Can you really hear the sound details over 20kHz?

[Wes]

*I* and *we* who have followed the tiresome zombie 'debate' on this for decades, are quite aware of the published studies (emphatically *plural*) on this topic.

It isn't an obscure topic nor is it hard to find discussion (and links to papers) related to it. Thus it's *you* making the extraordinary assumption.

I'm not doing this work for you. It's work you should have done before you posted.

Sorry krabby but I am a scientist, and expect people to support their assertions; otherwise, zombies proliferate.

 

[NTK]

There is a relatively recent (2016) paper looking at 18 previously published studies. Below is the abstract with my highlights. You can download it from Linkwitz's (RIP) website.
http://www.linkwitzlab.com/Links/Audibility of high rez.pdf

 

[Blumlein 88]

There is a relatively recent (2016) paper looking at 18 previously published studies. Below is the abstract with my highlights. You can download it from Linkwitz's (RIP) website.
http://www.linkwitzlab.com/Links/Audibility of high rez.pdf

View attachment 110560

A very definite maybe kind of sort of confirmation.

So to be high fidelity, do we go to 88.2 khz or 96 khz? That is the big question.

I fear the conclusion will be if we have PROOF that 96 and 192 khz are better then we have to assume the higher the better no matter how our hearing works. Which is why we need to hold out for 4096x DSD just to be sure. DXD will be merely adequate for the time being.

 

[BDWoody]

There is a relatively recent (2016) paper looking at 18 previously published studies. Below is the abstract with my highlights. You can download it from Linkwitz's (RIP) website.
http://www.linkwitzlab.com/Links/Audibility of high rez.pdf

View attachment 110560

From linked pdf:

. The meta-analysis also did not pursue ques-
tions regarding specific implementations of audio systems,
such as the choice of filtering applied, the specific high
resolution audio format that was chosen, or the influence
of the various hardware components in the audio record-
ing and reproduction chain (other than assessing potential
biases that might be introduced by poor choices).

Any chance that these issues, particularly the filters used, are more important than recognized?

Seems a lot more is needed to make anything conclusive.

 

[paulraphael]

Amir said in this interview that he was fairly consistently able to distinguish high-res recordings from version downsampled/downquantized to red book in ABX tests. But he suspected the differences had to do with requantizing artifacts (if I remember right). He was not suggesting he could hear the bat frequencies. He also made clear that he could not form a preference, and could not even tell which of the files was the high-res version.

 

[Tim Link]

I like this website because it has a scrolling bar to adjust frequencies, very useful. Just lower the volume to some 30-40% before hitting that play button.
https://www.szynalski.com/tone-generator/

Nice! Playing that through my cheap Labtec headphones revealed severe differences in frequency response between the left and right channels. It also confirmed that 16kHz is my hearing limit. My sound level meter confirmed that the headphones are playing a signal at 18k but I can't hear it at all.

 

[NTK]

Any chance that these issues, particularly the filters used, are more important than recognized?
Seems a lot more is needed to make anything conclusive.

Here is another 2020 paper looking at brain activities (and the results were negative). It seems that some of the Japanese researchers are still quite interested in this topic. I guess not everyone is considering this a settled subject.
https://www.nature.com/articles/s41598-020-78889-9

 

[andreasmaaan]

There is a relatively recent (2016) paper looking at 18 previously published studies. Below is the abstract with my highlights. You can download it from Linkwitz's (RIP) website.
http://www.linkwitzlab.com/Links/Audibility of high rez.pdf

View attachment 110560

I don't know if anyone cares much about my opinion, but I recently read took the time to read through every study discussed in that meta-study, and wrote up my notes here.

If that post is TL;DR for you, my conclusion was essentially that:

1. The authors' claim that training increased the ability to discern high-res audio "dramatically" is a bit misleading. Certainly, in some studies, trained listeners were significantly better able to discern differences than untrained listeners, but there were also many studies/tests in which trained listeners were not able to discern any difference.
2. In those studies in which statistically significant results were obtained, confounding choices to use simple truncation or non-noise-shaped dither were made, rendering it difficult to extrapolate from these to real-life cases in which redbook audio is dithered with noise-shaping. In the only study in which this issue was not present, the stimulus was not music, but rather 0.125s pulse trains, and even in that case, the result was only just statistically significant.
3. My takeaway would be that these studies demonstrate that redbook audio is skating very close to thresholds at which noise or other artefacts may become (marginally) audible in specific listening conditions, although they do not quite establish that properly-processed redbook crosses these thresholds with music as the program.

 

[NTK]

I don't know if anyone cares much about my opinion, but I recently read took the time to read through every study discussed in that meta-study, and wrote up my notes here.

