Multichannel DC-DC Converter

Today I'm reviewing another kit for building a multi-output voltage converter.
Some time ago, at readers' request, I ordered several different voltage converters. I picked this one myself, just out of curiosity—I wanted to see what it was like, and thought it might be useful to someone.
Photos, schematics, tests—the usual :)

As I mentioned above, I ordered this converter just for the sake of it, so to speak—to check it out and show it to others. It has no practical use for me.
What initially caught my interest was that it can provide 5 output voltages (two of them negative) from an input voltage of 5–24 V.
That really was intriguing.

Dual-polarity power supplies are fairly uncommon, used for example in audio equipment or when working with sensors.
But since this is a kit, I thought it might be interesting, so I ordered one.

The converter arrived in a thick polyethylene envelope, wrapped inside in a sheet of polyethylene foam. This is pretty standard packaging for this store; I'll just say that nothing was damaged.


Inside was a small resealable antistatic bag. As you can see in the photo, everything inside was just loose in the bag.


I emptied the bag onto the table and got this little pile. By the way, be careful when unpacking: the kit includes components that like to roll around on the table and may escape over the edge :)


Since this is a kit, most of the review will focus on assessing it as a kit. I'll test the finished result separately.

Here's what comes in the kit.
The first thing that surprised me was the quality capacitors made by Samwha—and they're from the RD series, too.
I use these capacitors all the time; they're quite reliable and inexpensive.
There's also a bunch of different connectors. The kit is designed as a fairly versatile solution.

I have no complaints about the contents: everything is there, nothing is damaged, and they even included a few small heatsinks.


There was no schematic in the kit, since all the necessary markings are printed on the circuit board. It looks like there are a lot of components, but in fact, most of them have the same value.


There are no markings on the back of the board because no components are installed there; it's a single-sided board.


I also have no complaints about the circuit board's build quality: it's excellent, the silkscreen is crisp, and everything lines up. The board is made of fiberglass laminate.


There was no schematic in the kit, and I was curious to see what the Chinese engineers had come up with, so I drew up a schematic for the device.
If anyone happens to buy this kit, it may come in handy.

The two power connectors are shown on the left; the USB connector can be disconnected.
In other words, if you connect the second connector while USB is plugged in, the board will draw power from the second connector.

The input capacitor is only 22 µF, which caused me some trouble during the review. For example, my bench power supply showed an average current of 0.5–1 A, but the current-limit LED still came on, even though the limit was set to 2.5 A. Increasing the input capacitance helped, but didn't completely solve the problem.

The use of electrolytic capacitors for C3 and C5 was a little unexpected; ceramic capacitors are usually used in circuits like this. I don't know whether it was a cost-saving measure or a sound design choice—probably the former. Or perhaps they did it to make the kit easier to assemble.


Since the circuit looks complicated at first glance, I drew a block diagram to make it easier to understand.
Everything on the right side of the diagram is just five ordinary linear voltage regulators, so I don't see any point in discussing them—I'm sure everyone knows what they are.
The left side of the circuit is much more interesting.
Yes, it's another SEPIC, in an almost classic two-inductor configuration. I've done several reviews of this type of converter, and I wasn't planning to talk about it in this review, but it caught up with me here too (they've clearly been reading my reviews and are now following me).

At first I thought the board simply had a boost converter plus a negative-voltage generation circuit. I thought that because I hadn't noticed that the input voltage range goes up to 24 V, and only a SEPIC can work like that.

Components L1, L2, C1, C3, VD1, and the PWM controller shown as S1 form a classic two-inductor SEPIC converter (I'll be reviewing another converter using a different circuit topology soon).
There's not much to say about this part; I've already done several reviews where I explained everything I could.
But the designers went further and added a negative-voltage generation circuit.
It's highlighted in red in the block diagram.
Capacitor C2 extracts the AC component, and then the circuit works like a regular step-down converter, but with a negative output—the diode is installed the other way around.
That's the whole trick.


