Small DC-DC converters with a USB connector

I got hold of several compact step-down converters. I bought them just “to try out,” since they’re very cheap. Today I’m reviewing a pair of almost identical, yet different, boards.


It all started when a friend asked me to buy a few dozen small converters. I bought one for myself too, just because, with no particular purpose in mind.
The boards cost $0.68 each - link. They arrived as almost a whole sheet, which then had to be broken up into individual boards.



The boards are very compact: 15 × 26.5 mm including the connector.


They look very neat, but it’s worth noting that these boards don’t support QC; the output is 5 V only.


I was about to write a review, but didn’t have the time, so I tossed my board into a desk drawer. Then, while placing another order with a different seller, I came across the same boards in two versions. One cost 80 cents and the other 78; they’re both currently selling for $0.67.
Having successfully forgotten that I already had one of these boards, I ordered one of each version. So I ended up with three boards: two of one type and one of the other (in the little bag).


I’ll stop trying to build suspense and explain.
On the left is the first type of board. I ordered one earlier and the second recently; it cost 80 cents. I’ll call it “type one.”
On the right is the second type, ordered along with one of the first two. It cost 78 cents; I’ll call it “type two.”

The key difference is the inductor type. The seller lists the first board as “3A USB Output good” and the second simply as “3A USB Output.”


Both boards are built around the same chip and use the same component values. They both have an output TVS diode; the only difference is the type of inductor used.


The inductance is the same too, but the inductor on the second board is a little shorter, presumably rated for a lower current.


One of the sellers has an image with more details about the components.
The boards also have an input diode for reverse-polarity protection, plus a 0.1 Ω resistor in series with the input for some reason. The resistor’s losses aren’t particularly high, unlike the diode’s, but I still can’t figure out what it’s for. The only thing I can think of is that it acts as a fuse.


It turns out the boards also differ in their PCB layout, so the inductor isn’t the only difference. Perhaps one is an older version and the other a newer one. “Type one” is on the left, “type two” on the right.


For testing, I used a power supply and an electronic load. I tested load capacity and efficiency using a switching power supply, and measured ripple using a linear one to rule out its influence on the results.


Before getting to the tests, here’s some brief information about the boards from the seller’s page. Below is a screenshot with Google Translate, but in short:
1. The components are claimed to be new, though that usually doesn’t matter much.
2. Maximum short-term current: 3 A; continuous current: 2.1 A.
3. Ceramic capacitors instead of electrolytic ones, so the claimed ripple is 10 mV, compared with up to 500 mV for competitors. We’ll check that.
4. High efficiency, up to 94–97.5%, with a caveat: that figure is without the fuse and protection diode. Since they’re on the board, I’ll measure with them in place.
5. Inductors wound with a flat strip rather than wire. I can’t verify that without taking them apart.
6. Output TVS diode. Both boards have one, so no questions there.
7. Some kind of USB identification circuit. It isn’t actually there: the USB data lines are connected together and tied to a voltage divider.
8. Gold-plated USB connector. Unlikely, but I’ll measure the voltage drop across the connector too.
9–11. Various other nonsense.
12. No-load current consumption: 0.85 mA. I didn’t measure it specifically, but the power supply did show zero current with no load, even though its minimum display reading is 1 mA.



I’ll start with “type two,” since that’s the one I placed the second order for.


The first run showed a maximum current of 2.9 A at an input voltage of 12 V, but I wasn’t convinced, so I ran several more tests. In all of them, the maximum was 2.8 A. The shutdown is abrupt rather than a gradual drop in voltage, which can be useful in some situations.


The board can deliver 3 A, but only for a couple of seconds; after that, the voltage drops and it shuts down.


With an input voltage of 24 V, the board also managed a maximum of 2.8 A. The output voltage does drop as the current increases, but regulation is very good: the difference between the maximum current and no-load operation is just 0.1 V, including the drop across the connector.
The measured voltage drop across the connector was 6+14 mV (20 mV total) at 1 A, or 60 mV at 3 A, so the drop excluding the connector is only 40 mV.


Efficiency measurement table; each step is 200 mA of current.


And here's the graph, which I think makes things clearer.


It gets hot as a furnace, that's for sure. I first tested it at 1.25 A, and it reached 63 degrees. Then I wanted to run it for a while at 2.5 A, but the board ended up overheating and shutting down even at 2.3 A.
Below are three thermal images: 12 V at 1.25 A and 2 A, and 24 V at 2 A.


Ripple at an input voltage of 12 V and load currents of 0, 1, 2, and 2.8 A. Ignoring the spikes, it's around 30–40 mV.


But at 24 V input, things are noticeably worse. At load currents of 0–1.2 A, it's about the same as before, but at 2.8 A it's around 110 mV, excluding the spikes. The last screenshot was deliberately taken at the same scale to make the difference clearer.



Now for the “type one” board.


The output voltage holds up just as well as on the previous board, but the maximum current came out to 3 A, both in the first run and in subsequent tests. The shutdown is abrupt here too.


At an input voltage of 24 V (I forgot to change the heading), the output current even reached 3.2 A, noticeably more than the previous board's 2.8 A.


Efficiency measurements, using the same 200 mA steps, but this board managed one more step at 12 V input and two more at 24 V input.
For some reason, connector losses were lower here. On the previous board, the drop at 1 A was 6+14=20 mV, while here it was 2+11=13 mV. The drop on the ground conductor is lower because the connection also runs through the connector housing.


Efficiency shown as a graph.


It also runs noticeably cooler. The first two thermal images were taken as before, at 12 V input and load currents of 1.25 A and 2 A. The third was taken at 24 V, with a load current of 2.3 A, and the board didn't shut down.


Ripple is also noticeably lower here, about half as much, and around 20 mV excluding the spikes, although still well above the claimed 10 mV. Load currents: 0, 1, 2, and 3 A.


At 24 V input, the results are worse than at 12 V, but noticeably better than with the previous board, especially at maximum load current. All the waveforms here are shown at 10 mV/div; the previous board's were at 20 mV/div.


A comparison of the efficiency measurements for both boards. The 12 V results are on the left, and the 24 V results on the right.
At 12 V, the boards have roughly the same efficiency, but at 24 V the “type one” board clearly outperforms the “type two” board.


Later, just out of curiosity, I calculated the efficiency without accounting for the input resistor-fuse, the protection diode, or the voltage drop across the connectors. I won't include the full formula, but at an input voltage of 24 V, the result was 1–2% higher, depending on the load current.
The graph shows the results in blue without accounting for parasitic losses in those components, but even so, I still couldn't get 97%.


To sum up, both boards work, but in my opinion the “type one” board is noticeably better. If you're buying one, I definitely recommend that one.
You can also noticeably improve efficiency by removing the reverse-polarity protection diode and replacing it with a jumper. The diode gets quite hot during operation and, in turn, heats up the PWM controller, which eventually shuts down due to overheating. I think that after removing the diode, the board should be able to deliver 2.5 A continuously without any problems.

That's all from me. As always, I'll be happy to answer questions or just read your comments.