Sunday, 22 October 2017

MQA: FINAL "Final" comments... Simply put, why I don't like MQA.

Looks like an interesting art exhibit :-).
I know... I said the previous post was "final". But I just had to kick the horse one more time. :-) Seriously, if you're sick of MQA chatter already, don't bother reading this - you've been warned!

This past week, I added my views to this thread on Computer Audiophile (from about page 198 in an epic thread with >5000 comments!). I know that forums can get pretty heated and ugly, but it is good fun sometimes so long as you don't let it get personal - it never has to, we're talking audiophile stuff after all - first-world problems, nothing life or death. Since forum threads and comments tend to get lost easily, I thought it would be reasonable to just enumerate some main talking points in the hopes of furthering clear debate (if still necessary)...

Suppose we start with an original 24/192 hi-res file which I think most of us would agree would be able to encode all that human ears/minds can perceive and likely more (sure, we can argue this point but remember that few recordings are even true 24/192). Now, suppose we want to stream this because that's the music distribution model we see growth potential in and we want to decrease the data rate to an equivalent 24/48 file (as per TIDAL's target data stream) as a reasonable Internet transfer speed.


We have 2 options currently:
1. We simply downsample to 48kHz while maintaining 24-bit resolution and give up the ultrasonic frequencies above 24kHz = STANDARD downsampling. 
2. We sacrifice 24-bit depth to "typically 15.85 bits" (Bob Stuart's words), and encode the ultrasonic frequencies from 24-48kHz in a lossy fashion = MQA encoding & decoding. [Throw in some stuff about "de-blurring" while you do this of course and claim you can recover everything else you "need" back to the "original" 192kHz. Turn on a LED/indicator telling us MQA decoding is happening, that there's no error in the stream and it's the "original" resolution (meaningless, but that's fine).]

Which of the 2 do you choose? Do you think there's going to be a massive difference in sound quality?

Personally, I think Option 1 is just fine and have said this from the beginning. Lossless 24/48 audio sounds great and in many cases would be easier to compress than MQA for streaming. Heck, we could zero out the last 4 bits and maybe compress a 20/48 stream for more data savings without worrying about anyone complaining. Plus, since time domain performance is linked with bit-depth, one could argue that maintaining true 24-bit resolution provides better time-domain performance below Nyquist. It will be "open" for easy adoption by manufacturers and not impose any licensing fee.


But TIDAL and MQA presents to us option 2. Here are my concerns:

1. It adds complexity and cost (unnecessarily):

a. We already have many high-resolution DACs out there. Do we need a new "format" that's not "fully" backward compatible?
b. Only MQA-certified software / hardware available for decoding. This reduces options for consumers and likely always will especially in the audiophile world where small manufacturers may find it hard to justify licensing costs. Furthermore, small independent developers, free software and Linux distributions would not be able to develop for MQA decoding unless under some kind of license. Ultimately this reduces innovation.
c. MQA versions of downloads such as what we see on 2L (US$19 for 24/96, $23 for 24/192, $24 for MQA, and $30 for studio master DXD) often costs more than downsampled very high quality 24/192. Does MQA encoding actually add any value or are we seeing the added licensing fee passed on to consumers? Is there any justification for the existence of MQA for file downloads when we're not faced with the bitrate limitations of streaming over the Internet?

2. Technical concerns:

a. It reduces the actual bitdepth to the aforementioned "typically 15.85 bits" and up to "17-bits" resolution when decoded. These numbers are from Bob Stuart. I know they want us to ignore this and focus on the analogue output rather than care about the digital technical values. We can argue about this of course; but the point is that a resolution limit has been imposed which is lower than standard 24-bit digital audio.
b. Reconstruction of the first unfold into the 24-48kHz audio frequencies is lossy in nature compared to the original. Again, they may argue that there is no audible difference, and yet again, a limitation has been imposed that would not exist with 24/96, much less from a 24/192 source.
c. MQA upsampling is done with "leaky" filters resulting in weak ultrasonic suppression of aliasing. Remember that DAC designers can easily program their devices to do this if felt to be preferable for their designs. Also, we as consumers can choose to do this ourselves with software upsampling if we really think this is a good idea (software like HQPlayer perhaps or even SoX).
d. For those who want to do advanced DSP like room correction filters, ambisonic processing (like this one), or surround processing, we cannot have access to the full digital resolution because of the proprietary MQA decode process. (This is a big deal IMO that limits flexibility and progress as we aim for better sound quality for hi-fi enthusiasts.)
[Given the technical concerns, what is the meaning of advertising claims of "de-blurring", how can it "fundamentally change the way we all enjoy music", represent a "new paradigm", or "reveal every detail of the original recording"?]

3. Minimal audible difference:

a. Digital subtraction tests show little difference. 
b. Blind listening test with 83 listeners show no clear preference. (In fact, in some situations, standard hi-res was preferred.) 
c. Inter-DAC comparisons with MQA decoding show no evidence of significant "authentication" of a standard "studio sound" in the analogue output beyond the expected DAC differences.

4. Unclear Digital Rights Management (DRM) implications:

a. Already as in (1), MQA certified products such as an MQA DAC adds expense and there are fewer software playback options. It locks us into a proprietary mechanism right from the start. This is a backward step considering the years of open and free (for consumer) lossless file formats like WAV, FLAC, ALAC, WV, APE, etc...
b. Potential future stronger mechanisms such as encryption or scrambling to enforce playback only with licensed products at the expense of significant sound quality if one does not "buy in". This is not an issue currently if you feel undecoded MQA sounds as good as CD resolution. However, there are hints in the MQA firmware that more severe forms of DRM can be imposed. It doesn't take a genius to see how that MQA "authentication" indicator can be linked to such a mechanism. Do we even need to entertain this possibility as consumers in 2017 after years of rejected attempts at DRM and digital watermarking?

I might have missed other arguments but these are the top-of-mind for me. I'm certainly happy to be wrong if evidence can be provided and explained how some of the technical pitfalls like item 2d can be circumvented without loss of resolution.

