Hearing & Hi-Fi Specs

[Aitch]

While we may not be able to hear frequencies above 20kHz directly, the way higher audio frequencies may impact upon lower, audible frequencies, may have subtle colouring effects.

OK, the above is something that hi-fi bods come out with regularly, as a reason for buying expensive hi-fi or keeping their MP3 files at extremely high sampling rates. But I don't see how it can be true.

Assuming a high frequency tone affects a lower one, the resulting sound can, like the original sound, be represented as a collection of sine-waves (Fourier analysis?) including the ones in the affecting tone. So, if you can't hear those higher tones, what you can hear is only the collection of sine-waves that make up the original sound. That is, it is unaffected.

I'm not an expert on acoustics/mathematics, so is anyone here able to confirm or confound this argument? Thanks.

BTW in my case it is more like 13-14KHz rather than 20, which makes it more relevant in real life.

 

[TheDaver]

By all means, allow me to confound things!

I think you’re talking about using a high lowpass filter value or none at all – not increasing the sampling rate.

CDDA (CD-audio) is itself usually lowpassed at 20kHz or so during mastering, to eliminate aliasing. Forcing your MP3 encoder to include frequency bands that aren’t even active is kinda pointless. And MP3 is a lossy encoder, so actually, the less frequency bandwidth it has to encode, the better. Some people insist on cutting frequency content above 16kHz even at higher bitrates.

 

[Aitch]

By all means, allow me to confound things!

I think you’re talking about using a high lowpass filter value or none at all – not increasing the sampling rate.

CDDA (CD-audio) is itself usually lowpassed at 20kHz or so during mastering, to eliminate aliasing. Forcing your MP3 encoder to include frequency bands that aren’t even active is kinda pointless. And MP3 is a lossy encoder, so actually, the less frequency bandwidth it has to encode, the better.

As per usual, I explained in an ambiguous way

The people on the board I was referring too WERE talking about sampling rates, not filters, not low-pass filters (it's a music board - lots of music experts, even the odd known composer) and whether they should keep there music files (possibly not MP3) at high bit rates to keep the high frequencies.

Some people insist on cutting frequency content above 16kHz even at higher bitrates.

Ah, remember the days when stereo radios had a button to filter out the 19KHz pilot tone? Or are you all too young?

 

[a_unique_person]

JJ knows all when it comes to topics like this. I don't know if he will add to this topic when it is clearly evident that high frequencies don't change low frequencies. It's like saying large rocks can alter how grains of sand come through a sieve.

 

[casebro]

I know that air molecules can vibrate in only one direction at a time. Same holds for speaker cones.

So, if two out of sync sine wave have peaks and valleys concurrent, they ought to cancel each other out, making the one pulse lower volume. (That is the way sound-canceling head phones work) If the two waves peak at once, maybe the one peak will be higher volume? Of course, it would happen erratically if the sound waves are different frequencies, as in this thread. And, can YOUR ear here it?

I heard the start of a desert race once. 200 motorcycles all started on the first kick. I expected to be deafened, but only heard a muted humm. But as the pack left,#201 who had flooded his bike got it going, impacting my ears much worse than the other 200.

 

[shadron]

Whenever there are two acoustic sine waves apinge on an ear, they create what are called beat frequencies. Thus, if you are listening to a 1kHz tone and there is a 50kHz tone present, then there will be tones generated by the beat of 51kHz and 49kHz. Now, obviously, if the cutoff of hearing is 20kHz, then neither of the beat frequencies will be heard, but what about the case if 20.5kHz and 1kHz? Then, an audible tone will be generated at 19.5kHz. See beat acoustics^WP.

These beat tones of x+y and x-y Hz are caused by responses in the ear and are not actually tones that exist per se (and therefore do not themselves contribute to secondary beat tones, and so on).

The theory of beat tones is essential to the understanding of FM radio (super-heterodyne) demodulation and in piano and other instrument tuning, among other uses.

 

[Aitch]

Whenever there are two acoustic sine waves apinge on an ear, they create what are called beat frequencies. Thus, if you are listening to a 1kHz tone and there is a 50kHz tone present, then there will be tones generated by the beat of 51kHz and 49kHz. Now, obviously, if the cutoff of hearing is 20kHz, then neither of the beat frequencies will be heard, but what about the case if 20.5kHz and 1kHz? Then, an audible tone will be generated at 19.5kHz. See beat acoustics^WP.

These beat tones of x+y and x-y Hz are caused by responses in the ear and are not actually tones that exist per se (and therefore do not themselves contribute to secondary beat tones, and so on).

I see what you are saying, I think. But if the effect is in the ear and the ear doesn't respond to anything above 20KHz (for example), then how can there be beats based on a frequency above 20KHz?

 

[marting]

Whenever there are two acoustic sine waves apinge on an ear, they create what are called beat frequencies. Thus, if you are listening to a 1kHz tone and there is a 50kHz tone present, then there will be tones generated by the beat of 51kHz and 49kHz. Now, obviously, if the cutoff of hearing is 20kHz, then neither of the beat frequencies will be heard, but what about the case if 20.5kHz and 1kHz? Then, an audible tone will be generated at 19.5kHz. See beat acoustics^WP.

