[Ableza]

  noise shaping/dither achieves infinite quantization and temporal resolution

Well heck, divide by zero and you get infinite, right??

 

[earnmyturns]

  Well heck, divide by zero and you get infinite, right??

Not sure which smiley to reply with... BTW, the dither Wikipedia page is really good, covers both audio and imaging. Wikipedia is my friend when I am trying to recover from memory all this stuff that I've not looked at in ages.

 

[Chris J]

[COLOR=0D0D0D]2016, Chapter 5:[/COLOR]
[COLOR=0D0D0D]The Subjectivist/Objectivist Synthesis[/COLOR]

[COLOR=0D0D0D]Color me stupid, but I’m going to wade right into the subjectivist vs objectivist debate, and see if both groups can find a happy place.[/COLOR]

[COLOR=0D0D0D]In the process, I’m going to (attempt to) distill the objectivist/subjectivist debate into a few lines, give some examples and anecdotes (yes, I know, not data) from my experience, call out some interesting factoids to think about, outline (what I see to be) the stuff that both objectivists and subjectivists get right and wrong, and attempt to find a synthesis in the end.[/COLOR]

[COLOR=0D0D0D]Now, don’t get me wrong. I don’t know where this is going to end. Like most of the stuff I write for this book, this chapter is written more as a stream of consciousness, usually over a single evening, or maybe two. Don’t expect a scientific paper. It won’t be endlessly footnoted and cited. It will call on you to do some Googling if you want to dig deeper into the stuff I bring up.[/COLOR]

[COLOR=0D0D0D]Spoiler alert:[/COLOR] [COLOR=0D0D0D]both subjectivists and objectivists are wrong on lots of things, at least in my opinion. Both camps have members that get way, way, way too into their dogma, both camps have advocates that really need to chill out and have a beer (because, let’s face it, the best mod that improves musical enjoyment is probably alcohol…it will almost certainly beat the pants off of any car-priced DAC), and both camps need to realize that, even if one camp “wins,” the world will spin on, despite any histrionics.[/COLOR]

[COLOR=0D0D0D]Yes, I know I am completely insane. And yes, I know this may not end well.[/COLOR]

[COLOR=0D0D0D]Ready? Let’s get started.[/COLOR]


[COLOR=0D0D0D]The Subjectivist/Objectivist Debate[/COLOR]

[COLOR=0D0D0D]If you’ve never experienced a subjectivist/objectivist debate, you must be new to the internet. Sorry about your sanity. Here’s how pretty much every objectivist/subjectivist debate goes, in 12 lines or so:[/COLOR]

[COLOR=0D0D0D]Subjectivist:[/COLOR] [COLOR=0D0D0D]“I think my new Arglebargle X1000 sounds way better than the Craphound PST-1.”[/COLOR]
[COLOR=0D0D0D]Objectivist:[/COLOR] [COLOR=0D0D0D]“No, if they both measure 20-20K flat, have THD below 0.1%, and have a low output impedance, they have to sound the same.”[/COLOR]
[COLOR=0D0D0D]Subjectivist:[/COLOR] [COLOR=0D0D0D]“I think my experience trumps your measurements.”[/COLOR]
[COLOR=0D0D0D]Objectivist:[/COLOR] [COLOR=0D0D0D]“No, humans can’t perceive anything beyond that, see (insert links to tests here.)”[/COLOR]
[COLOR=0D0D0D]Subjectivist:[/COLOR] [COLOR=0D0D0D]“Well, I hear a difference and so does (insert anecdotes about friends, spouses, dogs, fish, etc).”[/COLOR]
[COLOR=0D0D0D]Objectivist:[/COLOR] [COLOR=0D0D0D]“Anecdotes aren’t data! You’re fooling yourself. (Insert words about scientific method and significant results here.)”[/COLOR]
[COLOR=0D0D0D]Subjectivist:[/COLOR] [COLOR=0D0D0D](Sigh.) “Just leave me alone to enjoy my Arglebargle with the other folks I’m talking to here.”[/COLOR]
[COLOR=0D0D0D]Objectivist:[/COLOR] [COLOR=0D0D0D]“No! Don’t you see you’re being taken advantage of by evil companies selling overpriced gear?”[/COLOR]
[COLOR=0D0D0D]Subjectivist:[/COLOR] [COLOR=0D0D0D]“You probably just can’t afford good gear!”[/COLOR]
[COLOR=0D0D0D]Objectivist:[/COLOR] [COLOR=0D0D0D]“You’re nothing but a shill for the man!”[/COLOR]
[COLOR=0D0D0D]Subjectivist:[/COLOR] [COLOR=0D0D0D]“Ad hominem!”[/COLOR]
[COLOR=0D0D0D]Objectivist:[/COLOR] [COLOR=0D0D0D]“Ad hominem!”[/COLOR]