Now that we've looked at the circuit and how it works, let's move on to assembling it and describing the process.
This board is well suited for learning, since it includes both regular components and SMD parts. At the same time, it doesn't use tiny components, which can be harder to learn to work with.

The kit included an XL6008 PWM controller XL6008. It's a boost converter controller (the power switch is connected to ground), or it can be used for a SEPIC converter.
The built-in power switch is rated for a maximum current of 3 A, but since this is a boost converter, that's not much: the input current is always higher than the output current.
The maximum voltage is 60 V, but that doesn't matter in this circuit, since it's configured for a lower voltage. Personally, I think a 30 V, 6 A chip would have been a better choice.

The kit also included five linear voltage regulators: 12, 5, 3.3, -12, and -5 V.
There probably won't be any questions here, since this is a classic setup.


Since this review is partly educational (after all, we're reviewing a kit), I'll also say a little about soldering and the specifics of installing the components.

I start by installing the active SMD components. In general, for safety, it's recommended to install them last, but my experience shows that modern components have enough protection to be installed at any stage.
For larger components, I solder only the flange first, since it acts as a heatsink. Use a soldering iron with a large tip or good temperature control. Alternatively, raise the temperature a little; otherwise, you'll be soldering for a long time and may end up overheating the component (although modern parts handle that pretty well too).


Once I've secured all the components to the board this way, I solder their remaining leads. You can solder each component completely one at a time, but the leads need a lower tip temperature, so it's easier to solder all the large parts first and then do everything else.
Install the diodes as shown in the photo, with the stripe on the diode aligned with the wide stripe on the board.


When installing large SMD components, be careful not to end up with one component already in place and blocking the installation of another.
I've marked the problem area in the photo. If you install the inductor on the right first, it will get in the way when you try to install the one on the left.
Of course, you can still solder it in, but it'll be less convenient and sometimes quite difficult, so it's best to think ahead about the easiest installation order.


For example, in this case it's better to install the components in either the order 1-2-3 or 1-3-2.
I think the idea is clear.
Here, I also soldered one pad on each component first, then soldered all the rest. But I did it for different reasons.
The thing is, if an inductor is installed the wrong way around or gets in the way of another component, it's much easier to desolder one side than both. Solder usually wicks deep into the joint, and large inductors have a lot of thermal mass, so heating both pads at once is awkward. Heating them one at a time is unsafe, too—you can simply break off the leads.


The kit included several resistors, and some of them are quite accurate—1%.


It's pretty hard to mix them up: there are three 1 kΩ resistors, and telling the 2 kΩ and 12 kΩ resistors apart isn't difficult. Still, just in case, here's a chart for identifying resistor values.


The kit included seven ceramic capacitors, all with the same value, 100 nF. Very convenient—you don't have to hunt around the board to figure out which one goes where.


For convenience, the manufacturer also included indicator LEDs.
One indicates that voltage is present at the input; the other two indicate voltage at the +5 V and −5 V outputs.
The long lead is the anode, the short one is the cathode. In the photo on the right, they're positioned as they should be soldered.
This applies to all single-color LEDs.
While soldering the LEDs, I remembered the Soviet-made AL307s. If I recall correctly, the polarity markings differed between the red and green ones, though after all these years I may have forgotten.


Electrolytic capacitors.
There are a lot of them here, but only three values: 22 µF, 47 µF, and 470 µF. These components are usually installed near the end of assembly, since they're often among the tallest parts.


And finally, the parts usually installed last on the board: the connectors.
Connectors are usually fitted last not because they're large and tall, but because the board is washed before they're installed.
If you wash the board after installing the connectors, flux dissolved by the cleaning solution can seep into a connector, and then you'll spend a long time tracking down the source of a bad connection.
Since I wasn't planning to wash the board, I installed them right away.


I completely forgot to mention one of the components I haven't covered in my reviews before.
I'm talking about resettable fuses.