These beat tones of x+y and x-y Hz are caused by responses in the ear and are not actually tones that exist per se (and therefore do not themselves contribute to secondary beat tones, and so on).

The theory of beat tones is essential to the understanding of FM radio (super-heterodyne) demodulation and in piano and other instrument tuning, among other uses.

You are confusing non-linear responses, which creates modulation sidebands with beat tones which do not. For example, if you are listening to two audible tones .01 Hz apart you are not hearing a .01Hz tone but the same two tones going in and out of phase at .01 Hz. If you have a transducer that can measure down to .01 HZ it will show no response. This .01 Hz beat sounds nothing like a .01Hz tone which isn't audible. A .01 Hz tone can be created by getting in an elevator and going up and down floors at a 100 second rep rate. The two are quite obviously very different.

 

[theMark]

Beat / Interference: not just "in the ear", are they?

These beat tones of x+y and x-y Hz are caused by responses in the ear and are not actually tones that exist per se (and therefore do not themselves contribute to secondary beat tones, and so on).

I'm not sure about that - the wikipedia article states that this only happens if (e.g. by using headphones) one ear receives a completely different signal than the other. The usual "interference" of two signals "in the air" is something readily measurable, and e.g. mixing 9900 kHz and 10000 kHz would produce an audible result of 100 Hz (at least it does if i play back these two waveforms on left and right channel in an audio software).

However, for the beat tones of x+y or x-y to fall into the audible range, both need to be within the range the device can handle. The impact of a 48 kHz signal on a 1 kHz signal is either out of range of hearing (49/47 kHz), or an alternative like 48 +/- 47 = 1 kHz and 95 kHz would need a playback system that actually can handle this.

Any other "subtle coloring effect" that falls within the normal hearing range will already be in the recording of the original sound, so ... no need to tax you stereo (or quadro system with those ultra-high frequencies

Disclaimer: i'm not an audio pro, so i'm ready to be corrected on the accuracy above statements.

 

[rockoon]

OK, the above is something that hi-fi bods come out with regularly, as a reason for buying expensive hi-fi or keeping their MP3 files at extremely high sampling rates. But I don't see how it can be true.

One thing I would like to point out is that bit-rate is not sample-rate.

To be clear. MP3's do not store time domain samples.

The signal is converted to the frequency domain before encoding. The bitrate relates to the precision of each encoded frequencies phase and amplitude. With MP3 (and most other lossy audio encoders) the phase information takes a much bigger loss than the amplitude information because the human ear isnt all that good at distinguishing between different phases.

A high bitrate primarily increases the precision of the amplitude information. With some kinds of music/audio this can be important.

 

[shadron]

I'm not sure about that - the wikipedia article states that this only happens if (e.g. by using headphones) one ear receives a completely different signal than the other. The usual "interference" of two signals "in the air" is something readily measurable, and e.g. mixing 9900 kHz and 10000 kHz would produce an audible result of 100 Hz (at least it does if i play back these two waveforms on left and right channel in an audio software).

That's right. It is the effect that piano tuners use - they blow a reed note (fixed by the geometry of the vibrating reed and play the same note on the piano, and tighten or loosen the string listening for the beat frequency, and adjusting the string until unison is achieved and the beat goes to 0Hz.

You are confusing non-linear responses, which creates modulation sidebands with beat tones which do not. For example, if you are listening to two audible tones .01 Hz apart you are not hearing a .01Hz tone but the same two tones going in and out of phase at .01 Hz. If you have a transducer that can measure down to .01 HZ it will show no response. This .01 Hz beat sounds nothing like a .01Hz tone which isn't audible. A .01 Hz tone can be created by getting in an elevator and going up and down floors at a 100 second rep rate. The two are quite obviously very different.

Yes, you are right - I knew there was something in there about the non-linear effects of the ear itself (but "which do not..." what?) In any case, the effect is that a supersonic note might render itself audible by such a mode. One thing to note about these effects, though, is that supersonic frequencies become harder to produce at all as the frequency goes up - for energy reasons, for inertial reasons. It would take a fine hearing to discern them or their beats.

I used to be able, in my younger days, to hear the flyback transformer laminations vibrate that their 15,750 Hz resonance. Now, that is long in the past. So does age (and the Astronauts live) make deaf of us all.

 

[TheDaver]

The people on the board I was referring too WERE talking about sampling rates, not filters, not low-pass filters (it's a music board - lots of music experts, even the odd known composer) and whether they should keep there music files (possibly not MP3) at high bit rates to keep the high frequencies.

Okay, was it sampling rates, or bitrates?

CDDA’s sampling rate is always 44.1kHz. It’s a fixed spec. Resampling (e.g. to 48kHz) is a lossy operation.

High frequencies (near the threshold of human hearing) are tough to encode accurately. The more bitrate you have to work with, the better. Encoders tend to use lowpass filters at lower bitrates to cut off the frequencies that don’t have a chance of being encoded properly.

Ah, remember the days when stereo radios had a button to filter out the 19KHz pilot tone? Or are you all too young?

I suppose I am.

 

[Ivor the Engineer]

Resampling (e.g. to 48kHz [from 44.1kHz]) is a lossy operation.

Incorrect.