Skip to main content
Topic: Listening audio content above 20kHz (Read 11206 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Listening audio content above 20kHz

Sample audio for discussions in threads:

http://www.hydrogenaud.io/forums/index.php?showtopic=106200

http://www.hydrogenaud.io/forums/index.php?showtopic=106196

Original 24/192 vinyl recording: [attachment=7953:01-03-14...703_2250.flac]

[attachment=7957:24192_sp...m_full_2.png]

and what was recorded above ~20kHz  [attachment=7954:02-140703_2256.flac] (24/48)

[attachment=7958:24192_spectrum_2.png]

Latter file was created using this method (probably not the best possible method):

- 24/192 vinyl recording

- added 384th order HP @ 20kHz (steepest I could find) (RubberFilter by Budde&Stilianos)

- rendered to 24/192 WAVE (i.e. only content above ~20kHz)

- changed the WAVE file samplerate info from 192000 to 48000 (hex editor)

- imported the 48kHz WAVE back to original 24/192 project

- export to FLAC

EDIT: changed the spectrum images

Listening audio content above 20kHz

Reply #1
Pardon my intermission, but wouldn't this kind of claim be as wacky as another one claiming, say, "image samples for seeing in infra red/ultraviolet" given either is way beyond our naturally-human league?
Listen to the music, not the media.
Qualidade em MP3

Listening audio content above 20kHz

Reply #2
Try applying a 20khz highpass before "hearing" these files, so the audible frequencies don't disturb de hearing of the inaudible ones.

Listening audio content above 20kHz

Reply #3
Try applying a 20khz highpass before "hearing" these files, so the audible frequencies don't disturb de hearing of the inaudible ones.


He's done that and then slowed down playback so the ultrasound becomes audible. 

What's interesting about this test is that while there is some content up there, its largely uncorrelated with the music aside from some rhythmic clipping.

Listening audio content above 20kHz

Reply #4
Quote
Try applying a 20khz highpass before "hearing" these files, so the audible frequencies don't disturb de hearing of the inaudible ones.


Spectrogram of the audio before lowering the samplerate:

[attachment=7960:24192_sp...bove_20k.png]



Listening audio content above 20kHz

Reply #5
...

What's interesting about this test is that while there is some content up there, its largely uncorrelated with the music aside from some rhythmic clipping.


There are some synth overtones as well. I checked this with 2 different (decent) headphones. Files are in time sync so if you load them say in the beginning of Reaper track 1&2 ...

Distortion might be partly result from normalization (-0.1) I had done for the source file ... I'll check if distortion can be lessen by not normalizating.

Listening audio content above 20kHz

Reply #6
Sorry, that should have read "clicking", which would be the overtones.


Listening audio content above 20kHz

Reply #8
Pardon my intermission, but wouldn't this kind of claim be as wacky as another one claiming, say, "image samples for seeing in infra red/ultraviolet" given either is way beyond our naturally-human league?

I don't think it's about audibility. I think he's trying to refute my claim that the sounds above 20 KHz (or thereabouts) in a vinyl rip aren't musical content, but rather are a by-product of record damage or distortion otherwise introduced in the mastering, manufacturing or playback processes. I'm unsure what to make of it.

It's difficult to find anything but folklore and unverifiable comments from former cutting lathe operators on what the physical limits really are. A search I ran just now turns up some lathe specs say they have a response of up to 25 kHz, and one mastering house says, vaguely, "With modern cutting equipment, the bandwidth of the signal that can be used for cutting is more than 20 kHz". However, it goes on to say that the higher frequencies (it's not clear exactly how low this range goes) must be mastered quieter and must be cut into a narrower groove, and the groove must be played with a properly aligned non-spherical stylus to avoid damaging the record. Other sources allude to the upper limit decreasing as you get closer to the center of the record, but there are no actual numbers provided.

Here's a topic of interest from our forum: ([a href='index.php?showtopic=98178']click[/a])

(Also, I'm never really sure whether to capitalize the K in kHz...)

Listening audio content above 20kHz

Reply #9
Ran the same test without normalizing the gain for the source file (as it was normalized in the 1st test). Here are the spectograms got from new WAV files then:

Source file [attachment=7961:24192_no...liz_192k.PNG]

After removing content < ~20kHz [attachment=7962:24192_no...hf__192k.PNG]

After changing the file samplerate [attachment=7963:24192_no..._hf__48k.PNG]





Listening audio content above 20kHz

Reply #10
Distortion introduces harmonics.
Real instruments have harmonics.

