 |
| Topping DX9 DAC sitting on Linn Klimax DS/2 which we'll talk about another time. |
Hey guys and gals, thanks for the comments and feedback over the years on this blog. It really helps to refine the testing and of course explore broader ideas about audiophilia and what it is we're after in this hobby. I think it's important that we're free to discuss and critique not just the hardware, but also the ideas (including the philosophy and psychology) that "we" as individuals believe in. There's so much unsubstantiated audiophile cultural beliefs out there based on myths such that everything should be up for examination regardless of who says what. It's not unusual to find that audiophile review "high priests" are wrong, and some "heroes" of the audiophile world are nothing more than false prophets.
It's also good to get feedback suggestions on the published measurements, for example in the recent series on the Topping DX9 flagship DAC (based on rare AKM AK4499EQ chip), Nick in the Part III comments suggested that I consider this:
"Hi Archi,
Prior to version 5.40.78 beta rew had a gross error in IMD vs Level plotting, the results were underestimated by about 12 dB. It looks like the version you used has this error.
IMHO there are not enough performance measurements at high frequencies for a complete picture. If still available, it would be very interesting to see THD (without noise) vs frequency performance, accurate THD measurement at 10kHz and wideband fft (without weighting) with a weak signal such as the signal for the DR test.
Thanks!"
Excellent feedback and suggestions Nick, let's focus on checking this out in this post!
I. Updated REW IMD vs. Level graph
Okay, so I updated REW to 5.40.Beta 85-Windows 11 64-bit (June 1, 2025) which is the latest version as of this writing for both my data-capture and analysis computers. As usual, a big shout-out to John Mulcahy for the fantastic software that continues to evolve with even more features! For audiophiles and home theater hobbyists, this is an essential free tool for understanding our systems including performance in rooms with the ability to capture/analyze multi-inputs through the reasonably-priced Pro upgrade.
Previously, I was using 5.40.Beta 50. Let's see, with the update to Beta 85, what an IMD vs. Level graph looks like down to -120dB using the "Linkwitz IMD" tones - 1:1, 1.0/5.5kHz.
Performed at hi-res in 32/96, distortions analyzed between 20Hz-20kHz, 128k FFT, left channel using the E1DA Cosmos ADCiso + Cosmos Scaler as measurement combo:
II. THD vs. Frequency at different output levels and wideband!
Okay, for this second part as suggested by Nick, let me run THD vs. Frequency sweeps, from 20Hz to 20kHz at hi-res 32/192. Let's calculate the THD up to the 10th harmonic, FFT size increased to 256k for accuracy. Turned off the 20kHz low-pass filtering (as well as 20Hz high-pass) such that at 10kHz, harmonics up to 90kHz would be included in the calculations, and at 20kHz, we'll calculate harmonics up to 80kHz (4th harmonic).
I trust that's wide enough bandwidth well beyond human hearing, or the ability for consumer speakers to reproduce with any accuracy! 🙂
Let's look at the THD vs. Frequency 1/6-octave sweeps overlaying 0/-3/-6/-12/-20/-40/-60dBFS graphs:
Interesting morphology of the graphs. Notice that at around -20dBFS, we see the most frequent-dependent increase in THD; in particular, there's an increase from 5kHz to 10kHz.
Let's have a peek at the 10kHz THD(+N) FFT -20dBFS to see that distortion pattern in detail:
As expected, at 10kHz, the 2nd harmonic is already out at 20kHz. Notice that the highest distortion levels are out beyond 50kHz.
For comparison, here's the graph of 10kHz THD(+N) FFT 0dBFS:
 |
| Peak 3rd harmonic at 30kHz, merely -118dB! |
Notice much less distortion/harmonics >50kHz compared to the -20dBFS graph above; interesting.
On the other extreme, here's the 10kHz FFT at -60dBFS. Very low output level around 10-bits of attenuation, often used in DR/SNR measurements:
Nothing scary at such low output levels. There are a couple of high frequency noise peaks out beyond 80kHz, but even these are at most -60dB below the primary signal (or -120dB below an actual 0dBFS level!); nothing that would even be remotely of concern assuming we used a wideband amplifier and ultrasonic speaker for bats!
