Sanmim GPA40B power supply, 12 V 40 W

I left this power supply, as they say, “for dessert,” partly because it stands out from the rest of the units shown earlier, and partly because I simply liked it more than the others. In any case, that’s not particularly important, and today I’ll tell you about a power supply that needs no modifications.


First, let me mention a problem I’ve run into more than once. If you read the review title carefully, you’ll see it says 12 volts, but follow the link and you’ll find a power supply with a completely different output voltage. And yes, I wanted the version shown at the link; the store doesn’t carry a 12 V version at all.
I’ll write to the store manager, of course, because in my view, mistakes like this not only cause confusion but also hurt both customers who receive the wrong item and the store, which has to refund them for an item sent by mistake.


The power supply comes in its own box, which offers at least a little protection during shipping. There’s also a sticker with the part number, so the manager can see what I received and what I ordered.


This isn’t my first review of this company’s power supplies, and I’ve mentioned before that, in my view, just as a theater begins at the cloakroom, a reputable company should provide a datasheet for its product. The only problem is that it’s in Chinese :(


And here’s a spot-the-difference challenge for the sharp-eyed: find the ten differences, as they say. This one isn’t the store’s fault.
Actually, it’s simple: both photos really do show an SM-GPA40B power supply, but from different revisions. The one on the left is what you usually see on store pages and even in the datasheet; the one on the right is what I received, and what’s shown on the manufacturer’s website.
Well, that’s how the cookie crumbles...

There are differences. For example, one unit has a varistor, while the other has a second dual-winding inductor in its place. The second version also lacks the additional Y capacitors, but has two interwinding capacitors instead. Anyway, I’m getting ahead of myself.


So, we have a datasheet in Chinese, from which we can tell that this is a 40 W power supply. I only found out it was a 12 V model a little later, after checking the markings on the transformer and output capacitors.
Current up to 3.5 A, efficiency up to 86%, ripple below 50 mV, operating voltage range 85–265 V.


It really does look very neat. Not only is everything soldered straight, but the board itself has also been cleaned. Honestly, it looks just as good as, or maybe even better than, the Mean Well units.



The power supply uses a 2 × 4-inch form factor (51 × 101.6 mm), equivalent to the size of two matchboxes placed end to end, since a matchbox is almost 2 inches (50.5 mm) long.


I call it a form factor for a reason: other companies make power supplies in this size too. For example, to the right of the unit under review are two Mean Well supplies in exactly the same size, and farther to the right, for comparison, is the “people’s choice.”
Below are some other power supplies from Sanmim.


Here are all the dimensions on the drawing. It’s for a board from a different revision, but the board dimensions and the spacing between the mounting holes are the same.


Now let’s take a closer look.
The power supply has an almost complete input filter: a fuse, a pair of inductors, a thermistor, and an X capacitor. The only thing missing is the varistor, which was fitted on the board from the other revision.
There are pads for Y capacitors, but they’re unpopulated. Power is connected through a connector.


1. The input capacitor is rated at 82 µF. To its right, you can see the PWM controller’s supply capacitor and an inductor, which is unusual, but I’ll get to that later.
2. A transistor on a small heatsink. The transistor has an insulated package, and there’s no thermal paste, but a ferrite bead is fitted on the middle lead and secured with glue.
3. There are two Y capacitors connected in series, which should significantly improve safety.
4. There are two feedback optocouplers: one for voltage feedback and a completely independent one for overvoltage protection.


Four filter capacitors are installed on the output: a pair of 1000 µF, 16 V capacitors before the inductor and a pair of 470 µF, 25 V capacitors after it. The rectifier diode assembly, like the high-voltage transistor, is in an insulated package. This is done because both heatsinks are connected to power circuits: the transistor heatsink is connected to the negative rail of the high-voltage section, while the diode assembly heatsink is connected to output ground.


All the capacitors are made by Aishi, and the high-voltage one is also below its rated capacitance, while the output capacitors are all fine.
I don't know why. They seem like a decent company, but still...


The board is clean and the assembly is neat. The clearance between the primary and secondary sides is sufficient; the only thing you could really fault is the lack of slots under the optocouplers.


The PWM controller is an OB2269; nothing new here. To its right you can see a thermistor, which is essentially located almost directly under the high-voltage transistor.


The output section is also the same as before: a TL431 in the voltage regulation circuit, and just a zener diode and transistor for overvoltage protection. They did add a small ceramic capacitor across the output, though :)


The circuit almost completely matches that of the GPT70A-12 power supply, right down to the component reference designators. There are, of course, a few differences—not just in the numbering, but in the component values too—but overall they're very similar.
The most significant difference is the LC filter added to power the PWM controller. What's more, by some coincidence, this is the third power supply in a row where I've found two capacitors in the PWM supply circuit, isolated from each other by a diode whose purpose here remains a mystery to me.


The output voltage isn't adjustable and is slightly higher than the stated 12 V. No-load power consumption is about 0.3 W, which indicates good efficiency. The power supply also makes none of the extraneous noises typical of the controller operating in "green" mode.


