Mythbusting Tidal’s MQA Format – How Does It Measure Up?

MQA is an audio format that claims to use a unique “origami” algorithm, promising better quality and more musicality than other formats. At times, it’s been claimed to be a lossless format in so many words, and lauded by the streaming services that use it as the ultimate format for high-fidelity music. With the format being closed source and encoders not publicly available, these claims are hard to test. However, [GoldenSound] wasn’t born yesterday, and set out to test MQA by hook or by crook. The results were concerning. (Video, embedded below.)

To actually put the format through its paces, the only easy way available was to publish music to the Tidal streaming service, which uses the format. [GoldenSound] went this route, attempting to get some test files published. This hit a brick wall when the publishing company reported that the MQA software “would not encode the files”. The workaround? [GoldenSound] simply cut some audio test content into the middle of an acoustic track and resubmitted the files, where they were accepted without further complaint.

Testing with the content pulled from Tidal, [GoldenSound] found concerning evidence that the claims made around MQA don’t stack up. Significant amounts of added noise are often found in the MQA-processed files, and files served from Tidal are clearly not lossless. Additionally, MQA’s “blue light” authentication system, designed to guarantee to listeners that they’re listening to a identical-to-studio release, is demonstrated to be misleading at best, if not entirely fake.

Upon writing to MQA to get a response to his findings, [GoldenSound]’s test files were quickly stripped from Tidal. The company eventually disputed some of the findings, which is discussed in the video. The general upshot is that without open, transparent tools being made publicly available to analyse the format’s performance, it’s impossible to verify the company’s claims.

We’ve had fun looking at audio formats before, from the history of MP3 to musing on digital audio at truly ridiculous sample rates.

[Thanks to Chris for the tip!]

105 thoughts on “Mythbusting Tidal’s MQA Format – How Does It Measure Up?

  1. Audio encoding seems to be a lot like cryptography in the respect that every so often, someone comes on to the scene, touting a “breakthrough technology” that’s orders of magnitude better than the current standard…and then, someone takes the time to do some analysis, and demonstrates that some of the claims made by the newcomer are…optimistic.

    Very much worth watching and the evidence shown is persuasive.

      1. It was never universally applicable to start with.

        For most people it worked well enough for casual listening and anyone who claimed to be able to hear a difference was labelled a ‘Purist’ and/or other far less polite things…

    1. Audio stuff at the digital level always baffles me because, um… it’s kindof a pathetically low information rate – which is a point made in that video. 24-bit, 192k sampling is just… tiny. I mean, c’mon, we shuffle data vastly larger than that all the time, and I can buy chips that can transfer that data point to point on the crappiest cable ever with no effort. Even time transfer point to point (jitter) is essentially a solved problem at that level.

      The difficult part of audio is *all* analog. It makes *zero* sense to try to save space/bandwidth after you’ve gone through all the work to actually get something with 140+ dB dynamic range.

      1. It’s the compression thing…which is not unlike cryptographic systems (though the goal is different). There’s a lot of prestige to be gained by building a better compression scheme. What people often forget, is that compression is only half of the problem, decompression needs to restore the information with minimal distortion.

        Look up “Sloot Coding System” for a humorous tale of a “breakthrough compression technology”

        1. Sure, but the hilarious thing about this is that if I’m an audiophile equipment manufacturer or a recording studio I *wouldn’t ever want* a lossy compression scheme (which obviously is what a system where there’s any compression/decompression distortion is) because you’d never be able to show that it’s actually *better*. There aren’t any fancy pointless tests you can do, for instance, to justify higher-end, expensive equipment.

          I mean, if I’m trying to sell people stuff I’d start with a container that contains full multichannel studio recordings + mixing/selection information in some cryptographic container plus some dedicated ASIC/FPGA “reconstruction processor” outputting to DACs. Yeah, the end result would be exactly the same, but *there* would be justification for some magic “blue light.” Plus now your filesize is like, 20 times larger and you can charge more for it or something.

          1. Going back to the audio domain, it begs the question as to why any audiophile would put up with the decompression ‘algorithm’ applied to vinyl recordings.

            I am of course referring to the RIAA equaliser. What that does to group delay is anyone’s guess!

      2. Yep. With multiple terabyte-size HDDs now being standard, there is no reason anymore to save anything in a lower quality. 24 bit, 192k, flac’d.
        Sometimes i downsample to 16/44/flac, for on my phone or whatever where i still have tens of gigabytes of space.
        Maybe it’s different for streaming, but if i care so much about good quality i just download the thing.