How do you know which you're seeing?

That's kind of rhetorical - you need expertise and knowledge, but even then you can't be 100% sure.

So, for example, in some of these samples you can see real high frequencies (and also the lack of distortion creating fake high frequencies) simply because the brick wall filter of the digital master is visible in the vinyl capture...
http://www.mediafire.com/?59m192uz8mq09
...from...
http://www.hydrogenaud.io/forums/index.php...mp;#entry683458

Whereas I have many LPs copied from 78s with a 5kHz low pass filter, and distortion harmonics are visible out to 25kHz.

To train your ears, take a CD, and run your trick on the 15-20kHz region. This is real content. What does the result sound like?
You can, with skill, low pass filter than CD at 15kHz, put an aural exciter or similar effect on it, and run your trick yet again. This is now all fake content. What does the result sound like?

Cheers,
David.

Listening audio content above 20kHz

Reply #11
Distortion introduces harmonics. Real instruments have harmonics.  How do you know which you're seeing?  That's kind of rhetorical - you need expertise and knowledge, but even then you can't be 100% sure.


I understand your point but (that's why I wanted to do this test), as those spectograms I linked in my prev post are just for to show that when normalization to -0.1 isn't present in source audio then the data of audio file derived from it looks cleaner in spectogram (or did the blue colour vanish because of some other reason (I used those attached FLAC files for those previous spectograms)?).

Listening audio content above 20kHz

Reply #12
Pardon my intermission, but wouldn't this kind of claim be as wacky as another one claiming, say, "image samples for seeing in infra red/ultraviolet" given either is way beyond our naturally-human league?

I don't think it's about audibility. I think he's trying to refute my claim that the sounds above 20 KHz (or thereabouts) in a vinyl rip aren't musical content, but rather are a by-product of record damage or distortion otherwise introduced in the mastering, manufacturing or playback processes. I'm unsure what to make of it.

Thanks for the explanation.
(Also, I'm never really sure whether to capitalize the K in kHz...)

According to the wikipedia entry:
Quote
This SI unit is named after Heinrich Hertz. As with every International System of Units (SI) unit whose name is derived from the proper name of a person, the first letter of its symbol is upper case (Hz). However, when an SI unit is spelled out in English, it should always begin with a lower case letter (hertz), except in a situation where any word in that position would be capitalized, such as at the beginning of a sentence or in capitalized material such as a title. Note that "degree Celsius" conforms to this rule because the "d" is lowercase.— Based on The International System of Units, section 5.2.

I hope it helps.
Listen to the music, not the media.
Qualidade em MP3

Listening audio content above 20kHz

Reply #13
(Also, I'm never really sure whether to capitalize the K in kHz...)

According to the wikipedia entry:
Quote
This SI unit is named after Heinrich Hertz. As with every International System of Units (SI) unit whose name is derived from the proper name of a person, the first letter of its symbol is upper case (Hz). However, when an SI unit is spelled out in English, it should always begin with a lower case letter (hertz), except in a situation where any word in that position would be capitalized, such as at the beginning of a sentence or in capitalized material such as a title. Note that "degree Celsius" conforms to this rule because the "d" is lowercase.— Based on The International System of Units, section 5.2.

I hope it helps.


He was asking about the k.  It's small in kHz.

Listening audio content above 20kHz

Reply #14
...  What's interesting about this test is that while there is some content up there, its largely uncorrelated with the music aside from some rhythmic clipping.
  There are some synth overtones as well. I checked this with 2 different (decent) headphones. Files are in time sync so if you load them say in the beginning of Reaper track 1&2 ...  Distortion might be partly result from normalization (-0.1) I had done for the source file ... I'll check if distortion can be lessen by not normalizating.


Checked another recording which was done as 24/96. On Reaper, after normalization of the hf audio file, it was quite clear that the content correlates the source fully (singing parts, percussions, keyboards) ... highest peaks was caused (in this other case) by sibliants (ssss's).

There also was a static 22.2kHz signal but it's peak was ~-21dB from the highest peak (-0.1dB). I've seen this type static noise on many LP's but at different fr positions (but almost every time it is above 20kHz). Could this be some surface noise?

In the end, using this method, it's hard to say in practice how disdorted the audio above 20kHz is... least there's something there which reminds the data found below 20kHz ...?

Listening audio content above 20kHz

Reply #15
highest peaks was caused (in this other case) by sibliants (ssss's).
That's very likely to be due to distortion, rather than real content. Not 100% certain, but very likely.

