[eq1849]

It doesn’t really contradict the previous study, it just expands on it. The only thing it arguably contradicts is the “suggested benefits”, which were merely a suggestion anyway and based on correlation between certain mental/emotional states and certain brainwave types but of course correlation does not indicate causation. For example, a higher alpha wave power is associated with (results from) higher arousal but higher alpha wave power may not cause higher arousal.

Sure, it only investigated (and effectively eliminated) the filter ringing and >22kHz aspects of “hi-res”. The other aspect of “Hi-res” (16bit vs 24bit) was not investigated but has been by other studies and is somewhat easier because it’s a relatively simple question of physics/acoustics and noise floor levels/thresholds rather than frequencies which affect brain wave patterns.

Not sure I understand the question. If they were aware that they were “hearing hi-res over lower resolution” then they wouldn’t have been “blind(ed)”.

G

I think you and I are on the same page regarding Ultra High Frequencies(UHF) not appearing to effect EEG when listening to music.

Is that an accurate statement?

But you also admit, there is more to high resolution audio than simply UHF, correct?

high resolution audio also includes: improved quantization, which results in reduction of errors, reduced noise floor, reduction of artificial digital noise added to the signal as compared to 16 bit, etc. More samples give more points of time reference to the DAC, perhaps improving timing and filtering.

 

[bigshot]

There's no audible difference between high bit/sampling rates and 16/44.1. All those extra zeros and ones might as well be packing peanuts. They are as useless as teats on a bull hog. Ultrasonic frequencies only exist at very low levels in commercially recorded music, if at all. And the small amount that may exist in cymbals is masked, so even if you could hear it, you couldn't hear it. Perhaps a SACD of gamelan might have significant levels, but I don't know of any SACDs of gamelan, do you? 16/44.1 is audibly transparent on any metric you want to point at... frequency response, errors, noise, timing... There's no difference to be heard. It doesn't get any better.

Sound quality is best judged by sound you can actually hear, not sound you can't. That seems self-evident, but people still argue in favor of some etherial magic mojo that can't be heard or measured as being the key to a good sound system. Of course the argue that while having large response deviations, transducer distortion, the unnatural sound presentation of headphones, or complete lack of room treatment with speakers. They focus on thin air instead of the elephant in the corner.

 

[GearMe]

There's no audible difference between high bit/sampling rates and 16/44.1. All those extra zeros and ones might as well be packing peanuts. They are as useless as teats on a bull hog. Ultrasonic frequencies only exist at very low levels in commercially recorded music, if at all. And the small amount that may exist in cymbals is masked, so even if you could hear it, you couldn't hear it. Perhaps a SACD of gamelan might have significant levels, but I don't know of any SACDs of gamelan, do you? 16/44.1 is audibly transparent on any metric you want to point at... frequency response, errors, noise, timing... There's no difference to be heard. It doesn't get any better.

Sound quality is best judged by sound you can actually hear, not sound you can't. That seems self-evident, but people still argue in favor of some etherial magic mojo that can't be heard or measured as being the key to a good sound system. Of course the argue that while having large response deviations, transducer distortion, the unnatural sound presentation of headphones, or complete lack of room treatment with speakers. They focus on thin air instead of the elephant in the corner.

Wait a minute; the Elephant is in the corner?

Well...that's progress!

 

[bigshot]

Do you have a cartoon about "Teats on a bull hog"?

 

[71 dB]

high resolution audio also includes: improved quantization, which results in reduction of errors,

No, because proper dithering ensures zero quantization error/distortion.

reduced noise floor,

True, a noise floor further under the threshold of hearing in any practical/reasonable listening scenario. The audible benefit of this is zero, because if something is already totally inaudible, you can't make it even more inaudible. About 13 bits is enough in consumer audio meaning 16 bit is already enough and then some. 24 bit belongs to audio production where it has practical benefits.

reduction of artificial digital noise added to the signal as compared to 16 bit, etc.

As mentioned above, the added noise aka dither is inaudible in any practical/reasonable listening scenarios at 16 bit.

For the "etc." I ask what exactly?

More samples give more points of time reference to the DAC, perhaps improving timing and filtering.

As unintuitive as it sounds, higher sample rate don't improve "timing." The temporal resolution ∆Tres of digital audio is calculated using this formula:

∆Tres = 1/2*𝜋*fsig * Nsteps​

where fsig is the signal frequency and Nsteps is the amount of quantization steps in the signal. If we have audio in the bandwith 0-20 kHz covering human hearing range, it doesn't matter if the sample rate is 44.1 kHz or 96 kHz or 192 kHz. ∆Tres will be the same. Bit depth affects ∆Tres thou. However, 16 bit offers about 1000 times better ∆Tres than our hearing.

Higher sample rate can improve filtering. This is why almost all DACs use oversampling. We don't need the audio be at high samplerate all along. All we need is to oversample it for filtering/processing if need be.

 

[gregorio]

I think you and I are on the same page regarding Ultra High Frequencies(UHF) not appearing to effect EEG when listening to music.
Is that an accurate statement?

I’m not sure, not really. I don’t dispute that ultrasonic frequencies affect EEG. I just dispute that it makes any kind of practical difference, it cannot be heard in any sense, the effect on EEG cannot be perceived and does not cause any behavioural or psychophysical response which we can experience.

high resolution audio also includes: improved quantization, which results in reduction of errors, reduced noise floor, reduction of artificial digital noise added to the signal as compared to 16 bit, etc.

