Head-Fi.org › Forums › Equipment Forums › Sound Science › Can someone explain audio resolution to me?
New Posts  All Forums:Forum Nav:

Can someone explain audio resolution to me? - Page 3

post #31 of 77
Quote:
Originally Posted by UltMusicSnob View Post


The point to bring on board is that 44.1 kHz digital data **also** undergoes changes in the path from bits to sound waves. It's not a question of with or without. 44.1 has trade-offs and multiple parameters just like any other digital file, when processed back to analog.

 

It's odd for me to use the term "distortion", when the effect in question could be "provides improved phase alignment after D/A, when compared to a lower sample rate". "Distortion" implies a departure from a preferred norm, but in this case the processing involved could be moving from "norm" to "new, more-preferred". I am suspending an assumption which seems to be made a lot, which is that 44.1 kHz already represents a quality cap ***in all respects***. The science (see the papers cited above) does not support that.

Not too sure where the references to the audibility of "phase alignment" (the correct term is group delay) is coming from, but from research, group delay becomes audible when it is greater than 1ms at 2KHz, which is the band greatest sensitivity in human hearing.  No current DAC filter, at any bit rate, has even 1/10th that much.  

 

Then, there's this:

https://secure.aes.org/forum/pubs/journal/?ID=2

 

If you don't want to buy the paper and read it, it basically shows that high-resolution files played through a 16/44.1 "bottleneck" were indistinguishable from the high resolution originals.  The paper would seem to refute the 16/44.1 "quality cap".  

 

Group delay caused by reconstruction filters is nothing like the dual-driver experiment used in the paper, "Audibility of temporal smearing and time misalignment of acoustic signals" by

Milind N. Kunchur, cited above.  In that test, two discrete wave-fronts were presented, skewed in time.  The test signal was an artificial, analog-generated 7KHz square wave with, essentially, infinite odd harmonic content, not musical signals with natural and normal spectral content.  In short, the paper is interesting, shows a quality of perception that bears further study, but doesn't relate at all to the mechanism of group delay in anti-aliasing or reconstruction filters.  It may be reasonably good science (don't know yet, I haven't finished the paper), but is the study of something that doesn't occur within a digital recording/reproducing system.

post #32 of 77
Quote:
Originally Posted by jaddie View Post

Not too sure where the references to the audibility of "phase alignment" (the correct term is group delay) is coming from, but from research, group delay becomes audible when it is greater than 1ms at 2KHz, which is the band greatest sensitivity in human hearing.  No current DAC filter, at any bit rate, has even 1/10th that much.  

 

Then, there's this:

https://secure.aes.org/forum/pubs/journal/?ID=2

 

If you don't want to buy the paper and read it, it basically shows that high-resolution files played through a 16/44.1 "bottleneck" were indistinguishable from the high resolution originals.  The paper would seem to refute the 16/44.1 "quality cap".  

 

Group delay caused by reconstruction filters is nothing like the dual-driver experiment used in the paper, "Audibility of temporal smearing and time misalignment of acoustic signals" by

Milind N. Kunchur, cited above.  In that test, two discrete wave-fronts were presented, skewed in time.  The test signal was an artificial, analog-generated 7KHz square wave with, essentially, infinite odd harmonic content, not musical signals with natural and normal spectral content.  In short, the paper is interesting, shows a quality of perception that bears further study, but doesn't relate at all to the mechanism of group delay in anti-aliasing or reconstruction filters.  It may be reasonably good science (don't know yet, I haven't finished the paper), but is the study of something that doesn't occur within a digital recording/reproducing system.


Thank you for the reply and comments! These papers are put forward for my purposes not in order to establish their very specific claims per se, but to demonstrate that scientific understanding of all these factors is ongoing and far from completed. Interesting work on phase and human physiological detection/discrimination specifically is available here: http://www.physics.sc.edu/~kunchur/papers/Temporal-resolution-by-bandwidth-restriction--Kunchur.pdf

post #33 of 77

And of course, as further demonstration of the ongoing nature of research on these questions, this more recent paper is also on-topic: http://www.aes.org/e-lib/browse.cfm?elib=15398

Signal Detection Theory has a lot of light to shed on the ABX problems of testing. http://lsbaudio.com/publications/AES127_ABX.pdf In particular, the logic of ABX testing does not provide the ability to conclude that no perceptible difference exists; it can only prove that a difference does exist, or show that no difference was detected in a particular set of experiments (the 'black swan' problem).

post #34 of 77
Quote:
Originally Posted by UltMusicSnob View Post

In particular, the logic of ABX testing does not provide the ability to conclude that no perceptible difference exists; it can only prove that a difference does exist, or show that no difference was detected in a particular set of experiments (the 'black swan' problem).