For completeness, here's the THD+N vs. Frequency sweeps overlaid. Since we're going wideband without filtering the ultrasonic noise, the overall +N will be elevated, drowning out much of the underlying harmonic variability.
I hope this extra discussion answers questions about whether with a DAC like this there's much to be concerned with when it comes to frequency-dependent distortions out to the ultrasonic range. While ultrasonic harmonics would not be audible, it's still worth looking at this once awhile just to make sure there's nothing there that could be causing intermodulation into the audible range.
In short, there's nothing here to be concerned about with the Topping DX9 DAC. As per discussions on "perceptibly perfect" DACs, we're at a point in history where objectively high-resolution DACs are clearly achieving very low noise, very low jitter, and very low distortions such that we're simply not going to be able to differentiate between them in volume-controlled blinded listening tests.
Of course, some subjective-only folks who never try controlled or blinded listening (especially those acting as advertising mouthpieces), as well as companies that make such devices (especially in the luxury "high-end" needing to justify prices) will want to insist that sound quality is significantly different between objectively hi-res DACs.
Unless verified, I trust that the "rational audiophile" will be able to ignore nonsensical chatter and calmly debate basic truths using facts when he must (such as on the audio forums), without getting into needless flame wars. 😉
--------------------
 |
| The emotional dynamics of subjective reviews and comments. Let's be clear, sometimes (maybe even often!), unhappy, frustrated comments are warranted. Unhappiness could very well be justified. |
I noticed that in the last month or so, Stereophile's website has started to curtail the opportunity to leave comments for a number of their articles.
For example, it looks like audio show post (such as this and other recent Munich 2025 items) no longer are open to comments, while subjective pieces like music reviews which typically are less prone to criticism remain open to feedback. Unfortunately, it looks like recent hardware reviews like for the Amphion Krypton 3X speakers, and the Fezz Equinox DAC are now closed to comments also. I see a few snuck through with the latter post, suggesting that change happened after or on June 1st.
This looks like a continuation of the deterioration in reader-engagement on the website. They shut down their forum section I believe without notice at least a year ago.
I don't think it's surprising that this has happened. In fact, Stereophile has kept comments open for longer than other traditional magazines like The Absolute Sound or Hi-Fi+ that have closed their comments section years ago (years before 2020 at least). Other sites and channels have also been selective about whether comments are allowed (like this discussion about Darko.Audio back in 2020).
I think it's actually quite logical that over time this is happening as audiophiles become more knowledgeable about the performance of the gear and the utility of being able to assess quality based on objective parameters rather than testimonial subjective recommendations. As I've expressed before, there is value in subjective opinions when it comes to the look-and-feel of devices, quality of the user interface, opinions on robustness, ideas about quality of customer service, etc. However, when it comes to sound quality, I believe there is little that subjective magazine reviewers can meaningfully add to discussions if they don't verify such beliefs. For knowledgeable audiophiles, many reviewer claims can be easily tested and commenting on this can be interpreted as criticisms against the subjective reviewer (sure, sometimes it's so obvious that it's mind-boggling that reviewers don't bother checking before making obviously unbelievable claims!).
The old-skool hierarchy where "respected" subjective reviewers like Michael Fremer, or John Atkinson, or Robert Harley, or others having much power in shaping sentiment, driving the narrative towards particular products, is long gone. These days, we also have a host of YouTubers who wish to take their place; but this too is going to be rather fruitless I suspect, without providing objective context.
Stereophile, as a media outlet sponsored by the Industry through advertising, does not look good if reviews of US$100,000+ speakers are typically given an enthusiastic subjective write-up, but prove to have middling measurement results. Routinely, the open comments section works to amplify such discrepancies. No wonder then that open feedback can be seen as a thorn in the side of many of these reviews.
We will see... My suspicion is that not only might comments be further curtailed, but we might see the diminution of measurements in the pages of Stereophile in the near future. John Atkinson has been doing them for decades and I imagine full retirement can't be far off; he already stepped down from Editor-in-Chief in early 2019.
