How High of a Sample Rate is Enough?

[Krunok]

Okay one zip file. Opens into FLAC format. 30 seconds per file. The fundamental is -6 db for each. Potentially tweeter busting levels. Be easy on the volume till you choose an appropriate level.

Bonus file. 7 khz and 14 khz without 21 khz.

Thank you!

I'll report back when my sons wake up to do the test as properly as I can.

 

[Krunok]

So, my younger son (16 years) was easilly able to hear the difference between 11kHz and 11kHz with harmonics. According to another test his hearing range extends to 20kHz.
Older son is still sleeping, so I will test him later.

I can't hear the difference between those 2 files but I can hear the difference between 7kHz and 7kHz with harmonics, although the difference is very small to me and I have to listen very carefully to catch it.

Btw, I checked your files and they are perfectly well composed.

 

[andreasmaaan]

So, my younger son (16 years) was easilly able to hear the difference between 11kHz and 11kHz with harmonics. According to another test his hearing range extends to 20kHz.
Older son is still sleeping, so I will test him later.

I can't hear the difference between those 2 files but I can hear the difference between 7kHz and 7kHz with harmonics, although the difference is very small to me and I have to listen very carefully to catch it.

Btw, I checked your files and they are perfectly well composed.

This sounds like it makes sense so far. Usually hearing tests use frequency bands that are an octave apart, so if your son passed the 20KHz test, it's likely to mean that he has some hearing slightly above that frequency.

 

[andreasmaaan]

Okay one zip file. Opens into FLAC format. 30 seconds per file. The fundamental is -6 db for each. Potentially tweeter busting levels. Be easy on the volume till you choose an appropriate level.

Nice work

 

[Krunok]

This sounds like it makes sense so far. Usually hearing tests use frequency bands that are an octave apart, so if your son passed the 20KHz test, it's likely to mean that he has some hearing slightly above that frequency.

My older son (20 years) also passed the test differentiating between 11kHz and 11kHz with harmonics but with much less flying colours than younger one. According to hearing test his hearing range extends to 19kHz.

Btw, my hearing range barely extends to 13kHz.

EDIT: not 13kHz, but 12kHz. God, I'm really half deaf!

 

[andreasmaaan]

My older son (20 years) also passed the test differentiating between 11kHz and 11kHz with harmonics but with much less flying colours than younger one. According to hearing test his hearing range extends to 19kHz.

Btw, my hearing range barely extends to 13kHz.

Nice. Which hearing test did they take?

 

[Krunok]

Nice. Which hearing test did they take?

Simple one, available on this page: https://www.audiocheck.net/audiotests_frequencycheckhigh.php

 

[andreasmaaan]

Simple one, available on this page: https://www.audiocheck.net/audiotests_frequencycheckhigh.php

What sample rate did you use when (a) testing Blumlein's wave files and (b) conducting the hearing tests?

 

[Krunok]

Volumio switched my D10 DAC to 192/24 when playing this files. Properties of these files indicate the same bitrate.

 

[andreasmaaan]

Volumio switched my D10 DAC to 192/24 when playing this files. Properties of these files indicate the same bitrate.

Just speculating that maybe given the hearing test files are in 44.1KHz/48KHz only, the levels around the top of the test range might not be a full 0dBfs due to the filter, and this might have resulted in your kids getting results slightly lower than they are actually capable of hearing?

A test like that should really be at 88.2KHz minimum IMHO.

 

[Krunok]

Just speculating that maybe given the hearing test files are in 44.1KHz/48KHz only, the levels around the top of the test range might not be a full 0dBfs due to the filter, and this might have resulted in your kids getting results slightly lower than they are actually capable of hearing?

A test like that should really be at 88.2KHz minimum IMHO.

I fully agree.

 

[March Audio]

IME most digital signal generators suffer from aliasing, which can (will) be audible with square waves.
This website offers pretty clean test signals (the basic ones are free) for download. Highly recommended.
https://www.wavtones.com/functiongenerator.php

..... Mmmmm........ Not sure about this. The simple fact is that you need infinite bandwidth to replay a square wave perfectly. You never have infinite bandwidth. The waveform shape will be defined by the number of harmonics you can fit in your available bandwidth. That doesn't make the resultant wave shape wrong, it just is what it is - the result of putting that signal into the available bandwidth. I'm just not sure what they think they are achieving or why.

A square wave without any of its harmonics is a sine wave.

