• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Audibility thresholds of amp and DAC measurements

MC_RME

Technical Expert
Technical Expert
Manufacturer
Joined
May 15, 2019
Messages
173
Likes
727
I think what Serge meant is that the difference file shows the 'distortion' aka changes of the original to the DUT file, as in english language instead of english technical description. He did not meant distortion as in non-linear deviations, and refers to his own DF-metric test which uses broadband signals (like music) as stimulus, thus giving broadband (wideband?) difference files.
 

j_j

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
565
Likes
794
Location
My dining room.
The difference files will be broadband, but it would be extremely unusual for the difference file to have no harmonic structure even then. Well, maybe for pure drum machines :)
 
Joined
Dec 7, 2019
Messages
191
Likes
75
I think what Serge meant is that the difference file shows the 'distortion' aka changes of the original to the DUT file, as in english language instead of english technical description. He did not meant distortion as in non-linear deviations, and refers to his own DF-metric test which uses broadband signals (like music) as stimulus, thus giving broadband (wideband?) difference files.
Yes, I use the general definition of distortion [https://en.wikipedia.org/wiki/Distortion]: Output = Input + Distortion
 
Joined
Dec 7, 2019
Messages
191
Likes
75
The difference files will be broadband, but it would be extremely unusual for the difference file to have no harmonic structure even then. Well, maybe for pure drum machines :)
The difference signal do has some harmonic structure, which becomes less pronounced with decrease of its amplitude/level.
 
Joined
Dec 7, 2019
Messages
191
Likes
75
I've given up on fixing wikipedia. What you're talking about is the "error signal".
“Error signal” is mostly used in control theory. And, to be honest, I think this term and the whole approach of control system theory is very appropriate to the audio engineering tasks. Amplifier + transducer is an electro-mechanical system aimed to deliver sound wave with some predefined accuracy. Such systems should be designed as a whole. So, yes, for me they are equal: difference signal = error signal = distortion [in the first article (1999) I used the term “Error level” instead of “Difference level”]
 

j_j

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
565
Likes
794
Location
My dining room.
Serge, the difference is the error signal. Yes, control systems also use this term.

Distortion is reserved for nonlinearities.

You will continue to confuse people with your definition. Wiki has no excuse, but that's what you get from mob rule.
 
Joined
Dec 7, 2019
Messages
191
Likes
75
Distortion is reserved for nonlinearities.
The only difference between linear and non-linear distortions is absence of new spectral components in case of linear distortion. This keeps a pure sine wave undistorted, but more complex signals are corrupted in the both cases. Audibility of the corruption could/should be researched. The linear distortion can be less annoying but quite destructive for stereo imaging as an example. I don't think we should reserve the meaningful word “distortion” for non-linear one only. The difference between them is pure mathematical.
 

j_j

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
565
Likes
794
Location
My dining room.
"linear distortions" is a misnomer by definition. Goodness. Linear processes, in the absence of noise, are reversible.
 

j_j

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
565
Likes
794
Location
My dining room.
It doesn't matter what you prefer. The error signal is the total difference between two signals. Distortion is the result of nonlinear products. Noise is the result of stochastic processes. Linear modifications are reversible in the absence of noise. When there is noise, the issue is noise, not distortion.

These distinctions exist for a reason, and muddying them is not useful.
 
Joined
Dec 7, 2019
Messages
191
Likes
75
It doesn't matter what you prefer. The error signal is the total difference between two signals. Distortion is the result of nonlinear products. Noise is the result of stochastic processes. Linear modifications are reversible in the absence of noise. When there is noise, the issue is noise, not distortion.

These distinctions exist for a reason, and muddying them is not useful.
I think the distinction is clear now. As df-metric uses "black box" concept all types of distortion are treated the same - they are measured and assessed in concern of audibility of particular distortion. So, this distinction between linear and non-linear distortion is absent df-metric.

Below is a square signal distorted by pure linear process - all-pass filtering:

square-allpass.png


This is also distortion, "linear distortion" is not a misnomer.
 
Last edited:

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
4,075
Likes
7,456
Location
The Neverlands
I believe J_J's point is that you cannot undo non-linear distortion and noise but you can simply undo linear distortion.

It will show up in nulling as a difference signal but with the correct 'anti-EQ' you can get it back to normal so while it is a change in the signal it is not a distorted signal.
 
