Voltage converters: compact, powerful, and galvanically isolated

Today's review is about some uncommon, specialized devices that can nevertheless be very useful: powerful, compact voltage converters. They're specialized mainly because their 36–75 V input range is uncommon among DIYers. But I've been asked about converters like these several times, so I decided to fill that gap, since I haven't come across any reviews of them yet.


First, why would you need converters like these at all?
1. For the increasingly common electric transport: scooters, bicycles, and the like.
2. To power devices with higher reliability requirements for the converter.

I've been asked several times about converters for small electric vehicles. Their batteries often run at around 60 V, and most converters on sale can't operate at that voltage.
There's also the question of protecting against a shorted switching transistor in a step-down converter. That's why I usually recommend buying an isolated converter—not because isolation itself is necessary, but because the chance of something dangerous appearing at the output is orders of magnitude lower with these converters.

So I ordered three versions, especially since they cost next to nothing. I ordered two from one seller and one from another, so they arrived in two packages. Each converter was in its own slot cut from polyethylene foam. That's how they're originally packaged, just in larger sheets.


Here are the contenders in this test.
1. RBQ-31251 — 9.7 V output, up to 40 A, price: $2.81, link.
2. QBW018A0B — 12 V output, up to 18 A, price: $5.47, link
3. RBQ-8,2/45 — 8.2 V output, up to 45 A, price: $3.12, link.

According to their datasheets (more on those later), all the converters have the same 36–75 V input range and the same dimensions: 57.9 × 36.8 mm. At least two meet the Quarter Brick industry standard (57.9 × 36.8 × 10.6 mm).


These little units have some heft, weighing from 50 to almost 100 g.


I'll start with the RBQ-31251 and RBQ-8,2/45 models. I didn't notice at first, but once they arrived I realized they're essentially the same model, differing only in output voltage and the fact that the RBQ-31251 has a heatsink.


The undersides of the boards are identical.


You might expect the markings to be very similar too, but no: the only things they have in common are the manufacturer and the series, RBQ.
As for the manufacturer, as far as I can tell, these are products from the very well-known company Murata. The chance of them being counterfeit is vanishingly small: counterfeiting devices in this class just to sell them for a few dollars wouldn't make sense.


First up is the RBQ-8,2/45. Let's start with the manufacturer's datasheet. It's not for this exact model, but for another one in the same series. Incidentally, the datasheet is marked “Discontinued,” meaning this converter is no longer officially manufactured and the units for sale are old stock.


And here's what it looks like in real life.



PCB dimensions and pinout.


Despite the device's fairly simple functionality, it has quite a lot of components, some of which are covered by a small heatsink.


The transformer and inductor are also secured with sealant.


The contacts are quite substantial, clearly made of copper and apparently gold-plated. But their thermal conductivity is something else... I had to use a powerful soldering iron to attach the test leads to the output contacts, and much of the board visibly heated up in the process.
Input on the left, output on the right.


There are three contacts on the input side. The middle one enables the converter, and here's a small caveat: some converters are enabled by a logic high, others by a logic low, depending on the model suffix. At first I powered up the first converter, but it didn't work. Then I tried the second and thought it was faulty. When the third one wouldn't work either, I realized I must be doing something wrong. It turned out these models are enabled by a logic low, i.e. To turn it on, connect the middle pin to the input negative.

The culprit is the circuit from the datasheet, which shows the logic for a different type of control signal :)


The power supply is controlled by an LM5035 PWM controller.


This controller controls not only the primary side, but also the secondary side—more specifically, the synchronous rectifier.


Current is monitored using a dedicated transformer.


Essentially, this board is a conventional power supply, but designed for a low input voltage. It has the usual optocoupler and interwinding capacitor.
There are several spots for resistors on the secondary side. I assume they are used to set the output voltage.


The seller's page also has a few photos. The first shows a 2 kΩ resistor on the board.


Then there's a suggestion: replace this resistor with a 1.5 kΩ fixed resistor in series with a 4.7 kΩ potentiometer, and you can adjust the output voltage from 6 to 10 V. However, the converter with a 9.7 V output has a 6–10 V range, while the one with an 8.2 V output is shifted lower, to 3.4–8.7 V. For that one, though, the resistors should be 2 kΩ and 10 kΩ, not 1.5 kΩ and 4.7 kΩ.

So, in summary:
The 9.7 V output model can be adjusted from 6 to 10 V using 1.5 kΩ and 4.7 kΩ resistors.
The 8.2 V output model can be adjusted from 3.4 to 8.7 V using 2 kΩ and 10 kΩ resistors.


The output uses AON6240 transistors rated at 40 V, 85 A, and 1.6 mΩ.
On the input side, there are FDMS86322 transistors rated at 80 V, 60 A, and 7.65 mΩ.
And a few other small components.


