[krismusic]

I find this definition for psychoacoustics is the most accurate - the study of the perception of sound - I reckon it focusses on the brain processing aspects rather than the mechanical mechanisms of the ear but there are always grey areas when definitions are used to circumscribe anything. 

I've always taken the term to describe psychological factors such as placebo and expectation bias. Which I think really get in the way of objective evaluation.

 

[mmerrill99]

I've always taken the term to describe psychological factors such as placebo and expectation bias. Which I think really get in the way of objective evaluation.

I know but didn't we exchange these definitions already?
Agreed that perceptual testing is fraught with many confounding factors which are very difficult to deal with - so difficult that only tests done by professionals trained in the area of perceptual testing are of any note.
 
The recent moves towards EEG, MEG testing in neuroscience is a welcome departure from the previous, primitive, perceptual testing 

 

[RRod]

  The recent moves towards EEG, MEG testing in neuroscience is a welcome departure from the previous, primitive, perceptual testing 

 
Previous "primitive" perceptual testing has got us files that, to my ears, are indistinguishable from the originals at 128kbps. What kinds of perceptual abilities do you think the developers of codecs like Opus and AAC have overlooked? And why have such abilities not enabled people to distinguish high-rate lossy codecs from their original sources in blind testing? Give us a little bit.

 

[mmerrill99]

   
Previous "primitive" perceptual testing has got us files that, to my ears, are indistinguishable from the originals at 128kbps. What kinds of perceptual abilities do you think the developers of codecs like Opus and AAC have overlooked? And why have such abilities not enabled people to distinguish high-rate lossy codecs from their original sources in blind testing? Give us a little bit.

Sure codecs that use a psychoacoustic model are pretty good but not even the developers of those codecs would claim that they are audibly transparent at 128kbps. I believe that there are many blind test results on Hydrogen Audio showing differences are audible between 128kbps & original.
 
BTW, I consider them primitive because they have so many confounding variables that need addressing as is defined in the various published standards such as BS.1116 & others 
 
As to your second question - I'm wouldn't be sure that high rate lossy codecs have not been differentiated in blind testing?

 

[RRod]

  Sure codecs that use a psychoacoustic model are pretty good but not even the developers of those codecs would claim that they are audibly transparent at 128kbps. I believe that there are many blind test results on Hydrogen Audio showing differences are audible between 128kbps & original.
 
BTW, I consider them primitive because they have so many confounding variables that need addressing as is defined in the various published standards such as BS.1116 & others 
 
As to your second question - I'm wouldn't be sure that high rate lossy codecs have not been differentiated in blind testing?

 
You know for certain that none of the major codecs these days has been testing in an setup that meets the standards? I guess going to the source is the only way to find out.
 
128k was in reference to *my ears* with Opus, not everyone's ears with anything else. Still, give it a try yourself.
 
320k mp3 is probably differentiable for some folks with certain material. I have yet to see anything verified for 256AAC or similar Opus and Vorbis rates, but anything on a forum is suspect anyway, in either the negative or affirmative.

 

[mmerrill99]

   
You know for certain that none of the major codecs these days has been testing in an setup that meets the standards? I guess going to the source is the only way to find out.
 

I wasn't saying that the codecs are tested without adherence to the published standards for codec testing!!

128k was in reference to *my ears* with Opus, not everyone's ears with anything else. Still, give it a try yourself.
 
320k mp3 is probably differentiable for some folks with certain material. I have yet to see anything verified for 256AAC or similar Opus and Vorbis rates, but anything on a forum is suspect anyway, in either the negative or affirmative.

 


Yea, but it would not be really that surprising if these codecs were not differentiated by such blind testing, would it? - they were developed by using the same blind testing techniques, after all.
 
My point was that other newer brainwave testing in neuroscience may well reveal more than is possible with the blind testing used in perceptual testing up to now?  

 

[davidsh]

There are obviously pros and cons and limitations to either approach.
Your question is interesting, though I'm not qualified to answer, so I won't try

 

[RRod]

  Yea, but it would not be really that surprising if these codecs were not differentiated by such blind testing, would it? - they were developed by using the same blind testing techniques, after all.
 
My point was that other newer brainwave testing in neuroscience may well reveal more than is possible with the blind testing used in perceptual testing up to now?  

 
No doubt, but the question then comes back to the old "statistical significance vs. actual significance" problem. My feeling is that if a better-than-baseline signal doesn't actually translate into any conscious discrimination, then it's pretty much just an academic result. My view is that a positive neurological result should be a springboard into an actual sensory test. Perhaps you and others disagree. But I'm reminded of the whole Ōhashi brouhaha whenever this topic comes up.

 

[castleofargh]

I'm also on board with that. when I abx a file, I'm not trying to become a pioneer in science. I'm just checking if I should bother with a given codec that takes a lot of space or not. as soon as you go lossy, you go lossy and that's a fact. so the point isn't super max resolution, the point is storage space vs sound. it's pointless to argue about how a codec fails, if for me on my gears, it works fine at a conscious level. I already know I'm not getting the original file.

