[PiSkyHiFi]

OK then AAC 256 and other high bitrate lossy codecs are *audibly identical* to redbook and HD audio. Once it hits the point that it's audibly identical, the sound quality can't get any better. The file just gets bigger.

I would add a caveat that they are mostly audibly identical, because the variance as you have mentioned before in humans testing their own hearing abilities can be large.

Different humans have different hearing capabilities.
The same human can experience audio differently from one day to the next.
The function of memory and audio is not that well understood, even though there is much research, a lot of it conflicts because when we discuss what a human can and can't hear in an audio piece, we are also studying their consciousness, which is a really tricky one.
The dynamic range of human hearing shifts - ears adjust in a dynamic way to what they hear and humans can not hear everything at once.

All of that basically says that a study that finds that a bunch of people reported hearing no difference between lossless and lossy is actually going to have many uncertainties.

This is why I think it's important to have headroom by storing above your hearing capability to a fair degree, make those files a bit larger than just barely adequate to allow for potentially hearing things differently on different equipment at different times, even though it may not happen very often.

Also, as you said, there is no point compressing my collection if it's fine as it is as lossless and I can stream etc...

 

[shortwavelistener]

Is it possible that that ADPCM decoder is rendering to something other than 16 bit ? It's like the original was mastered with a low pass filter into 16 bit, showing nothing above 21KHz, probably deliberately.

It's like the ADPCM file has aliasing right up to 22050 Hz - either just noise from the codec or possibly rendered into floating point or something.

I recall using wav files that had details all the way up to 22050 and their AAC versions losing details above about 18.5 KHz - kind of the opposite of what I'm seeing here.

Edit: I guess you're just pointing out a nicely mastered recording having noise added by the ADPCM codec.

Try 576Kbps 24 bit ADPCM if you can, see what this shows. All we can tell with the graph is where aliasing is added really, accuracy of signal is not revealed just by looking at frequencies present.

Edit again, not aliasing, sorry, I mean artifacts.

BTW i transcoded the original WAV PCM file into ADPCM using Adobe Audition 1.5. I don't know if it was originally rendered into floating point, but by default audio files loaded into Audition are rendered as 32-bit floating point. But the ADPCM files are not dithered.

 

[bigshot]

I would add a caveat that they are mostly audibly identical, because the variance as you have mentioned before in humans testing their own hearing abilities can be large.

There is a huge body of evidence that shows that isn't true when it comes to lossy. If there are any people walking the Earth that can discern a difference, they are very rare birds indeed, and they still can't prove that they can hear a difference to statistical levels that establish that conclusively. I posted a couple of controlled tests that showed that neither age, nor being a professional sound engineer, nor a golden ear audiophile, nor listening with fancy "resolving" equipment makes any difference at all. Beyond a certain point, lossy audio achieves transparency so that any attempt to discern it falls into a typical bell curve that represents random chance.

Human hearing abilities are well established. They've been established for nearly a century. The figures you see commonly quoted about audible frequencies and distortion levels are almost always best case scenarios using test tones. They are already overkill. With music, the threshold can be an order of magnitude lower. Numerous tests have been done comparing different codecs and formats. There is a consensus there, and it's easy to find out for yourself by doing some controlled listening tests.

You can think that you can tell the difference, but you've never done a blind comparison. I offered you one and you said you weren't interested. You weren't interested in the controlled tests that I cited for you either. You've offered no proof to back up what you say, just semantic stuff about how your definitions of words are different than mine and smoke and mirrors about how "no one can ever know for sure". That says a lot.

I'm not angry or frustrated, just disappointed. I had hopes that you might actually listen. You clearly know some things about how audio works. But you refuse to accept any information from outside your sphere of comfort. That's a good way to build an impermeable intellectual bubble around your knowledge and not let anything new in.

And when I use the word impermeable, I don't mean "Maybe sometimes in certain cases it is permeable, but we can't ever really know for sure."

