Missing fundamental for a 15kHz signal - audible or not?

[pma]

If you really used three test tones with amplitude 0.9, mixed them together and exported to standard wav, then you simply clipped the signal. And yeah, that's exactly what you'd get:
View attachment 263897

Yes, we can see elementary mistakes and then flawed conclusion. Some people would do anything, even if incorrect, to support their illusion.

 

[Sokel]

As far as it goes the ONLY real world concrete result (which is what it counts,the rest is sterilized test-bench ones,rarely applicable in the real deal) was shown by @ctrl.
And the result is plain and simple,just inaudible in most cases.
The rest of the talk is only about fun,"if" theory and egos (as usual)

 

[Julf]

And the result is plain and simple,just inaudible in most cases.

"most cases"?

 

[Sokel]

"most cases"?

I'm counting @ctrl's Maranz too,who showed IMD,but no missing fundamental.
In real world that would be audible nevertheless and that's all I care for.

 

[Geert]

As to the OP, I don't see how this would work the other way around. If we are unable to hear the harmonics, how on earth can we "hear" the fundamental? Remember, either way we are not actually hearing it, we are just able to perceive it.

Sorry, theory is not accepted by the OP. That's why he's gone now. He didn't believe people can't fly.

 

[Julf]

I'm counting @ctrl's Maranz too,who showed IMD,but no missing fundamental.
In real world that would be audible nevertheless and that's all I care for.

Sure, I think we all agree IMD can be audible, but I guess the point of this discussion was the (nonexistent) missing fundamental.

 

[Sokel]

Sure, I think we all agree IMD can be audible, but I guess the point of this discussion was the (nonexistent) missing fundamental.

In every such discussion I have an only rule:audible or not.
I don't care if the hypothesis is wrong as long as something else emerges (even by accident) to put on the audible list in order to become useful for the real thing.
So...

 

[GXAlan]

In every such discussion I have an only rule:audible or not.
I don't care if the hypothesis is wrong as long as something else emerges (even by accident) to put on the audible list in order to become useful for the real thing.
So...

Spoken like a true scientist.

We have shown there is no missing fundamental effect but that the IMD generated can be audible across different real world products.

We have found new bugs in Multitone software.

I have learned that Audacity, running 32-bit floating point precision, doesn’t avoid clipping when exporting multiple tracks even though it does it in real time when doing a playback preview.

It’s been a pretty useful thread, in my opinion.

 

[Julf]

In every such discussion I have an only rule:audible or not.
I don't care if the hypothesis is wrong as long as something else emerges (even by accident) to put on the audible list in order to become useful for the real thing.
So...

I totally agree with that. Just worried that somebody jumping into this thread might read your "And the result is plain and simple,just inaudible in most cases" as "well, in some cases the missing fundamental might actually be audible".

 

[Sokel]

"well, in some cases the missing fundamental might actually be audible".

Nope,that's already out of the list,checked as inaudible.

Spoken like a true scientist.

We have shown there is no missing fundamental effect but that the IMD generated can be audible across different real world products.

We have found new bugs in Multitone software.

I have learned that Audacity, running 32-bit floating point precision, doesn’t avoid clipping when exporting multiple tracks even though it does it in real time when doing a playback preview.

It’s been a pretty useful thread, in my opinion.

(I'm no scientist,I just use whatever logic I have to discriminate the useful stuff and make a nice list to follow to the next of my applications.That's the value of tests)

 

[Julf]

Nope,that's already out of the list,checked as inaudible.

Indeed. Just wanted to avoid the chance of misunderstandings.

 

[pma]

BTW there is absolutely no problem to get always positive ABX result when testing the 22-33-44.wav file against the total silence file. Anytime the 22-33-44 file is started, there is a slight audible click in the beginning, regardless the fact the spectrum has only 22k, 33k and 44k components. Turning on the file of course brings the turn-on transient the has also low frequency transient spectrum. So it goes, we do not listen continuous signal if we start and stop the file. On the other hand, turning on "silence" gives no sound, no spectrum.

This is the analog output from DAC+headphone amp playing my 22_33_44.wav file and captured is the turn-on, start of the DAC signal. It is the initial transient that is easily audible as a click, at higher volume. It is captured with 12.5MHz sampling frequency, fast enough.

This is the spectrum of the first 10ms of the turn-on signal as shown in time domain here above. It has, of course, considerable low frequency content that is audible as a click.

