SANMIM GFA15A-5 power supply, 5 V, 15 W

Not long ago, Banggood sent me several different power supplies. They differ in almost every way: voltage, power, and manufacturer. After giving it some thought, I decided to test them and publish reviews in order of increasing power. Today, I’m reviewing the first one.
This power supply isn’t very powerful, and it’s compact, but I hope it’ll be interesting. Of course, you can read all about it in the review.


To be fair, I should probably mention that this is actually the fourth power supply from Sanmim (not Sanmin, as the product page says) that I’ve had my hands on. After testing the first three, I called the company something of a less well-known equivalent to Meanwell—perhaps a little cheaper, but with a narrower product range.

But let’s get on with the review.
The individual packaging is just a regular resealable bag, with only a sticker showing the part number as an identifying mark.


One important difference from the vast majority of “no-name” power supplies is that this one comes with a manual—though unfortunately, it’s only in Chinese.


The power supply under review is rated for an output voltage of 5 V and a current of 3 A. This model is part of a range of 15 W power supplies that also includes versions with output voltages of 9, 12, 15, 24, and 36 V. Because of its low output voltage, the 5 V version has the lowest claimed efficiency, just 80%; the others range from 83% to 86%.
The claimed output ripple for all these power supplies is no more than 50 mV p-p, which is pretty good. Naturally, I’ll check that myself.


It looks very neat, and I’d say it’s even tidier in some ways than well-known brands’ products. There’s no mess of hot glue, and the overall assembly looks more refined, if that’s the right word.


This power supply belongs to the low-profile series, which also includes the more powerful GFA20A-xx models rated at 20 W and offering the same set of output voltages. They’re noticeably larger, though what’s interesting is that both the 15 W and 20 W versions use transformers of the same size.



Speaking of dimensions, the datasheet includes a full drawing with the board’s overall dimensions and height, as well as the spacing between the mounting holes and other details.


And, of course, here it is next to a standard matchbox, which isn’t much smaller than the power supply.


A closer look at the individual sections.
1. All connections are made through a terminal block: the input power and ground go in, and the output voltage comes out.
2. The input EMI filter consists of a common-mode choke, one X capacitor, and a pair of Y capacitors. There’s a fuse for protection and a thermistor to limit inrush current at power-on. There’s also an interwinding Y capacitor, hidden under the transformer.
3. The thermistor, Y capacitors connected to ground, and the feedback optocoupler.
4. The input capacitor is rated at 22 µF and 450 V. The product photos show a board version with a 400 V capacitor, so both versions may be available.
5. A high-voltage switching transistor, MDIB6N70C (700 V, 5.0 A, 1.8 Ω), is tucked between the capacitor and the transformer. There’s no heatsink.
6. The output uses an EBR30L60 diode assembly, which, judging by the marking, is presumably rated for 30 A and 60 V. There are three 1000 µF, 10 V capacitors on the output, along with an inductor to reduce ripple.


All the capacitors are made by Aishi. The same was true of all the Sanmim power supplies I’ve tested.
They’re not Rubycon or Nichicon, of course, but they’re not no-name parts either. The input capacitor has an actual capacitance of about 19 µF, and the output capacitors add up to 3000 µF. I measured them directly on the board.


The board and assembly are double-sided, with most components mounted on the underside. There’s a safety slot near the interwinding Y capacitor, and a spark gap under the common-mode choke.
I’ll also note that there’s a bleeder resistor across the X capacitor, and a ceramic capacitor is installed in parallel with the main input filter capacitor—something you don’t often see, even in branded products.
I didn’t spot any narrow traces that could compromise safety.


The “hot” side of the power supply: the PWM controller, thermistor (R12), snubber, and the controller’s other supporting components.


The output section uses a classic design with an adjustable TL431 shunt regulator. An additional ceramic capacitor is installed in parallel with the output capacitor (the one after the inductor).


I sketched a little schematic to make it easier to understand. The schematic is approximate, since I didn't measure the values of the SMD capacitors. The PWM controller shown isn't the one actually installed; I couldn't find its proper name, so I picked the closest match based on the pin functions. The circuit itself is more than conventional.


