How high can you hear?

Topic: How high can you hear? (Read 16890 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

How high can you hear?

Reply #25 – 2002-04-01 09:11:14

What Frank meant is that the cutoff is not a hard limit, but a slope which is dependent on the exact playback loudness.

If you repeat a hearing test with a different playback volume, you will find a different cutoff. The louder, the further you will be able to go.

Unless we all calibrate to the exact same playback level, any comparison is meaningless.

Quote

i was able to hear all except the 8 hz sine, altho the 22khz sine was extremely faint, as was the 20khz sine. the 21khz sine for some reason seemed more audible than either of these.

You are most likely hearing distortion from you equipment, not the actual 22kHz wave. A CD player will lowpass at 20kHz, so it is not possible to perform hearing tests above that anyway.

The same (distorting equipment) could very well be true of any poster in this thread, including myself.

--
GCP

How high can you hear?

Reply #26 – 2002-04-01 10:00:21

Quote

Originally posted by Garf
You are most likely hearing distortion from you equipment, not the actual 22kHz wave. A CD player will lowpass at 20kHz, so it is not possible to perform hearing tests above that anyway.

The same (distorting equipment) could very well be true of any poster in this thread, including myself.

--
GCP

i'm agree in this. an 20kHz lowpassed signal sounds much clearer than a full range signal to me. but an 19kHz lowpass [or even 16kHz] like the most lossy codecs will do destroys the "real" feeling of an song.

Dezibel

How high can you hear?

Reply #27 – 2002-04-01 11:03:37

Some people thinks, that even if human ear is unable to hear >22Khz frequincies, human beeing still could feel them somehow. One sound producer was experementing with good analog recorder and violin, and found that people somhow could find the difference between original recorded signal, and 25Khz lowpassed signal.
This information was taked from "sound producer" journal, and, of course, it should be checked up. But... who knows? Hearing is very non-linear system, maybe we could not hear 22Khz sine, but could distinguish 11Khz sine from 11Khz triangle tone. the problem is that I can experement by myself, I just have no proper equipment...

How high can you hear?

Reply #28 – 2002-04-01 11:23:24

Quote

Originally posted by Dezibel

i'm agree in this. an 20kHz lowpassed signal sounds much clearer than a full range signal to me. but an 19kHz lowpass [or even 16kHz] like the most lossy codecs will do destroys the "real" feeling of an song.

This will not be true for a very very large amount of people.

I would really urgently urge you to do a blind test and see if you can reliably distinguish a 19kHz lowpass from a 20khz lowpasss on music. I would be surprised if you could do it. You have made several statements like this before and I am still waiting to see any evidence they are true.

Note that being able to hear pure tones up to those frequencies absolutely does not mean you will be able to detect the lowpass.

The MP3 designers did not design the format to go > 16kHz (which is why coding higher is so inefficient) because there was no evidence moving the lowpass higher resulted in an audible improvement to the average person at the time the standard was designed (I don't know if that is still the case).

I know for myself that I find distinguishing a 16kHz lowpass very hard, unless it's just after waking up.

--
GCP

How high can you hear?

Reply #29 – 2002-04-01 14:19:00

Quote

Originally posted by Frank Klemm
This thread is complete nonsense and shows perfectly that some people do not understand
backgrounds of hearing.

There's no cutoff. ATH increases continuously
with frequency (for high frequencies).

sorry, Frank, but I can't see how this would contradict my post (which you have quoted)

How high can you hear?

Reply #30 – 2002-04-01 15:20:24

Quote

Originally posted by vladimirovich
who knows? Hearing is very non-linear system, maybe we could not hear 22Khz sine, but could distinguish 11Khz sine from 11Khz triangle tone. the problem is that I can experement by myself, I just have no proper equipment...

Experiment done and commented at http://www.musicplayer.com/ubb/Forum3/HTML/000822-29.html page 29

I've played with GFB and speakers in physics degree, and can confirm that a 14 kHz square wave make a completely different noise than a 14 kHz triangle, or sine.