Experiments and investigations like this should not be done by incompetent persons, as they are prone to overlook mistakes and prone to false conclusions.

 

[Sokel]

This is the analog output from DAC+headphone amp playing my 22_33_44.wav file and captured is the turn-on, start of the DAC signal. It is the initial transient that is easily audible as a click, at higher volume. It is captured with 12.5MHz sampling frequency, fast enough.

View attachment 264017

This is the spectrum of the first 10ms of the turn-on signal as shown in time domain here above. It has, of course, considerable low frequency content that is audible as a click.

View attachment 264019

Experiments and investigations like this should not be done by incompetent persons, as they are prone to overlook mistakes and prone to false conclusions.

So,your file is bad?
(didn't use it yet,but I was about to to see if I have IMD in some gear now that I have my sacrificial gear in order.Will you make a good one?)

 

[pma]

So,your file is bad?
(didn't use it yet,but I was about to to see if I have IMD in some gear now that I have my sacrificial gear in order.Will you make a good one?)

No, your (and of many others) understanding is bad. Even if I had made fade-in file start, if you make a choice to abx compare the file somewhere in the middle the result would be the same and abx false positive. Please note that even just one impulse 192kHz sample (less than 5us width) is audible, if it has high enough level, because the spectrum of such impulse goes from DC to ~200kHz. The proper understanding of relation between time and frequency domain is needed here as in many other threads.
The low frequency portion of the spectrum disappears after certain time interval and then you hear nothing. The spectrum of 22-33-44 kHz is valid after the initial transient fades away. It is valid for the continuous stationary signal. Not at the moment of turn-on and turn-off.

 

[Sokel]

No, your (and of many others) understanding is bad. Even if I had made fade-in file start, if you make a choice to abx compare the file somewhere in the middle the result would be the same and abx false positive. Please note that even just one impulse 192kHz sample (less than 5us width) is audible, if it has high enough level, because the spectrum of such impulse goes from DC to ~200kHz. The proper understanding of relation between time and frequency domain is needed here as in many other threads.
The low frequency portion of the spectrum disappears after certain time interval and then you hear nothing. The spectrum of 22-33-44 kHz is valid after the initial transient fades away. It is valid for the continuous stationary signal. Not at the moment of turn-on and turn-off.

Bottom line (ignoring my bad understanding) not good for ABX or other test.
Got it.

(I only do this for fun,be nice,don't be an old grumpy )

 

[pma]

Digitally clipped (@GXAlan ) 22-33-44kHz signal. Please note false in-band components (5, 6, 11, 16 kHz).

Properly generated 22-33-44 signal

All measured at analog output from head amp driven from DAC. Scope resolution is only 8-bit but still enough to show the common mistakes.

 

[audiofooled]

are we confusing missing fundamentals with difference tones/beat frequencies?

No.

https://www.acs.psu.edu/drussell/Demos/superposition/superposition.html (bottom of page)

 

[tmtomh]

Has the OP left the building?

They typically do in such situations.

 

[Blumlein 88]

I have learned that Audacity, running 32-bit floating point precision, doesn’t avoid clipping when exporting multiple tracks even though it does it in real time when doing a playback preview.

It’s been a pretty useful thread, in my opinion.

When more than one other person get the same results and your results are bad you should try and find out what is being done differently. Audacity doesn't have a problem. Your procedure is faulty. If you have an older version, exporting more than one file can be combined into a single file upon export where it can be clipped. Newest versions handle this a bit differently. In any case, you need to learn how to use mix and render under the Tracks drop down menu. If you mix tracks and they clip due to excessive levels the mixed and rendered file will show this without export. You lower levels and mix again. Alternatively you can combine mono tracks into stereo and then render down to mono and Audacity will drop levels of each by - 6db to prevent clipping.

Here is a screenshot where I mixed and rendered 3 tracks to a single track and without exporting or anything else clipping is shown. I'd know at this point to lower all levels until the mixed track is not clipping.

Mixing Audio Tracks - Audacity Manual

manual.audacityteam.org

Tracks Menu: Mix - Audacity Manual

manual.audacityteam.org

 

[pkane]

We have found new bugs in Multitone software.

We did? Or did we find that you didn't use it properly by providing a sample file that was too short for the FFT size you picked? When you supply a very short sample and apply a large size FFT you are in effect dividing the limited energy of the noise floor in that recording across many more frequency bins, resulting in a much lower level per bin. That's not a bug, this is how this stuff works. If you were to use an 8M FFT, you would lower it a whole lot more.