My new article series on MQA.

Discussion in 'Audio Hardware' started by LeeS, Jan 9, 2018.

Thread Status:
Not open for further replies.
  1. tmtomh

    tmtomh Forum Resident

    Thanks for these links, @LeeS . I watched both Stuart ones (though I'd seen the origami one before) and the first Beekhuyzen one.

    I think others here should watch them too - though not exactly for the reason you intended.

    Let me first say that of course Stuart is a well-established figure and very knowledgeable and experienced; and that Beekhuyzen, though I don't know his background, generally is a very clear, effective explainer of a lot of the major concepts.

    That said, I'm a little shocked that you would recommend these videos for the purposes of engaging meaningfully in the level of argument that's happening here, primarily with @Archimago and others. Each of these videos repeats discredited canards about digital audio sampling, and/or states or repeats assertions about MQA that are not true.

    1. In "Audio Origami," Stuart goes to great lengths to talk about pre-ringing without calling it pre-ringing, and then simply says, "we use the right filters or we fix these filters," providing absolutely zero offers of evidence that MQA's filters are as good as or better than other filters; or that other companies' DACs are deficient in this area.
    2. Stuart argues for 96kHz sampling by claiming the audible range of music is harmed if frequencies between 20kHz and 50kHz are removed: "We know people can hear the difference if we remove that - scientific proof." That, to put it plainly, is bull****. To the best of my knowledge, that claim is based on a single Japanese study that never has been able to be replicated, and whose results are highly suspect to say the least, contravening (to my knowledge) pretty much the rest of the body of audio science. In fact, Beekhuyzen explicitly contradicts Stuart on this too, saying we know that sound above 20kHz is "not relevant."
    3. Stuart says that a sample rate above 96kHz - e.g. 192kHz - is beneficial because "higher frequencies allow the D/A converter to run at higher speed." Even if you subscribe to the idea that higher sample rates improve time-domain performance (more on that below), that's crucial in capturing the analogue sound at the A/D stage. The D/A stage has nothing to do with it - an oversampling DAC can run as fast as you want and it won't improve time domain performance one whit.
    4. If you pay close attention, Stuart implicitly admits that the reconstruction above 96kHz is not lossless, though he words things very cleverly so that you'll miss it if you're not already listening for it. He says MQA folding "will unwrap it perfectly" - but he's referring only to the initial unwrap. He then says, "once it's unwrapped we put the original sample rate to what it was, the original bit rate to what it was and you get the original sound restored" - which is IMHO just a shamefully untrue and misleading thing for someone with his credentials to say.
    5. As for Beekhuyzen, he repeats the canard that a higher bit rate has better low-level detail because it allows more different volume levels. This is another shockingly ignorant statement for someone in a position of information authority. 24-bit does indeed have more volume levels - but all the extra ones are just further below the noise floor. It's not like 24-bit lets you magically squish the bits closer together in amplitude. A bit of resolution is still about 6.2dB - you can sample at 64 bits and you still won't get more volume levels within the audible range of volume. If what he says were true, then SACD/DSD would be lower resolution than redbook CD (yes of course SACD/DSD uses noise-shaping, but that's not the point here).
    Finally, there is one claim Beekhuyzen makes that might actually have some truth to it. He says that only a 192k sample rate is fast enough to match humans' biological time-resolution. As he says, when we hear a twig snap, our auditory system starts reacting within 7 microseconds, and lower sample rates sample the sound less frequently than that.

    As has been noted above, there is no evidence that we can distinguish the timing of music at that level (as opposed to a single test tone or a snapping twig). But even putting that aside, if 96k sampling is not fast enough to match the human auditory system, then why on earth is MQA limited to a 96k sample rate? Remember - MQA only upsamples to 192k. A 192k PCM source gets destructively downsampled in the MQA encoding process (along with some of the precious bit depth too). And if there's one thing upsampling cannot do, by definition, it's restore the time-domain purity of the original, higher sample-rate source.