 

[PiSkyHiFi]

There is a huge body of evidence that shows that isn't true when it comes to lossy. If there are any people walking the Earth that can discern a difference, they are very rare birds indeed, and they still can't prove that they can hear a difference to statistical levels that establish that conclusively. I posted a couple of controlled tests that showed that neither age, nor being a professional sound engineer, nor a golden ear audiophile, nor listening with fancy "resolving" equipment makes any difference at all. Beyond a certain point, lossy audio achieves transparency so that any attempt to discern it falls into a typical bell curve that represents random chance.

Human hearing abilities are well established. They've been established for nearly a century. The figures you see commonly quoted about audible frequencies and distortion levels are almost always best case scenarios using test tones. They are already overkill. With music, the threshold can be an order of magnitude lower. Numerous tests have been done comparing different codecs and formats. There is a consensus there, and it's easy to find out for yourself by doing some controlled listening tests.

You can think that you can tell the difference, but you've never done a blind comparison. I offered you one and you said you weren't interested. You weren't interested in the controlled tests that I cited for you either. You've offered no proof to back up what you say, just semantic stuff about how your definitions of words are different than mine and smoke and mirrors about how "no one can ever know for sure". That says a lot.

I'm not angry or frustrated, just disappointed. I had hopes that you might actually listen. You clearly know some things about how audio works. But you refuse to accept any information from outside your sphere of comfort. That's a good way to build an impermeable intellectual bubble around your knowledge and not let anything new in.

And when I use the word impermeable, I don't mean "Maybe sometimes in certain cases it is permeable, but we can't ever really know for sure."

We've been through all of this already. You would have to admit that out of you and I, the one that insists that even though I can still always tell the difference between a sound system and a live performance, we've reached the pinnacle of transparency and that's the end of the story, is the more closed minded.

I draw the line at redbook, it has just enough overhead to mean I don't need to worry about re-encoding, I am free to focus on different aspects of the same recording without ever thinking, was that a compression artifact or bad master?
Even when I'm testing new equipment, I don't want to have to even consider that question.

Most of the time I don't mind listening to both lossy or lossless because it is mostly audibly identical for certain.

Done.

 

[PiSkyHiFi]

BTW i transcoded the original WAV PCM file into ADPCM using Adobe Audition 1.5. I don't know if it was originally rendered into floating point, but by default audio files loaded into Audition are rendered as 32-bit floating point. But the ADPCM files are not dithered.

I think it's just introduced artifacts because this particular master looks like it has used a low pass filter instead of dithering to avoid aliasing and then the ADPCM messed up the high end a fair bit - quite a lot really, it looks worse than what I remember from comparing AAC - I was trying to find the test results I did ages ago, I'll keep searching.

This strengthens my choice of AAC over Apt X at these rates.

I still prefer Aptx HD for Bluetooth until it can do lossless, I think it's closer to the PCM than AAC

 

[castleofargh]

We've been through all of this already. You would have to admit that out of you and I, the one that insists that even though I can still always tell the difference between a sound system and a live performance, we've reached the pinnacle of transparency and that's the end of the story, is the more closed minded.

I draw the line at redbook, it has just enough overhead to mean I don't need to worry about re-encoding, I am free to focus on different aspects of the same recording without ever thinking, was that a compression artifact or bad master?
Even when I'm testing new equipment, I don't want to have to even consider that question.

Most of the time I don't mind listening to both lossy or lossless because it is mostly audibly identical for certain.

Done.

the differences between a live experience and a recording are much more than a file format. when you argue that way you use the same sort of fallacies you used to argue that psychoacoustic models of data suppression, and plain compression, could be reduced to comparing the data rate to prove one superior. each time you dismiss the most important elements to focus on the one that can make you look right.
if you can pass a blind test with AAC at 256kbps vs redbook where the difference isn't a change in loudness or in mastering(you need to be sure of both or the test means nothing), then you can claim and prove that AAC is indeed not transparent to your ear. that in turn would absolutely legitimize your need for a higher format.
but if you fail to do that, like @bigshot, I, and so many others have failed to do, you're not looking to get higher resolution for audible transparency, you're looking to get it because you want it. which is fine, just a very very important distinction.

