24 V 8–10 A power supply

I got the chance to test some more power supplies, and today I’m reviewing the first one.
As always, the review will cover the inspection, analysis, circuit, tests, and conclusions.


I ordered this power supply mostly out of curiosity; I hadn’t come across this model before. The review will be relatively short, but that doesn’t mean it was quick to put together. I’ll just try to present the information concisely this time.

The packaging is as simple as it gets: a zip-lock bag wrapped in a sheet of soft material. Banggood often packs items this way.


Specifications listed on the product page
Input: AC 85–265 V
Output: 24 V DC
AC frequency: 50/60 Hz
Output current: 8–10 A
Output power: 192–240 W
Dimensions: 140 × 72.5 × 38 mm (measured)

It’s an open-frame design, so it looks like the “people’s” power supplies. I wouldn’t be surprised if they had the same designer or manufacturer.


The board is noticeably larger than the one in the “people’s” model. The 300 W model is on the left, and the “people’s” one is on the right.


The overall build quality is reasonably good and, in my opinion, on par with or slightly better than the units shown above.



1. It connects via screw terminals. There’s a fuse at the input, as well as a proper mains filter with both X- and Y-type capacitors. The only thing missing is a varistor; for some reason, a Y capacitor is fitted in its place.
2. There are two 82 µF input filter capacitors, but their actual combined capacitance is slightly lower, at 143 µF, which is low for the rated power. There’s also a thermistor at the input, which I recently posted a video about.
3. The product title says 24 V 10 A, but the board itself is marked 24 V 8 A.
4. I wouldn’t call the transformer large. The core measures 33 × 30 × 14 mm, which isn’t much for a flyback power supply. Of course, you could say that the size depends on the operating frequency, but supplies of this type usually top out at 100 kHz. Also, the windings use thick wire, although at this power level it would be better to use Litz wire.
5. There’s a 1 nF Y-type interwinding capacitor, so it shouldn’t give you much of a “bite.”
6. There are two MBR20200 output diode assemblies, mounted on separate heatsinks, which might initially make you think this is a dual-output supply.


The board layout is spacious overall, but the PWM controller section is packed very tightly.


1. It uses an NE1119J PWM controller. Unfortunately, I couldn’t find a datasheet.
2. The inverter transistor is a K16A60W. I didn’t look for its datasheet, but judging by the marking, it’s probably rated for 16 A and 600 V. The package is fully insulated; there’s no thermal paste.
3. The output has four 1000 µF 35 V capacitors. Interestingly, two are 10 mm in diameter and two are 13 mm, even though they have the same voltage rating. There’s a red LED near the output terminals.
4. The output terminal block is a standard type, with two contacts per pole. To the left is a connector for a fan, wired in parallel with the output, with no control circuitry.

The fan connector makes it clear that this supply is designed for active cooling. It’s fairly common to specify two sets of ratings, one for passive cooling and one for active cooling. But the heatsink layout makes forced-air cooling awkward: you can’t set up a “blow-through” arrangement, and simply blowing air onto the power supply is less effective and often inconvenient.


The board has two layers, but components are mounted on one side only. There are many vias between the layers, especially in the power circuits. The routing is a little crude in places, but generally sound.


There’s something resembling a datasheet, as well as a short description from a Chinese forum. Here it is in Google Translate, in case it proves useful.



Of course, I drew the schematic for this power supply, and not for nothing: it includes some solutions I hadn’t come across before.


Some unusual features:
Red — An input Y capacitor connected in parallel with the X capacitor, most likely soldered in just to have one. An unusual PWM controller startup circuit; one advantage is that it is powered only when input voltage is present.
Green — The PWM controller’s main power supply circuit, consisting of two diodes and two capacitors.
Blue — The PWM controller appears to be quasi-resonant; one of its inputs gets voltage directly from the power transformer, before the rectifier.


What I didn’t like even before testing.
First, one corner of the board had taken a hit. It’s not critical, but the packaging should be better. Also, the input choke’s core is loose. It doesn’t affect anything, but still.


The high-voltage transistor’s heatsink is connected to the negative rail of the high-voltage supply, but there’s no insulation underneath it. It’s not a major issue, but the PWM controller’s startup trace and the second contact of the current-sense shunt run under the heatsink.


