48 V, 125 W power supply: inexpensive but surprisingly good

During the ongoing project to build a device for testing and balancing large battery packs, I needed to arrange a power supply of around 90 V. To do that, I bought two 48 V, 2.5 A power supplies. They turned out to be surprisingly inexpensive yet quite decent, both in terms of performance and build quality.
As usual: inspection, schematic, tests, and conclusions.



Today's review will be relatively short, but that doesn't mean it came together as quickly as it might seem at first glance.
I've already mentioned that I'm slowly putting together a test bench for a friend to test lithium battery packs. Since the charger needed around 90 V, I had to find a power supply with a similar output voltage.
There were three options:
1. A power supply with the required voltage from the outset. It turned out there aren't many of these, and what's more, they're relatively expensive. Their power rating was either too low or excessive.
2. Get a regular power supply, say 24 V, and use a converter to step it up to the required 90 V. That seemed rather cumbersome to me.
3. Use two lower-voltage power supplies connected in series, as I did when building a high-power adjustable power supply.

In the end, I went with the third option, mostly because I spotted some relatively inexpensive power supplies from the well-known Chinese electronics hobbyist 100MHz. The product page included photos of the insides, and after looking at those, I made up my mind.

I ordered a pair of power supplies. They cost $4.69 each, plus $1.88 for shipping to the forwarding agent, for a total of $11.26, not including shipping to me.
The combined shipment included a separate box with the power supplies. It looks like it wasn't even opened; they just put it straight into the larger box. Each power supply was in its own bag.


The power supplies are identical, and it looks like they weren't even taken out of their bags, since there weren't any fingerprints on them. What you see in the photos is from when I unpacked them.
The product page lists each power supply as weighing 230 g, so the two should weigh 460 g. But according to the forwarding agent's report, they somehow came to a whopping 810 g with the packaging.


Since the power supplies are identical, I'll show just one in this review. I simply tested the other one under load, confirmed it worked, and put it back in the box.
External dimensions: 160 × 70 × 38 mm
Distance between mounting holes: 152 × 62 mm

There's a very minimalist label on the side listing the input voltage range, output voltage, and current. That's all the information you get.


The connector arrangement is a little unusual. I'm used to seeing the mains input on a connector like the one on the red and black wires, and the outputs on connectors like the ones on the black and white wires. Here, it's the other way around. The colors make it obvious, of course, but it still looks a little odd.
The wires are secured with sealant, but it looks like they were originally bent the other way, so the sealant doesn't hold them as well now. Though I suppose I'm just nitpicking.


The case is aluminum, and everything is neat. I just can't figure out why there are two large holes on the bottom...


All right, let's grab a screwdriver and take apart this marvel of Chinese engineering. They really went to town with the screws—there are 10 of them.


Just from looking at it, I can say it doesn't look like a Mean Well, of course, but it doesn't look dirt cheap either. At least it's neatly put together.



It also has an input EMI filter, which I'll show in more detail later.


The input capacitors are rated at 68 µF each, for a total of 132 µF. That's quite enough for the rated 125 W, since the input voltage range is “narrow.”
The transistor has a silicone insulator and is pressed against the side of the case.

The transformer is fairly large, and there's a piece of thermally conductive rubber on top. The power supply cover acts as a heatsink for the transformer—a very thoughtful design.


The output diode assembly is insulated too and is also screwed to the side of the case. There are four 470 µF, 63 V output capacitors, for a total capacitance of almost 2000 µF, which is enough for an output current of 2.5 A.
The capacitors are unbranded, of course, but what more can you expect for less than five dollars?


The interwinding capacitor is the correct Y type. You can see an optocoupler nearby, along with the power supply model information and the design date. There was also a manufacturing date near the transformer: May 2011.

On the output side, there’s an inductor to reduce ripple, as well as a mysterious current-sense shunt. At first I even thought it was some kind of clever jumper, since I wasn’t expecting to find one here.


We remove four more screws and take a closer look inside.