with all said and done, my own BT headphone is far from providing transparency, let alone the resolution AAC has to offers. I notice background hiss on silent passages or in a quiet room at quiet yet comfy volume level(the cause being a sub par internal amp). I measure distortions well above -60dB(some of it is caused by my microphone but I've measured stuff way below often enough to know that my BT headphone generate the higher levels I measure). I mention -60dB because that's the area where AAC usually starts removing some elements that the algorithm estimated will be hidden by auditory masking anyway. anytime a headphone reaches 1% THD somewhere, that's extra signal at 40dB below signal. -60db is 0.1%. a level we consider really great for headphones. and I won't start with the flawed stereo, the non flat FR and other issues still attached to typical headphone use to this day. if you want the sound like it was recorded, those are the stuff that need to be improved drastically, not file format.

now don't get me wrong, when @Monstieur was saying that AAC was clearly superior to other stuff, I was almost contradicting him because I haven't seen clean controlled experiments demonstrating that it is consistently so. personally, I feel that the gears I use, including the source, still have a lot of impact on which setting will work best for certain uses. but that's mostly gut feelings, I haven't thoroughly tested anything, I wanted to look up battery consumption but never even got to test that properly. and now it's too hot for closed headphones so I'm back on IEMs until next nuclear winter(or just winter, whichever comes first). so to stop talking for no reason, I do expect gears where AAC is the obvious choice, but I also expect gear combos and listening conditions where APTx or APTxHD or the sony stuff, will just work better(if only because it's the only thing they have to offer and they're good enough to motivate their use anyway). I really don't know what's audibly best, and something most certainly is. I'm just being overly cautious when I read general claims based on anecdotal experiences or half objective arguments.

 

[bigshot]

I had hopes tor this one. Sometimes people stumble in here with potential. I might have been able to learn something from him about codecs I'm not familiar with like Aptx. But with no interest in backing what he says up, and falling back on semantic arguments and deflection to validate his opinions, I don't see much hope. I redefined my wording to eliminate the word transparent that he wasn't able to understand, and he slipped back to "we can never know". If I provide evidence that we actually can know, he slips back to semantics. The clue has been given. You can lead a horse to water...

 

[PiSkyHiFi]

the differences between a live experience and a recording are much more than a file format. when you argue that way you use the same sort of fallacies you used to argue that psychoacoustic models of data suppression, and plain compression, could be reduced to comparing the data rate to prove one superior. each time you dismiss the most important elements to focus on the one that can make you look right.
if you can pass a blind test with AAC at 256kbps vs redbook where the difference isn't a change in loudness or in mastering(you need to be sure of both or the test means nothing), then you can claim and prove that AAC is indeed not transparent to your ear. that in turn would absolutely legitimize your need for a higher format.
but if you fail to do that, like @bigshot, I, and so many others have failed to do, you're not looking to get higher resolution for audible transparency, you're looking to get it because you want it. which is fine, just a very very important distinction.

with all said and done, my own BT headphone is far from providing transparency, let alone the resolution AAC has to offers. I notice background hiss on silent passages or in a quiet room at quiet yet comfy volume level(the cause being a sub par internal amp). I measure distortions well above -60dB(some of it is caused by my microphone but I've measured stuff way below often enough to know that my BT headphone generate the higher levels I measure). I mention -60dB because that's the area where AAC usually starts removing some elements that the algorithm estimated will be hidden by auditory masking anyway. anytime a headphone reaches 1% THD somewhere, that's extra signal at 40dB below signal. -60db is 0.1%. a level we consider really great for headphones. and I won't start with the flawed stereo, the non flat FR and other issues still attached to typical headphone use to this day. if you want the sound like it was recorded, those are the stuff that need to be improved drastically, not file format.

now don't get me wrong, when @Monstieur was saying that AAC was clearly superior to other stuff, I was almost contradicting him because I haven't seen clean controlled experiments demonstrating that it is consistently so. personally, I feel that the gears I use, including the source, still have a lot of impact on which setting will work best for certain uses. but that's mostly gut feelings, I haven't thoroughly tested anything, I wanted to look up battery consumption but never even got to test that properly. and now it's too hot for closed headphones so I'm back on IEMs until next nuclear winter(or just winter, whichever comes first). so to stop talking for no reason, I do expect gears where AAC is the obvious choice, but I also expect gear combos and listening conditions where APTx or APTxHD or the sony stuff, will just work better(if only because it's the only thing they have to offer and they're good enough to motivate their use anyway). I really don't know what's audibly best, and something most certainly is. I'm just being overly cautious when I read general claims based on anecdotal experiences or half objective arguments.