[COLOR=0D0D0D]And then you repeat the last two lines, ad infinitum, ad nauseum.[/COLOR]

[COLOR=0D0D0D]Even in this oversimplified distillation, we end up with some really interesting questions. Why does this particular objectivist* feel he has to insert test data as the ultimate arbiter of human perception into a discussion that was simply among excited audio-gear owners? Why does this particular subjectivist* reject the idea that measurements can offer anything of value? Why can’t the two groups simply ignore each other and both live happy, but separate, lives? Does the objectivist think they are “saving” the subjectivist from exploitation, or is the motive more inward-focused? Does the subjectivist really honest in his perceptions, or are they colored by the cost of the component? How did the objectivist come up with their measurement thresholds that certify transparency? What mood was the subjectivist in when they made their assertion? What experience does the objectivist have with audio test equipment? Did the subjectivist try the product in the same exact location as their own equipment and use the same exact music they are familiar with?[/COLOR]



[rule][COLOR=0D0D0D]*Not all objectivists feel the need to insert themselves into subjective conversations, just as not all subjectivists reject measurements. Groups are always a continuum of individuals. As I said, this is an oversimplification.[/COLOR]

[rule]

[COLOR=0D0D0D]Yeah. It gets murky, real fast. And it gets even murkier when you really start looking at both sides. Here are some factoids to consider. (And yes, thank you, I know that the plural of “anecdote” is not “data”—I believe I’ve already mentioned this.)[/COLOR]

[COLOR=0D0D0D]Consider:[/COLOR]

1. [COLOR=0D0D0D]Every engineer I personally know who is working in audio is a subjectivist—they think that stuff that measures good sounds different, without exception. Just protecting their jobs? Perhaps. But many of these same engineers started like me, as an objectivist…until they had a subjectivist experience or two that changed their minds. And yes, I know, I don’t know every engineer on the planet.[/COLOR]
2. [COLOR=0D0D0D]It seems that many of the most fervent objectivists aren’t working as engineers, or aren’t engineers working in audio, and most don’t seem to have any experience with the measurement gear they cite measurements from. Again, I know there are exceptions to this rule. Maybe their fervor is simply Robin Hood syndrome—trying to help the poor deluded subjectivists? [/COLOR]
3. [COLOR=0D0D0D]Most of the subjectivists I’ve met are unwilling or unable to factor in their own rationalization when proselytizing for gear, when the reality is we all rationalize what we have chosen as the best—and the more we’ve invested, the more powerful the urge to rationalize. Nobody wants to look like a fool who spent car money on audio gear that makes no difference. [/COLOR]
4. [COLOR=0D0D0D]Some objectivist thought leaders—the kind that have written books—don’t portray their findings in a very (ahem) objectivist manner. Instead, they use emotionally charged language, actively demean subjectivists, and are dismissive of any approach that doesn’t align with the path that they have codified. This shrill, “die-heretic-die!” tone is not exactly conducive to rational debate…and (in my opinion), it doesn’t seem to indicate security in their belief systems.[/COLOR]

[COLOR=0D0D0D]“Ah, you’re not being fair,” some will say. “You’re painting with a broad brush, constructing a straw man, I am gonna haul out all my captain-of-the-debate-team tricks on you.”[/COLOR]

