Can you really hear the sound details over 20kHz?

[krabapple]

Let me explain how science works. The person who makes an assertion bears the burden of producing some data showing that the claim is not false, and is likely to be true. So, if you claim that humans cannot hear above 20 kHz you will conduct hearing tests. That has been done for humans and a variety of other animals, using simple sine waves ("tones"). If you then cite those studies on sine waves to show something else, such as musical waveforms, you may be extending those studies too far. The most likely factors involved would be different frequencies in combination with each other, or fast transients. You can do some hand-waving and invoke masking or a Fourier decomposition, but neither is as compelling as an actual study of the phenomenon at issue. Such gap filling isn't illegal but doesn't establish the assertion very well either. Worse is when people become highly emotional and muddy the waters with invalid arguments or even name-calling.

First, to be precise, no one should claim human hearing cannot extend above 20kHz. The commonly accepted threshold based on experimental data is a few kHz higher than that, in the young and in exceptional adult cases.

Second, you seem under the impression that the only experiments pointing toward such a limit are based on 'tones' as the probe. That simply isn't true, as a perusal of the literature on 'hi rez' -- feel free to use the bibliography of Reiss's 2016 metanalytic review as a starting point -- shows.

Others have noted your claim, repeated above, that 'tones' are insufficient as test signals to establish the bandwidth of human hearing, and you freely hypothesize 'likely factors' and dismiss criticism as 'hand waving', all the while evincing an ignorance of the literature and extensive discussion of these issues online over a period of decades. Perhaps you see why some find your approach curious from someone demanding scientific rigor.

 

[Nango]

Ultra high frequency energy is very low. No way the skin nervous system could pick it up, and integrate the info with your auditory perception. You also can't feel the spot of a flashlight pointed at your skin. Not to mention it wouldn't explain the supposed ability to detect ultra high frequencies via headphones.

Some people believe in bone conduction, but according to different studies that also doesn't apply to loudspeakers. (It does apply of course to head phones designed for bone conduction, or other transmitters attached to your skull).

Exactly, feel it on your teeths.

 

[andreasmaaan]

Unfortunately the "clicks" remaining are mostly about filter impulse response.

The residual after this LPF filtering is an impulse response + ringing at the cut-off frequency:

How can the output ring at the filter cut-off frequency if the input contains no content in that region in the first place?

 

[krabapple]

Exactly, feel it on your teeths.

If our teeths rattled when we play 96/24 audio, hi rez audio would have died at birth

 

[andreasmaaan]

The most likely factors involved would be different frequencies in combination with each other, or fast transients.

Why would the involvement of these factors be "likely"?

 

[Nango]

If our teeths rattled when we play 96/24 audio, hi rez audio would have died at birth

Sir, I wasnt talking about false teeths, the natural ones, why not?

 

[Geert]

Sir, I wasnt talking about false teeths, the natural ones, why not?

Put ear protectors in your ears and a headphone on your mouth and report back.

 

[Pio2001]

Not hearing it doesn't mean not feeling it (elsewhere).

Right.
Do you feel it elsewhere ? I don't.

My point is that it is correct to include any other perception channel in addition to the auditory system, but this is done naturally. When a listener is asked "can you hear this", the answer may be "well, I'm feeling it rather than hearing it", and that would be ok.

 

[rdenney]

They might also be seeing it in their minds--hearing a difference they expect to hear.

Rick "the reason for blind testing" Denney

 

[weasels]

Let me explain how science works. The person who makes an assertion bears the burden of producing some data showing that the claim is not false, and is likely to be true. So, if you claim that humans cannot hear above 20 kHz you will conduct hearing tests. That has been done for humans and a variety of other animals, using simple sine waves ("tones"). If you then cite those studies on sine waves to show something else, such as musical waveforms, you may be extending those studies too far. The most likely factors involved would be different frequencies in combination with each other, or fast transients. You can do some hand-waving and invoke masking or a Fourier decomposition, but neither is as compelling as an actual study of the phenomenon at issue. Such gap filling isn't illegal but doesn't establish the assertion very well either. Worse is when people become highly emotional and muddy the waters with invalid arguments or even name-calling.

I think you're misidentifying the null hypothesis, and therefore where empirical evidence should be required.

 

[Pio2001]

Filter definition? Impulse/step response? Pre-ringing?

The characteristics of the filters are in the screen captures posted above.
In short, they are "reject" filters, 100 dB / octave, linear phase.
Their impulse responses are attached here.

