Saturday 8 October 2022

MEASUREMENTS: "CheapDAC'22" - Cheapest DAC in 2022? US$3.57 S/PDIF DAC from AliExpress. (Does objective analysis correlate with sound quality? Of course! And Lavorgna's DUT comment - "HiFi" is not art.)

"CheapDAC'22" - box, pamphlet, optical cable, USB 5V power cable.

While we've recently been examining high performance, high-fidelity DACs with amazing measurements, once awhile, I think it is good to have a peek at what "low-end" performance looks and sounds like. All the more important I think because audiophile magazines and "high end" online sites often seem to be reviewing increasingly more expensive "bling" these days, typically with just subjective comments. Needless to say, many expensive products do not appear to actually improve fidelity even if externally they might look great.

BigGuy in July pointed out to me one of the cheapest DAC packages ever seen on AliExpress - US$3.57; no tax, free shipping, arrived from China in about 5 weeks! For simplicity and lack of a better name, let's just call this generic device the "CheapDAC'22" (for 2022).

[With inflation lately, I see the price has increased since July as of this writing.]

The DAC is capable of both S/PDIF (Coaxial + TosLink) and Bluetooth input even (appears to be basically SBC codec, nothing fancy). The advertising even mentions 24/192 input.

Well, at this price, how could I resist not giving it a listen/test dear music lovers and audiophiles!? :-)

As you can see in the image above, the package is basic. Thin black plastic box with white lettering to indicate the in/outs. You connect the included USB cable to the 5V power connector. It even includes the thinnest TosLink cable in the world; how very thoughtful of them.

For completeness, here's the recent AliExpress ad:

Notice the $3.00 coupon offer... You might be able to even get this DAC for $1?

The Specifications tab did not offer much insight.

On the other side of the box are the Coaxial and TosLink S/PDIF inputs, along with a tiny red power LED when plugged in:

There's a single screw holding the box together, so removing that we can see the very simple (as expected) circuit board:

Other than the 26.000MHz clock and a few assorted passive components, it's basically a single chip board. The chip labelling is small and I can't quite identify the part number. Best I could make out, the chip is labelled "PTQA15E5B / UH7595 / 2141" across three rows. The "A3" logo tells me it's from Shenzhen Bluetrum Tech, but I don't see any online reference to this specific part.

I did not test the Bluetooth input but can confirm that my phone sees the pairing device named "YQ-88C".

Let's see how this performs on the test bench and have a listen later!

Here it is sitting on the testbench, powered on with a 5V lithium battery pack (not shown). I will use the Logitech Squeezebox Touch as S/PDIF coaxial / TosLink source for much of these tests except when running stepped measurements at which point I used the S/PDIF outputs of the Topping D10s plugged into my measurement computer. For ease of measurements, I'll use the RME ADI-2 Pro FS for these tests; as you'll see soon, given the level of performance, the E1DA Cosmos ADC and APU were not needed.

Connecting the Touch's TosLink output to the CheapDAC'22, here's the 0dBFS 1kHz sine output and -3dB square wave into the digital oscilloscope:

The 0dBFS sine output has a voltage of around 875mVrms, a bit lower than the typical 2Vrms we see from RCA outs of better DACs these days. Channel balance looks okay. We can see from the bandlimited (up to 20kHz) and unlimited 1kHz square waves that they are using a linear phase filter with pre- and post-ringing. I notice that there appears to be fluctuating relative phase between the left (yellow unlimited squarewave) and right (cyan limited squarewave) channels resulting in inconsistent overlap between the two. I have not seen this before in other DACs.

Now, let's have a look at the "Digital Filter Composite" (based on the Reis Test):

I took the opportunity to modernize the test a bit by using REW and 8x average instead of WaveSpectra which I have used for years. Notice I've put the cursor close to the Nyquist 22.05kHz. (Used the RME ADI-2 Pro FS for measurement here, hence rising ultrasonic noise level into 90kHz.)

We're seeing evidence of a slowish roll-off filter (about -12dB at 22.05kHz) with overload characteristics (0dBFS curve). The 19 & 20kHz test tone is showing quite a bit of intermodulation and harmonic content into the ultrasonic frequencies.

Let's have a look at the 16/44.1 impulse response:

As expected, linear phase impulse with some pre- and post-ringing. Notice that there's a little asymmetry with a slight DC shift coming out of the impulse response which we do not see with better DACs. It does maintain absolute phase.