Just like with the TAS articles recently, when we look at forum debates, one essentially never finds evidence-based arguments in favour of MQA (like the Danny Kaey video from Rocky Mountain Audio Fest last week as well). IMO, unless the proponents can look at that list above and provide clear answers, it's really not much of a discussion, is it?

Even worse I think is when proponents avoid the evidence and facts, instead start using a pejorative tone like calling those arguing against MQA "malcontents" in a debate or resorting to a mindless philosophy of "Are you having fun?" to substitute for valid points. I trust as adults, our emotional and intellectual lives are more nuanced, complex, and capable of grasping grander values, concepts and desires. Even if we think "it's just audio", the way one argues tells us a bit about the depth of the person that we are.

Unless we get past these issues, we'll just continue to talk past each other primarily with opinions and idiosyncratic ideas rather than simply face up to what MQA is and what it represents; a closed vision of "studio sound" as presented through the business model and technical lens of MQA Ltd. Accessible fully only thorough devices and software licensed within their controlled proprietary mechanism.

---------------------------------

Hopefully this FINAL "final" post on MQA wasn't too painful for everyone. I'm done.

Alright folks, I'm literally getting ready for my flight overseas as I post this :-). Looking forward to some warmth, sunshine, good food, sights, sounds and stimulating discussions with friends and colleagues.

Have a wonderful few weeks and I hope you're all enjoying some great music!

ADDENDUM:
Fascinating comment from Charles Hansen of Ayre. He has been an outspoken opponent to MQA from within the audio industry for awhile. And a critic of the industry's direction - check out this post particularly along with the rest of that thread addressing the audiophile press! Holy smokes.

It's one thing for consumers like myself to express criticisms of what I see. But there's certainly another level of courage needed to say what must be said as a well-known insider member of the industry.

ADDENDUM 2:
Wow. Quite the "industry insider" weekend...

Here's Bruno Putzeys on Facebook discussing Bob Stuart's MQA AES 2017 presentation that just happened this past Friday! Brilliant comment - a stab right at the heart of the technical issues...

Hmmm... "No Scientific Tests Were Done, Says MQA Founder" :-)

Any bets as to whether this would ever show up on the pages of Stereophile or The Absolute Sound or What Hi-Fi? or Hi-Fi News & Record Review? Have they ever asked Bob Stuart or company representatives questions about this over the years? Plenty of opportunities since December 2014, right?

ADDENDUM 3: (November 7, 2017)
Just wanted to add with a comment I made to Charles Hansen on AudioAsylum. Might be of interest to readers here...

Greetings from Shanghai, Charlie, 
Just wanted to add given that you're also a manufacturer... 
I think we all appreciate the importance of making money! Nothing wrong with capitalism, corporations, and the profit motive of course. When companies make good products, this is to be lauded and rewarded. 
The issue IMO with MQA is that it's bogus. It was marketed in a bogus fashion, it is not capable of what it claims, and even if we don't call it all-out DRM, they have not addressed the observation of such an underlying intent or to clarify what limits they seek to protect consumer interests. 
This is IMO bordering on fraud (again - this is my opinion). There are IMO lies being spoken of in the advertising and claims. When this happens, then absolutely, as audiophiles and music lovers, we must come out fighting against such nonsense whether perpetuated by the company itself or in the moronic audiophile press. 
As audiophiles, I think we can all find and see products being advertised demanding that we be skeptical of. For the most part, we can probably just let it pass... But in the case of MQA, there's a real threat here of that corporate "greed" infecting the music production and distribution chain that provides NOTHING to the consumer; in fact, it's counterproductive for the consumer. This is a major problem with MQA and again, demands thoughtful audiophiles to respond in kind.

56 comments:

  1. For what it's worth: I enjoyed this blog post more than the last 4 ones ;-)

    ReplyDelete
    Replies
    1. Thanks Techland.

      Always good to end essentially on a summary!

      Delete
  2. Excellent summary, archimago, thanks
    How anyone can claim that a piece of paper that has been folded and unfolded is closer to the original than one that has not been tampered with, eludes me. No wonder "no scientific tests were done" during development.
    The intention of MQA to deliver better high resolution audio through limited bandwidth is laudable, but I firmly believe it does not matter beyond 20/48. The influence of the mastering itself is so much more important, even at 16/44.1, that any improvement by the folding / unfolding process would be dwarfed in relation.
    In fact I believe the added complexity of the MQA processing at both ends probably does as much damage as any potential improvements would bring.
    Keep it simple.

    ReplyDelete
    Replies
    1. The paper folding analogy (which of course came from BS originally) is an interesting one: the folded paper has to be compressed so that it becomes the thickness of a single sheet of paper; when unfolded, both halves of the paper are clearly damaged and show print-through of ink from one side of the fold to the other!

      Delete
    2. Hi Jacobacci & Bob,
      Yeah, stands to reason that it's impossible to have *everything* intact. And this is what was illogical from the start. The promise of hi-res, studio sound, hear everything, all in a neat 24/48 package regardless of whether the music started life as 24/96 or 24/384.

      Ultimately, something had to give.

      The argument at the end of the day is no different than any lossy format. Ultimately choices have to be made as to perceptual limits of human hearing. MQA went with bitdepth reduction and lossy ultrasonics. And threw in smoke and mirrors around the filtering and "time smear" / "de-blurring" in the form of track-selectable 16 upsampling filters.

      That's all it ever was IMO.

      Delete
  3. "bit about the depth of the person"... haha - nice! can I haz t-shirt plz: personal bitdepth unfolded :D

    ReplyDelete
    Replies
    1. Feel free to put it in on a T-shirt Gimme :-).

      Maybe put it up for sale on Amazon or Baidu or whatever... I'm sure it'll sell!

      Delete
  4. When MQA first became available on Tidal, I quickly signed up. My early listening impressions were very favorable. Several 2L sourced MQA recordings had sound quality that I considered to be the best digital recordings I had ever heard. I was sold on MQA !

    A couple months later, I discovered the 2L download site and downloaded 2L content of the same recordings that I was so impressed with on Tidal. These files had similar sampling rates to the MQA processed Tidal content and also higer DXD sampling rates. I quickly discovered that the downloaded files were at least as good as if not better than the MQA processed "deblured" Tidal content. The DXD files sounded noticably superior ! Sadly, I had erroneously attributed the Tidal high quality to the MQA process. In fact, the key factor determining ultimate sound quality is the recording studio itself. MQA seemes to be designed primarily to extract licensing fees from everyone. Sad.

