[kiteki]

 
A thread about custom IEM's was going off-topic into the hypersonic effect (inaudible frequencies at 20kHz+), I couldn't find any recent or alive threads on the issue so I created a new one.
 
Feel free to voice your opinions a la "it's just marketing hype and pseudo-scientific nonsense, what's the point when you can't even hear it..?" or "it's the apotheostic life-blood of hi-fi, where's my portable SACD player..?" et cetera.....
 
The topic of discussion was mainly revolving around IEM's that extend to 26kHz, and academic essays like this one -> http://jn.physiology.org/content/83/6/3548.full
 
Whatever your take on hypersonics is, it sure is fun to read essays like that one, and part of me wonders if there is a cult following of hypersonic enthusiasts out there in their respective Hi-Fi rooms, and if that niche will ever transgress into portable audio.
 
 
 

 

[tinyman392]

When you have multiple sounds (frequencies) being produced at once, you get what's called a beat (the physics term) where you'll hear a tone over and over again so  long as both frequencies are playing.  This requires 2 frequencies though.  With multiple frequencies you hear a huge number of beats.  As long as the frequency is played, you'll hear the beat (even if you can't hear the original frequency). 
 
Let's say you have a 26000Hz sound and a 22000Hz sound.  You'd hear a 4000Hz (26000 - 22000) beat (audible) faintly.  Now, some beats dig even deeper than that, lets say you have a 26000Hz sound and a 26008Hz sound playing simultaneously, the beat you'd here now is 8Hz (can't be heard, but felt infrasonicly). You can also have beats formed by sonic sounds, like 18000Hz and 1000Hz.  This would result in a 17000Hz beat.
 
Even deeper into the subject, lets say you ahve a 26000Hz sound, 21000Hz sound, 18000Hz sound, you'd hear the following beats: 5000Hz, 9000Hz, 3000Hz sound.  Things get even more complicated when you add in more sounds, that even if not audible, will still form a beat. 
 
To delve even deeper, the timbre of the instrument doesn't go at a straight tone, it varies depending on instrument, this is why you can tell instruments apart.  Since they change constantly, so do the beats.  You get a more textured, riveting sound that you can't get with a 20-20000 area.
 
So what's the use of extending beyond 20000Hz?  You get cool beats.  What's cool about extending below 20Hz, you can get some cool infrasonic beats (like the 8Hz beat earlier).
 
Don't believe me? I took this out of a physics book (but made up my own numbers).  Sorry for going off topic into infrasonic sounds though   I think they should be in this topic as well.
 
PS: it's not hypersonic, it's ultrasonic (that is the correct term).

 

[kiteki]

Is 26 a magic number? If you read the essay they used 26kHz as the frequency cut-off for high and low frequencies in the testing, it's also what the Spiral Ear 5-Way custom IEM apparently extends to, it's also what the JH3A amplifier from JH Audio extends to I think, I guess someone decided 26kHz is the 'true' upper limit of human hearing then, with all things considered.
 
From the essay "26 Japanese volunteers (15 males and 11 females, 18–31 years old) participated in the psychological experiment."
 
More random trivia, 5x5 = 25, 3x3x3 =27.
 
26 is the only number that exists that is directly between a square, and a cube.

 

[tinyman392]

Quote:

Is 26 a magic number? If you read the essay they used 26kHz as the frequency cut-off for high and low frequencies in the testing, it's also what the Spiral Ear 5-Way custom IEM apparently extends to, it's also what the JH3A amplifier from JH Audio extends to I think, I guess someone decided 26kHz is the 'true' upper limit of human hearing then, with all things considered.
 
From the essay "26 Japanese volunteers (15 males and 11 females, 18–31 years old) participated in the psychological experiment."
 
More random trivia, 5x5 = 25, 3x3x3 =27.
 
26 is the only number that exists that is directly between a square, and a cube.