    Now, Lee, I am an amateur enthusiast. I teach close reading for a living at the college level, and I've always had a good aptitude in math and in logical analysis, so I like examining this stuff. But basically I'm just some guy. And if even I can see the above glaring holes in Stuart's and Beekhuyzen's narratives, then I would certainly expect someone in your position to see them too, and as @Agitater rightly notes, to stop using the argument-from-authority technique (in reference to Stuart and Beekhuyzen, but also in reference to yourself) as a crutch and a cudgel in this discussion. Enough with that, please.
     
    Last edited: Jan 22, 2018
    basie-fan, Dave, LarryP and 7 others like this.
  2. tootull

    tootull Looking through a glass onion

    Location:
    Canada
    Because you don't have your Music Quackery Association diploma. :laugh:
     
    McLover likes this.
  3. Steve Hoffman

    Steve Hoffman Your host Your Host

    You've posted on this one thread over 60 times. The system thinks you're a troll probably.
     
  4. ribonucleic

    ribonucleic Forum Resident

    Location:
    SLC UT
    Oh, artificial intelligence. :hide:

    (Just kidding. I love all you guys.)
     
  5. LeeS

    LeeS Music Fan Thread Starter

    Location:
    Atlanta
    This is a very weak argument against the posted videos as your points 1-5 have no substance. Let's go through each:

    1. Pre-ringing has been shown to be an audible problem and Stuart is the recognized leader in creating filters that solve this problem. Indeed, it is likely that MQA grew, in part, out of his success there. Way back Meridian was the innovator in this area so I suppose Bob feels he doesn't have to do anything other than refer to his papers on the subject.

    2. This is just opinion on your part as most audio researchers have found that frequencies above 20K matter. See James Boyk's work and many others including the papers from McGill's audio research lab.

    3. Again, just your opinion. Stuart has written that higher sampling rates at both the ADC and DAC stage matter in terms of improved timing but also using easier filters. His early Coding2 paper which I have linked to repeatedly here in the past proved the mathematical case for hirez. The DAC certainly is a conversion that benefits from the higher sampling rates as well. That is not in dispute.

    4. My research is showing that the MQA encoder is perfectly lossless in terms of encoding a 24/192 and bringing it back. The reason that some feel it is lossless, I believe, is that the MQA encoder is focused on that magic triangle where all the spectral content of the music is. Because it is a different way of encoding compared to traditional PCM encoding, there is a bit of apples to oranges going on here in terms of the comparison to traditional PCM.

    5. Beekhuyzen is correct in saying that the extra bits allow for more octaves to be captured. That's basic signal theory. What people initially thought was that there were some limits as to what people were able to hear. New research in fact shows we can easily hear those extra bits. Of course, the applies to both PCM and DSD and no recording engineer who has done a simple 16/44 to 24/96 has felt different.
     
  6. LeeS

    LeeS Music Fan Thread Starter

    Location:
    Atlanta
    Tater, you keep repeating this lie but anyone who has been at these major MQA demos knows that most people easily heard the difference. The LAAS demo was near unanimous from people in the room after Peter played his recordings in MQA. The MQA file was obvious in its clarity and extra presence. It wasn't even close.

    Please stop repeating this lie that no one has heard an improvement from MQA. It's simply not true.
     
  7. ServingTheMusic

    ServingTheMusic Forum Resident

    Location:
    SoCal
    "4. My research is showing that the MQA encoder is perfectly lossless in terms of encoding a 24/192 and bringing it back. The reason that some feel it is lossless, I believe, is that the MQA encoder is focused on that magic triangle where all the spectral content of the music is. Because it is a different way of encoding compared to traditional PCM encoding, there is a bit of apples to oranges going on here in terms of the comparison to traditional PCM."

    Utter and complete bunk. MQA CANNOT produce any content above 96 Khz and it DOES NOT produce 24 bits.

    You are spreading lies.
     
    McLover, Brother_Rael and j7n like this.
  8. ServingTheMusic

    ServingTheMusic Forum Resident

    Location:
    SoCal
    This is a fabrication. I know several listeners who were there any they concluded nothing of the sort.
     
