The Optimal Sample Rate for Quality Audio

[mwheelerk]

Hi Lavry Tech - While this paper may be interesting and very valuable from an engineering standpoint, your surrounding statements really hurt your credibility. If you post here on CA in an effort to educate there is no need to tout Dan as "One of the world’s top converter designers..." or to begin your post with "Interested in the facts?"

I can find several engineers and AES Fellows who contradict much of what Dan says. My point is there's not one set of facts.

Very much agree with this statement. That header line raised an immediate flag for me.

[Miska]

A fair amount of academic researchers have expressed concerns about Kunchur's work.

How about Oohashi & co brain imaging results when sound with ultrasonics produces different brain responses than sound without ultrasonics?

[Jud]

Originally Posted by Julf
A fair amount of academic researchers have expressed concerns about Kunchur's work.

How about Oohashi & co brain imaging results when sound with ultrasonics produces different brain responses than sound without ultrasonics?

Just like Kunchur, people were expressing concerns about Oohashi's work four years before it was done! ;-p

[esldude]

James Johnston has expressed doubts about Kunchur's results. I don't know if anyone has exactly replicated his experiment. But very similar experiments came up with fairly consistent results that don't agree with Kunchur. I am not sure it matters all that much either way. Kunchur basically said timing differences were audible to about half the accepted amount. Even 44.1 khz will have the timing accuracy for that. I believe, though might be remembering it wrong, that Kunchur is one saying 44.1 khz couldn't time accurately enough, and that is more where other people in the field disagreed with him.

Now Oohashi's work is much more interesting. Especially as the ultrasonics don't go through the normal hearing channel of the ear. Instead they apparently stimulate the inner ear or some related area of the brain yet the ultrasonics enter through the eye sockets. Would like to know more about how high a frequency is needed, what ranges it covers etc. Maybe you should start a thread just on that Jud.

[Jud]

James Johnston has expressed doubts about Kunchur's results. I don't know if anyone has exactly replicated his experiment. But very similar experiments came up with fairly consistent results that don't agree with Kunchur. I am not sure it matters all that much either way. Kunchur basically said timing differences were audible to about half the accepted amount. Even 44.1 khz will have the timing accuracy for that. I believe, though might be remembering it wrong, that Kunchur is one saying 44.1 khz couldn't time accurately enough, and that is more where other people in the field disagreed with him.

Hi, Dennis. First of all, thanks as usual for referring me to something relevant I hadn't seen before. Having fun and learning (actually the latter pretty much assures the former) are my reasons for hanging out here.

Now, re Johnston and Kunchur: If you're referring to discussions in the Stereophile forums, I think Johnston's disagreement may be more with how Kunchur's work has been characterized than with Kunchur's specific conclusions. Johnston says 44.1kHz is good enough to specify the time difference in two signals to an arbitrarily short length. That's elementary sampling theory. As I explained above, I don't disagree; neither, I think, does Kunchur. In 2007, Kunchur was (at least as I minimally understand from reading the paper) talking strictly about audible range material. More recently, he has begun to talk about human ability to sense sharper amplitude rise times than occur in sound waves at the highest audible range. He has referred to this ability with the descriptor "ultrasonic," but always in quotes, i.e., not to be taken literally. I haven't seen anywhere that Johnston has questioned this, but he may have; frankly the signal-to-noise ratio at Stereophile was too much for me to make it to the end of the forum discussions. Anyway, you may be interested in slides from a 2010 presentation by Kunchur and others: http://www.physics.sc.edu/~kunchur/p...ime-scales.pdf

Have a look and let me know if Johnston has weighed in on this aspect of Kunchur's work specifically.

Now Oohashi's work is much more interesting. Especially as the ultrasonics don't go through the normal hearing channel of the ear. Instead they apparently stimulate the inner ear or some related area of the brain yet the ultrasonics enter through the eye sockets. Would like to know more about how high a frequency is needed, what ranges it covers etc. Maybe you should start a thread just on that Jud.

I'm considering it, but have some trepidation that more heat than light may result.