          1. That’s a common misconception.

            There’s very little difference in the amount of data transferred between streaming and downloading audio which uses the same encoding method and quality settings, e.g. bitrate.

        1. The reason is well… reason. Even if we assume there are people who can hear difference between 44,1/16 and 192/24 that requires really high quality equpiment plus well prepared environment. Our home audio and mobile phones are nowhere near to that. Waste of space and money.

          1. I have a modest system by audiophile standards and hear and feel a huge difference between 16/44 And 24/192. 16/44 can sound great, but 24/192 can be phenomenal

          2. Nope. No way.

            If all you do is just *play* the audio, sure. Fine. But it’s similar to the reason why you need better color resolution than RGB888 gives you in games, even though it’s essentially fine for photos: when you try to algorithmically screw with stuff, you get obvious color banding. Similarly, if you *manipulate* that audio, say, rebalance it, alter it, mix it with other things, a higher-fidelity master would clearly be audible.

            I mean, the obvious example would be just (digitally) turning up the volume on an extremely faint portion of a recording, possibly to compensate for changes in one’s hearing over the years.

      1. How so, pls provide evidence of how it’s flawed otherwise do you seriously expect others to take you, a stranger online, at your word alone (especially after ironically telling others to not believe what they read online). And when I say evidence I mean measurements, calculations, data that contradicts what you are claiming is false, not just hand waving, angry diatribes and emotional arguments.

      2. No I did not say that it was flawed. I said that it would be insufficient to answer all questions/potential situations.
        MQA is a lossy codec. The point of the video was to prove that. The MQA encoder would NOT have fallen over in the way it did had it been lossless.

        The point is that MQA (and Tidal) are advertising either lossless or “better than lossless” content, and that is NOT what the user is getting. It is false advertising.

        How good/bad MQA is as a lossy codec isn’t really the point, because I wouldn’t want to have to use a lossy codec anyway, and seeing a lossy codec advertised as a lossless one is pretty bad.

      3. He changed the file first in area not in the triangle. That was his first blatant effort to break the system, when it is claimed they focus on the heard triangle. Then he claims he is able to just download the MQA file. That if false, or we’d have millions of music pirates grabbing from Tidal . It’s amazing to me how many fools just believed that golden one clown even while he was showing false or tanted methods, then to find out he changed the file? FTG!

    2. If you read MQA response they pretty much show the tester doesn’t know what hecis doing. And if you google his name he is on multiple sites ranting about MQA. The problem with his analysis is what we care about is the sound. Charts don’t tell you about the sound. And MQA sounds good to me. Sorry but this article is crap.

      1. So in a nutshell, discredit someone based on personal preference and smearing their name as opposed to looking at their data and charts which are inherently unbiased (much more so than a corporate entity who financially gains from spreading false info). But no, let’s crap on someone who provides actual data to instead be swayed by bias and emotion. If all you look for in music is that it sounds good then by all means good on you, but for them to make unsubstantiated marketing claims that can be dis-proven and attempt to stifle any attempt to do so is not just dubious, it’s actually illegal. Sorry but your opinion (and it is an opinion without any evidence beyond vaguely saying some stranger online doesn’t know what they are doing, as sort of “pretty much” “proven” by a questionable entity and that an article is crap) is simply crap.

        1. “ If all you look for in music is that it sounds good then by all means good on you”

          What on Earth else are you looking for in your music?!
          If it sounds good, listen to it. If it doesn’t, don’t; perhaps it’s over-compressed, or perhaps it was just rap music to begin with. Either way, who’s going to listen to it if it doesn’t sound good other than a pretentious audiophile?

          1. As a user that’s all that matters, that it sounds good to them. But as a company making marketing claims, they are held to the truth of their claims or they should be fined for false advertising. And any attempt to hide or prevent verification is unacceptable.

          2. I want my music in a lossless format for archival purposes. If I want to put a lot of music on a portable player without a lot of space, I can compress it as ogg at q5 or q6, and get around 5x as much music into the same space as I can with flac. Then if I want to play it on my friend’s player in his car that only works with mp3, I have a lossless source, and I can encode it for that.

