500 W Switching Power Supply for Audio Amplifiers

Many of you know how much I enjoy taking a closer look at different power supplies. This time I have a somewhat unusual one on my bench—at least, I haven’t tested one like this before. In fact, I don’t recall seeing any reviews of this type of power supply, even though it’s an interesting design, and I’ve built similar supplies myself.
I ordered it purely out of curiosity, thinking it might come in handy. More on that in the review.


I should probably start with a little background. Many years ago, I was really into audio equipment. I went through everything from fully DIY setups to “hybrids” using power amplifiers rated up to 100 W from the Yuniy Tekhnik store, along with a partly dismantled Radiotekhnika UKU 010, 101 and Odyssey 010. Later, there was the Feniks 200U 010S.
I even tried building Sukhov’s power amplifier, but something went wrong back then. I can’t remember exactly what anymore.


I had various speakers too, both DIY and off-the-shelf—for example, the Romantika 50AS-105 and Kliver 150AS-009.


But the ones I remember best are the Amfiton 25AS 027, although I’d modified mine a bit. Along with some minor changes to the circuit and construction, I replaced the original 50 GDN drivers with 75 GDN drivers.
These and the previous photos aren’t mine, since I sold my equipment a long time ago. After that I switched to Sven IHOO 5.1, and eventually ended up listening only to small computer speakers. Yes, quite a regression.


But lately I’ve had the urge to build something—maybe a power amplifier, perhaps just for the sake of it, and maybe do everything differently. In the end, though, I decided to order a power supply. Of course, I can build one myself. In fact, in one of my reviews I did just that and even posted detailed instructions. I’ll come back to that later; for now, let’s get on with the review.

Let’s start with the claimed specifications:
Input voltage: 200–240 V
Output power: 500 W
Output voltages:
Main: +/-35 V
Auxiliary 1: +/-15 V, 1 A
Auxiliary 2: 12 V, 0.5 A, galvanically isolated from the others.
Dimensions: 133 x 100 x 42 mm

The +/-15 V and 12 V outputs are regulated; the main +/-35 V output is not. I’d like to share my take on that.
I’m often asked which power supply to buy for a particular amplifier. My usual answer is that it’s easier to build one yourself using well-known IR2153 drivers or their equivalents. The first question that follows is always: “But those don’t regulate the voltage, do they?”
Personally, I don’t think voltage regulation is just unnecessary for a power amplifier—it can sometimes be harmful. A regulated power supply usually produces more high-frequency noise, and there can also be problems with the regulation circuitry because a power amplifier doesn’t draw power evenly; it draws it in bursts. We listen to music, not a single frequency.
An unregulated power supply usually has slightly higher efficiency because the transformer always operates under optimal conditions. It has no feedback, so it’s more like a conventional transformer, but with lower winding resistance.

And here’s an example of a power supply for power amplifiers.

The packaging is soft, but it’s wrapped so thoroughly that it’s unlikely to get damaged in transit—though the battle between the postal service and sellers will probably never end.


It looks nice on the outside; there’s not much to complain about.



It’s fairly compact, especially compared with a conventional transformer of similar power.


The product page in the store gives more precise dimensions.


1. The power supply has an input connector, which turned out to be quite convenient.
2. It has a fuse and a proper input filter. The only thing they forgot was a thermistor to protect the mains, diode bridge and capacitors from inrush current. That’s not good. There are also jumper pads near the input filter that need to be shorted to switch the power supply to 110–115 V. Before powering it up for the first time, check that the pads aren’t shorted if your mains voltage is 220–230 V.
3. The diode bridge is a KBU810. Nothing wrong with that, but it has no heatsink, and at 500 W one would be advisable.
4. The input filter capacitors are rated at 470 µF; their actual capacitance is about 460 µF. Since they’re connected in series, the total input filter capacitance is 230 µF, which is on the low side for a 500 W output. Incidentally, the board also has provision for a single capacitor. In any case, I wouldn’t recommend increasing the capacitance without installing a thermistor. There’s even a spot for one to the right of the fuse—you just need to solder it in and cut the trace underneath.


The inverter uses IRF740 transistors. They’re far from new, but I used them extensively in similar applications in the past. As an alternative, the IRF830.
The transistors are mounted on separate heatsinks, and there’s a reason for that. The heatsinks are connected to the transistor cases—not only at the mounting points, but also through their mounting tabs, which are connected to the circuit board. In my opinion, this is a poor design, since it will cause unnecessary RF emissions at the switching frequency. At the very least, I’d isolate the inverter’s lower transistor (the one farther away in the photo) from its heatsink, and the heatsink from the circuit.

