AK4495SEQ + USB converter: building a DAC from a kit

I've built lots of different kits: amplifiers, an oscilloscope, a signal generator, all sorts of measuring instruments and power supplies. Now it's time to try something new for me: building a DAC. In this case, we're talking about an audio DAC (digital-to-analog converter).
As usual, in this review I'll show you the kit itself, the assembly process and some tests. But I'll also try to fit the whole thing into a case afterward :)



The seller offers this DAC in two versions: fully assembled or as a kit for you to build yourself. The price difference is about $7, but in my opinion, the kit is more fun to buy and assemble. Of course, it depends on what you prefer.

I ordered both the kit and the USB converter, but made a mistake: I ordered the converter from a cart where it had been sitting for probably a couple of months, and in the meantime the seller had taken it off sale. In the end, I had to place a separate order with the same seller. The difference is that when ordering through an agent, you'll have to pay a couple of extra dollars to ship the second item. If you buy them together, you won't have to do that: you pay for one parcel whether it contains one item or three. Keep that in mind.

So I received two packages instead of one.


I'll start with the larger package and get to the smaller one when it's time to install its contents.
This one contained a circuit board and two fairly large bags of components.


All the SMD components, voltage regulators and a couple of chips were already soldered onto the circuit board.


The voltage regulators are identical: LM2941
AK4118 — a 192 kHz, 24-bit transceiver made by Asahi Kasei Microdevices
AK4495SEQ — a 32-bit DAC from the same manufacturer.

I'll mention well-known manufacturers by name quite often below. I can't say that all these parts are genuine—in fact, quite the opposite is probably true—but since the names are printed on them, that's what I'll call them. For some components, though, I'll probably say outright that they're counterfeit.


There was no documentation in the kit. For assembly, I used the component values printed on the circuit board and the photos on the product page. The seller has a link to all the documentation on the page, but I couldn't download it because it requires a password. Still, everything seems to have worked out without it; I'll explain the finer points as I describe the assembly.


There are only a few passive SMD components on the underside of the board. The board also had a layer of flux, which was partly helpful and partly a nuisance.


The second package also contained circuit boards, parts, ribbon cables and various other useful things. But I'll tell you about those when I get to assembling them.


The main package contained the components. There were quite a lot of them, and at first I thought assembly would be difficult. In reality, though, it turned out to be fairly straightforward. The power supply board I built recently was probably more difficult.


I'll put the assembly instructions under a spoiler, since not everyone will be interested in them, especially as you can buy the finished product instead.


The assembled setup: the control board connects via a fairly long ribbon cable, so I think it should fit in just about any case. The button board also has a long cable, so it can be placed quite far from the display.
Overall, it looks reasonably good, although I do have some complaints about the assembly and especially the components used.
Some components don't match the markings on the board, and the capacitors (at least) are clearly not genuine.


I applied power and.... nothing. :(
It turned out that the control board can control power to the system, but there's no "power" button, no control transistor, and no connector for a relay.
What's more, there's a spot on the board for an IR receiver, and they even included the parts needed to connect it (a resistor and capacitor), but the receiver itself is missing. It's unclear how to control the power, and shorting the contacts on the board didn't help.
In the end, I simply shorted the pads for the missing transistor, and only then did the board come to life.


For the initial tests, I used the familiar TPP270 transformer with its windings configured as 12+12+(15+15) V.
Don't be alarmed: I just used the headphones to check that it worked at all. Most of the tests were done with an old pair of TDS-15 headphones.


And then I ran into a second problem. The board is quite sensitive to cable quality—or maybe my cable is just complete junk.
The one on the left is the first one I tried. It caused constant clicks in the audio, which got worse as I increased the sample rate, and the device would periodically disconnect from the system.
In the end, I dug a new cable out of a drawer. It looked noticeably better quality, and all the problems disappeared completely. I think the issue was that the XMOS board is powered by the computer over USB, and the thin conductors in the cable were causing these problems.