            If I’m listening at home with my good quality headphones, I can hear it as clearly as possible, by listening to the original lossless material.

            a good quality ogg or mp3 file will generally sound good enough most of the time, but it is objectively, and subjectively worse than lossless, and something that has been transcoded from one lossless format to another is almost always obvious on any reasonably decent headphones. (decent, not necessarily expensive)

            I often hear tracks on spotify with obvious weird distortions, which 9 times out of 10, aren’t present when I listen to the CD, or flac file I have at home.

            high bit depths and sampling rates, OTOH, are a total red herring.
            I doubt anyone could tell the difference between a properly mastered 44/16 lossless file, and a 192/24 of the same material. (double blind ABX between 192/24, and same file properly resampled to 44/16)

            There may be many cases where the “high def” versions are re-mastered with better levels, less dynamic compression/distortion, etc. But in these cases, when it “sounds better” it sounds better because of the better mastering, and not the higher resolution.

            The extra sampling rate/bit depth does nothing but waste space.

          3. The company is not making false claims. The claims are being misinterpreted. Yes, there is added noise. That’s the secret sauce. The noise is shaped. This is nothing new. What is new is that the format has a fingerprint which can be used to authenticate the master recording that was used. Also new is the unique folding system that allows the format to degrade to lower resolution without modification and to play back as standard 16 bit 44.1K PWM. For the tester, if all they are interested in is what the file looks like, they are going to say the test fails. It is not lossless and no one has ever seriously claimed that. It competes with lossless. They’ve shown that you don’t need huge bandwidth to recreate the sound. The huge bandwidth of high res files is encoding mostly noise actually, and mostly inaudible. MQA is encoding noise that makes sense, and far less of it. Noise is not always bad. ADCs often add dither noise to the final result so that at lower volumes you can’t hear the bits popping on and off. As others have said, all that’s important is how it sounds. Admittedly that is hard to test, because it’s very subjective. Much like the claims of audiophile twats who insist that vinyl sounds better than digital.

        2. You sound like golden ones lover. He tried to break it, he lied, and does not show a way to actually get the file back for a proper test. Questioning something is one thing, ranting everywhere and trying to break something is all together different. Clean off your nose!

      2. I have to say after over a decade of being on this site, I’ve never seen this degree of hand waving for a product. “Not a hack,” par for the course. “He’s wrong because I said so,” very rare. That crap may work over at Reddit or the YouTube comments but here, show your work. You’re dealing with people that with timers, transistors, capacitors, and free time can reproduce the ADC/DAC needed to record and play back high quality sound with plywood or anything else from the low bandwidth analog world. Are there flaws in his work? Only from a sample size for the audio and hardware, but that doesn’t invalidate his findings. It’s not like comparing two waveforms is mathematically difficult and he’s not comparing some low/high pass filters, it’s a straight input vs output from the magical black box.

        Show some respect.

      3. I think I found the same guy commenting here ( https://hackaday.com/Anno1543/De-revolutionibus-orbium-coelestium ):

        “If you read the Vatican response, they pretty much show Copernicus doesn’t know what he is doing… the problem with his analysis is what we care about is subjective perception. Science-y maths, equations and charts, don’t tell you about subjective perception. And it is obvious to me that the sun goes around the Earth. Copernicus’ article is crap.”

      4. Er, how do you think Bob Stuart invented mqa, or in fact how did anyone invent a working transistor. They probably looked at charts. The earth is not flat, and that’s a fact.

      5. Why someone bothers to listen to streamed sheit when listening to music is beyond me. If I listen to music seriously i use flac and headphones. When playing sounds to drown out background noise the compression carries little weight for me since im only noiseblocking anyways. The only thing that bothers me is joint stereo on mp3.

        1. “Why someone bothers to listen to streamed sheit when listening to music is beyond me.”
          Well, I actually found a lot of music I never would have found “only” listening to flac and CD …
          Streaming is wonderful to learn new artists, new albums, new styles.

        2. How is sending your DAC streamed music to decode, in a lossless format (such as the streamed FLAC files Qobuz offers), any different than feeding it to your DAC via a file on a hard drive?

          This is the whole ruse of MQA to begin with, which became an obsolete endeavor before it was ever widely adopted—although its creators were savvy enough to grab up licensing contracts with labels super fast before they discovered its obsolescence. With gigaspeed internet and massive solid state drives becoming widely adopted over the past couple of years, those truly invested in lossless audio with capable wired headphones and a dedicated DAC don’t need to save space or bandwidth anymore.

          I’m afraid streaming is here to stay, as are cloud services, And given the failure rate and degradability of physical media I’m not sure what you’re upset about.