I also run the experiment with much more care at home on my hifi.

Result : the difference previously heard at the labo was just distortion.

I'm unable to hear any difference between a 6000 Hz sine and a 6000 Hz square (which is 6000 Hz sine + 18000 Hz sine + 36000 Hz sine...), though the cheap microphone of my soundcard can perfectly record it.

Experimental setup : http://pageperso.aol.fr/Lyonpio2001/ultra.htm

In the first link above, we tried to hear intermodulation between audible and inaudible frequencies (at 12 kHz for a 6 kHz square wave). It's easy when the frequencies are melted into a mixer, but as soon as we play them into two separate speakers, we can't hear any intermodulation.

I've just ran the test again on my new hifi, writing this post.
14 kHz+18kHz easily audible difference with 14 kHz
14 kHz left channel + 18 kHz right channel. The difference nearly vanishes... the intermodulation of this ampli is very high
I still can hear a little difference, but it can well come from interchannel intermodulation into the ampli, the test must absolutely be run with two separate devices, one generating the audible sounds, the other generating the unaudible ones, otherwise, all we hear is audible distortion !

How high can you hear?

Reply #31 – 2002-04-01 15:38:43

Quote

Originally posted by Garf
You are most likely hearing distortion from you equipment, not the actual 22kHz wave. A CD player will lowpass at 20kHz, so it is not possible to perform hearing tests above that anyway.

The same (distorting equipment) could very well be true of any poster in this thread, including myself.

I've measured the frequency response of my Yamaha CDX-860 Cd player (450 €) with sinewaves
It plays 20 kHz @ -0.5 db
It still plays 22 kHz with a level difficult to measure, in the -2 db range.

Here's the whole experiment : http://pageperso.aol.fr/Lyonpio2001/cdr/cdr.htm

Here's the 22 kHz picture :

The level is uncalibrated because of the spectrum analysis process. It's the recording of the CD Player analog output into the Marian 24/96 soundcard. Compared to the analysis of the original wav that was burned on CDR, the peak doesn't appear attenuated.
This picture means nothing by itself, please read the experiment if you want to comment on the method used.

According to the whole experiment, there is no problem with CD players and DACs. Distortion may only occur in amplis and speakers.

How high can you hear?

Reply #32 – 2002-04-01 16:22:35

Age: 33
Cutoff: 16.5 kHz

How high can you hear?

Reply #33 – 2002-04-01 16:33:08

I don't know if I'd put too much credence in the tone sweep thing. Probably better is using the various lowpassed music samples to see where you can't hear a difference. And better still is to measure the curve across frequency where you just barely start to hear a sound (your ath). With the last method you get to see how your hearing falls off over frequency instead of looking for a magical cutoff.

If I were to generate my own ath test, I'd probably generate repeating tones which increase in volume at the rate of 1 dB per second, at each frequency of interest (starting from perhaps 10 kHz and incrementing in 2 kHz steps to 20 kHz). Using a sample rate of 48 kHz is probably better than 44.1 to get better sound near 20 kHz.

ff123

How high can you hear?

Reply #34 – 2002-04-01 19:53:06

Quote

Originally posted by Garf
I would really urgently urge you to do a blind test and see if you can reliably distinguish a 19kHz lowpass from a 20khz lowpasss on music. I would be surprised if you could do it. You have made several statements like this before and I am still waiting to see any evidence they are true.

i've made many tests with winamp random function on a playlist with the same song as mp3 --alt-preset standard/--alt-preset standard --lowpass 20/--aps -k/mpc -standard/xtreme/insane/wav/flac/ape...

i also burned them to a cdr and play these files on my cdplayer in living room/studio. i checked these files under windoose using winamp/mediaplayer and under linux using xmms and nautilus. with a friend we "played" with these files and let advise the other with wich codec the file was encoded.

