Can you hear the difference between 16bit and 24bit audio files?

[sandyk]

See reply 39

It's the methodology that sucks, as well as his assumptions that any residuals that he measures are always below the audibility threshold, whether by themselves, or when interacting with other areas.

It's like an over 50s speaker designer like Anthony evaluating, by listening, speakers with a usable frequency response well past 20kHz when he can now only hear a sine wave response at a reasonable level below 15kHz. Perhaps he can give meaningful evaluations, but how can he possibly do it according to presently accepted hearing knowledge ?

TBH, I don't doubt that he is able to, without simply relying on measurements, but how does he do it ?

[beetlemania]

Me neither, which why I think if you record in PCM, you better do it at 384 in order to compete on SQ with DSD64

To my ears and in my system, PCM sounds as good or better than DSD. IMO, the DSD-hype is much ado about nothing.

[tne]

You keep ignoring the numerous posts and threads in this forum and elsewhere, that say DBT is fatally flawed for Audio, unlike medicine etc.

 

I am a bit late to this party, and I am one who believes that 24 bit (when done properly) can benefit sound quality, and also I understand some of the flaws of DBTs as currently practiced in audio.

 

However the statement quoted is quite intriguing. Without recycling from other threads, please explain how DBTs can be a valid approach for medicine and other phenomena, but not for audio. I can believe that the way that they are currently practiced in audio could be flawed, but I am not sure how the fundamental well-practiced approach should be different.

[sandyk]

Without recycling from other threads, please explain how DBTs can be a valid approach for medicine and other phenomena, but not for audio.

 

DEMAND IGNORED. I am not a pupil in your classroom/lecture room.

[gmgraves]

I am a bit late to this party, and I am one who believes that 24 bit (when done properly) can benefit sound quality, and also I understand some of the flaws of DBTs as currently practiced in audio.

 

However the statement quoted is quite intriguing. Without recycling from other threads, please explain how DBTs can be a valid approach for medicine and other phenomena, but not for audio. I can believe that the way that they are currently practiced in audio could be flawed, but I am not sure how the fundamental well-practiced approach should be different.

 

 

Without going into a big "thing", as I see it, the difference is that in medicine, the results of a DBT can be measured, and observed on an ongoing basis. The results are fairly unambiguous. In audio, all we have are peoples' temporal observations. After the test, there is no way, other than via consensus to verify the results. There remains nothing to observe and note, and nothing to measure. That means that all we have to rely on is what people said they heard, and what people think they remembered hearing. Human aural memory being as short as it is and the human penchant to be influenced by mob pressure, make these observations, by definition, less reliable than, say, disease symptoms going away, measurable and often visible and measurable improvements in vital functions, while no such visible or measurable results are noted in what turns out to be the control group being given placebos. There may be other reasons as well, but that's what comes to my mind.

[Boris75]

I find the results to be quite compelling.

 

I too, but I still buy high-res files.

[Musicophile]

Without going into a big "thing", as I see it, the difference is that in medicine, the results of a DBT can be measured, and observed on an ongoing basis. The results are fairly unambiguous. In audio, all we have are peoples' temporal observations. After the test, there is no way, other than via consensus to verify the results. There remains nothing to observe and note, and nothing to measure.

I probably should keep out of these discussions, as they never get anywhere.

 

However, having no objective measure is also true for every single pain killer out there that has ever been tested clinically. So probably the FDA would disagree with your statement.

 

Have a look here:

 

 

1. Efficacy endpoints in an analgesic trial should reflect a direct rating of pain intensity by the subject for all settings in which subjects can communicate in a reliable manner. We recommend the use of a well-defined and reliable PRO measure of the subject’s pain intensity.
    
   

 

 

 

 

http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM384691.pdf

 

(A "PRO" is a patient reported outcome).

[Musicophile]

I too, but I still buy high-res files.

Same here.

 

To my ears, with an admittedly upper-mid-fi system, the differences between even a well done 256kbit AAC and a 24/192 files are small at best.

 

However, if I do spend money on an album I really care about, those small differences do matter, otherwise we may as well all just buy Spotify and are done with it.