          1. It’s about much more than saving file download bandwidth. It’s a lot more about impulse response of the DAC and reducing ringing. It’s not necessary to capture ultra high sample rate and ultra high bit depth to get high fidelity and dynamic range. Most of the high frequency content at, say, 192 KHz is noise. Pure junk. The DAC has to work extra hard to restore that and for no good reason. MQA is designed and optimized to reduce side bands, which in turn will reduce temporal blurring, giving a more accurate reproduction. Mythbusing, eh? Mythmongering really.

          2. To add to my previous comment, impulse response of the ADC at time of capture is also important, and WQA optimizes that as well. With the right settings, as verified by a mastering engineer in the studio, the results are supposed to be indistinguishable from the original. I say supposed, because I’ve never even heard an MQA. I just trust the word of world renowned mastering engineers like Bob Ludwig. There is work involved. You’re not going to get it right just with default settings.

    3. Musician and engineer here (always a risky combination admittedly). Look… no one cares about any of this. 99% of can not hear the difference between an mp3 and a wav. Trust me, they can’t. Music is a social good, not a quantum mechanics lab. I am fine with innovation, especially when it comes to music creation. But the format changes have had some consistent attributes no one here seems to give a shit about: they lower the take for creators, while they lower costs and increase control for publishers. At the same time, it has become very difficult to exist as a creator without a large amount of PR. Organic scene building is almost impossible when everyone’s eyes and ears are on their phones. Out here where people MAKE music, there are big problems which the format changes drive. MQA is bullshit, sure, and it is bullshit in a context of exploitation, algorithmic wage theft, and concentration of capital. Maybe go outside more?

      1. You assume it’s bullshit. Have you ever tried to hear the difference between a high res or analog master and an MQA? Bob Ludwig says he can’t. That’s good enough for me. I’ve never even heard an MQA but I’m excited about it.

        1. I have an MQA decoding device on a high end system. I can’t hear the difference between an MQA file and a lossless FLAC.

          This is why MQA is bullshit. There are licensing costs associated with MQA that are not present with FLAC. Bandwidth and storage are readily available to the vast majority of users. There is simply no reason for MQA to exist. We don’t need file compression and that is all that MQA offers over a lossless file. There is simply no need for any of us to have to pay to have its proprietary algorithms in our music files or equipment.

          It is totally straight forward and uncomplicated. Getting caught up in arguments about noise and origami is getting pulled into a pointless argument about pointless things.

          Cancel Tidal and refuse to support closed source proprietary codecs. Open source is best for consumers and artists.

          1. I think now I’ve heard it all. Your complaint is that you can’t hear the difference? It’s good that you can’t hear the difference between lossless and MQA. What difference are you expecting? Lossless should sound better or worse? What quality is the lossless? FLAC can encode better than CD quality. It’s not about bandwidth and storage size. It’s about optimizing the ADC and DAC converters so that the impulse response has minimal side-bands and reduces ringing. Ringing introduces temporal blurring. High sample rate digital encoding encodes too much information that is just noise and then the DAC has to work to decode that unnecessary information to the detriment of the information we can actually hear. Good quality ADCs and DACs are expensive. These can handle high rates with minimal smearing. If your system costs $10,000, then maybe there is no difference. Average consumers don’t have access to such equipment, so the DAC in their playback equipment will not cope as well. I don’t give a squat about Tidal or any suspect business motivations. I think the technology is very interesting and it sounds promising.

          2. If you look at the CO2 footprint of music streaming, then compression is VERY relevant.

            It’s not about storage space on your system, but about storage space on the cloud and total network bandwidth.

            Admittedly, the whole audiophile streaming segment will most likely pale compared to the footprint of something like TikTok, but it is still relevant.

            The OSS argument is a good one, though. But for streaming, I’d still go for OGG over FLAC unless listening on a high-end system.

      2. Oh lord. If being a musician and engineer is a risky combination, then why bother invoking it to shore up your claims? I’m a musician myself, but even if I weren’t, you don’t have to be able to lay an egg to know if it tastes good.

        Anyways, I’m not sure what equipment you’re using, but to converse in jargon that surely a musician/engineer should be able to appreciate, I can assure you with utmost confidence in the following:

        1. With my iPhone speakers and maybe the shower running, I probably could not tell the difference between an mp3 file and a lossless track.

        2. With my ES9038-based Topping D90se DAC, my iFi-Audio Pro iCAN Signature amp with NOS Bendiz 6385 matched valves, and my beryllium driver Focal Utopia headphones, I can definitely appreciate a stark difference between 320 mp3 files and lossless 192/24.

