Saturday 1 June 2024

"High-End" DAC Blind Listening Results - PART III: Subjective Descriptions

In this last part of the "High-End" DAC Blind Listening Survey write-up, (see Part I, Part II) let me document the more qualitative aspects of the responses I received from listeners. These come from the comments section where listeners described what they heard. You'll also get to see in context some of the descriptions of the hardware used in the evaluation.

Comments will be posted verbatim other than removing any identifying information and names unless the person specifically says it's OK. Not all comments are posted, only the ones with a fair amount of subjective experiential content or description of evaluation procedure. It's interesting knowing which city/country some folks are from so I'll leave that information if mentioned.

So as not to neglect those who said they heard no difference (yet I know many spent a good amount of time on the test), let's start with that group of respondents... I'll add some comments/responses along the way.

Reminder of the identity of the tracks:
Sample A = Apple USB-C Dongle ($10)
Sample B = Linn Majik DS + Dynamik Power Supply (~$3,000)
Sample C = Linn Klimax DSM/2 (~$20,000)

I. Those who heard no difference (n=19):

Playback: Genelec 8351b active speakers, GLM room-corrected, 3m listening distance, moderately treated room, digital inputs fed from Mac mini-based music server using locally stored lossless files. I did NOT listen closely near the speakers for any differences in background noise/hiss but heard no noise or hiss at my listening position. Listened to the first part of all three samples repeatedly - thought I heard plenty of subtle differences but then when I would return to the previous sample those differences would disappear. Then I put my player app on shuffle and repeat and blindly pressed the Next Track button twice, so it could be any of the three samples and it could possibly be the same sample twice. Then I would guess which one it was each time, and each time I was wrong. So I conclude they all sound the same. Perhaps I could hear a very subtle difference here or there if I concentrated harder and listened a lot longer, but (a) I doubt it, and (b) if I had to work that hard to hear any difference, it wouldn't be worth the trouble anyway, and even then I doubt I could say one was "better" and the other "worse" based on such a miniscule difference. Oh - and thanks for running this test and survey!

I used Foobar's abx and beside some occasional artifacts switching like a madman while grasping at straws, I didn't manage to get anything conclusive, or even the feeling that I noticed a clear difference in "soundstage" or whatever audio qualities I find in all reviews. A few times I felt like between 1 and 3, I could sense a delay with 3 lagging behind, but: 1/ I checked the part where I thought I noticed the delay, and while it had among the most shift and in the direction I was "feeling it", the actual delay made me think I'm full of crap. 2/ I still didn't manage to get a meaningful result using that in another ABX trial, so, trying to be clever doesn't work for me. The first question should probably be removed from data as I didn't pick anything and don't think I could. Well, sample 1 is both called number one and is slightly heavier, so you can say I picked that based on totally legit rational... ^_^. I'm castleofargh on Headfi, and I feel fine being mentioned for whatever reason (like how amazing my English is). Thanks for doing this.

Excellent English!  😉

I downloaded Foobar2000 and its ABX utility specifically for the purpose of participating in this survey. I sat down expecting to do a long, in-depth, comparative study using this software installed on a laptop streaming over wireless to the router/ethernet to a DLNA/UPnP endpoint. However, I gave up after half and hour or so as I could not tell any difference whatsoever between the files. (BTW, this makes the file ranking in question 1 meaningless - I wouldn't rank them at all, but had to rank them in order to submit the survey form.)

Hey thanks for giving this a try with ABX Comparator! Appreciate the effort! 

Great music samples, thanks for sharing. Ultimately, all three samples sounded identical to me at moderately loud volumes using around ear, closed-back wired headphones. I would classify myself am a life-long music lover who enjoys a balanced sound profile with a slight bass emphasis/tilt (warm). I loaded all three samples into Audacity and listened to multiple ~20 sec loops switching between each music sample. I thought I heard a difference once or twice, but when I re-listened the difference was non-existent to very very slight. Thanks again for all your effort putting this together.

In the first attempt, I heard differences: A Most dynamic. Uncontrolled. More detailed. Harsh treble. Brighter sound. B Controlled. More precise bass. Dry sound. C Better sounding dac. Dynamic but acurate. Engaging sound. However, when I gave another try, the differences dissapear. Now, I can not say what sample am I hearing in a blind test. Audio gear: LMS, picoreplayer 3B+, topping D20, TubeCube 7, KEF 105.2 reference series. LG v35 cellphone, HiBy player, AKG k553 pro & Truthear Hexa. Thank you very much. It was so much fun.

Glad you had fun and was able to use a variety of hardware for listening.

At first listening when switching between the files and scanning back and forth over the different parts of the tracks seemed to expose significant audible differences. Which surprised me a bit because i had given up years ago on wasting my time and money on DAC’s. But when i took more time for repeated and longer more concentrated listening at a later moment, i could not replicate those first impressions. The differences that i thought i heard at first had disappeared when i listened to the same parts over and over again. It was as if what i heard at first glance was just out of reach every time i tried to pin point and isolate it. As if i was looking at the same scene but this time without my glasses on, unable to see the subtle the fine detail that caught my eye before. For me repeated listening erased all differences and often times threw up new ‘audible' differences and those also, without exception, were irreplicable. I was never able to find and hear it again. Maybe I’m too old by now (59)? Maybe my hearing is not refined enough or kaput? But really i am more and more convinced that our senses and hearing is not as exact as we like to believe and although i love nice looking equipment (Luxman) and high quality speakers (Quad and Elipson) and i enjoy spending (foolishly) my disposable income on such items. But spending it on DAC’s is not a joyful experience for me anymore since i bought my last cd-player, a Sony XA30ES over more than 2 decades ago. I thought Sony had nailed digital reproduction by then and nothing after since had changed my mind. But this test had me going for a minute but reality checked back in under scrutiny. Thank you for putting so much time and effort in keeping your wonderful and very informative website going, i check in regularly to see what’s new here and i enjoy reading your articles. It find that it is getting increasingly difficult to find reliable information on any subject on the internet . Your website is an oasis in an ever expanding desert of disinformation, fake news and BS. Kudos to you for that!

Glad you gave this a good go at it! I think you described it well in terms of how slippery the auditory experience can be when trying to pinpoint differences especially with music. Nothing wrong with enjoying the toys we have!

Linkwitz LX521.4 speakers, Hypex NCore amp modules, SMSL SU-8 DAC, really nice looking cables :) Sometimes thought maybe B was a little louder or brighter, but not reliably so.