True, but that’s all the same thing. As explained by @71 dB, dither eliminates all quantisation error resulting in a raised digital noise floor but it’s raised from -144dB with 24bit to around -120dB with 16bit. But the lowest noise floor we can achieve in practice with 24bit is around -120dB due to analogue components in the ADC/DAC, which is roughly the same as with 16bit and way below audibility anyway.

More samples give more points of time reference to the DAC, perhaps improving timing and filtering.

44.1/16 provides a timing accuracy down to around 110 picoSecs, human hearing on the other hand has a threshold down to around 6 microSecs. So, 16/44 has roughly 50,000 times more timing resolution than we can hear, what benefit do you think you’d get from even more timing resolution?

Filtering is improved in the sense that the anti-alias and anti-image filters can be implemented in the digital domain and therefore any filter artefacts with 44.1kFS/s reduced to inaudibility, which as 71 dB also mentioned, is why some consumer DACs have had oversampling since CDs were first released and virtually all DACs and ADCs have employed oversampling for more than 30 years. Higher than 44.1kFS/s just makes the filter artefacts that are inaudible to humans also inaudible to a wider range of animals, such as dogs, although you might need to use 192kFS/s for bats! Lol

G

 

[GearMe]

Do you have a cartoon about "Teats on a bull hog"?

Hmmm....think I'll pass on that one

 

[eq1849]

No, because proper dithering ensures zero quantization error/distortion.

At the expense of the signal's accuracy...

By modifying the original signal, it is by definition less accurate.

There's no free lunch on this.

 

[gregorio]

At the expense of the signal's accuracy...

Errr you have that backwards! As the quantisation error has been eliminated then the “accuracy” is literally perfect. Your assertion doesn’t make sense, how can you get a more accurate signal than a representation with zero error?

By modifying the original signal, it is by definition less accurate.

No, again the opposite! By applying dither we’re removing the (quantisation) error and therefore by definition making it more accurate! Effectively, the error is converted into uncorrelated noise.

G

 

[eq1849]

Errr you have that backwards! As the quantisation error has been eliminated then the “accuracy” is literally perfect. Your assertion doesn’t make sense, how can you get a more accurate signal than a representation with zero error?

No, again the opposite! By applying dither we’re removing the (quantisation) error and therefore by definition making it more accurate! Effectively, the error is converted into uncorrelated noise.

G

I don't think a modified signal on a random integer would not be more "accurate" than the Unmodified signal on the closest integer to the unmodified signal.

You give up some "accuracy" for less noise.

The original signal lies along certain integers, no? The dithered signal will lie along slightly different integers, no?

 

[VNandor]

By modifying the original signal, it is by definition less accurate.

To clear up on this: the "original" signal is the analog signal that gets encoded. If no dithering is used you get what's called "quantization error". This error can correlate (move together) with the signal which means that it can effectively cause nonlinear distortion. This nonlinear distortion gets reduced if the quantization uses more steps. Dithering on the other hand entirely removes the correlated error and replaces it with uncorrelated noise. More quantization steps reduce the noise. There's no "free lunch" so to speak of, the price to pay for eliminating distortion is the increased noise. However, the increased noise is practically always negligible compared to both the noise level of the recording and the listening environment. To put it in perspective if the recording had -40dB of noise and the quantization caused -80dB of noise, this would sum up to a total of -39.91dB of noise.

 

[71 dB]

At the expense of the signal's accuracy...

At the expense of a little bit higher noise level, which isn't a problem because:

Dither is more "pleasant" noise than granulating quatization error noise (if volume raised to insane levels actually hear it!).
The noise is far below audibility at realistic listening scenarios anyway.

By modifying the original signal, it is by definition less accurate.

This is kind to true, at least theoretically, but the "accuracy" we lose is far below audibility. We give out accuracy the smartest way possible so that we only suffer increased noise level, which IS NOT A PROBLEM, as I explained above. You could say we lose "accuracy" we never needed.

There's no free lunch on this.

Actually in this case there is free lunch. That's the miracle of dithering and digital audio. It is possible because 24 bit is ridiculous overkill in consumer audio. 13 bit is about what we need, 16 bit is enough and then some and 24 bit is over the top.

 

[71 dB]

I don't think a modified signal on a random integer would not be more "accurate" than the Unmodified signal on the closest integer to the unmodified signal.

You give up some "accuracy" for less noise.

The original signal lies along certain integers, no? The dithered signal will lie along slightly different integers, no?

If you round 7.8 to integer, you get 8 every time and you are systematically wrong by 0.2, but if you add "dither" to the 7.8 before rounding, you get 80 % of the time 8, but 20 % of the time 7 which means you get statistically 7.8 (the average). You are statistically 100 % correct, but there is dither noise on top of your signal. If you can't hear the dither, you just enjoy being correct!

 

[gregorio]

Modified signal on a random integer would not be more "accurate" in my book than Unmodified signal on the closest integer to the unmodified signal in my book.

How is a signal distorted by quantisation error more accurate than the same signal but without that error? I just can’t see how anyone could arrive at that conclusion? Fortunately, no one uses or has ever used “your book”.

G

 

[bigshot]

Any recording is limited and/or inaccurate at the extremes. If that limit falls below the threshold of what human ears can hear, it doesn’t matter. Wasting time worrying about inaudible sound is silly. Audiophiles can be very silly.