Unfortunately the "clicks" remaining are mostly about filter impulse response.

What do you mean ? The filter IS an impulse response.

A given frequency response curve and phase response curve completely define an impulse response, and conversely.

These two curves and the impulse response are two representations of the same filter. The Fourier transform is the operation that gives one from the other.

Impulse response -> Fourier transform -> phase and amplitude curves vs frequency
Phase and amplitude vs frequency -> inverse Fourier transform -> impulse response.

Below is the result with Chebyshev 15kHz, two runs, as per image, zipped flac attached.
View attachment 110742

The residual after this LPF filtering is an impulse response + ringing at the cut-off frequency:
View attachment 110751
We are skating on a thin ice here.

Yes, the sharper the filter, the longer the ringing.
If the filter is linear phase, like mine (the phase response curve is flat), there is ringing before and after any impulse.
If the filter is minimal phase, (the phase response curve depends on the frequency response curve in a special way), there is no ringing before an impulse. Only after. It looks like it is the case with yours.

However, the result that you show is not an impulse response. It is just your lowpassed signal.
The impulse response is what you see if you apply your filter to a single sample surrounded by digital silence.

The wav file that you get this way can be used to apply your filter to any other audio content using a convolver. That's how I filtered your signal : I asked Rephase to export the impulse responses of the reject filters that I was needing. It outputted the two wav files attached to this message.
Then I asked Foobar2000 to convert your signal to another wav file, but with the Convolver DSP applied in between. The convolver was setup with the attached wav files.
All wave files must have the same sample rate (I did everything at 44100 Hz, thus there were in fact two DSP during my conversion : resampler (to 44100 Hz) + convolver (with one of the attached files)).

 

[tmtomh]

Let me explain how science works. The person who makes an assertion bears the burden of producing some data showing that the claim is not false, and is likely to be true. So, if you claim that humans cannot hear above 20 kHz you will conduct hearing tests. That has been done for humans and a variety of other animals, using simple sine waves ("tones"). If you then cite those studies on sine waves to show something else, such as musical waveforms, you may be extending those studies too far. The most likely factors involved would be different frequencies in combination with each other, or fast transients. You can do some hand-waving and invoke masking or a Fourier decomposition, but neither is as compelling as an actual study of the phenomenon at issue. Such gap filling isn't illegal but doesn't establish the assertion very well either. Worse is when people become highly emotional and muddy the waters with invalid arguments or even name-calling.

First, your argument doesn't get any stronger by virtue of becoming more condescending in tone.

Second, you are making a fundamental logical (or perhaps empirical) error, which @weasels has described above as "misidentifying the null hypothesis." To put it another way, the fact that someone (or in this thread many people before and beyond me, as you well know) cites established knowledge does not mean they are "making a claim" that must have evidence provided to support it. You are the one making a claim, which is that the commonly established human hearing limit (of 20kHz or a little higher - the exact frequency doesn't matter for the purposes of this particular point) is not necessarily actually a limit since testing has all allegedly been with sine waves and not with either "real music" or "fast transients."

As there is no reason to believe that the number of tones in a sound, or the speed of the rise in volume of a sound, impacts what frequencies of sound we can hear, the burden is on you to provide a hypothesis for how or why the type of sound should impact the detectable frequency range of human hearing. The first difficulty I see in that regard is that, based on the prior comments of several members, your claim that the 20-20kHz range has been established only with sine waves is not true. So if the accepted range of human hearing has been determined experimentally with a variety of sound types, then your entire argument is invalidated from the start.

So what you're demanding of others here is not that they "do science" - you're simply stubbornly refusing to admit that what you thought was the case is not, when it comes to how scientists have established the range of human hearing.

 

[Wes]

Show me the data.

 

[Blumlein 88]

Show me the data.

What about those hearing tests that fire an impulse into the ear and then listen for what comes back. Shown to accurately depict hearing range. In infants you get a little beyond 20 khz. In adults typical results are 15-17 khz. No sine waves just a wide band impulse for a test signal.

ABR or auditory brainstem response for infants uses fast clicks at a high and low sound level. This one monitors the brain stem's nerve response.

OAE Ota acoustic emission test. Impulse fired into ear and microphones record what kind and how much sound comes back out.

Though less common OAE can also use tone bursts or two high frequency tones to find ear distortion via how much of the lower difference frequency is present.