Already, we can tell that the CheapDAC'22 is not a particularly "high fidelity" device. I'm seeing relative time domain fluctuating between right and left channels when playing different signals in each channel, distortion is higher than one would like to see, and certainly not a strong digital filter.

DAC Resolution:

Let's go deeper into the resolution of this DAC. As usual, we start with 16/44.1 using RightMark Audio Analyzer with comparisons to old CD players (Technics and Sony CDP), an inexpensive but good DAC (Topping D10s) and basically a state-of-the-art fidelity device (Sabaj A20d-2022):

As you can see the CheapDAC'22 is not like the others. Most egregious is that nasty looking frequency response with significant bass roll-off! The noise floor is not great but passable as a 16-bit device and the distortion amount is clearly above the comparison devices.

Let's move to "Hi-Res" 24/96. This time, we can throw in the Cayin RU6 R2R dongle (oversampled mode), and the Chord Mojo 1 into the mix.

I've tried to make it an "apples-to-apples" comparison using RCA outputs only and for the Chord and Sabaj, used TosLink S/PDIF rather than USB.

Even though the CheapDAC'22 can accept 24/96 data, clearly the frequency response is unable to extend much beyond 20kHz nor is the noise floor improved. There's also an unusual notch in the frequency response above 10kHz. In effect, this isn't a hi-res DAC. Notice how noisy the Cayin RU6 R2R DAC is along with even higher IMD than the CheapDAC'22.

Finally, since it's advertised, let's send some 24/192 audio test data over, compared to Topping D10s:

Again, this is not a hi-res DAC - even though it can accept 24/192, the data appears to be resampled and maybe truncated. Frequency response sucks, noise level is around 16-bits. Notice the significant amount of IMD and THD distortions compared to the D10s. So let's delve into these distortions now...

1kHz 0dBFS THD+N:

Here's a more detailed look at a 1kHz 0dBFS sine tone:

We're looking at a THD+N of around -70dB. Lots of "skirting" at the base of the 1kHz tone bodes poorly for jitter. Obviously not "good" in any way compared to more capable modern DACs. Notice that the coaxial input results are a little worse than TosLink in my sample here. We see this trend in the RightMark results above also. Given this, I'll just focus on the slightly better TosLink input with most of the tests below.

I was curious if pulling the volume down would improve things like the skirting - here's 1kHz -6dBFS, right and left channels overlaid using multichannel REW Pro:

Alas, no real improvement. The right and left channels measured about the same as you can see the red and blue colors overlaid. I guess it's good that the harmonic "cascade" generally follows a pattern of decreasing intensity into higher order.

Stepped Frequency & Amplitude Tests:

Time to run some of the more complex stepped tests. First, let's have a look at the frequency step (20-20kHz) graphs at different output levels from 0dBFS to -12dBFS:

Harmonic distortions are high across the frequencies, but higher in the bass frequencies likely related to the roll-off. 2nd and 3rd harmonics tend to be highest as we saw in the 1kHz FFT. Notice that the harmonics do not drop much as the output level is reduced down to -12dBFS. High quality DACs tend to show harmonics getting buried into the noise floor by -12dBFS.

Now let's have a peek at the 1kHz level step:

Again, this is clearly not a high-resolution DAC with minimal signal-to-noise by -90dBFS. With this DAC, we see about 1dB linearity deviation by -80dBFS. This is poor. 

Multitone Tests:

Here's the Triple-Tone TD+N (using my typical frequencies 48/960/5472Hz, 1:1:1):

Lots of intermodulation and harmonic distortion products on display. Notice the bass roll-off resulting in reduced strength of the 48Hz peak.

TD+N average result of around -51dB is poor, the kind of result one might see in a lower quality, utility amplifier.

Here's the 1/10-Decade Multitone 32:

The noise and distortion-free range is around 50dB across the audible spectrum. Again in the multitone signal, we see the unfortunate bass "sag".


Ugly J-Test. The FFT is littered with anomalies which are not just jitter sidebands. Lots of "skirting" with the 24-bit J-Test suggesting a significant amount of low-level random jitter. Not sure I've seen anything worse than this over the years in my measurements.

How audible this would be when playing music I think is still hard to know.

Subjective Impression & Summary:

Normally with these reviews, I'll have a "subjective" section talking about a few recent albums that I used for the DAC review, etc. For this device, let's dispense with my personal listening impressions and instead let me offer the Archimago's Musical Performance Track (AMPT) from the CheapDAC'22 (TosLink input) to listen for yourself:

So, what do you hear?