The transistors are controlled by an unknown module, but judging by the presence of a power resistor—and simply based on my experience—I don’t think I’d be far off in saying that it contains a basic IR2153. Why they made a module like this remains a mystery to me.


The inverter uses a half-bridge topology, but the midpoint is formed not by the junction of the filter electrolytic capacitors, but by two 1 µF film capacitors (the two mounted parallel to the transformer in the photo). The primary winding is connected through a third capacitor, also 1 µF (the one mounted perpendicular to the transformer).
This is a well-known and convenient solution, since it makes it easy not only to increase the input filter capacitance, but also to use a single 400 V capacitor, which can be useful for an upgrade.


The transformer is quite small for the claimed 500 W output. I’ll test it under load as well, of course, but I can already say that in my opinion its actual continuous power is no more than 300–350 W.

The store page listed the following among the key features:

In other words, the transformer uses a winding made of 100 strands of oxygen-free wire, each 0.1 mm in diameter, to reduce heat and achieve efficiency above 90%.
I’ll check the efficiency later, but the winding is definitely multi-strand. I didn’t count the strands, of course, but it’s a pretty decent bundle, and this winding method really does improve the transformer’s performance in particular, and the power supply’s overall performance.


They didn’t forget the capacitor connecting the “hot” and “cold” sides of the power supply, and they used the correct type (Y1).


The output rectifiers for the main rails use MUR1620CTR and MUR1620CT diode assemblies (16 A, 200 V). The manufacturer didn’t cobble together any “hybrid” arrangements, but used the proper pair of complementary assemblies: one with a common cathode and the other with a common anode. Both assemblies are mounted on separate heatsinks and, as with the transistors, aren’t electrically isolated from the components. In this case, though, the only concern is electrical safety. If the case is closed, there’s nothing to worry about.
The output filter uses two 1000 µF, 50 V capacitors, which seems a bit low to me.


There’s also an inductor between the capacitors to reduce ripple, and the capacitors after it are additionally bypassed with a 100 nF ceramic capacitor.
The product page also said:

According to the translation, all the capacitors have low impedance to reduce ripple. And that is more or less the case: Cheng-X capacitors are used, but they’re basically just slightly improved versions of ordinary Chinese capacitors. I’d rather have seen my favorites, Samwha RD or Capxon KF.

There are no discharge resistors across the capacitors, even though there’s room for them on the board, so you may get some “surprises”—the charge holds for quite a long time.


The additional power outputs are connected to their own transformer windings, and the 12 V output is galvanically isolated from the others.
Each output has independent voltage regulation, inductors to reduce interference, and ceramic capacitors at the output. But you may have noticed that the rectifier has five diodes. The 12 V output uses a half-wave rectifier.


There are terminal blocks at both the input and output, and they’re quite good in terms of both quality and design.


The product page has a top-down photo that shows everything at once. I only noticed later that all the store photos show mounting standoffs, but my kit didn’t include any :(


The circuit board is double-sided and very well made. It uses fiberglass laminate rather than the usual phenolic paper laminate. A protective slot has been cut into one of the narrow sections.
I also found a pair of resistors on the underside. I assume they form a basic overload protection circuit, sometimes added to IR2153 drivers. But honestly, I wouldn’t rely on it.


The underside of the circuit board is also marked with the output terminals and the output voltage options these boards are made for. Two things caught my attention: two identical ±70 V options and a custom option.


Before moving on to the tests, I’ll say a little about my own version of a similar power supply.
About three and a half years ago, I posted a review of an adjustable power supply, which used a power supply built using roughly the same circuit.


Assembled, it looked quite similar too. Sorry about the poor photo quality.


If we strip my version of all the “extras”—for example, the circuit that adjusts the fan speed according to temperature, the beefed-up transistor driver, and the auxiliary power circuit fed from the inverter output—we end up with the circuit used in the power supply under review.
Essentially, it’s the same power supply, just with more output voltages. The circuit is very simple; only a basic self-oscillating converter would be simpler.


The power supply under review also has a basic output power limiting circuit. I suspect it’s implemented as shown in the highlighted section of the schematic.


But let’s see what this circuit—and its implementation in the power supply under review—can actually do.
It’s worth noting that the main output voltage isn’t regulated, so it depends directly on the mains voltage.
With an input voltage of 223 V, the output is 35.2 V with no load. Power consumption is 3.3 W.


The resistor powering the transistor driver gets noticeably warm. It’s rated at 150 kΩ, which at 300 V means it dissipates about 0.6 W. This resistor heats up regardless of the power supply’s load.
The transformer also gets slightly warm. This photo was taken about 15 minutes after power-up.


