[castleofargh]

With bone conduction, a deaf person can detect the 25 Khz frequency. Normal hearing persons can detect ultrasonic frequencies around 50 Khz and I have read young healthy subjects can hear as far as the 150 Khz frequency. I am surprised this have not been discussed, or it has and I did not see it anywhere on headfi. There is a lot to read about this topic and I just give this random link here:

http://www.tinnitusjournal.com/arti...dy-implications-for-highfrequency-therapy.pdf

Bone conduction transmission occurs constantly as sound waves vibrate bone - Wikipedia

https://www.head-fi.org/search/1549633/?q=bone+conduction&t=post&c[nodes][0]=133&o=date
Not sure if that link works for others? just in case, it involves searching for "bone conduction" in the Sound Science section

But beside reading the paper, there isn't much to talk about. A headphone probably won't have that effect because we use comfy pads, not jackhammers. And because as I understand it, the more a headphone itself shakes, the less the diaphragm is moving as it should=> more distortions. so we might need bad headphones for something like that to work well.
Most headphones won't extend or remain loud enough at those frequencies.
Typical music content probably doesn't have the needed ultrasonic energy even at unreasonably loud listening levels. If/when it does, it might just end up audibly masked by the rest of the music(perhaps with the exception of poorly filtered DSD, but then that's all noise-shaped crap and nobody wishes to perceive that).

When the right conditions are met for ultrasounds to be heard from bone conduction( AKA not using typical headphones), listeners seem to be hearing something similar to a 12 to 14kHz tone that doesn't correlate with the actual frequency of the ultrasound. So it might be something closer to your windows resonating from your woofer, than to music content. and of course nothing in the research is suggesting improved soundstage, improved details or whatever random stuff claimed without any sort of evidence by audiophiles.

the important part here, is that it's not the empty claim of some random guy on the web saying "trust me I heard it" under sighted conditions. it's been properly demonstrated that bone conduction of ultrasounds can lead to hearing something. there is no indication, so far, that the studies were flawed, so we believe that it's a fact. that's how it works.

 

[ScareDe2]

You can detect high frequency noise at negative decibel value and it will be converted to a noise closer to where the music plays.

 

[pfzar]

WHAT?

Man, this has been interesting to follow. Certainly a lot of theories going on. I think there a lot of scientists rolling over in their graves.

 

[danadam]

I must admit that given some of the reception my post got I do doubt you will accept my evidence, but I went ahead and did it anyways. I created the 16/44 from the 24bit version using sox -V3 -v 0.99 -G -b 16 rate -v -L 44100.

]$ sox "09 Don't Bring Me Down.flac" -n stats
             Overall     Left      Right
...
Pk lev dB      -3.08     -3.08     -3.30
RMS lev dB    -18.72    -18.46    -18.99
RMS Pk dB     -15.07    -15.07    -15.51
RMS Tr dB     -99.03    -98.14    -99.03

]$ sox "1644 Don't Bring Me Down.flac" -n stats
             Overall     Left      Right
...
Pk lev dB      -3.33     -3.44     -3.33
RMS lev dB    -18.82    -18.56    -19.08
RMS Pk dB     -15.17    -15.17    -15.60
RMS Tr dB     -94.56    -94.09    -94.56

Here's the null of your files:

Try the conversion without "-v 0.99 -G". Then you'll get something that's a better match:

]$ sox "1644 proper.flac" -n stats
             Overall     Left      Right
...
Pk lev dB      -3.24     -3.36     -3.24
RMS lev dB    -18.73    -18.48    -19.00
RMS Pk dB     -15.08    -15.08    -15.51
RMS Tr dB     -94.47    -94.04    -94.47

And the resulting null:


And as others said, for ABX test convert it back to 24/96 and compare that with the original.

 

[sonitus mirus]

And as others said, for ABX test convert it back to 24/96 and compare that with the original.

Fully agree with this part as has been mentioned by @bigshot as well.

