Discrimination of High-Resolution Audio without Music: http://www.aes.org/e-lib/browse.cfm?elib=20690
They used test tones, not music. Can anyone with access to a PDF copy provide more information about their methodology etc.?
The Stereophile article about it is here: https://www.stereophile.com/content/hi-rez-audio-distinguished-blind-testing
The typical reason why sampling at a higher rate than 44,100 could produce better results is the distortion some sound cards or DACs can produce at that rate. It could also be the lowpass [filter] for a device that makes the results less clear. But when it comes to the math concerning frequency and time domain, when you're dealing with 16 bits or up, the phase information is very much intact leaving only the Nyquist limit—which is also removed during sustained lowpass.
There is, however, a case where high frequency transients can leak through a lowpass, but that only occurs when the lowpass isn't sustained—in software or with an energy-efficient DAC that shuts off after completion.
My ears are good enough that I can sometimes tell the difference between 44K and 48K, but that's usually because of the lowpass. It's easier cutting off at 22KHz sampling at 48KHz.
They used noise and impulse signals, and so the subjects were listening to noise. And so dither characteristics would be relevant. As High resolution audio sampling can be dithered ultrasonically it is quite reasonable that they could hear a difference.
I'd be cautious when trusting a japanese study done about limits of human hearing. The older japanese studies related to this were full of holes and lots of missing checks of the test enviroment itself.
In Japan it seems the audiophile crowd still loves their CDs and these "studies" may only try to push HiBit.
The AES history of high resolution audio papers lately is not a prime example of bullet proof setups also.
Can anyone with access to a PDF copy provide more information about their methodology etc.?
Try this one: https://www.audiosciencereview.com/forum/index.php?threads/new-paper-on-high-resolution-audio-perception-without-music.10425/
Discrimination of High-Resolution Audio without Music
Duh
How long has AE$ been trying to validate the industry’s push toward so-called hi-res?
Maybe pulling climate change argument would stop them: more res = more processing = more power draw = more fossil fuels burnt. Multiply that by number of audio commodities and Greta Thunberg might get interested in the topic.
Privilege, old, white, male, denial. I don’t think that would work.
I should add hard of hearing to the list.
Based on the discussion on here, reddit and audiosciencereview, the study and results seem very suspect. Sorry for wasting everyone's time :) . Glad to see that TOS 8 is still valid and going strong!
Maybe pulling climate change argument would stop them: more res = more processing = more power draw = more fossil fuels burnt. Multiply that by number of audio commodities and Greta Thunberg might get interested in the topic.
Gotta CHUD up other topics too, I see.
Hypothetically speaking, the difference in listening enjoyment between, say, 96 kbps opus (what I use for listening purposes) and, say, 24/96 FLAC (if there even is a difference) is likely at least several orders of magnitude smaller than the difference in file sizes unless the opus encoder messes up badly and has some sort of annoying artifact.
Maybe pulling climate change argument would stop them: more res = more processing = more power draw = more fossil fuels burnt. Multiply that by number of audio commodities and Greta Thunberg might get interested in the topic.
Still insignificant compared to the power consumption of e.g. bitcoin mining.
But we're getting way off-topic here.
Pretty sure he’s just playing the troll.
One possible explanation -- and I do have a very weak anecdotal example (almost all complaints about 44.1k/16 result from anecdotal rather than truly controlled scientific comparisons), that there MIGHT be reasonable possiblity of detecting differences in distortion due to realtively dirty audio files. Most 'hi res' files that I have seen do seem to have lots of wierd noise in them, and also the characteristic NR decoding splats (that is, IMD & MD created by fast gain control producing synthsized audio that extends from the normal audio frequencies to beyond 20kHz). The distortion products of significant level at all can be difficult for common amplifiers to deal with very cleanly... Also, transducers can create variables. (Op-amps often don't perform as well as data sheets imply due to input resistance/capacitance changes and other 2nd order effects.) We all know these things, just providing an anecdotal result from someone who is a 44.1k/16bit skeptic (I happen to believe in 44.1k/16bits, but prefer 48k.)
