Another power supply, this time 48 V, 0.5 (1) A

As I wrote in the previous review, I got several different power supplies, and today I'm reviewing the next one. Last time it was a 12 V power supply, but today's model has a less common output voltage of 48 V, while still having its own area of application, which I'll discuss at the end of the review.
As usual, there are tests, schematics, and answers to some questions people asked in the comments.


I've already talked about how much I enjoy tinkering with different power supplies, so I'll get straight to the review.

I also ordered a lot of three this time, and, just like with the first power supply, received three separate large zip-lock bags.


According to the product page, this power supply is rated at 48 V, 1 A, which should give a total output power of 48 W.
The word “Disassembled” has disappeared from the name, but “LED” has appeared instead.


It looks very neat and is quite compact.


This power supply measures 75 × 38 × 25 mm. It's certainly larger than the previous one, but not by much. The previous model was rated at 12 W, while this one is rated at 48 W. We'll check that claim later.


A heatsink is mounted along one of the long sides of the board, and it's connected to the negative side of the power supply's “hot” side. There's insulation near the “cold” side, and the heatsink itself is about 8–9 mm shorter than it might look in the photo at first glance. In other words, the insulation extends well beyond the edge of the heatsink.


A couple more general views of the board. I just really like how it looks :)



A closer look at the components.
1. The input has a 2 A glass-body fuse and a small thermistor, but no varistor. There's also an X-class interference-suppression capacitor.
2. The input filter consists of an X-class capacitor and a common-mode choke. The bridge rectifier is made from discrete diodes.
3. The input capacitor has a capacitance of 47 µF. That's cutting it very close for 48 W, but more on that later.
4. I've taken apart a lot of power supplies, but this is the first time I've seen a ceramic capacitor in parallel with the input electrolytic capacitor. I may have come across this before, but I'm not sure.


1. A high-voltage transistor in an insulated package, marked STK0465. I haven't checked the datasheet, but the part number suggests it's rated for 4 A and 650 V. The mounting hardware is also covered in varnish. I tried to remove it, but was afraid I'd just tear the heatsink off, so I gave up—especially since I already knew the transistor type.
2. The transformer is quite compact, thanks to the use of a non-standard core rather than the usual E-shaped one. You can see the windings coated with varnish inside.
3. The PWM controller's supply capacitor is located nearby.
4. An X-class capacitor connecting the power supply's “hot” and “cold” sides is tucked between the heatsink and transformer. I didn't try to get to it because removing the heatsink would be difficult, but I could read its value: 2.2 nF.


1. Interestingly, the feedback circuit uses a PS2561A optocoupler instead of the usual PC817, though I don't see much difference. You can also see that the board was clearly designed to accommodate two types of output diode: the one shown in the photo and a TO-220 package. The latter would most likely have used a heatsink. That makes sense, since the output diode is mainly limited by the output current, which isn't very high here.
The output has two 220 µF, 63 V capacitors connected in parallel. All the capacitors in the power supply are made by Nichicon: the input capacitor is from the KXG series, and the others are KY series.
The output filter uses a two-winding common-mode choke. Also worth noting is the P6KE51A Zener diode in the output circuit, which provides additional protection for the load.


These boards are clearly newer this time, judging by the 2012–2013 date code. I couldn't find out anything else, so I'll have to determine the specs experimentally.


I'd rate the soldering quality as average. There are some flaws, and a few spots are rather untidy.


The input section of the power supply and the PWM controller. The PWM controller's marking is very hard to read (LzP32), so when drawing up the schematic I simply found the closest equivalent based on the pin functions. The diode in the PWM controller's supply circuit is completely unmarked, on all three boards.


Along with the usual components, the output circuit also includes a Zener diode. The thing is, the TL431 adjustable shunt regulator has a maximum voltage rating of 37 V. Even with the optocoupler taken into account, that only gets us to 40 V, while the power supply’s output voltage is 48 V. So in cases like this, a Zener diode is connected in series with the optocoupler—in this case, a 24 V one. It “clips” the voltage to a safe level. It’s marked ZD2 in the photo.


