has a forward voltage drop of 0,6 V when the current is 2 A, but a typical multimeter uses a test current of about 1 mA. Figure 2 only shows the forward drop down to 100 mA, where the voltage is 0,34 V. A forward voltage drop of 0,27 V @ 1 mA therefore seems very reasonable and I doubt that the Schottky diode is the problem.

I think you're on to something. As a quick test, I took my bench power supply and put 4.5A through the diode. With my multimeter probes across it, I get a voltage drop of 0.78V. Which does correspond with the data sheet.

I'll take a closer look around the resistor. Although I am not confident in my reverse engineering skills.

[tunk]

Although I am not confident in my reverse engineering skills.

No expert, but I think it's fairly standard, maybe close to the
schematics in this link:
https://danyk.cz/impulz_en.html
I'm sure you know that the primary voltage is more than 300V,
discharge the big capacitor before working on it.

[Gyro]

Well for a "total noob" you seem to have had a very good shot at analysing and presenting the problem.

The secondary side electrolytics may have bulged due to overvoltage (depending on duration) but they could also have failed due to poor quality and high ripple current. It's a very common failure mechanism on SMPSs.

I agree with the others, to understand the reason for the burned R16 it will be necessary to reverse engineer the TL431 to opto circuit. It might be a symptom (rather than a cause) of the excessive output voltage but it does took pretty toasted and you don't normally see such high dissipations in that part of the feedback circuit. It would be a good opportunity to download the TL431 datasheet and see if the voltages are as expected. Understanding R16's position will give you more understanding of whether the LED in the optocoupler might have been damaged too.

As tunk says, be extremely careful of the primary side, it will severely bite you given the opportunity. I would recommend (after checking that the primary electrolytic has discharged) that you securely wrap the primary half of the SMPS [Edit: Including the heatsink] in card or flexible plastic and tape to reduce the chance of accidents.

[NeedMoreCoffee]

I'm back!

First of all, thank you all for your concern regarding the high voltage side and its cap. Yes, I have been properly discharging it before handling the PCB. If the time ever comes where I have to handle it while it's running (probing etc.), I'll take proper safety measures.

Now, I tried to figure out the circuit (on the secondary side). I drew the circuit with that Falstad circuitjs online thingamabob. Can't really run a simulation because it's incomplete, but it was an easy way to cobble together a diagram. So this is the circuit diagram on Falstad. I also attached an PNG with the diagram. The simulator wants a value for each component, so if I didn't know the actual value, I marked it as such.

If anything doesn't make sense, it's probably because I made a mistake. I only understand bits and pieces here and there, so I cannot sanity-check this diagram. For reference, I also attached a photo of the bottom of the PCB (only the secondary side) where I painted an overlay of the different tracks. The AC side is different shades of purple. Or magenta. Or whatever. I'm like an old EGA card, I only know 16 colors. Positive is shades of red, negative is blue. The components on the top are in bright yellow. Sometimes, tracks go under components or are obscured by the silkscreen, so this may also contain errors.

Still not sure why the PSU is behaving like this, though.


EDIT: Forgot the tiny little diode that goes from the other end of the AC to the positive. Fixed.

[NeedMoreCoffee]

Another update: I think I found the culprit.

I went with the brute-force approach to just replace everything in that part of the circuit. I had a broken ATX power supply in my junk box and salvaged some parts from that. It had no TL431 but a couple of AZ431. And their data sheets looked so similar (to me, so caveat emptor) that I thought they might be a drop-in replacement. The AZ431 worked, but didn't change anything. The problem persisted.

Then I grabbed the optocoupler. As mentioned in the first post, I have no information about the one in the wall wart. But the one from the ATX unit was a seemingly common Cosmo 1010. I went "no guts, no glory" and just popped it in. And lo and behold, the PSU came good. It now provides a stable 12.3V on its output, and the resistor no longer feels hot.

I connected a 2.4A load and let the unit run for about half an hour. Aside from some coil whining (which I understand to be a normal side effect of inductors doing their thing), everything seems fine. The output voltage is still good and the resistor is barely above ambient. Stuff on the the primary side (transformer, cap, and especially the heatsink for the switching mosfet) was noticeably warm after the stress test (I only checked after unplugging and discharging the cap), but not to the level of being painful to touch. I think I can chalk this up as normal, given I was pushing the unit to its limits.

Are there any other tests I should run or can I call this a successful repair?


TLDR: It was the optocoupler.

[magic]

Looks like it's good now.

If the optocoupler is bad, the 431 will sink some 100mA or more through it, desperately trying to get that internal LED to light up and signal overvoltage to the primary. This may have smoked that tiny resistor.

For peace of mind, you may check voltage across the new resistor, calculate current and power dissipation.

[NeedMoreCoffee]

For peace of mind, you may check voltage across the new resistor, calculate current and power dissipation.

Good idea. It came out as 0,8 mW (520 mV across a 330Ω resistor), so I think that's ok.

[magic]

Yeah, that's not a lot. But if TL431 goes effectively short circuit, the same resistor would dissipate over 1W with 20V across it.

There is chance that the original TL431 survived unharmed, but I wouldn't test it in this PSU and risk further problems if it's bad.