A Rather Unusual 25/5 V Power Supply

Not long ago, one of my regular readers suggested in the comments that I review an unusual power supply. It caught my interest, so I thought, why not? It certainly stands out from the crowd of common power supplies.


First, I’d like to thank Vladimir for this interesting specimen. I haven’t come across one like it before :)

Unfortunately, this power supply is no longer available at the link. Maybe that’s temporary, or maybe it’s gone for good. The point is, you can sometimes find interesting things on Taobao for relatively little money, and even after adding shipping, they can still be a good deal.

The first thing that catches your eye is the unusual shape of the circuit board. This power supply was clearly designed to be built into some kind of device, and the board’s shape is dictated by the shape of that device’s enclosure.

The board weighs 222 grams. Its maximum length is 126 mm, minimum length 89 mm, width 82 mm, and height 62 mm.



The high-voltage section of the power supply is on one side.


The low-voltage section is on the other.


Yes, it’s made up of two circuit boards connected by headers. You may have seen something similar in my review of the Novatek power supply.
There’s nothing particularly new about this approach. It’s used when it’s difficult to fit everything on one board, but it’s still not all that common.


1, 2. The input has a full mains filter, including an inductor, X and Y capacitors, a fuse, a varistor, and a thermistor. The filter capacitor has a capacitance of 82 µF.
3. This design uses a PWM controller with an integrated high-voltage transistor, which makes the circuit even simpler than usual.
4. The output uses the classic TL431 and an optocoupler.


STR-Y6753 PWM controller


There are hardly any components on the underside.


I don’t see much point in redrawing the schematic, especially since it largely follows the datasheet. Even the output is dual-channel.


The second board is a little more interesting.
1. As I mentioned, the SMPS has two output voltages, so there are two diode assemblies, a set of capacitors, and two inductors to reduce ripple.
2. There’s also an inductor on the board whose purpose I didn’t understand at first.
3. The board also has a relay and the same type of connector as the mains input.
4. What’s more, there’s a spark gap here, too!

A separate note about the spark gap and relay.
First, the relay breaks the neutral line if you connect power to the board according to the markings.
Second, the spark gap is connected in series with a 10 kΩ resistor and sits between neutral and the board’s common output.
Third, the spark gap is connected through a separate set of contacts, so it’s connected to the board’s mains output only when the relay is energized.


The purpose of the relay and power connector became clear after some experimenting. It turned out that the first pin of the large connector sends a signal to energize the relay, which in turn switches mains power to that connector. So the device must have had another mains-powered component that was switched on by a control signal.
1. The connectors also carry 24 V and 5 V, as well as signals for switching on and dimming the LEDs, and for controlling the fan. The fan control signal, however, simply passes straight through from one connector to the other.
2. Once I spotted the LED control signals and the LED output itself, the purpose of the inductor became clear.


There are clearly more components on the underside of this board than on the high-voltage board.


1. The BIT3251 step-up PWM controller boosts the voltage, regulates the current, and adjusts the brightness of the connected LEDs. It’s used in TV backlights. On the right, you can see a pair of transistors connected in parallel and the current-sense resistors for this output.
2. The board also has a 5 V regulator. It turns out this rail is around 7.5–8 V, with the 5 V supplied by a linear regulator. The transistor to the right controls power to the fan.


All in all, this looks like a board from either a TV or a monitor, although the purpose of the switched mains output is unclear. And 5 V at such a low current would only be suitable for auxiliary circuits, standby power, and the like. So it’s quite possible that the rest of the power came from a second power supply, which would have been powered from the additional output.
Or maybe it was some kind of clever LED light fixture...

1, 2. Initial power-up: 24 V on the main output, with about 1 W of no-load power consumption.
3. The second output, before the regulator, is at 7.9 V, but it will of course rise slightly when the main output is loaded.
4. I couldn't get the LED output working properly. If I periodically apply 5 V to the LED EN input, there's a voltage spike of about 70 V on this output. But voltage is present only while pulses are being applied to the control input; apply 5 V continuously and the output shuts off. Adjusting the LED ADJ input made no difference. To be honest, I wasn't too concerned about this output, and I think it can be made to work, especially since the chip's datasheet is available.


Since none of the power supply's output voltages are brought out directly, I had to connect wires straight to the contacts on the circuit board.


The load test showed that the main output can supply up to 4.4–4.6 A; beyond that, the protection trips. It's a latching protection, so the power supply has to be disconnected from power for 1–2 minutes to reset it.
Voltage regulation is excellent. At the start of the test, the load showed 24.28 V; at 4.5 A, the voltage had risen by 10 mV.


Efficiency was measured over a load-current range of 0.5 to 4.5 A, and the results are typical: no better or worse than other power supplies.
The horizontal scale is in 0.5 A increments.