I loved your different music samples- they demonstrated a really gorgeous range of different sounds and recording feelings. Also thank you for naming all the specific songs details so I can save a couple of these as new great songs to listen to on great speakers. I did many different types of specific listening tests from all the way through to ABX to blind switching back and forth. Almost no audible difference. It was incredible to (not) hear for my ears.

Hello from Germany! I did the test three times and I could not really hear any difference. But then I know that my hearing is not what it used to be, and aural memory… My tinnitus does not exactly help either. - One suggestion, it might have been easier if each of the tracks had been one after the other. Maybe I would have been more able to hear differences. Equipment: Synology NAS, Limetree Bridge, TacT 2150 amps (biamping), Eminent Technology Lft 8b. Best...

Nice gear man! Human echoic memory is brief (3-4 seconds typically quoted) when trying to listen to very fine nuances, so definitely using tools like the ABX Comparator can really help. Unless two pieces of gear very clearly sound different, it's hard to imagine hardware reviewers for example would be able to accurately compare the sound quality unless the devices are side-by-side and played back immediately one after another! Silly sometimes when people compare the sound of hi-fi devices they haven't heard in weeks or even months yet insist they can still remember nuances.

My system consists of Spatial Audio Lab M5 Sapphire open baffle speakers, Schiit Freya+ pre-amp, Schiit Aegir monoblocks, Denafrips Ares II DAC, and iFi Zen streamer. My playback software is Audirvana and also Foobar using the ABX Comparator. (Thanks for the suggestion!) I could not detect differences in the samples. One thing I found interesting was that Audirvana indicated that samples B and C were MP3 320 kbps. (My experience as a bass player in rock bands is somewhat limited, but I have 40 years experience as a video editor which has included a lot of audio mixing of network television shows with and without music.) I don't consider myself to have "golden ears" and my age limits upper frequency hearing, but I am trained to be discerning and was surprised by this listening survey. Thanks for providing it!

Strange that Audirvana thought Sample B and C were MP3!? I wonder why or how the program would detect such a thing if fed FLAC content.

First listening, somehow I was leaning towards C. Honestly, I don’t think it made any difference that I can randomly tell one is better than the other. Even with the same DAC, you tend to prefer one over the other subjectively. Used AirPod Pro Gen2 with IPhone 15 Pro Max. Tried with Spatial on and Off .

Wow, I'd like to start by thanking you for creating this beautifully simple survey; the sound quality was sumptuous and the music choices fantastic, I don't think I've heard my setup sounding better! It's also been the perfect opportunity to establish what the children have been telling me for years — 'turn the telly down, you need to get your hearing checked!" Given my age and the children's sensitively delivered diagnosis I was overjoyed to be able to enjoy your music regardless of what I might be missing. Fortunately, I've been too impecunious to embark on a fools errand and chase down the audio gold at the end of the rainbow, I'd have spent a fortune and it would have ended in tears. Thank you for everything, you're a ⭐️! Roon - Chord Mojo/Poly - Dan Clark Audio Aeon RT (Closed X from Drop)

Wonderful spirit and great gear! Cheers... 

RME ADI-2 Pro FS R BE, Sennheiser HD650 I can tell that from the differences I _saw_ on the spectrogram I like DAC A the least :-)
Yeah, it's amazing how even a little screen showing some objective parameters or the dancing FFT might let us know stuff even if we might not be able to hear it!

II. Those who thought the Linn Klimax DSM/2 (US$20,000, Sample C) Streamer sounded best (n=31):

Next, let's have a look at the comments where the listeners chose the most expensive and highest objective performance streamer as sounding "best". Within these comments in theory would be the "Golden Ears" if one believes this is indeed the superior high-fidelity DAC. Of course being subjective, as per the graphic at the top, one doesn't have to agree that the $20k streamer sounds best!

Thank you for this well-presented test. I used the Desktop PC setup I use to tag up my flac Library prior to transfer of music to my main system. Not costly or resolving apparatus: Windows PC > Cambridge DACMagic 100 > Cambridge Topaz AM10 amplifier > very old Wharfedale Dovedale speakers. I think I can hear differences. My first impression is that A is inferior and should be a cheap DAC like a Topping. C is more resolving and has thunder (= good). I held back thinking B might be more "natural" nevertheless. On a little more listening C won. I use foobar anyway but didn't use ABX. I spent very little time on this. Just scanned up and down the tracks a bit for a minute or two. I am not convinced the samples are volume matched effectively. C won to my ears but also sounds slightly louder. Thank you again. This sort of exercise is a great contribution to the hobby community.

Hi Archimago! I have listened very carefully to your test samples of DAC's and I have come to the opinion that they are in order from worst to best, that is, the worst is number 1 and the best is number 3. However, this is only my impression and I do not know if this is actually true . On my system it sounds like I wrote. DAC: EverSolo Z-6 Amp: BV Audio PA300SSE, vertical bi-amp Pre: BV Audio Pre1 Speaker: Dynavoice DF-5 Thanks for the opportunity to test my own ears and have a nice day!

My gear: DAC Audio-GD R1, Amp Schiit Mjolnir 3, HP Hifiman HE1000 Stealth My name is Kris To me sample A sounded somehow a bit dull and fuzzy and sample B maybe overly sharp and very hard hitting. C seemed like a nice, natural sounding compromise, even though I would assume that something between B and C would suit my hearing, gear and preference best. Overall I would see B and C generally over A.

"Sample B" is pretty artificial and not pleasant to listen to. "Sample A" gives a bit more information overall and is a bit more pleasing, I could happily live with that. Not being a close call, "Sample C" came in first though. Much more mature and stable, tonal colors are way better, overall much mor natural sounding. I so much hope it is the mid-price design... - By the way, thank you for doing this. It is really really cool!

Alas friend, I'm afraid you chose the high-priced product. I guess you're going to have to shell out the big bucks. 😦

HE400se + Topping NX4 + Harman-based EQ Sample A stands out in a negative way (a bit muffled, less detailed, maybe noisier) B and C are difficult to tell apart (I'm sure I would be happy with either DAC) Cheers...

DAC C - best separation, and balanced frequency reproduction. DAC A - sounds a little muddled, not great with separation, but ok frequency response. Wild guess that this DAC has a minimum phase filter. DAC B - had a harsh high end / general spiky sound, not neutral. I've used an embarassingly low-fi setup to listen to the samples, wouldn't be surprised if my comments are an outlier...