If that post is TL;DR for you, my conclusion was essentially that:

1. The authors' claim that training increased the ability to discern high-res audio "dramatically" is a bit misleading. Certainly, in some studies, trained listeners were significantly better able to discern differences than untrained listeners, but there were also many studies/tests in which trained listeners were not able to discern any difference.
2. In those studies in which statistically significant results were obtained, confounding choices to use simple truncation or non-noise-shaped dither were made, rendering it difficult to extrapolate from these to real-life cases in which redbook audio is dithered with noise-shaping. In the only study in which this issue was not present, the stimulus was not music, but rather 0.125s pulse trains, and even in that case, the result was only just statistically significant.
3. My takeaway would be that these studies demonstrate that redbook audio is skating very close to thresholds at which noise or other artefacts may become (marginally) audible in specific listening conditions, although they do not quite establish that properly-processed redbook crosses these thresholds with music as the program.

For me I have concluded long ago that all these are way too subtle to make any difference in my personal music enjoyment. I couldn't tell when I put on my headphones whether I am listening to 256 kbps MP3 or FLAC. Don't ask me how I know

 

[andreasmaaan]

For me I have concluded long ago that all these are way too subtle to make any difference in my personal music enjoyment. I couldn't tell when I put on my headphones whether I am listening to 256 kbps MP3 or FLAC. Don't ask me how I know

Same here, lol. The redbook/high-res question is of academic interest to me, but in practical terms, I'm happy with high-bitrate lossily-compressed music (I've always said 320, but who knows, maybe 256 is equally fine!).

 

[Aerith Gainsborough]

Btw, youtube cuts HF at a certain point. I suggest this site.
https://www.audiocheck.net/audiotests_frequencycheckhigh.php
They start from the top and go down and as soon as you hear it that's obviously the highest frequency you are able to hear.

13K it is then. :S

 

[Geert]

I don't know if anyone cares much about my opinion, but I recently read took the time to read through every study discussed in that meta-study, and wrote up my notes here.

If a meta-study about the safety of COVID-19 vaccines would show comparable results and reservations I would not sign up for vaccination.

 

[tmtomh]

Sorry krabby but I am a scientist, and expect people to support their assertions; otherwise, zombies proliferate.

The hypothesis that human hearing generally extends from 20Hz to a maximum of 20kHz, with very few people being able to hear beyond 20kHz, is backed up by decades of data.

The hypothesis that a significant percentage of humans can hear above 20kHz is backed up by no data.

You are doing what flat-earthers, Creationsists, and audiophile-cable defenders do: you are equating the highly probable (few humans can hear above 20kHz because tons of empirical data suggest that's the case) with the highly improbable (many humans can hear above 20kHz but we just haven't found them yet and/or every study has employed flawed or limited methodology), based on the fact that the highly probable is not actually 100% certain while the highly improbably is not actually zero percent certain. Then, based on this false 50-50 "we don't really know" claim, you are demanding that other people do your Googling for you ("find me the evidence that humans can't hear above 20kHz" aka "do you personally know the studies that prove the earth is round?") and strongly implying that anyone who feels confident that humans cannot in fact hear above 20kHz as a rule is somehow close-minded, beholden to unquestioned assumptions, and unscientific (in notable contrast to what you claim about yourself).

That's bush-league nonsense, especially at a place like this.

 

[Blumlein 88]

Plus if you look at how the ear functions, it is like something designed to work to 15 or 16 khz. The extra above that is an edge case from the upper response having a leaky filter if you will. Is it 18khz, is it 20 khz, is it 22 khz, for all of those you have rather elevated levels. The significance of those is minor, on top of minor at best.

 

[Pio2001]

Hello, are you a moderator in Hydorgenaudio? I suppose moderators in HA should have high degree of audio knowledge so my question here:
https://hydrogenaud.io/index.php?topic=114816.msg946524#msg946524
If you read the subsequent replies someone said that it is impossible, or highly unlikely that playing a "malicious" audio file can result in equipment damage.

On the other hand some people here said my test signal can cause equipment damage:

Hi,
Yes I was a moderator in HA, but I am no more active.

I remember that when the "udial" test tone was posted on Hydrogenaudio, two members reported a fried tweeter on their system after having tried the test tone !
Since then, I always post a warning.

Udial is a test signal designed so as to make the slightest amount of digital clipping or aliasing obviously audible. It features very quiet tones associated with a full scale signal vobulated around 20 kHz.
It was later remade with stronger tones in order to avoid people turning the volume too loud to hear them while not hearing the ultra-powerful 20 kHz sine.

 

[restorer-john]

Hi,
Yes I was a moderator in HA, but I am no more active.

I remember that when the "udial" test tone was posted on Hydrogenaudio, two members reported a fried tweeter on their system after having tried the test tone !
Since then, I always post a warning.

Udial is a test signal designed so as to make the slightest amount of digital clipping or aliasing obviously audible. It features very quiet tones associated with a full scale signal vobulated around 20 kHz.
It was later remade with stronger tones in order to avoid people turning the volume too loud to hear them while not hearing the ultra-powerful 20 kHz sine.