[COLOR=0D0D0D]Yes. Which is why I said this is my experience, and that it’s not absolute. But if it’s fair for an objectivist to ask a subjectivist if they’ve done a blind comparison of their gear when they anoint the Arglebargle as the greatest and bestest thing on the planet, it’s fair for a subjectivist to ask an objectivist how they determined the measured limits of human hearing, and their experience with making said measurements on an Audio Precision, Stanford, or dScope.[/COLOR]

[COLOR=0D0D0D](And—editorializing more heavily here—it’s entirely fair for both sides to ask each other to lay off and have a drink instead, because these endless debates become tiresome in a big hurry, IMO.)[/COLOR]

[COLOR=0D0D0D]Now, I can see the attraction of both sides. Subjectivity seems to be the more holistic, natural, touchy-feely way of finding the best gear for you—never mind understanding the measurements, how does it sound? Objectivity seems to be the most rational and comforting—no need to spend megabucks, just check the measurements.[/COLOR]

[COLOR=0D0D0D]But subjectivity can drive an unnatural mania for more and more gear, while objectivity is the best tool for companies trying to make a quick buck—if it measures well, get it done, push it out, don’t worry about the sound.[/COLOR]

[COLOR=0D0D0D]So how do you reconcile the two? Is there a middle ground?[/COLOR]

[COLOR=0D0D0D]Well, for me, there’s been nothing but middle ground, at least since the early days at Sumo. Once I realized things really did sound different, I accepted that measurements can’t tell the whole story. But I didn’t throw out measurements. Nor did I throw out the subjective experience. And, from time to time, I’ve tried to correlate measurements with sound quality, or come up with measurements that relate more closely with perceived quality.[/COLOR]

[COLOR=0D0D0D]“Enough about you,” some might be saying. “Is there a middle ground for me?”[/COLOR]

[COLOR=0D0D0D]I don’t know. Only you can answer that. But I hope both sides can learn something from the other. To get to this synthesis, let’s look at what (I think) each gets right and wrong.[/COLOR]


[COLOR=0D0D0D]The Right and Wrong of the Subjective and Objective Approaches (IMO)[/COLOR]

[COLOR=0D0D0D]If you’re closed-minded, you may want to skip the rest of this chapter. Because, as I’ve mentioned before, everything is a continuum. There is no black and white. Objectivists are not “good.” Subjectivists are not “evil.” Both sides have some very good points—at least in my opinion.[/COLOR]

[COLOR=0D0D0D]What most objectivists get right:[/COLOR]

• [COLOR=0D0D0D]They provide a foundation to work from—objective measurements will tell you whether your gear is broken, performing to spec, or you’re fooling yourself[/COLOR]
• [COLOR=0D0D0D]They provide a leveling effect, in that inexpensive gear can measure and perform similarly to very expensive gear in some cases—this keeps ALL of audio from ascending into gold-plated Bentleyphilia[/COLOR]
• [COLOR=0D0D0D]They are absolutely right in that some aspects of performance should be measurable, and that these measurements should be done[/COLOR]
• [COLOR=0D0D0D]They hold manufacturers to certain standards and help prevent true stinkers[/COLOR]
• [COLOR=0D0D0D]They provide some common-sense rules on how to get better performance from your products, such as impedance matching, power requirements for specific output levels, etc.[/COLOR]

[COLOR=0D0D0D]What some objectivists get wrong:[/COLOR]

• [COLOR=0D0D0D]Instantly dismissive of everything except measurements, leading to endless arguments over meaningless specsmanship, like whether or not 0.0007% THD is worse than 0.0005%[/COLOR]
• [COLOR=0D0D0D]A dogma-based assertion that performance above or below certain standards (frequency response, THD, noise, output impedance, etc) assures complete transparency—and that all transparent gear sounds the same[/COLOR]
• [COLOR=0D0D0D]A dogma-based assertion that all aspects of human hearing are known[/COLOR]
• [COLOR=0D0D0D]A dogma-based assertion that we are measuring everything we need to measure in order to fully characterize an audio system[/COLOR]
• [COLOR=0D0D0D]Blind faith in blind testing and the results of flawed studies using blind testing[/COLOR]
• [COLOR=0D0D0D]A seeming need to preach to audiophiles uninterested in engaging with them, coupled with an inability to realize when to say, “when”[/COLOR]