        3. With this setup I can even discern differences in soundstage and detail retrieval between a Red Book CD at 44.1/16 and a lossless FLAC file at 192/24.

        4. Finally (and perhaps most importantly), I can without question tell the difference between the junk masters low-end streaming services like Spotify utilize to save cost on their MP3 catalog, and the high-provenance masters premium services like Qobuz use for theirs.

        Seriously, what motivates otherwise reasonable folks such as yourself to make Luddite claims like this—particularly given that you obviously not only have a passion for music, but a vested interest in the technology that reproduces it? You’re not the first I’ve heard make these claims—my only explanation for it is fashionable contrariness, officious sanctimony, or just plain controversy for its own sake. C’mon, surely you have more respect for the human ear than to accuse it of such feebleness?

    4. Ah, no. Cryptography relies on transparency, among other factors. See Kerckoff’s Principle. Literally, assume your adversary knows what you know about the underlying security (eg cipher). The security is in the “secret”, an index of functions known as a “key”. It appears that the MQA format is proprietary for the same reasons Verance is proprietary. But, the back story is more interesting if the references are examined.

    1. My favorite audiophile nonsense was a device that was intended to play four identical ordinary CDs at the same time, apparently because you’d only lose 1/4 as much audio signal that way.

      Not sure what I2BP is though – any more info on that one?

  2. The fact they rejected the original test files, and then delete the later test files tells all I need about how honest their claims are.
    If they had problem with the tests, they could have kept the summit-ed files and post what they thought was wrong in the tests, instead they try to hide the fact the tests even exist.

    1. The AK4191 is a digital modulator meant to interface to a separate ADC. The 64bits refers to internal digital filtering signal path and not the bit depth of the ADC.

      1. The AKM factory also burned down, so you won’t be buying that in quantity anytime soon. They make great DACs and it is a shame. Research ESS hump. There is now one main player in the market with design problems built in. AKM was the better choice.

  3. MQA has been the the reason for me to cancel my Tidal subscribtion after many years. After a bit of looking around on the internet, it seems I am not the only one. It is a premium price for lossless, while actually it isn’t lossle
    Without ‘MQA equipment’ you definitely get less than standard CD quality, since they “mis”use some of the bits for MQA ‘noise’ .

    1. I agree with you 100%. I am about ready to cancel my tidal as well. Between the lies about MQA compatibility and the never ending lies about atmos native content, I feel like I’ve been paying for years waiting for the payoff that’s been promised and it just never comes. There’s questionable sound quality, horribly questionable support for true multi channel atmos and what hardware will support it, and seemingly no end in sight. It makes me want to sue over the years of paying for a service that has yet to deliver on all the things promised in the bill of sale.

    1. This is mostly a news aggregation site, no? I don’t see the problem. If you want new and original content, maybe you could make some and submit it to the tip line. Just don’t post it anywhere else or people like you will complain.

  4. Thank you for objective view on MQA, it is a bit of black box and you do get the feeling that the quality is better. My observation is some are better than others, but again i think it comes back to if the original is from a master studio.

    I would like to get your recommendation on which one is better out of the streaming services, or maybe one should wait for Spotify HIFI to come out. From a hardware i got the KEF LS50 Wireless 2 and the sound quality on these are amazing. Yet i want to match it with some service which is high quality. Its shame the Tidal does not provide more information about MQA and equally force MQA to be more transparent.

    Thanx

    1. I think audio CDs are most cost effectively streamed to the listener by postal services. Couriers are often more expensive, but may not be as lossy.

      A pair of wires, combined with a CD player, is likely to be less lossy than anything streaming to a KEF LS50 Wireless 2 and can be plugged into the Aux jack on the back of the speaker. You’ll also get decent dynamic range.

        1. What’s your point? When it comes to sound, nothing is truly lossless. A microphone distorts the sound of vocals. There is no truly lossless means of capturing audio, which is a mechanical phenomenon, into electrical energy. Digital signals can be preserved on a medium which will not degrade. CDs are lossless from the perspective of a 44.1 KHz 16 bit PCM master. Analog media gets a little bit worse every time it’s played back. Vinyl is certainly lossy.

    1. Unfortunatelly real life shows that marketing crap works perfectly and science is helpless against it. Look at the comments here on HAD – some people still defend the crap and even expect Golden Sound to correct his “mistakes”. I can only imagine what happens on audio voodoo sites.

  5. Testing? Objective, reproducible measurements?

    Waste of time! That’s going about it all wrong.

