Jump to content
IGNORED

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


Recommended Posts

Double blind studies are the gold standard for all sorts of psychological and perception testing. There are a number of double blind studies that demonstrate people can't tell the difference between high rez and standard cd resolution. Please provide a citation to a study that shows different results.

 

P18-6 Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz—Amandine Pras, Catherine Guastavino, McGill University - Montreal, Quebec, Canada

 

AES E-Library » Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz

Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz

 

It is currently common practice for sound engineers to record digital music using high-resolution formats, and then down sample the files to 44.1kHz for commercial release. This study aims at investigating whether listeners can perceive differences between musical files recorded at 44.1kHz and 88.2kHz with the same analog chain and type of AD-converter. Sixteen expert listeners were asked to compare 3 versions (44.1kHz, 88.2kHz and the 88.2kHz version down-sampled to 44.1kHz) of 5 musical excerpts in a blind ABX task. Overall, participants were able to discriminate between files recorded at 88.2kHz and their 44.1kHz down-sampled version. Furthermore, for the orchestral excerpt, they were able to discriminate between files recorded at 88.2kHz and files recorded at 44.1kHz.

 

Authors: Pras, Amandine; Guastavino, Catherine

Affiliation: McGill University, Montreal, Quebec, Canada

AES Convention:128 (May 2010) Paper Number:8101 Import into BibTeX

Publication Date:May 1, 2010

Subject:Audio Coding and Compression

Main listening (small home office):

Main setup: Surge protector +>Isol-8 Mini sub Axis Power Strip/Isolation>QuietPC Low Noise Server>Roon (Audiolense DRC)>Stack Audio Link II>Kii Control>Kii Three (on their own electric circuit) >GIK Room Treatments.

Secondary Path: Server with Audiolense RC>RPi4 or analog>Cayin iDAC6 MKII (tube mode) (XLR)>Kii Three .

Bedroom: SBTouch to Cambridge Soundworks Desktop Setup.
Living Room/Kitchen: Ropieee (RPi3b+ with touchscreen) + Schiit Modi3E to a pair of Morel Hogtalare. 

All absolute statements about audio are false :)

Link to comment
Did you use the words ad hominen to make it sound like you actually have electronics training ?

Using such Latin phrases here doesn't impress as many as it might in Hydrogen Audio.

 

Are you high?

 

Ad hominem - Wikipedia, the free encyclopedia

 

An ad hominem (Latin for "to the man" or "to the person"[1]), short for argumentum ad hominem, is a general category of fallacies in which a claim or argument is rejected on the basis of some irrelevant fact about the author of or the person presenting the claim or argument.[2] Fallacious Ad hominem reasoning is normally categorized as an informal fallacy,[3][4][5] more precisely as a genetic fallacy,[6] a subcategory of fallacies of irrelevance.[7] Ad hominem reasoning is not always fallacious, for example, when it relates to the credibility of statements of fact.

 

Ad hominem arguments are the converse of appeals to authority, and may be used in response to such appeals.

 

...

 

Abusive ad hominem usually involves attacking the traits of an opponent as a means to invalidate their arguments. Equating someone's character with the soundness of their argument is a logical fallacy. Mere verbal abuse in the absence of an argument, however, is not ad hominem nor any kind of logical fallacy.[8]

 

Link to comment
I may be wrong about this but doesn't the number of bits (16 versus 24) only apply to the dynamic range of the music. If so, it would seem that the extra bits would only be needed for highly dynamic music.

 

It’s not just dynamic range but the important small intricate details in music within the dynamic range called micro-dynamics.

 

16 bit has 65,536 possible levels while 24 bit has 16,777,216 possible levels. 24-bit offers considerably more intermediate steps within the dynamic envelope. Giving 24-bit the potential of 256 times the resolution of 16-bit.

 

Thus even with a small dynamic range a 24 bit recording has the potential of considerable more resolution because of the greatly increased number of levels within that dynamic range.

I have dementia. I save all my posts in a text file I call Forums.  I do a search in that file to find out what I said or did in the past.

 

I still love music.

 

Teresa

Link to comment
Short answer: no, you can't.

