60 V 5 A Power Supply and a Few Notes on Modifying It

By chance, another power supply landed in my hands for a while—or, more precisely, one of the versions of a model I had already reviewed once. As it turned out, though, I had to make a few modifications before testing it. As always, the details are in the review.

As usual, a reminder: to keep up with new posts on the blog, I recommend following me on Instagram. I’ll post notifications about every new topic there, and might also write about something interesting from time to time — link to my account

First, let me explain how this power supply ended up with me. A friend bought it for a DPS5005 converter, so he needs a power supply with an output voltage of around 55 V and current up to 5 A. Since the output voltage isn’t adjustable, it has to be changed at the hardware level.

The power supply was bought on Taobao, but similar ones are also available on AliExpress, so I thought a description of the modifications might be useful too.


It looks pretty good overall. There are a few flaws, of course, but nothing I’d call a deal-breaker.


There’s a large heatsink on the output side, but it’s clearly designed for active cooling. Even so, it works a little better than a solid block of aluminum in this power supply.


For comparison, here’s the previous model. I didn’t bother digging it out of the boxes, so this photo is from the previous review.
As you can already see, the key difference is that the PWM controller and its surrounding components are on a separate board here, whereas in the unit under review they’re on the main board.


The input also has an almost complete mains filter, which is perhaps missing only a varistor.
The input capacitor is 220 µF, 450 V.


1. Terminal block—awkward and not to my liking. The fuse is next to it.
2. GBU808 bridge rectifier.
3. 5D-15 thermistor; a pair of high-voltage capacitors is visible next to it.
4. Pay attention here: the capacitors are connected to one of the mounting holes. In use, this hole must either be grounded or have no electrical contact with the case at all!


The layout is almost identical to the previous model, and yes, the transformer core is really crooked.


The PWM controller’s markings have been erased. It’s also interesting that there are two electrolytic capacitors on the high-voltage side, rather than one. To get ahead of myself, the power supply takes a very long time to start—several seconds after power is applied.
There are also components under the transformer, and a pair of resistors immediately catches the eye. They’re presumably in the PWM controller’s power supply circuit.


The power supply uses a single-ended forward topology, so there’s a storage inductor next to the transformer.


1. 17N80C3 transistors, two connected in parallel.
2. Y capacitor between the primary and secondary sides.
3. There are two diode assemblies in parallel at the output, but more on that later.
4. The output has three 470 µF, 63 V capacitors, with no inductor to reduce ripple.


There are no components on the underside. The board is relatively clean and has safety slots, so everything looks fine here.


First power-up, and the first oddities.
To start with, I was surprised to see the output voltage gradually rise from about 57.5 V (the photo was taken a little later) to 60.2 V. But that’s not all: the power drawn from the mains was around 10–11 W.
At first I thought there might be a problem with the electrolytic capacitors (I’ve seen that happen), but even then, with no load, the power supply should bring the output voltage up immediately.


But as we know, if a device is consuming power with nothing connected to its output, all that energy has to be dissipated somewhere. Below are photos taken a few minutes after power-up, then another 10 minutes later, and finally after half an hour of warm-up.
It turns out the diode assembly heats up to around 83–85 °C, even with no load. The heatsink is too hot to hold.


I immediately understood what was causing the heating, but first I decided to fix the voltage drift, adjust the voltage, and confirm that my theory about the heating was correct.

As expected, the power supply has a feedback circuit and a voltage divider. The upper divider resistor is marked in blue, the lower one in green, and the pair of adjustment resistors in red. They are connected in parallel with each other, but in series with the lower divider resistor.
There is also a footprint on the board for a trimmer potentiometer, but replacing the adjustment resistors is much easier.
My calculations showed that replacing these 750 Ω resistors with 1.5 kΩ ones would get me close to the target.

Also, the adjustment resistors are clearly ordinary resistors, not precision ones, which explains why the voltage drifts as the unit warms up.
I've often seen people mention sorting through a batch of ordinary resistors to find ones with the desired resistance, then using them in precision circuits. But precision resistors don't just have a smaller tolerance; they also have a lower TCR, meaning their resistance varies less with temperature. That's exactly the effect we're seeing here.


To show what I mean, here's the section of the circuit responsible for feedback; the colors match the ones marked above.
I'll also mention right away that, ideally, the voltage on the TL431 output should be limited.



After replacing the adjustment resistors, the voltage dropped to about 53.5 V. More importantly, it now barely changes as the unit warms up: the first photo was taken just after power-up, the second after half an hour of warm-up. The original drift was about 2.7 V; after replacing the resistors with proper ones, it fell to 30 mV—almost a 100-fold improvement!
The lower voltage also reduced power consumption, which further confirmed my theory.


Now for the second problem: the output diode assemblies were getting too hot.
After removing the heatsink, I found that the manufacturer had installed MBRF20200CT assemblies, rated for 20 A and 200 V, which is far too little for a power supply like this. In this case, though, the diode doesn't break down; instead, its reverse current simply increases, heating up the assembly.
To fix this, the diode assembly needs to be replaced with one rated for a higher voltage. It also needs to be fast (Ultrafast, Hyperfast).