    No format will be truly audiophile quality until someone makes a vacuum tube based computer fast enough to encode/decode it!

  6. This really isn’t that hard to prove… Well, its quite trivial really.

    The moment audiophiles give up trusting their ears, they can start using algorithmic methods to prove lossy/lossless codecs.

    [Here is a Gnuradio flow](https://imgur.com/a/DMLR3ey) comparing the same song, one that is “high rate MP3” and the next is “64KB mp3”

    And if you don’t want to download and set up Gnuradio, here’s a [Peertube video] https://videos.trom.tf/videos/watch/457b99a5-e3ee-4cf7-951c-e250b6840724 of me showing how to do it.

    The expectation for a lossless encoding would be NO outputted data, since their waveforms would have annihilated each other. Any deviations are emitted as waveflow. And if you listen to it acoustically, is effectively white noise (expected with XOR). Note that even tiny deviations would show the codec is lossy – as in data is being thrown away. We don’t care the quantity or quality – a non-zero number of deviations is enough to show us their claim is blatantly false.

  7. Back in the day, I used oscilloscope to compare output from CD, MiniDisc and mp3.
    mp3 is crap below 8MB with high got cut off. Anyone with half a brain will hear the different.
    MD is pretty close to CD with some smoothing.

    1. MD (or whatever AAC it was) was surprisingly good.

      MP3s of that era were audibly crappy b/c of low bitrate. (Cymbals are a dead giveaway, if anyone puts you in a low-bitrate MP3 listening test.) But at 192k, they’re not so bad either, really.

      I store all my audio in FLAC. It’s really hard to beat bit-for-bit identical with the source, and 30% compression.

    2. ATRAC1 was pretty lackluster, but by the end of the MD’s effective life upgrades in the decoders made it sound pretty good.

      All of the ATRAC codecs were designed for RAM and CPU efficiency while being better quality per bitrate than ADPCM

      1. ATRAC3plus was pretty damn amazing; you could get audio roughly the same quality as commercial MP3 files (they were all around 192kbps or so at the time), but the file size would end up being anywhere between 1/2 and 1/4 the size.

  8. At least it’s not like that old image file compression scam that “compressed” images to incredibly tiny sizes, while shuffling most of the data off to a hidden file.

  9. When you suffer from a slight form of tinnitus anyway, fuck it.

    For me, digital audio is engineered to an end. It is a tiny amount of data. done. it is good. no loss, great. Don’t start about it again. thanks.

    Great tho someone actually tries to debunk false claims.

    1. Exactly. Most people have no idea about M-S mastering (Hackaday great article idea BTW) or compression or Eqing. Mastering engineers are solely there to make it sound good and sometimes that means ADDing harmonics or distortion.

  10. Are there any software-only decoders for this MQA thing or do you need a hardware decoder even on a Windows PC?
    Maybe someone just needs to do what has been done to many other proprietary media formats over the years and reverse engineer the thing…

    1. Full MQA “benefit” is a result of software unfold followed by proper rendering by an MQA licensed DAC. MQA does not license any DAC that has a post-processing digital-out. This should tell you everything you need to know.

  11. Did you confront What*HiFi magazine with your findings? They have repeatedly considered Tidal the best sounding streaming service and seem to push MQA masters as the pinacle of audiophile enjoyment. I do not agree (at best, it’s just incredibly loud) and prefer lossless flac from Qobuz or HRA streaming (Germany). Thank you for adressing MQA’s flaws this thouroughly.

  12. I see the author and many commentators missing the point. If your test is to look at the resultant audio file and compare it to the original waveform, the test is going to fail here. No one has ever seriously claimed that MQA is lossless. It competes with lossless formats. Yes, MQA adds noise. For that matter, dither adds noise. Noise is not automatically bad. In high bandwidth high res file formats like 24 bit 192 KHz PCM and their lossless compressed derivatives, most of what is being encoded is noise and mostly inaudible. That’s not even the highest bandwidth available. This really is the emperor’s new clothes. You cannot hear the difference. You cannot even really hear the difference between 16 bit and 24 bit audio at the same sampling rate unless you have a very high dynamic range and it’s played back really loud. The difference you hear is the noise floor.

