MQA again

MQA has once again floated to the surface of the perfectionist-audio pond—not belly-up as some have hoped but forced there by relentless pursuit by anti-MQA predators posing as impartial jellyfish.

In case you've been living amidst the seaweed at the bottom of the pond, MQA is a thing that's done to digital audio data—sort of a codec that is said to reduce the "blurring" repeated digital conversions cause. For high-resolution audio, MQA compresses data into a much smaller file or stream while retaining, it is said, the sonic benefits of the original high-rez audio (and also of "deblurring"). MQA appears to be genuinely clever and legitimately new, implementing post-Shannon developments in sampling theory that have not previously been applied to digital audio.

MQA is controversial, for good reasons. In the interest of making the sound better, it alters the sound the mastering engineer and musicians signed off on. It would replace open standards with proprietary ones—a big step backward, especially for internet-libertarian types who think information should be free (a perspective I respect but don't fully share).

On the positive side, MQA partisans (including esteemed recording engineers Peter McGrath and George Massenburg, multi–Grammy Award–winning mastering engineer Bob Ludwig, and hi-fi and music critic Jason Victor Serinus of Stereophile) love how MQA sounds. What's more, MQA takes small steps toward addressing provenance issues in digital audio (although in this respect it does not go far enough).

This new attack on MQA comes from "GoldenSound" on YouTube (footnote 1). His findings may seem damning, but there's less than meets the eye. I am not a partisan, for or against MQA. If I'm partial to anything, it's fairness, and GoldenSound's critique is unfair. MQA isn't perfect, but here it has been falsely maligned.

"GoldenSound" managed to get two music files accepted by Tidal and encoded in MQA. (Kudos; if I had known this was possible, I'd have done it myself a long time ago.) He presents results from two tracks: one with a sampling rate of 44.1kHz, the other at 88.2kHz. Inside these tracks, he embedded an impulse, white noise, a squarewave, a 32-tone test signal, "the entire RightMark Audio Analyzer (RMAA) test sequence," and a 1kHz sinewave at –60dBFS. The 88.2kHz file added two ultrasonic sinewaves (with summed amplitude very close to full-scale) and a series of full-range sweeps (right up to the Nyquist frequency, at levels that appear to be about –6dBFS).

Let's talk first about the high-rez (88.2kHz) data, since that's the easiest to dismiss. The 88.2kHz file embeds high levels of ultrasonic information, including those two sinewaves, the full-range sweeps, squarewaves, and white noise right up to 44.1kHz.

Ever since its announcement in October 2014, Bob Stuart and the MQA team have been clear that MQA is intended for music. It is not intended to deal with high levels of ultrasonic information (which are not found in music)—certainly nothing close to full-scale. Also well-established—nothing new—is that MQA by design allows more aliasing than other methods. It's hardly surprising that, in a file containing so much ultrasonic information, high-level aliased signals appear in the undecoded analog signal in the audible range. The high levels of noise seen in GoldenSound's data, too, would be lower with data that more closely resembles actual music. (Apparently realizing this too late, GoldenSound submitted a third file with far less ultrasonic content, but by this time MQA and Tidal had grown suspicious; GoldenSound's original files had been removed, and the upload failed.)

What about the 44.1kHz data? GoldenSound's test-file–embedded music tracks still do not resemble real music. The test file has very high levels of information at the very top of and above the audible range—right up to the 22.05kHz Nyquist frequency. In the test spectrum shown in the video, at 6:30, you can see the white noise going right up to 22.05kHz at what appears to be about –6dBFS. Parts of the RMAA test include full-scale information, also up to the Nyquist frequency. The squarewave's harmonics do the same, although with decreasing intensity as frequency increases, as squarewave harmonics do. Full-scale data at the Nyquist frequency is not something MQA, which is intended to optimize time-domain behavior and so is very careful about what it does near Nyquist, is intended to deal with; in contrast, a conventional DAC that employs a traditional brickwall filter would have little trouble with such data, although it would do the usual damage (if damage it be) in the time domain. Perhaps there is real, electronic music in the world that pegs the meter near 20kHz, but I wouldn't want to listen to it.

GoldenSound's –60dB, 1kHz sinewave is a low-level signal that is intended to test dynamic range. The result in his test file was high-frequency noise that was "only" 45dB below the signal level. This noise level is probably higher because of the issues I've already discussed, but even so, high-frequency noise 45dB below a –60dB signal—itself barely audible—won't be audible at all. (The much higher level of high-frequency noise in the 88.2kHz track was surely a result of the excessive levels of ultrasonic information in the file.) What GoldenSound's tests show, then, is that a bull in the china shop can damage the china. The solution is not to throw out the china but to keep out the bull.

GoldenSound's tests are a missed opportunity.

Footnote 1: MQA's (the company's) rebuttal can be found at here Also see the excellent comments from Pablo Banñados of Santiago, Chile ("Mieswall"), here and elsewhere in the same thread.

smetzger's picture

Why do this when bandwidth no longer requires compression? Seems like an answer to a problem that does not exist. Why find a solution to a problem you don't have?

NeilS's picture

Why find a proprietary lossy "solution" to a problem solved by non-proprietary lossless, e.g., FLAC.

jhwalker's picture

FLAC (and I'm assuming ALAC, as well) allows for discarding / optimizing bit-depth to match the needs of the source material (while remaining lossless), if bandwidth savings is needed. I want my music lossless, not the newest form of lossy MP3-like perceptual encoding (along with potential DRM limitations).

jhwalker's picture

MQA potentially allows some disk space / bandwidth savings on the distribution end, but I'm paying for hundreds of Mbytes of download bandwidth, so I'm happy to download or stream full-sized lossless files, thanks.

Anton's picture

If MQA does, in fact, improve the sound, then make it an ‘after process’ as people download or stream lossless and let MQA process the original file on my end after it arrives.

I can stand or fall as a stand alone processor, no need to insinuate itself between the listener and the ‘seller.’

I listen to MQA via my Meridian DAC streaming Tidal, so I am not a hater, I just want consumers to have the choice to ‘fold’ or not ‘fold.’ (Descriptions of Tidal remind me of the spice space ‘folders’ in Dune.)

As mentioned, as streaming hits lossless, MQA should become discretionary.

Let MQA make its case as an aftermarket toy.

Terozzz's picture

It does, the pre echo is eliminated. Sure you can do with regular DAC too BUT the filtering is HARD to make and mostly cost a lot.
Ofc you need real full MQA decoder to do so but if your other equipment is up to the task you can spot it. On my +4000€ STAX set it's easy to spot.
Every note is sharp like drum hits and guitar picking.

TimorEtPax's picture

Because streaming lots of ultrasound for no particular reason makes no sense.

John Atkinson's picture
TimorEtPax wrote:
Because streaming lots of ultrasound for no particular reason makes no sense.

See my comments on the meta-analysis discussed by Joshua Reiss in his AES presentation, where he showed that hi-rez audio can be distinguished from 1Fs/16-bit audio:

When I master my own commercially released recordings for release on CD I spend a lot of time experimenting with decimation, noiseshaping, and dithering strategies so that the Red Book files sound as close as possible to the hi-rez originals. The final difference can be very small but it is still audible.

John Atkinson
Technical Editor, Stereophile

TimorEtPax's picture

I didn't expect JA answer. I didn't criticise Hi rez in this particiluar problem but only as an answer to the first statement. I disagree with that since every bit consume energy. I will stop here since I'm not that much in green apology. I understand that MQA uses high sample rate because it needs sample time density according to the latest findings about time sensitivity of out hearing organs. For that reason MQA uses high sampling rate, and to avoid so called brick wall filter issue. The result is capturing lots of ultrasound that contain no musical information. Thus it is irrelevant for the music content wheather it is lossless or lossy. At least, that is an offical explanation. It terms of streaming that content, it makes a lot of difference. Correct me if I understood something wrong.

dschamis's picture

I had this same thought for a while. People are streaming Netflix day and night which requires much more bandwidth than a hi-res FLAC file. It's as if someone identified a problem in 2005 and it took 10 years to "solve it" with MQA, and no one noticed that the world changed in 10 years.

tnargs's picture

It’s interesting that you get to point this out, while Mr Austin prefers to focus on criticising the criticisms, instead of the products themselves.

Why doesn’t Stereophile say MQA is the missed opportunity, instead of focusing on Goldensound’s test?

While musing upon that question, my eye wandered to the right-hand-side of the Stereophile page I am typing on. “Silver Sphere — the AUDIOPHILE’s HDMI” it proclaims. Ahhhh.

skris88's picture

Because a lot of people go for the hype, not common sense! For example when I tell people Full HD 1080 movies are just 2 megapixels stills 24, 25 or 30 times a second they don't want to believe that and so continue to keep taking 20 megapixel stills and block my Internet connection their HUGE rubbish - iPhones being the worst culprit (there's almost zero apps to allow you to shoot in 1920 x 1080 resolution) - PLUS Apple loves to take 10 shots per click. What a great way to fill up phone memories and sign up larger and larger iCloud storage!

martynbaker1's picture

Seems prudent to include Goldennote's response (though it may be missing here due to publishing dates).

DDvinyl's picture

I am not an "internet-libertarian types who think information should be free" and I am willing to pay for streaming, CDs, and vinyl. But I do take issue with a format that trades lossless files for lossy ones with REDUCED choice. Pre-MQA, one could choose her/his filter of choice, implemented on the "master" file - be it 44.1 or a higher rez one. MQA does away with user's flexibility, while at the same time also manages to charge us (either directly or indirectly). So, I don't mind paying, but when I do, I expect to get more, not less.

tedmingo's picture

Why in the world would I want to use any lossy proprietary compression that limits my choice of DAC. I'm sure there is a reason you don't see MQA on any FPGA or R2R top of the line DACS.

TimorEtPax's picture

There is no limitation for MQA file, it can be streamed without any decoder like it probably is streamed since most of the Tidal users streams it throught their phones without mqa decoder. FPGA can and is adjust for MQA like Paul McGowan stated several months ago in his daily video. Otherwise, mqa is in flac, alac ,aiff or wav I suppose.

JohnG's picture

I recently acquired a dCS "Bartók" streaming DAC and it does indeed do MQA. The dCS Ring DAC is an FPGA device.

tedmingo's picture

Ok, I should have said it's very rare to see a top of the line DAC with MQA decoding. Why DCS chose to add the MQA filter had to be marketing decision. You know that you bypass the FPGA circuitry when the MQA filter is used. Why would anyone choose do do that on a DAC of that quality.