"the differences between a live experience and a recording are much more than a file format"

I totally agree, and I am fairly certain the scientists that drew conclusion about AAC's "audible transparency" pretty much ignored their unkonwn 18 bit DAC from 1991, it's cheap power supply, their Pioneer Amp and B+W speakers or Sennheiser HD 450 in terms of if they could even allow for the differences to be possibly heard - where does that leave audiophiles that know this is relevant?

You should go with your gut, it's merely saying there is some uncertainty at the boundaries that should be avoided if you don't want to repeatedly ask yourself if it really is enough, every time you hear something new.

 

[bigshot]

A lot of us here have verified the transparency of high data rate AAC for ourselves. We aren’t depending on old or obsolete or cheap equipment. In fact I cited a published test that showed that AAC was audibly transparent on current high end equipment too. Did you read that?

You’re in the wrong forum for advising to go with your gut.

 

[PiSkyHiFi]

A lot of us here have verified the transparency of high data rate AAC for ourselves. We aren’t depending on old or obsolete or cheap equipment. In fact I cited a published test that showed that AAC was audibly transparent on current high end equipment too. Did you read that?

You’re in the wrong forum for advising to go with your gut.

Scientifically, one should provide error bars or a degree of certainty based upon assumptions, *if* you are making claims of relative certainty.

I am the one making claims that the error wasn't included or analysed well enough to be certain about it, which means gut is sufficient to cast doubt about how these things were done.

We're talking an industry where they were mostly focussed on how to get AAC to sound better than the others at 96Kbps for quite a while, because their standards for audibly transparent were mostly concerned with telephony.

Yes, I'd like to read that article - I missed the link, please direct me.

 

[PiSkyHiFi]

Scientifically, one should provide error bars or a degree of certainty based upon assumptions, *if* you are making claims of relative certainty.

I am the one making claims that the error wasn't included or analysed well enough to be certain about it, which means gut is sufficient to cast doubt about how these things were done.

We're talking an industry where they were mostly focussed on how to get AAC to sound better than the others at 96Kbps for quite a while, because their standards for audibly transparent were mostly concerned with telephony.

Yes, I'd like to read that article - I missed the link, please direct me.

Also, what about that dynamic range argument, don't you think it makes a mockery of mastering to redbook using dithering to achieve a slightly higher dynamic range, if you then throw it away by remastering to AAC from this dithered version?

I guess you could argue that mastering with dithering is overkill compared to the mostly audibly transparent AAC, but I think I've heard differences in mastering quality anecdotally and so I still err on the side of caution.

One cannot master to AAC without using PCM, it's impossible - not saying you said that, I'm just saying, AAC is an afterthought in mastering and it's to do with saving space mostly.

Mastering to 24/96 and then using AAC, well, now we're talking - that's like using a 4K screen on 13 inch laptop because you just don't want to even know that pixels exist - bring it on I say, that's an excellent way to use AAC, much better than let's take redbook, make it slightly worse and then argue that redbook was overkill - not saying you said that @bigshot , I'm just saying.

 

[bigshot]

Scientifically, one should provide error bars or a degree of certainty based upon assumptions, *if* you are making claims of relative certainty.

The best way to find out for yourself for certain is to do a controlled listening test yourself. Then you don't need to worry about statistics or uncertainty. You'll know.

For lots of great articles on listening tests of various parameters of sound equipment, see the first post in Testing Audiophile Claims and Myths. It's the most useful and interesting post in this forum. https://www.head-fi.org/threads/testing-audiophile-claims-and-myths.486598/

Also, what about that dynamic range argument, don't you think it makes a mockery of mastering to redbook using dithering to achieve a slightly higher dynamic range, if you then throw it away by remastering to AAC from this dithered version?