LOL, 26 isn't a random number...  26000Hz was a well rounded number that you cited and would be easy to do math with   That was the main reason it was used.  I doubt 26000 is the true upper limit (I did use 26008Hz in one of my examples ), but I don't see anything going higher than that. 
 

 

[goodvibes]

Wouldn't the theory be that those beat frequencies be produced whether the driver itself can reproduce them directly or or not so there is no need for the extra bandwidth other than perhaps some margin on transient response. Even the transient part is theoretically not valid but I suspect there may be some minor lag of acceleration on a limited bandwidth driver. That a driver needs that bandwidth to reproduce frequencies we can't hear and will exist in use anyway sounds a bit off to me. That the better transient behavior may be audible wouldn't surprise me or that something else finding it's way into a test like filter sound, amp character change with more bandwidth or even the digital end of things becoming more noticeable with more bandwidth introduced. The problem with this sort of test is that as much as you try to keep things equal and they may have, there may be other variables introduced that aren't readily apparent. If the theory is that we need the bandwidth to reproduce natural beats in the audible range, it may be valid but since none of the recorded material we use for playback has those frequencies intact, it's rather moot.

 

[kiteki]

There are of course other variables, but the testing scenario they used to validate the hypersonic effect in the linked essay seems much more well thought out and with a higher degree scientific rigour than the tests I've seen which have claimed to disprove it.
 
I'm a little bit skeptical of both sides of the coin, I feel they put in so much effort (moreso in proving the hypersonic effect) by using all kinds of advanced equipment and going to Bali to record gamelan music (what is that anyway?) with HF data that the university director must have felt disappointed if the brain scans didn't show up anything at all, I mean where did his "hunch" come from? Was he on a holiday in bali and listened to music and felt a special sensation he couldn't recreate listening to CD's? Or was he just trying to make a new format in the Hi-Fi world and then sell it to Sony? that's my skepticism of the initial tests (to be honest I don't understand all the statistical evidence lingo in these academic essays so I can't make a fair distinction), but I'm more skeptical of the "disproving" tests where critics have said they're not using ideal equipment, ideal surroundings, and they're simply "asking" the subjects if they could "hear" any difference! It's as if they were expecting an audible difference along the lines of 192kbps MP3 versus 24/96 lossless, and even in a test like that with random participants simply "asking" them which one sounded better, I'm sure they could still conjure up a stastistical "flipping the coin" result, and die-hard audiophiles would just laugh at it, since the difference is clearly audible with the correct equipment and you can't just ask random university students "which sounds better" when the average joe's typical attitude towards hi-fi is a sub-woofer, or perhaps a $50 headphone with a V-shaped equalizer curve listening to the latest britney spears dub-step remix.

 

[tinyman392]

Quote:

Wouldn't the theory be that those beat frequencies be produced whether the driver itself can reproduce them directly or or not so there is no need for the extra bandwidth other than perhaps some margin on transient response. Even the transient part is theoretically not valid but I suspect there may be some minor lag of acceleration on a limited bandwidth driver. That a driver needs that bandwidth to reproduce frequencies we can't hear and will exist in use anyway sounds a bit off to me. That the better transient behavior may be audible wouldn't surprise me or that something else finding it's way into a test like filter sound, amp character change with more bandwidth or even the digital end of things becoming more noticeable with more bandwidth introduced. The problem with this sort of test is that as much as you try to keep things equal and they may have, there may be other variables introduced that aren't readily apparent. If the theory is that we need the bandwidth to reproduce natural beats in the audible range, it may be valid but since none of the recorded material we use for playback has those frequencies intact, it's rather moot.

 
It is true that these  beats would be audible in the recording.  However, beats are really soft and at a low amplitude (low volume) and are a little harder to pick up on there own (played around with tuning forks to listen to them).  With a pair of headphones that can produce the frequencies required for the beats, you'll get those amplified a little more so you can hear them.  It adds a little more detail and mix into the signature.  Not everyone will like them, but it does add something unique to a pair of headphones.
 