[spdif-usb]

Timing resolution is not the same thing as timing performance. Timing information is not restricted to ongoing phase differences at low frequencies, but can also be obtained from the envelopes of high-frequency sounds (Henning, 1974).
The localization of sound sources by humans has been studied alot in auditory neuroscience. One thing that plays a very important role in sound localization is transient information (i.e. the initial attack of a sound), which can be smeared over time due to filter ringing and which, especially if the transient is very steep, cannot always be accurately reproduced by low, or medium, resolution digital audio. As I already mentioned in another thread a while ago, the white paper on the HDCD standard very very clearly shows how and why Redbook falls completely short on this.
On top of this, binaural hearing not only relies on interaural time differences (ITDs), but interaural level differences (ILDs) also, as well as monaural spectral cues related to the cavities in the external ear, reflections and shadows of the head, the torso, and reverberations of objects such as the walls in a room.
Sound localization is largely influenced by visual cues and other neural interactions, for example information processed in the cortex. Interaural differences also play a role in the identification of sounds against background noise. The more familiar (or the more natural) a sound, the more accurately the human brain can usually localize and detect it.
Simply due to the way psychoacoustics work, blind listening tests cannot provide any solid proof as to whether two sounds are completely identical. For example, if someone listens to music in Redbook format, then listens to the same music in Hi Res right after that, the Hi Res version might reveal subtle sounds that were inaudible in the Redbook version but when the same person switches back from Hi Res to Redbook, these subtle sounds can suddenly be heard in Redbook even though they were inaudible at first. So, because psychoacoustics can never be ruled out of the equation, those who still claim measurements are the be-all and end-all in sound quality are ultimately wrong IMHO.

[esldude]

Simply due to the way psychoacoustics work, blind listening tests cannot provide any solid proof as to whether two sounds are completely identical. For example, if someone listens to music in Redbook format, then listens to the same music in Hi Res right after that, the Hi Res version might reveal subtle sounds that were inaudible in the Redbook version but when the same person switches back from Hi Res to Redbook, these subtle sounds can suddenly be heard in Redbook even though they were inaudible at first. So, because psychoacoustics can never be ruled out of the equation, those who still claim measurements are the be-all and end-all in sound quality are ultimately wrong IMHO.

Much of psychoacoustics was determined with blind tests. To say they can provide no proof of two sounds being identical is rather ridiculous. Perhaps some phenomena will require particular conditions, but that doesn't prevent a blind test being possible. I am not sure I believe your redbook hi rez example of what might be called innoculation. Once innoculated with the sacred hirez you then hear more in redbook? Even it could be tested blind if desired. I do understand one could hear a more clear rendition of sound, and once picking something out then go back to a more muddled result, and knowing what to listen for find it more easily. But it has to be in the less clear rendition at an audible level to be discerned. An interesting test would be an example of a recording available in the same form, same mastering where you could hear something first in hi-rez and then pick out the previously inaudible artifact in the lo rez form. I think that is conjecture on your part based upon the idea hi rez is higher resolution and easier to hear into for the listener. Something which may or may not be true in my opinion.

As for measurements being the be and end all, I don't know that really being the aim. Listening itself will always be the final arbiter. But one must include all that is known about how listening can be fooled or well you will get fooled. Measurements can provide a consistent background in many areas. Measurements are an excellent short cut as in ruling out known factors. It also can usually be adapted to measure any new result coming to light to help figure out just what is going on. It can sometimes eliminate factors as well.

[Miska]

he has begun to talk about human ability to sense sharper amplitude rise times than occur in sound waves at the highest audible range. He has referred to this ability with the descriptor "ultrasonic,"

I believe this and Oohashi's work match quite well, and also with my own experience on passive sonar and such. It's not about hearing ultrasonic sine waves alone, it's about sensing waveform shapes and changes steeper than "audible" (as continuous sines) frequencies. I believe hearing is closer to wavelet analysis filter banks than Fourier frequency breakdown. (I've spent quite some time on WVD analysis and etc, perfectly matching these)

[Jud]

I believe this and Oohashi's work match quite well, and also with my own experience on passive sonar and such. It's not about hearing ultrasonic sine waves alone, it's about sensing waveform shapes and changes steeper than "audible" (as continuous sines) frequencies. I believe hearing is closer to wavelet analysis filter banks than Fourier frequency breakdown. (I've spent quite some time on WVD analysis and etc, perfectly matching these)

Kunchur is not a neurologist, but he does cite work by others regarding different parts of the brain being used to process continuous tones versus changes in amplitude. My recollection of reading some of the work he cites, and of reviewing the cited papers in Google Scholar, is that this (different parts of the brain being used) is not controversial.

