A (relatively) simple repair of a couple of power supplies

Some time ago, I was sent several different pieces of equipment for repair, including a couple of power supplies. The repairs themselves aren’t particularly complicated, but they cover quite a broad range of common failures. I’d like to use them as examples to show what’s what, what you should avoid doing, and what you absolutely should do.

As usual, a reminder: to keep up with new posts on the blog, I recommend following me on Instagram. I’ll post notifications about all new topics there, and might also write about something interesting from time to time - link to the account

You can also subscribe to updates and new articles on the Telegram channel - https://t.me/KirichBlog


First, an important point: I strongly recommend that you don’t try anything described in this article unless you’re confident in what you’re doing, have at least a basic understanding of how a switching power supply works, and have some hands-on experience. It’s dangerous for you, the “patient,” and your equipment. Power supplies contain dangerous voltages—at least 300–320 V.

First up is a power supply rated at around 300 W that will already be familiar to my readers. I’ve published both a review of a similar power supply (24 V) and a review of its modification (a 60 V power supply). What’s more, I later had to modify another 60 V power supply, so it seems this model is fairly popular.

This one is a 48 V power supply. Since these power supplies are rated at 300 W, the current is around 6 A.
Symptoms: it doesn’t work. Before it failed, a voltage (presumably) from the load appeared at the power supply’s output while the power supply itself was switched off. When mains power was applied, the power supply failed.



We always start a repair by carefully inspecting the components for damage, and the circuit board for soot and burn marks. In this case, everything looked fine except for the missing fuse, so we can go straight to taking measurements.
1. First, check the input capacitor. I usually recommend discharging it beforehand, since it can retain a charge in some situations. Here we have a short circuit, so the problem is either the inverter transistors, the bridge rectifier, or the capacitor.
2. Checking the drain-to-source path (the transistor’s middle and right-hand leads) also shows a short circuit. There are two transistors connected in parallel here, with their gates isolated from each other by resistors.
3, 4. Here are the first results: the resistance between the source and gate leads (the two outer leads) is about 2 kΩ on one transistor, while the other shows a short circuit.


The transistor with a short across its leads is probably blown, so we desolder it.
That’s right. After removing it, the resistance on the board returned to something close to normal; there are no obvious short circuits. Below are the results of checking the drain-to-source and source-to-gate paths with the meter connected in both polarities.
Why didn’t I just test the transistor? I could have, of course, but I’d still have had to check the board afterward, since the problem might have remained on the board.


At this point, a beginner electronics hobbyist might ask, “So, do we replace the transistor and get on with it?”
Of course not. This is just the beginning.

The gate circuits are isolated by low-value resistors. This is standard practice, and these resistors often burn out—which is exactly what happened here.
The left resistor is open, or close to it; the right one is fine.


Someone might ask, “Okay, can we replace the resistor, transistor, and fuse now?” Nope, not yet. We need to keep checking.

When a transistor fails, especially if it shorts to the gate, the components in the gate circuit often fail too, such as the resistor. But the resistor usually doesn’t burn out on its own.
By the way, the owner of this power supply was very lucky: the current-sense shunt survived, which means the current-measurement circuits survived too. More often, the failure also takes out this circuit, and in 90% of cases that means replacing the PWM controller. You still have to find one, and the markings are worn off, so you have to work out the pinout, find a substitute...

Let’s check the transistor control circuit. A similar 60 V power supply that was sent in for modification came in handy here.
At the top are the continuity-test results for the driver output in both directions on the power supply being repaired; at the bottom are the results for the working unit. The anomaly is immediately obvious. I decided to keep the photo of the working power supply’s measurement results just in case—it might come in handy for someone doing a repair.


Oh, and it turned out the Zener diode had failed—not by shorting out, which would have been easier to diagnose, but by partially breaking down. In other words, it tests almost like a good one, but no longer works properly. Incidentally, when a high-voltage transistor fails, the gate circuit is particularly vulnerable and should be checked first.