    BTW, here an interesting post regarding MQA - https://benchmarkmedia.com/blogs/application_notes/163302855-is-mqa-doa

    ReplyDelete
  5. "No Scientific Tests Were Done, Says MQA Founder" :-)

    Any bets as to whether this would ever show up on the pages of Stereophile or The Absolute Sound or What Hi-Fi? or Hi-Fi News & Record Review?

    Well, one time blind test denialist Robert "The blind (mis)leading the blind" Harley, is now a huge proponent of "the latest neuroscience and psychoacoustics research" (which of course requires blind tests) as espoused in his article, titled "Let the Revolution Begin MQA and the Overthrow of 20th Century Audio".
    Or so it seems...

    ReplyDelete
    Replies
    1. Hello Ammar,

      Yup. Stuart and Harley cannot have it both ways. To make a claim of a wholesale "revolution" and that this is a big deal, the burden of proof is on them to show the difference.

      Preposterous that they would make such a big deal about it all and splash it on shiny magazine article pages yet have nothing to show for it except the testimony of a few.

      IMO, unless there is some decent evidence, might as well ignore for the most part...

      Delete
    2. Not sure if you are an AES member, but some folks at McGill U are proposing just such a venture (no reason to think BS & Co. would ever do such a test, any more than Sony for DSD etc). http://www.aes.org/e-lib/browse.cfm?elib=19340
      I have already asked the relevant question in the Comments section.
      But as you know, such highly controlled lab conditions, often using headphones, etc, still won't answer the question - whether believers can hear anything in their 2ch home audio/loudspeaker rigs in their reverberant rooms. Precisely why M&M conducted their "Hi Res" investigation using audiophiles/own systems/music choices.
      So my plan is to utilize members of my own audiophile club, play back MQA files of their choice on a "high end" system, using a Mytek Brooklyn for "full unfold" blah, blah...but simultaneously run a 2nd output through an ADA loop (16/44 etc)...and compare the pure MQA "deblurred" file vs its "blurred" self. IOW, an actual test of MQA, NOT mastering shenanigans.
      That should be interesting ;-).

      Delete
    3. Hi Ammar,
      Nope, I'm not an AES member but I did see this "engineering brief" last week. It will be interesting how they proceed with this out in Montreal.

      Absolutely, if the only way to prove statistically significant results is in a pristine, controlled lab setting with presumably high-quality equipment, then that is actually not impressive and in fact suggests insignificance in the real-world where MQA is being used.

      I like your idea of a test with your audiophile club - I hope you document it well, take lots of photos, and write it up for us to see :-). If there are no differences, then we can safely say that the "origami" and "de-blurring" makes no difference. I'm of course particularly interested to see if the results are as I suspect based on what the Internet Blind Test recently has shown!

      Have fun with that!

      Delete
  6. Yep, this MQA spectacle was very fishy from the start.
    OK, compressed inaudibles, check. But as you clearly pointed out why those tracks have to be more expensive than 24/192 ??? Wasn't 24/192 the source material for those MQA encodes?
    Several months ago I noticed that on 2L as well.

    If this hobby has vinyl and tube enthusiasts, then I am RedBook lover.
    Crystal clear audio, no BS, small size and unprecedented compatibility. Over the past 35 years we managed to fix all its problems. Sure, early players had audible aliasing, but we managed to fix those with oversampling.

    In responsible hands 44.1/16 or 48/16 can sound truly spectacular.
    Personally I prefer 48/20 for recording, as -24dBFS equals 16bit DR, and I rarely go that low. So even 20bit for recording is enough for 16bit result.

    ReplyDelete
    Replies
    1. Nelson Pass recently stated that we’ve pretty much hit the end of science with regards to amplification and that his main focus lately has been the “art” of design. When Archimago showed that an IPhone 6 will hold its own even on 24 bit files he pretty much confirmed that unless we invent better ears we passed the end of science digitally a long time ago (2010?).

      So it gets back to the art of the "responsible hands" and the art of the content. And I’ll add I’m a passenger on Pass and other’s “art of simple” bandwagon. I don’t need MQA and origami folds for those; FLAC is just fine.

      Delete
    2. Greetings boys,

      I think it's really a slap on the face for the folks who built the free and open lossless compression formats like FLAC or APE or WavPack that MQA should even be losslessly compressed with them! Talk about an uncomfortable dissonance. LOL. Even if legally the authors of these formats cannot prevent anyone from using them, it would certainly be good if they spoke out about this travesty!

      As for hi-fi as a whole, yes, we have to appreciate the limits of human hearing and the limits of the goal of high-fidelity reproduction; especially 2-channel.

      This of course is not to say that we cannot or should not expand audio into new frontiers of immersion - multichannel, object-based encoding, beam-forming technologies, marriage with virtual reality, ultimately possibly mind-machine high bandwidth connections...

      I believe as rational audiophiles:
      http://archimago.blogspot.ca/2017/05/musings-on-being-audiophile-rationality.html
      we can appreciate that a point exists not only of diminishing returns, but that likely even before that plateau, we can subjectively reach a level that's simply not just "good enough", but realize that "I'm not missing anything of significance"!

      George: iPhone 6 was 2014. But clearly since 2010, we know that >16-bit resolution and extremely low jitter were already within the reach of very reasonably priced DACs. By 2013, with asynchronous USB essentially entrenched, there was nothing to worry about with most reputable devices.

      Delete
    3. 20bit recording equipment came before 1995 and 24bit WAVE support was implemented in 1998, followed by WaveEx shortly afterwards. By 1999, DAC/ADC market already had contenders.

      Delete
    4. My 2010 reference was as a hobbyist and not as an engineer or anything. I was thinking that 2010 was about the time that NWAvGuy posted graphs of his ODAC. Not a breakthrough design in anyway but it showed that an inexpensive DAC could handle 24/96, output 2.0vms Redbook, -100db SNR, no THD, and have plenty of headroom to do volume control. For me that was the end of "you need high quality (read expensive) DAC."