Comparatively, the less detailed measurement results from sister publication Hi-Fi News and Record Review would be a downgrade if just copied and pasted into Stereophile; I see they've started sharing results such as for the dCS Varèse DAC already. Furthermore, for years, Atkinson's speaker measurements have been superseded by the Klippel Near Field Scanner data produced by ASR and Erin's Audio Corner. Even if Stereophile continues to produce detailed measurements, they should update the testing procedure and graphs especially for loudspeakers.
If (when?) the measurements section weakens or even disappears, there would be little to differentiate Stereophile from the likes of The Absolute Sound.
Addendum: June 18, 2025
I checked Stereophile last night and noticed that ALL comments seem to have been removed from the website for all articles!
Just like them shutting down their forum without notice, I think this is actually very disrespectful to readers because most commenters are not trolls and have spent time to think and compose the comments. Oh well. Maybe this is really the end of what made Stereophile at least the better magazine for audiophile information here in North America, fine job Jim Austin.
 |
| Clearly no severe compression or peak-limiting being applied. |
Hey Arch, great post, I always enjoy these in-depth tests and measurements, although admittedly a lot of it goes over my head :) And I probably misunderstand a big chunk of the rest :D
ReplyDeleteCurrently I'm looking into the odd distortion behaviour of Cirrus Logic CS431xx series DAC chips. Very luckily the dongle I got recently - Shanling UA1 Plus - is not one of the affected devices! It sounds amazing to my ears, although apparently it still doesn't have ideal/perfect behaviour w.r.t the distortion issue. I find it fascinating, and there's a great in-depth post about the topic on ASR:
https://www.audiosciencereview.com/forum/index.php?threads/on-the-distortion-of-cirrus-logic-cs431xx-based-devices-a-comparative-review.63038/
So while your quest for the perfect DAC has likely ended with the DX9, mine continues!
Btw, I guess you missed this, but a few weeks ago fgk asked a fascinating question which I'd be very curious to get your expert opinion on. Re-posting it below. Thanks Arch!
>>
"Digitally generated test tones are often used to measure D/A converters;
it is important to choose test frequencies that are not correlated with the
sampling frequency. Otherwise, a small sequence of codes might be
reproduced over and over, without fully exercising the converter.
Depending on the converter’s linearity at those particular codes, the output
distortion might measure better, or worse, than typical performance. For
example, when replaying a 1-second, 1-kHz, 0-dBFS sine wave sampled at
44.1 kHz, only 441 different codes would be used over the 44,100 points. A
0-dBFS sine wave at 997 Hz would use 20,542 codes, giving a much better
representation of converter performance. Standard test tones have been
selected to avoid this anomaly. For example, some standard test frequencies
are: 17, 31, 61, 127, 251, 499, 997, 1999, 4001, 7993, 10,007, 12,503,
16,001, 17,989, and 19,997 Hz."
This quote is from Ken Pohlmann's book "Principles of Digital Audio".
Archi, don't you think that in view of the quoted above testing a DAC with 1 kHz sine wave may not test it properly?
The test files are usually 24-bit, so the change from 441 to 20542, when expressed in % of all codes tested, is from 0.003% to 0.12%. Not sure if it is worth the trouble.
DeleteIf you really wanted to test a significant number of codes, you could generate a tone like 999.991 Hz, which is nicely self-dithered, but you would have to measure for a long time. Like, after 5 min still only ~50% codes would have been used.
And besides, other tests, which are more sweep-like (e.g. distortion vs frequency, distortion vs level) surely use many more codes than 441 or 20542.
Btw, there was some discussion about self-dithered signals in this thread:
https://www.audiosciencereview.com/forum/index.php?threads/yamaha-cdx-393-review-cd-player.57600/page-2#post-2110987
> Btw, there was some discussion about self-dithered signals in this thread:
DeleteBut note that the context there is testing CD Players, so 16-bit is enough.
Hey there guys, interesting discussions on the use of single-tone testing and the PCM values that are actually represented in the signal. Has anyone seen an example these days where a 1.000kHz tone tests beautifully but something like 0.991kHz clearly resulted in significantly more severe distortion?
