Mechanism for making 20kHz+ frequencies relevant in high end audio?

[daqq]

Hi guys,

I've always dismissed frequencies above 20kHz specified in audiophile gear as BS, but thinking about it further, I think that there might be a way that they could actually contribute to the overall quality of the reproduction.

Assuming you would get an interference between two speakers, each sounding at one different frequency, two high frequencies (say, 30kHz and 30.2kHz) would effectively substract, leaving only the difference (200Hz), which would be in the audible area. Neither of the speakers would reproduce the 200Hz wave, so if that frequency was desired, the speakers would indeed have to go all the way up to 30+kHz.

Could such a mechanism work?

[hamster_nz]

Yes, but in general it relies on non-linear distortion occurring to 'create' the low frequency output. If the system is completely linear then no 200 Hz tone will be heard.

[Richard Crowley]

https://en.wikipedia.org/wiki/Sound_from_ultrasound

[edavid]

Assuming you would get an interference between two speakers, each sounding at one different frequency, two high frequencies (say, 30kHz and 30.2kHz) would effectively substract, leaving only the difference (200Hz), which would be in the audible area. Neither of the speakers would reproduce the 200Hz wave, so if that frequency was desired, the speakers would indeed have to go all the way up to 30+kHz.

Could such a mechanism work?

It doesn't matter, since we know there are no 30kHz tones in live music (unless you are recording a bat orchestra).

[Lightages]

When I was much younger, I could hear to 23kHz reliably. I don't know what I can hear now but it is much lower. I could not be in certain buildings that had ultrasonic alarm detectors. 20kHz is just an arbitrary cutoff for the general population.

[magetoo]

It doesn't matter, since we know there are no 30kHz tones in live music (unless you are recording a bat orchestra).

You beat me to it.

There could theoretically be, coming out of the instruments, frequencies above the accepted range of human hearing, but the sound recording equipment won't pick it up; the microphones commonly used rarely go above 20 kHz, anything that is specified for much above that is specialized lab equipment and not the sort of thing that is marketed for "sounding great with drums" (or whatever you're recording).

When you think about it, it's surprising that so many are looking at the reproduction side and so few at the recording side.

[fivefish]

There's few energy left in 16Khz to 20Khz in most finished/pressed CD recordings.  But if you boost these range of frequencies say +15/+20dB during the recording and/or mixing and mastering phase of music production, there is noticeable and audible difference... not much, very subtle, but it's there and it makes for a nicer listening experience (more airy as we say it). Also, it depends on the musical instrument being recorded. (Sampled synths won't benefit from this process, but recorded acoustic guitar, cymbals, percussive instruments and recorded voice benefit from this slight boost of the "air band".)

[Someone]

Hi guys,

I've always dismissed frequencies above 20kHz specified in audiophile gear as BS, but thinking about it further, I think that there might be a way that they could actually contribute to the overall quality of the reproduction.

While sustained tones in that range are considered unimportant for reproduction it is possible to hear very small changes in relative phase or arrival time in stereo/multichannel systems.

It doesn't matter, since we know there are no 30kHz tones in live music (unless you are recording a bat orchestra).

Many acoustic instruments have a lot of energy above 20kHz, even strong resonant modes that appear as "tones" in spectrograms.

When I was much younger, I could hear to 23kHz reliably. I don't know what I can hear now but it is much lower. I could not be in certain buildings that had ultrasonic alarm detectors. 20kHz is just an arbitrary cutoff for the general population.

Yes, people come in a great variety of specifications! And while much/most of the population is happy with highly compressed mpeg audio and video some of us want something more.

It doesn't matter, since we know there are no 30kHz tones in live music (unless you are recording a bat orchestra).

You beat me to it.

There could theoretically be, coming out of the instruments, frequencies above the accepted range of human hearing, but the sound recording equipment won't pick it up; the microphones commonly used rarely go above 20 kHz, anything that is specified for much above that is specialized lab equipment and not the sort of thing that is marketed for "sounding great with drums" (or whatever you're recording).