I have some anecdotal information from an audiophile friend who claims that my near perfect 'remastering' of Crime of the Century -- some people believe that it blows away other hi-res versions and even better than the original vinyl, sounds just as clean at 44.1k/16bits. This person/friend is the kind who doesn't believe in 44.1k/16 at all, and likes the results at CD rates. The difference between my home-brew remasters and those normally available is that the signal is extremely clean. Still has the over-enhanced vocal sound from the album, but the details are pristine. (I do the so-called remastering for testing purposes, not really for distribution -- people have asked me about them though.) One note: I have fixed the mild over-enhancement on the Supertramp albums also, but that is another story for another time -- it does appear to be a DolbyA decoding issue.
I don't get into the discussions when someone claims that they can percieve the differences -- however, I'd like to see some studies based upon clean audio signals rather than the noisy stuff that is often available.
John
My ears are good enough that I can sometimes tell the difference between 44K and 48K, but that's usually because of the lowpass. It's easier cutting off at 22KHz sampling at 48KHz.
So, you claim you can hear higher than 22 kHz?
My ears are good enough that I can sometimes tell the difference between 44K and 48K, but that's usually because of the lowpass. It's easier cutting off at 22KHz sampling at 48KHz.
So, you claim you can hear higher than 22 kHz?
I don't see him claiming that. He says he can hear the difference in filters. Don't know all the conditions of his claim, and there's no mention of TOS 8 validation.
My ears are good enough that I can sometimes tell the difference between 44K and 48K, but that's usually because of the lowpass. It's easier cutting off at 22KHz sampling at 48KHz.
So, you claim you can hear higher than 22 kHz?
I don't see him claiming that. He says he can hear the difference in filters. Don't know all the conditions of his claim, and there's no mention of TOS 8 validation.
Then pray tell how
you determined whether ajp9 claim is for <22k or >22k being "heard"?
I see a question mark at end of Rollin's post.
How is going SAM? Still defending your D-K pal little Johnny?
People aren't necessarily hearing the actual difference between the capability of 44.1k vs 48k vs 96k, but instead if you start with normal 96k material, there is sometimes a lot of HF stuff in there that might screw with even the best op amps. There could certainly be very subtle differences are not just sample rate effects or frequency response effects -- other than the distortion created by subsequent analog circuitry. Some distortion is maybe possible if any math is done in 16bits, therefore further damaging 16bit results.
Was just thinking about if any filters would be implemented in 16bits, can there be nonlinearities when doing the individual tap multiplications where each step isn't dithered? If there are *possible* issues there (I havent' sat down and done the calculations -- not sure), then any filtering might even have enough error to be noticeable.
There are so many strange effects, if you think about every step in the math, that I'd suspect that there have been few, if any, fully controlled experiments to find out where the audible differences can come from...
I wouldn't trust any analysis without carefully reviewing EXACTLY what is going on, controlling for HF effects on analog circuitry, controlling for any odd effects of truncated math, etc.
Of course, this is all assuming that a proper ABX testing for comparison is being done.
John
My ears are good enough that I can sometimes tell the difference between 44K and 48K, but that's usually because of the lowpass. It's easier cutting off at 22KHz sampling at 48KHz.
So, you claim you can hear higher than 22 kHz?
I know I can hear a little above 20KHz, being able to tell Billy Joel's "Big Shot" from the track cut off at 20KHz. (Intense high frequencies with the cymbals.) But above 22KHz, I wouldn't bother. You get into territory where the loudness required may be damaging to the ear.
The point really was with the roll-off filter applied, and to avoid aliasing some can have an excessive transition band. Or some may assume that at such a high sample rate no filter is necessary.
Was just thinking about if any filters would be implemented in 16bits, can there be nonlinearities when doing the individual tap multiplications where each step isn't dithered?
At 16-bits, factors like precision and dithering are less likely to have an audible effect on sound quality. For DSP, it's factors like echo (unsuitable window type) or an inadequate number of polyphases.
All in all, if the tests were done properly there would be no audible difference between 44 and 48K, speaking from a mathematical standpoint of knowing impulse response filters and how they preserve signal energy. And that most people can't adequately hear above 22KHz.
My ears are good enough that I can sometimes tell the difference between 44K and 48K
I know I can hear a little above 20KHz, being able to tell Billy Joel's "Big Shot" from the track cut off at 20KHz. (Intense high frequencies with the cymbals.) But above 22KHz, I wouldn't bother.
All in all, if the tests were done properly there would be no audible difference between 44 and 48K
:-\