[COLOR=0D0D0D]What most subjectivists get right:[/COLOR]

• [COLOR=0D0D0D]They are open to new ideas, like the idea that there may be something more than what we can measure[/COLOR]
• [COLOR=0D0D0D]They (usually) don’t feel the need to go out and convert audio objectivists into subjectivists[/COLOR]
• [COLOR=0D0D0D]They tend to talk more amongst themselves, compare more gear, experiment with different kinds of gear, tweak gear, and in general push the limits[/COLOR]
• [COLOR=0D0D0D]They encourage manufacturers to continue to improve their own products, both relative to their current line and with respect to the competition[/COLOR]
• [COLOR=0D0D0D]They are usually more willing to take a chance on new and unproven companies and technologies, which can help drive innovation[/COLOR]

[COLOR=0D0D0D]What some subjectivists get wrong:[/COLOR]

• [COLOR=0D0D0D]They blow subjective differences way out of proportion, even when the actual differences are tiny[/COLOR]
• [COLOR=0D0D0D]They don’t take into account their mood/feelings/physical condition/intoxication when passing judgment on gear (like, if you get something in a crushed box with a broken knob on a day when you’ve just gotten passed over for promotion, versus a day when you have a wonderful first date, discovered a nice bonus in your checking account, had a wonderful dinner, and a few scotches…)[/COLOR]
• [COLOR=0D0D0D]They frequently make snap judgments based on brief listening at meets/friends houses/with different gear/under different circumstances, and these snap judgments can endure for months or years[/COLOR]
• [COLOR=0D0D0D]They are subject to rationalizing their purchases, especially when those purchases are very costly[/COLOR]
• [COLOR=0D0D0D]They can get obsessive about gear and go crazy spending, spending, and spending even more, trying to find that “last 1%” of performance[/COLOR]

[COLOR=0D0D0D]Again, if you’re feeling a bit in a huff right now, go back and read those “some” and “most” disclaimers on the summary headings. And consider that there are some who place themselves firmly in the subjectivist and objectivist camps. I’ve said many times before that I would never want to design audio gear without extensive testing (see the test and measurement chapter for more info on that one—by far the longest chapter to date, by the way), yet at the same time, I wouldn’t want to have the blind pursuit of measurements be the sole arbiter of my designs. I believe that the different stuff I make sounds different.[/COLOR]

[COLOR=0D0D0D]Yet, at the same time, I try not to blow these differences out of proportion. Will any of our amps fundamentally transform your headphones into something completely different? No. If you don’t like your headphones, you should be seeking new headphones long before an amp or DAC.[/COLOR]

[COLOR=0D0D0D]And I live by my beliefs that our products sound different, as well. My “main stack” is Mjolnir 2/Gungnir Multibit, not Ragnarok/Yggdrasil. Mjolnir 2 is a warmer, wetter, “happier” amp than Ragnarok, and Gungnir Multibit is also a bit more euphonic than Yggy. The combo may not be the absolute ultimate word in resolution and transparency, but I like the way it sounds better than our top stack.[/COLOR]

[COLOR=0D0D0D]However, note this is a subjective choice…and also note that I’d still be happy with Ragnarok/Yggdrasil. The differences are, as I mentioned, relatively subtle. [/COLOR]

[COLOR=0D0D0D]So don’t get upset if you don’t fit the broad categories above. I’m just listing what I noticed off the top of my head. It isn’t meant to be comprehensive, or definitive.[/COLOR]


[COLOR=0D0D0D]Bad Data All Around[/COLOR]

[COLOR=0D0D0D]“Wait a sec,” some of the sharper-eyed readers are saying. “I noticed you cited ‘flawed studies’ up in the list of objective wrongs. What’s up with that?”[/COLOR]