The seller offers a couple of transformer options for powering the DAC (a large bovine animal), with the following voltage requirements:
One center-tapped winding: 16+16 V
Two independent windings: 8–9 V.

In either case, the transformer is rated at around 30 W, and the windings are rated for 0.6–0.7 A.


I kept wondering why the board needed so much power and what it was supposed to do with it. After all, 30 W is enough for a small power amplifier, not a small DAC.
It turned out that:
1. The dual-rail supply is used only for the op-amp, and the current drawn from the transformer is just 28 mA with no audio and up to 36–38 mA with audio.
2. One of the 8–9 V supplies powers the digital section, i.e. the DAC itself. It draws 164 mA in operation and about 15 mA in "standby" mode.
3. The other 8–9 V supply powers the control board, which draws about 70 mA regardless of operating mode.

So the actual power requirement is around 3.5 W. Even with plenty of headroom, let's say 7 W. 30 W seems like a lot.


I also don't understand why the small op-amp's supply has 2×8200 µF of capacitance. They could just as well have used a battery of capacitors like this; it wouldn't have made much difference. Though I may be wrong.


The controls are very simple: two buttons cycle through the inputs—coaxial S/PDIF, optical S/PDIF, and USB.
When an input is active, the sample rate is shown at the top right, and arrows at the bottom right indicate that a signal is present.
There are also four digital filter modes, cycled through with a single button:
Short Slow
Short Roll
Slow Roll
Sharp Roll.

In practice, two buttons would have been enough, since the inputs are cycled through anyway; the buttons only determine the order of switching.


Before moving on to the tests, I decided to install the whole assembly in a case. I grabbed some old satellite receiver; in this case, it was a Star Track.


It was already half dismantled, just lying around and waiting for its fate.


I removed the old internals and test-fitted the new board.


Of course, nobody expected everything to line up. The only holes that actually matched were those for the RCA connectors, so I used those as my reference and adjusted the rest to fit.


I also had a second case option, probably even more convenient inside, but there's nowhere to put the display in this one. Well, I’ll say that it’s needed, and a few LEDs would really have been enough, but I couldn’t get them connected.


The seller’s page also has a case option offered for boards like this, although this particular version doesn’t have a hole for the USB connector.


It certainly looks nicer than my version, and the space inside is laid out more sensibly. The case is also fairly inexpensive, but it’s heavy, so shipping would wipe out any savings.


For obvious reasons, the holes in the rear panel didn’t line up with the ones I needed, especially the USB and 3.5 mm jack holes.

A quick note about the connectors. S/PDIF and USB are fairly self-explanatory, but here’s some more detail on RCA and 3.5 mm. They’re simply analog audio outputs. I tried plugging headphones into the jack; they work reasonably well and even get quite loud, although that’s not really what the line-level output is intended for. I connected a power amplifier to the RCA output.


Grab a conical drill bit and a needle file, and “saw away, Shura, saw away.”


Of course, it looks pretty rough, and a trim panel wouldn’t hurt. I’ll have to come up with something, but it’s more than good enough for testing. I later secured the RCA and optical connectors with screws as well.


The board is mounted on 6 mm standoffs, although I had to shorten one to 5 mm by filing off the excess. I mounted the control board on the standoffs that came with it.


Next up was the front panel. I removed it and took out the original board.


And then came the disappointment. Not only does the plastic inside partially cover the window—I cut that away later—but, worse, part of the plastic in the front window is painted black :(


But no need to get upset—we improvise :)
In the end, I positioned the display so the main window shows the selected input and filter mode, while the smaller window shows the sampling frequency. It was a very tight fit, of course, and part of one character got cut off, but this was the best solution I could come up with.
The display included in the kit is the cheapest kind, and it’s almost impossible to read through the tinted filter, even with the flash on.