It should be quite evident that this is not a high-resolution DAC as discussed above with the objective measurements. For me, the loss of low-frequency content is obvious when listening for example to the bass on Benjamin Clementine's "Winston Churchill's Boy". Clearly missing the deep "body" in the percussion backing track and his vocals. At times, the treble for instruments like on "Japanese Roots" sounds "nervous" to me, lacking a bit of the nuances and not as well localized soundstage placement. The clapping of the live audience in the Ray Brown Trio track is not as clean nor conveyed a sense of "presence" in the venue. Eva Cassidy's vocal tonality sounds harsher with unpleasantly accentuated sibilance.

Given the audible differences, we can even look at the waveforms and note the differences between the CheapDAC'22 and others:

Each of the waveforms have been normalized to the same output level at 1kHz. Notice the similarities of the old Technics CD player and the other good high-fidelity modern USB Topping DACs. In comparison, we see quite a difference with the CheapDAC'22! In part, this is due to the frequency response but notice also a lack of "control" during playback of the last 2 tracks which have been dynamically compressed; the CheapDAC'22 shows stronger spikes in the output during peak-limited portions of the audio.

Given that the non-flat frequency response is quite audible, if you have one of these DACs, you can apply a simple low-shelf EQ to boost the bass. This setting in Roon should get you essentially flat down to 50Hz and around -6dB by 20Hz, which isn't bad:

Make sure not to push volumes too high, adding +13dB EQ will further increase distortion in these low frequencies. Human hearing is quite tolerant of moderate amounts of bass distortion.

I think it's fun measuring and listening to a device like this once awhile to compare with the best performing DACs I've come across like the Topping D90SE, SMSL DO100, and Sabaj A20d 2022 recently.

There's a huge improvement comparing an inexpensive but good DAC like the Topping D10s (~US$100) with the CheapDAC'22. If we then compare the D10s performance with those more expensive DACs linked above, we'll notice that the objective improvements are certainly not staggering. Given the limits of human hearing, I trust this is a good example using objective data of the concept of "diminishing returns" when it comes to price increase and evident fidelity (let's ignore the silliness of "accelerating returns").

Clearly at US$4, there's no way this DAC would have been competitive with a "high-fidelity" device. Nonetheless, I think it's good to ask ourselves whether the sound quality can still be "enjoyable". Are you surprised by the sound of that AMPT recording? Do you think there are situations where this kind of sound would actually be good enough, or might even be "synergistic" within a system?

Subjectively, some might actually prefer the "timbral/harmonic richness" of the sound from this box, and the higher noise level might even be experienced as "atmospheric" especially with acoustic recordings with minimum bass content! There are similarities between the CheapDAC'22 and the modded tube Oppo from a number of years back.


Over the years, there have been questions raised about the value of objective measurements and audio. Often it has been suggested that "measurements don't matter" and that "objective results don't correlate with subjective experience". IMO, anyone who spends even just a little bit of time listening will very quickly realize that of course objective results can be audible.

The issue is not whether it does, but rather, for any one person, with one's equipment, and in one's sound room, how much the objective change matters. Let's highlight these items:

1. What kind of equipment is in the audio system? Yes, sometimes we might not hear differences because our equipment isn't of a high enough resolution. No, this is not the same as when people argue that "Cables make a difference, you can't hear it because you gear is not high enough quality!" (there is no good objective evidence that decent cables sound different). Rather, as a more objective audiophile, one can be more specific in predicting audibility and measurement results. For example, if you're a bookshelf guy with a pair of BBC LS3/5a speakers (or KEF LS50 Meta for that matter) and no subwoofer, the CheapDAC'22 bass roll-off might not be too noticeable.

2. Related to the above, we need to make sure our room acoustics, which really is an important part of the total system, doesn't obscure audible differences. Rooms with high reverberation times will affect how well we perceive transients. Rooms with high ambient noise will not let us truly appreciate low-level nuances in high-resolution recordings.

3. How are we as listeners? Human ears and brains are non-linear devices (as discussed years ago). Furthermore, hearing deteriorates over a lifetime. Our minds are flexible and can easily accommodate to subtle and not-so-subtle imperfections. This is not a bad thing! It allows us to get past the surface noise of vinyl, and still enjoy the music even if there are also crackles or scratches here and there. As we get older, potential noise-induced high frequency loss and insidious presbycusis might affect acuity and perceived tonal balance. Even if we were to have "Golden Ears" at one point in life, this is obviously not expected to be indefinite.