  9. thrivingonariff

    thrivingonariff Forum Resident

    Location:
    US
    You're evidently overlooking the word that he italicized:
     
  10. ralf11

    ralf11 Forum Resident

    Location:
    Earth
    speaking of lying... Lee, please stop

    journalists are supposed to be accurate
     
  11. LeeS

    LeeS Music Fan Thread Starter

    Location:
    Atlanta
    They must be deaf. It was very obvious.
     
  12. ServingTheMusic

    ServingTheMusic Forum Resident

    Location:
    SoCal
    On this page alone there are half a dozen fabrications.
     
    Brother_Rael and tmtomh like this.
  13. ServingTheMusic

    ServingTheMusic Forum Resident

    Location:
    SoCal
    Obvious to shills.
     
    Brother_Rael likes this.
  14. tmtomh

    tmtomh Forum Resident

    Well, @LeeS , I've tried to take you at your word and engage in good faith, but if this what you come up with when directly challenged on specifics, perhaps some of your more vocal detractors have a point. At any rate...
    1. I never claimed pre-ringing is not a problem, and you know it. "Stuart is the recognized leader" - just your opinion, as you would say, and irrelevant to boot. The point is that one doesn't need MQA, or anything like it, in order to apply a digital filter at the DAC stage that minimizes pre-ringing.
    2. I dispute this point, and will leave it to others to comb through the actual research and decide for themselves.
    3. Complete hogwash: An oversampling DAC cannot improve time-domain accuracy beyond the sample-rate of the original ADC - which is relevant because MQA destructively downsamples 192k PCM to 96k. So when "we put the sample rate back" at the end, as Stuart says, that upsampling process cannot reverse the time-domain losses incurred when the 192k original was downsampled to 96k.
    4. If your research shows that the MQA encoder is "perfectly lossless," then either your research or your understanding of the meaning of "perfectly" and "lossless" is deficient. And you never addressed the fact that Stuart himself implicitly admits that the MQA render stage is not lossless but rather just upsampling and filtering.
    5. Beekhuyzen doesn't say extra bits allow for more octaves to be captured. He says that higher sample rates allow for more octaves to be captured - and he says that the higher octaves are "not relevant" for human hearing. He says that higher bit rates allow for better low-level detail, but that's only true if you are capturing music with more than 96dB of dynamic range - and really only true if you're capturing music with more than 120dB of dynamic range, because of the influence of noise-shaping. Which once again is why DSD can be 1-bit and still produce approximately 20 bits' worth of dynamic range/snr. Let me clarify that I am all for recording in 24-bit (in fact, in 32-bit float or even 64-bit float if one wants). And I certainly have no objection to all digital music, whatever its sample rate, being released in 24-bit instead of 16-bit. But virtually no music ever ends up being mastered and released with 120dB of dynamic range, because it would unlistenable except perhaps for demo and testing purposes. And in any event, it's already been established that MQA doesn't even achieve 24 bits of depth anyway.
     
    Last edited: Jan 22, 2018
    j7n, basie-fan, Edgard Varese and 5 others like this.
  15. Metralla

    Metralla Joined Jan 13, 2002

    Location:
    San Jose, CA
    This is more fun than I have had in a long time.
     
    ds58, dartira, LeeS and 3 others like this.
  16. ServingTheMusic

    ServingTheMusic Forum Resident

    Location:
    SoCal
    His answer to #4 clearly demonstrates he has ignored all verified independent information on MQA encoding.

    It clearly establishes a clear agenda to ride the MQA train through the tunnel even it means making things up.

    The fact he can post that MQA is perfectly lossless when Paul Miller's, Archimago's, and even John Atkinson's measurements
    have shown it to be CLEARLY LOSSSY is just appalling.
     
    McLover, Brother_Rael, j7n and 5 others like this.
  17. tmtomh

    tmtomh Forum Resident

    Gotta agree with you - it's ridiculous, and unnecessarily so: He could defend MQA without resorting to such statements.
     