DeleteMaybe I'll give this a try. In any event, I would agree with the general sentiment that a "one number objectivism" assessment is incomplete. So while it's a fine "fidelity screen" that we can compare and rank THD+N/SINADs of devices, it's of course important to have a look at multitones, frequency responses, IMD, sweeps, etc...
I wasn't aware of the magnitude of the "Cirrus hump", MB. Interesting to see. Yeah, these days I'm very happy with the high quality later-generation devices based on AKM or ESS chips. :-)
Hi amigo, interesting article. It'll be fascinating to see what happens with subjective-only review sites and paid sponsored magazines—not just in the audiophile world, but across every other hobby. As AI becomes more popular, people will lose all trust in written content. Nothing written will ever be completely trustworthy again. This could spell the end for many YouTube stars and other influencers as valuable sponsors.
ReplyDeleteHow long will it be before this current generation of children grow into sceptical adults, having seen nonsense social media pages from birth promoting rubbish through obvious lies? Will this constant exposure numb them, or will it force them to become sharper thinkers—more analytical, more questioning, better able to spot falsehoods and separate fact from fiction? Growing up in the midst of noise and manipulation might not weaken them, but instead prepare them to navigate it with more clarity than we ever managed.
Ironically, this shift could lead to higher standards in education. Truly intelligent, well-studied teachers may end up being more respected and valued, simply because they'll have to sift through far more misinformation than we ever did as children or adults. Their ability to guide others through the noise might become one of the most valued skills in society.
It also raises the question: what does it mean to be a creative human? Where exactly is the line between authentic, human-created writing and AI-produced content? Take this very text as an example. The ideas are mine. The perspective is mine. But the wording, the clarity, the structure—much of that comes from AI. It’s not pure automation, but it's not purely human either. So is this a collaboration, or is it something else entirely? At what point does something stop being a product of human expression and become more of a synthetic construct generated by a lifeless programme with no understanding of anything?
We're approaching a time when this distinction won't just be philosophical—it’ll affect our sense of identity. If everything becomes co-written by machines, what becomes of authorship? What does it mean to say, “I wrote this”, when so much of the final result comes from an external intelligence?
Hey there Dan,
DeleteThanks for the great comment and thoughts about this important topic of "trust"! This is going to be very interesting to watch and follow as it unfolds.
Trust forms the fabric of society and in the big picture, it is what holds use together. I hope that you're right in the idea that watching social media erode with untrustworthy content might actually result in intelligent humans compensating. Developing vigilance, strengthening critical thinking skills would be great and rewarding the teachers and skilled thinkers would be nice to see! That would certainly be an improvement over superficial "influencers" and conspiracy-theory prone podcast hosts.
Maybe the luxury "high end" audiophile hobby could serve as model. After all these decades of falsehoods and bizarre beliefs, folks are wondering "How Long Will the 'High End' Continue to Exist?".
While I don't think hobbyists desiring high-fidelity sound will completely disappear any time soon, I think the subgroup of companies that sell snake oil products like cables, tweaks, crazy-priced DACs/speakers/amps might be ripe for a major "correction" under their own weight of unbelievable claims and skirting with pseudoscientific attestations.
As for the topic of authorship, those are intriguing points. When it comes to the comment you wrote with the help of AI, I would consider that a genuine reflection of your thought process because I presume the AI writer expressed what is in your mind and you proof-read it for accuracy. In that sense, the AI "tool" just formulated the words and polished the sentence structure.
Where I think it gets scary is with image and audio creators. Clearly when I tell the image generator to create a "Cute bunny rabbit running through a field on a sunny summer day." That idea translated to a fully formed hi-res image with the bunny, grass, trees, the sun, maybe throwing in a lake in the distance, was mostly determined by the AI's processing than the contribution of that simple statement. In that sense, I would attribute the AI as being the creative agent just like if I commissioned an artist to draw such a picture.
As such, I would already attribute creativity to AI. Even if the human user is the executive agent supplying the prompt for creating the intended output. The effect of this on the arts and the need for certain types of "artists" is already in the midst of massive displacement!