The comments above are good examples of the Linn Majik and Klimax being difficult to tease apart, yet Sample A, the Apple USB-C dongle stood out negatively. We saw this result reflected last week in the ranking of the dongle being placed last primarily by headphone listeners.

Ferrum stack (Hypsos, Wandla, OOR), Su-6 DDC, HD800 phones, track 3 (C) seem the most natural sounding for my old ears, also bit more dynamic on some passages, seem better vocal and strings texturing and some faster transients in places...

For listening were used e1da 9039s/smsl dl200, hifiman he400se, beyerdynamic dt770 pro and topping pa5 for speakers. I believe that the track C is the resampled original AMPT. The most bright differences was noticed on Eva Cassidy recording. Thanks...

Hmmm... Not sure what is meant by "resampled" original AMPT track. They were all directly captured from the analog outputs using the RME ADC. Maybe you thought I was trying to fool folks and just digitally resampled C because it sounded the cleanest?

Speakers, B&W 683 and Sub, SVS SB-1000. Headphones HiFiMan HE400i. I came to this with a very clear bias of "all DACs sound the same". I did feel that A and C sounded the same though on headphones at high listening level I did establish a slight preference for C as an 'easier' listen. B consistently sounded very slightly muted - to these very old ears - less HF energy. Good or bad? - I don't know. On many tracks B actually sounded preferable to my ears, as the tracks themselves (except 1&3) have an 'over-produced' sound - not to my taste.- Thanks for arranging this interesting test. ... Retired BBC sound engineer.

Cool, an ex-BBC guy! Thanks for participating.

Sample C track 1, the bass was very deep and tight as it reverberated in my room . Track 2, the drum had a nice solid weight to it, I could clearly hear the musician take a breath when playing the flute. Track 3, the live recording was fast pace with great PRaT. Track 4, Eva's voice was crisp but not overly sibilant. I could almost picture her singing in front of me. Track 5, was QSound used at the beginning of this track? Vocals were nicely placed between the speakers. Track 6, put on your dancing shoes and let's dance! Marantz SR7012 Paradigm Premiere 800F speakers SVS PB2000 sub Super Clarity Acoustic Panels

Oh, almost at the last minute I discovered this new test. So I'm sending quickly my evaluation. I don't think this test can discover golden ears. The worst sample can be guessed with 33.33% probability. Putting all samples in the expected order can be done with 16.67% probability. But on what basis is the best sample determined? Price? Hopefully not! Here is my short summary. Sample A is the absolute worst. It's about the level of a 20 euro USB stick. Sample B and C are very similar. Both have a big distance from A. I suspect they may be from an identical DAC using different filters. At a cost of about 2000 euros or so? Emotionally, based on the music sample 3 (Ray Brown Trio), I preferred sample C. Since I have the original SACD/DSD64 , I compared sample 3 with the original stream as well. There the difference is already audible clearly. All samples are very unsuitable for the test, but that would be a long explanation. Arch, I wish you a lot of energy in your nice work in the audio field. I appreciate very much your fight with the nonsense called MQA. Excellent work! My recommendation to you, don't fight against 96k/24b because it is really a significant jump from 44k/16b. It's a really big difference. I have no way to prove it except with my ears. That's why I'm offering anyone a million dollar bet that I can correctly determine 12 times in a row whether they're playing the original 96/24 or a copy resampled to 44/16 and back to 96/24. Any software resampler is allowed. One mistake and I lose, I think that's a fair offer. Rich people like Elon waste a lot more money on less meaningful things than this exact proof of the audibility of audio formats. Notes: You can publish my post if you want. ... Translated into English using AI with a little manual corrections.

Fascinating man. Amazing ears. I'm impressed!

Anyone want to take this gentleman from Slovakia up with his $1M challenge? He's offering 12-times-in-a-row level of accuracy between original hi-res 24/96, and a version that has been downsampled to 16/44.1 then back to 24/96! Maybe you can leave a comment telling us what hardware you used? I see your loudspeakers and your system total was >$30k. Maybe knowing this context first will help if we can get a blind test going with money on the table; I might be willing to chip in a few bucks. 💰

BTW, I don't think I fight against 24/96 - even back in 2013 I've advocated for hi-res for my favourite music if available. If the music is recorded well like excellent uncompressed classical and jazz, I'm certainly all for 24-bits and 96kHz. My only issue is when the recording technically does not seem to benefit from the larger hi-res "bit bucket".

Not really a lot of differences depending on music. Absolutely sounds as good as my big rig.

FIIOK7 - MEZE109 pro - PC/FB2K Impressions: sample B had overpronouced low end, indifferent mid/high, less detail. sample A/C are quite close. C sounds more natural to my hearing. could live with both


III. Those who thought the Linn Majik DS + Dynamik Power Supply upgrade (US$3,000, Sample B) Streamer sounded best (n=29):

I have a cheap system. PC, Topping E30 lite Dac, Altec 14 speakers, Nad350 amplifier. My choice was BAC. Small but audible differences

My test system: Acer Aspire 3 Windows 11 PC, THX Onyx DAC, Truthear x Crinacle Zero:RED IEM Sample B sounded the smoothest and best to me, but overall, I had trouble determining which sounded best.

B rules, A & C are crap

A and B are nigh indistinguishable from each other. C has a narrowed stereo image that I'm assuming is due to higher crosstalk compared to A and B in listening, it is a very subtle effect that also draws my attention to what sounds like a slight bit of clipping happening at the peaks of C that are not audible in A & B.