[COLOR=0D0D0D]Okay. Deep breath. Let’s get one thing clear from the start: I don’t think either the objectivists or subjectivists have definitive, clear-cut proof positive of the limits or capabilities of human hearing.[/COLOR]

[COLOR=0D0D0D]The reality is that there’s not a lot of money in proving audibility, one way or another. Audio isn’t another $100B drug, nor a breakthrough in cheap and clean energy, or even a new cleaning product that could break $100M in sales every year. (And this is an important fact to keep in mind.)[/COLOR]

[COLOR=0D0D0D]As a result, the studies I’ve seen are relatively fringe, ad-hoc, limited, flawed, or all of the above. The most frequently cited, Meyer and Moran, used sources that were not actually high resolution, and is directly countered by Bob Stuart’s paper on the audibility of digital filters. At the same time, there is physiological evidence that the brain responds differently to sounds containing inaudible supersonic components (google “Inaudible High-Frequency Sounds Affect Brain Activity: Hypersonic Effect”) and there is significant evidence that experienced listeners can hear the differences in amplifiers (but not DACs) in Innerfidelity’s recent top-end gear shootout.) Further confounding things are unproven assertions like our own Mike Moffat’s statement that he believes hearing is integrative, rather than differential.[/COLOR]

[COLOR=0D0D0D]Like I said before, I’m not going to go into minute detail in citing every paper, nor in picking them apart. If you’re more interested in the subject, I suggest you start Googling, reading papers, refreshing yourself on what a statistically significant result is, and coming to your own conclusions.[/COLOR]

[COLOR=0D0D0D]The point I want to make is: there isn’t any study large enough, perfect enough, and significant enough to end all doubt on the subject of what is, and what isn’t, audible. Nor will there likely be one in the near future. There’s simply no payoff in it.[/COLOR]

[COLOR=0D0D0D](And, come on, admit it: even if there was an ideal study, a certain tinfoil-hat crowd would claim that there is some giant deceitful conspiracy behind it all!)[/COLOR]

[COLOR=0D0D0D]However, from reading many papers, viewing the results of many tests, hearing many anecdotes, observing my own reactions and the reactions of others, I’m willing to bet on what a potential “perfect” test might find.[/COLOR]

[COLOR=0D0D0D]Here’s my bet on a one-line synopsis:[/COLOR]

[COLOR=0D0D0D]There are some people who can discern subtle differences below the commonly accepted limits of human hearing, and some of this group of people find those differences meaningful enough to care about.[/COLOR]

[COLOR=0D0D0D]“So you’re saying that if I can’t hear the difference, I’m a tin-ear?” someone is shouting. “Well, I never! What an insult! I’m gonna grump off into a corner!”[/COLOR]

[COLOR=0D0D0D]No, I didn’t say that at all. All I’m saying is that different people have different capabilities. You wouldn’t want me flying your plane, at least without glasses. Nor would I expect to pass a sommelier’s deductive tasting evaluation.[/COLOR]

[COLOR=0D0D0D]And (cue us never selling anything ever again), I don’t think my ears are the last word in golden-ness. There are plenty of times I think I’m fooling myself. I doubt what I hear.[/COLOR]

[COLOR=0D0D0D]But (big sigh of relief), that’s why I have other listeners I trust, both inside and outside of the company. Some of these listeners are much more discerning than I am…they have picked out tiny details that I’ve never heard, and called me out when I brought in “ringer” products for a listen.[/COLOR]

[COLOR=0D0D0D]“So why don’t you guys do a study?” someone is surely asking. “You can have the perfect data, you can confirm or disprove your hypothesis, you can end the arguments forever.”[/COLOR]

[COLOR=0D0D0D]Well, setting aside that I doubt we’d be able to do a study with the scope and rigor necessary to be considered definitive, you’re talking about:[/COLOR]

• [COLOR=0D0D0D]A heckuva lot of time[/COLOR]
• [COLOR=0D0D0D]A heckuva lot of money[/COLOR]
• [COLOR=0D0D0D]A heckuva diversion from what we’re doing[/COLOR]

