SANMIM GPT70A-12 power supply, 70 W, 12 V

This is the third power supply from SANMIM and, in my personal opinion, the most interesting of the three. It also looks very much like another power supply model that I think many of my readers will recognize. But you can read all about it in the review.


So, this is the third power supply. The voltage is still 12 V, but the power is higher: 70 W, compared with 12 W and 20 W for the previous two. It’s also the only one of the three that comes in its own box. A small thing, of course, but still.



As with the previous models, there’s a datasheet with technical specifications and operating notes. In my view, that already shows the manufacturer takes its products seriously. The manufacturer also has a website, though it’s a bit... rough around the edges and, naturally, in Chinese.


There are two other models in this power supply series: the GPA50A, rated at 50 W, and the tiny GPA15A, rated at 15 W.
Mine is the 70 W model, which I think offers the best balance of power and price. I bought it along with the previous models at the Kharkiv Cosmodrome.


The GPT70A series includes six power supply versions with output voltages of 5, 9, 12, 15, 24, and 36 V. At the moment, though, the store only has the 12 V and 24 V versions. I’d add the 5 V and 36 V models, since 9 V is rarely used, and you can get 15 V from the 12 V version if you really want to.


I can sum up my impression in one phrase: it feels good in your hands. Everything is clean and neat; nothing is loose or dangling. Just a solid 5 out of 5.


I once saw a comment saying, “The Chinese are at it again: they make the board and forget the mounting holes.” Here it’s exactly the opposite: there are no fewer than 8 mounting holes, and they’re elongated, which I especially like. It’s genuinely well thought out. I don’t even remember seeing that on Mean Well units.


A drawing with all the overall and mounting dimensions, though unfortunately it isn’t very clear.


And, of course, a comparison with a standard matchbox. What would a review be without one? :)


Here’s a more interesting comparison: with the “people’s” power supplies. These are 24 V models, but that’s not important. I recently pulled the first one out of an electronic load; the second has been lying around for a long time. They differ in PCB revision and PWM controller type. The one on the right is the unit under review.
The power supplies look very similar. The GPT70A is slightly larger, and the mounting holes don’t line up, but otherwise they’re much alike. I’d even guess the GPT70A came first, and was later heavily simplified and “optimized,” fitted with lower-quality components, and turned into the “people’s” power supply—which, in reality, isn’t all that much cheaper. I’ll inevitably compare these two versions from time to time.


The input has an almost complete mains filter; the only thing missing is a varistor. The input capacitor is 150 µF, made by the same manufacturer as before: Aishi.


The layout is very tight. The screw securing the high-voltage transistor almost touches the capacitor body. Nearby, you can see a TVS diode and a Y capacitor connecting the inverter’s negative supply rail to the copper pour around one of the mounting holes.


The transformer is also fairly compact. Two heatsinks handle cooling: the transistor heatsink is connected to the negative terminal of the input capacitor, while the diode heatsink is connected to the power supply’s negative output.


Between the heatsinks are an interwinding Y capacitor, the PWM controller’s supply capacitor, and two optocouplers. Yes, this power supply has overvoltage protection on the output.


The output has two 20 A, 100 V diode assemblies connected in parallel, four 1000 µF, 25 V output capacitors, and an inductor to filter ripple. There’s also a 22 µF, 50 V capacitor at the output, plus a 0.1 µF ceramic capacitor. The last thing really surprised me. It looks like they accidentally left in the filter circuit used when measuring ripple—very unusual...


The capacitor values are a little on the low side: 125 µF on the input and 865 µF on each output. But even so, with our mains voltage, the input capacitor has plenty of headroom. The extra capacity is there because the power supply is designed for a wide input voltage range. Still, it only delivers 100% of its rated power from 110 V upward.


Since I was desoldering the input capacitor anyway, I checked which high-voltage transistor was used and the suppressor's rating.


The board is double-sided, and some of the high-current traces are duplicated on both sides. More importantly, the vias connecting the two sides are doubled up for greater reliability.
There are protective slots at all the narrow points, and you can tell just by looking at the board that someone put thought into the routing instead of simply running an autorouter.


The PWM controller is OB2269CP, with a thermistor for overheat protection visible to its right.
The feedback circuit uses the classic TL431 adjustable shunt regulator. Above it is the output overvoltage protection circuit: if the voltage rises to 15 V or higher, the PWM controller shuts down.


