Small Isolated DC-DC Converters Rated for up to 50 Watts

Not long ago, I received another little package from Tao and am gradually starting to post reviews of the items in it. I’ll start with some small converter boards, especially since people have been asking me about them.
This review covers boards with an input voltage of 36–75 V and power up to 50 W.


Not long ago, I posted a review of three converter boards with the same input specifications: 36–75 V, what’s known as a distributed power system. This is where you have a very powerful power supply, often at 48–60 V, and all the secondary voltages are generated from it using small converters. Similar voltages are also used in various electric vehicles.
Those were high-power models capable of handling around 200–300 W; this time we have a simpler option, rated for just 50 W. By the way, those models were cheaper, but if I’d ordered them from Tao, their greater weight would probably have brought the total cost to about the same.

Of course, I can already hear the question: we’ve seen reviews of converter boards that are several times cheaper and rated for around 25–50 W, so why are these so expensive?
It’s simple: these boards are galvanically isolated and are essentially small power supplies, but with a lower input voltage. Because they’re galvanically isolated, their outputs can be connected in series without any problems to get other voltages or a split supply. But in my opinion, the most important thing is that the chance of a failure caused by a board burning out is about two orders of magnitude lower than with ordinary DC-DC step-down converters. And you have to admit, that’s worth paying for.

I ordered a lot of different converters, but today I’ll be looking at two bought from the same seller. They actually had three options:
3.3 V, 15 A
5 V, 10 A
12 V, 4.1 A

I can’t even think of a use for 3.3 V at that current around the house, but 5 V and 12 V are much more useful to the average user, so I ordered those two. In the store, they’re distinguished by color, in the order listed above: blue, green, red. Though they’re almost the same colors in person, too—although I’d say the 12 V board looks more yellow than red.


The forwarding agent packed them in the usual bag, and the converters themselves were stuck into a piece of polyethylene foam.
The first thing you think when you see them is how tiny they are: about half the size of a standard matchbox.


The boards were made by ROAL Electronics. The 12 V module is marked mod 223E, and the 5 V one mod 223.
The company does indeed exist. It makes various interesting but highly specialized power supplies, as well as... LED drivers. Unfortunately, I couldn’t find a datasheet for this series of modules, and finding documentation for their other series is a problem too.
The production years are presumably 2007 and 2004.


The differences between the boards are minimal and mostly visible only from above, and even then only on the secondary side. In fact, the 3.3 V and 5 V models are almost identical, but I don’t have the 3.3 V module.


They also weigh almost the same and not very much, which matters when ordering from Taobao.


A few photos from different angles. You can see the small auxiliary transformers.


The power terminals on the output side have a noticeably larger diameter, and the converters are designed to fit into special panels for quick replacement.


As before, the design uses a planar transformer, with its windings formed by a multilayer circuit board. I talked about these in a little more detail in my previous review.


Interestingly, although the modules are galvanically isolated, they don’t have the usual optocoupler for feedback. Instead, there are two transformers, and I suspect all the control signals are transferred through them from the secondary side. Each transformer has 6 pins, and they look identical.


The boards look a little scuffed, and there’s some slight oxidation on the flanges of the power transistors. Perhaps they were stored outdoors, even if they were in their packaging.
1. Primary and secondary sides of the 12 V board
2. The same, but for the 5 V board.


Well, I won’t keep you waiting—let’s move on to the tests. As with the previous boards, these have an enable pin. To start the board, connect it to the input negative terminal; in other words, the board is enabled by pulling the pin low. The voltage on this pin is about 8 V.

And, just like with the larger boards, we see the same problem—or “feature”: high current consumption with no load.

12 V board. At the top, input current in “sleep” mode; at the bottom, with the output enabled. The input voltage is 36 and 62 V, respectively.
Standby consumption is still more or less reasonable, but 1.5–1.8 W when running is a bit much.


The following tests were carried out with a powerful power supply, an oscilloscope, and an electronic load. This time I decided not to use a multimeter. The little board looks microscopic next to all that. :)


The converter starts up at 33 V and, with no load, shuts down when the voltage drops to 30 V. The specified operating range starts at 36 V, but it can handle the full load at 33 V; above about 60% load, though, the output voltage starts to drop slightly.
You can see in the photo that once it starts up, there is noticeable noise on the output. It’s clearly not coming from my power supply, since everything is quiet before startup.