[Leonardo Piras]

However, if I do spend money on an album I really care about, those small differences do matter, otherwise we may as well all just buy Spotify and are done with it.

 

+1

[EuroDriver]

Sorry, but this sounds like you are looking at it from the perspective of a human trying to draw a best-fit-curve on a graph, rather than sampling theory.

 

As long as the signal is properly bandlimited, the signal will be reproduced correctly.

 

 

 

Just because it doesn't produce pretty waveforms when you display the data incorrectly, doesn't mean that the signal is not being properly represented.

 

Many thanks for the link to the video, I enjoyed it very much. I plan on down loading a text on DSP and taking my time to work through it

[Le Concombre Masqué]

I took the test ; who did ?

You have to be very brave since the 1rst track is so awful I could kill to make it stop

You have to free your mind from the "there should be improved transparency, ambiance, definition" you've read in the instructions. Cause it ain't no DBT : Mr Archimago chose and knows...

 

I first took the test casually playing the flac from A+ on a busy Mac and extra length of usb cable, then I went through my usual process : wav Alloy (eQ within +- 3 dB) osx on a sd card ram disks etc.

 

basically I heard the same phenomena :

 

A+/casual

 

tracks ABB were lighter but options I could not rule out as not being the correct answer to "there should be improved transparency, ambiance, definition" since anyway I can't bear the sound of straight flac through A+ :I didn't like anything so why not the A option for the unbearable 1st track ?

Then the Full Monty on an expensive system going straight and clean below 20 (but not above 20K):

I could physically sense a greater density of the air moved by tracks BAA ; ABB sounded lighter and just far more remote from experiencing a performance.

 

Please keep buying Hires and helping it staying alive. I wish Mr Archimago and everybody who goofed to be able to offer oneself a good system one day.

 

If someone has a mediocre equipment and tries to convince himself and everybody that it's enough it is a very despicable attitude...

 

Nevertheless I don't doubt that Mr Archimago has demonstrated than on a chosen set of tracks he could have half of the people deciding that the 16 bits tracks were a better answer to " there should be improved transparency, ambiance, definition".

 

 

So... buy SD cards, create ram disks, convert to wav for a very cheap starter to a better SQ and keep backing Hires !!

[Paul R]

I am a bit late to this party, and I am one who believes that 24 bit (when done properly) can benefit sound quality, and also I understand some of the flaws of DBTs as currently practiced in audio.

 

However the statement quoted is quite intriguing. Without recycling from other threads, please explain how DBTs can be a valid approach for medicine and other phenomena, but not for audio. I can believe that the way that they are currently practiced in audio could be flawed, but I am not sure how the fundamental well-practiced approach should be different.

 

Well, except in some specific cases, there is a difference between then. Audio testing is purely perceptual - and the best research in that is in the marketing area. Medical research - even when testing perceptual reactions - almost always has corrobative evidence to assist with judging the responses. Blood pressure, for instance, can be statistically correlated to pain among other things.

[Paul R]

Care to substantiate your claims? When using a decent DAC and ADC, you have a nearly transparent loop - even at 16/44.1:

 

Gearslutz.com - View Single Post - Evaluating AD/DA loops by means of Audio Diffmaker

 

The results diff down to -77db for the best DAC/ADC loop. I personally doubt that there can be anything audible that far down, in particular when only going the DAC part of the loop.

 

 

You know as well as I do that it is far from a perfectly transparent loop in all practical circumstances. Higher sample rates allow for better reproduction for many reasons, and the hotly defended 16/44.1 rate is more a compromise of the technology of its time than anything else.

 

You also know perfect reproduction by Shannon/Nyquist requires infinite bandwidth limited signals over an infinite period of time. I suspect you know that ignoring the time domain in audio reproduction, as so many are wont to do, also has audible consequences.

 

There is a difference between engineering and theory- one must know and account for both.

 

On the other hand, nobody is forcing anyone to buy or listen to high res music, so if someone wants to listen to Redbook, or MP3s even, bless their little hearts and pass the earbuds...