I'm involved a few enthusiast hobbies; golf being one of them. I see similar trends to the high fidelity market in that there are a few equipment reviewers emerging that are doing good and detailed objective reviews to show the measurable differences between various pieces of gear. I think this is a good trend. Certainly good for consumers that desire to be better educated in their recreational areas.
ReplyDeleteFantastic to hear Doug,
DeleteWhen technology matures, at some point companies/industries will tend toward selling things based on sentiment because there's not much gains to be had to differentiate between products of various companies objectively any more - or we just achieve the upper limits of "diminishing returns".
Hobbyists, users, and I would argue the news media itself will need to "self-regulate" when they see too much airy-fairy nonsense and get the pendulum to swing back towards reality. I think this is why we've seen the rise of ASR, of measurements on YouTube, more "objective"-leaning commenters, headphone measurements, or blogs like this one.
I believe there is a hunger out there to simply get back to the truth of how things work, and what actually makes a difference given all the disappointments and false promises based just on opinions and biased assertions we've been exposed to. This is a good thing to see; to me it represents a maturity out of wishful fantasies without concrete demonstration of actual performance.
Good stuff as always here, for years now! Also, the comments often bring great value. I hope that they can remain here in the future even when others shut them down.
ReplyDeleteDespite the superb performance of the DX9, I decided against it because I'd like to have an analog input, like that Luxsin X9 on ASR.
I did get some Canadian music last night - my first ever concert with Alanis Morissette. 30 years since Jagged Little Pill, and it's still one of my favorite albums. The sound wasn't very good, but she still sings great, so overall I really enjoyed it.
Greetings Freddie,
DeleteHey that's great. The Luxsin X9 looks nice with that RCA in and ARC as well I see. Plus a sub-out is great. For my needs, I'll be pairing the Topping DX9 with my AV receiver so analog input along with sub processing will always be going through that.
Yeah Alanis has been a star of the Canadian pop/rock music scene for a long time now... Along with the likes of Bryan Adams, Celine Dion, Shania Twain, maybe even Justin Bieber if he's still a Canadian citizen. 😁
Why do you buy expensive dacs like this topping if you believe there is no audible difference. You d better highlight cheapset ones that are perfect. Actually on your amtp files i don't like It very much. It's paradoxal to be enthousiaste about expensive products that give nothing more that cheap ones. Truth is that each chip handles peaks others differently and some DSP can have bas effectue and amplify this problem. At full volume with most music that is too lound mastered dacs will handles differently intersamples and sound différent. As lot of people are listening with cds and as streaming is even loudness and as people are not educated with this problèmes it's very probables that intersample is an explication of différent sounds of dacs.
ReplyDeleteMaybe Review, the intersample overs are a problem for those who listen to highly compressed music extensively.
DeleteI generally don't find it a problem since I use ReplayGain routinely which will protect against that. I also seek out better masterings with higher DR if possible. And then with multichannel content (which can be easily folded down to 2-channels), we're again better protected as discussed.
As I explained in Part I of the Topping DX9 review, I bought it because it looked cool, seems well-built, uses the rare-in-2025 AK4499EQ DAC chip inside, and is a great headphone amp!
I'm happy to spend the money since none of this has to be related to sound quality; yeah, I think $100 DACs can sound the same to my ears but won't give me the same "pride of ownership" if we want to lump the reasons above together into one psychological category! To me, that's what luxury and "high end" has always been about even if magazines and subjective reviewers claim it's "about the sound quality".
The solution for intersample other is to have headroom and It would be Nice to have a recommandation for each dac of a Level of atténuation recommanded. And to know if the digital atténuation of the dac is before the filter and can solution the problem
ReplyDeleteWell, if you use digital volume control, attenuate by -3dB and the vast majority of problems will be solved. I don't see this as a major complication these days.
DeleteModern hi-res DACs can easily accommodate this without losing audible resolution. Alternatively, use ReplayGain (as discussed) and this likely also will solve the majority of problems. Bennett Ng addressed much of this years ago.
Sure, every once awhile there will still be clipped samples here and there with -3dB attenuation (Bennett in his testing of many of his albums showed a good number of tracks with +3.5dBTP and above), I don't think it'll be audibly significant the majority of the time. As audiophiles, highly compressed albums are by nature poor and one of the domains in which we have to be mindful of in order to achieve best quality. The real solution is to fight for better quality recordings from the artists and record labels!