I have almost no questions about the power supply’s circuit design. Unlike the previous PSU, this one uses a separate PWM controller and a powerful high-voltage transistor. This approach has both advantages and disadvantages.
The advantage is that the PSU’s power rating is almost independent of the type of PWM controller.
The disadvantage is that it’s harder to implement overheat protection.

At the input, there’s a discharge chain of three 1.5 MΩ resistors that discharges the CX capacitor. The familiar feedback circuit, with a Zener diode added.
There are also a couple of minor details:
1. The exact IC type is unknown, but the closest equivalent is the FAN6862, which has a temperature-sensing input for an external sensor. In the PSU under review, this input is used for overvoltage protection. If the feedback loop is disconnected for some reason, the PWM controller’s supply voltage will rise, and so will the voltage at pin 3 of the PWM controller. As a result, the PWM controller will start limiting the output voltage. At least, that’s clearly how it was intended to work.
2. There’s an unpopulated spot on the board for a thermistor, marked NTC2. At first I thought it was meant to be part of an overheat protection circuit, but the way it’s connected (and the thermistor’s location) is rather unusual, so I didn’t quite understand the purpose of this design. This circuit is shown in red in the schematic.


Let’s move on to the tests.
As I mentioned, the store page claimed that the PSU has an output voltage of 48 V at currents up to 1 A. Last time, the product title hinted at a different current rating, but here 1 A is stated everywhere.


All the connections were made just as with the previous PSU; the only difference was the electronic load. Testing this power supply was a bit tricky because its output current and power aren’t very high, but with an output voltage of 48 V, I couldn’t use the load shown in the previous review. I had to use a more powerful one, but it was also less precise.
At no load, the output voltage is slightly below the stated value, but in practice this is absolutely no cause for concern, since the difference is less than 0.1% :)


Actually, when I picked up the board, my first thought was that this was clearly a power supply rated at around 25 W. That was based on the following observations:
1. The board dimensions
2. The input capacitor’s capacitance
3. The transformer’s size.

Of course, the transformer’s size depends on the converter’s operating frequency, but since the frequency is usually in the 66–133 kHz range (more often 100–133 kHz), the difference in size isn’t all that great. There are exceptions, of course, but not in this case—the circuit design was clear at first glance.

So I tried to run the test carefully, even though I had two more “spares” on hand.
The photos below show loads of 200, 400, 600, 800, 1000, and 1050 mA.
The last value was chosen for a reason: at 1.1 A, the PSU reliably trips its protection and shuts off the output. Once the load is removed, it returns to normal operation.
So, it does deliver 1 A, even for several minutes in a row :) I didn’t test it for longer, since that wasn’t the goal at this stage.

The last two photos might make it look as if something is wrong with the output voltage, but everything is fine. As the PSU warms up, its output voltage rises slightly. Since the last two test stages didn’t happen instantly, the increase is more noticeable than in the first four steps.


I’d rate the high-frequency ripple as very low: 40 mV at full load on a PSU with a 48 V output is less than 0.1%.


In the previous review, someone asked me to check the ripple level at 100 Hz. I decided the information could indeed be useful, so I measured that too.
The waveforms were captured at load currents of 200, 300, 400, 500, 600, and 700 mA. The largest ripple amplitude occurs at 300–500 mA (15–25 W), although I expected it to increase further as the current went up.


But as always, a thermal soak test gives a more accurate picture of the power supply’s actual power capability.
The procedure is standard for my reviews: each step lasts 20 minutes, at 200, 400, 600, and 700 mA. The load current for the final step was chosen based on the temperature measurements from the previous step.
I noticed that the output voltage rises as the PSU warms up; this will be visible in the table. At the very end, I suddenly removed the load and checked the no-load voltage while the PSU was still warm.
Before the thermal soak on the left, after on the right. In reality, the difference turned out to be smaller than I expected, and the voltage essentially reached the specified value.
In any case, voltage regulation accuracy and thermal stability are both quite good.