[COLOR=0D0D0D]And the payoff would be…what? Half the audio world disbelieving and (probably) reviling us, no matter which way the study ended up?[/COLOR]

[COLOR=0D0D0D]“And maybe killing your own business, if nobody can tell the difference,” someone snidely remarks.[/COLOR]

[COLOR=0D0D0D]Actually, I don’t believe that. I think that even if one great study proved complete audio objectivity, the audiophile world would continue spinning, completely unaffected.[/COLOR]

[COLOR=0D0D0D]So, sorry. We’re not going to try to prove our hypothesis. We’re simply going to keep acting “as if.” Which seems to be exactly what most of you want.[/COLOR]


[COLOR=0D0D0D]So What Can Each Side Learn From the Other?[/COLOR]

[COLOR=0D0D0D]“Wow, you really went off into left field, didn’t you?” some are probably saying. “Is there any point to all of this blather?”[/COLOR]

[COLOR=0D0D0D]Well, like I said, this exploration might not go anywhere.[/COLOR]

[COLOR=0D0D0D]But maybe there is something here. If you zoom way out, you can kinda look at the subjectivist position as being the exploratory one—the drive that keeps us trying to make increasingly better-sounding gear. At the same time, the objectivist position as being the foundational one—it keeps us from going off into cloud-cuckooland as we explore the fringes of audio perception.[/COLOR]

[COLOR=0D0D0D]I think both sides have valuable insight. The subjectivists remind us that thinking “all is known” is not a good bet. Look at the guy who wanted to shut down the US Patent Office in the 1800s because “everything had already been invented.” The objectivists remind us that there are fundamental rules (like impedance matching) and that our quests into the unknown can be costly, frustrating diversions.[/COLOR]

[COLOR=0D0D0D]If I was a pure subjectivist, here’s what I’d take from the objectivist side:[/COLOR]

• [COLOR=0D0D0D]You have no reproduced audio without science and the scientific method, and it’s worth learning more about this[/COLOR]
• [COLOR=0D0D0D]Those “huge” differences others are talking about may not be so huge at all, or they might even be cognitive bias[/COLOR]
• [COLOR=0D0D0D]Spending big is not always the answer when it comes to great sound—being more discerning may keep both your ears and wallet happy[/COLOR]

[COLOR=0D0D0D]If I was a pure objectivist, here’s what I’d take from the subjectivist side:[/COLOR]

• [COLOR=0D0D0D]Science hasn’t fully characterized everything, whether you’re talking perception, medicine, physics—question your own hypotheses and be open to revising your position[/COLOR]
• [COLOR=0D0D0D]No difference in perception to you may be significant to someone else with different perception—and cognitive bias cuts both ways[/COLOR]
• [COLOR=0D0D0D]Spending big on audio is a largely harmless pastime that doesn’t affect you—let them be, and concentrate on something that makes you happy[/COLOR]

[COLOR=0D0D0D]And to both sides:[/COLOR]

[COLOR=0D0D0D]In a thousand years, when godlike AIs are unearthing the data-foundations of the Human Internet, they’ll be shaking their metaphorical heads at the silly stuff we get into arguments about. Audio isn’t a cure for cancer, it’s not a new physical particle, it’s not a hyperloop transport system. It’s a fun pastime that helps you enjoy music to its fullest.[/COLOR]

[COLOR=0D0D0D]Sit back. Relax. Buy each other beers. Laugh at yourselves. Try something new. Ask questions. Read and digest. Interact and learn. Because, let’s face it, if we’re really here to share, and if everyone really has something to contribute, we should all be reading a whole lot more than writing.[/COLOR]

[COLOR=0D0D0D]Happy objective subjectivizing, or subjective objectivizing...[/COLOR]

[COLOR=0D0D0D]…or simply sitting back, listening, and enjoying some great tunes.[/COLOR]

Sorry for the mega quote......
.........IMHO, this is the greatest post I have ever seen on Head Fi!
Or in any audiophile forum.
And, to be brutally honest, is the main reason why I do not like participating in the Sound-Sci Forum....with apologies to all the Sound-Sci- Forum participants.