Joined
Dec 7, 2019
Messages
191
Likes
75
I think if a distortion has happened then the main/only question is whether it is audible or not. Researching this question one can use various approaches and concepts. The distinction between linear and non-linear distortion exists only within traditional audio metric. Df-metric uses the most general approach for handling distortion and the above distinction has no sense within it. In other words the distinction is analysis-dependent.
 

j_j

Addicted to Fun and Learning
Audio Luminary
Technical Expert
Joined
Oct 10, 2017
Messages
565
Likes
794
Location
My dining room.
Ok, now that's a crock, the distinction between linear and nonlinear is a fundamental distinction based in mathematics. It has nothing to do with "traditional audio metrics" which as usual are somewhat behind the times.

Linear operations (in the absense of noise) are reversible. Your all-pass filter can be convolved with the time-reverse of the allpass filter, and presto, the "distortion" is gone. That's because IT IS NOT DISTORTION, it is a linear filtering operation.

When a nonlinear operation happens, the result is lossy, and information is lost. This is a fundamental difference, and one you need to learn.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,221
Likes
1,679
Location
North-East
When a nonlinear operation happens, the result is lossy, and information is lost. This is a fundamental difference, and one you need to learn.
I'm trying to learn, but having a hard time. Any transfer function that produces a one-to-one mapping between the input and output can, in theory, be reversed. Including non-linear. Information is lost only if the mapping is many to one (e.g., clipping/compression) or one to many (e.g., noise).

Are you then defining non-linear distortion as that which is generated by a subset of all the non-linear transfer functions that are lossy?
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
4,075
Likes
7,456
Location
The Neverlands
So you can take a tube amplifier with lots of certain types of distortion and use DSP (which also introduces some distortion) and make it sound like SS again ?

That would take quite some characterization of the amp in question (lots of measurements) and then write something to reverse it.
With DAC's and aliasing involved as well it would take quite some code to correct that.

It could be your next software... 'the un-doer' that corrects every non-linear device that is characterized first of course so would also need very high quality ADC and DAC to ensure it doesn't (or compensates itself as well) add distortions of its own.

The distinction between non linear and linear is not for nothing.

Of course ... when one doesn't know whether linear or non linear distortion is there (including phase distortion) and this is simply nulled all of these changes will pop up as an amplitude difference in the null and cannot be separated unless one analyzes the result compared to the original.
 

pkane

Major Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
1,221
Likes
1,679
Location
North-East
So you can take a tube amplifier with lots of certain types of distortion and use DSP (which also introduces some distortion) and make it sound like SS again ?

That would take quite some characterization of the amp in question (lots of measurements) and then write something to reverse it.
With DAC's and aliasing involved as well it would take quite some code to correct that.

It could be your next software... 'the un-doer' that corrects every non-linear device that is characterized first of course so would also need very high quality ADC and DAC to ensure it doesn't (or compensates itself as well) add distortions of its own.

The distinction between non linear and linear is not for nothing.

Of course ... when one doesn't know whether linear or non linear distortion is there (including phase distortion) and this is simply nulled all of these changes will pop up as an amplitude difference in the null and cannot be separated unless one analyzes the result compared to the original.
What do you think DeltaWave does to compute the null? It corrects for distortions found in the comparison file to match it to the reference. This includes measuring and correcting for both, linear (with apologies to j_j) and non-linear distortions, as long as these are not lossy.
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
4,075
Likes
7,456
Location
The Neverlands
Yes, I know.
It is something different than writing a DSP program that undoes the damage though.

Have you characterized a device and fed the opposite of it in your distortion program ?
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
7,837
Likes
9,417
So you can take a tube amplifier with lots of certain types of distortion and use DSP (which also introduces some distortion) and make it sound like SS again ?

That would take quite some characterization of the amp in question (lots of measurements) and then write something to reverse it.
With DAC's and aliasing involved as well it would take quite some code to correct that.

It could be your next software... 'the un-doer' that corrects every non-linear device that is characterized first of course so would also need very high quality ADC and DAC to ensure it doesn't (or compensates itself as well) add distortions of its own.

The distinction between non linear and linear is not for nothing.

Of course ... when one doesn't know whether linear or non linear distortion is there (including phase distortion) and this is simply nulled all of these changes will pop up as an amplitude difference in the null and cannot be separated unless one analyzes the result compared to the original.
Isn't that rather what feedback does? The distortion undoer signal. Some audiophile company can create a story about that, and suddenly feedback will be good again.
 
Top Bottom