The heating situation is a little ambiguous. For some reason, I initially expected the transformer to overheat, but it turned out I was wrong: the output diode was the first to reach the finish line. It’s worth mentioning that the output uses a regular fast-recovery diode, not a Schottky diode, since Schottky diodes are rarely used at these voltages. I think replacing the output diode with a faster one could increase the continuous power a little further.
In any case, I can already say that this power supply’s actual continuous power is around 25 W, just as I thought at the very beginning, although it can briefly deliver about 45–50 W.


Thermal images from two angles. You can also see here that all the heat is concentrated around the output diode.


I was also asked to test the power supply’s input response to an impulse disturbance. I should say right away, though, that I’m quite skeptical about this test for several reasons:
1. The test conditions are not standardized.
2. The power supply’s input filter prevents interference from the power supply from getting into the electrical grid, but more on that below.

The test “setup” was assembled according to the diagram below. The principle is very simple: when you press the switch button, the contacts briefly break apart, with the upper contact already open and the lower one not yet closed. Since an inductor is connected as the load, this generates an impulse disturbance.


In reality, it looked much more intimidating. I used a 60 W transformer and wanted to find an 80 W choke for a fluorescent lamp, but I’d apparently either thrown it out or put it somewhere else. The latter is more likely.


Then came the test. I spent a long time repeatedly pressing the button, and sometimes the interference could be heard through the computer speakers sitting nearby. But the power supply itself apparently didn’t quite get what I was trying to do, since I managed to capture the disturbance only about once every 30–50 button presses. And since the disturbance is generated twice (when pressing and releasing the button), that worked out to one trigger for every 60–100 pulses.
In the end, I did manage to capture an output spike a few times. The maximum peak-to-peak amplitude was about 1 V, which is only 2% of the output voltage for a 48 V power supply.


Since this is the second power supply among my recent “test subjects,” I decided to run the same test on the previous one as well.
The “setup” is almost the same as above; I only replaced the electronic load with the one I used last time.


The results here are roughly similar. I again spent a long time trying to generate interference and eventually got several output spikes with an amplitude of about 0.2–0.3 V. Given the 12 V output voltage, that’s almost the same 1.5–2.5% as in the test above (the first two screenshots).

I also checked the ripple level at 100 Hz. Everything was excellent up to 900 mA (the second-to-last screenshot), but at a current of 1 A the ripple increased sharply and the power supply’s protection kicked in.


A few more words on why I was skeptical about the impulse-disturbance test.
First, let’s imagine a simplified version of a household electrical grid. If we assume the disturbance (say, from a refrigerator) is generated at point 2 and our power supply is at point 1, we could get significant interference at its input (Sinclair users will remember). But if we move the power supply to point 4, the interference level will drop many times over, because along the way it will encounter:
1. Wires that are definitely not made of superconductors
2. Circuit breakers, whose current coils are small inductors, even if only small ones.
3. A load, for example at point 3. This could be an ordinary heater (a resistive load) or a computer power supply (a capacitive load).

In other words, you can’t tackle the problem head-on, because the grid is a fairly complex and somewhat inert load. That’s why I see the transformer test more as a “spherical generator in a vacuum.” No, of course it does have some value, but in my opinion it’s only a limited one, since input filters also vary.


In general, the electrical grid is constantly subject to impulse spikes from refrigerators, sparking contacts, and high-power loads (especially inductive ones). But natural factors have a greater effect, such as thunderstorms or wires crossing on utility poles. Again, the latter is much less common in cities, since power cables are laid underground. However, in places with poor conditions, it is recommended to use a surge protection device (SPD), essentially a large varistor.