The second converter is almost an exact copy. The key difference is that it has a small heatsink covering one side of the board.


Under the heatsink is the same board shown above; this photo is from the product page.


It has additional standoffs, and the connection leads are longer.




You may already have noticed the unusually thick circuit board. It measures about 3.5 mm thick, compared with 1.5 mm for a standard board. As you can imagine, there's a reason for that.


The reason is that the transformer and output inductor in these converters are integrated into the circuit board. In other words, there is no conventional wire here; its job is done by the multilayer board.


This kind of solution is used when you need a compact, highly efficient design, as well as in military equipment.


There are several advantages.
1. Better coupling between the windings
2. Lower leakage inductance
3. The product can be made thinner
4. Easier to manufacture.

You can also integrate a cooling system for the windings, so to speak. More details here.


E-core magnetic cores are used most often.


There are also several ways to make them.
1. The transformer is a separate component.
2. As in option 1, but the magnetic core is recessed into a cutout in the circuit board.
3. Hybrid: some of the windings are on a separate board, while the rest are part of the main board. In my opinion, this is better suited to high-voltage products.
4. A fully integrated transformer/inductor, with all the windings forming part of the main board. The converters under review use this option.


Special low-profile magnetic cores are also made for applications like these. You can read more here; it's a very useful article on the website of a company in Kharkiv.


Above, I mentioned improved coupling between the windings. Let me explain this in just a few words.
1. The worst possible winding arrangement is to put the windings on different parts of the same core. This is unsuitable for switching power supplies. If you try to build a transformer for a switching PSU this way, its life will probably be bright, but short.

Next is a cross-section of the bobbin, showing the winding layers.
2. The simplest arrangement, with one winding on top of the other, works, but only at low power. The problems themselves don't depend on power; they just become more pronounced at higher power, usually as increased snubber (parasitic spike suppression) heating. Because the coupling between the windings is poor, energy transfers to the load more slowly, and a fairly large portion goes back to the primary side (I've simplified things a lot).
3. To improve things, the primary winding is often wound in two sections: first one part of the primary, then the secondary, and then the other part of the primary. This way, the primary winding sort of "embraces" the secondary and provides better coupling between the windings. In other words, when the inverter transistor switches, current flows to the load faster (if you can put it that way), leaving less work for the snubber.
4. And here's an example of a planar transformer "winding". You can achieve something almost never done with conventional transformers: a multilayer sandwich. With conventional winding, this is difficult to manufacture, but it's no problem for a planar transformer. The coupling between the windings is almost ideal, so this design has higher efficiency, as well as fewer spikes on the primary side and less work for the snubber.

By the way, incorrect transformer winding is a common mistake among beginners who don't yet understand that it's not just the turns ratio that matters, but also how the windings are arranged. The higher the PSU power, the more pronounced this problem becomes. It's basically the same as PCB routing: you need not only to connect the components according to the schematic, but also to account for how the traces affect one another.


Let's move on to the second type of converter. This one is designed for the more popular 12 V output and provides up to 18 A.
In this case, things are a little simpler: there's the original datasheet.


It doesn't look much different from the previous one: the form factor is the same, as is the input voltage.


By the way, about the input voltage: the description gives a range of 36–75 V, while the maximum values listed are also 75 V, or 80 V if the converter is in "sleep" mode. Some sellers advise against applying more than 60 V. I'd recommend staying at or below 65 V, with 70 V as the absolute maximum.


The transformer here is smaller, but the power is also almost half as much, around 200 W.


There are no heatsinks on this board.


Since the converter has a standard form factor, its dimensions are the same as the previous one.


The connection terminals here are a little simpler. Interestingly, the end of the terminal conducts current, but inside it looks like a fiberglass rod wrapped in foil. It can't be soldered, but it has excellent thermal conductivity.


The board has exactly the same contact layout and pinout, so you can desolder one module and easily solder in another if you need different specifications.


Some converters of this type can also be connected in parallel. Certain versions have a dedicated pin for synchronizing their operation, so the boards can be stacked like a sandwich.
By the way, while looking for an example, I came across a board like the one shown earlier, made by Murata.


As I already mentioned, there are no heatsinks here, so all the components are "right out in the open". The input section is on the left, the output section on the right.


The module markings are visible in the photo above. Interestingly, I first found a datasheet for the same module made by a different manufacturer, and only later downloaded the datasheet from the seller's page. As they say, spot the difference.
Although the Tyco datasheet was published 4 years earlier, this is an interesting situation overall: these high-efficiency voltage converters have been around for 14 years, yet what usually reaches the masses are super-budget versions.


A planar transformer and inductor are used here as well.


There isn't much to say about the components. I'll just note that the board has two optocouplers: one for feedback, and the other for overvoltage protection on the output.