Hey Archimago! Long-time reader, first-time respondent! Now I want to preface before providing my thoughts that although I did detect a very slight difference, it was likely placebo and I wouldn't expect anybody to hear a difference unless they have golden ears and maybe an extremely high-end rig and is actively looking out for differences. As a teenager, I don't consider myself to have a set of golden ears. I can only hear up to 17kHz and I personally don't think I can differentiate between 16/44.1 and 24/96. What I do think is important is the quality of techniques and filtering used when making music. In my experience a lot of processing and effects behave differently depending on the sample rate. Instead of high-resolution audio (past 20-bit 48kHz) I think substantial improvements for fidelity come from high dynamic range (assuming the mixes are good), and more importantly, multichannel audio, which I go bonkers over. I'm also a digital die-hard (lossless digital, anyhow). Now for the listening setup for this survey I used an Air 192/4 DAC paired with a Phillip SHP-9500. I do have a much higher-end 5.1 speaker rig but I do a lot of my listening on this combo so I thought I'd do the test on here because that's where I'd realistically evaluate a mix critically as I'm extremely familiar with the sound profile of the headphones. I'm also using a custom EQ curve with PEACE because I love "smiley-curve" responses, much to the dismay of some audiophiles. I'm just used to that type of sound profile the most, so I have an easier time mixing and critically evaluating compared to a flat-tuned response. For my evaluations I skipped straight to tracks 2 and 3 and evaluated them first because I expected them to be recordings that would easily highlight any flaws. Whilst I could not tell much of a difference between Samples B and C, Sample A, at least to my ears, had something weird with the high-end. It was slightly brighter, by a miniscule amount, and was initially going to prefer it (I love brightness, and so it messes with my evaluations), but I felt that there was this odd brittleness in a way. I felt the fidelity was somehow compromised in a sense. So I ranked it last. B and C were very close, but C was too smooth. That usually would be a good thing, right? But on the Dua Lipa track, I felt it was blurring some of the harsher elements in a way. When going for accurate reproduction, IDK. If I could kind-of have a tie between B and C I would, but I felt B had the very slight edge. In the normal course of life, I wouldn't expect anybody to tell a difference between the three. If critically evaluating, I wouldn't expect anybody to be able to rank between the three. I am able to hear that there is a very slight change in sound amongst all three, but to which is actually better, is very hard to tell. You'd have to have a reference to compare to in a sense, at least that's what I think. I'm always here for more surveys like this, as well as anything about multichannel audio! Interested to see the survey results and whether I got anything right (I doubt I did). Also, shoutout to the QQ forum.

Thanks for the very detailed discussion on the procedure used and subjective descriptions of what you heard! Another good description of the relative inferiority of Sample A (Apple dongle), yet similarity between the two Linn streamers B & C.

I listened through two systems, the one I referred to when asked about price was Rega DAC-R > Violectric DHA V226 > ZMF Aeolus, and the other one was Rega DAC-R > Rega Brio > Beyerdynamic DT880 Edition 600 Ohm (less than half the price). Came to same conclusion with both: sample A sounded clearly inferior with less dynamics. There was much less difference between sample B and C, but I found B more musical and coherent and perhaps more agile with the details. Also, I answered no about being a musician because it was at such an amateur level a long time ago, but yes I found myself on small stages a handful of times performing with a band.

System: foobar2000 with various DSP + Topping DM7 DAC + Neumann KH310A. What I heard: sample B 2:00 “Japanese Roots” the decay & reverb really stood out, the attack was really clear and sounded really good. 2nd place goes to sample C. Followed by sample A which was a little bright, and a little muddy in the lower mids.

Interesting the words "bright" to point of "harsh", and also "muddy" have been used by listeners to describe the Apple dongle a few times here and also in live listening when I asked for example my kids to listen to the actual device compared to a better headphone DAC+amp. 

System: VLC version 3.0.20 Vetinari (Intel 64bit) running on a 3.6 GHz 10-Core Intel Core i9 iMac through SoundSource 5.6.3 (to ensure it was bit perfect) to a USB hub then USB A (3.0) to USB C cable connected to a Chord Mojo2 DAC (no adjustments enabled) to Focal Elegia. Differences: Sample A was my least favorite by far. Didn't matter the type of music, it felt veiled. Now, as for the difference between B & C? That was nearly impossible to determine. Sample B one in the end by having what I took to be just a bit more presence, but if you ask me tomorrow that answer might change.

1 - sample C is lackin bass, Sample B is small win margin compared to A 4- sample B is above all (high More distingushable) - 6 sample C is lacking bass and B has best high distinguishability. All in all I prefer B over A and difference margin is medium, and I really like the sound of B. Sample C is clear looser on high/mid/low.

Your point is well made. In the age of really well designed electronics it was difficult to justify the minor differences between the 3 dac's. I played them in my modest system with a pair of DIY open baffle full range (50-16K) point source speakers with additional passive base compensation drivers (not a subwoofer) DSP (30-90htz) both with class D amplification thru my $1000 delta sigma sabre dac and also with my new relative inexpensive Chi-fi dac <$300 with AKM's new chipset 4499ex/ 4191 balanced connection and to my ears it sounded great. The tracks were streamed thru Roon using an IFI Zen Stream (as endpoint) and using its coaxial output to the dac. Thanks for the exercise.

Nice to get feedback from knowledgeable DIY guys with interesting gear! 

Hi Arch, Frustrating test especially using headphones. Louder volumes on my speakers made it slightly easier to discern differences although in all fairness these differences were barely noticeable. My eldest daughter found it easier to hear differences and they did align with my preferences. Starkest difference between B and A. C and B were quite similar and on some tracks almost indistinguishable. Especially B sounded more dynamic and less harsh/fatiguing compared to A. The huge price disparity does not align well with the sound quality of all three. Equipment; Speakers Dali Grand Coupe, Dynaudio S 1.3 SE and Dali Menuet SE. Headphones: Beyerdynamic DT1990 Pro. Aune AR 5000 Amps; Atoll Preamp and Atoll Poweramp. Desktop: WiiM Amp/Fosi V3 . Headphone amp; Little Dot mk 2 upgraded and SMSL SH-9. Sony TA-AN1000 AV amp. Streamer: WiiM Pro Plus DAC: 2 Upgraded X8 XVIII MAGIC DAC FPGA. All Roon integrated with Roon Server on PC. Cheers... in Sweden

Another interesting description of similarity between the Linns (B & C) as superior to the Apple USB-C dongle. 