      My current 9018K2M based DAC can handle 32/384 and DSD with a -112.5 SNR. All nice to have but most of my files are 16/44.1 with 24/96 being my largest. I guess the only real difference is it doesn't need near as much power; my Android tablet has plenty of juice for a couple of hours.

      Delete
    5. Hi SUBIT and George,

      On the studio production since, indeed 20/24-bit implementations have been around before the new Millennium. Of course, on the consumer side, achieving >16-bit DAC performance in a reasonably priced product (say around $1000) didn't really happen until after 2000. I discussed a little about this here from back in 2013:
      http://archimago.blogspot.ca/2013/07/measurements-dac-waveform-peeping-903db.html

      Although I have never used an ODAC, clearly for a low price of around $100 barebones, >16-bit sound was easily achievable. Very reasonable of you George to note this as a milestone in 2012 I think.

      Of course, the ODAC was just before the onslaught of inexpensive and easily-integrated asynchronous USB chips. Over the last 7 years, the ODAC's Tenor TE7022 USB (adaptive isochronous) has been superseded by asynchronous USB 2 designs which essentially removes the effect of jitter from the source (ie. no need for expensive USB streamers). And at <$100, something as simple as the SMSL iDEA can now play back anything up to 768kHz and DSD512.

      Hardware design has brought us to a place of an "embarrassment of riches"
      in terms of quality and options for samplerates for example in the world of DACs.

      Better speakers.

      Better music production.

      These are really the limitations for high fidelity audio these days. (There are other individual issue of course like one's listening room, or the limitation of just 2-channel sound.)

      Delete
    6. Archimago,

      I totally agree on your conclusion,
      "Better speakers.
      Better music production.
      These are really the limitations for high fidelity audio these days."

      In fact, that's where research and development should be done. And in fact, some development has been done, like the subwoofer/satellite configuration (The wavelength of bass frequencies around 100Hz are 3 meters, so there is not much of a wavefront propagating in a listening room, so bass does not provide much of directional cues for stereo imaging.), bass-reflex and active speakers.
      Or the technique of the missing fundamental to exend the precieved range of bass reproduction by adding harmonics which would imply a much lower fundamental. (Similar concepts can be found in pipe organs.)

      Also, the room correction found in AV receivers really improves sound and fidelity.

      What I would like to see/hear would be a speaker system which provides a stable stereo image regardless of listener position.

      Delete
  7. Isn't 16/44.1 still slightly overkill or has it been proven we need 24/88 or 24/96 and above? Much less special ways of compressing the unhearable parts...

    ReplyDelete
  8. AES conducted a listening test where so-called "Hi-Res" material from SACD was either routed through a 16/44.1 AD/DA converter chain, or directly routed into the amp.
    80 subjects, but no statistically significant aural differences found.

    ReplyDelete
    Replies
    1. Yup, the Meyer & Moran study (Boston Audio Society) as found here:
      http://drewdaniels.com/audible.pdf

      Indeed, they found that SACD and DVD-A material could not be differentiated in a significant fashion even when run through a 16/44 ADC/DAC step. However, remember that this test was run back in the 2005 time frame using albums available then. Notice that a number of the recordings used at the time were typical old analogue material such as the 30th Anniversary "Dark Side of the Moon" SACD, Steely Dan's "Gaucho", Steve Parsons' "I Robot" although some of the other recordings are better. Obviously, these were not hi-res recordings so perhaps not a big surprise that a difference was not heard. In any event, the results certainly suggest that there was no reason to spend more money on these SACD or DVD-A's (for 2-channel at least)!

      For more info on the test, equipment and music used, go here:
      http://www.bostonaudiosociety.org/explanation.htm

      Fast forward 10+ years now and I agree that the human ear/mind has not evolved further :-). And on the whole, there is no great rationale to go hi-res. Personally, as I expressed back in 2013 here:
      http://archimago.blogspot.ca/2013/07/musings-is-cd-sound-quality-1644-pcm.html

      I've always seen "hi-res" as "insurance" when I know that the provenance is of a good pedigree. I can feel better about owning a version that has all the resolution that the studio can get out of the recording, and in the future, if I upgrade equipment, there just might be the chance that I can appreciate more of the source file... Almost universally, the only music I buy in hi-res these days are in the classical genre.

      And like insurance, one NEVER pays much more than the underlying value of the item. A 10-20% premium, maybe even 30% premium *might* be reasonable if the recording is something I truly love and will listen to many times. Otherwise, I'm happy with an actual CD in hand if I can get it at a good price off Amazon.

      Delete
    2. Heh, I usually got the "hi-res" version just because I didn't like the noise shaping algorithm they used for their 16/44.1 files. So I got 24/48 and dithered it to 16bit on my own. :)

      Delete
    3. Yes,

      but on the other hand, the AD/DA converters used in the AES test were simple 16bit/44.1kHz converters with no noise-shaping at all. And still...

      I understand your "insurance" attitude, but do not follow it.
      Even, I argue that 16bit/44.1kHz would be fully adequate even for production. I don't see no reason, to use more than 16 bits.

      Some argue, the 24 bits (or, resulting 20 bits) of "hi-res" were required or beneficial for production, when recordings would be dynamically compressed, so that soft portions would be raised and brought up in volume.
      But, that does not hold water. Usable dynamic compression have threshold values at around -6dBFS down to maybe -50dBFS (which already is very large). But, nothing ever at -120dBFS gets compressed up into the audible range. And, that -120dBFS would be where the lower of the 20bits of "hi-res" would be busy.

      Also, for recording, to get all the 20 bits busy, you need an audio signal with a dynamic range of 120dB, that would require a sound pressure level of approx. 130dB SPL. Because, every microphone has a self-noise, which can be expressed as an equivalent sound pressure level. The best microphones have self noise equal to around 10dB SPL (where 0dB SPL is about the limit of human hearing). (The 10dB figure depends on the type of microphone, and the weighting filter used.)