For the load test, I put together a setup consisting of two electronic loads, an oscilloscope, and a multimeter.
The multimeter measured one power output; the other was monitored by the voltmeter built into the electronic load, which was connected with short wires.


I won’t bore you with a long list of tests, so let’s go straight to the oscilloscope traces.
1, 2. Different points at the power supply output, before the diode assemblies, shown with different time bases. The inverter operates at 70 kHz.
3, 4. Ripple before and after the inductor on the 12 V output. After the KR142EN regulator, the waveform is completely smooth, but there’s a problem: at this point, the voltage is only about 14.5 V with no load on the main outputs, and 13.6–13.8 V under load. That’s too low for a 12 V regulator.


The load tests went like this:
First I loaded one output to 50%, then the second to 50%, then increased the load on the first to 100%, and finally did the same for the second. The result was four load levels: 25-50-75-100%.
First, the high-frequency output. In my opinion, it looks pretty good: ripple is minimal, and adding an extra inductor could reduce it to almost zero.


At 100 Hz, though, things look rather bleak: the input capacitance is too low, much too low.
The peak-to-peak ripple at an output power of 500 W is about 4 V.


Load tests. Since the voltage sagged under load, I increased the load current as needed so the output power would roughly match the 125-250-375-500 W sequence.
1. Channel one: 0 W, 42.4 V; channel two: 126 W, 33.75 V
2. Channel one: 125.6 W, 32.21 V; channel two: 130 W, 32.32 V.
3. Channel one: 247.8 W, 29.86 V; channel two: 127 W, 30.64 V.
4. Channel one: 236 W, 29.44 V; channel two: 240 W, 29.58 V.

You may have noticed that in the first test, the voltage on the unloaded channel was over 40 V. This is due to voltage spikes, and since there was no load at all, the voltage gradually rose. Even a small load brought it back to normal.


I also measured power consumption, but since the output power measurement has a relatively large error, the calculated efficiency figures are approximate as well.
1. 25% load, 89.3% efficiency
2. 50% load, 91.6% efficiency
3. 75% load, 90% efficiency
4. 476 W, about 95% load, 88% efficiency
5, 6. Just out of curiosity, I measured the power factor at 50% and 100% power.

Overall, the results are roughly in line with the claimed 90%.


The tests showed that the power supply performs pretty well, and everything would be great if it weren't for the usual fly in the ointment: heat. Right at the start, I estimated the PSU's power rating at around 300-350 W.
During my usual test, gradually warming it up in 20-minute intervals, I found that at 250 W the PSU performed excellently. Component temperatures were roughly as follows:
Bridge rectifier - 71
Transistors - 66
Transformer (core) - 72
Output diodes - 75

But when I raised the power to 75% (375 W), things looked quite different after 10 minutes:
Bridge rectifier - 87
Transistors - 100
Transformer (core) - 78
Output diodes - 102 (the more heavily loaded channel)

Trying to figure out the problem, I found that the transformer windings were overheating badly. This heated up the core, reducing its saturation flux density and causing it to start saturating. As a result, the transistors heated up rapidly (I later recorded temperatures as high as 108 degrees), so I stopped the test. However, the tests at 500 W when starting from cold went fine.

Below are a couple of thermal images: the first at 25% load, the second at 75%, taken after half an hour (20+10 minutes), respectively. The winding temperature reached 146 degrees, and there was a noticeable smell of overheated varnish.


So, here are some conclusions, some of them rather disappointing.
The overall build quality is very good, but there are a few design quirks, such as mounting the transistors without insulating them from the heatsinks. The wide range of output voltages is a plus: for example, 35 V to power a power amplifier, 15 V for a preamp, and an independent 12 V supply for various auxiliary devices.

There are some circuit design shortcomings, such as no input thermistor and too little input capacitance.
The specs say the additional 15 V channels can supply up to 1 A, but in reality I wouldn't expect more than 0.5 A without extra cooling for the regulators. The 12 V channel probably won't deliver more than 200-300 mA at all.

But these issues are either not critical or easy to fix. The biggest problem is heat. The PSU can deliver up to 250-300 W continuously; 500 W is only sustainable for a relatively short time, unless you add active cooling.

While I'm at it, I have a question for the community. I'm thinking of building my own amplifier and reviewing it, too. Which would be more interesting: a power amplifier or a preamp? If a power amp, what output power, and so on? I don't really need one myself, but I'm in the mood to tinker. The PSU under review has little to do with this :)

This PSU on AliExpress - link, and another one.

That's all for now. I hope this information was useful, and as usual, I look forward to your questions in the comments.