THD has a much bigger effect on sound than you think

[Pdxwayne]

Some parts of the singer will certainly be masked by your voice. You just don't realize it because the voice is dominant. The moments the voice is quiet you will hear the singer.
It would be another matter if the singer did not sing but said the exact same words as you in the same tone and a bit earlier to compensate for the time delay of 600m. Then you won't hear that second person... THAT is masking.
When that person sneezes and the person close to you doesn't than that sound won't be masked unless you sneeze as well.

You need to read up on masking not theorize away on plots that have no relevance and doing absolutely pointless self-tests at unrealistic levels.

I did read up on masking. It looks like you are arguing with theory too without any proof of your own. On one hand, you talked about masking to fit your arguments. Once I said your masking theory is not always correct, you then changed your tones and argue that 1200hz at 60db down is audible when 250hz is playing........

 

[FrantzM]

Oh boy!

 

[pma]

i think it's much more audible when you're listening to it this way (on headphones with low distortion/low excrsion at low volume) than a particular peak on a track on speakers (high volume thus higher masking and very transiently).

yes 3% is reasonable for a lot of bookshelves but not necessarily all the time..

I tried it with speakers and the audibility is similar, though differently “colored” than with headphones. Yes it is obvious, because speaker adds its own coloration.
I know I am bringing a bit “academic” example, however it works and is good for educational purposes. We need to be careful when talking about audibility. With the same system nonlinearity, something may be audible and something not. This applies to music as well. Depends. It is more than clear that nonlinearity aka distortion is a bad thing and should be reasonably minimized. There is no reason to tolerate levels of distortion created for example by SE tube power amplifiers as an inevitable phenomenon. The example I have posted is exactly of this category. We still have non-optimum nonlinear behaviour of speakers and this is the challenge in future.

Edit: spectrum of the signals from the post #176

THD has a much bigger effect on sound than you think

I think you may need a little tutorial to understand distortion better, what causes it and how it manifests itself. Yes, DISTORT will produce different levels of distortions at different sound levels. This is how distortion works. DISTORT works by computing and simulating the non-linear behavior...

www.audiosciencereview.com

attached. 16bit non-dithered BTW. No effect on audibility of IMD.

 

[solderdude]

I did read up on masking.

You clearly did not read up on it enough and I explained why your 'test' had nothing to do with masking at all nor was it even remotely representative.

 

[Frgirard]

The sound energy in music in the 1-4khz range is 30-40db lower than bass, 15-30db than midrange.

When i read the spectral contain on Fabfilter Pro q (EQ) i do not read that for most records

example My last purchase : bernard herrmann the film scores on phase 4.

 

[pkane]

Would you please help confirm my findings? @pkane, please note and comment.

I found that the distort app does not perform distortion setup correctly. For example, I give -45db for second harmonic, I get -55db for second harmonic. I give -103db for 3rd harmonic, and I got -115db.

I took a 1KHz tone and put it though the app. -45db 2nd harmonic, -103db third harmonic, -99 forth harmonic.

Here is the GUI of the distort app:
View attachment 160825


When I use deltawave to compare the distort app created file with original 1khz file, I see the issues. As you can see, all are like 10db or more lower than requested.

View attachment 160824

I have no idea what you're trying to do, but here's a 1kHz sinewave with distortions generated by DISTORT:

Here's the same 1kHz sinewave played back in a loopback after saving it into a file, measured by REW. Notice the magnitudes of harmonics in the top left corner. Seems exactly right, no?

 

[peng]

So, in that worse case amp, the multitones test basically telling us the harmonics added to all freq is at about -50db?

View attachment 160792

That's an interesting thought, but not really. Let me do some read up on whatever I can find about the 32 tone test first before I say anything else. For now, I just know it doesn't exactly work like that..

 

[Pdxwayne]

I have no idea what you're trying to do, but here's a 1kHz sinewave with distortions generated by DISTORT:

View attachment 160866


Here's the same 1kHz sinewave played back in a loopback after saving it into a file, measured by REW. Notice the magnitudes of harmonics in the top left corner. Seems exactly right, no?
View attachment 160865

How about try duplicate what I did?

Take a 1KHz tone, then do -45db 2nd harmonic, -103db third harmonic, -99 forth harmonic.