Cheers,
David.


Listening audio content above 20kHz

Reply #16
! Something might went wrong in this HF content test??? 

RubberFilter, which I used for all HP-filtering, seems not to work properly with 192kHz audio (can anyone confirm this?) ... to cut off everyting below 20kHz happens when Hz value is set to ~5k ... so, how could this effect in this test (files in my 1st post) ... my recording was therefore done with a steep HP set to cut below ~100Hz (explains missing lows) and the HF audio file is generated actually from above ~40kHz tones ? 





Listening audio content above 20kHz

Reply #17
I don't know, but does it have to be perfect? If you want to do more than just highpass, you can maybe resample the original to 44.1, upsample it back to the original rate, then mix-paste that 50/50 over the original. Highpass that (doesn't have to be perfect/steep) and then slow it down.

Anyway in your original files, the fact that there's such a raspy, scrape-y aesthetic to the highpassed, slowed-down content makes me think it must be awash with distortion, but it's hard to know what's distortion when it's mostly percussion (noise instruments) anyway.

However, when tonal instruments are playing, their harmonics do seem to extend higher than expected. I'll grant you that. Still, it would  be nice to have some samples of purely tonal instruments rather than this one which is full of percussion...

In my search for something other than anecdotes regarding the lowpass filter protecting the cutting head from overheating, I ran across something interesting. I probably shouldn't try to glean any facts from poorly translated product literature for a $900 phono preamp described by an impenetrable wall of audiophile gibberish, but this one I ran across does say the following:

Quote
This filter has in case of Ortofon 60kHz, early Neumann´s had 30kHz and 50kHz, where up from the 60´s the Neumann´s had only 50Khz. Scully or Westrex have low roll-off values of 20-30kHz.


They go on to say they consider 50 kHz the most common rolloff point, and they say that Neumann schematics show that the way these filters are implemented, it's basically flat out to 20 kHz (maybe 0-2 dB drop by that point) but introduces a phase offset of ~15º at 10 kHz and in excess of 25º at 20 kHz. They consider this a problem and apply a modified RIAA curve with a bit of a treble boost to compensate somewhat. I'm not convinced it matters.

Anyway, this analysis seems to suggest that if any ultrasonic content makes it through recording, mixing, and delivery to the vinyl mastering stage, then it's not necessarily going to be filtered out on its way to the cutting head, at least not with these high-end lathes that have been in use since sometime in the 1960s.

In other words, I am backpedaling slightly.

Listening audio content above 20kHz

Reply #18
-Except for a very small number of percussion instruments such as the crash cymbal, acoustical instruments simply don't make these frequencies.
-The few instruments that do always have concurrent, stronger content in the octaves just below these so it would be perceptually masked [pretending for the moment that some human on earth can hear these frequencies when presented in isolation].
-Typical studio microphones don't record these frequencies, at least not very well.
-Typical studio recording consoles, processors, tape decks, and digital recorders don't record these frequencies.
-Many (most?) cutting lathe setups use a digital delay line to control variable pitch (wider groove spacing for impending loud, bass heavy passages) which truncates these frequencies from ever reaching the cutting head.
-Cutting lathe heads add distortion, noise, and artifacts which are related to the music in these frequencies.
-Most turntable cartridges have poor response in these frequency ranges.
-Content recorded in these frequency ranges are fragile and often can't stand being played more than a few times before the stylus damages and distorts the signal by plowing through it like a rake, permanently damaging the ultra-HF content. [JVC's "super vinyl" may be an exception.]
-Most one inch or .75 inch tweeters have very poor dispersion at these frequencies, assuming they can even reproduce them at the same level as the audible content on-axis, so the actual acoustical power response measured in the far field is very poor.

Oh, and here's a big one: humans can't hear it!

Perhaps we should consult bats and dolphins to ask if this spurious junk we see in spectrograms is "musical content" as opposed to simply noise, distortion, and artifacts, as I strongly suspect.

Listening audio content above 20kHz

Reply #19
There's no argument here about whether it's audible; it's certainly not. What Juha is wondering about is whether it's true that a vinyl rip is not going to have anything but noise above ~20 kHz. I asserted, for the same reasons as you, that it's just noise, but when he cut out all the lower frequencies and slowed it down to shift the pitch down to the audible level, there was a moment where a tonal instrument (I want to say organ/keyboard?) was audible and had evenly spaced stripes on the spectrogram. My question, then, is whether that means anything, or if that's just what happens "up there" on the vinyl and/or as an artifact of playback. It seems too non-random to be distortion, but then, I am not an expert, so I'm putting it to those of you who are.