Let's move on to the tests.
The power supply's output voltage is fixed, and changing it requires a soldering iron and additional components.
In my case, the output was 5.06 V, though I expect there may be slight variations from one batch to another. In any case, it meets the stated specification.
Note that starting up with completely discharged capacitors takes quite a while, around 5 seconds. When I first switched it on, I even wondered whether it was working at all.


The rest of the tests were carried out using a ZKE EBC-A10H electronic load. I checked the load characteristics, efficiency, ripple, and temperature rise.


The very first test surprised me.
The power supply is rated for a maximum current of 3 A, so I set the load test a little higher, at 4 A.
But the power supply handled it without any trouble, and held the voltage remarkably well: it was 5.05 V at the start of the test and stayed there even at 4 A.
I tried a few more times, and even at 5.6 A—almost twice the rated current—the power supply didn't trip its protection and continued to hold the output voltage just as steadily. It seemed not to notice at all. All right, I raised the load current to 5.6 A: no reaction whatsoever, the voltage stayed rock steady, and the power supply didn't trip.


Out of curiosity, I measured the shunt resistance: just as specified, about 0.5 Ω. Then I took another power supply from the same series. It also measured 0.5 Ω at 20 W, and in my tests that unit shut down at around 32 W.
I set the current to 6.6 A and started the test. The power supply shut down at that current, which works out to about 32–33 W.

Some might say, “That's a good thing—look at all that headroom.”
Actually, it's not a good thing at all. For a 15 W power supply, this threshold should be around 20 W, 25 W at most. The load could suddenly increase, the power supply could keep “pushing through,” and eventually fail from overheating.


I first checked efficiency at output powers of 0, 3, 6, 9, 12, and 15 W. Since the power supply could actually handle more, I later added tests at 18, 21, and 24 W.


Here's the same data as a graph. I measured efficiency of up to 77%, but I should point out that wattmeters are noticeably inaccurate at such low power levels, and as far as I know, they tend to read high. At least, mine does. I think the actual efficiency was around 79–80%, but only at the maximum rated power.


I checked the ripple amplitude in six operating modes, two of them outside the rated range, and I think the results are excellent. Even in the worst case, I measured just 20 mV against the specified 50 mV.
1. No load.
2–6. At 5, 10, 15, 20, and 25 W.


I also checked the output using a slower sweep rate, so you can see both the high-frequency ripple and the 100 Hz ripple. The results are a little worse, around 30–35 mV, but still good.
1. No load.
2, 3, 4. At loads of 8, 15, and 25 W.


According to the datasheet, the power supply can deliver its full rated power over an input range of 85–240 V, provided the ambient temperature is between -30 and +60 °C. The power must be reduced at higher temperatures.
In my tests, both the temperature and the voltage were within the range where it can deliver 100% of its rated power.


I started by running it for about half an hour at 50% load, then increased the power to 100%.
1. 50% load.
2, 3. 100% load. The hottest component was the output diode assembly, at around 95 °C.

After warming up, the output voltage rose by about 20–30 mV, which indicates that the feedback circuit resistors are stable.


That didn't seem like enough, so I raised the power to 20 W. Once again, the hottest parts were around the output circuitry: the output inductor reached about 100 °C, and the diode assembly 113 °C.


Conclusion.
Overall, this is a decent power supply. It holds the output voltage very well, even at twice the rated load; has low temperature drift, low ripple, a compact size, and thermal protection. But, as is often the case, there are a few “caveats.”
This power supply has one serious problem: its protection is set to trip at far too high a power level, around 32–33 W. This problem is easy to fix by replacing the 0.51 Ω current-sense resistor with a 0.7–0.75 Ω one. The power supply will then shut down at around 20 W.
The second issue is the long startup time if the power supply has been switched off for a while. To reduce its no-load power consumption, the manufacturer increased the startup resistor values from the usual 1–1.5 MΩ (2 × 510–820 kΩ) to 3 MΩ. This can be fixed by replacing these resistors (R13 and R14 in the schematic).
Even with these quirks, I liked this power supply, just like the three others I tested earlier. It sits somewhere between no-name power supplies and, say, Meanwell units—probably closer to the latter.
Also, I think that with a heatsink fitted to the output diode assembly, you could easily draw 20 W from this power supply continuously.

And a couple of words about the price. In my opinion, it's a bit steep. For example, here in a local store, exactly the same power supply costs about a dollar less.

That's all from me. I hope you found this useful.