      If the sound you want to record has less sound pressure level, i.e. is softer or lower in volume than ca. 130dB SPL, some of your 20 bits are idle and unused.
      Of course, you could turn up the pre-ADC gain, but then the lower bits were just encoding microphone self-noise and the noise of the analog electroncis before the ADC.

      So, therefoe, I don't see much reason in 24 bits even in production. However, everybody would use 24 bits recording, because, it is there and doesn't cost any additional cent. It's like your "insurance" argument.

      Delete
    4. Reasonable arguments Tim...

      Even if the actual recordings from mics don't need >16-bits, at least working in the 24-bit domain would maintain the best possible final dithered output after all the mixing, DSP special effects, volume normalization, etc... And that's pretty well I think the only argument there is.

      Like I said, it's only "insurance" for me and this only applies to a small number of recordings I would want in 24-bits.

      SUBIT: Yup, that's one way to do it :-). This is usually what I do with vinyl rips. Get them thru the ADC in 24/96 and use my own downsampling and dithering (usually just a light iZotope MBIT+) down to 16/48 for the music library. Never noticed a difference from the original 24/96 with vinyl :-).

      Delete
    5. > Also, for recording, to get all the 20 bits busy, you need an audio signal with a dynamic range of 120dB, that would require a sound pressure level of approx. 130dB SPL.

      And that's enough to damage many mikes.

      Recording at >16 bits is a matter of convenience as it gives more latitude in setting gain levels without risk of clipping while still capturing the softest sounds.

      Specifically 24-bit is again a convenience, fitting neatly into three 8-bit bytes.

      Delete
    6. Actually I'd disagree here, recording at 16bit is a bad idea, 20bit should be kept as minimum. 16bit is enough but with peaks nearing 0dBFS and TPDF dither. Actual usable DR of 16bit recording is 90dB (at -96dBFS there's no signal so the last 6dB doesn't really count).
      Usually, recording levels are set so that average level reaches -18dBFS to allow headroom for peaks.

      -18dBFS translates in 16bit recording to just 72dB DR, and -66dBFS dithering noise level. That's already audible level of noise.

      On the contrary full 24bit when captured with noise nearing its limits is quite versatile, as you can leave the mics set to fixed level and do all the gain adjustment in digital domain. With 16bit you'd have to constantly fiddle with analog faders to compensate for triple piano flautando in strings (which is nearly ambient noise level) and multiple drumsets.

      Delete
    7. Subit,
      I respect your arguments, and myself I also use the "24bit" mode, simply because it's there, and because of the "insurance" attitude Archimago described above.

      But, just for the curiousity for the technic, I would like to ask again...

      Suppose, we want to record an acoustic guitar which produces peak sound presure levels of - say - 85dB SPL in the distance of the microphone.
      The microphone has a self noise equivalent of 10dB SPL, give or take a few dBs.
      Then, the S/N ratio of that guitar sound can only be 75dB at best, no matter how many bits or how you set the input level control of your AD converter.
      So, in this example, had you calibrated the A/D converter so, that 100dB sound pressure level equals 0dBFS on the digital side, you would be able to record every sound from 10dB SPL (noise floor of mic and analogue circuitry before the ADC) up to 100dB SPL, without loosing any S/N ratio or resolution.

      What is wrong with this way of thinking? Could someone point it out for me?

      Appart from this, there have been excellent sounding digital recordings made with 16bit ADCs. So, 16bits is not really a bottleneck at all, but I agree, 24bits (20bits) would be more convenient for the recording engineer. The only source of noise probably is the mic or the mic pre-amp.

      Delete
    8. Whoa wait , hold on Arch! It's one thing to note that Meyer & Moran used many 'analog sourced' hi rez releases in their tests (and oh how the 'golden ears' scrambled to moved their goalposts when they learned *that*), and to note that when these 'hi rez' releases were converted to Redbook, their was no audible difference. But that doesn't mean there was no reason to buy SACDs or DVD-As! As M&M themselves note, they conceivably *could* sound better (or at the very least, different) than their CD counterparts, if the *mastering* was better than the CD version. At which point, one should rightly ask the industry, why not just release that better mastering on *CD*? Hmmm, I wonder what the an$$$wer would be? ;>

      Delete
  9. Hi there,

    Take this thread as an example:
    https://hydrogenaud.io/index.php/topic,114401.0.html

    Of course, it's a bit lame to use an inexpensive portable recorder as an example, but the idea is you need a very quiet environment (e.g. anaerobic chamber) and very low noise mic/preamp to make use of the lower bits.

    As for the alignment level, my recommendation is "better safe than sorry", otherwise you can get something like this:
    https://forum.cockos.com/showpost.php?p=1852077&postcount=106

    Ancient professional open reel tape machines like those from Studer, Ampex and Otari usually have max level at +26dBu. With the standard calibration 0VU = +4dBu the machines themselves have 22dB of headroom, but the tapes gradually saturate and produce more distortion at higher level. As my observation as a restoration technician in the past, the tapes I digitized seldom have peaks higher than +20dBu.

    How good are those tapes?
    https://youtu.be/BYTlN6wjcvQ?t=45m20s

    Later, people use dBFS, RMS and so on, for digital metering

    Nowadays VU is replaced with LU for digital metering.
    http://www.itu.int/rec/R-REC-BS.1770
    https://tech.ebu.ch/docs/r/r128.pdf

    Surprisingly the "new" standard is pretty close to the old one. It recommends -23LUFS as target level and -1LUFS as maximum permitted level, which is also 22dB of headroom.

    So think about it, regardless of technology advancement, people are dealing with similar amount of dynamic range, at least at the starting point -- recording. But thanks to the loudness war, what consumers got are finished product of lower DR. If you think in this way, the argument of 16 vs 24-bit recording is pretty useless. People use 32-64bit of precision during mixing and mastering anyway.

    ReplyDelete
  10. Dear Golden Ears,

    I have tested personally vinyls vs all types of digital formats of the same recordings on high quality equipment. I have personally read zillions of blogs and papers on MQA,DSD, HD Flac, Redbook, you name it digital formats. Never found any paper or research that would seriously claim that digital version sounds musically better than vinyl. Are we missing or adding something in the digital world. Please make cost-no-object technical breakthrough (MQA, DSDxxx, 48/1024 FLAC,...) that sounds better to music lovers than vinyl in controlled, scientific laboratory setting. Then we can move forward. We all want Tidal/Spotify/Itunes type of interface with excellent sound.