 

 

[goodvibes]

Most recordings are multi mic-ed so no, those natural beats would not be in them to the extent described. I've done live stereo pair recordings and they should then exist but really if so, the next time you would play the recording with the extra bandwidth it would just create more and be again unnatural if you could hear them, at least in free air. In an IEM, there is effectively no beating so maybe there's a binaural application for a perfect simple recording done at 24/96 with mics that go that high etc. etc. etc.
 
Basically, there's something wrong with everything. IEMs are a bit unnatural. Multi way speakers with good phase from 1st order networks beat more than steeply sloped ones, etc. etc. It's always trade offs and execution. I'm aware of many technical aspects of sound and how to measure but I get a bit perturbed when somebody finds that meaningful measurement or insists on others to prove something they've heard with measurement. It's really too complex for that.

 

[kiteki]

For anyone with a spare hour on their hands for reading up on the hypersonic effect, here's some articles for you.
 
Here is an essay on the studies: http://jn.physiology.org/content/83/6/3548.full
Here is another one: http://www.cco.caltech.edu/~boyk/spectra/spectra.htm
 
Here is the wikipedia entry http://en.wikipedia.org/wiki/Hypersonic_effect "Attempts to independently reproduce these results have so far been unsuccessful.^{[size=x-small][4][/size][size=x-small][5][/size]}"
 
If you click on 4 and 5, you are linked to these articles:
4. http://www.hificritic.com/downloads/Archive_A10.pdf
5. http://www.nhk.or.jp/strl/publica/labnote/lab486.html
 
 
 

 

[a_recording]

I bring this up quite often just because of my love for my Akira blu-ray and the slightly mad scientist theories about the hypersonic effect in the accompanying booklet. Sadly though even in the original papers they weren't able to find a measurable 'hypersonic' effect when using headphones.

 

[kiteki]

Hehe right! It only works with speakers.
 
I should pick up Akira on blu-ray.
 
 
EDIT: You could watch Akira on blu-ray with headphones and the TakeT BPP..?

 

[gregorio]

Let's say you have a 26000Hz sound and a 22000Hz sound.  You'd hear a 4000Hz (26000 - 22000) beat (audible) faintly.

No you wouldn't. For the ear to hear a beat frequency it must be able to hear both tones. So if you had two sine waves with frequencies of 15kHz and 25kHz, you will not hear an additional 10kHz tone you will only hear a single 15kHz (providing you are able to hear up to 15kHz).

Things get even more complicated when you add in more sounds, that even if not audible, will still form a beat

The beat is always lower in amplitude (volume) than either of the two tones. So if you can't hear one of the two tones, you definitely can't hear the beat frequency.

To delve even deeper, the timbre of the instrument doesn't go at a straight tone, it varies depending on instrument, this is why you can tell instruments apart.  Since they change constantly, so do the beats.

No, you are mixing up two totally different audio processes. A single note played on an instrument does not produce a single audio frequency, it produces a fundamental and then a range of harmonics at different frequencies, the balance of all the harmonics gives the instrument it's timbre and is how you can tell a trombone from an electric guitar even if they are playing the same note. Harmonics are a different phenomenon and not directly related to beat frequencies.

Beat frequencies can only be produced by non-linear devices, like amps and synths. If you create a beat frequency in a synth, say you modulate a 25kHz and a 15kHz wave, you will get a 25kHz wave, a 15kHz wave and a 10kHz wave. If we record this using a 96kHz sample frequency and our amp and speakers are capable of replaying the 25kHz wave, what we would hear is a 15kHz wave and a 10kHz wave. If the amp or speakers only have a frequency response up to 20kHz then the 25kHz signal will not be reproduced and all we would hear would be a 15kHz wave and a 10kHz wave (unless the 25kHz signal has caused the amp or speakers to distort!). If we record the synth output at 44.1kHz sampling rate, the 25kHz wave will be filtered out by the anti-alias filters and what we would hear is a 15kHz and a 10kHz wave. Anyone spot the difference between these three? .... Good, because there isn't one!