Those of us who work with test tones all the time recognized the potential for such evil many decades ago with digital. A low level recording with a "trojan horse" at high level, with inaudible frequencies to kill expensive tweeters. As such, dubious audio files downloaded from the internet should be treated with absolute suspicion until viewing a spectrum.

It's just a phenomenally dumb idea to put anything near 20kHz at 0dBFS on any digital file and encourage people/audiophiles to download and "listen" on speakers. Such signals should stay in the electrical domain.

 

[Wes]

The hypothesis that human hearing generally extends from 20Hz to a maximum of 20kHz, with very few people being able to hear beyond 20kHz, is backed up by decades of data.

The hypothesis that a significant percentage of humans can hear above 20kHz is backed up by no data.

You are doing what flat-earthers, Creationsists, and audiophile-cable defenders do: you are equating the highly probable (few humans can hear above 20kHz because tons of empirical data suggest that's the case) with the highly improbable (many humans can hear above 20kHz but we just haven't found them yet and/or every study has employed flawed or limited methodology), based on the fact that the highly probable is not actually 100% certain while the highly improbably is not actually zero percent certain. Then, based on this false 50-50 "we don't really know" claim, you are demanding that other people do your Googling for you ("find me the evidence that humans can't hear above 20kHz" aka "do you personally know the studies that prove the earth is round?") and strongly implying that anyone who feels confident that humans cannot in fact hear above 20kHz as a rule is somehow close-minded, beholden to unquestioned assumptions, and unscientific (in notable contrast to what you claim about yourself).

That's bush-league nonsense, especially at a place like this.

You obviously did not read what I have posted here.

ONE MORE TIME: The "decades of data" are simple sine waves.

IF you have something on fast transients in mixed musical waveforms. LMK. Until then, that claim is not proven. ASR needs to be clear on what is science and what isn't, but is simply likely or a gap-filler.

I am saying this a scientist.

 

[rdenney]

Amir said in this interview that he was fairly consistently able to distinguish high-res recordings from version downsampled/downquantized to red book in ABX tests. But he suspected the differences had to do with requantizing artifacts (if I remember right). He was not suggesting he could hear the bat frequencies. He also made clear that he could not form a preference, and could not even tell which of the files was the high-res version.

Right. Detecting a difference is not the same thing as declaring a winner. Amir also mentioned in his interview that he required specific training in knowing what to listen for.

In an online test from somewhere (I didn’t note the source such that I remember it now), I could distinguish between lossless and lossy MP3 files ranging from 128 to 320 kbps. With headphones in a good headphone amp, I could identify them reliably enough to be better than guessing. But I had to listen to them over and over again, looking for specific details of rendering. And those differences are undoubtedly greater than comparing redbook and high-res. Nothing of what I heard between the better lossy formats and redbook could attract the sorts of transformational experiences people claim. Without the AB comparison, none of what I was listening for would have been noticeable. I consistently cannot hear a difference in my own files (24/96 FLAC needledrops downsampled to 16/44 redbook). But I suspect some downsampling can show aliasing effects or quantizing artifacts if done poorly.

In photography, I always work in raw because I want to have the headroom to make big adjustments without posterization or other artifacts. In sound, I record in 24/96 for the same reason (and my Benchmark ADC is certainly good for 20 bits). In ancient analog days, I recorded live music on VHS HiFi or onto a Tascam 80-8, even when I knew the distribution would be on cassette. Same principle—one might need to boost a track during mix-down or add an effect. But when mixed down from the masters to digital files to put on CD, I could hear no losses even back when I could hear 17 KHz with no difficulty.

But I can hear it when high-res files and CDs are mastered (I said mastered, not mixed) differently, just as when comparing LPs and CDs. But that comes across as differences in EQ.

For nearly all people, this seems like a vacuous argument. I go with my own ears. So far, I can’t hear the difference in lossless formats when mastered the same way.

Rick “who has done his own AB testing” Denney

 

[krabapple]

Sorry krabby but I am a scientist, and expect people to support their assertions; otherwise, zombies proliferate.

I'm a scientist too, and thus know how to do literature searches. Did they skip that part of your training?

 

[krabapple]

There is a relatively recent (2016) paper looking at 18 previously published studies. Below is the abstract with my highlights. You can download it from Linkwitz's (RIP) website.
http://www.linkwitzlab.com/Links/Audibility of high rez.pdf

View attachment 110560

Uh huh...and after that the wheel kept turning....that paper provoked its own criticism. Hint: a pitfall of meta-analysis is that results depend on what data are included or excluded.

Next up Oohashi et al. . Then perhaps Kuncher (ask JJ about him). Rinse and repeat.

This is all new stuff to some of you, apparently. To some others of us, the molehill birthed from this mountain of work could hardly be tinier.

Now, think...even generously granting that Reiss is correct , that he did not exclude any important data, and assuming that what is heard is greater bandwidth and not playback artifacts ... if hi rez was the audible game changer its numerous fans say it is, would we even need a meta-analysis of decades worth of papers (as done here) to finally unearth a positive signal? And would the signal be 'small but statisticially significant'?