We were also lucky in part because the PWM controller doesn’t drive the transistors directly, but through an emitter follower (inside the green rectangle)—a pair of transistors just to the left and below. Of course, I could have desoldered them and tested them, and that’s exactly what I’d recommend doing. But I just checked them in-circuit. Besides, the resistor that supplies power to it is still intact; it’s just below.
The second high-voltage transistor is intact too, which means the zener diode protected the driver, while the gate resistor acted as a fuse.


The driver output readings in diode-test mode should look like this. I’ve included them just in case, since this kind of information can sometimes speed up troubleshooting considerably.


And of course, some beginners are probably exclaiming right now, “So, shall we replace the components and fire it up?”
No, we’re not firing it up. I still don’t know what caused the failure, and with power supplies you also need to check the output stage. We’ll skip the usual dead-capacitor scenario, especially since this PSU hasn’t seen much use and has several capacitors.

Let’s check the output. There’s another short circuit there, most likely in the output diode. Since this is a forward-converter power supply, that means the diode assembly.
That’s exactly what it is. The diode assemblies here are also connected in parallel, and one half of one assembly has shorted. So after repairing the primary side, at best we’d have triggered the protection; at worst, we’d be off to buy more components.
The original assembly is on the left; the diode assemblies are desoldered on the right.


So, the failed components are:
1. Transistor
2. Diode assembly
3. Gate resistor
4. Transistor protection zener diode.

Though my personal take is that the diode assembly failed first. Then the protection either didn’t react in time or couldn’t handle it, the transistor failed, the zener diode broke down, the resistor burned out, and finally the fuse blew.


Now let’s test the power supply. Of course, someone will ask, “How are we supposed to test it when half the power components are missing from the board?”
Easy. First, install the protection zener diode, then make a small modification to the board by adding an insulator under the heatsink, and then fit the heatsink with one working diode assembly.


Power it through a light bulb (this is essential) and check the output. And the output is fine. In fact, this PSU’s design means it has effectively gone from 300 W to 150 W, and it’s still perfectly usable.


The second unit I diagnosed was an apparently ordinary enclosed power supply. There must be millions of these around, and the reported fault was a classic too: loss of power. In other words, the PSU seems to work, but shuts down as soon as you put a load on it. It had originally been installed in a 3D printer.


When I opened it up, I expected to find the familiar TL494-based PSU, but it turned out to be something clearly newer. To be honest, I was glad; for some reason, I find TL494-based PSUs less enjoyable to repair.


1. It even has an input EMI filter.
2. The PWM controller, marked LO1216 BE0747LE, controls its operation. I didn’t look up what controller it is, and didn’t need to, since it seems to be working fine.
3. What really surprised me was the transformer. I’ve seen plenty of power supplies, but never one using a flyback transformer core as its magnetic core... That was quite a surprise. Well, it’s basically the same ferrite, and it can certainly be used this way, but this is the first time I’ve seen it in a factory-made PSU.
4. The output has a pair of “ordinary” 2200 µF, 35 V capacitors, rated for just 85°C.


We connect it and run a load-capacity test with the cutoff threshold set to 22 V. At 2.8 A, the load shuts off. We lower the threshold to 20 V, and it now shuts off at 5.4 A. Well, that makes it clear: I desolder the input capacitor and find that it’s gone open-circuit.


For a quick check, I temporarily swap in some used capacitor of roughly the same capacitance and test it again. The PSU now handles 11 A without any trouble, despite being rated for 10 A.


I don’t repair power supplies very often, perhaps because I don’t get asked much. And when I do, it’s usually someone asking about remote repairs. Or maybe it’s because repairs often aren’t economical, so I don’t keep a stock of common components at home.
So after diagnosing them, I sent the customer a list of what needed replacing and how much it would cost, got approval, and placed an order for both PSUs at once, along with a few things for myself to save on shipping. By the way, shipping a 100 g envelope via Nova Poshta came to... 75 UAH.