And here’s the mains filter. What can I say? It’s almost textbook: all the necessary components are there—two common-mode chokes, a fuse, a varistor, a thermistor, X and Y capacitors. It’s surprising to see a proper filter in a cheap power supply.
The only thing I’d point out is that the varistor is rated for 560 V, or 394 V RMS. It’s not really there to protect against overvoltage, since it probably wouldn’t even trip if you applied 380 V. It’s meant to protect against transient surges, but I’d replace it with a 430–470 V part.


The resistors in the snubber on the “hot” side and the RC network on the “cold” side were also chosen “with love”—rated for at least 2 W. As it turned out, that’s just right for the high-voltage circuit.


What’s more, even across the power supply output, after the inductor, there’s not just a tiny ceramic capacitor but a decent 0.47 µF, 63 V capacitor.


The underside of the board looks neat. In general, a power supply topology with the hot and cold sides arranged parallel to each other is far from optimal in terms of size, since they need some clearance between them, and here that clearance takes up space along the long side.
But the “cold” and “hot” sides are separated, albeit without isolation slots. This is partly offset by the generous clearance and the fact that the PSU was designed to be grounded from the outset.


A closer look at the high-voltage section with the PWM controller.


And the low-voltage section.


1. They even remembered the resistors across the input X capacitors. Good on them for not being penny-pinchers.
2. The PWM controller is a TEA1733. I knew that before buying it, though—it’s listed on the product page. You can tell right away that the description was written by an electronics hobbyist :)
3. The high-voltage transistor is a P10NK70, rated for 700 V and 8.6 A.
4. The output diode assembly is an HBR16200, rated at 16 A and 200 V according to the marking.
5. I also found a dual op-amp, an LM2904, which I really wasn’t expecting to find here. At first I thought it was a TL431, but that’s on the other side of the board.
6. There’s another TL431 here, in a SOT-23 package.


The TEA1733 PWM controller may look simple, but it’s actually pretty good. The datasheet does, however, position it for power supplies rated up to 75 W.


I also wasn’t quite sure what the description meant about thermal protection. The block diagram shows it, and the description says it trips at 140 degrees, but it also mentions the option of connecting an external sensor to the universal Protect input.
In any case, whether it has internal thermal protection or not doesn’t really matter. If the chip heats up to that temperature, the other components will already be at a temperature “incompatible with life,” so we can safely assume there’s no thermal protection.



I drew the schematic here just out of curiosity. I wanted to understand what the designers had cobbled together and why it needed a second reference voltage source, an op-amp, and a shunt.
It turns out this PSU supports not only the usual CV mode but also CC mode, i.e. constant-current regulation. I might have thought it was an LED driver, but the product page said PoE, and the voltage is a good fit. Besides, I’ve never come across LEDs rated at exactly 125 W.
There’s also a somewhat odd RC network in parallel with the output diode assembly. And while carefully inspecting the circuit board, I noticed the traces are actually routed quite properly. Overall, I’d give it 4.5 points for now—not quite 5, but definitely 4.5.



This next bit isn’t really relevant to the review, but the designer even took the trouble to mark the diameter and length of the jumper wire.


I put most of the power supply back together, connected the power, tester, load, and oscilloscope.
The output is 49.82 V. There’s no provision for adjusting the voltage, so I’ll have to change the divider resistor values to get the 45 V I need, but I’ll do that later.


A rough estimate of efficiency and output voltage regulation.
1. 33% load: 42 W output, 49.1 W input—about 85.7% efficiency, output voltage 49.798 V
2. 66% load: 83 W output, 94 W input—about 88.3% efficiency, output voltage 49.775 V
3. 100% load: 125 W output, 141.3 W input—about 88.4% efficiency, output voltage 49.743 V

Efficiency is decent, but I wouldn’t call it especially high for a power supply with a 48 V output. Usually, the higher the output voltage, the higher the efficiency.
Output voltage regulation, on the other hand, is good.


Since the power supply was working steadily, I decided to continue the test.
1. Load current 3 A: 150 W output, 169 W input—about 88.7% efficiency, output voltage 49.722 V
2. Load current 3.03 A: 119 W output, 134 W input—about 88.8% efficiency, output voltage 39.175 V

The second part of the test shows that the power supply switched to constant-current mode. The transition is very abrupt: at 3.02 A, the output was still almost 50 V, but at 3.03 A it dropped below 40 V. Efficiency remained at a normal level.