 

[Roly1650]

Yea, but it would not be really that surprising if these codecs were not differentiated by such blind testing, would it? - they were developed by using the same blind testing techniques, after all.

My point was that other newer brainwave testing in neuroscience may well reveal more than is possible with the blind testing used in perceptual testing up to now?  

So it's fair to say you have no view of or are unaware of null testing, neuroscience or abx are the only viable choices? I would argue that null testing is likely to be more precise than neuroscience, with digital audio it's purely mathematical. And after 150 years of research, science has taught us what the finite limits of human hearing are, so null test results are pretty unequivocal, which is probably why null testing is carefully avoided by those not seeking real answers.
It would seem to me that the neuroscience solution is not likely in a living room near me anytime soon, whereas both abx and null testing can be carried out by anyone, anywhere, testing whatever, providing them with results meaningful to them, (unless of course they don't get the result they expect, god forbid they find out they're under the bell curve along with the rest of us).

 

[uchihaitachi]

So it's fair to say you have no view of or are unaware of null testing, neuroscience or abx are the only viable choices? I would argue that null testing is likely to be more precise than neuroscience, with digital audio it's purely mathematical. And after 150 years of research, science has taught us what the finite limits of human hearing are, so null test results are pretty unequivocal, which is probably why null testing is carefully avoided by those not seeking real answers.
It would seem to me that the neuroscience solution is not likely in a living room near me anytime soon, whereas both abx and null testing can be carried out by anyone, anywhere, testing whatever, providing them with results meaningful to them, (unless of course they don't get the result they expect, god forbid they find out they're under the bell curve along with the rest of us).

Brain scans will be a source epic confounders. How would you possibly begin to isolate psychoacoustic effects

 

[mmerrill99]

   
No doubt, but the question then comes back to the old "statistical significance vs. actual significance" problem. My feeling is that if a better-than-baseline signal doesn't actually translate into any conscious discrimination, then it's pretty much just an academic result. My view is that a positive neurological result should be a springboard into an actual sensory test. Perhaps you and others disagree. But I'm reminded of the whole Ōhashi brouhaha whenever this topic comes up.

Ah, but not being able to consciously identify a difference in a blind test does not mean we aren't affected by such difference - in other words we may feel uncomfortable/un-engaged/etc when listening - something that doesn't necessarily show up in A/B testing but yet makes a substantial difference to our listening.
 
This is one of the reasons why I asked the question I started with - do we know enough about the workings of auditory perception to be able to pin down measurements?
 
Recent research is providing strong evidence that the auditory system summarizes the temporal details of sounds using time-averaged statistics. This has some consequences for perceptual testing - for instance when two dissimilar "sound textures" are being discriminated, repeats & longer duration samples aids in the discrimination. But when similar "sound textures" are being compared, using repeats & longer duration samples, discrimination performance declined. The explanation appears to be that such statistical representations produce good categorical discrimination, but limit the ability to discern temporal detail.
 
We still have a lot yet to be discovered about auditory processing - not just peripheral issues but I believe some fundamental issues

 

[mmerrill99]

So it's fair to say you have no view of or are unaware of null testing, neuroscience or abx are the only viable choices? I would argue that null testing is likely to be more precise than neuroscience, with digital audio it's purely mathematical. And after 150 years of research, science has taught us what the finite limits of human hearing are, so null test results are pretty unequivocal, which is probably why null testing is carefully avoided by those not seeking real answers.
It would seem to me that the neuroscience solution is not likely in a living room near me anytime soon, whereas both abx and null testing can be carried out by anyone, anywhere, testing whatever, providing them with results meaningful to them, (unless of course they don't get the result they expect, god forbid they find out they're under the bell curve along with the rest of us).

When it comes to perceptual testing - null testing has no place! It's the equivalent of the bridge analogy I gave already - you can prove that the bridge works as designed but it can still make people nauseous crossing it due to the frequency of foot falls & the resonance set up in the bridge structure. 
 
 
Ah you highlight exactly the problem I'm talking about (although you cite it as an advantage) - "whereas both abx and null testing can be carried out by anyone, anywhere, testing whatever, providing them with results meaningful to them, (unless of course they don't get the result they expect, god forbid they find out they're under the bell curve along with the rest of us)."

 

[mmerrill99]

Brain scans will be a source epic confounders. How would you possibly begin to isolate psychoacoustic effects

To the layman, this may appear to be the case but it's not necessarily the case

 

[uchihaitachi]

To the layman, this may appear to be the case but it's not necessarily the case

You mentioned an EEG, which can still fail to show differences even for individuals who have epilepsy. It is hardly foolproof. I'm curious how you would go about implementing these tests to test for psychoacoustic perception.