(DIY Audiophile streamer: Fanless, Pentium GoldP 5400, LPS, Matrix-Audio USB card, internal Crucial SSDs [5TB] storing all music files, Debian 12.5 with custom kernel to allow native DSD for SA-10, MPD 0.23; not used for this test) - USB flash-drive direct into Marantz SA-10 - Musical Fidelity KW550 integrated amp - Bowers & Wilkins 802-D3 speakers. (didn't compare with headphones: Sennheiser HD800S/Audio-gd NFB-1 amp) Key difference was preferred recordings sounded more analogue or natural/life-like - improved stereo imaging, natural timbre and smoother/clearer - less boom 'n' tizz and compressed sound! These are the sort of differences you can hear between PCM any resolution and DSD, the latter preferred in almost all cases up to DSD256 with the SA-10 in native DSD mode from the music server (better than DoP if custom kernel not used since Linux doesn't support the SA-10 player) BTW: I also used an Intel NUC with LPS running same software but sound not nearly as good as using a proper ATX mobo for the server due to the noisy environment inside the NUC and its internal VRMs! ... - London UK

Playback: LMS->raspberry pi4/Squeezelite->ADI-2DAC->ATC40A+Arendal1723-1S sub Sample A - more low bass/slight emphasis on high-mids, a bit "HiFi"/turned up to 11 Sample B - best balance Sample C - slight lack of bass/dynamics and definition Differences subtle - all sound good! Opening made me Jump :-)

Yes, have to be careful with the Reference Recordings track's dynamic range at the start! 

SMSL_do100 DAC + Topping_L30II HP Amp + Sennheiser_HD800s/Truthear_Nova headphones/IEMs I've used ABX plugin in foobar2000. It was difficult, but possible to distinguish A from B (12 from 16) and A from C (11 from 16), but it was really hard to hear the difference between B and C (7 from 16). I think the A has a little less enjoyable impact on bass, too low distortion maybe?

Playback PC - Oppo 205 Dac/Pre - Minidsp Flex - Boulder MKI monitors - JLAudio 12 Subs. Sample B more detail more music. A wasn't far behind B, but C was definitely worst.

concerning how much difference: for me B was best, but the difference between A & C was not so great


IV. Those who thought the Apple USB-C Headphone Dongle (US$10, Sample A) sounded best (n=26):

computer (foobar2000)>ifi DACver2>AKG K371 slight difference between A and C, A having the better soundstage (spatial sepertion) for me and instruments detail were clearer

PLAYBACK SYSTEM: Apple Mac mini M1 (Sonoma 14.3.1) with MCRU No.91 EU Power Cord (Belden 83803 / Furutech FI-E11 (Cu) Plugs) -> Cockos REAPER -> Intona ULTIMATE USB Cable -> Gustard DAC X18 with Synergistic Research Purple Quantum Fuse and MCRU No.91 EU Power Cord -> 2x XLR Cables (self made from Belden 89207 cable and Neutrik ECM-Series XLR Connectors) -> 2x APS Klasik Studio Monitors (the old ones, not the 2020s) with Synergistic Research Purple Quantum Fuses and MCRU No.91 EU Power Cords. THE DIFFERENCES: For me, the most important factors are the „realism“ and „holography“ of the recorded audio samples. Sample A is by far the most transparent for me, judging by these criteria. Samples B and C sound significantly less transparent and kind of „processed“, but I still find B slightly better than C. If, say, the Sample A DAC is worth 1000$, I would say the Sample B DAC would be max. 300$ and the Sample C DAC max. 150-200$. This is only meant as an example for price relations, not for the absolute value in USD that I would actually spend on it.


in my hearing, the difference between A and B is more noticeable, A and C are more similar, and the classic music can offer more difference like sample 1 and 2, the sample 4,5,6 are hard to differ

I did not use the compressed tracks 5 and 6 for the comparisons. I find Sample A distinctly better, with more details and separation between instruments. Sample B sounded muddy and flat at times. Sample C is slightly better. Maybe Sample A makes use of a different filter that I like better. Subjectively, I find the Eva Cassidy song more moving and engaging in Sample A, and very bland in Sample B. The bass drum in track 1 was weaker in Sample B. I used EQ'd AKG K702 headphones directly into my MacPro and listened under Audacity. I think that even though critical listening reveals some difference, all 3 samples would be acceptable to a more casual listener.

My playback system is: Laptop running MPD on linux as bitperfect source SMSL SU-10 DAC Marantz PM-11 amplifier KEF R900 speakers There's no huge difference in sound quality between the tracks. For me, it was easy to evaluate from the beginning since I very much like the recording of Stravinsky's Firebird from Minnesota Orchestra. Track A was the familiar sound I get from my system. Track C was so close to track A. Only track B was in a way significantly different than the other two, at least to my evaluation of what I hear or better, what I believe that I hear. ... from Istanbul, Turkey.

Equipment: Violectric V800, Topping D90SE, Topping A90 Discrete, Dan Clark Audio Stealth, Acoustic Energy AE1 Active. Very subjective difference. A is tasty and natural in a gentle way. B didn't touch me at all. C may be not so accurate, but powerful sounding. It is hard for me to choose between A and C actually. I don't want B, but probably want to keep both A and C.

Thank you Arch. This was a cool test/survey. Your choice of music was awesome and made the countless replay/A-B listening tests bearable 😂 I have always been interested in higher-end DACs and headphone amps compared to what I own and I was immediately aware of two things as I played these tracks: (1) with my modest kit, and my old ears, everything sounded awesome, and I enjoyed the music from all three test cases. I liked them all, but there were some differences, nonetheless, I enjoyed the music from all three tracks on my kit (i.e., I didn’t find anything egregious with the DACs) so I probably don’t need that DCS Bartok or a Chord Dave with mScaler; and (2) I did have a preference for brighter sounding (less muffled) music so with my kit that was A or C and B was last in my preferences. Prefered Ranking: 1-A 3-C 2-B * iPad Pro 12.9” ($900) > mojo 2 ($650) * Sony IER-Z1R iems ($1600) * total chain costs: $3150 * Lumin U1 Mini ($2000) > RME ADI-2 DAC ($1200) * Sony MDR-Z1R headphones ($1800) * total chain costs: 5000

Playback Chain: HP Laptop -RME ADI 2 DAC fs(ESS chip) - NAD C298 (Purifi Class D modules, Eigentakt technology) - ProAc Response DT8. The differences were pretty clear for me. Sample A, just sounds right to me. Balanced. It gave me the goosbumps and i enjoyed the listening. In sample C, things got a little worse, and in sample B really bad. Not in one aspect. Low end, resolution, soundstage, tonality, placement, i thing they all go together. Thanks for this opportunity, because you really have to listen to the 3 samples one after the other , to really hear the difference. ... Greece.

Harbeth M30.2, McIntosh MC275,Chord Hugo TT2+Mscaler, Aurender N200, two REL T9x all dialed in using REW. Sample A was my strong preference as it was more dynamic better resolution and clarity, similar to my Chord gear. Sample B was my least favorite because it was easily the worst at those characteristics. Pure guess but sample B sounded line an R2R dac which is not my preference. Sample C was good and seemed closer to sample A than B for me. Nice music selection!