JohnG's picture

I didn't know that. I'll have to try a few recordings in both formats and see if I can tell the difference. dCS also says their player/music management system Mosaic sounds better than Roon.

AJ's picture


A Comparison of Clarity in MQA Encoded Files vs. Their Unprocessed State as Performed by Three Groups — Expert Listeners

This paper aims to examine perceived clarity in MQA encoded audio files compared to their unprocessed state (96-kHz 24-bit). Utilizing a methodology initially proposed by the authors in a previous paper, this study aims to investigate any reported differences in clarity for three musical sources of varying genres. A double-blind (listening) test is conducted using three groups—expert listeners, musicians, and casual listeners—in a controlled environment using high-quality loudspeakers and headphones. The researchers were interested in comparing the responses of the three target groups and whether playback systems had any significant effect on listeners’ perception. Data shows that listeners were not able to significantly discriminate between MQA encoded files and the unprocessed original due to several interaction effects.

The authors would like to thank.....Bob Stuart and Meridian for providing the MQA-encoded versions of our audio source materials

skris88's picture

Why not 320kbps MP3 as well? 99% of listeners would not have noticed a difference there either!

AJ's picture

Dunno why other lossy codecs beside Master Quality Aliasing weren't tested, maybe they weren't reinventing the wheel. KISS?
I'm just the messenger AJ, not McGill U or BS.
Do note that only unauthentic Master Quality Aliasing has the blue Pavlov light for their subjects, so somewhat unique.



AaronGarrett's picture

Libertarians are all about property rights, intellectual or otherwise and privatization. Do you mean "internet anarchist types"?

dcolak's picture

"It's better than lossless!"
"It's bigger than infinity!"

TimorEtPax's picture

If you carry a sack with small package in it, wouldn't be wise to fold it for easier transport since it is full of air? A what does that do to the package itself? Nothing.

vintagemintage's picture

I'll admit I'm not an audio engineer but if MQA has nothing to hide (and it very much seems they do ) then why didn't they let GoldenSound upload that third musical sample to vindicate themselves? They and I suppose you as well would rather folks pay for 'useless' at best algorithms and the chips/licenses to decode them.

Anton's picture

I saw an absolutely genius thing in my newest Stereophile!

The “Reviewers Playlist” on Qobuz is a fantastic move! I hope it draws many new subscribers!!!

DH's picture

What a tendentious biased column:
A) GO didn't "attack" MQA he calmly and politely tried to demonstrate that it isn't lossless. The fact that you call that an "attack" - reveals a lot about your bias.
B) Wow, you appealed to authority to "prove" how good MQA sounds. So what?
Other experts don't like it. Neither proves anything. Is that your level of argument? Why do feel you need to "prove" MQA sounds good?

You wrote back in 2018 that MQA needs to be properly tested. Yet MQA has never let anyone do that. Why do you think that is? Maybe they have something to hide?

vintagemintage's picture

This might be the most cheerleading article I've ever seen, but then again I don't read stereophile that often anymore

Archimago's picture

- Considering Internet bandwidth these days can easily accommodate true lossless hi-res audio.
- Considering that lossless (hi-res) audio CODECs for years can encode anything that comes along without the limitations of what MQA has deemed to be "music" based on their limited testing.
- Given that Qobuz has been streaming lossless FLAC for years up to true 24/192 without arbitrary limits like Tidal & MQA.
- Considering that Apple is going with ALAC for Apple Music and Spotify is coming out with their lossless stream (no evidence that it'll use MQA).
- Considering that MQA has never shown evidence for actual sonic "de-blurring" or that their DSP added any value to sound quality.

MQA is very much "last decade" technology developed prior to 2014, aiming at data bandwidth limited streaming and is now anachronistic. Why would audiophiles who by definition want the best quality want MQA now?

Why is this even being talked about in a mainstream audiophile magazine that's supposed to be promoting state-of-the-art fidelity? Jim, if you're trying to be helpful, why don't you start encouraging Tidal to stream true lossless audio and get with the future like everyone else and encourage Tidal to reverse MQA's data manipulation that has affected even 16/44.1 streams?

We know these days that MQA processing has been applied to the vast majority of the files with no actual oversight by the artists or production team so the claim of "what the artist intended" is meaningless.

The fact that GoldenOne had to submit his files to Tidal to get it encoded to MQA was the only way non-insiders (unlike John Atkinson) got a chance to examine test signals all these years. If MQA is all that great, why didn't they just encode stuff submitted to them including some true hi-res tracks by folks like Mark Waldrep years ago? I am personally aware of some other test signal material that would likely, relevantly demonstrate significant limitations of the CODEC but of course they would never allow these to be encoded/tested.

Face it Stereophile, audiophiles as represented by the comments above IMO have done the right thing and rejected this abomination which we never asked for. Letting this go "belly up" is exactly what respectable media desiring to promote the interests of the readership need to face up to unless as the title of this message suggests, you can give ONE good reason why audiophiles should support this now. (Promoting the agenda of a highly questionable company so they can make money to pay off their debts is certainly not an appropriate reason of course.)

supamark's picture

Not everyone in the USA, let alone the world has access to broadband internet. Most people access the internet via mobile devices, which tend to have data caps and limited bandwidth. By compressing the data MQA gives access to higher quality audio to a larger group of people than would otherwise have access to it.

MQA is also kinda useful when reviewing DACs. Comparing two DACs with MQA, the differences are almost entirely down to things other than the digital filter with MQA files which is about impossible otherwise. I also find it, for my personal taste at least, to be one of the better sounding filters - I'm not a big fan of linear phase filters generally which are really common, and prefer a slow minimum phase or no filter (NOS).

Mark Phillips
Contributing writer, Soundstage! Network.

Bonus section! If you're curious why Adele's 21 sounds so bright and etched it's not just EQ - it wasn't mixed in the box so there's a lot of extra D/A/D conversion steps which screwed up the tonal balance. Oh, and way too much compression of course, but that's Andrew Scheps' thing (he mixed it). He seems to have an intense hatred of dynamic range... did it kick his dog or something?

Archimago's picture

1. If Internet mobile is that poor and data caps can cost $$$, then just stream 16/44.1 FLAC and turn off hi-res streaming (Apple Music will do this). Heck, stick with MP3 lossy in some places. Before the pandemic when I did some work and travels in Asia, many places were strictly lossy for streaming. [Since even 24/96 has barely shown benefits in listening tests, why bother with "hi-res" regardless if one is so data-capped when mobile?]

2. I don't agree that MQA provides better sound quality in general based on my own listening. That's my subjective opinion of course.

3. The MQA filters are really poor quality. In my testing, almost every modern DAC can perform better filtering natively. Remember that their algorithm caters to low-power microcontrollers dating back to the first Dragonfly DACs or the old Meridian Explorer 2. I routinely test a DAC's filter quality and these days ASR does it as well and John Atkinson's measurements also do although not as stringently.

4. For your own listening, why not upsample yourself with Roon for example or even better yet HQPlayer if you want the best quality? In fact, if you use Roon, then select the upsampling there and voila, you can easily compare the same filtering performed on multiple DACs that accepts high sample rates. If you like NOS, then MQA will be of no help to you!

Thanks for the bonus on Adele ;-). Interesting background, will have to pull it up again and have a listen!



supamark's picture

You didn't ask if it was a good reason for everyone, but it is A reason. One likely reason I like the MQA filter is that it's a minimum phase design (no pre-ring) with a slow roll off so there's little post ring. It's also great for apples to apples comparisons between MQA capable DACs. I seem to be more sensitive than most to digital filters' ringing, and they can do really weird things to my perception of tone and imaging. I once told my editor that one filter setting on a DAC felt like it was punching me in the ear, but put it more politely in the review of course.

I'm actually reviewing a ladder DAC right now and prefer it with no filter even after I know what the lack of reconstruction filter looks like on a 'scope (ugly). It does have an OS mode with a couple linear phase filters but they sound kinda chalky in comparison. I've been curious what using something like the Chord M-scaler with a NOS DAC would sound like but doesn't the act of oversampling require a filter and create ringing? I'm not an expert on digital audio theory (recording engineer and biochemist by training/education) so I'm genuinely curious.

As far as why not just listen to mp3's, in my experience hi-rez files tend to be mastered with less compression/limiting (I really like dynamics, hate the loudness wars) so even with a less than perfect MQA codec the listening experience is better for me. Some hi-rez files are still stomped on but it's far less common than CD quality masterings.

I don't have Roon (yet), but then neither do most music listeners. Most of my music library is still on physical media (vinyl, CD's) though I've been using Tidal a lot more lately so Roon probably won't be useful to me until I get around to ripping my CD's at full rez. It's really surprsing how much MQA encoded material is mastered at 24/44.1, like why bother other than to get that "MQA" tag?

Mark Phillips
Contributing writer, Soundstage! Network

John Atkinson's picture
Archimago wrote:
Considering that MQA has never shown evidence for actual sonic "de-blurring"

I discussed this subject at length with Charley Hansen of Ayre before he passed away. Based on those conversations I published an analysis of time-domain issues in A/D and D/A conversion in 2018:

The slow-rolloff "Listen" antialiasing filter in Ayre's QA-9 A/D converter introduces no ringing at high sample rates - see fig.11 in my article. Charley had developed an experimental, complementary reconstruction filter for Ayre D/A converters that, in conjunction with the QA-9's "Listen" filter, would result in a digital recording/playback system that would perfectly reproduce transient information. (The tradeoff is that both filters are slow-rolloff types, so there will be some aliased energy present in the reconstructed analog signal. However, at sample rates >48kHz, this will be inconsequential.)

Charley sent me the firmware for that filter (which Ayre never released), but I no longer had access to an Ayre D/A converter in which it could be installed. However, I found that reconstructing the time-perfect digital data with the slow-rolloff MQA filter in a Mytek D/A processor did result in excellent transient performance - see fig.16 in my article.

