[skyscaper]

Since normal human hearing is the range 20Hz-20KHz, why do some headphones have a frequency response up to 30kHz and beyond?

[NicolasB]

Because humans can perceive sound components well above 20kHz.

If you play a sine-wave signal at 30kHz, you can't hear anything. But if you have something like a snare-drum hit, the sound goes from nothing to about 130dB in almost no time at all. If you apply a 20kHz filter to the signal, you lose the steepness of the initial rising edge - and that does actually make a perceptible difference to the sound.

This why some hi-def audio formats sample at 96kHz or even 192kHz (thus reproducing elements of the recording up to 48 or even 96 kHz accurately) - if the sound system is capable of reproducing ultrasonics, there is an audible difference. If there wasn't there'd be no point in higher bitrates.

It's also partly just a question of "headroom". If the headphones are going to do a good job of reproducing stuff at 20kHz, they need to be doing something at 30kHz; if the cut-off is too low it has a bad effect on parts of the signal that just squeeze in under the bar.

[severnsource]

Originally Posted by NicolasB

Because humans can perceive sound components well above 20kHz.

If you play a sine-wave signal at 30kHz, you can't hear anything. But if you have something like a snare-drum hit, the sound goes from nothing to about 130dB in almost no time at all. If you apply a 20kHz filter to the signal, you lose the steepness of the initial rising edge - and that does actually make a perceptible difference to the sound.

This why some hi-def audio formats sample at 96kHz or even 192kHz (thus reproducing elements of the recording up to 48 or even 96 kHz accurately) - if the sound system is capable of reproducing ultrasonics, there is an audible difference. If there wasn't there'd be no point in higher bitrates.

It's also partly just a question of "headroom". If the headphones are going to do a good job of reproducing stuff at 20kHz, they need to be doing something at 30kHz; if the cut-off is too low it has a bad effect on parts of the signal that just squeeze in under the bar.

You are joking aren't you?

If you're under 20 and haven't ruined you're hearing with too many clubs or personal stereos it's possible that you may be able to hear a little higher than 20K, but not by much and only at high SPLs. You are saying that people can detect ultrasonics even tho' they can't hear them; although I have seen the claim made I have never seen any reliable validation of the claim even though it shouldn't be very difficult to prove. Hence it is reasonable to take the claim as being false. Even if it were true the effect would be at such a low level as to be unimportant to musical perception.

If you are relying on commercial music sources there won't be any HF information above 22KHz. All CDs are produced with a steep low pass filter before mastering. Low bitrate compressed formats will have a cut-off frequency lower than that.

The so-called high def. formats were originally proposed as an ideal, to ensure that there were absolutely no limitations on sound quality imposed by the distribution medium. If you have a higher sampling frequency any passband ripples caused by the steep cut anti-aliasing filters will be well outside the audio band. It is possible that the passband ripples of CD anti-aliasing filters may be audible to some people.

Even if your material came from a high-def source it is extremely unlikely that there would be much ultrasonic information on it, as most microphones used in commercial music production don't have a good enough HF response to be able to record it.

The answer to the original question is marketing. Bigger numbers look better.

Bill

[skyscaper]

Originally Posted by NicolasB

If you play a sine-wave signal at 30kHz, you can't hear anything. But if you have something like a snare-drum hit, the sound goes from nothing to about 130dB in almost no time at all. If you apply a 20kHz filter to the signal, you lose the steepness of the initial rising edge - and that does actually make a perceptible difference to the sound.

So, this would suggest that the Grado SR80s aren't so good, having a frequency response of only 20Hz-20KHz?

[AndyC_772]

Originally Posted by severnsource

The answer to the original question is marketing. Bigger numbers look better.

Bill

I don't agree with this - it's not marketing, it's science. Produce a transducer with a flat, low distortion response up to 20kHz, and chances are that its -3dB point will be at 30kHz or higher.

Whether or not the fact that it responds between 20kHz and 30kHz is useful, or whether that response is accurate, is neither here nor there.