 

Perhaps you can. If you like classical music download a BIS 24/44.1kHz music file from eclassical.com it also comes with the 16/44.1kHz version so you can compare them before believing the results of this test.

 

If you already have a 24/44.1kHz music file on your computer you can change the bit depth to 16 using XLD. Only ALAC (Apple Lossless) offers this option in XLD.

I have dementia. I save all my posts in a text file I call Forums.  I do a search in that file to find out what I said or did in the past.

 

I still love music.

 

Teresa

Link to comment

Normal people don't use phrases like Ad hominem in everyday speech or on forums, and your average person on the street wouldn't know WTF you were talking about if you did. Even most doctors write prescriptions in plain language these days, not a language from centuries ago.

It just highlights the gap between stuffy Academics , some of whom suffer from a huge superiority complex , and normal people.

 

How a Digital Audio file sounds, or a Digital Video file looks, is governed to a large extent by the Power Supply area. All that Identical Checksums gives is the possibility of REGENERATING the file to close to that of the original file.

PROFILE UPDATED 13-11-2020

Link to comment

If you already have a 24/44.1kHz music file on your computer you can change the bit depth to 16 using XLD. Only ALAC (Apple Lossless) offers this option in XLD.

 

I did try something myself, in no double blind fashion. The key here is to take some high res files and downsample/dither them to 16/44.

You must compare the same files at different resolutions.

 

Though I have what I consider a high end and revealing system, I can't really tell the difference between 24/44 and 16/44.

 

Basically because maybe (maybe!) 65536 levels of volume are well enough for any kind of music.

 

Rock, pop and modern jazz are all so compressed that the levels of volume are way, way below that threshold.

But I guess that 16bits would be enough for classical as well.

 

Basically, most of the sound quality of a recording derives from the way it has been recorded, mixed, mastered and produced.

 

Sampling rate, DSD and so on are a different story... and there are some other tests and papers about them.

 

I do think that labels and studios should keep delivering music at the highest possible resolution.

 

But the highest resolution, in my opinion, should match the highest resolution that recording was tracked at.

Link to comment
I think selling 24/44.1 files as high resolution is a bit of scam for many kinds of music.

 

I just compared the 16/44.1 version of Lana Del Rey's Born to Die album to the 24/44.1 version sold on HDTracks. They sound identical to my ears and Jriver says the dynamic range of the first six songs is the same (6, 5, 5, 6, 4, 4). This suggests zero benefit from the extra 8 bits of headroom, for this album at least.

 

BTW, I have a vinyl rip of this album (24/96) that kicks both 44.1 versions to the curb.

 

It's Dynamic Range (R128) you should be looking at to determine the required bit-depth, not Dynamic Range (DR) which is not actually a measure of dynamic range at all - that's a measure of peak-to-loudness ratio.

 

1 LU = 1 dB. 6dB = 1 bit.

 

This is a weighted measurement though, between the 10th and 95th percentile, so you do require more dynamic range than it suggests, but based on the R128 values, nothing in my library should even come close to requiring 16-bit.

 

The highest R128 value I have is 27.5 LU (<5 bit)

Less than half a percent of my library has an R128 value greater than 12 LU.

 

 

And Dynamic Range (DR) values are not valid for vinyl rips, so higher numbers do not necessarily mean that the vinyl rip is a better source.

 

This must explain why most of my old Telarc cd files sound better than the DSD I just downloaded.
Mastering is the most important thing for sound quality in my opinion.

A well-mastered CD sounds a lot better in my system than most of the remastered high res releases with a compressed dynamic range.

 

Some people seem to care more about the format than anything else - there's a member here that flat-out refuses to listen to 16/44.

 

That just seems insane to me. While I'll take a higher quality version of the same master if it's available, the format it's available in, as long as it's lossless, is a much lower priority for me than the mastering quality.

 

...

You really need 3 samples for the first half of the wave, and 3 samples for the second half of the wave, and then when you curve fit the 6 points, you will have a reasonable recreation of the wave form

...

If you sample at 192 KHz you will be capturing frequency content below 96 KHz, but with extremely poor fidelity if you look at the wave form on a scope. You can see the lack on fidelity on a scope, and well trained ears can hear the difference as well. In the trumpet example there will be about 4 data points from which you need to recreate the full wave form, that's not enough to do a decent job

...