John Atkinson
Technical Editor, Stereophile

AJ's picture

Now all we need is old guys demonstrating hearing "actual sonic" 22k filter temporal "blur" issues, rather than aliasing back down in the "actual sonic" bandwidth.
We know that the highest weighted "trained" listener test in Joshua Reiss/LandR infamous "Meta Analysis" - Jackson, et al., “Further Investigations of the Audibility of Digital Audio Filters in a High-Fidelity Playback System,” J. Audio Eng. Soc., 2016...turned out to be a complete bust. Never made in past review (yet somehow still part of the Meta Analysis).

SNI's picture

That would take many hours of sleep I guess :-)
Besides what has already been discussed, it is also appropriate to look at the need for very high sampling rates in the first place.
As You surely all know, sampling rate is the same as frequency range.
So higher Fs means larger frequency range.
But do we actually want this?
The answer is mostly yes, because this eases the task for the reconstruction filter, which can be made more simple.
But there is a downside to that, and that is DAC performance.
Looking at datasheets for a bunch of DAC chips, you´ll soon find out, that performance peaks @ arround 100KHz. Going any higher than that, performance will drop on most DAC´s. ESS is an exception, because the chip itself includes an onboard ASRC, besides that you cannot download their datasheets for study without NDA an so on.
Beside that the whole talk about pre - and postringing seems to be pretty much exagerated or widely misunderstood I think.
The passband ripple of an almost 20 years old DAC chip as the TI PCM1792 is "Pass-Band Ripple: ±0.00001 dB" :-)
And this is a very deep digital filter with an astounding stop-band attenuation of 130 dB.
In the analog days a taperecorder had easily audible pre-echo due to magnetic reasons, just as vinyl records also has this "feauture" due to mechanical reasons.
So what are we actually discussing?
I hope everyone knows that the shape of the impulse in a typical phase linear digital filter is stop-band behavior, and thus has nothing to do with actual music signals. It is the filter behavior towards unallowed frequencies way beyond human hearing.

Anyways disregarding the fuss, I´ve been experimenting a little with the MQA audio experience.
I have both Qobuz and Tidal subscriptions, and using Audirvana for both streaming services, the outputs are in both cases treated the same way as ASIO.
So software differences between Qobuz and Tidal software should be elliminated this way.
Then finding some tracks with MQA @ Tidal and true Hi-rez @ Qobuz is pretty easy. I quickly found something I know pretty well.
You just can´t use classical music, as Tidals catalog is really bad compared to Qobuz.
Anyways a lot of people prefer MQA it says, but I´m not one of them.
What came to mind first was the artificial "airiness" of the MQA stream.
It is some strange subtle kind of whizzling, that surrounds voices.
The 3D experience tends to be the same most of the time, MQA gives you a wide and pretty high soundstage, but does so on all tracks.
Switching to Qobuz you instantly hears the dryness and no BS sound of music with details as a part of the music, not as music in it self.
Qobuz is clearly more holistic, and the 3D experience varies from track to track, and is more subtle too.

As I see MQA, it is a way to manipulate music files in a certain direction, chosen by some peoples taste of music. In this case Bob Stewarts.
For me that is absolutely nogo.
If Bob Stewart could prove what is actually wrong with PCM and the sampling theorem, one might consider to do something about it, but this has never happened, and if true, it might even result in the great honour and glory of a Nobel price for proving Nyquist/Shannon wrong.

All in all MQA seems to be a solution to a problem, that does not exist in the first place, which is the basic reason it will do no good at all.
Maybe it could have been justified 10-15 years ago or so, but not today. If you have DSL bandwith problems, you will probably not be able to stream anything else than MP3 anyways.
MQA reminds me in some way of HDCD, which also needed special decoder hardware, but just as MQA has been overtaken by cheap broadband internet connections, HDCD was overtaken by SACD and DVD audio.

May the force be with us all, and correct this matter ASAP

Best regards


hfvienna's picture

To defend a technology, which is maybe not the favourite one, but misused by a self-proclaimed expert show real quality of a reviewer. Looking into the details and answers not only from MQA themselves turned golden sound into golden shower. And I am surprised how many of comments ride on losslessness, without ever checking the exact process as described many times in its basic procedure. Also I am surprised how many people dismiss the limits of bandwith still happening in a lot iof areas. And deblurring time smear makes a mqa file better than a lossless blurred one, like it or not .

hb72's picture

I suppose, phase-correction might take place also on-board in real-time, instead of the multi-stage MQA unfolding process, i.e. using original flac or similar converted by conversion SW that reads required AD parameters from the flac file (meta data comes to mind).
That would be a rather open approach (definitely realistic if a contemporary PC was used in RT-phase correction process as is AFAIK already required for some levels of the unfolding process). The makers of MQA however have chosen not to follow that one.

Advantageously, an on-board / real-time phase-correction would permit direct comparisons of original vs corrected music peace, as the reference flac can be validated as such.

ok's picture

net-working and net-studying, when all family members are constantly connected, bandwidth does matter for many of us; not that mqa is the last word in bandwidth saving, mind you. I personally find irritating the fact that mqa does not allow for full software rendering, even though it should't matter soundwise. I also think that the "lossy" tag is rather misleading: if I understand correctly, any loss of information takes place exclussively at ultrasonic territory. Allegations for indecency etc do not really concern me as an end-user as long as I can stream freshly sounding hi-res music without having to sell an arm and a leg. Artists' consent and royalties are mostly nominal of course as they always have been. By the way Jack Dorsey (of twitter) is the new Tidal owner, so I guess the whole thing is here to stay one way or another.

hb72's picture

disqualifying tests which involve non-musical signals in turn means disqualifying development and oem-internal testing, and, most peculiar, practically all component test results ever published in Stereophile.

I believe your column here requires some corrections.

tx and best regards

Ivano's picture

My Chord DAC does not support MQA, so I do care about purity of non-MQA encoded files. If GoldenSound's finding is correct that Tidal's non-MQA stream is in fact encoded, that would be a huge downside to Tidal for me, I may very well reconsider my subscription.

Stefan@airmusic's picture

I have the same concern and it is interesting that Tidal responded by giving subscribers in Australia a choice between MQA and non-MQA files by changing the subscription options.

Ivano's picture

Using Roon you may choose between MQA and regular files. But GoldenSound claims that "regular" file hass exactly the same content as MQA encoded one, but with no MQA flaq, so it does not get decoded. That means music is degraded by encoding process.

Stefan@airmusic's picture

The credibility and trust of MQA and Tidal are at stake in this dilemma and the lack of transparency is counter to current business ethics. Take consumers for a ride and they will respond with their wallets. The sarcastic and condescending tone of this article makes me question how much Stereophile is invested in all of this. A technical debate is one thing but taking sides is another. I think it is a missed opportunity for Stereophile not for GoldenSound. I was waiting to see how Sterophile would respond and it is a disappointing response.

Stefan@airmusic's picture

I guess that if this is true " ..that a bull in the china shop can damage the china." I would be careful not to be seen as being too close to the china or be spending too much time protecting the china shop. Collateral damage is a real risk for Stereophile who is nurturing those who are in the perfectionist-audio pond with the kind of products they review on a regular basis.

zimmer74's picture

who thinks "everything should be free." I pay for my music. But I am not willing to use a proprietary data format that restricts the equipment I can use to listen, and which demands royalties for that "privilege."

tiagoramossdg's picture

I used it when I had Tidal for a few months (before Qobuz was released in the U.S., I think). It was fine, but I did not notice anything so extraordinary. Then again, I was listening to it on my LG phone at the time, a V30, if memory serves, and Audio Technica headphones. Nothing esoteric or bleeding edge, but reliable for sound reproduction. In any case, I later switched to Qobuz, left MQA behind and have not missed it since.

I have, however, casually followed the discussions around the technology since then, and to be honest, I tend agree with a lot of people commenting here: what is the point?

I'm open to have my mind changed, but it seems that MQA is, as we might say in my native Brazil, a truly ingenious, high-tech, sophisticated, machine to straighten bananas.

thethanimal's picture

“The data gets hashy above 22 kHz!”
- Are you a bat, or concerned about the delicate ears of the bats in your attic? Does your amplifier even cover that range? Can your speakers even reproduce that signal?

“There’s low-level noise below the low-level signal!”
- That’s already lost in the noise floor of your amplifier? Further lost by the distortion of your speakers? Further lost by the noise floor of your room? Or do you listen at a level that reproduces pink noise at 110 dB-A?

MQA is very clear that it chops up the signal and repackages the data in areas where the music signal is only a few bits deep — at the higher frequencies. So GoldenSound’s tests seem designed to exploit and exaggerate this design using signals that aren’t representative of anything anyone listens to for musical enjoyment. Using Bach’s Toccata and Fugue seems a better full-frequency-range and high dynamic range test file. Konstantin Reymaier has a recording of it from 2020 in MQA on Tidal and 24/96 on Qobuz. Maybe GoldenSound can do a follow-up analysis of both of those files and then listen to them and decide which he likes better.

And isn’t which one you like better the point? Because neither version is “accurate” and neither version will put you in St. Stephen’s Cathedral in Vienna, but perhaps in my room and system MQA sounds best to me and perhaps it’s non-MQA for you, and how can you dictate to me which I perceive to be best?

As for Tidal, almost every album for which I can select MQA I can also select a 16/44.1 version. If the CD quality version is actually encoded in MQA my Bluesound Node would show the MQA signal, but it doesn’t so that argument seems like a red herring.

vintagemintage's picture

I believe the flag to start MQA decoding is just removed for 16/44.1 but the original data is still the MQA'd sample

Anton's picture

In his reply to the reply, Golden Sound repeatedly entreats MQA to "work with him,' of "let's work together."

I like how he only wants to be their colleague.

boxster233's picture

It's a great counterargument to the anti-MQA crowd about this latest bit of research. I enjoy MQA in how it sounds, but also agree people should have a choice and they do.

Our industry is rife with people wanting to achieve "lossless", "original", "as intended" or "perfect sound". I haven't found any of those yet, but in MQA I have found filters and a sound I enjoy and appreciate that.

Archimago's picture

You do realize that MQA stands for "Master Quality Authenticated", right?

Phrases like "exactly as the artist intended the track to sound", or just the word "authenticated" were introduced by MQA themselves.

MQA are the ones who claim some of those things you "haven't found".

Nothing wrong with enjoying the sound or liking the filters, to each his own.