When you think about it, it's surprising that so many are looking at the reproduction side and so few at the recording side.

Thats drawing a lot more into the discussion, I'm regularly working with audio chains that are designed to capture an extra octave and have -3db points out beyond 30kHz but it has remained a specialist area and delivery is still constrained by the 44/48kHz broadcast rates and considerations for listeners (the majority) who want to listen on their clock radio in the kitchen etc.

[vk6zgo]

A lot of the "you need to go up to 50kHz plus " stuff comes from the days of tube amplifiers with the output transformer included in the Negative Feedback Loop.
The idea was that fairly sharp cutoff above the audible range could lead to phase reversal in the feedback system,hence oscillation at ultrasonic frequencies (they called them "supersonic" frequencies,but that is no longer PC in Engineering circles).

[SL4P]

Also read about slew-rate & transient response.

A recurring tone may not occur or exist at that frequency, but having the energy and ability to change level/state at 'faster than the reproduced frequency' is significant - otherwise the edges will appear 'blurry'.
Nyquist squeezes in here (hear) again!

[German_EE]

Yes, the frequencies above 20 KHz matter but not in the way that you think. Feed an 18 KHz square wave into an amplifier with 250 KHz bandwidth and you will get a pretty good square wave out, feed the same 18 KHz square wave into an amplifier with 25 KHz bandwidth and your output will be a sine wave. The higher frequency signals are needed for accurate timbre and without them you don't get accurate reproduction.

[Tim F]

Yes, the frequencies above 20 KHz matter but not in the way that you think. Feed an 18 KHz square wave into an amplifier with 250 KHz bandwidth and you will get a pretty good square wave out, feed the same 18 KHz square wave into an amplifier with 25 KHz bandwidth and your output will be a sine wave. The higher frequency signals are needed for accurate timbre and without them you don't get accurate reproduction.

You are aware that the harmonic frequencies which make a 18kHz square wave... square... are ultrasonic and therefore inaudible? Both an 18kHz sine wave and 18kHz square wave sound the same to a human, because our ears don't respond to the  54kHz, 90kHz and so on odd harmonics present in an 18kHz square wave. The ears will turn that 18kHz square wave into a sine wave just as the 25kHz amplifier did. Unless your audience are mice no one is going to complain about 'lost timbre'

If you consider a lower frequency tone, say at 1kHz then square and sine waves will sound different because we can hear the harmonics of the square wave occurring at 3kHz, 5kHz, 7kHz and so on.

[daqq]

Thanks guys,

So, the only possible effect of anything audio that has a rating above the audible would be marginal at best by enabling faster transition times, timing, etc? Not so much the actual sound but rather the in between?

Yes, the frequencies above 20 KHz matter but not in the way that you think. Feed an 18 KHz square wave into an amplifier with 250 KHz bandwidth and you will get a pretty good square wave out, feed the same 18 KHz square wave into an amplifier with 25 KHz bandwidth and your output will be a sine wave. The higher frequency signals are needed for accurate timbre and without them you don't get accurate reproduction.

Wouldn't a square wave at 18kHz sound to a person who hears up to 20kHz sound exactly the same as a sine wave at 18kHz, since the harmonics would effectively be cut off?

[Jr460]

Yes, in some ways stuff above 20K can be heard.....

Send into a cheap A/D converter with sloppy input filter.  The things above 20K get wrapped around and back below 20K.  And since they are no longer harmonicly related, they stick out to our brain.

Make that input filter very sharp and you start running into phase and other issues in the passband below 20K.

The standard CD rate of 44.1 just makes it hard to cut things off.  I like 48K, to me it sounds better.

I have heard from a respected audio engineer, he makes his living running a studio and mixing albums, that something he couldn't hear made a difference.  For some reason an input to his board, one of the last large format completely analog boards made, had some ringing when hit hard.  The ringing was about 19.5K.  No one in the studio heard it, and it only lasted a short time when it happened.  But it did make the compressor on that channel think it had a valid signal it had to deal with.  That caused him to set the compressor differently.