Next come the various tests, starting, of course, with a check of its load capacity.
The results are similar to those for the GPT70A-12: the protection kicked in at 5.7 A, which is 70 W against the rated 40 W.
The voltage remained stable right up until the overload shutdown.


The stated efficiency was 86%, and the power supply reached that figure at close to maximum power. Under overload, it got up to 87%.
Below is the graph for load currents from 0.5 to 5 A; each horizontal-axis division corresponds to 0.5 A.


Honestly, this power supply had already surprised me during the initial tests. I couldn't resist taking a photo while measuring efficiency: with no load, the output was 12.083 V; at 5 A—150% of the rated current—it changed by just 1 mV.


During the thermal test, the power supply seemed not to understand what I wanted from it at all. First, I loaded it to 50% and 100% for 20 minutes each. Even at 100%, the hottest spot—the high-voltage transistor—reached only 73°C.


Then I ran it for another 20 minutes at 125% of its rated power. When it handled that just fine, I added another 20 minutes at 150% load. Only at one-and-a-half times the rated load did I see relatively high temperatures—or rather, a high temperature, again on the high-voltage transistor: 93°C. For a semiconductor, that's still considered perfectly normal, broadly speaking.


All in all, the power supply passed the test with flying colors. Even at a 60 W load, none of the components exceeded the arbitrary 100°C threshold. The thermal protection stayed silent too. I later thought I should probably have kept heating it at higher currents, but the current was already close to the overload protection trip point.


As usual, I checked how the output voltage varied with both load current and temperature.
1. Cold power supply, no load
2. At the end of the test, at 20 W
3. At 40 W
4. At 50 W
5. End of the 60 W test
6. Hot power supply, no load

Overall, I'd give its regulation an excellent rating. The output voltage varied by just +/-2 mV throughout the test, i.e. In practice, it was unaffected by the load, the temperature, or both factors at once.


The specifications state that the power supply's ripple is no more than 50 mV. I wouldn't call that particularly low, but it's worth noting that Mean Well often specifies even higher ripple.

To minimize its effect on the test results, the load ran autonomously and was powered by a 50 Hz transformer.
Here you can see the test at 5 A.


1. With no load, there are small spikes at around 2.5 kHz. The “green mode” is active, but as I mentioned earlier, there was no noise.
2, 3. At 50% and 100% load, the main ripple component was 5–6 mV; including the short spikes, it was 30 mV.
4. At 150% load, the main component was unchanged, while the switching ripple rose to 35–37 mV.
5, 6. At 100% and 150% load, with a slower time base, the power supply still easily stayed within the specified 50 mV—even at 150% load.



I was about to wrap up the tests, but curiosity got the better of me: how would the power supply behave at a lower input voltage? So I hooked up a Variac and continued experimenting.

First, let's set the input to 115 (117) V.


The only change in its behavior was that the protection kicked in a little earlier, at 5.2 A instead of 5.7 A.


All right, let's try the minimum specified input voltage: 85 V.


Again, the only difference was the protection trip current. It shut down the power supply at 4.8 A (58 W), while the output voltage remained rock-solid right up until the overload protection kicked in.


I could have run temperature tests at these input voltages too, but they're extremely rare in practice. Besides, there are plenty of restrictions in this mode. For example, at an 85 V input, the output power must be reduced to 70% of its rated value. The graph even shows operation down to 70 V, but at just 50% power.
By the way, similar restrictions apply at higher input voltages. They're shown in the graph on the left; the graph on the right shows the same thing for DC input.

The manufacturer also provides a graph of maximum power versus temperature, which indicates that output power should be reduced linearly to zero over the temperature range from +55 to 90.



This is probably one of the few power supplies that's genuinely hard to fault, since it has:
1. High-quality construction. It's not TracoPower, of course, but in my opinion it's certainly no worse than Mean Well.
2. Excellent voltage regulation. In fact, the variation I measured across the entire load and warm-up range was within the last digit on my multimeter's display.
3. In an open-frame setup, it easily handled 150% load continuously.
4. Ripple below the specified limit, even at 150% load.
5. Protection against overload/short circuit, overheating of the primary side, and output overvoltage.

One of the few “flies in the ointment” is the input capacitor, whose capacitance is lower than marked. That's a little odd given the overall results. This is essentially the main difference between this power supply and Mean Well: Mean Well uses components from better-known manufacturers, typically Rubycon, Nippon, or Lelon capacitors.

In short, I'd put it this way: if you have “spare” money and don't want to add or replace components, install heatsinks, and so on, this is a good choice—as is the Mean Well mentioned above, though Mean Well is still a bit better. If you'd rather save money and have the time, opportunity, and inclination to tinker, it's better to choose something from the “people's choice” range, since the unit reviewed here definitely doesn't qualify as “budget.”


That's all from me. And, by the way, there's another very interesting power supply waiting to be tested—also from Mean Well :)