Georgios's picture

I had compared MQA on Tidal and Spotify (Premium) with the same tracks (Yo-Yo-MA Bach Trios) with the result of minor differences between the 2 streaming providers. So I prefer Spotify due to the better prices for streaming.
If MQA has one better resolution on high frequencies (> ??kHz) that is one other story, considering my age (60 I don't care), the differences I recognize between the 2 providers are minimal.
My opinion about Hi FI enthusiasts regarding recordings is the following.
Only small ensembles can be reproduced in a nearly perfect way.
A big Orchestra unfortunately never will be reproduced at home as live. Here your HiFi system after one visit to one live concert and you will close it.

Leotis's picture

"I am not a partisan" does not ring true. Gratitude to GoldenSound for exposing MQA. Canceled my Tidal. Qobuz sounds better to my ears.

Audiochrome's picture

On one side we get "relentless", "predators", "impartial jellyfish", "internet-libertarian types", "attack", "unfair", "bull in the china shop".

And on the other side "genuinely clever", "legitimately new", "post-Shannon developments", "controversial, for good reasons", "the interest of making the sound better", "esteemed", "Grammy Award–winning", "love", "very careful".

rt66indierock's picture

The people you don't want to see the video are sharing it, executives in the music industry.

bigasherm's picture

My issue with MQA is that it is a proprietary technology that adds costs at various points in the music chain. MQA also gives control of the MQA processes and standards to whoever owns MQA. What would happen to MQA if Google, Sony, Apple, or some other large corporations bought out the rights to MQA and changed the MQA specifications or jacked up the MQA licensing fees? That would not be good for audiophile listeners or musicians. I will stick with Qobuz/FLAC.

cgh's picture

I hesitate to even comment, it feels like wasted breathe. This whole MQA issue is part of a broader issue, arguably intertwined with the broader currents of IP, earnings, monetization, etc. The (hifi) audio industry has arguably been on its way out, catering to a diminishing demographic (in the mortality sense), saddled with a reality that innovations towards a "golden sound", catering to the "golden ear", is diminishing. We can argue this last point, but speakers and amps and tonearms, et cetera, haven't really changed, while prices have more than kept up with inflation. And we talk about masters and sound quality and other, frankly, non-falsifiable attributes, at least on the consumer side. In a parallel universe SPhile could have lead this discussion and evolved with the external realities, with merit or not. I am not saying MQA is the cable debate. But for someone that can tell trickle-down use cases, etc., it seemed superfluous, lacked transparency, suffered from hyperbole in the press, etc. In a lay-cum-scientific space there's a simple attribution from lossless to MQA that could have been championed by it's creators and adherents and, while still debated, less so. Maybe if there wasn't all this history it would have been easier. Maybe a CBA validates all this nonsense from BS, Meridian, Tidal's perspective. But it doesn't help. The credibility isn't there. I can't help but think this echoes the current "beyond experiment" argument in fundamental physics, in that audio (im)perfectly timed having it's moment when the rest of the collective unconscious was ready for "beyond experiment" and subjective trumps reality. This will all play out, arguably predictably, but it just seems stupid to me.

TimorEtPax's picture

I think the main reason for tech guys is the problem in measuring because all the measuring is mostly based upon sine function. According to the Bob Stuart, he does not use sine function, but function that correlates to the natural sound, then treated by SN theorem of double the frequency for the desired frequency range. And that is only one part of the tech apart of the other features that are discussed lively. But nobody discusses about that part - the function. And I have never understand that SN theorem apply only and exclusevly to the sine function. It is the difference between engineering and fundamental math. Prople who knows hiw live instrument sounds are fans of MQA. Like JVS who is mentioned in the article. So, he uses knowladge the recognize what is wrong with adopted trueth and that is always the hard way no matter how does it end. So it is common conflict between faith and ratio.

hb72's picture

"all the measuring is mostly based upon sine function"
No, it is not, here in Stereophile and many other sources, Dirac impulse is part of the measurement catalog. see here for example:

"And I have never understand that SN theorem apply only and exclusively to the sine function." - I suppose you mean Nyquist Theorem with "SN"?
if so, well it is about frequencies, so a cyclic function of 1 frequency it must be, which can be reproduced from digital source (PCM) of 1 sampling frequency.

What is more important about Nyquist's Theorem is that mostly the fact of infinite time as a requirement for reproduction is neglected. So, solid reproduction of a very short sequence of high frequency components within audible range (say 20kHz, although..) is done better on much higher than 44.1kHz sampling frequency.

TimorEtPax's picture

Very good, I thought Shannon- Nyquist as SN theorem. I understand that there is a pulse measuring device but the reference is still sine function to comapre with. Comparison between the reference and measuring results gives e.g. THD based of an assumption that we know how it should be or look alike. Don't we? I think that Bob Stuart made adjustment to the reference based upon his research about propagation of sound throught the air, 1/f rule and sensitivity of human auditory system which can recognize something unnatural. So it is adjustment of computing power in reconstruction to satisfy human ear, not measuring device. That is my opinion. Since the human hearing is not regular, why are seeking so much regularities and rules. That is how we satisfy our rational part of the brain which wants order. Order in hearing and sounds is very different (fletcher munson curve for e.g.). I think that paradigm of BS goes in the right direction. Time will show. I always say - math is a tool, not the goal.

hb72's picture

In this context, music for human sound perception vs Test-Signals for test purposes, isn't there a grey area? Am wondering what this MQA does with music of Kraftwerk or Giorgio Moroder's famous plip (the latest recorded versions of which might have never seen an AD-Converter and it's limitations).

TimorEtPax's picture

It doesn't do anything that mastering engineer do not want. It all came out of question: what is wrong with digital music reproduction? It offers some of the solution. If one find no value in it, it can simply listen the same file in cd quality so every one can hear for themselves and decide. I think it is a fair offer. Blip, and similar can only benefit from it since the mastering engineer can use mqa tools to improve the quality of sound. If there is none of improvement, then let it be. We will hear soon.

rt66indierock's picture

Bob Stuart himself told me the encoding process changes the music at the LA Audio Show in 2017. A mastering engineer would need an encoder or a plug in to duplicate MQA output.

What is wrong with digital recording is people and processes. I can make very good digital recordings with an ancient Pacific Microsonics Model 1 (24/44.1 ADC/DAC) or a Sony Digital Recorder and I'm a professional in another field.

What will hear about MQA soon? Another press release to feed to the audiophile press.

TimorEtPax's picture

MQA alters the sound. Fact. But mastering engineer can hear the difference before decides, not after. And it has to authenticate it. Thus the name. I'm not an advocate and fan of mqa, but challenge my understanding of mqa with the others to understand it better. So far, it is a rare critic that I found useful. As a trial user of Tidal out of curiosity, I can say that there are lots and lots of title with mqa label so it is in progress. Lately,in his interview BS said that 10 millions aongs are already in mqa and they are signing and expanding in radio bussiness. Those are just fact and you can hear for yourself in his interview with Lucy Hegdes or something similar.
I believe that you are able to make a great recording (by your own standard) and that means that you have great future in your bussiness. By "time will show" - I think that the market will decide about mqa so I'm not in any kind of worry or emotional attachment. Simply, I don't care becauss I will continue to listen the music and enjoy myself. What I don't understand is bitterness and heavy critics on the formar which is matter of choice. Some people pay for qobuz, some for spotify, others have amazon, apple... if people do not find added value in mqa, it will be a history. Like sacd. So far, srill in game. In avaiting the final outcome.

Good luck and make many good recording for us to listen regardless of mqa.

TimorEtPax's picture

I think that MQA is an acronym for the Master Quality Authenticated and what I have read so far itmeans that Mastering engineer gives his final blessing upon end version of what the audience should hear. General misconception among audiophiles is that all the music then should be audiophile grade. Misconception. But MQA ia a tool for musicinas and crew to express themselfs and their values. So MQA does not alter the music, quite the opposite - mastering engineer authenicate its version and locks it.

Kursun's picture

It’s a good thing Stereophile pushes MQA.
I take note of the reviewers that try to push MQA and I ignore them later. I do not read any review written by them.
I see the real faces of some reviewers.

dschamis's picture

Is that really what MQA is trying to do? So it's not just trying to replicate hi-res in a smaller file than the equivalent FLAC (still not sure why, but OK) - but it's actually trying to make the original master "sound better"?

Defending MQA keeps getting harder and harder.

ok's picture

so many scientifically inclined people who happily wear masks like second skin and do what they are arbitrarily told to do interested in Bob Stuart conspiracy theories.

TimorEtPax's picture

Very good!

monetschemist's picture

Jim writes:


MQA ... would replace open standards with proprietary ones—a big step backward, especially for internet-libertarian types who think information should be free (a perspective I respect but don't fully share)

The issue with proprietary formats is not that the information they contain is free or otherwise.

The issue with proprietary formats is that they are bad for the consumer. They limit the consumer's choice in that only devices and software that license the the format and technology can be used to fully enjoy the content provided therein.

FLAC (for example) specifically does not limit ANYONE's choices, which you can easily see since MQA uses FLAC to package its data.

People who build software and hardware to decode FLAC (for example) don't have to worry about whether the good people who invented FLAC are around tomorrow or not, because all of the FLAC intellectual property is free for anyone to use.

People who buy music encoded in FLAC don't have to worry about whether or not some company will be around tomorrow, like those of us with HDCD encoded music who have to rely on reverse-engineered HDCD decoders of suspect quality to deal with HDCD format data.

ArthurK's picture

Entertaining exchanges! the meantime and while waiting for the outcome of the 'war', I will put another record on my ClearAudio deck..

Happy days!

C_Hoefer's picture

GoldenSound says he was trying to evaluate what MQA does, and whether it can merit the marketing claim of "lossless". His long video discusses this question on many fronts, not just in the context of the trick files he uploaded in order to probe what MQA does. JA2 only responds to a few of the points that GS makes (and the same is true of MQA's response).

Jim does not discuss facts such as these: (1) MQA claims to use knowledge of the production chain to implement its "de-blurring" magic, but in most cases this is clearly impossible (many aspects of provenance unknown, multiple D/A/D steps in the original mastering, etc.), and in the case of GoldenSound's files, MQA never contacted him to ask about the equipment used in creating his digital masters. (2) Files that are altered and damaged can still light up the blue "authenticated" light, which shows that it is not authenticating anything, it is just checking for a flag inserted during MQA encoding. (3) MQA's claims about the alleged virtues of the encoding have evolved over the years, while never leaving the realm of impossible-to-verify marketing-speak. I could go on, but hopefully I've made my point.