    Contrast our hifi-situation with photography. HD Digital photos are better than film based photos. What is so difficult with audio?

    Yours

    Eero

    ReplyDelete
    Replies
    1. Just checking in from overseas, Eero,

      Hmmmm, where has there been "vinyl in controlled, scientific laboratory" showing that analogue is preferred over the digital versions? In fact, when volume controlled, I would expect that digital recorded analogue playback (say master tapes) likely would be indistinguishable in A/B testing.

      I can certainly say that although I like my vinyl collection, and enjoy showing it to visitors; many of whom have never seen an LP in operation, I cannot imagine honestly preferring the sound of vinyl compared to good digital. The key of course is "good" digital mastering. I can enjoy the cuing up and playback process, love the physicality of vinyl, large artwork, enjoy the collectability of the pressings... It certainly is not because the sound is actually "better" IMO. This judgement is ultimately subjective of course. Some might prefer the sound that the vinyl process creates (including of course the distortions introduced).

      Delete
    2. Didn't even Michael Fremer once admit that a digital capture of a needle-drop was indistinguishable from direct vinyl playback?

      Delete
    3. Ironically, yesterday, I did a demo for the local audio club using vinyl vs it's "digitized" self. TT >dual output phono preamp, one output direct to stereo preamplifier, the other through a 16/44 ADC/DAC loop > 2nd input on preamplifier, V out matched, to speakers. "Real time" vinyl playback switching, clicks, pops and all. Care to guess the results? ;-)

      Delete
  11. Dear Archimago,

    Thanks for your excellent essay and nice comment. Måns, thanks for the comment.

    Archimago, I think you just defined the exact test that I had loosely referred to. There are actually two levels. First, Analogue vs Digital test (all-out): all analogue best possible recording and master tape vs. best possible digital recording and digital file from the same performance. Which sounds better? Second, Analogue vs. Digital (home entertainment): best possible analogue recording pressed in vinyl vs. the same recording digitalized from master tapes and played from commercially available HD format. Which sounds better?

    I don´t think subjectivity is an issue here. At least the first test can be done in strictly scientific blind A/B/X long listening times fashion among professional musicians and hifi enthusiasts. With Måns´twist maybe even the second test can be done in strict scientific blind fashion.

    Where I am coming from is interest in whether MQA finally solved, or not, the biggest Hifi problem of today. Is it technically and commercially possible to have practically all the recorded music streamed to your home through a nice library program with great sound quality that is indistinguishable or better than vinyl and commercially available downloadable digital HD formats.

    I have listened to some of the best available MQA streamers and so far, Linn streamer (without MQA), sounded better. So which streamer should I buy (below astronomical price range)?

    Yours

    Eero

    ReplyDelete
    Replies
    1. "Where I am coming from is interest in whether MQA finally solved, or not, the biggest Hifi problem of today."

      No, absolutely not. It's just believer nonsense with 2 channels. The biggest problem with "HiFi" has been known since the advent of stereo. It IS stereo. Not "blurring" or any of the other bogeymen associated with digital, meant to frighten adolescents.
      MQA is just catering to old deaf guys with a mountain of 2ch audio jewelry in their rooms, looking desperately for any way to "improve" their "Hifi", but too dense to comprehend what's missing. About 90% of the Soundfield!
      Even the patriarch of audiophiledom knew this long ago https://www.stereophile.com/asweseeit/1107awsi/index.html
      MQA is just more 2ch idiocy that does zero to address any real issue, though it surely caters to the imaginary ones that have plague audiophiles, since the advent of "digital".

      Delete
    2. "the exact test that I had loosely referred to. There are actually two levels. First, Analogue vs Digital test (all-out): all analogue best possible recording and master tape vs. best possible digital recording and digital file from the same performance. Which sounds better?"

      Interesting question. If it is a test of accuracy then there is no debate in audio science, the all digital path will exceed analog pathways (ie lossless digital at least 16/44 just to put it beyond any dispute). That is the objective test which can be verified with a null signal test.

      Subjectively though, the question is do listeners prefer accuracy to source or a distorted euphonic colour? I prefer accuracy but sometimes a bit of euphonic colour can take the edge of a bad recording.

      But back to your question. Subjective listening tests have been done between all analog and all digital recordings of the same live event like a concert. The best known is the Geringer, J., Dunnigan, P. "Listener Preferences and Perception of Digital versus Analog Live Concert Recordings." in the Bulletin of the Council for Research in Music Education. 1 Jul. 2000, Number 145: 1-13.

      It was a proper double blind controlled test, ie digital and analog recordings of the same concert performance, recorded unequalized and unmixed for the test:

      "Results showed that music major listeners rated the digital versions of live concert recordings higher in quality than corresponding analog versions. Participants gave significantly higher ratings to the digital presentations in bass, treble, and overall quality, as well as separation of the instruments/voices."

      Note that even though the majority preferred the digital recording there was still a minority that preferred the analog recording. This is expected given the complex factors that make up our hearing perceptions and subjective preference.

      There are not many other peer review papers on digital vs analog comparisons simply because objectively, digital is always going to be more accurate to source so these tests are a waste of time (much the same reason why we don't do tests of viewer preferences of analog video vs digital video - no doubt a well produced analog film could look better than a poorly produced digital film which has nothing to do with the format). A preference for analog is therefore a preference away from accuracy and if that inaccuracy is wanted for any reason, it could be duplicated digitally - eg a digital copy of a record through a quality needle drop will sound identical to the record.

      Delete
    3. Dear Prep 74,

      Thank you for solid arguments and references. I assume that people prefer the real music performance to digital or analog reproductions of it. If this is the case then accurate reproduction of the real performance (not test signal) is the key. Let´s suppose that most people prefer digital to analog reproduction. Then the question becomes which digital format sounds best. Some results seem to suggest that people cannot really hear the differences between CD quality, DSD or HD PCM . Are there any scientific test on MQA vs non-MQA versions of the same master tapes?

