[Poll] Is no-distortion really better than any distortion?

[lashto]

Pray tell, which part is 'utter nonsense'? Here are the Sony MH755's full measurements, and here is a detailed description of the site's IEC-compliant testing methodology.

How exactly did you test that those HPs "don't distort audibly"?
Or how does anyone do such a test for any transducer!?

 

[Robin L]

Pray tell, which part is 'utter nonsense'? Here are the Sony MH755's full measurements, and here is a detailed description of the site's IEC-compliant testing methodology.

It's a cheap IEM, give me a freakin' break.

 

[bobbooo]

I'd say so too.
There are hundreds of blind tests where people could not diff between amps/dacs/etc but I am not aware of a single test where people failed to ABX two transducers. Maybe there are some!?
Transducers have HD spikes at around -50db in the most sensitive ranges. Or even worse. Plus audible diffs in directivity, FR, etc. Hard to believe that any two transducers will sound similar enough to be confused.

Take note @Robin L , this right here is utter nonsense. We're talking about whether all transducers audibly distort @lashto , and then you suddenly start talking about blind ABXing transducers that differ in other attributes like frequency response etc. I'm baffled why you think that's relevant here. I've literally just linked to measurements of a transducer that has THD about 0.05% around our ear's most sensitive frequencies. This is ~-65 dB, well below your -50 dB value.

 

[bobbooo]

It's a cheap IEM, give me a freakin' break.

And your point is? What does the price of the transducer have to do with your claim that 'all transducers audibly distort'?

 

[Robin L]

Take note @Robin L , this right here is utter nonsense. We're talking about whether all transducers audibly distort @lashto , and then you suddenly start talking about blind ABXing transducers that differ in other attributes like frequency response etc. I'm baffled why you think that's relevant here. I've literally just linked to measurements of a transducer that has THD about 0.05% around our ear's most sensitive frequencies. This is ~-65 dB, well below your -50 dB value.

No, there will be many aspects of sound that will be different. That difference is distortion.

 

[bobbooo]

No, there will be many aspects of sound that will be different. That difference is distortion.

Nope, distortion is specifically a nonlinear modification. The other modifications are linear or stochastic. I've already linked to this twice already in this thread, but here's hoping this is third time lucky and it will sink in this time:

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.

This is from audio luminary James D. Johnston, one of the pioneers of perceptual audio coding that was used in the development of MP3 and AAC. I think he knows a bit about distortion and its audibility.

 

[Robin L]

Nope, distortion is specifically a nonlinear modification. The other modifications are linear or stochastic. I've already linked to this twice already in this thread, but here's hoping this is third time lucky and it will sink in this time:


This is from audio luminary James D. Johnston, one of the pioneers of perceptual audio coding that was used in the development of MP3 and AAC. I think he knows a bit about distortion and its audibility.

Stop it right there. If it doesn't sound exactly the same as the source, it's distorting. And an IEM is as far from the sound of reality as a rabbit to a diamond.

 

[j_j]

Stop it right there. If it doesn't sound exactly the same as the source, it's distorting. And an IEM is as far from the sound of reality as a rabbit to a diamond.

It is helpful to separate out the changes due to nonlinear from linear from noise sources.

Your highly aggressive "stop right there" is simply refusing to actually attempt to determine the cause of an impairment.

 

[Robin L]

It is helpful to separate out the changes due to nonlinear from linear from noise sources.

Your highly aggressive "stop right there" is simply refusing to actually attempt to determine the cause of an impairment.

An IEM is stuck in one's ear. Resonances will be different because ears are different. The imaging will be different because DUH! Come one, get real. If it doesn't sound the same as reality, it's different, full stop. I don't really need to go any further when the difference is so obvious. I don't need to determine the cause when the results are so clearly off. That's someone else's problem.

 

[j_j]

An IEM is stuck in one's ear. Resonances will be different because ears are different. The imaging will be different because DUH! Come one, get real. If it doesn't sound the same as reality, it's different, full stop. I don't really need to go any further when the difference is so obvious. I don't need to determine the cause when the results are so clearly off. That's someone else's problem.

Aaaah, that's classic. Changing the subject, now? I didn't say anything about IEM's which are, by the way, by and large linear modifications.

And, of course, you'd be surprised what good processing can do with an IEM. Don't assume that everyone has the same weakness as whatever you personally have used.

However, distortion remains the result of nonlinear processes. Linear processes are by and large reversible within the bounds of noise floor. Noise is noise, not distortion, unless it's noise-like signal created by complex intermodulation, which isn't hard to accomplish.

Then there's signal-modulated noise, where the envelope follows the signal but the noise is random.

But the question of an IEM is a DIFFERENT, separate question. Don't conflate things and try to change the subject.

 

[MRC01]

No, there will be many aspects of sound that will be different. That difference is distortion.

Stop it right there. If it doesn't sound exactly the same as the source, it's distorting. And an IEM is as far from the sound of reality as a rabbit to a diamond.