JA2 is the one who missed an opportunity here, to frankly discuss the full range of awkward questions raised by the GS video, and keep Stereophile neutral on this incendiary issue. (Or move it towards neutrality at least... the early history of cheerleading for MQA has not faded from my memory.)

Maybe MQA encoded files do, in general, sound as good or better than lossless counterparts. If this could be demonstrated in an unbiased context, that might sway me to keep Tidal. But as things stand, I will be moving to Qobuz this month. The straw that broke the back of this camel was learning that Roon has to pay MQA something every time I stream an MQA encoded song in Roon. I would rather keep Roon financially healthy than MQA!

LouJetlag's picture

JA2, good article, thanks. When listening to Tidal, I usually prefer MQA. However, I'd been reflecting lately on good outcomes from the pandemic, thinking that one of those good outcomes was that "The Great MQA Debate" had been killed off...but I see now that it had just been lying dormant, or perhaps mutating like a virus variant, judging by the number and length of the comments...but I shouldn't be surprised...I just "upgraded" part of my digital sources, and now I'm torn by the fact that greater detail is (apparently, to my ears) now accompanied on many Tidal recordings with some harshness and, now I'm off to Qobuz to check and compare...such is the nature of the Audiophile hobby...the variants always get you...
Thanks again and Cheers!

P.S. Thanks to you and all the staff for ALL that you do for us, music, and our hobby! Much more than only equipment, I have discovered so much new and wonderful music thanks to your reviews, as well as the comments sections. Blessings!!

Lou G in Clinton, NJ

Wavelength's picture

The biggest problem I find with the GoldenSound video is they used only one dac to do the testing.

Personally I like the sound of MQA. If you don't or find a differential in sound between say a download, MQA or other streaming services then sure.... go with what you like.

Really all this bickering is dragging down this hobby. Why are people spending thousands of hours pissing and moaning over stuff when they could be doing something positive.

This is suppose to be fun!!!

PapaCharlieBravo's picture

I have a few issues with your article attacking the Goldensound review of MQA.
1) You say "I am not partisan, for or against MQA." You are clearly defending it in this entire article without any criticism. This reads much more like a damage control article written by MQA.
2) If your product does what you say it does, it should be able to be peer reviewed and the reviews should confirm its performance. Instead they lock it down in such a way it is nearly impossible to do so, and when someone finds a way to test it they shut it down and hide behind marketing wank.
3) When anyone or any company says "trust me" but dont verify you should be highly suspicious.
4) If you dont like how Goldensound tested this product, perhaps you can work as an intermediary to allow a test that both parties can accept, since you seem to be on MQA's payroll. Prove your truly impartial.
5) There has been many other reputable people critical of MQA, ie. Paul McGowan (PS Audio)

ChrisS's picture

...makes cars, then get a Toyota.

Like everything else reviewed here, MQA is a consumer product.


rom661's picture

I'm not familiar with the samples referred to by the article. It sounds like they are probably BS. I dame to the same conclusion about this article within the first two paragraphs as he defends MQA by listing the people who endorse it. Hey, remember Von Karajan's "All else is gaslight" when the first mostly wretched DG CD's came out? I've heard it sound good. I've heard it sound like you are spending money for a log. I think it's largely marketing but don't really care that much about it either way, this after attending several MQA demos. The DAC's are never the same. How the hell do you really tell anything. My 2/192 stuff sounds great, I don't need MQA for storage savings so why worry about this article? Because it struck me as biased and somewhat dishonest, much like what it is criticizing.

John Atkinson's picture
rom661 wrote:
I'm not familiar with the samples referred to by the article. It sounds like they are probably BS.

You can find Stereophile's articles on MQA and related subjects at

rom661 wrote:
Hey, remember Von Karajan's "All else is gaslight" when the first mostly wretched DG CD's came out?

I was at the press conference in Salzburg, Austria when HvK uttered those immortal words!

John Atkinson
Technical Editor, Stereophile

thethanimal's picture

It’s all just semantics. JA1’s previous articles have pointed out that nearly half the bits in a 24/88.2 file are noise, much less the higher sample rates. So MQA claims lossless because they don’t lose any music data while origami folding in those noise bits. GS, and others, claim it isn’t lossless because you lose those noise bits. Both claims from MQA and GS are true, from a certain point of view. None of that is obfuscated by marketing.

But what matters for sound quality: musically lossless or empirically lossless? Judging by the comments on this thread and the dearth of comments on the music reviews also posted within a day of this article, most of us care about empiricism rather than music. That’s fine, but loving our 24/192 FLAC doesn’t mean others can’t love MQA or dragging a rock through a ditch.

What JA2 was pointing out is that you can’t embed a full-scale test signal at 20kHz as a test of MQA and then call the results into question; it just proves you didn’t take the time to learn how MQA works in the first place. Questions on IP, long-term support, provenance, etc. may remain, but do not suddenly validate GS’s flawed technical hypothesis or results.

By the way, I listened to the organ music I referenced in my earlier post via Tidal Masters and Qobuz at 24/96. Tidal won using iPad > DragonFly Red > Sony MDR-7506. Rhiannon Giddens’ “Calling Me Home” was also more natural through Tidal Masters. Both those and Ma/Thile/Meyer’s “Bach Trios” were a draw on my main system, but I also know my acoustic environment is still lacking optimization.

skris88's picture

MQA is just another scam for our hard earned $$$. Spare cash? Send it to a charity that looks after the millions of poor and suffering people who make up the majority of the world!

Try this test -

90% of us won't hear a difference.

And that's before even MQA!

skris88's picture

PEOPLE CANNOT HEAR THE DIFFERENCE. This is how I've proven it: I show my audiophile friends my collection of 16/48 CD rips in raw WAV but play the 320kbps MP3 versions for them. They leave saying how great my system is!

cgh's picture

Not sure who is actually following the original content. I had watched the original video around the time it came out. Based on a response from MQA there's another, more recent, video. It also touches on some of the Stereophile position.

agim's picture

i haven't heard MQA yet, but i'd like to know how mr. fremer thinks it compares to LPs. after many years of ignoring vinyl, i've discovered that with a modestly good turntable and phono preamp and cartridge (total under 1K), even high rez digital doesn't compare to a clean record. an $8 2-LP set of The Doors vs the remastered hi-rez of the same is easy to demonstrate to anyone: the hi rez sounds wonderful, and is a great experience of listening to a recording of The Doors; however the vinyl is a great experience of listening to The Doors actually performing, as a band of musicians, and somehow what each of them is doing is much more obvious and less "blended". does MQA bring any of that back? i don't know if the technical explanation is due to noise, time-domain performance, transient response, or what. but when michael listens and compares, what does he hear? i'm willing to invest in MQA if that same "un-blended" performance can come through in a digital format.

MikeSmith's picture

As mentioned before, MQA is a solution for a nonexisting problem.

There is no such thing as post Shannon. What a nonsense. I suppose Austin studies digital signal processing.
Its absolutely clear that MQA is damaging the signal.
HighRes, the aim is to extend frequency range and bit depth. People claim you can hear ultrasonics. So HighRes PCM just reproduces ultrasonics in a perfect manner. MQA does not.
If you do not want HighRes, PCM gives you that as well. CD 44.1/16Bit or 44.1/24Bit.
So you get what you ever wanted with PCM today. Hardware (DAC) are perfect in regards of the ability to hear. So technically there is no need whatsoever to go for MQA or any othere strange stuff.

Jim Austin's picture

There is no such thing as post Shannon. What a nonsense.

Do you really think they stopped research on that subject after Shannon? "Yep, Shannon finished the subject, there's no more to do." If you think that, you know nothing about how science and math progress.

No, I'm not a DSP expert, but yes, I have studied DSP. No, I'm not a signal-processing engineer, but as a former physicist who long ago earned a PhD, I do know enough math to get the gist of some pretty technical stuff.

Start here.

That article has been cited by others almost 900 times.

Jim Austin, Editor

MikeSmith's picture

Teh theory is still valid and will not change even if politics sys so.
I hold a Phd in electircla engineering and I am a member of IEEE, the organization you cited the article from. I have not seen any IEEE aricle who denies the sampling theorem. You may claim different as do the 10001 perpetuum-mobile inventors.

I worked in signals processing, AD and DA conversion. There is no free lunch. If you violate the sampling theorem you get aliasing whatever clever modern tricks you may use.
So there are no wonders to be seen. The progress in science and math has not changed the sampling theorem whatsoever.

MQA is one of the worst claims I have seen. Just stay away of MQA.
Just nonesense and useless.

Jim Austin's picture

I certainly didn't, and I'm quite sure the MQA folks didn't either. To suggest otherwise is just plain dishonest.

What I did was suggest that research continued after Shannon, that meaningful progress was made, and MQA appears to incorporate some of that research.

Even though you're a member of the organization that published that article, I'm betting you'd never seen it before.

You wrote, "There is no such thing as post Shannon. What a nonsense." It's astonishing that someone with an EE PhD who had worked in DSP could write something that was so obviously false.

Jim Austin, Editor

MikeSmith's picture

Sure there is work after Shannon. But this work does not change anything at the Shannon theorem. So I am astonished you try to suggest that.

So what is clearly false you say I mentioned?

Clearly false is that MQA is improving the signal qulality.

Jim Austin's picture

"There is no such thing as post Shannon. What a nonsense."

What is there to misinterpret? That is what you wrote.

Jim Austin, Editor

MikeSmith's picture

I repeat: There is no news to the sampling theorem. It is as Shannon postulated.

That is what I said.

MQA is not changing it whatever MQA claims.

Jim Austin's picture

MQA never claimed to change the sampling theorem. That was your misunderstanding.

Have you read any of the more recent work in the field?

Jim Austin, Editor

MikeSmith's picture

Its just needless to discuss it anymore. The facts are crystal clear.

MQA is worsening the signal. Point.

And as here an op already mentioned, I need MQA as urgent as a whole in the head :)

ChrisS's picture

....Ford cars?


adamdea's picture

It is disingenuous to say that MQA did not claim to change the sampling theorem. There is no point startign off with innuendo and then defending with pedantry. The express "Post Shannon" is bandied around as though it means something, ie that in some sense Shannon has been supplanted. of course any fule no (pace post modernism)that the prefix "post" can have many meanings.
If it merely means "coming in chronology after Shannon" then pretty much all work in information theory is "post Shannon".