i don't use this abx utility because i can't find windoose binaries yet. you don't have to be stubborn. i know what i hear! and if i say "i can hear an difference between an full range/20kHz lowpassed and 19/16 or whatever kHz lowpassed signal" then this is a fact. i'm not an mp3 kiddie who downloads 5Gb music every day and hears them on €20 speakers.

there are people who can hear differences between low-cost and high-end cdr's [sony/traxdata etc.]. these guys are not mad. it is also true that an pressed cd sounds different than a burned cd. i had situations in studio who i had to lower midrange frequencies on guitar -0,5 db so the keyboard sounds transparent.

on the other hand i know that serious musicans use 128kb mp3 samples in they're songs and no one hears it when byuing this cd.

best regards
Dezibel

How high can you hear?

Reply #35 – 2002-04-01 21:14:22

I can hear +20kHz, although I cannot hear much between 17-18kHz for some strange reason... I believe it has something to do with the fact that my right ear's "auditory/Eustachian tube" is much more narrow than my left ear's... I´ll have to "consult"/contact a doctor later on... :/

And also, I´m 21...

How high can you hear?

Reply #36 – 2002-04-01 21:34:30

age : 35 ; cut-off for tone ( sinus ) : 16.5 KHz

cut-off for transients : about 80 KHz

I still do believe its essential for a Hi-End chain to be able to playback frequences of up to 80 KHz ... not because you hear 'sound' or 'tones' at this frequencies but because i believe our sound receiver organs ( must not be inner ear !! ) can reproduce sound location very precisely by analyzing transients in the >50 KHz frequency domain .... but i know Frank will jump on me now as he did in the German newsgrous years ago ... and i cant really discuss with him due to lack of deep background information that he truely has ...

How high can you hear?

Reply #37 – 2002-04-01 22:18:10

What do you call a "80 kHz transient" ?

How high can you hear?

Reply #38 – 2002-04-01 23:13:37

Quote

Originally posted by Dezibel
i don't use this abx utility because i can't find windoose binaries yet. you don't have to be stubborn. i know what i hear! and if i say "i can hear an difference between an full range/20kHz lowpassed and 19/16 or whatever kHz lowpassed signal" then this is a fact. i'm not an mp3 kiddie who downloads 5Gb music every day and hears them on €20 speakers.

Try: http://www.pcabx.com/program/ABX173_setup.exe

I'd like to see your results. No cheating then please..

How high can you hear?

Reply #39 – 2002-04-01 23:44:51

Quote

Originally posted by ff123
I don't know if I'd put too much credence in the tone sweep thing. Probably better is using the various lowpassed music samples to see where you can't hear a difference. And better still is to measure the curve across frequency where you just barely start to hear a sound (your ath). With the last method you get to see how your hearing falls off over frequency instead of looking for a magical cutoff.

If I were to generate my own ath test, I'd probably generate repeating tones which increase in volume at the rate of 1 dB per second, at each frequency of interest (starting from perhaps 10 kHz and incrementing in 2 kHz steps to 20 kHz). Using a sample rate of 48 kHz is probably better than 44.1 to get better sound near 20 kHz.

ff123

I agree. I was going to point this out, and even send people to ff123's site for the link to the guy with these type of test files. Though I think to do it well, you want to make many of them, closer together in frequency, like ff123 is saying.. and make one down at 3-3.5Khz (in the most sensitive area) with the same steps to "calibrate" your threshold, like stated in the referenced link.

The problem I have with the wav everyone is using, is that it goes too fast/its too short, it's hard for me to be exact on what frequency does it "fade out" for me... a half second error is a lot of frequency change.

But I think the approach of fixed tone files, diminishing in amplitude would be more "accurate".