You're using "distortion" to mean any difference between the sounds. But "distortion" has a narrower meaning, referring to specific kinds of differences that are non-linear and non-random.

So if you hear a difference, it might be distortion but it's not necessarily distortion. Indeed, changing the volume is a difference that can be heard, but a proper volume control is linear so it's not distortion. Adding a bit of uncorrelated noise is a difference that can be heard, but that's not distortion either.

PS: is non-flat frequency response considered distortion? It seems it would be, since it's non-linear. That's the most commonly audible flaw in most transducers. If it's not considered distortion, what is the proper term for that kind of flaw in a transducer?

 

[SerpensCaput]

PS: is non-flat frequency response considered distortion? It seems it would be, since it's non-linear. That's the most commonly audible flaw in most transducers. If it's not considered distortion, what is the proper term for that kind of flaw in a transducer?

It's not and that's because distortion always introduces spectral components to a reproduced signal that were not present in the original signal. Non-flat frequency response doesn't introduce anything that wasn't there before, at least not in itself.

 

[MRC01]

OK so if non-flat frequency response isn't distortion, is there some other technical term for it?
I can make one up: "linear response error" but that is rather long-winded.

 

[j_j]

Somebody remind me how to post 3 wave files please. I have 3 files, one of which is noise (to avoid any copyright issues), one of which is filtered noise, and the third of which is inverse filtered, filtered noise.

This clearly demonstrates the reversibility of linear processing.

I will also note that the inverse filtered, filtered noise minus the original results in errors around 1e-15, well under 16 bit level.

 

[Robin L]

Aaaah, that's classic. Changing the subject, now? I didn't say anything about IEM's which are, by the way, by and large linear modifications.

And, of course, you'd be surprised what good processing can do with an IEM. Don't assume that everyone has the same weakness as whatever you personally have used.

However, distortion remains the result of nonlinear processes. Linear processes are by and large reversible within the bounds of noise floor. Noise is noise, not distortion, unless it's noise-like signal created by complex intermodulation, which isn't hard to accomplish.

Then there's signal-modulated noise, where the envelope follows the signal but the noise is random.

But the question of an IEM is a DIFFERENT, separate question. Don't conflate things and try to change the subject.

Someone else changed the subject. I said that transducers are inherently distorted compared to the signal that goes in. They are. There's no escaping that. And I really don't care for all this semantic parsing of the meaning of distortion. Distortion can be frequency response deviation, it can be resonances, it can be THC, it can be IM, but what ever it may be, it is not the same as what went in. You can keep going on proving I'm wrong, but from this point on, you'll be talking to yourselves.

 

[SerpensCaput]

OK so if non-flat frequency response isn't distortion, is there some other technical term for it?
I can make one up: "linear response error" but that is rather long-winded.

Uh, is it too far fetched to just call it filtering? After all that's what filters do, change the magnitude of output signal. Or maybe non-ideal transfer function?

 

[j_j]

Someone else changed the subject. I said that transducers are inherently distorted compared to the signal that goes in. They are. There's no escaping that. And I really don't care for all this semantic parsing of the meaning of distortion. Distortion can be frequency response deviation, it can be resonances, it can be THC, it can be IM, but what ever it may be, it is not the same as what went in. You can keep going on proving I'm wrong, but from this point on, you'll be talking to yourselves.

I'm not responsible for your lack of interest in precision. The fact that you refer to TH(D)? and IM as different things suggests you're not yet to the necessary page in the book. And I presume you meant THD, not THC. Big difference.

Resonances remain linear, of course, but you appear to be missing a basic step here.

 

[j_j]

Uh, is it too far fetched to just call it filtering? After all that's what filters do, change the magnitude of output signal. Or maybe non-ideal transfer function?

Filters are what I call them. As opposed to 'distortions' or 'noise sources'.

 

[Blumlein 88]

Somebody remind me how to post 3 wave files please. I have 3 files, one of which is noise (to avoid any copyright issues), one of which is filtered noise, and the third of which is inverse filtered, filtered noise.

This clearly demonstrates the reversibility of linear processing.

I will also note that the inverse filtered, filtered noise minus the original results in errors around 1e-15, well under 16 bit level.

Zip them and you can post them as a .zip file.
@j_j

 

[MRC01]

... I really don't care for all this semantic parsing of the meaning of distortion. ...

I'm not responsible for your lack of interest in precision. ...

I wonder, since the term "distortion" is often colloquially used to refer to any alteration of the sound, whether, when the OP first started this thread and the poll, he meant this term broadly in its colloquial sense or narrowly in its technical sense.

The context of the question seems to raise the issue whether some alterations in the sound can be euphonic, whether or not these alterations are linear. If so, that suggests the question may use the term "distortion" in its general informal sense. For example, in some cases, adding the right amount of the right kind of noise can subjectively improve sound quality (indeed, it can objectively increase resolution, but that's a different subject). Strictly speaking, that (adding noise) is not distortion. But I speculate that and potentially other examples may be within the purview of the OP's question.