4. Then there are the truly subjective cognitive and emotional preferences we all harbor. This could be in the form of conscious prejudices whether it be positive ("I love tubes!") or negative ("I hate solid-state sound!"). I suspect there are just as many subconscious and non-declarative biases. For example, I suspect at some point in all our lives as audiophiles and music lovers, we played around with the "smiley curve" EQ and enjoyed the accentuated bass and upper-end sparkle until at some point we discovered a more natural, "accurate" sound. We might have chased "loud" until the day we discovered the beauty of nuances in the music and now aim to collect well-mastered recordings with better dynamic range. Whether this "change of heart" was purely the result of our own explorations or could have been something we picked up hanging around like-minded audiophiles, I suppose it doesn't matter and we might not consciously know when these changes in preferences happened.

When it comes to subjective preferences, it's obviously not about "right" or "wrong" just as much as nobody can tell us which genre of music we "should" enjoy or whether a girlfriend or boyfriend "should" be beautiful in one's eyes.

I bet you somewhere out there, there are folks with excessively bass-heavy systems in reverberant rooms who would listen to this CheapDAC'22 and be amazed by the clarity of the mids and treble frequencies. ;-)


Over the years, I have not been shy about expressing issues with the ideas promoted by people like Michael Lavorgna (like here, here) and others in the audiophile press; there's no need to be shy since a lot of this stuff demands to be questioned and clarified.

The other day, against my better judgment, a reader asked me to have a look at his "DUT: This Is Not A Test" article. I rarely call equipment "DUT" on these pages, simply because the terminology never spoke to me either (I'm not an engineer after all!). That's fine. The problem is that he doesn't seem to believe that the role of the hardware reviewer is to "test" the device, apparently.

So that we're all on the same page, let's find a definition for "test":
test /test/
1. a procedure intended to establish the quality, performance, or reliability of something, especially before it is taken into widespread use.
take measures to check the quality, performance, or reliability of 
(something), especially before putting it into widespread use or practice.

That sounds right both as noun and verb. When I read reviews, am I not interested in knowing about the "quality, performance, or reliability" of something? Once I finish reading the review, should I not walk away having learned something factually true about these characteristics and hopefully the reviewer has done a good job contextualizing the findings so I can practically apply what I learned? So why aren't the best reviews actually "tests" of the gear that answers these questions? Lavorgna would prefer that we follow the path of reviews being "experiential in nature, so the value of a hifi, painting, movie, or meal lies in their consumption" (emphasis mine). Herein lies the confusion.

He seems to indicate that in order to judge the value of something, we need to directly experience the art ourselves which is what paintings, movies, and even meals are. That is true. But doesn't "hifi" belong in another category? We can put "album" and "music" with "painting" and "movie" but not "hifi" (the equipment used to reproduce the art)!

Let's use a visual analogy. In the image below of the Mona Lisa in the Louvre, where is the art to be experienced, and which are the support structures used to display the art?

Obviously, the art is in the center, da Vinci's Ms. Mona Lisa herself. Everything else around it is meant for the presentation (and safety) of the work of art.

So too with "hifi" - is the "art" the audio system with its speakers, amp, source, and cables? Would bigger speakers, massive amps, fancier cables, more expensive streamers make the "art" that much better? Not necessarily because there is a threshold beyond which sonic differences are simply trivial!

Our "hifi" gear is not the Mona Lisa painting, it's the railings, the bulletproof glass, the size of the room. Just as the frame around the painting might be beautifully embellished, some audiophile gear can also look wonderful as pieces of industrial design. However, I think we must be careful not to lose track of the music itself, otherwise we might end up in a rather unhealthy "fetishistic" state of adoring the gear rather than the art which it was meant to reproduce.

When a "high fidelity" audiophile (which can be defined as one seeking "transparency" to the source material) sees the conditions of a room looking like the one above, we know quite well what we need to do to best experience the art (painting)! We would take away the front railings so we can get closer, get all those people out (ambient noise level too loud, too much pushing and shoving, distracting!), and perhaps most importantly, literally "remove the veil" of the slightly dark-tinted bulletproof glass. Each of these dimensions can be "measured" like how loud the room is, how close we are to the art to fill our field of view to see nuances, how much light attenuation there is between the viewer and the artwork through the glass.