    McLover and Brother_Rael like this.
  18. ServingTheMusic

    ServingTheMusic Forum Resident

    Location:
    SoCal
    In general, when you have an anti consumer, pro industry agenda, facts are meaningless. When you have to ignore facts and fabricate,
    it clearly shows where you stand.
     
  19. Mal

    Mal Phorum Physicist

  20. ricko01

    ricko01 Well-Known Member

    Location:
    Blue Mountains
    I dont have a dog in this race.. my digital collection is stable, I add to it rarely these days and I dont generally chase the next new thing.

    I do however have opinions on MQA that generally dont differ from the general consensus on this thread so I wont bother repeating what are now ad nauseam (but valid) questions about MQA.

    But here is what I dont understand.

    LeeS is basically taking a beating on this thread (not saying he is losing because he certainly keeps engaged and responding) but a beating in the sense that he has convinced no one on this thread to see his view point.

    So what is his motivation?

    I dont know LeeS but from what I have gleaned from this thread he is somehow part of the "industry".. a part time reviewer which potentially entails some industry "entanglements".

    Thus LeeS is not some Joe Blow audio enthusiast who basically has a hard on for MQA...he is an industry insider.

    So that means two potential motivations to keep coming back for more:

    1-He has seen the light (much as Blues Brother Jake did) and as such is on a genuine and personal mission from God to save the great unwashed on this forum from their ignorance

    OR

    2- He is on a [paid] mission from Bob to save the great unwashed on this forum from their ignorance


    While I dont post much on forums, I am certainly an active voyeur on half a dozen audio forums and I have rarely seen a guy in a forum keep banging away, pressing his view point as tenaciously as LeeS and certainly not on what has been for several years now the most controversial product discussion for many a decade... so maybe the reason is more of (2) than (1)

    Peter
     
  21. ribonucleic

    ribonucleic Forum Resident

    Location:
    SLC UT
    In fairness, what Lee wrote was "perfectly lossless in terms of encoding a 24/192 and bringing it back". [emphasis added]

    Similar asterisks are found in Bob Stuart's own comments:

    A Comprehensive Q&A With MQA's Bob Stuart

    So he appears to be saying: MQA isn't lossless. But it's better than lossless.

    Carry on.
     
    Kyhl, soundQman, LeeS and 1 other person like this.
  22. LeeS

    LeeS Music Fan Thread Starter

    Location:
    Atlanta
    I'm not convinced that MQA "destructively downsamples" to 96k. Based on my research, the evidence is that the 96-192k is folded down by the MQA encoder. Also, I differ with Hans on the frequencies above 20K not having an impact on the audible range.

    Read this: There's life above 20 kilohertz! A survey of musical instrument spectra to 102.4 kHz
     
  23. ServingTheMusic

    ServingTheMusic Forum Resident

    Location:
    SoCal
  24. tmtomh

    tmtomh Forum Resident

    Thanks for this. If one defines lossless as, "We start with 24/192 and we end up with 24/192, and what happens in between is not lossless, but it's a better kind of lossy than what typically happens inside a DAC," then I can see how that could be a rational claim in favor of MQA, but still a misleading and IMHO dishonest claim about being lossless.

    Furthermore, if you have a delta-sigma DAC, then it's never going to be bit-perfect internally - but an MQA firmware update won't change that. So then you're left, once again, with MQA's claim that their algorithm applies filtering that corrects for the DAC's native sonic impacts; a claim that might or might not be true, but which does not need any of the other MQA stuff - it's just a digital filter not unlike the selectable ones available on some DACs and CD/SACD players.
     
    Dave, Kyhl and ServingTheMusic like this.
  25. LeeS

    LeeS Music Fan Thread Starter

    Location:
    Atlanta
    Yes, I think when MQA reps have said it is "lossless", it means that the MQA encoder and decoder are lossless. I guess they would argue that the encoder itself captures all the "spectral content" of the music so anything outside of the encoded then decoded file would not be audible.
     
Thread Status:
Not open for further replies.

Share This Page

molar-endocrine