Now, about the power supply's input filter itself. First, it's worth understanding that these filters can serve different purposes.
In the simplest version (apart from specially designed jumpers), there's an X capacitor and a common-mode choke. In this configuration, the filter protects the electrical grid from interference generated by the power supply. In other words, interference coming from the power supply is first attenuated by the choke, then essentially shorted out by a dedicated capacitor. Switching power supplies generate a lot of interference on the electrical grid, especially when the manufacturer has cut corners everywhere.
If interference needs to be attenuated in both directions, two capacitors are used, one before and one after the choke. In this case, the filter also attenuates interference coming from the grid that could otherwise get into the power supply.
Of course, some interference is also attenuated by the input capacitor after the bridge rectifier, but a dedicated capacitor immediately after the choke is more effective specifically against impulse noise.

This is all very simplified, of course, but I tried to explain it in simple terms.


But that's not all. Above, I wrote about interference that comes in over the mains wires between live and neutral. There is also interference relative to ground, so additional capacitors are placed after the choke. Since a failure of these capacitors could be dangerous, they are Y-type capacitors.


In general, a power supply's input filter can be a fairly complex device, consisting of a whole bunch of capacitors and chokes. The schematic below uses two-winding chokes of two types.

So, why am I telling you all this? Simply because it's important to understand what we're doing and why. If we're designing equipment where output interference could be critical, we use the full filter. If all we need is for our power supply not to "pollute" the grid, an X capacitor and a choke are quite enough. In our case, we have the second type of filter, nothing more.


Now, why would you need power supplies with such an "unusual" voltage in the first place? The photo below shows two boxes connected by a cable, with one box powered by the other.


One box is a power supply, and the other is a step-down converter. Together, they form a passive PoE equivalent, intended to power low-voltage devices over a local network cable. I put this setup together "in a hurry," so to speak, when I needed to power a router located 10 meters from the nearest outlet and had only a LAN cable available.


You can see in the photo that even the box is slightly melted. It was left over from some experiments, and I felt bad about throwing it away.
I usually made power supplies for 48-55 V, but this one is 24 V.


While the power supply shown above differs only in voltage, I did things quite differently at the remote end. Since everything was inside the apartment in this case, I used a simple DC-DC converter at the remote end.
In other cases, though, I made converters with galvanic isolation and PWM controllers such as TOP414 or DPA-Switch, depending on the required power. TOP412-414 was cheaper and simpler, but its power was limited to around 15 W, if I remember correctly. DPA-Switch is noticeably more powerful, with some models providing up to 100 W.
The setup shown here is the simplest thing I've made along these lines.

By the way, you can see a bulging capacitor in the photo, the notorious Capxon, although I really like their KF series. The converter shown below (like the power supply) is over 10 years old, and the capacitor was still fine when I removed it. The converter outputs 5 V at up to 3 A.


That's probably all; let's sum things up.
These power supplies are clearly newer than the ones shown earlier. I also got the impression that they hadn't seen much use. They have branded capacitors, a quality transformer, input and output filters, and, unexpectedly, even a ceramic capacitor in parallel with the input electrolytic capacitor.
The specs are decent. The power supply can really deliver around 25 W continuously and up to 45-50 W for short periods, which is quite good for its size. I think that if you get the 24 V version, you could easily make a compact power supply for a TS100 soldering iron (maybe I should get one and give it a try).

The store page lists an output current of 1 A, but doesn't say that this is a short-term current. So, just like last time, the short-term rating was presented as a continuous one.
Otherwise, I liked the power supply, even though it was used, which of course adds a certain "element of surprise". Unfortunately, I can only speak for the three units I have here.

As always, I’m looking forward to your questions and comments. I hope you found the review useful.

A small addendum.
In the review, I wrote that I’d tried to find a choke from a fluorescent lamp but couldn’t. But as often happens, you look for one thing and something completely different turns up—and sometimes it’s even more interesting.
That’s what happened this time: I came across a Celeron 266/66 processor and couldn’t resist taking a photo :)
I have a feeling that next time I’m searching, I’ll find a motherboard to go with it.


It’ll be 20 years old this year. Time flies so fast—the store where it was bought is long gone.