 

[SearchOfSub]

Schiit book chapters are fun to read and informative as well. Was gonna a purchase the book awhile back but forgot.

 

[earnmyturns]

I feel like you will appreciate this: https://www.meridian-audio.com/meridian-uploads/ara/coding2.pdf

I'd love to hear your thoughts if you would be willing to share, your maths experience goes well beyond my own.

Pretty informative overall. A few over-the-top statements (like that "infinite" resolution claim that is quite misleading), and some potentially self-serving arguments based on 20-year old perception studies without more recent potentially contradictory citations, but a few interesting conclusions: 1) 96kHz/20bit LPCM is more than enough for human hearing, even without fancy dithering or noise shaping; 2) bit streaming (DSD) is never beneficial; 3) 44.1kHz/16bit without proper input dithering/noise shaping may fall a bit short; 4) the only reason for fancier coding schemes than LPCM is to reduce bit budget. It's also interesting that dithering in the DAC is never mentioned (unless the DAC is doing additional truncation, of course), which of course is the right way to think if you understand the rest of the story (if the source was not dithered, correlated noise is already baked in, so trying to fix it post hoc makes no sense). My conclusion is that this paper is excellent evidence against MQA for higher-end uses. It is only a bit-budget optimization that most of us don't care much about, given that much of the music we listen to is 44.1/16, and any new material at 96/24 has plenty of headroom. In other words, more evidence that Schiit's market analysis is right on the money. In my own blunt words: if someone has to worry about bit budget, they should not be spending their scarce resources on high-end DACs, let alone ones capable of handling multiple encoding schemes.

 

[yay101]

Pretty informative overall. A few over-the-top statements (like that "infinite" resolution claim that is quite misleading), and some potentially self-serving arguments based on 20-year old perception studies without more recent potentially contradictory citations, but a few interesting conclusions: 1) 96kHz/20bit LPCM is more than enough for human hearing, even without fancy dithering or noise shaping; 2) bit streaming (DSD) is never beneficial; 3) 44.1kHz/16bit without proper input dithering/noise shaping may fall a bit short; 4) the only reason for fancier coding schemes than LPCM is to reduce bit budget. It's also interesting that dithering in the DAC is never mentioned (unless the DAC is doing additional truncation, of course), which of course is the right way to think if you understand the rest of the story (if the source was not dithered, correlated noise is already baked in, so trying to fix it post hoc makes no sense). My conclusion is that this paper is excellent evidence against MQA for higher-end uses. It is only a bit-budget optimization that most of us don't care much about, given that much of the music we listen to is 44.1/16, and any new material at 96/24 has plenty of headroom. In other words, more evidence that Schiit's market analysis is right on the money. In my own blunt words: if someone has to worry about bit budget, they should not be spending their scarce resources on high-end DACs, let alone ones capable of handling multiple encoding schemes.

Thanks! I cringed at that too, had to read their sources just to get a laugh out of it. The reason I wanted another person's opinion was so that I could check my takeaway because this paper was written by the main engineer behind MQA, but at least at this technical level I don't feel like it's a super compelling product from their own paper.

Getting those doing the recording to stop taking part in loudness wars etc is another thing altogether.

 

[Ableza]

paper was written by the main engineer behind MQA,

My money's on he (also) has an MBA.

 

[yay101]

My money's on he (also) has an MBA.

Hahaha it explains so much when you add that.

 

[AndreYew]

  Thanks, started to read it, got my guard up soon enough with the claim that dither achieves infinite quantization and temporal resolution, but I'll persist through the end. 

It's infinite in the sense that if the dither is done properly (ie. TPDF), there are no correlated error moments, ie. the noise is totally unrelated to the signal. This means that if you can listen long enough to average out the noise, you can hear a signal that is as low a level as you like as long as the signal's bandwidth is less than the inverse of your listening window's length in time.