Another thing to play with: http://www.marchandelec.com/fg.htm a little free frequency generator. With this it's easy to run up and down the scale, and increase and decrease amplitude with some relative accuracy, (if not absolute accuracy)

You can hear by using the other waveforms, the difference in harmonics generated by a non-sine waveform. That's another point in this to consider. If you don't have a pure sine wave, you will generate harmonics above and below the actual frequency, and may be tricking yourself into thinking you are hearing the fundimental when you're not. Also, you can generate harmonics if you drive your speaker/headphone/amp hard enough that it becomes non-linear, even if you use a pure tone source.

And finally, you may not know how accurate your speaker/headphone actually is. It might be rolling off the high end too, which will make your ear's rolloff look worse. As well as the mentioned problem of playback via wav files that are sampled at some frequency. The closer to the sampling frequency you get, (at 44Khz, a 22Khz tone has only 2 data points to get smoothed into a sine wave!) the more distortion occurs, creating harmonics to cloud the picture.

It gets tricky.. this measuring stuff.

And, for the record, I roll off real fast after 14.5-15Khz :-( but I can still detect louder things up to 18Khz or so, possibly better than at 16)
Age is 43 (vet of some pretty loud concerts and years of convertible driving)

How high can you hear?

Reply #40 – 2002-04-02 00:22:25

Hmmm.... I retested myself, and want to remove my lowest catagory poll entry!

First of ff123, is the sweep file on your site the same one everyone is using?

I realized that I had used my computer speakers long ago when I did my test, (tsk, tsk) and tried it just now with my new pair of Koss KSC-35 headphones, and a stopwatch.

I'm consistiantly getting about 3.2 seconds, which should put me about 15.5Khz. :-) I feel a little better...

Though playing with the freqency generator I mentioned in the last post, I still see my personal "rolloff" speeding up fast, around 14-14.5Khz.

How high can you hear?

Reply #41 – 2002-04-02 00:38:25

Quote

Originally posted by JonPike

And finally, you may not know how accurate your speaker/headphone actually is. It might be rolling off the high end too, which will make your ear's rolloff look worse.

Yes, this is an important point, also for soundcards. The Creative Labs Soundblaster Live! and Audigy series are not suitable at all, because they attenuate above 16 KHz. (see this thread i posted to 3DSS a while ago). IIRC the attenuation goes as far as -9 dB at 19.5 KHz.

How high can you hear?

Reply #42 – 2002-04-02 00:44:15

Oh, one more note.. as the sweep went up to the high end, and after I wasn't hearing anything, I heard downward sweeping tones, two sets of them. That must be the lower half of the aliasing distortion, caused by the 44.1Khz sampling rate. A fault of using this kind of setup (digitized wave file played back with DAC) but you can tell it's not the intended tone.

How high can you hear?

Reply #43 – 2002-04-02 11:23:38

Quote

Originally posted by JonPike
Oh, one more note.. as the sweep went up to the high end, and after I wasn't hearing anything, I heard downward sweeping tones, two sets of them.

This occurs with tones badly generated. I got the same effect (with countless aliases) generating a fixed sine in SoundForge 4.5 and applying a pitch bend effect. The sonogram shows a beatiful net of countless curves going up and down.
FF123's sample, however, is clean. The analysis shows one sweep only.
Those aliases may come from a poor resampling in the soundcard too, but I don't remember the SB live being so bad. I don't think it generates aliases (mine is no more in use).

How high can you hear?

Reply #44 – 2002-04-02 12:29:18

Quote

Originally posted by Pio2001
Those aliases may come from a poor resampling in the soundcard too, but I don't remember the SB live being so bad. I don't think it generates aliases (mine is no more in use).

SB Live! and Audigy both have this aliasing problem. Users with those cards should resample signals to 48kHz or use kX Project drivers.

How high can you hear?

Reply #45 – 2002-04-02 12:59:33

Quote

Originally posted by Dezibel

i don't use this abx utility because i can't find windoose binaries yet. you don't have to be stubborn. i know what i hear!

Yes, I'm pretty sure that you know what you think you hear. That's why we're asking for exact ABX results.

And please use the original wave and a lowpass filter that can do antialiasing. Using lame -k for example is not a good test - it will introduce distortion of its own.