"Well, if you use digital volume control, attenuate by -3dB and the vast majority of problems will be solved."
DeleteFor the purpose of attenuating the volume digitally to avoid intersample clipping, in Foobar I prefer to use AirWindows BitShiftGain VST plugin (https://www.airwindows.com/bitshiftgain/). It can **losslessly** shift the volume down by -6 dB (i.e. 1 bit). Obviously, if you do it with 16-bit audio, you then need to output your audio as 24-bit, because everything has been shifted by 1 bit. If you do so with 24-bit audio, then, accordingly, you need to output it as 32-bit.
In Foobar, we now can also control the digital volume losslessly with this Foobar component: Amplifier DSP (https://foobar.hyv.fi/?view=foo_dsp_amp), as the author of this component made it possible recently when he rolled out the latest update:
1.0.5 (2025-03-29): Changed 6 dB, 12 dB and 18 dB to switch to bit-accurate scaling - 6 dB will perfectly halve or double the signal, 12 dB will quarter or quadruple it, and 18 dB will divide or multiply it by 8.
Hi Arch! Thank you very much for the measurements. I can notice that the THD behavior at high frequencies of the ES9039Q2M (E1DA 9039S) is similar: -110dB@-18dBFS, -91dB@-24dBFS. The DR wideband test was interesting to see the ultrasonic noise level.
ReplyDeleteHey there Nick,
DeleteGood to hear of what you're getting. The newer generation, higher quality converters at least from AKM and ESS are performing excellently as long as no issues with the opamps and other parts of the DAC design.
Enjoy the music!
I used to enjoy the Stereophile articles, not because I trusted the reviews or had anywhere near enough funds to purchase any of the gear, but because I enjoyed reading audio based articles by Herb and Art. They few YouTube videos of those two simply chasing about audio were great.
ReplyDeleteI find it sad now that not only are the comments off, but also the quality of the writers is diminished. Further, sites like The Absolute Sound now simply credit "TAS Staff" for most of their articles.
Yup, I agree The_Fall.
DeleteSpeaking for myself, even if I disagree with a writer and might openly criticize the article/review, there was still joy to be found in reading the articles and in appreciating the artistry of the writing. I think engagement is good for the magazine because it also increases web traffic for the site.
I don't know if it's the quality of the writers these days or just I've read too much of the same-old hype and can't stomach a lot of it anymore. Looks like TAS is trying hard to get into the YouTube video space. They're also broadening into technologies like higher priced Bluetooth headphones. Away from focusing on longer, detailed articles - hence the generic "TAS Staff" writing maybe.
I wonder if Stereophile might also be wanting to try to pivot more into videos? Don't know if there's much charisma to tap when it comes to engaging presenters on their staff. 🤔
Stereophile is bad, but "The Absolute Sound" is a scary place. It is full of meaningless flowery prose and all the logical fallacies known to man, especially when it comes to digital audio. For some reason I occasionally read articles there though, and I regularly have to stop thinking through the implications of digital interconnects having subtle effects on sound, since this is so monstrously and fractally wrong that my brain threatens to core dump.
ReplyDeleteLOL gzost,
DeleteYup, TAS is already and maybe Stereophile in the days ahead will become the "shadowlands" of the audiophile pursuit.
Articles wherein our sound rooms and inside the audio hardware itself, ghosts, hauntings, UFOs, monsters exist but sanitized by using science-inspired physically-based terminology like "noise", "jitter", "distortion", "PRaT", "dynamics", "micro-detail", etc. usually with different, non-physical, unmeasurable meanings only defined by the subjective projections of the writers.
Of course, these demon-like deficiencies need to be rescued (at times exorcised) by way of typically expensive products from experts and audiophile heroes, sometimes using only the most advanced quantum technologies!
Occasionally I'll have a peek at TAS which can be entertaining as a guilty pleasure like how I might want to put on a "documentary" on UFO's or even the occasional horror flick. 😱 Not that I would actually *buy* anything based on their recommendation.