The components themselves came to:
SPA17N80C3XKSA1 transistor — 168 UAH. I could have found it cheaper, but then I’d have had to pay for shipping twice.
VS-60CPH03PBF diode assembly — 63 UAH
220uF 400V 25*35 105C 2000Hr capacitor — 84 UAH
2 TE 3300uF 35V 16*32 105C 5000Hr [Low ESR] capacitors — 2 × 25.20 = 50.40

I already had the other odds and ends, and I didn’t even bother counting them.


I’ll start with the 48 V power supply as well.
There’s no point showing the component replacement process, but I’ll explain a few things about the diode assembly.

Power MOSFETs can be connected in parallel without any problems, but diode assemblies are a different story. You either need to find an exact match—one that’s as close as possible—and install two identical ones, which means replacing both, or replace the two with a single, more powerful one. I went with the second option, especially since I’d done this before.
This avoids having to match the assemblies’ parameters, since there’s only one. Also, this diode assembly works fine in the 60 V version and has better specs.

The rest is straightforward and clear from the photos. I’ll just point out that it needs to go at the top of the heatsink, because if you install it lower down, its screw will hit the transformer. There’s no need to insulate it from the heatsink; I’ll explain why below.


About insulation: although these power supplies use insulated diode assemblies, the heatsink itself is connected to the output positive through the mounting screw, and in turn to the diode assemblies’ mounting flanges. This creates one unpleasant problem.
While working on another power supply, I mounted the assembly a little lower. It pressed slightly against the transformer, shifting the heatsink by just a fraction of a millimeter and shorting it to the ground plane. That’s why I strongly recommend adding insulation under the heatsink. You don’t even need to desolder the diodes: just loosen the mounting screw and slip a piece of cardboard underneath.


Unfortunately, the transistor I received is noticeably different from the one previously installed. I can’t say which one is actually genuine.
As you can imagine, at almost $5 apiece, replacing both would be quite expensive, so I’m replacing just one. It’s not as bad as it might seem: the only potential issue is a difference in turn-on and turn-off times, but there’s nothing I can do about that here.


The last thing I did was replace the fuse. Since it had left a bit of residue on the board, I removed the terminal block just to be safe and checked underneath, cleaning off some of the soot while I was at it. By the way, soot—and especially carbonized areas—conduct electricity well, so make sure to clean them thoroughly.
The fuse here is a small 5 A one. I didn’t have any like that, so I installed a regular fuse vertically.


After any repair, the device needs to be tested, so I connect it to a load and see how it behaves.


And its behavior isn’t encouraging: I found abnormal heating around the gate driver. Looking closer, you can see that the Zener diode and driver transistors are getting very hot.


I measured the PWM controller’s supply voltage and found it was about 19 V with no load and around 21–21.5 V at maximum load. Naturally, the 15 V Zener diode was having a hard time.
Ideally, I should lower the supply voltage itself, but that would mean more modifications. So I simply replaced the 15 V Zener with an 18 V one. That’s the highest Zener voltage you can use here, since the transistor’s maximum gate voltage is 20 V.

I forgot to mention: I didn’t have any 5.6 Ω resistors on hand. Well, I did, but they were 1206 size, while this board uses 0805. The pads could have accommodated 1206 parts, but their placement made it very awkward. For reference, you can use resistors in the 3–10 Ω range, but you need to replace both at the same time.


Let’s load it to 6 A and see what happens.


And that’s about it. The power supply handled 7 A just fine. I ran it at 6 A for a while, and no other problems showed up.


Clean the remaining flux off the board, and the repair is done.