Isn't that the same for audiophile reviews when we think about the equipment we buy? Don't we want gear that gets us "closer" to the art, that doesn't distort or dim the reproduction of the music? Is there then anything wrong with testing each device to make sure the resolving ability is high?

I really don't care what "art" Lavorgna enjoys (I find his taste in art rather uninteresting if not a little disturbing/grotesque at times). But when he brings gear into the "barn" for review, are his writings on those devices insightful? Is he capable of accurately judging the limitations of the equipment? Is he teaching us anything interesting or useful about the gear other than superficially repeating what the company says in most of those articles? Is he able to detect problems when he's "consuming" that "hifi"?

He wants us to believe rhetorically that "If Mr. X has heard 3 DACs and Mr. Y 300, whose opinion holds more weight in terms of how a given DAC stands up to the competition?". Is it supposed to be the guy with bigger numbers? Surely it's not that simple! I would rather hear the opinion of a thoughtful 25 year old who is listening to his 3rd high quality DAC than a geriatric fella well past the prime of his auditory acuity even if he has heard 300 DACs since the 1980's - likely most of those devices antiquated and out of production anyways. (Not to mention auditory memory isn't that precise to begin with.)

Mr. Lavorgna has not given any indication that his tastes are actually "hifi". He loves the TotalDAC stuff for example which I can't say is known to be "high fidelity" despite the price tag. For all we know, since Lavorgna has never shown himself to be bothered to anchor his opinions with objective reality, when asked to improve the Mona Lisa experience in the room above, he might switch out the clear bulletproof glass for something with a slight warm tint to satisfy some kind of "euphonophilic" subjective preference (like the TotalDAC). There's nothing wrong with that since we can be idiosyncratic if we choose, even though it's not strictly "hifi". This is the nature and limitation of pure subjectivism across the board.


To end, speaking of the importance of questioning and clarifying things - and I know I'm deviating into the thorny reaches of politics and ideologies here. Did you guys catch the recent Kanye "Ye" West interview on Fox News (especially part 2)? Does he know that Big Pharma isn't "Big Farma" (20:20) buying up little local farms as some kind of solution? There's much more to be said about the pseudo-intellect on gross display, but I'll just stop even at that.

Please Americans, let this not be your image of what conservatism or Christianity should look like with his disturbing prosperity gospel and weird "bipolar" viewpoint. And please, let's not unleash upon the world a dystopian future with this kind of character as the POTUS (which he states he'll be one day as if a prophesy). Musician, fashion designer, even reality TV personality are good enough roles for this unusual but entertaining man.

Like I said before, over the decades, with much of audiophilia turning into an extreme subjective exercise which deviated from reasonable reality-testing and the media not helping to educate hobbyists on the science, the hobby has created echo-chambers of crazy talk and bizarre beliefs. This has created a breeding ground for snake oil salesmen of all stripes over the years. While I hope to play my small part in "remediating" this hobby, I see it as a microcosm of what society at large could also become as a result of failing public educational systems, excessive emphasis on personal "subjective truth" and the increasing influence of "confidence men" seeking power and fame for egotistical satisfaction.

Let's hope that noble Western democracies truly do not devolve further into ignoble idiocracies. We have many real problems in this world already and do not need to create more.

Alright audiophiles, I hope you're all doing well and enjoying the music (above and beyond just the "hifi" system).

Happy Canadian Thanksgiving this weekend!


  1. Not sure this is useful aside from figuring out it is basically broken in regards to FR. Need to find the cheap DAC with a flat FR and then do the listening test

    1. Hey there Booty,
      Plenty of cheap flat FR DACs out there to try!

      Here are a couple over the years:

      Although they cost >$4 ;-).

      And we already pretty well know that it's hard to differentiate DACs within reason in a blind test:

      With predictable results based on the noise level in that instance (and frequency response here).

      Maybe CheapDAC'23 will have a flat FR at $4 as a demonstration of progress. ;-)

  2. Not that much worse than the Schiit Modi 2/2U which was $100+ from years ago.

    1. Nice one Arthur, LOL.
      Had a look at Amir's measurements of that from 2018:

      Looks like the CheapDAC'22 is still *that* much worse ;-).

      I assume the FR was at least decent on that old Schiit? Weird that in that old thread I don't see a simple frequency response - way more important than the distortion results.