      Personally I am wondering why after all these fine points many modern digital recordings through high quality playback systems still often sound harsh and unnatural. Is the real problem in over processing or leaving out 3-dimensional sound fields, or what?

      Yours

      Eero

      Delete
    4. Eero

      All recorded music is really just an illusion of the real event. Objective measurements, and various controlled tests, demonstrate that digital capture, storage and playback of recordings exceed human limits when the recording format is at least 16/44. That is the format itself is transparent to the source regardless whether it is CD, DSD or Hi Res.

      However, getting back to the illusion bit, no recording will sound like the original live event. Mikes are not ears and speakers are not instruments. Having said that, playing a top notch live recording that is well mastered through a transparent playback chain with reference grade speakers in a proper acoustically treated room is about the best we can get to a live performance.

      That is the gist I get out of the various test Archimago conducts, for good sound focus on the quality of the recording (try to find the best masterings of any particular album), focus on your speakers and room environment. Putting emphasis on whether MQA or hi res makes a difference without regards to its provenance is a waste of time and effort.

      Delete
    5. Hi Prep 74,

      Good point. I agree that speakers, mastering and room acoustics are really important in creating an illusion of a music performance.

      My hifi system is still missing a high quality streamer. If MQA or HD formats are not important issues, what do you think are the important criteria in looking for a good sounding streamer that can be plugged into an existing stereo system?

      Yours

      Eero

      Delete
    6. Eero
      There are many great streamers available these days. I'm not sure though whether you mean streamers for streaming services such as Tidal etc or streaming from your own files via a PC or a NAS. The only two that I have personal experience with is my Naim NDX and Bluesound Node. Both sound great, are plug and go, have easy to use interfaces and rock solid wireless connections. The Node streams all the paid (and unpaid) sites but I only use the streamers for my files stored on a NAS.

      The Naim is worth over $5k and the Node about $600. But here is the thing about how technology has brought the price of quality playback down, comparing the two side by side using their own internal DACs, the Naim does not really sound much better than the Node - more of a shade of grey than night and day. In fact on well mastered music I doubt anyone can pick the difference between the two. The Naim seems to perform better with low bit rate files but even so, that could just be my imagination looking at the impressively built unit.

      With my limited exposure to streamers I still would recommend looking into the Bluesound for streaming commercial services and your own music collection. It also does MQA.

      Btw, and MQA or Hi Res file of any particular can sound better than 16/44 as the hi res source often (but certainly not always) has a better master.

      Delete
  12. Dear Ammar,

    Thank you for your exciting comment and reference. We are getting into even higher level. Let´s add to the standard 2-channel analog vs. digital comparison a new challenger. All-out level: which technological system can best reproduce a live acoustic concert in the same concert hall. How close can we get to the real performance if you listen there blinded, not knowing whether it is orchestra or playback? Home entertainment: Does there exist a commercially available multi-channel music source, recording and playback system that sounds better than the best vinyl and digital 2ch competitors.

    In other words, 1) what is the state-of-the-art audio reproduction now ? 2) what is best home entertainment system and material that you can buy now?

    Yours

    Eero

    PS By the way I listened over 40 years ago "Ortoperspekta" a multi-channel system designed by Tapio M. Köykkä which played 3 channels: L+R, L-R, R-L. It sounded really good and had a nice sense of space.

    ReplyDelete
    Replies
    1. Btw, I should make very clear that I am a loudspeaker manufacturer, so beware that my opinions may/will contain biases/impartiality as such.

      "How close can we get to the real performance if you listen there blinded, not knowing whether it is orchestra or playback?"

      Well, that's a still ongoing process with MCH. There are systems with a whole bunch of channels (20+). There are also those who argue that many creates problems (like spatial aliasing etc.) that makes it not quite "real" and suggest less channels, even stereo!
      I'm somewhere in between. 4 minimum for envelopment, 5 minimum for "ideal", with proper encoding method...which require 7 minimum channels of input.
      http://www.aes.org/e-lib/browse.cfm?elib=9136 (membership required for full paper)
      Membership not required http://www.onhifi.com/features/20010615.htm
      That is about as close to "virtual reality" as I know.

      "In other words, 1) what is the state-of-the-art audio reproduction now ? 2) what is best home entertainment system and material that you can buy now?"
      This is a different question, because unfortunately, the best (PSR), is not available for consumers.
      My guess would be something like Auro3D. If instead of nonsense like MQA, we were given PSR recordings, that would be a true advance.
      But as Gordon Holt noted way back, the market for such is shrinking rapidly..and don't demand it in the first place.

      Delete
    2. Dear Ammar,

      Thanks for nice references and for insightful look at the state-of-the-art. Despair not. We consumers might be erratic but we will get there. As home tv´s, future hologram systems, car virtual displays, virtual reality etc. are becoming 3-dimensional and of higher visual quality, audio systems have to follow. I am convinced that true surround sound will become a natural part of our homes, offices, movie theaters and cars. When and how? Leading there, I think well-organized tests of most realistic and/or awesome high tech audio systems would be interesting reading for hifi, home entertainment and car enthusiasts. Someone will make money on these new high fidelity systems for sure.

      Then back to current topic of Archimago´s MQA musings. Let´s start from a consumer perspective on audio interface. I think "ITunes/Spotify/Tidal" type exciting interface with gigantic visual and audio library is state-of-the-art in delivery channels. What, if any, in your opinion is the best way to enjoy this new world of streaming. Is it ITunes/Spotify and damn the sound quality? Or is it Tidal with super streamers? Is Tidal experience better with top quality MQA or non-MQA equipment? Or do you feel that streaming is now of so low quality that it is better to stick to your home vinyl/HD files library?

      Yours

      Eero

      Delete
  13. Archimago,

    I agree with your conclusions.
    MQA does not provide any benefit to the music lover.
    And that's why they get so hysterical. And all their shills and sock puppets come out and try to de-rail discussions with off-topic and provocation.

    ReplyDelete
  14. Excellent blog!

    I’m an audio engineer/music producer (for context). I don’t work on commercial level recording, but I have been doing it professionally a long time. I’ve been loosely following this MQA debate, and the level of detachment from reality is surreal. The only hope for it as a business proposition is to hype it mercilessly.