Sample A stood out as best. Samples C and B were closer to each other, but from time to time it seemed C was better than B. Sample B had the smallest sound stage. Again B and C were close.

Indeed we can say that B & C are more similar sounding as Linn products and A sounds different being the Apple dongle. A good example of how subjective preferences are idiosyncratic. There's certainly nothing wrong with preferring the sound of the much less expensive device! And this means having more money in the bank account.

Source: CD/Media/SD/USB TASCAM CD200SB, PC Win 11 prof 64Bit, MSI-B450 A Pro MAX, foobar 2.1.4 / 64Bit, ABX, Behringer U-Phoria UMC204HD Preamp Yamaha C40 Speaker Neumann/K+H O300A Monitor Headphones AKG 721 --- Hi Archimago, i noticed only very small differences in Track 1 (A) - at medium and high frequencies.

Laptop (ASIO USB) -> Gustard x16 (DAC) -> Gustard H16 (Headphone Amp) -> Hifiman HE400se (headphones). Only focused on second song. I focused on echo trails mainly. Seems like sample 1 echo trails I like the most.

the poor man's audio system: nuprime dac10 (sabre9018) / zerozone irs2092 (class D monoblocks) / magneplanar mg1.7 in dedicated audio 2ch room VERY LITTLE DIFFERENCE, I preferred: B for instruments positioning and detail A very similar to B, a bit more brilliant, C no main differences with A, maybe on sibilants greetings from Italy... p.s. I LOVE NYC


V. Final thoughts...

It has been great reviewing the subjective impressions! Thanks again to all the participants.

Even though we can look at the numerical results as a whole and find patterns as discussed last week, subjective responses come in all flavors and there's no way we would be able to say that what was experienced by any one person can be "wrong" regardless of whether they preferred the Apple dongle, or the Linn streamers. 

I trust that based on the preference numbers, magnitude of audibility, and subjective descriptions in the feedback, we can agree that overall, the recording from the "cheap" Apple USB-C headphone dongle wasn't obviously "bad" sounding. Even if not the best, certainly the sound was not to the point where listeners could not enjoy music through this $10 DAC.

Imagine for a moment, if we didn't run this as a blind test. Would audiophiles in a sighted listening challenge ever admit that they thought the Apple dongle bested the Linn streamers!? Highly unlikely, right?

Imagine if I went to an audiophile meet-up and said to everyone: "Guys, the Apple USB-C dongle sounds pretty close to my Linn Klimax DSM/2; I think I'll just sell off the Linn and use that as my DAC!". I'd be laughed out of town, my audiophile membership card would be immediately confiscated! Forever ostracized as an "oxidized tin ear" laughing stock of the club who only deserves to listen to Sanyo boomboxes and white freebie iPod headphones.

This is what audiophilia can seem like when we enter forums, browse through the pages of magazines like Stereophile or The Absolute Sound, reading reviews from guys who somehow always seem to think that "high-end" gear with high price tags to boot will always sound "amazing", "phenomenal", "the next level", etc. Of course those guys who write for these magazines are all Golden Ears! Right? 🤔

Blind testing (double blind or otherwise) has had a bad rap among audiophiles. Feel free to review articles like this from Roger Skoff (RSX cable salesman) or forum comments like this. Like it or not, blind testing provides a form of necessary "honesty control". As emotional creatures, subjectivity comes easy to us and emotional biases are part of our lived experience. Every once awhile, I think it's good to try a blind listening test as an exercise either in affirming confidence in our hearing ability, or perhaps more often than we audiophiles might admit, to realize humility, recognizing that some classes of modern consumer audio products like DACs, even if inexpensive, have come a long way in achieving very good sound quality and our hearing ability is finite.

As for myself, I mentioned last week in the comments that I had a listen to the actual Linn Majik DS and Klimax DSM/2 streamers hooked up to linnrd's great sounding system consisting of the Manley Jumbo Shrimp preamp, First Watt SIT-2 amplifier, and Avantgarde Uno horn speakers + REL B1 subwoofer on the evening of recording these test samples. Music was streamed to both Linn devices simultaneously using Roon, basically instantaneously switching between inputs using the Jumbo Shrimp preamp. With volume controlled <0.2dB between the Linn devices, neither of us could hear a meaningful difference despite the large price differential. I did not try listening to the Apple dongle due to logistical issues; maybe next time at his place.

Here at home with my Paradigm Signature Reference S8v3 speakers, Paradigm SUB1, Topping PA5 MkII+ Class D amp, switching as fast as I can between the Apple dongle, a borrowed Linn Klimax DS/2+EXAKT (pictured above, ~US$19k MSRP back in 2015-2016), and the Sabaj A20d 2022 DAC, the only difference I might notice is a better noise floor on a very quiet evening using the better devices (Klimax and Sabaj).

Even if I do not believe there is an audible difference, there's nothing wrong to still prefer the more expensive DACs/streamers in the system than the Apple dongle. Life in general and audio gear selection would be rather boring lived just based on listening tests or objective measurements. Non-utilitarian benefits like how a product looks, or the pride of owning a well-crafted device can be meaningful for quality of life - an inherently subjective concept. Let's not forget that the ultimate goal is music enjoyment which is subjective anyways. Nonetheless, my philosophy remains that of favoring devices with higher objective performance (the reason behind hi-fi) and good value (with elements of rational insight, rather than being caught up in foolish consumerism) when adjudicating the hardware. With the desire of becoming a knowledgeable and wise hobbyist audiophile.

A big thanks again to linnrd and friends for providing me access to these Linn streamers - obviously a well known high-end brand with a long legacy among audiophiles.


Let's end off with mention of some music. An oldie but goodie that can be used as demo music beyond some of the tired stuff at audio shows - the "Little Wing" (Hendrix) acoustic performance by the Acoustic Hippies From Hell which is Def Leppard + Hothouse Flowers. Released initially as the B-side on the Have You Ever Needed Someone So Bad CD single back in 1992:

Great sounding DR13 track and disc. Highly "spatial" presentation especially when played back with crosstalk-cancellation DSP. On that CD single are "From The Inside" and "You Can't Always Get What You Want" (Rolling Stones) performed by these guys as well.

Def Leppard was awesome back in the days of my formative years through the '80s and early '90s. It's been said that their album Hysteria (1987) was the pinnacle of analogue multi-track recording at least for mainstream pop/rock (interesting article, also see Perfecting Sound Forever).