[sandyk]

"It's not about hearing ultrasonic sine waves alone, it's about sensing waveform shapes and changes steeper than "audible" (as continuous sines) frequencies."
Miska
I have always believed that it is about the steep changes, and have said so.Due to my age and hearing damage, I shouldn't be able to hear the things I do, yet I have a great deal of success with my audio projects elsewhere, where the constructors do not feel the need to modify the amplification sections.They say that I have got it right.
Regards
Alex

[spdif-usb]

Much of psychoacoustics was determined with blind tests.

That's true.

To say they can provide no proof of two sounds being identical is rather ridiculous. Perhaps some phenomena will require particular conditions, but that doesn't prevent a blind test being possible.

I never said a blind test would be impossible. It's the outcome of the test that would tell you the exact oppositte of what you would expect it to be. Rule number one is everything audio can be steered. During a blind listening test, people always listen in a different way than they would listen to the same sounds under a different circumstance. This has all been explained at the beginning of this video: Audio Myths Workshop - YouTube

I am not sure I believe your redbook hi rez example of what might be called innoculation. Once innoculated with the sacred hirez you then hear more in redbook? Even it could be tested blind if desired. I do understand one could hear a more clear rendition of sound, and once picking something out then go back to a more muddled result, and knowing what to listen for find it more easily. But it has to be in the less clear rendition at an audible level to be discerned. An interesting test would be an example of a recording available in the same form, same mastering where you could hear something first in hi-rez and then pick out the previously inaudible artifact in the lo rez form. I think that is conjecture on your part based upon the idea hi rez is higher resolution and easier to hear into for the listener. Something which may or may not be true in my opinion.

See the bottom of this page: TAS 194: Meridian Audio's Bob Stuart Talks with Robert Harley | AVguide
...And then continue to read on the next page. People aren't delusional, it's just the way the human brain works.

As for measurements being the be and end all, I don't know that really being the aim. Listening itself will always be the final arbiter. But one must include all that is known about how listening can be fooled or well you will get fooled.

That was my whole point from the start. Well, actually... Even if you do include all that is known about how listening can be fooled, you will still be fooled. This is because if you knew everything about how listening can be fooled, it would have probably won you the Nobel prize. lol

Measurements can provide a consistent background in many areas. Measurements are an excellent short cut as in ruling out known factors. It also can usually be adapted to measure any new result coming to light to help figure out just what is going on.

A shortcut inside the human brain? Hmmmm...

It can sometimes eliminate factors as well.

Meaning, it will usually eliminate the wrong factors until someone comes up with a whole new way of looking at things. That's science!

[Julf]

Julf, I don't understand what you intend by this.

My apologies - I made my posting while on the road, from a car ferry crossing from Holland to England, using a very slow satellite internet connection. As a result, I didn't recheck the Kunchur documents, and confused his findings with those of Kiryu and Ashihara. While Kiryu and Ashihara actually claim to show some ability of the human ear to hear frequencies above 22 kHz (addressed by the Griesinger presentation), Kunchurs experiments only shows that the ear can detect quite small timing differences. The problem with Kunchur is that he makes statements that, especially when taken out of context, imply that a 44.1 kHz sample rate system can't reproduce those timing differences (despite the fact that his experiments don't address that issue at all).

So, again, my apologies for a misleading link.

[Jud]

Kunchurs experiments only shows that the ear can detect quite small timing differences. The problem with Kunchur is that he makes statements that, especially when taken out of context, imply that a 44.1 kHz sample rate system can't reproduce those timing differences (despite the fact that his experiments don't address that issue at all).

Not to rehash old discussions, but: Kunchur's experiments reported in 2007 did indeed focus as you say on timing differences between two continuous signals. For audible signals, Shannon-Nyquist proves 44.1kHz is adequate to handle such timing differences to any arbitrarily brief length of time.