    Back to MQA: it adds noise. The noise is shaped. This is nothing new. What is new is the unique fingerprinting that allows authentication of the original master. Also new is the folding structure that allows the same file to gracefully degrade and be played back as standard 16 bit PCM without modification. That may not sound quite as good as a CD master, but it’s not intended to. The claims are not inaccurate, rather they have been misinterpreted. What sounds good or right is difficult to test because it’s very subjective, much like the unsubstantiated claims of audiophile fanatics who swear up and down that vinyl sounds better than digital. I don’t claim to have golden ears, but I am an audio engineer with a degree in electrical engineering which was focused on digital signal processing. MQA has world renown mastering engineers on board. Who am I, indeed who is anyone, to argue with Bob Ludwig? When setup correctly, MQA is designed to achieve maximum clarity by reducing phase smearing to an absolute minimum. MQA has shown that you don’t need huge bandwidth to capture the essence of a recording. Their response to this test is intelligent and intelligible. I provide a link here:

    https://pastebin.com/2YfT6vfZ

      1. I think you may be right, in the sense that the encoded high res information appears to be scrambled in such a way that it mimics pseudo-noise in playback without the decoder. But that’s not the same as a watermark. A digital watermark may have many uses, but I’ve never heard of one which offers an enhancement to users of the file. This is not some sort of hard to remove stamp to el or track the file. It’s an additional layer to be unlocked. I think it’s a real innovation that this container degrades gracefully into a format that still gives very good results.

        1. A digital watermark it is! It’s described generally as being part of steganography, data hiding, information hiding, obfuscation.

          Craven and Gerzon (his late partner in acoustics) describe a “buried data technique” in the 1990s. They use the LSBs, hardly “inband” if subject to perceptual compression (eg MP3, AC-1, AAC, etc). But it does introduce more detail on just what is being done in the file.

          SDMI, 4C, STEP2000, all had consistent definitions of the types of digital watermarks sought – mostly “copy control information” – but, forensic watermarking has a greater value and is associated with cryptography. Especially, authentication functions & keys (index’s of functions).

          The notion slow degradation of signals is the basis for VHS protection schemes offered by Macrovision.

          Another time …

          1. Hmm. I guess I’ve been holding onto an outdated notion of the purpose of a digital watermark. It IS data hiding, but not for the usual reasons. A watermark can add value. Good information.

  13. I have done null tests on mqa vs hi rez files & the best I could achieve was a 70db null. I am capable of complete nulls of 2 separate recordings recorded from same audio source through analog to digital conversion with same equipment. It takes me about an hour to to achieve such nulls as everything has to time aligned down to 1/4 of a sample width in time. This is not the same as taking two identical digital recording & nulling them as no real work is needed. These were recorded from analog input.

    There were differences in high frequencies that were audible even with the deepest null achieved when nulling MQA to a high resolution recording of the same but a complete null was achievable with 2 analog recordings of either the MQA or high resolution recordings made in the analog domain when nuĺled for example MQA to MQA & visa versa.

    This means MQA IS NOT IDENTICAL TO HIGH RESOLUTION files. However the null was sufficiently deep that it would be nearly impossible to hear the difference. Differences were at very high frequencies.

    These were conducted using the Audacity program which allows you to null two files together with all the tools needed to edit file in oder to align them down to 1/4 sample width.

    1. No one ever said that MQA is or is meant to be identical to high resolution files. Indeed, they are not. The authors of MQA have repeatedly pointed out that comparing wave forms is completely the wrong approach. Mastering engineers in the studio tweak settings on the encoder until they get a result they are happy saying is a reproduction of the original master. This very simple point is lost on so many.

      1. I in fact still see them being advertised as lossless. Much of the advertising copy currently on MQA’s website itself would tend to lead people to believe it is lossless. I.E. fully replicating the sound of the master. This is where there is an exception. By the way I didn’t just look at the resulting null waveform but listened also granted with a lot of amplification as the null was quite deep but far from perfect.

        Even the claims made in the white papers says that it is bit perfect to 17 bits resolution which means a null should have been achievable to 102 dB. This was not the case. With the amount of amplification used it should have still just been barely audible but it wasn’t just barely audible. It was quite audible with 20db amplification & just barely audible without amplification. You could tell it was music not noise but only frequencies above about 5KHz.

        Seeing as how I could achieve perfect nulls from 2 identical but separately recorded from analog recordings of these files as listed above if I recorded the same file twice which is the only way to test such files as you have to convert MQA files to analog to test.

        I repeated these null test proofs to test my skills at nulling daily during these tests to ensure I was doing the tests correctly & that I was on top of the skill needed to do them. These tests took a lot of time.

        1. “I.E. fully replicating the sound of the master.”