  3. Finally, a modern DAC that measures as poorly as the Cayin RU6! But surely the RU6 must have sounded subjectively better (at over 50 times the cost)? :-)

    1. Hey there AudioPhil,
      Yes, indeed the Cayin sounded better ;-).

      On just as serious a note, that unfortunate frequency response is really the problem here. If it weren't for that, this probably could sound excellent like a tube DAC at 1000x the price!

  4. Hi,

    I got some good measurements with Apple Headphone Dongle. It is my go to item to introduce people to high performance audio. With either my AKG K450 or K371 headphones. If they want a little more my LG V60 ThinQ (no MQA) does the job quite well.


    1. Hey R66indierock,
      Yup, the Apple dongle is pretty good and likewise the LG phones. I believe LG exited the market, right?

  5. Well, proves there is indeed a limit on price for a DAC where there will indeed be audible degredation of the sound output. But I don't think we are too far off from the day where a $15-20 DAC will deliver all the sonic performance human hearing is capable of rendering. The sooner the better IMHO. The day when a person can buy a $50 DAC/AMP for his $200 HD6XXs will be a breakthrough in the availability of true high end sound available for everyone.

    1. Hi Phoenix,
      Yeah, I think the time is fast coming when at $20, you should be able to get a decent headphone amp that is objectively capable of "hi-res" playback with clearly >16-bit dynamic range. I'll certainly keep an eye out for something like that!

      If anyone knows of one I should test, E-mail me (address in the upper banner).

  6. Which audio forums other than audiophilestyle are you active on?

    1. Hey REXP,
      I also hang out quite a bit at Steve Hoffman's Hardware forum:

      I find the denizens there quite balanced with a number of viewpoints and generally respectful, not ideologically "poisoned" by one-sided Industry propaganda. The "Music Corner" is also great with lots of knowledge about albums and masterings for example.

  7. The strange part is that this recording sounds much better than the YouTube audio I often listen to through my Topping EX5/A30Pro system.

  8. Found the datasheet here:

    It seems like its called AB5605B? I dont get their naming scheme. They have same naming problem on the JieLi bluetooth chips where physical designation differs from actual designation.

  9. I bought this DAC - or almost the identical one, case is slightly flatter, but the DAC chip and connections are the same - for $3 at a thrift store, and thought it would be fun to compare it to two very high-end late 90s Parasound DACs: The D/AC-1100HD and the TOTL D/AC-2000 Ultra Analog.

    I'm listening to it now via a CD player's coaxial output going into a Parasound P/HP-850 pre and John Curl-designed HCA-1000A amplifier. The speakers are rare, Italian Opera Terza IIIa's (solid walnut cabinet, roughly $2000 in 1999).

    Subjective impression with expertly recorded jazz album as the source? It sounds very good, with nothing audible I can find to complain about. The speakers in my listening room emphasize the bass around 50Hz, so there is nothing noticeably deficient in the low end.

    As much as I like to sometimes tell myself I can hear the differences between solid state audio components, whether they be DACs or amplifiers, I usually have to admit that I would very likely fail to discern any difference in a double blind test. I say this as someone with pretty refined and trained ears - I've been recording, producing, and mixing my own music for 30 years (Love and Japan on Spotify if anyone is curious).

    Looking at measurements is interesting because it gives me an idea of what level of measurement deviation it would take to make an audible difference. There's a YouTube video out there that goes through increasing levels of harmonic distortion so you can get a sense of the threshold at which you can detect it, and for most people that's going to be at much higher levels than the THD ratings of our gear.

    Anyway, interesting stuff here. Glad I found this blog, but unfortunate it appears to have ended in 2022. More posts coming?

    1. Correction: I was only seeing the links up to 2022 posts. This blog is still going in 2023. Apologies for suggesting otherwise!

  10. A little update: the culprit for poor low frequency behaviour is the pair of output coupling capacitors. I measured them, they are measly 0.1 uF MLCC capacitors. If anyone have the necessary tools - and eyesight - solder a 2.2-3.3 uF cap parallel with them. Assuming an average 22 kOhm load on the output of the tiny DAC, the lower cutoff frequency is 72 Hz with the original 0.1 uF cap, and 3.3 Hz with a 2.2 uF capacitor. The output level is also very low, (0.3V eff) so a preamplifier is necessary to boost the voltage up to a power amp's input level