The ripple was a pleasant surprise. Only in no-load mode was the ripple large, due to the “green mode”; at 1.5, 3.0 and 4.5 A, it peaked at 35 mV—and even then, the large ripple occurred only at 150% overload.
Ripple was measured with a direct connection, i.e. without a C+C filter.


At a slower sweep rate, problems became apparent at 4.5 A. There had already been a bit of “wobble” from the start, but now the cause was clear: the power supply was obviously overloaded.
At 1.5 and 3 A, though, everything looks fine.


Because its design wasn't very convenient for a thermal test, I had to position the power supply vertically. I also measured how the output voltage changed with temperature.


The test had four stages, each lasting 20 minutes, with load currents of 1, 2, 3 and 4 A, respectively. As you can see in the screenshot, the power supply didn't trip, and the voltage remained stable.


I measured the temperature of the main components with a pyrometer. Only at the very end did the PWM controller reach 100°C, so everything looks great here.


I also took thermal images at the end of each stage. I usually include only the final stages of a test, but this time I want to show them all.

The reason is that there was a component on the board that got very hot but wasn't included in the list above. I only found it later.
1. After 20 minutes at 1 A, the image clearly shows a hot spot near the PWM controller, but the controller itself clearly can't be getting that hot at such a low power level.
2. After 20 minutes at 2 A, the components had warmed up, and the thermistor was clearly visible, but the hottest spot was still in the same place as before.
3. Low-voltage side at the end of the second stage.


1. After 20 minutes at 3 A, the maximum power for this power supply, the components had warmed up, but there was still just one very hot component, in the same place as before.
2. After 20 minutes at 4 A, everything was the same, except the temperature of the component in question had risen to 112°C.
3. Low-voltage side of the board at the end of the fourth stage; nothing alarming here.


Naturally, I wanted to find out what was getting so hot. But the thermal camera has terrible focus at close range, so it took a bit of fiddling.
1. I repositioned the power supply for a better view and overlaid the thermal and regular camera images. The hot component is clearly visible.
2. The component was standing on edge, so I bent it over slightly, making it easier to see.
3. It turned out to be a 220 pF ceramic capacitor.


Here it is in the center of the photo.


Apparently, this capacitor is effectively connected in parallel with the high-voltage transistor. I've seen this kind of arrangement before, but this much heating seems strange. The measurement may have been inaccurate because the capacitor has a glossy surface; the thermal camera showed a slightly lower temperature. But even if the actual temperature was 20 degrees lower, that's still a lot, in my opinion.


Voltage stability is excellent, both under varying loads and across temperatures. With the power supply hot and a 4 A load, the voltage was 24.256 V. After removing the load, it dropped by 2 mV, and after cooling down, still with no load, it rose by 13 mV.
I think the main reason is that the feedback divider is located quite far from the high-temperature components.


The person who sent me this power supply wrote that one of the reviews included instructions for making it adjustable.


The suggested modification is to change several resistors in the feedback circuit and add a potentiometer for adjustment. According to the author of the modification, this should give an output voltage range of 10.3–32 V.


To be honest, I'm skeptical of modifications like this, since the power supply would then be operating outside its optimal range. The upper limit is constrained by the reliability of the circuit containing the high-voltage transistor and snubber, while the lower limit is set by the PWM controller losing power and dropping out.

There's no point testing the upper limit, but we can try the lower one. I won't modify the whole circuit; I'll just change the value of the divider's upper resistor.

1. I connected a 22 kΩ potentiometer in parallel with the 20 kΩ upper resistor, with a 1 kΩ fixed resistor in series. When I switched it on, I got about 14 V.
2. Reducing the potentiometer's resistance brought the voltage down to 10 V without any problems. Below that, the power supply started restarting. I ran this test with a low-power bulb connected to the LED output, which has the same 24 V through a diode when inactive.
3. I raised the voltage slightly; at 10 V, the power supply started working stably.
4. But as soon as I removed the load, the voltage rose by 0.4 V.

The result was fairly predictable: without special measures, the power supply won't operate quite correctly in modes like these, since changing the voltage by more than +/-10–20% is already a lot.



Conclusions.
This is definitely an interesting board, especially for the price. Its main appeal isn't the board itself, but its potential use in some kind of “smart” devices: it provides not only 24 V power, but also a low-voltage output, an LED driver, fan control, and a controllable power output.
The board delivers the stated specifications. The 8 V output (before the regulator) can clearly supply more than the specified 0.5 A as well.

But there are two things I don't like. First, the design is rather bulky; second, there are connectors in the feedback circuit. Yes, the contacts are duplicated, but if you power up only the high-voltage board, there will be no voltage at the feedback divider input, and the power supply could fail.

The board is currently out of stock, but for some reason I'm sure it'll show up again—if not from this seller, then from another. This power supply is also a good example of how, if you look around, you can find interesting things on Taobao for relatively little money, even with shipping included.

That's all from me, and thanks again for this interesting power supply.