 

[bennetng]

Interesting. I clicked on the "Save" button and analysed the saved files (= aliasing). Did you do the same ?

That website is funny. There is no aliasing in playback mode (I verified by recording the signal) but there is aliasing in download mode, so both of you are right.

 

[Kees de Visser]

A square wave without any of its harmonics is a sine wave.

Right. And a bandlimited square wave is a series of sines waves up to the cut-off freq. That's ok, but there should not be any artifacts, which isn't easy for digital generators (see this AP article). I have a nice, cheap signal generator app which is great for sine waves and noise, but the square waves are full of aliasing, so useless. I'm not a DSP engineer, but apparently it's not easy to do properly. Perhaps Amir can generate a nice square wave with his new AP ?

 

[Fitzcaraldo215]

Ok, very interesting and not surprising. But, thanks to all. However, haven't we for the past several pages been over focused on just the narrower question of the audibility of ultrasonic frequency response rather than a more complete understanding of hirez?

I suggest going back to this paper, now buried deep in this forum:

https://audiosciencereview.com/foru...h-resolution-audio-does-it-matter.11/#post-13

The author seems to know what he is talking about and I tend to agree with his conclusions. Worth a read or reread.

 

[dc655321]

And a bandlimited square wave is a series of sines waves up to the cut-off freq. That's ok, but there should not be any artifacts,

Please correct me if wrong, but a square wave will only present as a series (> one) of sine waves if the fundamental is less than one third the Nyquist frequency. For a band-limited 44.1kHz sample rate, that limit is around 7kHz. If there is no band-limiting, of course aliasing will happen.

 

[bennetng]

http://onlinetonegenerator.com/
That's what I got by recording vs downloading a 1367Hz square wave from the website. They are clearly different.

recorded:

downloaded:

 

[DonH56]

So it does not get lost in the discussions:

• Aliasing happens at the ADC side when any signals at or beyond the Nyquist frequency (22.05 kHz for 44.1 kS/s) are aliased, or folded, back into baseband so they appear as lower-frequency signals. Anti-alias filters at the ADC's inputs are to suppress aliasing.
• Imaging happens at the output of DACs as signals are replicated (imaged) at frequencies above the Nyquist rate, creating ultrasonic signals that are "mirror images" of the lower-frequency signals. Anti-image filters at the DAC's outputs suppress images.

Aliasing in the source material cannot be removed unless you know essentially everything about the input signal and can somehow filter out the aliasing artifacts. That is difficult if not impossible.

Images can be filtered but to whatever extent are present can cause problems such as IMD in buffers and tweeters that adds lower-frequency signals (distortion) and so forth. So in some cases the differences you hear may be due to other components distorting. For example, I ran a test once to look at images of a CD test signal, and discovered IMD was adding low-frequency content and that was why I was able to easily tell which test had images. Alas, my hearing had not magically reached the ultrasonic range...

Another issue is that, if the signal to the DAC is full-scale, then when you add another signal (like a harmonic), you must reduce the fundamental amplitude to maintain the same maximum signal levels. How much depends upon how large the additional signal. If you have two identical signals, then the original FS signal must be reduced to -6 dB when you add a second signal of the same amplitude, so the sum of the two reaches FS. That 6 dB change is readily audible.

HTH, FWIWFM, etc. - Don

 

[Blumlein 88]

Right. And a bandlimited square wave is a series of sines waves up to the cut-off freq. That's ok, but there should not be any artifacts, which isn't easy for digital generators (see this AP article). I have a nice, cheap signal generator app which is great for sine waves and noise, but the square waves are full of aliasing, so useless. I'm not a DSP engineer, but apparently it's not easy to do properly. Perhaps Amir can generate a nice square wave with his new AP ?

Audacity makes square waves with aliasing. Lots of it. However, under the section to generate waveforms there is an option to generate "square, no aliasing" which does indeed generate a nice square wave within the bandwidth limitations that does not have aliasing.

 

[Blumlein 88]

Something you can try with the signals I prepared is on those where your sons heard differences reduce the level a few decibels and see if they still hear a difference. The threshold for those higher harmonics like at 22 khz would be rather high. So reducing the signal say 10 db might then prevent them from hearing a difference as the higher frequency probably won't be heard by them at reduced loudness.

Also I'm curious did they hear a difference in the 7 khz + 14 khz signal vs the 7 khz + 14 khz +21 khz signal?