I really wasn’t happy with this display arrangement. I’m not claiming the end result will be a high-end device, but I wanted it to look a bit nicer, and the generic display spoiled the whole thing. Ideally, I wanted a VFD display—I love those—but they’re hard to find and the prices are ridiculous, so I bought a compromise option, OLED 1602. It costs about 12 dollars; the model is WEH001602ALPP5N00001, datasheet. I bought it from Kosmodrom in Kharkiv because I just couldn’t find them for sale in China.

I’d actually already used this kind of display in a powerful power supply, but that was the larger version.


Another reason it suited me was that the backlight doesn’t stick out beyond the metal shield, and the display is shorter. It fit the front panel better, and I thought I might be able to mount it somehow using the old board.


Wait a minute, you might say. That’s clearly not a DAC in the photo, but something else entirely. That’s right: it’s a transistor tester, which I also reviewed.

It’s simple: this display just refused to work with this DAC—not at all. The screen stayed completely black, and at first I didn’t even realize that it had no backlight, so there was no way to tell whether it was getting power. Then I tried it with other devices, and it started up without any problems with the old transistor tester.

The image doesn’t fade even when photographed almost straight on with the flash. That’s exactly what I wanted: sharp and high-contrast. Ideally, I wanted a white glow, but only yellow and green were available, so I had to go with yellow.


Next, I started testing it with different devices and reading the datasheet. Here’s what I found:
1. It only worked for me with the transistor tester.
2. When I tried connecting it to the 6005 board in the power supply (essentially the same as the 6020, where I used the larger display), the screen showed gibberish.
3. When I tried installing it in place of the 2004 in the LCR tester, it showed nonsense on the screen. It turned out that it was displaying the text mirrored, with some errors. A regular 1602 only showed parts of the first two lines, but displayed them correctly.
4. When I disconnected data lines 0–3, the DAC displayed a blinking cursor in the bottom-right character position.

I then started studying the datasheet and found that the display can operate in different modes, selected by jumpers (at the bottom left of the board). I compared their positions with those on the large OLED 1602; everything matches, but it still doesn't work.

I already knew about some initialization problems with these displays. In various DIY projects, though, people seem to have no trouble at all, and can even adjust the brightness by changing the voltage on the contrast-control input (with a small modification, I think). But I never expected this kind of snag :(


But the contrast and viewing angles are excellent. This is probably the kind of option you go for when you want something good and don't mind spending the money. It's a shame it wouldn't work here, but maybe someone can tell me what I'm doing wrong and whether there's still a way to get it running. I'd be very grateful.


Anyway, I managed to fit the original display somehow. At first I planned to attach it to the main front-panel board, but I had cut off the tab that was right where the display now sits, so I had to make do with hot glue. I shortened the board itself, since I still need the buttons.


Something like this. You can't see how the display is mounted here: I didn't save the photo, and the case is already assembled. The right side is simply glued in place (though I had to cut away a little of the case plastic); the left side is held down by a strip, also attached with hot glue.


I bought a few extra parts for the rest of the build. A prototyping board for a little over a dollar and a transformer for about five dollars. I had some capacitors lying around from various power supplies, and I desoldered the inductor, fuse, and connector from a power supply in a tuner of roughly the same type :)


I don't usually use prototyping boards, but in this case it was more convenient. I went with a 90×150 mm board so everything would fit on it and it would fit in place of the tuner's original board. I made it wider to accommodate the transformers.


The board is double-sided with plated-through holes, but I'll say right away that it will carry mains voltage, and it isn't designed for that. I'll replace it with a proper board later; I'll have to think of something else to put on it. It's best not to route mains voltage across it because the clearance between the pads is too small. I drilled out a few pads in critical areas to improve safety.


The transformer, on the other hand, was a perfect fit. I had one lying around for years, from a small subwoofer amplifier: 2×15 V, 0.6–0.7 A. So I bought a second one rated at 12 V. There was a 9 V transformer too, but it cost a little more. This transformer is very neatly made, and now I'm tempted to replace my old one with the same model, but rated at 18 V: that would give me 12+12 V for the control and digital sections, and 18+18 V for the analog section. The reason is that, in my opinion, 15 V is cutting it close for regulators that need to provide 15 V. My DC voltage came out to only about 21 V, while the capacitors are rated up to 35 V, so why not make use of that headroom?