[NicolasB]

Originally Posted by severnsource

If you're under 20 and haven't ruined you're hearing with too many clubs or personal stereos it's possible that you may be able to hear a little higher than 20K, but not by much and only at high SPLs. You are saying that people can detect ultrasonics even tho' they can't hear them;

No. I'm saying you can hear ultrasonic transients even though you can't hear a steady, ultrasonic sinusoid - or, more precisely, the presence of ultrasonic transients makes a perceptible difference to the sound, while the presence a long-term, non-varying ultrasonic played in isolation is not audible.

To put it another way: you can hear a difference between a 15kHz sinusoidal signal and a 15kHz square wave.

There have, I think, also been some claims that the ear may be able to detect beat-frequencies between multiple individually inaudible ultrasonic sounds if those beat frequencies lie within the normally audible range - I'm less familiar with that angle.

although I have seen the claim made I have never seen any reliable validation of the claim even though it shouldn't be very difficult to prove.

Actually it's quite tricky to prove because there's no way to create a signal that only contains the transient component. Testing whether people can hear a steady-state sinusoid is easy: you play the note and they either hear it or they don't. The only way you could test the effect of transients is to take two real recordings, one that contains ultrasonic components and one that doesn't, and then test to see if listeners can consistently hear a difference between them - which, inevitably, is a far more subjective question and far more prone to placebo effect.

Hence it is reasonable to take the claim as being false. Even if it were true the effect would be at such a low level as to be unimportant to musical perception.

That's an impressively vast assumption to make on the basis of no data whatever.

[severnsource]

Originally Posted by AndyC_772

I don't agree with this - it's not marketing, it's science. Produce a transducer with a flat, low distortion response up to 20kHz, and chances are that its -3dB point will be at 30kHz or higher.

That's a big leap of faith. If such a transducer had been produced I would agree, unfortunately it is extremely difficult to produce a transducer with a smooth response to beyond 20KHz without significant resonances. It is even more difficult to produce such a response in a headphone partly because the interaction between the transducer, its enclosure and the listeners ear is completely unpredictable. Reliable and consistent measurement of earphones is difficult at relatively low frequencies let alone above 10kHz.

Most headphone specs don't give amplitude limits, a claim of a response to 30kHz without specifying amplitude is meaningless, so any such spec will be primarily marketing speak.

Bill

[severnsource]

Originally Posted by NicolasB

No. I'm saying you can hear ultrasonic transients even though you can't hear a steady, ultrasonic sinusoid - or, more precisely, the presence of ultrasonic transients makes a perceptible difference to the sound, while the presence a long-term, non-varying ultrasonic played in isolation is not audible.

To put it another way: you can hear a difference between a 15kHz sinusoidal signal and a 15kHz square wave.

I trust you are aware that the lowest harmonic component of a 15kHz squarewave is 45kHz? A 15kHz squarewave having passed through any common digital system will become a 15kHz sinewave. So even if you were correct in your assertion it is irrelevant in the real world.

Originally Posted by NicolasB

Actually it's quite tricky to prove because there's no way to create a signal that only contains the transient component. Testing whether people can hear a steady-state sinusoid is easy: you play the note and they either hear it or they don't. The only way you could test the effect of transients is to take two real recordings, one that contains ultrasonic components and one that doesn't, and then test to see if listeners can consistently hear a difference between them - which, inevitably, is a far more subjective question and far more prone to placebo effect.

Actually it is quite easy to test. Make a recording with your super wide bandwidth microphone into your 24/96 PC card, make a copy of it and filter to your hearts content in Audition or similar and play the two files through ABX. You'll soon find if you can hear the difference or not.

Originally Posted by NicolasB

That's an impressively vast assumption to make on the basis of no data whatever.

Actually it is the application of Occam's razor. There is an awful lot of knowledge about the limits of hearing accumulated over decades. None of the reliable data indicates that there is any value in extending the HF response much beyond 20kHz.

Bill

[severnsource]

Originally Posted by skyscaper

So, this would suggest that the Grado SR80s aren't so good, having a frequency response of only 20Hz-20KHz?