First, here's what they look like. Many of you have probably seen them on motherboards, security systems, PBXs, and so on. Some of you may even have mistaken them for capacitors at a glance.


The way a polymer fuse works is both simple and complicated.
The simple part is that when it's overloaded, the fuse heats up and the current drops sharply. Once the cause of the overload is removed, the fuse cools down and starts conducting again. Convenient.

The complicated part is that a number of chemical processes take place inside this little thing, which is why these fuses aren't exactly cheap—at least, the branded ones.
This diagram makes the processes inside the fuse easier to explain.


And this graph shows what happens when the current exceeds the limit.
By the way, fuses are available for different voltages. The standard ratings are 30 and 60 (75) V, but there are also 250 V versions, though only for low currents.
The thing is, the size of a fuse is directly related to its trip current, so the highest rating I have at home (and have ever come across) is 9 A at 30 V. For 60 V, the maximum rating is already limited to 6–7 A.

There are three more things to keep in mind:
1. This type of fuse does not completely break the power circuit, so it is not entirely safe.
2. These fuses are slow. Yes, like any other fuse, its trip time depends on the overload current, but it operates more slowly, which you need to take into account.
3. These fuses get a little warm during operation, even when carrying their rated current, and get very hot when limiting the current. So don't install them near components that are sensitive to heat.


The kit included a fuse rated for up to 30 V (the number 30 at the top) and up to 3 A (the number 300 after the type designation, UF).
I usually bend the leads on larger or heat-generating components. Just a habit.


And that's it—the board is fully assembled.


The power input connectors are on one side of the board: a MiniUSB connector and a 5.5/2.1 connector.
The other side has screw terminals and matching pin headers. No need to put jumpers on them :)
Everything is labeled. There are three ground contacts: one for the +3.3 V output, and one each for the 5 V and 12 V pairs. Very well thought out.


I didn't mention this earlier, but while soldering I used only flux-core solder. I didn't use anything else to make soldering easier.
The board solders beautifully: nothing is oxidized, and the finish is excellent. Five stars for that.
This is especially important for anyone who isn't very handy with a soldering iron—the board practically solders itself :)))


There are plated through-holes under the IC and voltage regulators to improve heat transfer to the back of the board.
But there are only three holes per component, which seems too few to me, so I filled them with solder to improve heat transfer.
If I were the manufacturer, I would have done two things differently:
1. Added more vias
2. Used larger pads, especially since there's room on the board.


The board came with six small heatsinks. I don't know where the designer intended them to go, because they get in the way underneath and don't do much on top.


Each heatsink already has a piece of black, double-sided thermal tape stuck to it (if it really is thermal tape).


I decided to stick a heatsink on top after all. I had already added solder to the thermal pads underneath, and now it's much harder to attach a heatsink there because the pads are raised.


I wasn't mistaken when I said I glued on a heatsink, not heatsinks.
The photo simply shows what the board is supposed to look like. But I only glued a heatsink onto the PWM controller, since I didn't see much point in adding heatsinks to the other components.
No, the voltage regulators do get hot, quite a bit, but that didn't affect the test results, and I was happy with things as they were, so I just put the heatsinks back in the bag.


First test power-up. In this case, I powered the board from a USB hub without even checking it for errors beforehand.
But everything went perfectly: no magic smoke escaped, the board started up, and all three LEDs lit up, indicating that the converter was working.
I see that as another plus: solder it, switch it on, and it works.


Further testing was done with a bench power supply, since the board supports a wide input voltage range.
The board starts up at 2.66 V, but it only just starts—the output voltages are too low.


The IC enters regulation at around 3.2 V. Above that, the output voltage stops rising.
According to the datasheet, the IC's minimum input voltage is 3.6 V, which seems about right.
All the outputs are present and within tolerance. Since they're handled by off-the-shelf fixed-voltage linear regulators, I wasn't expecting any problems.