There's also some free space near the DAC board where I could install a proper headphone amplifier, probably a Class A design. But for now these are just ideas; maybe someone can suggest something else to improve it.


This is what the finished assembly looks like. The power supply consists of two transformers, and I'll keep it that way; I won't use one shared transformer for everything. I might even power the digital and control sections from separate transformers to completely eliminate any interaction between them, but that's more of a “thinking out loud” idea.
The mains input has a filter consisting of a 40 µH common-mode choke and two capacitors: 0.22 µF on the mains side and 0.47 µF on the transformer side. The capacitors are Pilkor.
The display is connected with a ribbon cable so the front panel is easier to remove for modifications. I soldered a pin header to the top of the board for this; the socket is on the other side, so I can test displays by plugging them directly into the board.
The wires aren't routed very neatly; I'll fix that. I'll probably also separate part of the power supply from the rest with a metal shield. One option is an L-shaped shield that separates the DAC from everything else in the case. I put a ferrite filter on the wire from the control panel to the main board, at the main-board end.
And I'll need a blanking panel, since my inner perfectionist won't put up with holes in the back of the case for long :)


I wouldn't say it's all fantastic, but considering the case was free, it's not bad at all. Two things bother me: the ugly control buttons and the blue display. It actually looks very clear; the flash makes it look dim, but I just don't like it. :(
The left/right buttons switch inputs, and Menu selects the digital filter mode. The power button isn't connected yet; for now, power is controlled by the switch on the back of the case. I'll sort this out either by installing a separate power-control module operated by the original power button, or by finding out how the original board handles it.

The previous owner took such good care of the tuner that I was the one who ended up peeling the film off the front panel. Many thanks to him for that :)


Since the tuner has an extra compartment for access cards, and there's nothing in it now, I plan to install headphone jacks here—both 3.5 and 6.3 mm, just in case—as well as a volume control, so I can adjust the analog signal rather than the digital one. I think that should work better. The little cover will come in handy, since I won't use it often.


To test the DAC, I had to download and install the drivers. You can get them here.
I tried two drivers and settled on this one: Thesycon JLsounds_USBAudio_v4.41


Information about the sound card and driver


I immediately disabled audio processing in the settings and switched the operating mode to 24-bit, 192 kHz.


As far as I can tell, the XMOS-USB-Audio-(v3.34.0) driver supports ASIO mode, but I couldn't get it to work.


And now for a few tests. I fully understand that RMAA tests aren't the gold standard for accuracy, but it's all I can use for now. Also, I ran the tests using my built-in sound card, since I don't have an alternative. I'll try to test it with better equipment when I can, but this is what I've got for now.
I tested the 16/44.1, 24/48, and 24/192 modes. The last two gave similar results, so I'll include them together in the summary table. After that, I'll show only the 16/44.1 and 24/48 tests.

By the way, there's an alternative review of a device with this board. It has that soft, fluffy sound :)




The analog output is driven directly by the LME49860NA op-amp. According to the manufacturer, it's a high-quality op-amp designed specifically for audio applications.


But knowing how much the Chinese love all kinds of fakes and clones, I bought an original—at least, that's what they confirmed at the same "Kosmodrom" store. I specifically asked, and they said they work directly with the manufacturers. The price also seems to back that up: about 7 dollars (200 UAH) — link. By the way, I came across one for 60 UAH (2 dollars) somewhere, but I have doubts about its authenticity.
To be fair, I buy quite a variety of components, and this store is probably one of the few where I've encountered counterfeit parts least often—and even then, it was quite a while ago. That's my answer to the question I get so often: where do I usually buy electronic components?