So why do you claim that MQA implements "post Shannon" developments? Why not post Shannon compromises?

So no, the reader did not misunderstand what MQA (or for that matter you) said, the reader understood the innuendo correctly. Of course it's rubbish and of course it should not be parroted.

If you wish to be taken seriously as an impartial commentator then it is you who should be apologising for using such a misleading (and marketign parroting) expression.

Jim Austin's picture

When Einstein devised his General Theory of Relativity, that didn't invalidate Newtonian mechanics. That's not a perfect parallel, but it's not terrible either: The original theorem remained valid--of course--but subsequent work generalized it, and in so doing, allowed innovation, new ways of thinking. This is most obvious in image processing. Shannon is now a subset--a narrower, special case of a more general result.

One way this matters is that it allows for a broader range of sampling kernels. Another is that the modern theory incorporates error in a more sophisticated way: Instead of just requiring perfect band-limiting, as Shannon does, error due to imperfect band-limiting (which in Shannon is left out of the calculation, since perfect band-limitation is merely assumed),now "undersampled" data can be treated and the error determined.

MQA has not been transparent about exactly what they're doing--most proprietary technologies keep secrets--so it's difficult to say precisely what aspects of more general sampling theory are included, but it does appear that MQA incorporates, if not B-splines precisely, then something similar. And it's obvious that they have relaxed insistence on perfect band-limiting (as others had before them).

Because I cannot say whether what MQA does is an improvement in a purely theoretical sense (because they are not transparent about the theory), I am not an advocate, as I wrote above. My position is, and has been all along, that with only a few exceptions, criticisms of the technology have been ill-informed and often just plain ignorant. Even the better critiques tend to approach it from a traditional, Shannon's-theorem perspective. And the slander and libel against Stuart and Craven, both of whom have earned respect via a lifetime of contributions to our field, contributions that many and perhaps most critics are unaware of--is shameful.

I am not an advocate of MQA because, based on the information available, I cannot fully understand it. But I, at least, have tried. I am an advocate of considerate, respectful, well-informed debate. What comes from most--not all--MQA critics is anything but. It has more in common with extreme religious or political commentary than it does respectful dialog focused on evidence and fact. One side says, "I like the way it sounds." The other side says, "it's an evil plot to seduce children and Bob Stuart is the devil." In effect.

Jim Austin, Editor

adamdea's picture

The problem with all of this, Jim, is that it is in effect a regurgitation of MQA's marketing. It is obvious that it is impossible perfectly to apply the sampling theorem since it requires infinite sampling time and a filter of infinite length, so in various fields it has been necessary to approximate to greater of lesser extents.
In audio, these days it isn't really necessary to do so and the filters can approach perfect sinc function allowing for the window and be almost arbitrarily accurate Various audio manufacturers have used quite different filters for many years and no one needed to call them post Shannon.
The post Shannon thing is just repeating the MQA blurb: there is no body of concrete science suggesting the human hearing would be assisted by non Shannon sampling.There is no post Shannon development. Just saying "post-shannon developments" doesn't work.

And second, who is to say, even if that made up science existed, that they are in fact applying it?

I'm sorry, but your answer seems rather to prove the point.

Jim Austin's picture

And second, who is to say, even if that made up science existed, that they are in fact applying it?

It's amazing that you continue to deny the existence of a whole body of scientific/mathematical work that is so easy to find that anyone can do it. That makes you typical of the attitude I criticized in my previous post: You seem to know some sampling theory, so the only explanation is that you're just not honest.

Jim Austin, Editor

adamdea's picture

Where is this body of knowledge? I have asked some real experts and none are aware of it.
And the only fancy MQA filters which appear to exist are either leaky minimum phase filters or all-pass filters which introduce group delay. Where is the "post Shannon" science to support this?

Jim Austin's picture

Did no one ever teach you how to do a literature search? The core MQA documents include enough references to get you started. Follow the trail of references and citations. It's what scientists do.

I'll point out once again that this admission--that you've never even bothered to investigate the scholarship that (apparently) underlies MQA--badly compromises any criticisms. Do your research and then maybe we'll listen.

Jim Austin, Editor

adamdea's picture

Admission? I've read the MQA blah papers and the footnoted references and I do not believe that they warrant any implied claim that there exists a "post Shannon" body of thought which mandates the application of anything other than orthodox sampling and reconstruction to audio. I have asked others far more expert than me and neither do they. And not one person has managed to provide any even slightly satisfactory explanation.

There could be other applications where you only want certain information from a signal and don't want to reconstruct it perfectly. I seem to remember the dspguide having an example where an electrocardiogram just wanted to locate the leading edge of an impulse with a quite low sample rate. But none of this has anything to do with audio and I've seen no evidence of any audio signal needing to be reconstructed using anything other than a Shannon compliant filter.

If you think otherwise do explain. What is is post Shannon thought, what does it mandate and how is it applied? Many very thoughtful people will be very interested to hear.

Jim Austin's picture

What is is post Shannon thought, what does
it mandate and how is it applied?

I have already covered some of this. I feel no inclination go to over it again. Just read back through if you care to. It is clear from your comments here though that you're an example of the breed I characterized earlier, more interested in dogma than knowledge. We're done here.

Jim Austin, Editor

cgh's picture

I am going to go out on a limb here and say that I think we can all agree that everything in general info theoretic, signals, or EE after 2001 can be considered "post-Shannon" ;-)

prof's picture

Sampling is ubiquitous in a wide-variety of digital signal applications. I read the article you linked to, and it tells me absolutely nothing relevant to the business at hand.

The quick summary (for those who don't have access to IEEE publications) is that

1. The space, V, of band-limited functions is a closed subspace of the Hilbert space ℋ=L²(ℝ).
2. The "shifted sinc functions", φⱼ(x)=sinc(x-j), j∈ℤ, form an orthogonal basis for V.
3. Any function f(x)∈V can be expanded as f(x)=∑ⱼf(j)φⱼ(x).

From that point of view, "Modern Sampling Theory" consists of playing around with:

1. The choice of Hilbert space ℋ.
2. The choice of closed subspace, V⊂ℋ.
3. The choice of basis φⱼ for V.

If you want to claim that MQA is using some different ("Post-Shannon") sampling theory, maybe you could spell out what (ℋ, V, {φⱼ}) are.

Jim Austin's picture

As I said, that paper is only the starting point, and I don't remember what's in that article and what's in others that I (kinda, sorta) read. The key thing here is 3, the choice of the basis. MQA has not said exactly what functions they're using; one statement, and I think it was recently, suggested that it was like B-splines but wasn't exactly B-splines.

Other work (presumably it's other work; again, I didn't look back at this paper) removes the band-limitation condition and directly treats the error resulting from imperfect band-limitation. So, the error imposed by antialias filters is now 'on the record,' incorporated into the theory instead of perfect band-limiting being assumed. We can now ask, what's the total error that results when we "under-sample"--that is, when the signal goes beyond the Nyquist frequency?

"Classical" theory has not been overturned, as some here have accused me of claiming. New theory, though, enables new and more subtle perspectives and new opportunities. Exactly what MQA has done with this isn't clear, nor is the significance of what they have achieved, but I do not believe that Stuart and Craven are frauds, as many zealous critics (few of whom can understand the math even as well as I can--not well) have insisted. MQA simply hasn't disclosed exactly what they're doing, but much of what they have disclosed seems consistent with recent developments (post-Shannon) in sampling theory.

In a 2019 paper, AES Fellow Jamie Angus of the University of Salford in the UK concluded,

Modern sampling has reconstruction filters that have short impulse responses whereas traditional sampling requires infinitely long impulse responses for perfect reconstruction.

For real time use modern sampling has a lower reconstruction error for a given anti-alias filter impulse response length when compared to traditional sampling methods.

Another paper to consult is by P.L. Dragotti of University College London, and two coauthors, from 2007, in IEEE Transactions on Signal Processing, "Sampling Moments and Reconstructing Signals of Finite Rate of Innovation: Shannon meets Strang-Fix". This does not explain MQA of course--it preceded MQA by several years--but it does show why post-Shannon sampling theory is important for signal processing and why it could matter for digital audio.

I don't know what your field is, but if you're a mathematician and can read the papers with more insight than I have, I suggest you take a deep dive and report back. Maybe we'll all learn something.

Jim Austin, Editor

prof's picture

As I said, that paper is only the starting point, and I don't remember what's in that article and what's in others that I (kinda, sorta) read. The key thing here is 3, the choice of the basis.

No, I think the key here is: what is V, the choice of subspace?

There's a whole raging debate between you and other commenters here about whether certain perfectly band-limited signals are nonetheless "illegal" (i.e., not in V):

GoldenSound's test-file–embedded music tracks still do not resemble real music. The test file has very high levels of information at the very top of and above the audible range—right up to the 22.05kHz Nyquist frequency.

Conversely, MQA purports to successfully reconstruct some ultrasonic (non-band-limited) signals.

So, V is not the subspace of band-limited functions. It's some other subspace of "music". What is that subspace?

You can keep your basis {φⱼ} and your reconstruction algorithm (the coefficients cⱼ in f(x)=∑ⱼcⱼφⱼ(x) ) secret, while still allowing me (or @GoldenSound) to test your system by feeding it signals f(x)∈V and seeing whether they are reconstructed correctly.

The frustrating thing about MQA's response to @GoldenSound is their insistence that V itself (i.e., what constitutes a "legal" signal) is a trade secret.

Jim Austin's picture

There's a whole raging debate between you and other commenters here about whether certain perfectly band-limited signals are nonetheless "illegal" (i.e., not in V):

I guess I never thought of it exactly that way. It's true that many commentators see aliasing and assume ineptitude, although there are other "filters" that tolerate aliasing that have nothing to do with MQA.

If, as seems very likely (as I've said repeatedly), MQA is based on these post-Shannon approaches, then obviously they are not assuming strict band-limiting. The sampling is "sparse," in other words, although that phrasing switches things around, focusing on the sampling rate and not the bandwidth. That may be the most important post-Shannon development: Giving up insistence on strict band-limitation. Whatever the critics say, MQA seems based on that kind of theory--which, these papers show, is quite legitimate in the post-Shannon world. (One paper showed that you can achieve near-perfect results even with sparse sampling.