But I ran into some difficulties with the next part of the testing.
Testing all 5 voltages under load in different combinations is very inconvenient.

But since all the voltages are generated by linear regulators, whose output current is the same as their input current (slightly higher due to the regulator's own consumption), we only need to test the DC-DC converter's output voltages.
And since there are only two of those, it's still inconvenient, but much easier than testing all 5.
For the test, I soldered wires to the positive and negative voltage outputs.
Red to positive and blue to negative. In general, I recommend sticking to a convenient color-coding scheme so you don't have to guess later—for example, which one is positive, yellow or purple? :)


We check the output voltage across the 3.2–24 V range on both rails, with no load.
So far, so good: the output voltages barely change.


Next, we load the positive output first, drawing 400 mA.
Both output voltages hold steady.
16.28 V positive and 16.68 V negative.
I apologize in advance for the poor photo quality. I couldn't find a convenient way to fit everything on the table, so I'll comment using the readings.


But loading only the negative rail caused problems.
At 400 mA, the voltage dropped below my 4 V limit (I set the electronic load's minimum protection threshold), and the load shut off.
At 300 mA, it tried to work, but was unstable.
It only worked reliably at 200 mA, but as you can see in the photo, the output voltage sagged from 16 V to 12 V. At that voltage, the 12 V regulator could no longer regulate, and the output was 10.7 V.


For further testing, I added a load to the +12 V rail using two 15 Ω resistors in series, drawing about 400 mA at 12 V.
With no load on the negative rail, there are no problems; everything works fine, with plenty of headroom (I'll explain later).


Now, point by point.
1. 12 V input, positive rail loaded to 400 mA, negative rail to 200 mA: everything is fine.
2. 12 V input, positive rail loaded to 400 mA, negative rail to 400 mA: everything is fine. The converter's voltage has sagged, but is still within the acceptable range.
3. After a while, the voltage started to drop slightly. You can actually see it in the input current: it was 1.35 A and fell to 1.06 A.
4. When the load current on the negative rail was reduced to 300 mA, the converter's output voltage recovered somewhat, but was still below the critical limit.

For a standard linear regulator, the critical limit is a difference of about 2.5 V between input and output. There are low-dropout regulators, but this board uses standard ones.
The 5 V and 3.3 V outputs will work without problems because they have plenty of headroom, but the 12 V output will be low. It needs at least 14.5 V at its input.

This is a good time to return to what I said above about having plenty of headroom.
In the previous photo, the positive rail is loaded to 400 mA, while the input current is only 700 mA.
During testing, I found that no matter how heavily I loaded the converter, its input current never stayed above 1 A for long. If it did rise above 1 A, it would eventually drop back down. This is caused by the PWM controller overheating (even though it has a heatsink). Even just blowing on the heatsink made the voltage start to rise noticeably right away.


After that, I ran a few more tests (I know, you're probably tired of my tests).
In all four cases, the load current on the negative rail was 200 mA.
1. 12 V input, positive output loaded to 400 mA, negative output to 200 mA. Almost everything is fine: the positive DC-DC rail measures 15.16 V instead of 16 V.
2. 24 V input, positive output loaded to 400 mA, negative output to 200 mA. Everything is fine, even better than in the previous test: the positive rail measures 15.96 V instead of 16 V.
3. 15 V input, positive output loaded to 400 mA, negative output to 200 mA. The result is somewhere between the first two cases.
4. 9 V input. Here the voltage sag is obvious: only 12.4 V remains out of 16 V.


Here you can see that when the input to the 12 V linear regulator is only 12.4 V (the last case in the previous test), its output drops to 10.73 V, which is pretty bad.
As I said, it's best to maintain a difference of at least 2.5 V between input and output.


And now a quick ripple test.
Since both rails are usually needed in audio or measurement equipment, ripple levels are quite important.