The chip on the left is from Taobao; the one on the right was bought at a store. The package material is slightly different, and the marking on the right-hand chip is crisp, while the one on the left is much harder to make out, even though I tried cleaning it with alcohol before taking the photo.


Honestly, I more or less expected RMAA to show no difference, since my sound card is noticeably worse than this DAC. As we know, measurement equipment needs better accuracy and resolution than the device being tested.
But I can also say that I did notice a difference by ear. I swapped the chips back and forth while listening to the same track, but that's a subjective test, so it doesn't really count as evidence :(





But if you think I stopped there, you're very much mistaken. Since I was already buying a display, transformer, and so on, I figured I might as well also picked up a few capacitors to replace some of the ones in this DAC. I just didn't like the originals, that's all.
In this case, I bought two capacitance values:
22 µF 63 V - link
220 µF 35 V - link.

Both are Samwha WL series. Unfortunately, there were no lower-voltage options available, such as 16 or 25 V, so I had to buy 35 and 63 V versions.


The WL series consists of low-impedance capacitors. They're a little better than the RD series I'm used to, but I use them less often, mainly because there's less choice available.


I measured their characteristics mostly out of curiosity. Since I planned to use them to replace the 22 µF 25 V and 100 µF 25 V capacitors, I'll compare the measured parameters right away:
22 µF - "Elna" - 16 µF, 2 Ω at 7.8 kHz
22 µF Samwha WL - 20.5 µF, 0.46 Ω at 7.8 kHz

100 µF - "Nichicon" - 87 µF, 0.216 Ω at 7.8 kHz
220 µF Samwha WL - 242 µF, 0.093 Ω at 7.8 kHz


Replacing them wasn't entirely straightforward: the ground plane on the board has quite a lot of thermal mass. Still, I think it was worth it, if not for the sound, then certainly for long-term reliability. I might replace the remaining capacitors later, except for the large 4200 µF ones; I'm quite happy with those, though who knows.
By the way, a bunch of 22 µF capacitors are used for decoupling the DAC's power supply, and almost all of them are connected in parallel, so replacing them could well affect its performance. They're also in the op-amp's power supply circuit.


As I expected, there's no difference in the results—or I just can't see it.




Summary of test results:
Red - original configuration
Green - after replacing the op-amp
Blue - after replacing the capacitors.


Video version of the review


Now I'll try to sum things up.
Overall, this is a decent kit. It goes together without much trouble and sounds pretty good. I tried it with various tracks, listening both through headphones and through the Dartzeel amplifier clone I reviewed recently. There are four digital filters, but I can't pick a favorite yet. I think I just need more time and a quiet setting to figure out which one I like best.
My biggest concerns were about whether the components used were genuine. For the most part, almost everything here is a clone, or so I think. First of all, I'd recommend using decent capacitors from the start. Since most of them are in the power supply rather than the signal path, I'd focus more on the capacitors' ESR. But that's just my opinion, and I may be wrong.

The control board is supposed to support power control, but that feature hasn't been implemented, nor has remote control. I'll still try to get hold of the assembly documentation and find out how to make all this work. As for the display included in the kit, I really didn't like it, and I couldn't replace it with a better OLED either. (

Speaking of the documentation, I don't understand why the seller password-protected the file downloads. In the end, I had to spend more time assembling the kit, sometimes figuring out what went where from the photos in the store :)

In short, the kit left me with more positive than negative impressions. The sound is pretty good, though I don't have much to compare it with yet. Maybe I'll get my hands on something better down the road and decide that this one is really bad, but for now I'm satisfied. The one real fly in the ointment is the quality of the components.

This review was sponsored by the forwarding service yoybuy.com, which covered the shipping costs.
The kit cost $46.88 including shipping to the forwarding service. The kit weighs 330 grams. The XMOS board cost $26.56, but I had to pay another $2 for shipping; it weighs 20 grams. Shipping from the forwarding service depends on various factors, such as the quantity and whether there are other items in the order.