He only found out later, after the album was mixed and mastered and sold.  Some people with bat hearing heard this little chirp at times.  The input gets hit, it rings after the signal is gone, then the compressor kicks in and yanks things down.  That is the basic story as I remember it from him.


To other's questions.   A square wave at 18k will sound just like sine wave at 18k, unless you can hear up to at least 54k.

Faster transition times, mean nothing in terms of hearing or timing of audio sounds.  They do require greater bandwidth.

[Richard Crowley]

...The ringing was about 19.5K.  No one in the studio heard it,... Some people [end-user listeners] with bat hearing heard this little chirp at times....

And that is the primary factor here. A vanishing number of customers (and producers for that matter) can hear anything near 20KHz. And an even more vanishing number of customers ever hear anything but compressed* audio, whether on their music players, on radio or on TV.  Does anybody sit in their acoustically-correct parlor anymore to listen to their stereo system playing back a 44.1KHz sample-rate CD?

* and by "compressed" I mean BOTH amplitude-envelope limited AND digital data lossy compression.

[Maxlor]

Monty explains a good deal of this very nicely. He talks about high sampling frequencies beginning at 11:18:

[Tim F]

Yes, in some ways stuff above 20K can be heard.....

Send into a cheap A/D converter with sloppy input filter.  The things above 20K get wrapped around and back below 20K.  And since they are no longer harmonicly related, they stick out to our brain.

Make that input filter very sharp and you start running into phase and other issues in the passband below 20K.

The standard CD rate of 44.1 just makes it hard to cut things off.  I like 48K, to me it sounds better.

It's almost never an issue with modern hardware which oversamples at several hundred kHz or even MHz. You are most likely to run into an audible issue with software doing a quick and dirty resample by simply doubling or discarding samples.

[Jr460]

Yes, in some ways stuff above 20K can be heard.....

Send into a cheap A/D converter with sloppy input filter.  The things above 20K get wrapped around and back below 20K.  And since they are no longer harmonicly related, they stick out to our brain.

Make that input filter very sharp and you start running into phase and other issues in the passband below 20K.

The standard CD rate of 44.1 just makes it hard to cut things off.  I like 48K, to me it sounds better.

It's almost never an issue with modern hardware which oversamples at several hundred kHz or even MHz. You are most likely to run into an audible issue with software doing a quick and dirty resample by simply doubling or discarding samples.

Very true.  The next thing you find that kills audio is cheap processing that doesn't do dither correctly, to in most cases doesn't really get what a sample is or isn't.  People find those poorly coded audio plugins, VST and AU, sound better when you are editing a session at 88k, or 96k.  They could just up sample within the code, but, like you said they do it so poorly that it makes a mess out of things.

Some other things that will kill you.  Poor converters, even if they over sample by a large margin.  The analog side just doesn't have the balls in most cases to put out a full scale sine wave at all frequencies.  They test the thing with small signal, and show that the response is flat 20-20k +/- .1dB or something like that.  Push that converter to the limit, and the analog side gives up. 

One thing that gets missed is people think that the signal is a stair step kind of thing.  It is not.  Each sample point is magnitude of a sinc function, which adds with all the other others.  This means if you have a sample that is full scale digital, and you have the converter set that it gives you 1V  at FS.  You will have an analog signal that exceeds that 1V level. 

Design the thing to put out consumer level -10dBm rather than +4, slap on top of that a crappy power supply, and then something that is not so linear as it gets to full scale (no extra headroom), and you get a not so great sounding converter.

[Richard Crowley]

Design the thing to put out consumer level -10dBm rather than +4, slap on top of that a crappy power supply, and then something that is not so linear as it gets to full scale (no extra headroom), and you get a not so great sounding converter.

Level is not the problem. Gigantic mixing consoles that have legendary reputations and cost as much as your house use unbalanced -10dBm internally.

[Zero999]

Yes, it's BS and this should apply to THD which is unimportant at frequencies above around 10kHz where the harmonics can't be heard.