This will best remove the behavior of the device that is playing back the different files. It is possible the Denon AVR is upconverting the 16/44 file to 16/48 internally, which could be creating a noticeable difference in the audio or perhaps some unassuming cue from the hardware. I know that the Pono player would play 16/44 files using a completely different type of filter that would roll off very early compared to the filter used with 24/96 files. While I might not be able to hear any difference with 1-2dB around 12kHz, some people surely can and the comparison is not apples to apples with regards to potential audible differences with the file types.

http://archimago.blogspot.com/2015/08/measurements-ponoplayer-another-mans.html

Thanks @jlawler for participating. You have my respect for actually taking the time and effort to do the test and post the results.

 

[KeithPhantom]

You can detect high frequency noise at negative decibel value and it will be converted to a noise closer to where the music plays.

What does that even mean? Negative decibel value relative to what? Decibels are a measure of ratios in logarithmic scales, they need a reference to make sense.

 

[bigshot]

I think he means that ultrasonic noise can create harmonic distortion down in the audible range if the playback equipment isn’t designed to work with super audible frequencies. He’s parroting something without understanding it.

 

[KeithPhantom]

I think he means that ultrasonic noise can create harmonic distortion down in the audible range if the playback equipment isn’t designed to work with super audible frequencies. He’s parroting something without understanding it.

He means intermodulation between aliased and passband frequencies when a simple reconstruction filter isn't used or not used correctly. It can happen, but you are hearing only the intermodulation that is in main audio band (20 Hz-20kHz) as distortion, not the supersonic frequencies, and that DAC has to be pretty much broken for this to be even noticeable.

 

[bigshot]

Audibility doesn’t matter to him I think. I’m not used to wrapping my head around fatal flaws that are imperceptible.

 

[ScareDe2]

Audibility doesn’t matter to him I think. I’m not used to wrapping my head around fatal flaws that are imperceptible.

 

[KeithPhantom]

If you are using Paul as evidence, you're talking about vested interests that are openly expressed by himself. Also, tell him he should accept the public debate Ethan Winer challenged him.

 

[ScareDe2]

I thought the debate was closed. If we use good over-ear headphones or good speakers with high definition audio (expensive or not) we detect the high frequency noises even if the signal is weak. That noise is being processed by the brain and relocated inside the frequency range of the music. I would say it is rather felt/perceived than heard. With noise filter it sounds more gentle on the ear and more enjoyable. By the way, I do not agree with everything Paul says I just think his videos are fantastic and deserve mentions.

 

[KeithPhantom]

we detect the high frequency noises even if the signal is weak.

What do you mean by "high frequency"? What is a "weak" signal in your example?

That noise is being processed by the brain and relocated inside the frequency range of the music.

There is a century of research showing how the upper limit of human hearing is 20 kHz, if you want to go against it, you better present evidence and your methodology to be scrutinized by the public.

With noise filter it sounds more gentle on the ear and more enjoyable.

What do you mean by filter? Are you using interpolation so you can use a less abrupt reconstruction filter? Is it linear-phase or minimum phase, or other type of filter? How do you know it sounds "gentler"?

Finally, why do you worry about differences that are usually -100 dBFS and can be masked by the distortion of your transducers? How do you separate transducer and electronic distortion by only "hearing"? Are ears more sensitive than AP analyzers?

 

[ScareDe2]

I have already answered everything.

As for the transducer distortion, it's close to the intent of the producer who himself worked with transducer distortion. The window resampling noise peak around 20Khz is a distortion you want to get rid of.

 

[KeithPhantom]

As for the transducer distortion, it's close to the intent of the producer who himself worked with transducer distortion.

What do you mean by this? So if the distortion is high in a specific transducer, the "producer" wanted to engineer something subpar?

The window resampling noise peak around 20Khz is a distortion you want to get rid of.

Can you show evidence that this is a problem? I can ask @amirm who actually was the director in charge of the Windows Audio stack and see if this is even true? Even if it is true, at what level is the distortion located?