Let’s move on to repairing the second power supply. This one is much simpler.
I ordered replacement capacitors for both the input and outputs, just to be safe.
1. The new input capacitor is rated at 220 µF, but actually measured about 193–194 µF. I checked it again with a multimeter, which gave a similar reading. Hmm... they’re 10% short...
2. My old used one. In principle, I could have left it in, but I don't like installing used parts in items I'm repairing, and only do so as a last resort, with the customer's approval.
3. New output capacitors. Unfortunately, the store only had this "brand"; the datasheet is here, and I didn't have anything suitable on hand. Personally, I think it's pretty expensive for what is essentially a no-name capacitor, but its capacitance is almost spot on, and my multimeter actually read exactly 3300 µF.
4. The old ones have even lower capacitance, and twice the ESR. I could have left them in, strictly speaking, but I wanted to do a better job.


You may also often see another capacitor on the power supply board: a tiny one in the high-voltage section. Its capacitance is usually around 10–47 µF, with a voltage rating of 25–50 V. If the power supply isn't new, replace it, especially if it's a no-name part. This capacitor is in the PWM controller's power supply circuit. It often dries out, causing the power supply to behave erratically and making it hard to start.


And here's what I don't like about repairing cheap Chinese power supplies: you diagnose the problem and fix it, then notice during reassembly that the transistor is clamped in a haphazard way. So, on top of the actual repair, you have to hunt down and fix other problems.
I had to add a fiberglass laminate spacer, with a layer of electrical tape on it to cushion things.


Put everything back together, replacing the thermal paste with fresh stuff along the way.


Always power it up for the first time through a light bulb, even if you barely did anything, even if you're 200% sure everything is fine, even if you have 60 years of experience (though you already know this) and repair five power supplies before breakfast. Always!
I didn't do that, and paid for it. Luckily, all it cost me was a fuse; it could have set me back 84 UAH + 75 UAH shipping. Guess why. The power supply on the left is powered up, by the way, but its own power consumption is very low, which is nice.


Connect it to the load and run the autotest with a 12 A limit. So far, so good.


Next, I load it at 10 A, its stated maximum, and wait. A few minutes later, I notice a smell of something overheating.


Actually, it turns out this is a "feature" of this power supply; it's simply designed that way. The thermistor and output load resistor get very hot, although 78 °C on the inductor and 92 °C on the transformer are hardly normal either. The load resistor reaches 120 °C; in the second photo, it's just partly hidden by the inductor.


And that's about it. Nothing particularly complicated, but as I always say, you learn to check everything after your first bucket of fried transistors.
Now, about whether repairing power supplies is worth it. In my opinion, the answer is close to no. How much do you think it should cost to repair these two power supplies, not counting the parts? Just curious.
Let's do the math. Cheap power supplies aren't worth repairing because... they're cheap. Expensive power supplies might be worth repairing, but they're often high-tech, tightly packed, and full of expensive components. So unless the problem is something simple, like dried-out capacitors or a varistor blown by an overvoltage, the repair can turn into a real headache. For examples of how modern power supplies are laid out, take a look at this one or this one.

So, yes, they can be repaired, but it's not always worth it. You have to assess each case individually, because sometimes repairing one really is easier, especially if you're doing it yourself.

And let me say right away: I don't really like questions like, "My power supply doesn't work. Tell me what I need to replace to get it working." First, remote troubleshooting takes a lot of my time and isn't very effective for the person asking, and it isn't always safe either. A technician can spot some things right away during an inspection, just by turning the unit over in their hands; photos aren't always enough. Plus, something I can check with a multimeter in five minutes can take hours, or even days, to sort out over messages.
That said, remote repair is an option, but it costs at least twice as much as if you did it yourself (at most, I can offer a quick tip for free if it's a common issue), and there's no guarantee of any kind. Sorry, my time is worth money too, and it's not always convenient to remember a couple of days later where we left off last time, especially when I may be handling several requests at once.

That's it. I really hope this was useful in some way. Maybe I'll share something similar again sometime. For now, I'm off to keep tinkering with the new board for converting a power supply into a UPS. I can already say it's working just as I expected :)