    What makes it ridiculous, is that the notion of “high fidelity reproduction” only makes sense for the very limited numbers of recordings that actually have an acoustic source, (and haven't been subjected to heavy post processing). Such records have such a tiny economic significance, that it’s probably not measurable.

    The point is, for the vast, vast majority of records, there is nothing represent with fidelity.

    The closest thing, conceptually, would be to try and develop audio production systems that could recreate the sonic environment that the record was produced in. The MQA marketing has made some noise about this. It is literally impossible.

    On a practical level, what mixers and mastering engineers try to do is shape the recording so that it “translates” as well as possible across the infinite potential playback systems the recording will be listened to in. This is mind boggling difficult, but many recordings do succeed. The playback environments found in the “real world” today are mostly awful. My step kids listen to music on their phones, the actual speakers! There are earbuds, Bluetooth speakers, home systems with with all the time DSP on, car systems with massive subwoofers. It’s crazy.

    This playback environment is already shaping the recorded product! You can't "back it out" of the production.

    On a practical level, while working on a recording, ultimately all you can do is make it sound “good” on the system you are working on (the particular monitoring system of the studio) and pay attention to issues that affect the ability of a kid to “translate.”

    Even in an acoustic genre, like classical, the sonic outcome will be heavily shaped by the studio monitoring system.

    When it comes to modern pop music, we are far afield of the concept of audio “reproduction.” Many of the sounds you hear will actually be first generation! Meaning the only ever existed as numbers in the transmission chain to your particular system, where they are finally converted to actual sound.

    Pop music also full of weird distortion from the dsp. There’s a decent chance that one of the sources might have been an mp3!

    So even if MQA offered some tiny “improvement “ in reproduction, it has no real world impact at all.

    What an audio producer is really creating is a string of data. My preference is that the streaming systems would simply deliver the actual data that the producer and artist signed off on. In the meantime, I mostly listen to Spotify premium, and it sounds OK.

    I love your blog, some of the best critical writing on audio I’ve come across!

    ReplyDelete
  15. Dear b1daly and Archimago,

    Thank you for very interesting blog discussion. Let´s think about acoustic, or at least real music that is also played in concerts - not completely artificial sounds. Are you in essence saying that 1) streamed music is so bad anyway that it does not matter whether you use non-MQA or MQA system. And the only true hifi sources are top quality vinyls/HD files ? Or are you arguing that 2) Streaming is also a high quality source material where top quality non-MQA system sounds as good as MQA system? Or are you in fact arguing, as Linn seems to do, that top quality non-MQA system is sonically better than top quality MQA system which distorts the sound?

    Yours

    Eero

    ReplyDelete
  16. My main point is: standard CD resolution audio is more than adequate for the vast majority of delivery options.

    The small amount of hypothetical "improvement" in sound from HD formats is dwarfed by actual quality of the production. I haven't heard an MQA playback, but from what I've read, whatever potential improvement it might make (or degradation) will be totally irrelevant to the real world listener.

    The vast majority of playback systems, and playback environments, have far more egregious problems than these fine details. For example So many playback systems now have excessive EQ built into their default settings. Never mind the sub-optimal listening environments.

    As a producer, you need to take into account this multitude of playback environments. As a simple example, if you have excess bass in your mix, it might sound just fine, great even, on your studio monitors. But on a cheap system, with artificial bass boost, it will overload the bass, and sound horrible.

    Dynamic range compression is generally a must, as most real world playback systems, will not successfully represent a recording with a large dynamic range.

    The bottom line is that the "quality" of the end product a user hears is the result of a huge amount of work, skill, talent, and artistic vision. Along with the inherent difficulty of creating a "good" recording, compromises have to be made to account for real world playback systems.

    I do think there are fine details in the issue of playback formats, bitrates, and resolution, that affect the quality of sound. If you truly are a "golden ears" audiophile, and you listen to acoustic music, recorded with minimal processing, then such fine details might have an impact on your listening experience.

    I really like Spotify, and the like, but if I had my way they would simply stream CD resolution audio. I want to get as close as I can the product the producer created, and that is the digital stream. As a producer, I would like to have confidence that the file I created will be transmitted faithfully to the listener's system. Unfortunately, many online systems are putting out music as low bitrate, low quality encoded music, and sound pretty bad.

    All of a sudden, some company wants to jump past CD audio quality, to some hypothetical hi-res system, and charge a bunch of money for it! No thanks.

    The placebo effect is very powerful!


    ReplyDelete
  17. Hi b1daly,

    So it seems the jury is still out finding whether MQA is or is not a significant improvement in sound. I agree it would be great to first have streaming in full original CD level sound. Isn´t Tidal actually doing this already or are the hifi quality Tidal steams not equal to original CDs?

    Yours

    Eero

    ReplyDelete
    Replies
    1. FWIW, I believe the non-MQA Tidal content is redbook streamed as FLAC, i.e. compressed but not lossy. I've heard people claim that they can hear a difference between FLAC and WAV, and even between different levels of FLAC compression. I certainly can't. In fact, I cannot reliably hear the difference between high-bitrate MP3 and redbook, although I am 56 and can't hear anything above 14k if my life depended on it.

      Delete
  18. Hello,

    to summarize the discussion:
    The jury has found out that MQA offers no audible, not to speak of 'significant' improvements.
    In fact, MQA is a lossy compression format.

    In mathematical theory, MQA may or may not be a clever invention.
    But, technologic theory aside, the listening tests revealed that MQA is just no sonic improvement.

    And after all, we listen to the music with our ears, not with MQA's marketing brochures. Just let your ears decide.

    ReplyDelete
  19. Hi Tim,

    Based on my own listening tests I am not quite sure. Do you have some references on blind, well-organized MQA vs non-MQA listening tests?

    Yours

    Eero

    ReplyDelete
  20. This comment has been removed by the author.

    ReplyDelete
  21. Today I got this article via Facebook that might be of interest for the "non-believers": http://fairhedon.com/2017/11/05/an-interview-with-mastering-engineer-brian-lucey/

    ReplyDelete