So what's the use of extending beyond 20000Hz?  You get cool beats.

No you don't. There is no benefit to extending beyond 20kHz!

I've heard the argument of can we hear over 20kHz many times over the last 20 years or so. The answer is NO, there is nothing above 20kHz which you can hear. In fact, unless you are a young child, there is probably nothing above 16-17kHz which you can hear and if you're in your 50s you are probably down to 12-13kHz. If there are signals above 20kHz which affect other frequencies in the hearing range then we can record, reproduce and hear those interactions but we can never hear the ultrasonic frequencies themselves. Also consider that even a high C on a piccolo flute is only 4.2kHz pretty much everything above this frequency is just harmonics of generally diminishing amplitude.

The other argument I've heard is that we might not be able to hear beyond 20kHz but maybe we can perceive it somehow. Sure, this is easy. Say I played a 30kHz sine wave at you at 200dB. Your skin would peel away from your body and a fraction of a second you would be dead but for that fraction of a second I'm pretty sure you would "perceive" 30kHz!!

You have to also consider that none of the common studio recording microphones are capable of capturing frequencies above 20kHz and most speakers also dramatically roll off after 20kHz. So it's very unlikely you could find many recording which contained anything but noise above 20kHz anyway.

Having said all this, there is a potential benefit to recording at 96kHz sampling frequency. It is easier (read cheaper) to create good quality filters inside the converters using higher sample frequency. There is a point of diminishing returns though, a sample rate of 192k has more noise and distortion than the 96k sampling rate. Recordings made on high quality professional converters at 44.1k still provide the highest level of audio quality.

G

 

[khaos974]

Suppose that you all a panel of listeners all incapable of hearing beyond 20 kHz, they could still pass an ABX test with the following 2 signals: nothing and plenty of ultrasonic content.
There could be plenty of inter modulation distortion that is below 20 kHz, and that would be equipment related issues.

 

[kiteki]

^ That's why one of the studies I linked to above presented the information to the participants like this
 




As you can see there's a dip between 20kHz and 23kHz or something like that, this was an ABX study where they asked the particpants which was which, they concluded the results in the end were not much better than flipping a coin, if you check this study they are not asking the participants which recording was which, they are doing an MRI brain scan.
 
Without another test using the EEG and MRI, the original test has not yet been replicated, so I fail to see how they wikipedia entry is correct:
 
 "Attempts to independently reproduce these results have so far been unsuccessful"
 
 
Actually on second thought it is correct, the attempts have been unsuccessful, I mean the attempting part!
 
The wikipedia entry links to the ABX study as the source for it's claim, and then it links to another article which comes to the final conclusion the we should use the widest bandwidth possible in recordings for the sake of fidelity and archival purposes!
 
I'm not really taking either side here, I'm just pointing out the wikipedia entry is incorrect.
 
 
 

 

[gregorio]

Suppose that you all a panel of listeners all incapable of hearing beyond 20 kHz, they could still pass an ABX test with the following 2 signals: nothing and plenty of ultrasonic content.
There could be plenty of inter modulation distortion that is below 20 kHz, and that would be equipment related issues.

So in other words a panel of adult human beings! Yes, they could pass the test but you would not be testing for the ability to hear ultrasonic content, you would be testing for the ability to hear distortion within the hearing range of 20Hz to 20kHz!

Kiteki:- The Oohashi study has become infamous and a real pain in the audio world. When it became apparent that the Oohashi experiment produced non-repeatable results (a basic criteria for scientific evidence) it's methodology was examined and found to be severely lacking in several areas. Due to these facts and the large amount of evidence (repeatable) the Oohashi paper, it has been discredited. The Boyt paper you linked to did not provide any evidence of hearing beyond 20kHz. With regard to hearing Boyt just referenced the Oohashi paper, Boyt's paper was written before Oohashi's paper have been discredited.

To date there exists no reliable scientific evidence that humans can hear beyond 20kHz but a substantial body of scientific evidence exists which indicates that we can't.

G