You shouldn't jump to that conclusion. Grado may just be being more conservative (realistic?) in their specs. Headphones are one of those items were specifications are more or less completely valueless. You need to listen to them and compare. The frequency response in band is vastly more important than having a theoretcal ability to reproduce 30kHz. Many people like Grado cans.

Bill

[skyscaper]

Originally Posted by severnsource

You shouldn't jump to that conclusion. Grado may just be being more conservative (realistic?) in their specs. Headphones are one of those items were specifications are more or less completely valueless. You need to listen to them and compare. The frequency response in band is vastly more important than having a theoretcal ability to reproduce 30kHz. Many people like Grado cans.

Bill

Indeed .

[NicolasB]

Originally Posted by severnsource

I trust you are aware that the lowest harmonic component of a 15kHz squarewave is 45kHz?

I meant 9kHz, sorry.

A 15kHz squarewave having passed through any common digital system will become a 15kHz sinewave. So even if you were correct in your assertion it is irrelevant in the real world.

But you recall that the reason why this thread began is that the original poster was asking why certain devices have a quoted frequency response that goes well above 20kHz. In fact, there are some devices go a lot higher than that. You can buy stand-alone "super tweeters", for example, which extend the upper range of conventional loudspeakers in the same way that a subwoofer extends the lower range - typically up to 100kHz.

What you might use as source material is another story, of course but the choice of 100kHz is to allow you to resolve all of the information in a 192kHz digital recording.

Actually it is quite easy to test. Make a recording with your super wide bandwidth microphone into your 24/96 PC card, make a copy of it and filter to your hearts content in Audition or similar and play the two files through ABX. You'll soon find if you can hear the difference or not.

Yeah, but not that many people have a 192kHz-capable microphone, or a high-quality supertweeter setup.

There is an awful lot of knowledge about the limits of hearing accumulated over decades. None of the reliable data indicates that there is any value in extending the HF response much beyond 20kHz.

The problem is that all of the research has been done by people who make exactly the same assumption that you're making. It's a natural enough assumption: people are used to dealing with long-duration wave-forms, and Fourier analyses.

But the ear isn't a linear device: it picks up sum and difference frequencies, for example. If you play 100Hz and 150Hz sinusoidal signals together and analyse the result, there's no component at 50Hz. But you hear a 50Hz component when you listen to it. You also hear tones corresponding to 250Hz, 350Hz, etc. despite there being no harmonics of the original signals at those frequencies. The fact that the ear responds differently to an ultrasonic component that is present for only a fraction of a cycle to the way it responds to a wave of the same frequency that persists for many cycles is not really that surprising, but it's only very recently that it has occurred to anyone that it would be worth checking - and, indeed, only very recently that there has been any possibility of acquiring both a recording and playback equipment that are actually capable of reproducing ultrasonic frequencies.

Perhaps an interesting thing to consider is the design philosophy for SACD - why did Sony's engineers bother to make it capable of reproducing information up to 100kHz or higher if there is no benefit in doing so?

[AndyC_772]

Originally Posted by NicolasB

Perhaps an interesting thing to consider is the design philosophy for SACD - why did Sony's engineers bother to make it capable of reproducing information up to 100kHz or higher if there is no benefit in doing so?

An even more interesting thing to consider is whether SACD's response at high frequency is in practise actually anywhere near its manufacturers' claims. Here is a very interesting article which explains some of the science behind SACD and DVD-A, and suggests that SACD has severe inherent limitations.

My own belief is that SACDs sound good - terrific, even - because it's relatively simple and straightforward to design a really good DSD converter. A really good PCM DAC is a complex product with numerous technical challenges to overcome, problems which DSD simply doesn't have. The benefit to the consumer, therefore, doesn't so much come from the fact that the format is inherently superior, but that it's easier in practise to produce a good quality player.

DVD-A is another question entirely. But, given that PCM is inherently much more space efficient than DSD, and the enormous storage capacity required for video has already been made available, then why NOT allow the DVD-A standard to sample up to 192kHz/24 bit? Anything less would just be wasting space, but it doesn't actually mean all those bits are audible.

They do keep some of the naysayers quiet, though, which is a very worthwhile reduction in background noise