Whenever you see specs, it's really good to know how those relate to what you actually hear. Do you know what the audible threshold is for a noise floor under music? And do you know what the dynamic range of the most dynamic recorded music is? Both of those figures are between -40dB and -50dB. The dynamic range of AAC 256 VBR is more than enough to achieve audible transparency.

Redbook is overkill. It was designed to cover every extreme and unlikely circumstance. 16 bit was selected for the convenience of the math, not because 16 bit was actually required. The threshold of transparency is somewhere between 12 and 14 bits.

When you work with sound and test it and look at the measurements in relation to how it actually sounds, you have context. Looking at numbers on a page and assuming that bigger is always better is just plain wrong. Human hearing has definite limits. You can shove higher quality sound and bigger numbers into your ears, but it won't sound any different.

Audiophiles know a lot about technical stuff. They read white papers, they compare numbers, they do math to calculate differences... but it's all theory because they spend no time at all researching the thresholds of human perception. That's what we're talking about here. Stuff you can and can't hear. Not abstract numbers. Everyone knows that "lossy" has less information than "lossless". But the information that is missing is inaudible. You can't hear it. That means that to human ears, the sound of audibly transparent lossy and lossless is the same.

 

[PiSkyHiFi]

The best way to find out for yourself for certain is to do a controlled listening test yourself. Then you don't need to worry about statistics or uncertainty. You'll know.

For lots of great articles on listening tests of various parameters of sound equipment, see the first post in Testing Audiophile Claims and Myths. It's the most useful and interesting post in this forum. https://www.head-fi.org/threads/testing-audiophile-claims-and-myths.486598/



Whenever you see specs, it's really good to know how those relate to what you actually hear. Do you know what the audible threshold is for a noise floor under music? And do you know what the dynamic range of the most dynamic recorded music is? Both of those figures are between -40dB and -50dB.

Redbook is overkill. It was designed to cover every extreme and unlikely circumstance. 16 bit was selected for the convenience of the math, not because 16 bit was actually required. The threshold of transparency is somewhere between 12 and 14 bits.

When you work with sound and test it and look at the measurements in relation to how it actually sounds, you have context. Looking at numbers on a page and assuming that bigger is always better is just plain wrong. Human hearing has definite limits. You can shove higher quality sound and bigger numbers into your ears, but it won't sound any different.

Audiophiles know a lot about technical stuff. They read white papers, they compare numbers, they do math to calculate differences... but it's all theory because they spend no time at all researching the thresholds of human perception. That's what we're talking about here. Stuff you can and can't hear. Not abstract numbers.

You haven't listened to anything I've said about the scientific limits of ABX, you come across like a zealot. If this is all you have, I don't need it. I feel we're done.

 

[bigshot]

Blind testing with direct switchable line level matched comparison is the best way to determine if there is an audible difference between two samples. You can think differently about it, but you're wrong. That's how controlled testing works. It's designed to minimize the effects of bias and problems with auditory memory.

You're in Sound Science. In this forum we get to demand proof. A good way of providing that is to point to published controlled tests (which I've done to back up my opinion) and to do tests yourself (which I offered to help you with). Just pointing to abstract numbers and ignoring the thresholds of human auditory perception doesn't cut the mustard. We both agree that lossless contains data that lossy doesn't. Our disagreement is whether that data is audible. I've offered proof that it isn't. You've just pointed back at the numbers.

We're not going in circles because I'm ignoring what you say. The opposite is true.

 

[PiSkyHiFi]

Blind testing with direct switchable line level matched comparison is the best way to determine if there is an audible difference between two samples. You can think differently about it, but you're wrong. That's how controlled testing works. It's designed to minimize the effects of bias and problems with auditory memory.

I have already been through all of this. Please stop now. We disagree and I have nothing more to direct towards you. I'm here for specific issues.
I'm not going to say you're wrong because it's what you believe and you have apparently thrown away your originals, so be happy with what you have and leave the rest of us to explore beyond simplistic reductionist analysis.
Seriously, enjoy the music.