 

Likewise, it doesn't prove that a difference is an improvement. Only that it is different.

post #35 of 77
Quote:
Originally Posted by jaddie View Post

Then, there's this:

https://secure.aes.org/forum/pubs/journal/?ID=2

 

If you don't want to buy the paper and read it, it basically shows that high-resolution files played through a 16/44.1 "bottleneck" were indistinguishable from the high resolution originals.  The paper would seem to refute the 16/44.1 "quality cap".  

 

 

 

While in no way supporting the premise that there is any audible benefit in consumer-end high resolution audio vs 16/44.1 I have to point out that there are some methodological issues with the much discussed Meyer and Moran paper.

 

The biggest of these is that several of the high resolution discs used were not native high resolution but taken from lower resolution sources and resampled, several were born high resolution but not all. Thus testing the not really high res against the downgraded not really high res is only testing the downgrading not the actual content differences. Of course nobody managed a positive test with any of the proper high res discs vs red book either but it does mean that some of the results should not count. The interesting thing here is that low res sources resampled to high res formats could not be distinguished from low res resampled and then downgraded again. 

 

The 2nd lesser point is that 1 (possibly 2) of the 3 high res players were noisy, that is the SNR/Dynamic range was no better than a decent quality red book spinner. So that high res played back already had the extra dynamic range mangled by poor playback implementation so that the downgrading rather than lopping off 30db of dynamic range might have only lopped off 6db (obviously still a doubling of noise in effect)  possibly none. Of course no high res spinner actually manages the theoretical 144db DNR of 24 bits and few get over 120db in practice.

 

There is one AES paper that seems to suggest audible differences in live recordings simultaneously recorded at different sample rates and /or post-hoc donwsampled, it is the "Sampling rate discrimination: 44.1 kHz vs. 88.2 kHz" Amandine Pras, Catherine Guastavino 2010 AES convention paper. It is interesting but the stats are possibly a bit iffy so it is not the final word. it is discussed at length here http://www.hydrogenaudio.org/forums/index.php?showtopic=82264 it is copyright so I cannot post a link to my copy.

post #36 of 77
Quote:
Originally Posted by bigshot View Post

 

Likewise, it doesn't prove that a difference is an improvement. Only that it is different.


No argument there. That gets into subjects' internal representation of what they're hearing, and in fact no currently known scientific procedure that could establish such an internal mental state as fact. There might be brain scanning someday, but right now it's in what philosophers call "privileged epistemic access", similar to a personal report of pain. If I report pain (or, the pleasure of improved perceived quality--of coffee, a poem, a steak, or a musical sound), then no one else has standing to call into question my mental state. Nor do I have any means of demonstrating objectively to others that I actually have that mental state.

 

The most ABX-ing can do is prove that I do detect an audible difference, within the bounds of a confidence level. If I report an improvement, that's my report to others of my internal mental state. Given our limitations in knowing others' minds, we can at most infer indirectly from other behavior. If I'm in pain, I may grimace, take pain medication, go to the emergency room. In the case of musical quality, other indicators of my internal experience are quite indirect. For this platform, the easiest indicator is the amount of time/effort I must have expended to write these posts--not very good data, by a long shot.

post #37 of 77
Actually, once you establish that there is a difference, the next step is to measure what that difference is. Compare waveforms, test the equipment you're using, etc. Once you determine exactly what the difference is, then you can judge whether it's an improvement or not. Until then, it's just a difference. And the odds are usually on the side of it not being an improvement, becuase it's a lot easier to mess up sound than it is to improve it.
post #38 of 77
Quote:
Originally Posted by bigshot View Post

Actually, once you establish that there is a difference, the next step is to measure what that difference is. Compare waveforms, test the equipment you're using, etc. Once you determine exactly what the difference is, then you can judge whether it's an improvement or not. Until then, it's just a difference. And the odds are usually on the side of it not being an improvement, becuase it's a lot easier to mess up sound than it is to improve it.


I agree, measure what the difference is--as best we can. There are plenty of audiophiles who swear by tube gear, even vinyl LP's (!). That's their own subjective preference, so even if I go through and measure the outputs of these various systems, "higher quality" may be associated with measurements which appear to indicate more limited capabilities.

I'm going through this right now with selection of monitors to take the next step up from my low-middling current pair. Highly accurate reference monitors tend to reveal flaws which might otherwise be less apparent. Depending on a listener's goals, practices and musical tastes, "improve it" could be associated with objective measurements going either direction.