Now for the tests. I'll start with the RBQ-8,2/45 model. I don't see much point in testing the RBQ-31251, since they're like twins. The only exception might be the version with a heatsink, which could deliver a little more power without forced cooling.

And here's the first disappointment: power consumption.
In standby mode, it draws about half a watt. The higher the input voltage, the lower the power consumption. That's quite a lot, in my opinion, especially when running on battery power. I also hoped the start input could be used to switch it into a low-power mode, rather than disconnecting the power supply. You might ask, why not just break the power supply circuit? That's the problem: the contacts in ordinary switches and relays are rated for a maximum of 30 V DC, and this can easily be twice that.


Connect the control input to the negative supply terminal and the converter starts up. The output is 8.271 V.



The no-load power consumption really surprised me: about 4 W. As the input voltage rises, the current drops, but the power still increases slightly.
Pretty much the same thing happens under load: power drops slightly as the input voltage decreases, and rises as it increases.


Now we can roughly estimate the efficiency and how the output voltage varies with load current.
The input power is 70.3 W, and the output is 65.1 W at a load current of almost 8 A, giving an efficiency of about 92.6%.
At 16 A, the input power is 136.7 W and the output is 130.8 W, for an efficiency of about 95.7%.


After 20 minutes, the converter had heated up to almost 90 degrees, and the temperature stopped rising after that.


To test it at higher load currents, I had to use a more powerful power supply and a load capable of drawing up to 40 A at up to 300 W.
The results:
Output current: 23.9 A. Voltage: 8.195 V. Efficiency: about 98%.
At 31.9 A, the voltage dropped to 8.163 V, and efficiency came to 97.7%.

At these power levels, active cooling is necessary. The converter is highly efficient, but it still starts to overheat after a while.


The highest load I could manage was 37.9 A, at about 300 W, with an efficiency of roughly 97.9%.

Of course, the efficiency calculation has a significant margin of error because it depends on the accuracy of four measurements, but the result is still pretty good.


With a small fan blowing on it and an output power of 300 W, the temperature was about 60 degrees.


The ripple amplitude and waveform are almost identical both at no load and at 38 A.
The oscilloscope traces were taken at currents of 0, 8, 16, 24, 32, and 38 A.


The tests of the second converter will be a little shorter.
1. In standby mode, it draws twice as much power, which is pretty disappointing :(
2, 3, 4. At no load, its power consumption is higher than the previous model's and also depends on the input voltage.


With no load, the output voltage is slightly higher than specified, at almost 12.2 V.


Efficiency and output voltage at load currents of 6, 12, and 18 A.
1. 95.4%
2. 95%
3. 95.5%


The ripple waveform and amplitude remain completely unchanged, both with no load and at maximum current, at around 100 mV.


The overheating test was performed at 10 A with passive cooling.


The results are so-so: after about 15 minutes, the converter heated up to 102 degrees, and a little later to 111; after that, the temperature barely changed.


Heating also affects the output voltage: the difference between the cold and hot states was about 0.1 V.


At half load, the converter operates reliably over an input range of 36–60 V. Its power consumption stays the same from 48 to 60 V, and drops slightly at 36 V.


At full load, however, the converter operated reliably only with an input of 44–45 V; lowering it further triggered the protection. Power consumption (and therefore efficiency) is the same across the 45–60 V range.


Among other things, the second converter was accidentally tested for short-circuit protection. It worked perfectly: the converter shut down and waited for its power to be restarted.

There are also versions of similar converters with a larger heatsink, capable of delivering up to 500 W.


Time to wrap things up. When I started writing this review, I thought it would be about three times shorter, but somehow it grew unexpectedly :(

Now, about the converters.
The first one doesn't offer very convenient output voltage options, but at least they were available and inexpensive. On the plus side, these are branded units with high efficiency, high output current, and quality construction.
At room temperature, the RBQ-8,2/45 can deliver up to 15–16 A without additional cooling.
I think you can expect 20 A, perhaps even 25 A, from the RBQ-31251, since it has an additional heatsink.
And the price speaks for itself: for this money, you usually get something based on Chinese PWM controllers, with worse specs.

The QBW018A0B (12 V, 18 A) doesn't fare as well. I doubt it can deliver more than 8 A for long without forced cooling. One option is to install a heatsink on it.

I have just one significant complaint about all these converters: their power consumption in standby and while operating with no load. In my opinion, 0.5–1 W when switched off and around 4 W with no load is far too much. The first figure is something you can just about live with, or simply add an electronic switch, but there's no way to fix the second—the models are clearly designed to operate at high currents.

That's all from me. I hope I've gathered some useful information, and that it will be useful to someone.


This review was sponsored by the yoybuy.com forwarding service, which covered the shipping costs. Affiliate link to sign up; new users get a 10-off-50 coupon.