Ripple measurement on the +12 V output, with only the positive rail loaded.
What can I say? Not great: ripple is around 60–80 mV, which is a lot.
Still, this shows that the linear regulator doesn't handle high-frequency input ripple very well. Our PWM controller runs at a fairly respectable 400 kHz, and SEPIC is also the “noisiest” type of DC-DC converter.
A small inductor between the converter output and the regulator would have improved things noticeably, but there isn't one here.


The same test, but with the negative rail loaded.
Ripple rose to almost 100 mV, unfortunately :(


Now let's see what's happening on the -12 V rail.
The load current is 200 mA, while the positive rail is loaded with the same 400 mA.
At a low input voltage, the noise is lower and has a different character, but the output voltage is clearly too low.
If we raise the input voltage to the point where the linear regulator starts working properly, the ripple increases again to the same 70–90 mV.
Overall, the situation is exactly the same as in the previous test: an inductor is needed.
Or rather, two inductors are needed: one between the converter's positive output and the linear regulators, and another between its negative output and the corresponding linear regulators.

That wraps up the testing, since we can already draw some conclusions. But first, here's the usual photo with a matchbox to give you an idea of the board's size.


Here's what I can say after assembling and testing it.
As for the assembly itself, it's excellent, no question. The board goes together easily, all the holes line up, and there's no need to file or modify anything. It's a real pleasure to assemble. And the board worked as soon as it was assembled.

But I can't sum up how the finished device performs in just a few words.
Let me explain.
The manufacturer says (irrelevant text omitted): 1. Output power = output voltage from various quarters * total output current, input power should be at least> Output power /0.9. For example, a standard power supply 5V, 1A, and to supply the unit, it can provide up to about 4.5W of output power.
2. The machine total output power should not exceed 10W, can provide short-term high current single output, but this time be sure to pay attention to the heat sink temperature, not too high so as not to damage the IC. If you do not have experience, please control the output current of each channel to less than 0.2A is appropriate. Excessively large output current may cause damage to the IC overheat
In plain English, that means:
1. The converter's output power must not exceed the input power multiplied by 0.9. For example, with 5 V and a current of 1 A, we won't get more than 4.5 W at the output.
2. The maximum total output power must not exceed 10 W, or it will overheat. For safety, limit the output current to no more than 0.2 A.

But the specifications also state:
Input voltage: 5-24V DC
Output voltage: + 12V, -12V, + 5V, -5V, +3.3V.
Output current: 300mA (per channel)
The input and output voltages are clear enough, but what about the stated 300 mA per channel?
Since it uses linear regulators, we can simply calculate the power at the converter's 16 V output. That gives us 16 V × 1.5 A (300 mA × 5 channels) = 24 W. Something doesn't add up.
All right, the manufacturer says something about 200 mA, so let's recalculate using 200 mA. That gives us 16 W, which still doesn't work, since the stated maximum power is 10 W…

Anyway, let's leave the math aside and see what the tests tell us in practice.
1. The converter always limits the input current to 1 A because of overheating.
2. The negative rail is weaker than the positive one and needs a load on the positive rail to work properly. In other words, it's best not to load it on its own, or the rails will become unbalanced.
3. In real-world tests, the positive rail can deliver about 400–550 mA on its own.
4. With the negative rail in use, you won't get more than 400 + 200 mA, and even that requires an input voltage of at least 12 V (15–24 V is better).
5. At an input voltage of 5 V, the output current will be just 50–100 mA, at most.

All currents are calculated as the sum of the currents across the channels. For example, a 400 mA limit means something like: -12 V at 100 mA + 5 V at 200 mA + 3.3 V at 100 mA = 400 mA.
Load one channel and you have to reduce the load on another. It's a mess.

In short, this is a very weak converter. It does have its uses, but only with light loads, and the available load current varies. You also need to account for the ripple and voltage drop when only the negative supply is loaded.

As a kit: 5 out of 5. As a product: a 3 minus.

I hope you found this review useful. As always, I look forward to your additions and questions in the comments.
Sorry for the wall of text.

The product was provided by the store for review.