Load regulation at an input voltage of 36 V.


The same test at 62 V input: the output voltage is rock solid, measuring 11.878 V at a 100 mA load and 11.85 V at 4.5 A (10% overload).


The output ripple is considerably worse, though. All the oscilloscope traces shown below were captured at four load levels: no load, 33%, 66%, and 100%. The oscilloscope’s 20 MHz bandwidth limit was enabled, the probe was set to 1:10, and the corresponding compensation was applied.
At 36 V input. For the last screenshot, I even had to increase the volts per division; otherwise, the trace wouldn’t fit on the screen.
In short:
No load — 450 mV p-p
33% — 900 mV p-p
66% — 1.4 V p-p
100% — 1.5 V p-p


At 62 V input, things are a little better. All screenshots were taken at 200 mV/div.


Now for the temperature—this is where things get more interesting. The store page says the maximum power without cooling is 15 W, with a very subtle hint that there are modules with the same power rating. I assume the subtext is: buy the cheaper modules that also deliver 15 W, and which the seller probably has more of. :)


The test was done with the module already warmed up from the previous tests. First I loaded it to 15 W, waited about 10 minutes, then checked the temperature again: 60°C maximum.


All right, I thought, what if I push it harder?
I loaded it to 25 W; after 8 minutes it was at 72°C. That looked good, so I increased the load to 35 W. At that point it started heating up noticeably, reaching 97°C. All of this was without additional cooling.


And what about efficiency? First, here’s a simple table with four key load points: 25%, 50%, 75%, and 100%.
36 V input
25% — 88.6%
50% — 91.6%
75% — 91.3%
100% — 90.0%

62 V input
25% — 86.5%
50% — 89.8%
75% — 90.1%
100% — 89.1%

As you can see, efficiency is slightly higher at the lower input voltage. Here are a few photos at the extreme load points.


The 5 V board. All tests were identical to the previous ones, with the load current adjusted accordingly.
Power consumption at different input voltages, in standby and running with no load.


Output voltage regulation is excellent at both 36 V and 62 V input.




Ripple: I’m gradually getting used to the new oscilloscope, and I hope the traces are easier to read this way.
But the ripple itself is certainly high, measuring:
At 36 V input:
No load — 800 mV p-p
33% — 850 mV p-p
66% — 1.2 V p-p
100% — 1.4 V p-p


At 62 V input, things got even worse, although the previous converter showed the opposite behavior. The first three screenshots are at 200 mV/div, the fourth at 500 mV/div.


The input voltage during the test was 48 V.
At a load of 15 W (3 A), the temperature is still 60°C.


At 25 W and 35 W, it runs a couple of degrees hotter than the 12 V version.


And of course, efficiency measurements.
36 V input
25% - 89.3%
50% - 91.3%
75% - 90.7%
100% - 89.2%

62 V input
25% - 87.8%
50% - 90.4%
75% - 89.5%
100% - 88.6%

On average, they're about the same: a little better in one operating mode, a little worse in another. But note that across a load range of 2.5–10 A and an input range of 36–62 V, the output voltage stays rock steady.


And of course, here's a group photo alongside its more powerful “siblings” :)


This is the power supply designed to power these and similar modules: up to 2000 W, with a 48 V output. Special thanks to one of my regular readers for this gift :)


So, what can I say in conclusion? First of all, both converters work, which is already a good start. They can deliver up to 25 W without additional cooling, and up to 35 W with a little airflow. They can handle full power continuously with proper cooling. For short periods, they easily deliver their full 50 W, with very stable output voltage at input voltages from at least 36 to 62 V.
The only complaint is the output voltage ripple: it's really high, and I wouldn't use these without an additional LC output filter, at least with anything remotely sensitive.
The advantage of converters like these is full galvanic isolation, which makes them safer for the load they power.

Otherwise, it's a very useful little device. If you're buying from Taobao, it's worth adding one to another order or buying several at once; the price then works out very well. They're very light, too: you can fit about 70 in a kilogram, so even at $20 per kg for shipping, each one comes to just $4.5.