[Boris75]

However, if I do spend money on an album I really care about, those small differences do matter, otherwise we may as well all just buy Spotify and are done with it.

 

+1. My admittedly highly subjective perception of high-res vs. CD is that I have a very hard time distinguishing between them when switching between them, but I get ear fatigue after a couple of hours listening to 16/44 files while I can listen to high-res for hours without ever getting tired of the sound (at reasonable sound pressure levels of course).

 

By the way, I am surprised to read that you describe your system as "upper mid-fi". Your signature line suggests that it clearly belongs to the hi-fi category in my understanding of the word. Then, there is the "high-end" category, but I never really understood if a "high-end" is one that sounds extremely well whatever its price or if it has to be very expensive to be called "high-end", whatever its sound quality may be. One thing that is clear is that electronic components and drivers have to be hosted by extremely expensive cases to get the "high-end" label.

[Musicophile]

 

By the way, I am surprised to read that you describe your system as "upper mid-fi". Your signature line suggests that it clearly belongs to the hi-fi category in my understanding of the word. Then, there is the "high-end" category, but I never really understood if a "high-end" is one that sounds extremely well whatever its price or if it has to be very expensive to be called "high-end", whatever its sound quality may be. One thing that is clear is that electronic components and drivers have to be hosted by extremely expensive cases to get the "high-end" label.

Sorry for the OT: yep i had that discussion before, and actually know my rig sounds pretty good even by high-end standards. All I wanted to say that there are certainly systems out there that are way more resolving than mine.

[RealAudio]

 

You also know perfect reproduction by Shannon/Nyquist requires infinite bandwidth limited signals over an infinite period of time. I suspect you know that ignoring the time domain in audio reproduction, as so many are wont to do, also has audible consequences.

 

 

And that would only matter if human hearing had infinite bandwidth.

[DigiPete]

I suggest that

 

Recording is done

Minimum in 24 / 192

Optimum in DXD 32bit float

 

Distrimution is done

Minimum in 20 / 60

Optimum in all resolutions, chosen at download and/or later but paid as a whole package

(Like when buying a 2L bluray album with a SACD stereo/surround)

 

 

And then let this squabble lie.

[mayhem13]

Same here.

 

To my ears, with an admittedly upper-mid-fi system, the differences between even a well done 256kbit AAC and a 24/192 files are small at best.

 

However, if I do spend money on an album I really care about, those small differences do matter, otherwise we may as well all just buy Spotify and are done with it.

 

......and I'm patiently waiting for a stateside streaming service at 16/44 with a very extensive catalog which would probobly make my local streaming and personal catalog obsolete. My hope is that the Apple/Beats merger is heading in that direction but i remain skeptical as such a streaming service would only damage Apples iTunes market share.......now. But iTunes sales are constantly dwindling so I'm shure they're considering how and when to make the transition.

 

For me, 16/44 reliable streaming is the end game. I firmly believe it's the rest of the gear and the room that's most important and with today's advances in DSP, even the room acoustics are becoming less and less a factor. Soon the day to great sound without the computer, external DACs, USB cables and adapters..........and all the other nonsense fueling the audiophool machine.

[jhwalker]

......and I'm patiently waiting for a stateside streaming service at 16/44 with a very extensive catalog which would probobly make my local streaming and personal catalog obsolete. My hope is that the Apple/Beats merger is heading in that direction but i remain skeptical as such a streaming service would only damage Apples iTunes market share.......now. But iTunes sales are constantly dwindling so I'm shure they're considering how and when to make the transition.

 

For me, 16/44 reliable streaming is the end game. I firmly believe it's the rest of the gear and the room that's most important and with today's advances in DSP, even the room acoustics are becoming less and less a factor. Soon the day to great sound without the computer, external DACs, USB cables and adapters..........and all the other nonsense fueling the audiophool machine.