          That does not mean lossless from a digital file perspective. As I’ve said elsewhere in these comments, the claims are not misleading. The claims are being misinterpreted. Please tell me where in the AES paper you see something that you interpret as a claim that it is bit perfect to 17 bits resolution. Bit perfect in comparision to what? I see some mentions of a 17 bit noise shaper, but nothing like what you are suggesting. It is not at all being marketed as lossless. The wiki page describes the encoding as lossy, but points out that is uses perceptual encoding based on psychoacoustic models. Bob Stuart has said many times that you will see a different looking waveform, and that this should not be the criteria. We do not consume music with our eyes but with our ears.

          1. “Replicating the *sound* of the music”. Sounds like all those digital music companies that said “CD-quality”.

            Reality is that the noise shaping is unnecessary if they actually had a better perceptual model for the processing of the authentication data – aka CCI AKA watermark information &/or message – it’s tough to noise shape the dependency of the carrier with additional “data” – mostly bc of recoverability in a reliable & authenticable manner.

            Example, wanting to authenticate a sound recording so, having to preprocess or include the encoding process in the authentication as the encoding necessarily changes, at the very least one bit, in the recording. That is, to differentiate between all other watermarked copies in a group of copies, authentication has to work for the copy in question.

            Refer to earlier papers on “buried data” to get a better sense of raw measures of dB versus actual subjective listening tests done over 20 years ago in the SDMI & related standards tests.

          2. Ironic Slip, I got the impression that in MQA each copy is authenticated the same way. I may be wrong, but as this is not DRM the authentication should just be a check that it decodes properly. From the source (https://bobtalks.co.uk/blog/mqa-philosophy/mqa-authentication-and-quality/):

            “* The MQA ‘Studio’ (blue light) gives confirmation directly from mastering engineers, producers or artists to their listeners. MQA Studio authenticates that the sound you are hearing is exactly as played in the studio when the music was completed and, by implication, that this is also the definitive version of the recording at that point in time.
            * A second level, ‘MQA’ (green light) is available to indicate that although the stream is genuine, provenance may be uncertain or that it is not yet the final release.”

            I don’t know what’s involved in the blue light authentication. Maybe it’s some sort of digital signature. There would have to a registry I think. There would also be the issue of revocation and how that should be handled. The green light seems simple enough. Neither suggests that this is tied to a particular copy.

  14. Thanks to flayman for giving the best info on this yet. So many of these commenters decide their own opinion on what the claims of MQA lossless are is annoying. Just a bunch of blabbermouths that show no proof, or will even consider attacking this in a way that MQA explains. Most just want to do analysis on the file, rather then listen. Only Ronald L. Brant (sp) did actual analog recording testings, and shows some high end changes,can we even hear those? Yet he still wants to argue it’s not lossless for unheard sound, fucking stupid, all of you doing that. I still have no proof, and sure, probably not good for the industry, but this shiz sounds freaking good man! Go listen to it and chill the F out.

    1. Be annoyed.
      The point of the discussion is obviously not directed at you. It’s a reckoning between what MQA principals say & what has already been developed, tested, and deployed in the wild going back to the early 1990s (Craven & Gerzon re: “buried data”).
      Many of these innovations are lost in the obfuscation.
      What was missing in this discourse is an understanding of cryptography & steganography.
      CBS did tests that were out of band & BBN was relied on by industry folk around the same time using a notch filter.
      That Major Labels pleaded with three letter agencies to “fix” piracy in the late 1990s, while destroying VC funded startups is telling, even today.
      Ask any musician!
      The disclosure of MQA is misleading.
      What “sounds good to you” may one day be erased because you didn’t understand that DRM was actually involved & it was not “you”-facing technology.
      Even InterTrust raised another round of financing. What did Meridian do to get an “authentication” gig?

      1. “What did Meridian do to get an “authentication” gig?!

        From what I can see, Meridian Audio was co-founded by Bob Stuart and MQA was developed within that company before it was spun off. There’s a lot to unpick in this comment. You seem to know your stuff, so I’m curious. What is misleading about the disclosure of MQA? This is the first time I’ve ever heard of anyone using DRM-like technology to encode value added content masquerading as dither. I don’t understand the push back. It has not exactly taken the music industry by storm. It has the feeling of a labour of love, but it is held back by irrational distrust based on criticism of the closed nature of it. Has anyone asked Bob Stuart why they went with trade secrets? I presume it’s to minimize tampering. It’s probably something to do with that blue light. Would an open standard win over these detractors, or would they still just continue to claim it’s pixie dust?

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.