 

[Pio2001]

Show me the data.

https://www.aes.org/e-lib/browse.cfm?elib=14195

Audio content used in this test : Many types of music and voice signals were included in the sources, from classical (choral, chamber, piano, orchestral) to jazz, pop, and rock music. (https://hydrogenaud.io/index.php/topic,57406.msg515700.html#msg515700)

 

[tmtomh]

Show me the data.

Provide a scientific definition of a "fast transient" that includes a qualitative distinction between a "fast transient" and a slow one. Then provide a hypothesis for how the speed of a change in volume could possibly impact humans' ability to detect sound at a particular frequency.

 

[Pio2001]

Provide a scientific definition of a "fast transient" that includes a qualitative distinction between a "fast transient" and a slow one. Then provide a hypothesis for how the speed of a change in volume could possibly impact humans' ability to detect sound at a particular frequency.

Non-linear distorsion, intermodulation, aliasing... that is not the point: the facts must come before the explanation. Ask first for data showing the audibility of fast transient beyond 20 kHz, then, and only then, an explanation can be proposed.

 

[restorer-john]

I also tried udial but no equipment damage, and I also commented about it here:
https://forums.dearhoney.idv.tw/viewtopic.php?t=52943&start=18
I suppose Google translate is good enough so I don't need to translate it myself, as my English is also broken anyway LOL.
Now everyone, especially @restorer-john , should read the instruction and inspect the signal clearly:
https://forum.cockos.com/showthread.php?t=201868
View attachment 110657

I suppose the original udial file looked like this. I don't have the file anymore, only a screenshot:
View attachment 110664

You can see that it is nowhere similar to udial, it has a fade in , and two out of the three tones are in the non-disputable audible range. When combined with my instructions, I don't see how someone can unconsciously damage their equipment by playing it, because the lower two tones will be painful enough to force people turn down the volume.

Therefore my conclusion is that people only read someone else's spectrogram (in log scale) without knowing the whole story, and accused me, I don't think it is entirely fair.



On the other hand in HA I commented about malformed floating point file may cause equipment damage. Why? Because I played those files unconsciously, and used IEMs with relatively fragile cable. I was scared that I quickly pulled them out from my ears, forcibly. The pulling action damaged the cable.

The bottom line is high level, high frequency tones of any type should only exist in the electrical domain. By all means play to your heart's content in that domain. I do it all the time.

But it's utterly stupid in the extreme to play such signals through amplification and connected loudspeakers. Amplifiers can and do oscillate and tweeters expire instantly. A typical tweeter can absorb between 1Watt to maybe 10W maximum and that's in a 50-200W rated speaker. A high level, high frequency/inaudible tone can deliver the full power of the amplifier to the tweeter. Most speakers these days have zero protection for the treble units in the form of fuses, polyswitches, etc. Protection has gone out of fashion unfortunately. Even 5% level of a typical 100W amplifier will fry the tweeter in less than a few seconds.

Just because it's trivial to create such audio files, doesn't mean it's a good idea to distribute them to typical audiophiles as some form of "test".

 

[Pio2001]

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.

It makes sense. With such a signal, we don't realize how loud the speaker is actually playing because we can't hear it.

I've just had a look at some speaker measurements and I saw that usually, the tweeter's impedance is at its lowest between 10 and 20 kHz.

The bottom line is high level, high frequency tones of any type should only exist in the electrical domain. By all means play to your heart's content in that domain. I do it all the time.
[...]
Just because it's trivial to create such audio files, doesn't mean it's a good idea to distribute them to typical audiophiles as some form of "test".

The problem is the playback level. If pure high frequency sines were not to be used at all, we could not equalize or measure anything. REW plays high frequency sines all the time during measurements. As long as the playback level is moderate, it is ok.

 

[Wes]

What about those hearing tests that fire an impulse into the ear and then listen for what comes back. Shown to accurately depict hearing range. In infants you get a little beyond 20 khz. In adults typical results are 15-17 khz. No sine waves just a wide band impulse for a test signal.

ABR or auditory brainstem response for infants uses fast clicks at a high and low sound level. This one monitors the brain stem's nerve response.

OAE Ota acoustic emission test. Impulse fired into ear and microphones record what kind and how much sound comes back out.

Though less common OAE can also use tone bursts or two high frequency tones to find ear distortion via how much of the lower difference frequency is present.

Good. Too bad it took this group 8 pages to get here.