 

I'm in agreement with this. I've already reached that "almost Nirvana" with lossless 16/44 streaming and downloads, and am only buying selected new releases in high-res. I am unconvinced I can actually hear the difference between Redbook and higher-res offerings, but for new / old favorites, I'm willing to pay a bit more to get the "best" possible release / re-release just for my own satisfaction

[Leonardo Piras]

I'm in agreement with this. I've already reached that "almost Nirvana" with lossless 16/44 streaming and downloads, and am only buying selected new releases in high-res. I am unconvinced I can actually hear the difference between Redbook and higher-res offerings, but for new / old favorites, I'm willing to pay a bit more to get the "best" possible release / re-release just for my own satisfaction

 

Qobuz is coming to the States.

It may be your Nirvana.

I have been a long time subscriber myself.

Then I've found that local file playing with dedicated software allowed me for both room correction and better listening.

Qobuz is a great compromise...although their software is far from perfect.

[Leonardo Piras]

I suggest that

 

Recording is done

Minimum in 24 / 192

Optimum in DXD 32bit float

 

I theoretically may agree.

However, every time I've recorded the engineers were never using more than 24/96.

When you've got to layer track after track after track and bounce after bounce, disk space suddenly becomes a limit, even at today's prices.

 

And what's more, every time we experimented with higher sampling rates or bitrates, we could never hear any appreciable difference over 24/96. We agreed it was "good enough".

 

When recording, the difference btw 16 and 24 bits is dramatic. It allows for higher dynamics and better noise floor.

The difference btw 24 and 32 however is waaaay more subtle, and even in the most "large budget" projects we never used 32 bits.

 

Never worked on those blockbuster recordings, so these are just my 2 cents.

Never worked on DSD recordings. Would love to

[kumakuma]

When recording, the difference btw 16 and 24 bits is dramatic. It allows for higher dynamics and better noise floor.

 

Now I'm confused. Didn't you start this thread saying the exact opposite?

[Leonardo Piras]

Now I'm confused. Didn't you start this thread saying the exact opposite?

 

There's a whole world of difference between the recording / mixing stage of a recording and the final playback.

Capturing an uncompressed instrument sound at 24 bit allows for a lot more headroom and better signal to noise ratio.

 

Once a whole record, of which each song is usually made of 100+ individual tracks is mixed and mastered, and dithering has been properly applied, everything changes.

 

So we need to separate recording from playback. Again, two completely different worlds.

 

Starting this thread, I just reported that there's an interesting crowdsourced test by archimago that demonstrates you can't hear differences between 24 bit and 16 bit versions of the same audio file

[Le Concombre Masqué]

There's a whole world of difference between the recording / mixing stage of a recording and the final playback.

Capturing an uncompressed instrument sound at 24 bit allows for a lot more headroom and better signal to noise ratio.

 

Once a whole record, of which each song is usually made of 100+ individual tracks is mixed and mastered, and dithering has been properly applied, everything changes.

 

So we need to separate recording from playback. Again, two completely different worlds.

 

Starting this thread, I just reported that there's an interesting crowdsourced test by archimago that demonstrates you can't hear differences between 24 bit and 16 bit versions of the same audio file

 

Wrong. if it demonstrates anything it is that a set of files chosen by someone who seems to dedicate himself to prove that nothing can beat what he can afford is chosen by half of the respondents as fitting the criteria that "there should be improved transparency, ambiance, definition". As I reported earlier...

[gmgraves]

I probably should keep out of these discussions, as they never get anywhere.

 

However, having no objective measure is also true for every single pain killer out there that has ever been tested clinically. So probably the FDA would disagree with your statement.

 

Have a look here:

 

 

1. 
   
2. Efficacy endpoints in an analgesic trial should reflect a direct rating of pain intensity by the subject for all settings in which subjects can communicate in a reliable manner. We recommend the use of a well-defined and reliable PRO measure of the subject’s pain intensity.
   

 

 

 

 

http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM384691.pdf

 

(A "PRO" is a patient reported outcome).

 

 

While in that case, there might be nothing to measure (not sure about that anymore, what with the availability of MRIs and other devices for imaging brain activity), there certainly is the fact that the participants know if their pain lessens or goes away after taking the pain killer. I realize that's a subjective opinion, but then again, drug tests of that type encompass a much larger sample size than does an audio DBT where ten listeners is bordering on too many for a comfortable test.