And, of course, the output-diode heatsink—or rather, the way the capacitors are installed. This isn’t a new solution, and you see it in Mean Well units too, but the quality of the capacitors they use is still different.

There are plenty of spots for ceramic capacitors in parallel with the output, but only one is installed.


Tests.

The power supply is initially set to a voltage slightly above the stated value. Again, this kind of variation isn’t particularly significant.
But the no-load power consumption really surprised me: it’s 0.1 W or less, because sometimes the wattmeter showed 0.0. That’s when I remembered the description on the Chinese forum —


Mostly out of curiosity, I decided to measure how long the PWM controller would keep running on the charge stored in the input capacitors. It lasted almost two and a half minutes. That is how long the output voltage stays regulated; then it starts dropping sharply because the output capacitors are no longer being replenished.


I measured the load regulation and efficiency at currents up to 10 A, which the power supply delivers without much trouble, although it gets noticeably hot.


Voltage regulation is mediocre: across the full range up to 10 A, the output voltage varied from 24.572 to 24.382 V, a difference of 0.19 V. Most of that drop occurred at load currents above 6–7 A.
Efficiency was about 86–87%.


I checked the overload protection threshold twice. The first time, I set the limit to 13 A and was surprised when the protection didn't trip. It turned out I had missed the trip point by only about 150–200 mA. Once the protection trips, the output shuts off completely, and the power supply restarts after a few seconds.


I planned to run the thermal test at load currents of 3.3, 6.6, and 10 A, for 20 minutes at each step. But after seeing how hot it got at just 6.6 A, I set the current to 8 A for the next step, and the overtemperature protection tripped literally a minute later. This told me two things:
1. It has thermal protection, which is good.
2. Without forced cooling, the power supply can't handle more than 6–7 A, which is bad.


Most of the heat is concentrated around three components: the high-voltage transistor, the transformer, and the output diode assemblies. Here are the thermal images taken after the runs at 3.3 and 6.6 A. For obvious reasons, there are no results at 8 A.


Since thermal images don't show the temperature of individual components, I also measured it with a compact pyrometer during the test. The high-voltage transistor got the hottest; the transformer still had some headroom, and the output diodes were nowhere near overheating.
I also measured the temperature of the output capacitors. It was only 8 degrees lower than that of the diode assemblies. That's not too much for branded capacitors, but you can probably guess how the no-name ones will fare.


Voltage stability with temperature is excellent, too. If there was any difference, it was around 10 mV. The power supply is hot on the left; on the right, it's shown about half an hour after the tests ended.


And of course, ripple. What kind of power supply review would this be without checking what's going on at the output?
1. With no load, the power supply switches to a mode with very short, infrequent pulses (around 1 Hz), so I had to increase the sweep time. The unloaded output voltage shows no noticeable fluctuations—just the same 10 mV I mentioned above.
2, 3, 4. Load currents of 3.3, 6.6, and 10 A. It's clear that there's no output filter inductor: the peak-to-peak amplitude of the main ripple is about 400 mV, but there's also a switching component, bringing it closer to 650 mV.
5, 6. At 5 and 10 A, but with a slower sweep. Given the relatively small filter capacitance, the 100 Hz ripple isn't too high, which is good.


I'll keep the conclusions fairly short, and they aren't encouraging. The idea is good, but the implementation is still a bit rough around the edges.
I wouldn't draw more than 150–160 W continuously without active cooling. Improving the cooling of the high-voltage transistor could raise that limit a little; I think even simply adding thermal paste would be enough, but it wouldn't make a huge difference.
The only way to significantly increase the power is to add active cooling, but as I mentioned, the power supply's design doesn't lend itself to effective cooling.

I wouldn't call the ripple excessive, especially considering the output voltage, but an inductor between the pairs of capacitors would help. The downside is that adding an inductor would increase the ripple current through the capacitors between the heatsinks, bringing us right back to the design's shortcomings.

In short, it works as-is, but with limitations and caveats. After some modifications, it can be used without issue. The question is whether it's worth it for the price, or whether the cost of the modifications would bring it up to the price of a branded unit, such as a Mean Well.

That's all for today. I hope you found this review useful.