--
GCP

How high can you hear?

Reply #46 – 2002-04-02 14:32:28

Quote

Originally posted by JonPike
Oh, one more note.. as the sweep went up to the high end, and after I wasn't hearing anything, I heard downward sweeping tones, two sets of them. That must be the lower half of the aliasing distortion, caused by the 44.1Khz sampling rate. A fault of using this kind of setup (digitized wave file played back with DAC) but you can tell it's not the intended tone.

When playing f, a DAC also outputs some signal of the frequency
of fs-f (f=44.1 kHz, fs=20 kHz, f-fs=24.1 kHz). f and fs-f
normally mix in amplifiers and loudspeakers and the frequency
fi = fs-f - f = fs-2f occures (in the example above 24.1 kHz-20 kHz=4.1 kHz). Depending on the steepness of the antialiasing filter a DAC is not suitable to generate high quality signal above
fs/3. ALL DACs I tested at 44.1 kHz are not usable to generate
a 18 kHz tone, because the generated 44.1 kHz-2*18 kHz = 8.1 kHz is much louder. Pioneers have problems with 16 kHz, the
44.1 kHz-2*16 kHz = 12.1 kHz is much louder than the 16 kHz.

Another problem are normal power amplifiers. The hum increases
with the signal. You don't hear the 30 kHz tone, but the hum
pumping is clearly audible even with US$ xx000 amplifiers.
The only way to prevent this is to use headphone amplifiers.

How high can you hear?

Reply #47 – 2002-04-02 21:38:09

Made some sine waves in goldwave's expression evaluator. Used 100hz increments, ,I could hear up to 19300Hz clear as day (sounded like my television). After 19300, It got less audible and I had to turn my speakers up to hear much. It hurt my ears also.

How high can you hear?

Reply #48 – 2002-04-02 21:40:31

How loud are your 8 kHz tones ? How did you measure them

I've analyzed the analog output of a CD Player, and after some careful adjustments, I got perfectly clean sines until 20 kHz, then some distorded sines until 22 kHz

http://pageperso.aol.fr/Lyonpio2001/cdr/cdr.htm

With the Yamaha CDX-860, the main distortion are two tones placed at -/+ 4 kHz from the sine (eg : playing a 16 kHz sine, I get also a 12 kHz and a 20 kHz)

Their level is 86 db under the main frequency, that is -6 db full scale ! Therefore they can well be some artifacts due to the spectrum analysis process (they would be less than two steps peak to peak).

Aliases appear at 20 kHz :

20 k : -74 db @ 24 kHz
21 k : -31 db @ 23 kHz
22 k : 0 db @ 22.1 kHz

Making this experiment, I noticed that the main source of aliases was the tone generator itself in SoundForge or CoolEdit.
It is especially difficult to generate a pure sweep, since the program uses low end resampling techniques to sweep the tone.

The second problem come from the recording : you can well generate a lot of aliases if you digitally record a high frequency in order to analyze it. that's why I recorded at 96 kHz sampling rate.

And also, the spectrum analysis process can generate some alias too, I noticed some little alias shown in SoundForge that weren't shown in CoolEdit analysing the same file.

Always work at 96 kHz, there's no way otherwise.

How high can you hear?

Reply #49 – 2002-04-03 01:21:26

Quote

Originally posted by Pio2001
How loud are your 8 kHz tones ? How did you measure them ???
I've analyzed the analog output of a CD Player, and after some careful adjustments, I got perfectly clean sines until 20 kHz, then some distorded sines until 22 kHz

http://pageperso.aol.fr/Lyonpio2001/cdr/cdr.htm

With the Yamaha CDX-860, the main distortion are two tones placed at -/+ 4 kHz from the sine (eg : playing a 16 kHz sine, I get also a 12 kHz and a 20 kHz)

What quantization operator Q was used to generate the signals you analyzed? How was satisfied that
uncertainty is at least fs bps?

Notice

Poll