 

[jcx]

Yes the math is clear, of course the practical limits will still include finite wordlength arithmetic but nowadays is 32 or 64 bits
 
And yes the "infinite" doesn't really apply when the real world audio signals already have noise floors above most better studio ADC converting the signal from real microphones, or when played back by real electronics, thru real electroacoustic transducers in real rooms to limited noise floor human ears with perceptual 100 Hz+ (or ~20% of center frequency, whichever is greater) critical bands
 
For the digital audio dither theory: https://www.google.com/#q=wannamaker+dither is good for the EEs in the crowd
 
Less math heavy, with some practical listening demos at reduced bit depth:
 
http://audio.rightmark.org/lukin/dither/dither.htm
 
http://www.tnt-audio.com/sorgenti/dither_e.html
 
 
I have taken some heat from Mike for daring to discuss some dither theory with respect to DAC implementations when questions from other thread participants introduced the subject
 
 
I would love to give a critique of Jason's The Subjectivist/Objectivist Synthesis - its a fair attempt but still reads a bit biased to an "objectivist" with decades of EE experience, hobby amp builds, contributions to community projects, some reading in Psychoacoustics, occasional attendance at AES, Boston Audio Society meets, demos, studio tours...

 

[earnmyturns]

It's infinite in the sense that if the dither is done properly (ie. TPDF), there are no correlated error moments, ie. the noise is totally unrelated to the signal. This means that if you can listen long enough to average out the noise, you can hear a signal that is as low a level as you like as long as the signal's bandwidth is less than the inverse of your listening window's length in time.

Which is totally irrelevant for music, as the music signal is stationary long enough to allow the uncorrelated noise to be subtracted out. That's why claiming "infinite resolution" in this context seems disingenuous. Anyone interested in a spherical cow?

 

[earnmyturns]

   
I have taken some heat from Mike for daring to discuss some dither theory with respect to DAC implementations when questions from other thread participants introduced the subject

I don't see how dither is relevant to DACs, as opposed to ADCs. When applied at the source, dither replaces artifacts by uncorrelated noise. Once the signal is encoded, its bits are the most one can now about the original signal, and additional dither at the DAC would only increase the noise level. If the source was not dithered, tough, but the DAC has no way of "knowing" that, as any pattern of LSBs could be the result of just a bunch of unlikely coin tosses while dithering the source.

 

[Jason Stoddard]

  I don't see how dither is relevant to DACs, as opposed to ADCs. When applied at the source, dither replaces artifacts by uncorrelated noise. Once the signal is encoded, its bits are the most one can now about the original signal, and additional dither at the DAC would only increase the noise level. If the source was not dithered, tough, but the DAC has no way of "knowing" that, as any pattern of LSBs could be the result of just a bunch of unlikely coin tosses while dithering the source.

But...but...the resolution becomes infinite! We'll expand our perception into non-euclidian dimensions! Yog-Sothoth knows the gate. Yog-Sothoth is the gate. Yog-Sothoth is the key and guardian of the gate. Past, present, future, all are one in Yog-Sothoth!

 

[AndreYew]

Which is totally irrelevant for music, as the music signal is stationary long enough to allow the uncorrelated noise to be subtracted out. That's why claiming "infinite resolution" in this context seems disingenuous. Anyone interested in a spherical cow?

You can't subtract the noise out unless you know what it is exactly, and since it's random ... Instead you have to reduce noise with averaging since the uncorrelatedness will give you a 3dB reduction with each averaging. Is it useful for music? No idea, but if there is a narrow signal, and you (the human) listen hard enough, you can pick it out. It's about getting the technical stuff out of the way so you let the human hearing system use as much of its native abilities as possible.

 

[aamefford]

But...but...the resolution becomes infinite! We'll expand our perception into non-euclidian dimensions! Yog-Sothoth knows the gate. Yog-Sothoth is the gate. Yog-Sothoth is the key and guardian of the gate. Past, present, future, all are one in Yog-Sothoth!

Now I have to go read all of Lovecraft…