I couldn't resist drawing up the schematic for this power supply. It could well become a popular go-to model, so someone might want to adjust it for a different voltage or simply understand how the circuit works.


Since it connects via plugs rather than terminal blocks, I had to solder the wires in my case.
Powering it up: everything works, and the output voltage is just a little below the specified value.


Next, I grab the modified electronic load and move on to the tests. First, I measure the power consumption with no load: the meter reads 1 W.


I had to test the load capacity in several stages, but the results showed that the maximum current is 8.5–8.6 A. Beyond that, the overload protection trips. At that point, the output power is already over 100 W, despite the rated 70 W.
This is useful in practice when connecting a load that needs a high startup current, such as hard drives, motors, and the like.
The output voltage holds up extremely well from zero current right up to the protection trip point, with a difference of about 10 mV.


The power supply's efficiency is almost independent of the load, ranging from 80% to 87%. Peak efficiency is reached at an output current of 7 A.


Thermal test. While taking measurements, I also tracked how the output voltage changed with temperature. I found that the output voltage rises slightly as the unit warms up.
1. Load current: 4.5 A
2. 6 A
3, 4. 7.5 A, immediately after setup and 20 minutes later.


The total voltage drift from cold to maximum temperature was +15 mV, which is also an excellent result.


I measured component temperatures using my usual method: set a specific load current, wait 20 minutes, take readings, and if they're within limits, increase the load current by one step and wait another 20 minutes.
There were five steps in total, each increasing the current by 1.5 A, or the load from 25% to 125%. The test took 1 hour and 42 minutes in all.


Component temperatures stayed within safe limits even at 125% of the rated power. I didn't push it any further.


The datasheet does include a note about temperature: at an ambient temperature of 55°C or higher, the output power must be derated linearly. In my case, the ambient temperature was about 30°C, so at 55°C you can add 25 degrees to all the temperatures listed here. That makes it clear that 6 A really is the maximum current, unless you provide airflow over the power supply.


Thermal images at load currents of 3, 4.5, and 6 A.


And finally, at a load current of 7.5 A, at the very end of the test. The highest temperature was recorded around the transformer (95 degrees) and its windings (100 degrees).


Output ripple measurement. The oscilloscope was connected directly to the output connector, with the load on the other side of the board, beyond the oscilloscope connection point.


1. No load; the frequency is lower because the "green mode" has kicked in. Peak-to-peak amplitude: 20 mV.
2. 50%, or 3 A: 40 mV peak-to-peak, but the main ripple is only about 10 mV.
3. 100%, or 6 A: 50 mV peak-to-peak, with the main ripple at 10–15 mV.
4. 125%, or 7.5 A: much the same as at 6 A.
5, 6. The sweep frequency is lower. Results with no load and at 7.5 A show an overall maximum peak-to-peak amplitude of 55–60 mV.

The results are noticeably better than those of the "people's choice" power supply, which had up to 150 mV at maximum power. The overall result is pretty good in its own right, too, but adding ceramic capacitors at the output could reduce the ripple even further.


So, friends, what can I say about the results of testing this power supply, and about the SANMIM power supplies I've tested in general?
I really liked the overall quality, both in terms of component selection and build quality. The test results are also quite in line with the price. Compared with the well-known Mean Well, it's a little worse, especially in terms of heat dissipation, but noticeably better than no-name Chinese power supplies.

Now, about this model.
The circuit design is quite conventional, of course: there's no synchronous rectifier, APFC, or other "bells and whistles." But it does have a well-thought-out design, thermal protection, output overvoltage protection, a high maximum current, relatively low ripple, input and output filters, a quality circuit board with duplicated interlayer vias, and simply neat workmanship. The power supply left a positive impression, which is only slightly marred by its somewhat low efficiency and the resulting heat. A synchronous rectifier would improve things considerably, but that would put it in a different price category.
If I had to choose between the "people's choice" model for 7–8 dollars and this one for 12, I'd choose this one because it works properly "out of the box."


In short, it's a sort of "people's choice" Mean Well: the price is still reasonable, but the quality is already pretty good.
A similar power supply costs about the same on AliExpress.

That's all for today. I hope you found the review useful.