MQA admits error; it does not seek perfection. (This is my opinion, not something any MQA rep has said to me.) But remember that even with Shannon, error is imposed by the anti-aliasing filter; its just, as I've already said, "off the books." This, combined with the fact that, as what I quoted above shows, post-Shannon methods can achieve a better trade-off between time and frequency domains (via the choice of basis functions), and you've got a new approach that's worth trying.

Jim Austin, Editor

prof's picture

If, as seems very likely (as I've said repeatedly), MQA is based on these post-Shannon approaches, then obviously they are not assuming strict band-limiting.

That they're not band-limited was, I thought, the whole point of MQA.

One might have assumed, however, that the space V, of signals that MQA can successfully reconstruct, contained the space of band-limited signals as a proper subspace. What @GoldenSound's research showed is that that is not the case. Whatever V is, it does not contain the space of band-limited signals as a subspace.

Whether this "matters" is a different question. But, to carry the discussion forward in any intelligent fashion, we need some information that the MQA folks have refused to supply, namely what constitutes a "legal" signal as far as the MQA encoder is concerned (i.e., what V is).

99.9% of the "post-Shannon" literature is completely irrelevant to MQA. Which 0.1% might be relevant depends on the answer to that question.

Jim Austin's picture

Yes, I'm not sure it matters. One of the core ideas (perhaps the core idea) in post-Shannon sampling theory is to start by relaxing the requirement of perfect band-limitation and see where you can go with that. That doesn't, IMO, mean that any approach you take must deal perfectly with every band-limited signal. On the contrary, the whole idea is to explore the possibilities.

But at least we've now got a substantive conversation on the record establishing the reality and potential relevance of post-Shannon sampling theory--something that many critics have denied and will no doubt continue to deny. All I've ever really tried to do is to show that this is all reasonable and potentially viable and something that any legitimate critique should encompass.

It would be great if MQA disclosed more information about what it is and precisely how it works, but they're well within their rights to withhold it, and withholding it is not evidence of fraud.

To sum up: It's clear that, potentially, post-Shannon theory offers some modest advantages--advancements--and that MQA has implemented them in some form. To believe otherwise is to believe that legitimate, respectable scholars are engaged in fraud, which to me isn't credible. You don't have to like it or to believe that it's advantages outweigh the disadvantages of a closed technology. I just think that criticisms should be fair and well-informed. In the era of social media, that's obviously too much to expect. (Prof, I am not directing that at you.)

Jim Austin, Editor

prof's picture

Whether or not it matters, the fact that the space of band-limited signals is not a subspace of the space of "legal" signals in MQA is big news. That was totally unexpected, based on everything that had previously been written about MQA (including everything you'd previously written on the subject).

We have @GoldenSound to thank for that.

John Atkinson's picture
prof wrote:
Whether or not it matters, the fact that the space of band-limited signals is not a subspace of the space of "legal" signals in MQA is big news. That was totally unexpected, based on everything that had previously been written about MQA (including everything you'd previously written on the subject).

We have @GoldenSound to thank for that.

I wrote about this aspect of MQA in 2014; see fig.1 at, where I quote Bob Stuart as saying that "regardless of the type of music, this basic triangular shape in information space is characteristic, music and musical instruments having evolved to match our hearing sensitivity." This was also discussed in Bob Stuart's 2014 AES presentation.

To which I added: "That the musical information only occupies a fraction of the information space, generally a triangle-shaped area with the greatest dynamic range at low frequencies and the smallest dynamic range at ultrasonic frequencies. This is because of the fundamentally self-similar nature of music with respect to its spectral content."

GoldenSound didn't do his homework :-)

John Atkinson
Technical Editor, Stereophile

prof's picture

I wrote about this aspect of MQA in 2014; see fig.1 at, where I quote Bob Stuart as saying that "regardless of the type of music, this basic triangular shape in information space is characteristic, music and musical instruments having evolved to match our hearing sensitivity." ...

GoldenSound didn't do his homework :-)

I'm sorry. In that same piece, you wrote:

If you look again at fig.4, above, you can see the green line that defines the 16-bit noisefloor of the 16-bit PCM format. All the extra encapsulated information lies well below that limit. So if the file is truncated at the 16th bit, you have the equivalent of a normal baseband digital recording, sampled at 48kHz in this example. To a DAC or player that doesn't have MQA decoding, the MQA file will play as a normal 16-bit file.

which implies to me that any signal that can be encoded in an ordinary 16/48 PCM file can also be encoded in a 24/48 MQA file.

The former includes any properly band-limited signal (whether "music" or not), whereas according to @GoldenSound's research, the latter does not.

Perhaps that's not what you meant to imply by the above paragraph. If that's the case, then I am grateful for this belated clarification.

Jim Austin's picture

Prof, you're a breath of fresh air compared to many MQA critics here and elsewhere. But your agenda, and that you have one, could hardly be clearer. "Prof"--your screen name (and, I do not doubt, your profession), with all the clichés it implies, seems more and more to characterize your posts, which, instead of seeking insight and common ground, seem narrow, pointed, and rhetorical. You seek to score points, not to understand.

Academically, your point about spaces containing other spaces is well-taken. It seems reasonable as an academic argument. What you're missing is that reference is to just one paper in a wider field, and that the defining principle of post-Shannon sampling (or one important one anyway) is not what space contains what subspace in strict mathematical terms but simply relaxing the requirement of perfect band-limiting and then trying stuff and seeing how it works out. In some articles I've seen, they've come very close to perfectly reproducing a sampled signal even in "sparse" conditions.

Your point about the expanded space apparently not containing the space of perfectly band-limited functions would likely be important to a mathematician. Fine. (It would also be important to an anti-MQA partisan aiming to score points. These categories apparently overlap--although, as a former physicist myself, I have you pegged as a physicist, not a mathematician. Hopefully I'm not being overly generous. You could be a sociologist or an English lit prof for all I know. No criticism intended of those fine people, but their ability to understand sampling theory is likely to be limited.)

But I entered this conversation to make a single, simple point, or a handful of them: That "post-Shannon" sampling theory exists (in contrast to what certain anonymous posters, who have now conveniently disappeared, persistently claimed), that it's interesting, and that what MQA is doing is related to it. Closely related. Which is to say, for better or worse, for richer or poorer, like it or not, it's very likely real. Those who criticize it should (as you are doing, although not totally convincingly so far) engage with its substance (which few are capable of doing).

That point has been abundantly made, as I'm confident you'll acknowledge. That is all I ever aimed to achieve. My work is done here.

Jim Austin, Editor

prof's picture

I'm not sure why you think that casting aspersions on my screen name, my profession or my (alleged) motives is really the right direction to take this discussion.

Leaving that aside, let me try to clarify one point before this slides completely off the rails.

Your point about the expanded space apparently not containing the space of perfectly band-limited functions would likely be important to a mathematician.

The question of what is and is not a "legal" signal in MQA is important, not just as a mathematical curiosity, but as a practical matter.

To pick one famous example, is the cannon blast in the 1812 Overture a "legal" signal in MQA? It certainly doesn't adhere to the "triangular" spectral profile that JA1, quoting Bob Stuart (, says characterizes "real music".

I have no idea what the answer to that question is.

I learned some relevant facts about what signals MQA can (or, more precisely, cannot) encode correctly from @GoldenSound's experiments. (He also foolishly tested a bunch of things that obviously can't be encoded in MQA. But that kind of time-wasting is par-for-the-course in science.) I also (and I mean this sincerely) learned a bunch of relevant things from this discussion with you.

But, if we're going to go down the road of attacking people's motives and credentials, I think we should quit this discussion while we're ahead.

Jim Austin's picture

Here's what the first canon shot in the Kunzel/Telarc/Cincinnati 1812, in red, from the CD. Blue is a quiet passage from the same piece. Fuchsia is a random cello sonata (with piano). While the canon shot has a lot more high frequency energy than the other excerpts, it is still down 60dB at 20kHz relative to the low-frequency peak levels--and down by more compared to full-scale.

Jim Austin, Editor

John Atkinson's picture
prof wrote:
In that same piece, you wrote:
"If you look again at fig.4, above, you can see the green line that defines the 16-bit noisefloor of the 16-bit PCM format. All the extra encapsulated information lies well below that limit. So if the file is truncated at the 16th bit, you have the equivalent of a normal baseband digital recording, sampled at 48kHz in this example. To a DAC or player that doesn't have MQA decoding, the MQA file will play as a normal 16-bit file."

which implies to me that any signal that can be encoded in an ordinary 16/48 PCM file can also be encoded in a 24/48 MQA file.

Yes, if the spectrum of that 16/48 files conforms to the expected triangular shape in Shannon space. This was clear from what I wrote in 2014. And on the subject of the analog noisefloor, this is the spectrum of the noisefloor on one of my 24/88.2k choral recordings, with 0dB the peak of the musical spectrum:

I always try to make the recording in a quiet hall and spend time chasing down and eliminating sources of noise before the sessions start. My microphones have low self-noise and I use low-noise microphone preamplifiers from Millennia Media. Nevertheless, there is always noise present in the recording.

As you can see from this graph, the spectrum of the noise is not flat or "white," but is closer to pink. The peak level is close to -70dBFS in the low bass (thanks to distant traffic noise) and slopes down at around 24dB/decade to 1kHz and with a somewhat shallower slope in the treble.

An FFT-derived spectrum of dithered 16-bit silence with the same number of FFT bins would produce a flat spectrum with all the components lying around -130dBFS. As the music is always higher in level than the noise, you can see that the only part of the spectrum that would need to be encoded with >16 bits lies between 2kHz and 30kHz.

A 24-bit recording of music made in this church that peaks at 0dBFS therefore has spectral space available below the analog noisefloor in which to embed a buried data channel. This "steganography" is used by MQA to bury the ultrasonic spectral content, which can be encoded with very few bits, in the 24-bit baseband data. As the spectrum of the buried data channel can be made to be identical to the noisefloor of the recording, I haven't reduced the resolution of the music data and there is a negligible rise in the overall noisefloor. As I wrote, the sound without MQA decoding should thus be identical to that of the baseband content of the original file.

prof wrote:
The former includes any properly band-limited signal (whether "music" or not), whereas according to @GoldenSound's research, the latter does not.