 

[Pdxwayne]

You clearly did not read up on it enough and I explained why your 'test' had nothing to do with masking at all nor was it even remotely representative.

Fine, it is your "expert" credibility at stake here.

So, you are claiming no masking is at play because I can abx that two tones? You are arguing that "everyone" can hear 1200hz at 60db lower than 250hz, when both 250hz and 1200hz are being played the same time.

Since you are the one who made the claim and you are part of "everyone", please support your own statement by doing such an "easy" abx and post your result here.

 

[pkane]

How about try duplicate what I did?

Take a 1KHz tone, then do -45db 2nd harmonic, -103db third harmonic, -99 forth harmonic.

Why? What's so different about your test compared to what I posted? I just picked a random sequence for harmonics.

 

[pma]

How about try duplicate what I did?

Take a 1KHz tone, then do -45db 2nd harmonic, -103db third harmonic, -99 forth harmonic.

Looks OK to me. I did what you wanted and analyzed by independent SW. Do not forget that distortion depends on amplitude of fundamental.

 

[Pdxwayne]

Why? What's so different about your test compared to what I posted? I just picked a random sequence for harmonics.

View attachment 160890


View attachment 160892

Because there might be corner cases? You are using different sequence than mine. Best to duplicate exactly.

 

[pkane]

Because there might be corner cases? You are using different sequence than mine. Best to duplicate exactly.

No, there's no reason to believe you've lucked in on some specific sequence that just doesn't work. I posted your example working properly with DISTORT already, there is no problem or corner case. I suggest you try to figure out why your test isn't working instead of trying to find bugs.

 

[pma]

Because there might be corner cases? You are using different sequence than mine. Best to duplicate exactly.

No, You make a mistake somewhere, probably in the "custom" setting, check the box, fill in distortion values and do not forget to click on "Apply" button.

 

[Pdxwayne]

Looks OK to me. I did what you wanted and analyzed by independent SW. Do not forget that distortion depends on amplitude of fundamental.

View attachment 160893

I was too lazy to create a pure 1khz tone and I used your pure 1khz file (for the tube amp 0.85% distortion abx test) as the base. I wonder how it could be a problem....

Don't you think your chart harmonics also seems off by a few dB? Look especially at 4th?

 

[pma]

I was too lazy to create a pure 1khz tone and I used your pure 1khz file (for the tube amp 0.85% distortion abx test) as the base. I wonder how it could be a problem....

Don't you think your chart also seems off by a few dB?

The distortion settings in the Distort are valid only if the fundamental is 0dBFS. H2 -40dB will make 1% distortion only for 0dBFS sine. Transfer function is Y = Ax^2 + Bx + C (in case of pure H2) and distortion decreases with decreased sine input. This is normal, just math.

 

[Pdxwayne]

The distortion settings in the Distort are valid only if the fundamental is 0dBFS. H2 -40dB will make 1% distortion only for 0dBFS sine. Transfer function is Y = Ax^2 + Bx + C (in case of pure H2) and distortion decreases with decreased sine input. This is normal, just math.

Your graph showed 0dbs fundamental, right? So if you entered -45db for h2, why your graph showed about -48?

Also, if you enter -99 for h4, why your graph showed about -103?

 

[solderdude]

Fine, it is your "expert" credibility at stake here.

I'm not an expert nor do I claim to be an expert.
I do have many years of practical experience though.

So, you are claiming no masking is at play because I can abx that two tones.

No, I am claiming there is no masking at play in your experiment at all because the frequencies are too far apart and the amplitude difference is too small. Masking does not apply.

 

[Pdxwayne]

I'm not an expert nor do I claim to be an expert.
I do have many years of practical experience though.



No, I am claiming there is no masking at play in your experiment at all because the frequencies are too far apart and the amplitude difference is too small. Masking does not apply.

Sure, prove it and show us your own ABX results.

 

[Pdxwayne]

No, there's no reason to believe you've lucked in on some specific sequence that just doesn't work. I posted your example working properly with DISTORT already, there is no problem or corner case. I suggest you try to figure out why your test isn't working instead of trying to find bugs.

I tried with 250hz pure tone (I created with audacity) and your app worked just fine when assigning harmonics. Strange that using pma pure 1khz tone could cause such differences.

Anyway, is there a reason why using "L+R” option to save files would create files that sound much better than saving using “stereo”?