 

It would be nice to try and eliminate pure switching cues, since that comes up so often. Someone, can't remember if it was this forum, mentioned the possibility of a click which was louder one way than another on different sample rate files. That one, conceivably, could be measured by capturing the analog line out and watching for transients at the onset of playback, and then just comparing amplitudes. For what it's worth, I've *tried* to cue just on clicks and playback-start timing, and failed utterly. **Does not** prove a difference isn't there, but I report being unable to detect any such differences on my equipment.

post #39 of 77
Quote:
Originally Posted by bigshot View Post

I think it's safe to assume that if the upsample sounds noticeably different from the master 16/44.1, then the upsample has been distorted in some way. Then it's just a matter of determining whether the distortion was introduced in conversion, or whether it's an equipment problem during playback.

Agreed fully. As for Kunchur's claim that 44/16 is unable to resolve small time differences, the guys at Hydrogen Audio debunked that soundly. But rather than link to pages and pages of discussion there, Monty Montgomery proves the point nicely at 21:55 in this video:

Monty Montgomery explains digital audio

--Ethan
post #40 of 77
Quote:
Originally Posted by EthanWiner View Post


Agreed fully. As for Kunchur's claim that 44/16 is unable to resolve small time differences, the guys at Hydrogen Audio debunked that soundly. But rather than link to pages and pages of discussion there, Monty Montgomery proves the point nicely at 21:55 in this video:

Monty Montgomery explains digital audio

--Ethan


For the purposes of satisfying readers of forums, whatever evidence threshold they individually prefer is fine. As far as the scientific validity of claims in peer-reviewed empirical research, forums play no role. I look forward to seeing any rigorously conducted, reviewed, and published evidence which is on the record either for *or* against claims which are made about psychoacoustics and the audio engineering involved. If guys at Hydrogen Audio have opinions, that's great, but they have no bearing on the scientific validity. The same point is applicable to videos online--I'm glad folks make them, and I hope people watch them if they're interested and want to learn, but they have no bearing on the type of work being done in the discipline's research.

post #41 of 77
Quote:
Originally Posted by UltMusicSnob View Post

If guys at Hydrogen Audio have opinions, that's great, but they have no bearing on the scientific validity. The same point is applicable to videos online--I'm glad folks make them, and I hope people watch them if they're interested and want to learn, but they have no bearing on the type of work being done in the discipline's research.

If you had actually taken a single look you'd know that there is a proof. You can tackle the problem both from the theoretical side with maths or practical side with real signals. As such it is really trivial and surprising that people claim otherwise (ok, not so surprising if they don't understand digital audio).

post #42 of 77
Quote:
Originally Posted by UltMusicSnob View Post


For the purposes of satisfying readers of forums, whatever evidence threshold they individually prefer is fine.

 

Pssst! psst! you might want to consider who you're replying to there. He isn't a typical reader of internet forums!

post #43 of 77
Quote:
Originally Posted by bigshot View Post

 

Pssst! psst! you might want to consider who you're replying to there. He isn't a typical reader of internet forums!


Okay. This line [For the purposes of satisfying readers of forums, whatever evidence threshold they individually prefer is fine] was intended to refer just to "the guys at Hydrogen Audio", who  "debunked that soundly". I'll try to be more precise in the future.

Although...how am I supposed to know who I'm talking to on a forum with usernames??

post #44 of 77
Try googling Ethan Winer.
post #45 of 77
Quote:
Originally Posted by xnor View Post

If you had actually taken a single look you'd know that there is a proof. You can tackle the problem both from the theoretical side with maths or practical side with real signals. As such it is really trivial and surprising that people claim otherwise (ok, not so surprising if they don't understand digital audio).

"If you had actually taken a single look you'd know that there is a proof." -

 

At the video? If so, then you have mistakenly assumed that I haven't looked at the video--I've watched it all the way through twice, from a tip given by a user on another forum, and I am already familiar with the content. The video doesn't go to the points concerning the steps involved in detection of very small temporal changes which are discussed in Kunchur--it just makes a simple claim about the representation of the 44.1 kHz digital representation itself. Of course, no one hears a Redbook digital file directly--it has to be processed through both D/A and human physiology. And there are the further questions of temporal differential in the arrival of stereo signals, which is also not addressed in the video. Montgomery's video is a great presentation, it just doesn't go to the full range of questions about temporal resolution which are available to consider and study.

 

"You can tackle the problem both from the theoretical side with maths or practical side with real signals."

I'm not clear what this refers to. Montgomery's excellent video displays generated real signals, and the several papers mentioned above are empirical research with also real signals. In any case, previous threads in at least three fora have gone through this sort of discussion multiple times. Any further journal sources are of course welcome.  

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Can someone explain audio resolution to me?