Except that as I wrote and you appeared to miss, GoldenSound's content did not conform to the expected triangular spectrum.

John Atkinson
Technical Editor, Stereophile

prof's picture
prof wrote:The former includes any properly band-limited signal (whether "music" or not), whereas according to @GoldenSound's research, the latter does not.

Except that as I wrote and you appeared to miss, GoldenSound's content did not conform to the expected triangular spectrum.

I did not miss that. That wasn't the question.

16/48 PCM can encoded any properly band-limited signal, whether it conforms to that triangular spectrum or not. Not so (apparently) for 24/48 MQA. In other words, the latter is not just using the lower 8 bits for steganography (while the upper 16 bits function just as they do in conventional 16/48 PCM).

This was not clear from your 2014 piece. The impression left by your piece (and countless others, right up until @GoldenSound's work) was that the upper 16 bits of 24/48 MQA were functionally identical to conventional 16/48 PCM (i.e. could encode anything that 16/48 PCM could encode).

Thank you for clarifying.

As to the cannon blast in the 1812 Overture, its spectral content falls off with frequency, but much more slowly than "music" does. Does that slower falloff still satisfy Bob Stuart's "triangular" requirement?

prof's picture

Along the same lines, how about the gunshot at the end of 2 Chainz's "Forgiven" (on the album "Rap or Go to the League")? Does that satisfy Bob Stuart's "triangular" requirement?

epgawt's picture

it's been for some years I started to take music quality seriously. I've tried hard to improve the rig, and since digital became more affordable to not only buy but to understand I've learnt many things in the journey, like for example that dacs chips are not the only thing, I can now "see" my old cds with open source software and I can see what the big labels did with volume to fool us.

During these years I've seen many companies claiming better specs, SNR, THD, jitter.. some of them implementing new technical approaches for amplifying...

An interesting found was to finally admit that it is not a sin to use the EQ in my DAC, after all we all have a taste, and a little bit here and there is fine, that's how I like it, so what ?

What I have not seen in these years is what's happening with MQA. After carefully watching the GS videos, and some other staff, I already had it clear, MQA to me is suspect of trying to cheat the market, I'm not listening MQA, nor buying any hardware saying it supports MQA, I simply don't want to pay a bit of loyalties for something I don't want, I am not paying a bit to MQA, either directly or indirectly.

Anybody in 99% of cases can adjust treble or bass or loudness, or even do some EQ, to improve their experiences with our own tastes. These are honest adjustments, we know what we're doing, we know how to get more air, or bass or..., so I do not need anybody to do "I do not know what" saying "trust me this is better for your joy, but I can't tell you how I do it, because I want to earn money with that", I prefer to stay in the ground of professionals that say "I mastered this LP, and made it sound gorgeus", not the same thing.

I belong to the amount of people that looks forward to listen to music recorded the closest to honest original masterings, whatever the release, some of them sound better than others, but they all have on common one thing, the mastering process is "hand made", and I pay for that, they generate open music formats using processes widely understood and accepted by the community, where everyone fights improving their own implementations in all the phases of the path.

To me is pretty surprising that having access to all kinds of tools to adjust the music to our tastes, even including room conditions, people is accepting to pay for this obscure thing called MQA,

I am keeping an eye to GS, to see if anybody is capable to convince this guy of the contrary. Up to now what I have seen in his videos is light years ahead of any other stuff I've read, including the article that pushed me to write all this insane long post, this Stereophile one.


Andrei's picture

I have a fine LP rig, my digital system can play CD, SACD, DVDA, PCM (Red Book or Hi-Rez) and DSD. I need another format like I need a hole in the head.

John Werner's picture

Does anyone else who buys into more digital signal manipulation of a high-resolution file to be damning of digital period?

Oscar Hansen's picture

These frequencies don't occur in music, waah waah. This guy has clearly never heard of electronic music (or maybe he considers himself too good for that "filth".) I can find thousands of songs that don't follow these "rules". And this guy works at a magazine dedicated to music? Holy shit, allow me to laugh. I see why old media is dying with people like this around.

EDIT: I missed the sentence where he actually said this. Holy shit, of course I was right. That is hilarious. I am sure music exists out there that don't follow these archaic rules I just made up, but I am too good for that music. Get the fuck out of here, man. MQA is useless. Just like you and your "journalism".

EDIT2: This "magazine" is hilarious and employs children. They even edit others' comments when you trigger them enough (as they don't know what satire is, among a plethora of other things they don't understand.) If I type "These frequencies don't occur in music, waah waah" no reasonable person reads that as anybody actually saying literally that. But, sure, I'll make it "[These frequencies don't occur in music, waah waah]" and add an /s on top so even the children running this circus of a magazine understand.

John Atkinson's picture
Oscar Hansen wrote:
"These frequencies don't occur in music . . " This guy has clearly never heard of electronic music. . .

The electronic music I have examined doesn't have content - let alone high-level content - above 20kHz, so will be correctly encoded in an MQA file. To my surprise, I did find that distorted electric guitar has content above 20kHz - see - but this is still sufficiently low in level to conform to the spectrum anticipated by the MQA encoder.

John Atkinson
Technical Editor, Stereophile

cgh's picture

Thanks for the reference, I enjoyed your article. I've done many similar measurements in my forays building classical guitars and violins but I have never had the time or inclination to bridge the two worlds. I remember teaching an undergrad course - Mathematical Methods in the Physical Sciences, I think mainly using Boas for those that remember - and I would attempt to simplify and describe the fourier transform in t-space as "smearing" the info in t^-1-space and vice versa for its inverse. It would seem that all the non-linearity and complex relationships I've seen in guitar and violin measurement would be lost to some extent if sampling theorem applied, or any new-fangled recipe. It's not just where it lives on the axis but what it constructively or destructively interacts with that matters. While a significantly small aspect of the time domain is the onset transients chopping them have significant impacts. While the primary modes up to dipole of a guitar are almost 90% of the sound intensity the remaining modes are the difference between a really boring and poorly made flamenco guitar and a concert classical. As you said in your 2000 article, there's more to this all.

Jim Austin's picture

Don't use quotation marks unless it's a quote. Because quotation marks suggest that it's something someone actually wrote, which isn't true in this case: Nobody wrote what you put in quotes. If you continue to do it, I'll delete your posts.

Assuming your fake quotes reference the essay above, here's what I did write:

GoldenSound's test-file–embedded music tracks still do not resemble real music. The test file has very high levels of information at the very top of and above the audible range—right up to the 22.05kHz Nyquist frequency.

I went on to give specific examples (see above). An example I gave was white noise at –6dBFS at or just below 22.05kHz. If you can show me a single track, electronic or otherwise, that's –6dBFS at or just below -22.05kHz, I'll admit I was wrong. It has to be real, published music, not something you or someone else just threw together. Show us the data, and tell us where it comes from. If you can't find a track that meets those specifications, you'll admit that you're wrong--deal?

Jim Austin, Editor

Anton's picture

I am trying to think back to any of JA1's measurements of digital playback devices and can't think of any that used music. They are all based on test tones and non-musical parameters.

As I think about it more, I can't even recall any measurements of any gear in Stereophile that use music as a basis of measurement.

Not mean in a pejorative way. It seems no measurers actually use music. Why be mad at Golden Sound?

Hell, is he really doing anything different that most of those old vinyl 'track records' from Shure and other manufacturers used to sell?

I don't feed square waves into my amp, but JA1 loves to measure with them.


Related question: does Stereophile regard MQA as a high fidelity (high end) means of reproduction, or as an enhancer for compressed streaming when true Hi Rez is not available?

John Atkinson's picture
Anton wrote:
I am trying to think back to any of JA1's measurements of digital playback devices and can't think of any that used music. They are all based on test tones and non-musical parameters.

That's because I am using the signals in a diagnostic manner, forcing the device under test to reveal its inner workings. For example, when I test D/A processors, I feed them an "illegal" signal, "digital black" data with a single high sample. This will never be encountered in music, of course, but the processor's analog output with this signal identifies the type(s) of reconstruction filter the designer has implemented. And those filters do sound different with music.

Anton wrote:
I don't feed square waves into my amp, but JA1 loves to measure with them.

The squarewaves reveal problems with amplifier stability that may well prove problematic with music.

Anton wrote:
Not mean in a pejorative way. It seems no measurers actually use music. Why be mad at Golden Sound?

It's not that I (or Jim) are "mad at GoldenSound." It's just that he didn't understand the nature of what he was trying to measure and thus drew erroneous conclusions. Something I keep in mind every time I go into the test lab is what the late Raymond Cooke, founder of KEF, told me back in the 1970s: "There's nothing more dangerous in audio than a fool with a voltmeter."

Despite many years of measuring I still try hard not to be Raymond's dangerous fool. Others should do likewise.

John Atkinson
Technical Editor, Stereophile

scottsol's picture

In a recent interview with Bill Leebens , Bob Stuart confirms that MQA cannot be technically lossless and does use B-spline sampling for de-blurring.

atmfrank's picture

Sitting here reading everyone's comments about an utterly useless and entirely non-enriching controversy. Honestly who cares?

Sitting here, with a 30yr old Sennheiser HD600 strapped on, plugged into a LG V40 phone (w/ ES9038 DAC) running USB Pro and streaming a NAS stored version of Robert Plant's MQA re-mastered "Digging Deep". It sounds phenomenal!! Do I care about if Robert cares that his recordings may not sound the way he listened to in the mastering room? NO. Jimmy Page might care...

Do I care if MQA stays or goes away like so many other formats from the past, like HDCD for example? NO, I don't care. It is very predictable that it will fade. Good riddance...but in the meantime Robert Plant sounds simply AWESOME in MQA

So much wasted energy. Who cares, really.

R Johnston's picture

GoldenSound test files were intended to show that MQA was not a lossless encoding system. They did just that.
A true lossless system produces the exact same output file when decoded regardless of the input file. MQA does not. So MQA is not lossless system. Now the important question is if MQA's "modification" of the audio file is always a improvement or not. Again to be fair to GoldenSound MQA seemed to respond to that question with "trust us". That is also red flag.

mpb020479's picture

I guess keeping a format proprietary and requiring hardware companies to provide engineering insights, all while never fully demonstrating the benefit of the product may not have been the keen business decision that it seemed at the time.