Design mistakes found in switching power supplies

Actually, today’s post is about a power supply. The idea for this article came to me spontaneously while I was sorting through some junk at home and came across one of the power supplies.

I don’t remember how I got this power supply, but it’s clearly locally made. I used to build power supplies like this myself. What caught my eye was several obvious design flaws, and I’d like to show you “how not to do it.”
I’ll say right away that I’m not going to repair or modify this PSU, because I don’t see the point.

The board is relatively large, since I’d estimate the PSU’s power at around 35–40 W, with an output voltage of 12 V and a current of probably about 3 A.


The appearance—well, let’s just say it could have been a bit neater.


Some components are domestic (presumably dating back to the USSR), while others are modern imports.


The board’s build quality gets a C, maybe even a C-minus. It just looks untidy, and the soldering could have been neater too.


So I don’t just make accusations and get told, “Show us what you’ve built yourself,” here’s a PSU I made a long time ago (about 8 years ago) for a review. Its specs are roughly the same.



Anyway, back to the original example.
At the input there’s a thermistor and a common-mode choke. By the way, I think its windings aren’t separated but wound on top of each other. I used to use the same kind of chokes in a low-voltage converter.
The input capacitor is a 47 µF Samwha, which is perfectly fine.


The bridge rectifier looks rather undersized for a PSU like this. Next to it, you can see the snubber resistor, with a strange scorch mark on one of its leads.


The PSU came to me dead, with the high-voltage transistor literally having “blown its top.” By the way, the transformer is wound very neatly; I’d guess it was wound on a machine. I usually do it by hand.


The high-voltage section: what surprised me here was that a bunch of SMD resistors are connected in parallel with the snubber resistor mounted on the top of the board.
The PSU is built around the classic UC3842.


Mistake number one.
And this is the “fuse.” No, I’ve seen this kind of solution in other PSUs too, mostly Chinese ones (Chinese-designed, not necessarily manufactured), but it should only be used in a low-voltage circuit. Here’s why.
When a fuse like this blows, the circuit board material can char, and carbon conducts electricity—not as well as copper, but it still conducts. Then comes the heat, a fire, and another apartment renovation...


Let’s take a close look at the board and see its flaws, apart from the flux residue and poor soldering.


The transistor and diode assembly are mounted on the same heatsink. That’s not ideal, though it’s tolerable—even well-known manufacturers do this. In fact, in some cases they even insulate a transistor that already has an insulated package.


Mistake number two.
“Everything is perfect” here: at a glance, there are four problem areas that also affect safety.
The arrows mark several places where a breakdown could occur between the primary and secondary sides of the power supply. In some cases, the “hot” and “cold” sides of the PSU are simply too close together; in others, a breakdown could occur along the path: primary circuit – heatsink – secondary circuit.
They could at least have added isolation slots. That wouldn’t have eliminated the problem completely, but it would have improved things somewhat.


Mistake number three.
For some strange reason, they decided to squeeze the optocoupler’s leads closer together. In other words, the distance between the leads on each side was originally one thing; decent PSUs spread them apart a little more for safety, but here they’ve done the opposite and squeezed them together.


As a result, we have several more problem areas, both between the optocoupler’s leads and next to them. As a reminder, the minimum safe clearance is usually around 7–8 mm; more is preferable. In places like these, it’s also a good idea to add isolation slots so that contamination on the board won’t cause a surface breakdown.


And again, so I’m not just making claims, here’s an example of a Chinese PSU, that is, even if it isn't a name-brand or expensive one.
As you can see, all the problem areas have adequate creepage clearance, and slots have also been cut.


Mistake four.
A regular high-voltage capacitor is used between the windings, rather than a dedicated Y-class capacitor.


The Chinese used to use capacitors like this sometimes too, but I haven't seen this kind of blunder in a long time. They almost always use Y capacitors now.


Mistake five.
The output uses a completely ordinary general-purpose capacitor, even if it is a name-brand part. This is an SD-series capacitor from Samwha. These capacitors are well suited to low-frequency circuits, but not at all to the output of a switching power supply. And if it were after the inductor, that would at least mean lower ripple, but it's right after the diode.


To be fair, I sometimes use capacitors that aren't quite suitable too—also Samwha, but from the RD series. They're not designed for this kind of operating mode, but they have lower internal resistance and work perfectly well when used with plenty of capacitance headroom.


But if you want your power supply to last, it's still better to use more suitable capacitors.


The low-voltage section has a minor routing flaw too.
The output capacitors are connected in such a way that there are sections where current can flow in both directions.

I couldn't be bothered to investigate further, but the transistors in the output circuit near the TL431 shunt regulator caught my eye.


In simplified terms, the left-hand version is wrong and the right-hand one is how it should be. This power supply has something in between.


I'll let the output inductor, wound in a messy bundle, slide. But the 1 W resistor in the output snubber surprised me a little. Why use one with such a high power rating? I don't understand; 0.25 W is usually more than enough for a power supply of this rating, with plenty of margin.


So there you have it. They may seem like minor details and don't affect whether the power supply works, but they're still quite dangerous. Could these mistakes have been avoided during the design stage? I think so. There are practically no parts of this power supply where getting things right would have been a struggle. The only possible exception is the Y capacitor, but even then you can always use two in series, one near the hot side and the other near the cold side—if there's really no other option.

Anyway, just a recommendation: don't repeat mistakes like these.

That's all from me. I haven't disappeared; I just haven't felt much inspiration to write lately.