Listening audio content above 20kHz

Reply #20
I've made the test slightly differently to get something a bit more meaningful.


1) Imported flac to audacity (2.0.5, configured with WASAPI driver, and windows audio setup for  192Khz 24bits).
1.5) Ensure that the project is setup to 192000 Hz (bottom left part of the screen)

2) Applied 3 times the HighPass filter effect, at 20000Hz with 48dB rolloff to get the filtered file.  I don't hear this filtered file (with earphones, at the same volume that i listen to the non-filtered file).

3) Changed the filtered file's samplerate from within audacity ( click on the track name to show the dropdown menu, set (sample) rate option, 48000 Hz). This is equivalent to what the OP did when editing the file with an hex editor.

4) Split the stereo track and delete the right channel. ( click on the track name to show the dropdown menu, and use the split stereo track. do not use the split to mono).

5) Use the change tempo effect, increasing it by 300%.

6) import the original flac track into this project.

7) Split the stereo track and delete the right channel. ( click on the track name to show the dropdown menu, and use the split stereo track. do not use the split to mono).

8) Setup original's track as the right channel. ( click on the track name to show the dropdown menu, and use the Right channel option).

Play it. on the left channel, you hear the 20Khz-96Khz frequencies, shifted into the 5Khz-24Khz frequency bands.
on the right channel, you hear the music.


Playing it this way, it is much more easy to hear what's in there. In this piece, it is mostly the hihats, as one would expect, but also a bit the overtones of the distorted guitar.

The sound is a bit distorted overall. I don't know what to expect from other genres of music.

Listening audio content above 20kHz

Reply #21
[quote author=[JAZ] link=msg=869509 date=1405188409]Playing it this way, it is much more easy to hear what's in there. In this piece, it is mostly the hihats, as one would expect,[/quote]
The artifacts caused by the hihats. I assure you the studio microphone that was aimed at the hihats didn't do diddly above maybe 25/30 kHz or so.

Harmonic distortions of the cutting lathe and playback pickup are indeed related to the music content and not random in nature, so it is not surprising that when we pitch shift them down to audible ranges they seem tonal. These artifacts were never in the original studio master recording however.

Listening audio content above 20kHz

Reply #22
Quote
Still, it would  be nice to have some samples of purely tonal instruments rather than this one which is full of percussion...


I'll check if I can find suitable source from my LP collection (looks like it's quite common that there is some sort of high frequency cut done ... a suitable source which one I already found did have disruptive popping issue at high frequency (>~20kHz) area).
 


Listening audio content above 20kHz

Reply #23
Good luck finding a recording that doesn't truncate the octaves you seek due to the microphone used and the tape recorder it fed, but my understanding is that the king of ultra high frequency content from an acoustic source is jingling a set of keys. It is said to have the majority of its acoustical power above 20kHz!

Listening audio content above 20kHz

Reply #24
I pretty sure, some years back, I reported recordings from, and most likely provided a Spectral View screen shot of,
Cardas Frequency Sweep and Burn-In Record
tracks that have sweep tones which go up to 30kHz. The high end of one of them actually goes to 31kHz.

They were easily captured in my 88.2kHz sample rate recording. In addition there are four harmonic tones from the sweep fundamental that clearly reach the Nyquist limit and alias back down, plus three more higher order harmonics that fade out of level range before going that high. Thus, if there were music at that frequency that could be captured by the microphone, and the rest of the recording chain had no lower frequency limit cutoffs, the music at that frequency could end up on an LP.

The point is only that real signal well above hearing could be captured on LPs -- with the right set-up and equipment,  not that I think it has any relevance to music listening.

Just a note that may avoid another irrelevant harangue based on my statement above. I also reported, quite some years ago, that all the sound cards I could test (included a semi-pro card from M-Audio and one from Echo), and several more expensive professional audio cards other respondents tested, always produce aliasing from tones that exceed the Nyquist limit of whatever sample rate one uses. These show up quite well in a spectral view or a frequency analysis. These are most likely meaningless from a real music recording viewpoint. That is, the aliasing, and thus the aliasing distortion from any music with content high enough to produce such aliasing, is probably far too low a level to ever be audibly detected.

 
SimplePortal 1.0.0 RC1 © 2008-2019