While studio technology had already been changing very significantly by the late '80s, most albums released from that point in history understandably quickly shifted over to being mainly digital constructions. Def Leppard's follow-up album Adrenalize (1992) from which "Have You Ever Needed Someone So Bad" was on, was their last multi-platinum-selling album and an example of a full-digital production (SPARS code 'DDD'). Thankfully back in 1992, the "Loudness War" wasn't a problem yet and the album had a good DR11 (non-remastered).

For those unfamiliar with Def Leppard, Pyromania (1983) was their other mega-platinum mainstream album and the recent 40th Anniversary Bluray includes a multichannel/Atmos presentation that's worth checking out if you like this group and genre.

Well guys & gals, June is here! Time to look forward to some R&R with the family and maybe a trip here and there. Hope you're enjoying the music as we approach summer here in the Northern Hemisphere, dear audiophiles...


  1. Thanks for the test Arch! All in all it's rather reassuring that you don't have to spend an arm and a leg on a DAC these days to get a taste of top-notch sound quality :)

    I have an original 1980s CD copy of Def Leppard's Hysteria (made in West Germany!) I always thought the recording and mastering of this album in particular was fanastic. Superb dynamic range too - but the Apple Music version is terribly compressed unfortunately.

    I also recently got some awesome CDs from the 80s of Michael Jackson, Van Halen, The Cars, Pet Shop copy of Michael Jackson's Bad was made in Japan! Same deal as Hysteria - the CDs really sound much better than the lossless Apple Music versions. It's sad that the original dynamic versions are basically inaccessible to new listeners, I bet the music would gain a lot more fans if they were available in their original versions. I mean, I first heard Van Halen, Def Leppard, The Cars in just the past 4-5 years, and honestly they blow all modern music out of the water, especially when you listen to their original dynamic versions...

    Which reminds me of the recent Apple Music 100 best albums of all time. Good God, what a farce that was...but that's a discussion for another day ;)

    1. Hey Mr. MB,
      Indeed it would be fantastic if streaming services offered listeners the opportunity to access the original "non-remastered" versions of the albums before everything was converted loud.

      IMO, it's a travesty again art. I wonder if the musicians even care or have any power to insist on something like "Don't you dare compress my music below DR10!" when their albums are being remastered. Maybe they're just happy to get extra revenue when labels decide to release a remastered deluxe edition.

      Thankfully, at least the "spatial"/multichannel mixes have max loudness criteria as discussed over the last few months.

      I'll have to take a peek at that Apple Music 100 list. :-)

    2. Yes, making the originals available as a streaming option would be the ideal solution. Wouldn't be hard to do - just rip the original CDs! Sadly I don't see it happening anytime soon...the music labels don't seem to care much about DR issues.

      BTW, I left a couple of comments on your post "On Hardware Audiophilia and Wine Tasting" - I came up with a few intriguing points, if I may say so myself ;) Please do have a look when you have the time :)

    3. Thanks man! Will check it out... As you can imagine, been caught up with getting this series posted, work, and life in general!

    4. I totally understand :) I'm frankly astonished at the time and effort you clearly put into everything you've been doing on this site over the years, your thoughtful approach to this field, and it all being fueled solely by your passion for it all. It's very much appreciated :)

    5. A pleasure organizing and writing about this...

      I hope these tests, topics, discussions on the blog reflect and in some ways fill an area of "need" for audiophiles. As hobbyists, I hope we can all dive down not just into the low-level technicalities (noise levels, distortions, frequency response - IMO we need to know this stuff given the technological hobby) but also take a step back from the Industry narrative used to make sales and appreciate what matters for ourselves (what we each can and cannot hear) and in turn find wisdom based on knowledge and insight. Understandably, so much of the talk within the "professional" publications (magazines, audio websites, YouTube channels, etc.) revolve around things to buy rather than answering the questions of sound quality and whether products even make a significant difference!

      Like in most areas of life, I think wisdom is the most difficult thing to achieve; if that is one's desire of course. 🤔

    6. Your approach, insights, and contribution is definitely helping a lot of us get closer to wisdom, that's for sure. Thanks again Arch!

  2. Hey Αρχιμαγο! Thanks for this blind test! I made a complete fool of myself by choosing sample A as best sounding, with my RME - ProAc - C 298 stack! Hi hi hi!
    Still i hear the samples and i still prefer sample A!
    I guess it is not just the samples that matter. It's also the room, the speakers placement and the particularities of every space each of us is listening in, and the matcing of the system, that makes preffer one over the other.
    Surelly, we must close our eyes and listen, without taking into account the brand name or the price tag.
    Yoy made your point and i'm totally with you!
    Greetings from Greece! Dimitri!

    1. Hey there Dimitri,
      Absolutely no need to feel at all shamed for choosing "A"!

      Given the fact that you're using good speakers (ProAC DT8), accurate DAC (RME), good amps (NAD C298), listening in a room which will add its own character, this just means that the combined sound quality either overwhelms the differences between DAC recordings or that the character of "A" fits better for your space and personal preference.

      This would be absolutely acceptable based on the science of audio reproduction + physiological perception + individual preferences.

      Yeah, it's important to close eyes while listening. But if we already know what we're listening to, it's impossible to unbias our mind! Those who claim they can IMO are either liars or not human. :-)

  3. Thank you for setting up this test - I came to it a little late but my own personal experience over the last year has been a lot of work talking myself down from the heights of big budget audiophilia, particularly on the digital front end stuff. I had been working my way up the dCS DAC stack for a while... But then one day I did an a/b test (admittedly not blind) with a TEAC CD player's internal DAC vs. my main DAC (with a clock!): playing a CD and switching between the CDP's balanced outs to my preamp and then through the external DAC. I have to say, I really could not tell a significant difference. At that moment I really said to myself I need to reorient - pull back on the DAC, cables and the like.

    At the end of the day, I think it's incredible just how good a solid-quality DAC that costs a couple/few hundred dollars is these days. The esoterica will keep getting more and more expensive but the value in the more realistic spectrum is fantastic.

    I've still got audiophilia on the brain - boy that dCS DAC and clock looked GOOD and I can't help myself thinking that if the TEAC CD player sounds so good, then an Esoteric upgrade must be so much better... But now my entire system costs about half as much as just that DAC, it sounds amazing. And on the margin what I've been seeing with playing with DSP solutions is far and away more meaningful than plunking down five figures for a DAC upgrade. It's a process but thank you for doing these write-ups as I think without them I'd just be searching for the next more expensive thing.