But certainly by 2010, Kunchur was focusing more on the audibility of inharmonic transients with such a steep rise in so short a time that 44.1kHz sampling rates are not adequate to reproduce them. Kunchur sometimes refers to such signals as "ultrasonic" (always using quotes) to point out the paradoxical fact that humans appear capable of responding to them as audible stimuli even though the amplitude rises are steeper than those of sine waves at the highest audible range for continuous tones. It appears from neurological research, as Kunchur has pointed out, that different parts of the brain handle these transient stimuli than are used for continuous tones.

[Jud]

Meaning it [measurement] will usually eliminate the wrong factors until someone comes up with a whole new way of looking at things. That's science!

While we must be conscious that measurements are not the whole story, let's not disregard the fact that they're an inescapable part of figuring out why and how things happen as they do, and what might work to make them happen more along the lines you'd like them to. There are people like Keith Johnson in audio, or Richard Lenski in biology, who are extraordinarily painstaking and careful in designing and setting up tests or experiments and taking the data from them, and that data in turn assists immeasurably (heh) in helping these fields progress.

Intuitive leaps in science and engineering may have a grip on the public imagination, but to me, anyone who announces that there is a "whole new way of looking at things" has a virtually 100% chance of being a crackpot.

[Julf]

Not to rehash old discussions, but: Kunchur's experiments reported in 2007 did indeed focus as you say on timing differences between two continuous signals. For audible signals, Shannon-Nyquist proves 44.1kHz is adequate to handle such timing differences to any arbitrarily brief length of time.

Indeed, full agreement here.

But certainly by 2010, Kunchur was focusing more on the audibility of inharmonic transients with such a steep rise in so short a time that 44.1kHz sampling rates are not adequate to reproduce them.

The only fairly recent work I can find is the AES panel contribution you already linked to. I find it interesting that the only actual experimental research work he refers to is the original work about temporal differences - all the rest is pretty much just theoretical speculation. It might be right, but it is still pretty much unconfirmed by any actual observations.

[JPS]

Dear all,

I've been experimenting with CDs upsampled to 24/192 using iZotope SRC.
24/192 definitely sounds better than 24/96 or 24/88.2 (which are themselves better than 16/44.1).
I don't know if this can be attributed to iZotope being better than the DAC internal up/oversampling, to the absence of clipping due to the DAC internal up/oversampling, to the need for less filtering within the DAC, to increased jitter immunity or to something else, but my experience is definitely in line with Barry Diament claims that with 176.4/192 everything is significantly more natural.

[Jud]

The only fairly recent work I can find is the AES panel contribution you already linked to. I find it interesting that the only actual experimental research work he refers to is the original work about temporal differences - all the rest is pretty much just theoretical speculation. It might be right, but it is still pretty much unconfirmed by any actual observations.

I do remember being frustrated that I couldn't find more clear-cut information in Kunchur's published work about rapid-rise inharmonic transients specifically. I will say that the evident care with which he designed, set up, and conducted his experimental work published in 2007 (building his own equipment in some critical cases to ensure it met the desired specs) went some way in convincing me he wasn't just making stuff up, but was discussing something about which he had some actual data.

There's also other published research work on such inharmonic transients, and on the differences in how they are processed in the brain versus continuous tones, as well as remarks by audio engineers in various sources (for one very informal example, see Miska in this thread) so it is not as if Kunchur is all alone in this. If he were, no matter how careful he appears to be in his experiments, I'd be considerably more skeptical.

[Julf]

I will say that the evident care with which he designed, set up, and conducted his experimental work published in 2007 (building his own equipment in some critical cases to ensure it met the desired specs) went some way in convincing me he wasn't just making stuff up, but was discussing something about which he had some actual data.

I agree that the 2006/2007 work seems very solid, but I will personally continue to be a bit sceptical about his more recent thoughts until he publishes some actual data. As to the ears/brain being able to process ultrasound, I would definitely find work from actual neuroscientists more interesting (and credible), but it is definitely way outside my area of expertise.