I initially wanted to use assemblies in the same package, but it turned out to be difficult to find suitable ones for sale, so I had to order 60CPH03 in a TO-247 package.
This assembly is not insulated, so I used a piece of mica.
Of course, many people will say it would have been better to mount the assembly lower down, where it could dissipate heat more effectively. But the TO-247 package is thicker, and the transformer is very close, so it might have got in the way.


Next, I marked the positions of the old diode leads and made new leads to fit the existing holes. A quick explanation: a single-ended forward converter effectively has two output diodes—one from the transformer output, the other to the output common. Their cathodes are connected together and to the output inductor.
In this particular power supply, each diode assembly functions as those two diodes, and the assemblies themselves are connected in parallel. So removing either one won't affect operation. But some power supplies use each assembly as essentially a single diode; sometimes they're even rated for different voltages, so be careful.

I powered it up, and everything works normally. Power consumption dropped from 8 W to 3 W.


A comparison: both photos were taken 30 minutes after power-up, with no load. The left one is before the modification, the right one after.


Now I can get started with the tests. I'll begin by checking the load regulation and how accurately it holds the voltage.


Overall, everything looks great. In my view, the voltage drifts more as the unit warms up than it does under load. The results shown are with no load, at 2.5 A and 5 A, and then with no load again.


The power supply also handled a 6 A load just fine. I didn't push it any further because it isn't mine and I still have to return it, though I think it would probably have been fine in terms of overload protection.
I wouldn't call its efficiency particularly high, though. I got about 86–87% at peak; the scale along the bottom is in 0.5 A increments.


Output ripple was measured using a filter made from 1 + 0.1 µF capacitors, with no load and at currents of 1.5, 3, and 5 A.
Considering the output voltage, the results are actually pretty good. With the probe connected directly, there were of course some spikes, but they'll be attenuated by the wires to the load.


The low-frequency ripple isn't particularly concerning either. By the way, the power supply has an input voltage range of 170–260 V, so it's a “narrow-range” model.


Even at the initial inspection, I wasn't too impressed with the output capacitors, so I decided to try replacing them. Unfortunately, the capacitors shown in the photo below were the best I could find in the same store where I bought the diode assembly.
I chose the same capacitance and voltage rating as the originals. I wanted larger ones, but they simply wouldn’t fit on the board.


Even so, their ESR is 33 mΩ, compared with 56 mΩ for the originals.


To keep the comparison results brief, I’ll show only the tests at maximum load current. In every case, the left side is before replacing the capacitors, and the right side is after.
1, 2. Probe connected directly
3, 4. Probe with a filter
5, 6. Low-frequency ripple.

As you can see, the new capacitors had almost no effect on the short pulses, but reduced the main ripple by about half. That roughly matches the difference in ESR between the old and new capacitors.
The replacement also had no effect on the low-frequency ripple, which makes sense, especially since the new capacitors have slightly lower capacitance.


I checked the temperature in three stages: first at currents of 1.5 and 3 A, and then, after the power supply passed those tests, at 4.5 A. Each stage lasted 20 minutes, with the board simply resting on the table, heatsinks facing up.

The situation here is practically identical to the 24 V version: the output inductor gets extremely hot. I’ll also note the high thermistor temperature. I didn’t measure it last time, but I suspect it would have been the same.


Thermal images after the 1.5 and 3 A stages.


And this is after the 4.5 A run. Because the thermistor was so hot, I had to take two separate photos so the hottest spot wouldn’t shift focus onto it.
Apparently, I couldn’t measure the temperature of the thermistor and inductor accurately with the pyrometer: the first was around 175 degrees, the second 160. And the smell—there’s no way to describe it. Overheated Chinese varnish; that smell lingered throughout the room for a long time :)

The output diode assembly, on the other hand, works just fine, which means it was correctly selected both for current and switching speed.


Conclusions.
Well, the results are essentially the same as before: the power supply can run at full power only with forced cooling. In both cases, the maximum continuous power was around 250 W.

What I changed:
1. Replaced the feedback compensation resistors for two reasons: to change the output voltage and improve thermal stability.
2. Replaced the output diode assembly because the original one was rated for too low a voltage, causing excessive self-heating.
3. Replaced the output capacitors, reducing ripple by half. It would be good to install something from a reputable brand, but I couldn’t get any quickly.

What else should be done:
1. Add a zener diode to the optocoupler circuit, since the TL431’s maximum voltage is only 37 V, while the output is currently 53.5 V. This isn’t critical during normal voltage regulation, but transients could cause problems.
2. Replace the output inductor; judging by the tests, it’s overloaded.

Otherwise, I got the impression that this was originally a decent 24 V, 150–180 W power supply that was later “pushed” to 300 W. Then they started raising the output voltage, first releasing a 36 V version, then 48 V, and finally 60 V.
At this point, the modifications have cost around $2.5, which is a quarter of the power supply’s original price. In my opinion, that’s quite a lot for a finished product.

That’s about it for now. I’m not sure what else to add; I hope this was useful :)