    1. Hey there ruchottalex,
      Thanks for the note and describing your experience. I believe there is a wisdom to be gained in us audiophiles trying listening tests like these and it can be as straightforward as what you did with 2 devices and being honest with oneself.

      I have for years (more than a decade at least) been amazed at the apparent lack of wisdom shown by the audiophile press, including the aging ones (gee, I wonder why)... Thus it's left to us "grassroots" audiophiles to write about these experiences openly and honestly. And where we can, maybe try a test like this for ourselves.

      Having said that, yeah the dCS stack looks nice and I believe they objectively also measure well so if one has the desire and means, there's nothing stopping any one of us from purchasing that so long as we can appreciate that most likely, a much less expensive DAC/headphone amp will sound just as good - maybe even objectively more precise! I think that's just being honest, stripping away hype, since there are many other places our consumer dollars can go towards.

      Hopefully you're having fun with the clock 🤔. From an objective perspective I'm sure you're aware that external clocks are probably more likely to degrade performance objectively. Unlikely to be audible regardless, but useful for studio applications to synch multiple devices.


  4. I, too, mentioned in my comments, that I could hardly tell B & C apart, though I eventually chose C over B.
    But I chose A to be the better sounding, after a few trials. Perhaps because it was just "Different sounding".
    I even made the test "double-blind" by asking my daughter to rename the tracks at random, to see how consistent I was on my selection, and I was.
    What I am baffled with, is that you used 24 bit versions of your tracks for this test, when most of the originals were 16 bit, why??
    24bit version of a 16bit material is still 16bit at best!

    1. Hi Ken,
      Yes, true, most of the tracks are 16-bit source but a track like the Reference Recordings classical one was 24-bits and the Ray Brown Trio track was sourced from SACD so I started with the 24-bit PCM conversion.

      As shown with previous testing, I don't think 24-bits makes a difference but since I did have them available and each of the DACs, even the inexpensive Apple had no problems decoding, I might as well use the best quality just in case this has an effect one way or another. I think this is fair anyhow since these days I think many folks listening to streaming will have access to 24-bit audio either in their collection or streamed from Tidal/Qobuz/Apple/Amazon.

  5. Thanks for your blog. It prompted me to think what is "audibly transparent". Is there any "audibly transparent" DAC in reality?

    Here is my answer:

    What's your view? There is one? or there is none?

    1. Hey there sunjam,
      Interesting blog. My perspective after measuring and listening to a number of DACs is this:

      1. Never use a NOS DAC with 44/48kHz content as an example of high-fidelity audio :-). Sure, some people will prefer the sound but they are not transparent and will introduce a -3dB frequency response dip at 20kHz as discussed here:

      This, along with ultrasonic imaging are the distortions introduced by not having a proper filter in place and should not be used to argue against Monty's video that DACs can reproduce signals "perfectly". Within the limits of human hearing in 2024, a modern high quality DAC (doesn't have to be expensive) can already reproduce a digital signal within the bandwidth of the samplerate "perfectly" (as perfectly as humanly needed) already utilizing the appropriate filter.

      2. I can appreciate your comment of a "general public" definition of what is audibly transparent as opposed to the "audio science" version of this definition. However, ultimately it's the same, just a matter of thresholds. Each person will have his/her own threshold of where their hearing can no longer tell a difference. For some, maybe 12-bit resolution is all they need, others maybe 15-bits. Likewise, some will not be able to hear past 12kHz whereas the younger healthy listener can hear up to 16kHz, etc...

      So long as the DAC has the ability to reproduce the audio content up to and beyond the personal ear/brain resolution, then they would have achieved "audibly transparent / perfect" for that person. Note that environmental conditions also play a role. If I were listening on a quiet evening in my sound room with noise floor down at 30dB SPL at reference playback levels, I would need a more demanding DAC than in my living room during the day cooking in the kitchen.

      "Audio science" guys might throw out numbers like noise floor minimum -100dB below music peak, flat response 20-20kHz, etc... as "safe" values that suggest these would satisfy the broad spectrum of listeners from youths to geriatrics. Almost certainly, each of us would not need such extremes of "perfection" to saturate the limits of our abilities.

  6. Thank you for the test, Archimago! I can imagine the amount of work and efforts it took. Such tests have additional value for me as they help to compare subjective results with objective ones, which I can compute using my df-metric. For the purpose I used your reference test track and the three recorded samples. The graphic below summarizes the comparison of your listening test results to waveform degradation of the Samples tracks -

    1. Very cool Serge!
      Wondering is the df-metric psychoacoustically scaled? For example are there frequency weightings applied? I haven't published the full test suite for the Linn streamers and I know there are oddities with the Klimax that I did not see with the cheaper Majik that some folks might find interesting; this might make a difference on analysis of complex musical signals.

    2. To be honest I'm trying my best to avoid a psycho-acoustic scaling in df-metric. But this is naturally happening in some sense because the distributions of errors are computed using the real music signal where frequency components are already “weighted” naturally (and resembles standard PSN test signal). In other words, the most errors that we see on the histograms refer to the most loud freq.components in the music test track. And freq. profile of any music corresponds to human hearing freq. profile, so it's a kind of psycho-acoustic ”shaping" of df-metric results by design )).

      The real reason for avoiding a psycho-acoustic scaling of any objective measurements is the absence of necessity for it - today the very accurate audio devices can be produced cheaply. The devices that reproduce an initial music waveform with great accuracy. Lossless (transparent) audio tract is not a technological problem anymore. I'm sure that sooner than later such “inaudible” audio tract will overcome in popularity the current “psycho-acousticaly suitable” design exactly like lossless encoders have out-marketed mp3,aac,... in hi-fi because the bandwidth is no-longer a bottleneck. The precise reproduction of any music signals is no-longer a bottleneck for audio chip makers. DF-metric is well-suited for the case.

  7. The histograms have additional dimention - intensity of sound, coded with brightness. In each histogram bin the 50ms portions of the signal are sorted according to intensity (louder = brighter). It's visible that in Linn devices the louder portions of the signal are distorted the most. In Apple dongle such dependency is less pronounced. My guess is that the cause of such degradation pattern in Linn devises is analog circuitry (OAs and other active elements).