There’s practically no high-frequency ripple across the entire range. Below are the oscilloscope traces at load currents of 0, 0.83, 1.66, 2.5, 3.0, and 3.03 A.


At a slower timebase, though, I saw what I’d noticed during the first test: some “wobble,” where the otherwise clean high-frequency output was modulated by low-frequency ripple. In the second part of the test, I saw 15–20 mV of ripple at around 250 Hz. That’s a very odd frequency, but in any case, even 20 mV of ripple at maximum load current is very good. The decent output capacitance, along with the output inductor and film capacitor, clearly helps.


I figured it out later: the low-frequency ripple is a feature of the controller, which has modulation at 280 Hz.


The unit works well in constant-current mode, and is excellent up to 40 V. At around 37–38 V, it goes into protection. This may have been due to the electronic load’s behavior, but the power supply worked perfectly steadily down to 38 V.


As usual, the long-duration load test is split into 20-minute intervals, at the end of each of which I measure the component temperatures.
First, a test at 50% load.


Since the power supply didn’t get very hot, I just used the thermal camera to see what was heating up the most.
1. 20 minutes at 50% load.
2, 3. Another 20 minutes, this time at 100% load. The snubber resistor and the input thermistor get the hottest; the thermistor in turn warms up the bridge rectifier a little.


A quick note about the thermal image above. In the first image, the transformer is the hottest component. The power supply is designed so that some of the transformer’s heat is transferred to the aluminum cover, so I considered the first test—run without the cover—to be invalid.
In the end, I simply placed the cover on top and covered the remaining opening with a plastic insulator. After that, the transformer reached just 67°C at 100% load, whereas without the cover it had reached 65°C at 50% load.


I raised the output current to 2.97 A to make sure the power supply stayed in CV mode, then ran it for another 20 minutes. The highest temperatures were at the snubber resistor and the input bridge rectifier/thermistor—around 100°C in both cases. The transformer reached 71°C.


But I wasn’t satisfied with that and wanted to see how hot the transformer would get without any additional cooling. I removed the cover and ran it for another ten minutes :).
In the end, its temperature rose from 71°C to 83°C, while the other components got a little cooler. In any case, the power supply delivers the rated 2.5 A without any problems, and can also operate normally at the maximum current of 3 A.


As it warms up, the output voltage drops. On the left is the cold power supply with no load; on the right, after 1 h 10 min of testing. The difference is quite substantial, but since the output voltage is high, it looks normal as a percentage—about 0.1%.


So, what can I say after testing it? I was tempted to call this review “Chinese doesn’t mean bad.”
The thing is, the power supply uses some obviously cheap capacitors, which detract a little from the overall impression. But in terms of circuit design and performance, I’m completely satisfied. I can only nitpick a little about the output ripple, but even at 120% load it’s just 20 mV peak-to-peak, so I think we can ignore it.
Otherwise, we have:
1. A low price.
2. An aluminum case and passive cooling.
3. A proper input filter.
4. An output filter.
5. Adequate capacitor capacity.
6. Almost no high-frequency ripple.
7. Acceptable temperature rise and efficiency.
8. Not just short-circuit protection, but also CC mode.

If you want to modify this power supply, I’d probably replace the capacitors, though I think they’ll be fine as they are. I really like the protection: in practice, you can’t overload this power supply, so it’s also hard to overheat. Up to 3 A, it regulates the voltage; above that, it limits the current. If the output voltage drops below the critical point, it switches to short-circuit protection. All of this works correctly.

Overall, if you need a power supply like this, I can recommend it. I’m happy with mine. It’s just a shame they’re fairly heavy, so shipping will be relatively expensive :(
Shout-out to 100MHz, and thanks for the great power supplies :)


This review was sponsored by the yoybuy.com forwarding service, which covered the shipping costs. Affiliate link to sign up; new users get a 10-off-50 coupon.
The two power supplies cost $11.26, including shipping to the forwarding service. The cost of shipping from the service depends on various factors. The two power supplies weigh 807 g with packaging, according to the order page on the service’s website.