[bdiament]

Hi JPS,

Dear all,

I've been experimenting with CDs upsampled to 24/192 using iZotope SRC.
24/192 definitely sounds better than 24/96 or 24/88.2 (which are themselves better than 16/44.1).
I don't know if this can be attributed to iZotope being better than the DAC internal up/oversampling, to the absence of clipping due to the DAC internal up/oversampling, to the need for less filtering within the DAC, to increased jitter immunity or to something else, but my experience is definitely in line with Barry Diament claims that with 176.4/192 everything is significantly more natural.

Upsampling with iZotope's wonderful SRC algorithm can certainly be informative.
However, to my ears, it isn't close to *native* 24/192 (i.e., the original recording being at 24/192), which to me, is where the real magic happens and where, with those few converters that are actually up to the task, digital finally fulfills its promise.

Try upsampling the 16/44 or 24/96 version of one of the samples on Soundkeeper's Format Comparison page and compare it with the 24/192 sample to hear this for yourself.

Best regards,
Barry
Soundkeeper Recordings
Barry Diament Audio

[spdif-usb]

Intuitive leaps in science and engineering may have a grip on the public imagination, but to me, anyone who announces that there is a "whole new way of looking at things" has a virtually 100% chance of being a crackpot.

Again, my point exactly. The temporal resolution for monaural sound is about 2 ms, yes indeed but for binaural sound you're looking at an entirely different picture. Binaural measurements were achieved using only smallband tones. Complex, natural sounds are wideband. So then, who knows which way of looking at things will give you the least chance of being a crackpot?
Last time I checked, the Nyquist-Shannon theorem only applied to infinite duration, continuous signals. With a time limited, discrete signal you'll typically get both quantization and interpolation errors, which can have an audible impact. A modern filter not only causes post-ringing, but also changes the characteristics of background noise (i.e. noise that's already there before the filter is applied). Hence, the steepness of a filter does matter. Even, if using the best oversampling and upsampling techniques.

[Julf]

Last time I checked, the Nyquist-Shannon theorem only applied to infinite duration, continuous signals.

Indeed. Fortunately what happens with real, non-infinite signals is well understood. Because of finite amplitude resolution / signal-to-noise ratio, time resolution is also finite - but still more than sufficient. And yes, filters are always compromises. With infinite processing power, you can definitely avoid ringing, but most of us don't like the idea of a supercomputer bolted on to our DACs.

It is interesting that while people can discuss capacitor choice and cable quality until the cows come home, I see very little discussion of the processing power of the various DSP processors used in the DACs...

[Miska]

With infinite processing power, you can definitely avoid ringing, but most of us don't like the idea of a supercomputer bolted on to our DACs.

No you can't, if the ratio of passband vs Nyquist frequency is too small. I've spent enormous time to optimize filter to be as short as possible while still being "perfect", however it's just about math and it's impossible to optimize the filter to 1-tap and make the transition band with 144 dBFS attenuation to fit into 20 - 22.05 kHz band.

[Julf]

Hmm, still haven't figured out how to actually delete a duplicate posting... :-/

[Julf]

No you can't, if the ratio of passband vs Nyquist frequency is too small. I've spent enormous time to optimize filter to be as short as possible while still being "perfect", however it's just about math and it's impossible to optimize the filter to 1-tap and make the transition band with 144 dBFS attenuation to fit into 20 - 22.05 kHz band.

Sure, within those constraints it is impossible. Or, as a computer scientist would say, "hard"

But if you remove some of the constraints - allow massive oversampling and a practically-endless number of taps (both for filtering and for compensation), you have a situation where throwing CPU power at the problem is actually a solution.

Then there is the entirely separate debate about how harmful the ringing actually is - as long a it is post- and not pre-ringing.

[sjoc2000]

Interesting post, interesting comments. As is often the case in the world today everything comes down to "objective data", and conjecture-diagreement as to what it all "means". This "means", comes about because you always end up with a pesky human being variant at the end of the data stream. :0)

Now I was testing my hearing the other day (I'm getting a bit older these days), and discovered I'm deaf as a bat above 13 KHz. That's just my tympanic response. We have other variables, like the individual differences in brain processing of of audio signals, judgments based on psychological positions, etc. The point is that there is always the subjective reality of events. Somebody around here said you have your system just right, when you start tapping your foot, I wave my hands around in the air myself. :0)

Now I like and enjoy objective analysis, but don't get so wound up in it, you forget to tap your foot... :0)

Jim