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.

 

Increasing the sample rate makes the waveform easier for us to understand when plotting the sample points on a graph, but it does not improve the accuracy or "smoothness" of the audio.

 

I recommend watching the whole thing, but about four minutes into this video actually demonstrates this, with an analog signal generator going into an ADC at 16/44, the digital waveform being displayed on the laptop, and then sending a 16/44 signal back out to an analog scope.

 

The graph for higher frequencies is indecipherable to us, and yet it comes out as a perfect signal on the other end.

 

 

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.

Link to comment
I think selling 24/44.1 files as high resolution is a bit of scam for many kinds of music.

 

I just compared the 16/44.1 version of Lana Del Rey's Born to Die album to the 24/44.1 version sold on HDTracks. They sound identical to my ears and Jriver says the dynamic range of the first six songs is the same (6, 5, 5, 6, 4, 4). This suggests zero benefit from the extra 8 bits of headroom, for this album at least.

 

BTW, I have a vinyl rip of this album (24/96) that kicks both 44.1 versions to the curb.

 

As of Ultraviolence : 16 bits is ultra violence compared to the 24 bits version

I admit i have very expensive gear

Link to comment
P18-6 Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz—Amandine Pras, Catherine Guastavino, McGill University - Montreal, Quebec, Canada

 

AES E-Library » Sampling Rate Discrimination: 44.1 kHz vs. 88.2 kHz

 

I did read that paper and it seems like the authors might have messed up their down-sampling of the files, based on some of their odd results.

 

"Furthermore, our findings show that listeners were

more sensitive to differences between files recorded

at 88.2 kHz and their 44.1 kHz down-sampled

version, than to differences between files recorded at

different sample rates. As we down-sampled the files

through a single software program, further

investigation of down-sampling algorithms is

required to draw conclusions regarding the impact of

down sampling vs. recording at 44.1 kHz. "

 

With that said, these days you can't even guarantee that the hi-rez version of a recording even has the same mastering as the cd-rez version. A dirty trick indeed.

Link to comment

Has anyone ever tried this test as proposed by the venerable Steve Hoffman?

 

 

"The SACD/CD of Creedence WILLY & THE POOR BOYS that I did with Kevin a few years ago for Acoustic Sounds/Analogue Productions has a great example that you can try at home.

 

If one put on FORTUNATE SON in the SACD/DSD layer and just listen to the intro of the song, concentrating on the echo trail after each drum thwack you will hear how far "back" into the mix the echo of each distinct drum hit goes. Memorize that sound. If you don't trust the resolving power of your speakers, use your headphones...

 

Now, switch to the CD layer, cut by Kevin and I on the same day with the same mastering. Listen to FORTUNATE SON again, concentrating on the each drum thwack in the intro. Notice how the reverb vanishes much faster and is not as intense? Loss of resolution."

 

It is pretty easy to hear the difference. It is also easy to hear the difference with the new Hdtracks version.

Link to comment
That's garbage. I could easily hear the difference between 16/44.1 and 24/44.1, and 24/48 which to my ears, just seemed to come all together, for want of better words. BTW, 24/44. is also what HD Tracks is now supplying for some recent recordings.

 

Archimago makes too many assumptions based on incomplete data about human hearing.

Recent research shows that the rise time of the waveform does matter, not just whether oldies can hear even a 12KHz sine wave.

 

 

The answer is simple. Dithering the last bit in a 16 bit files affects the sound and so why

shouldn't a 24 bit file improve the sound? There are simply far too many different setups and

configurations in replies to make them consistent and the conclusion valid. Basically, the OS,

the Player software, buffer settings , streaming mode, and the computer all affect SQ.

 

AS an analogy, I was a surprised to discover that Snag It processed jpg pictures

appear a lot more realistic when the colour depth is set to 32 bit from the default 24 bit!

 

I cannot fathom why a guy would have spent so much of his time using iffy software

(Rightmark) and refusing to spend a small amount on proper, professional analysis software.

This has resulted in ALL of his measurements to show the same trends when it is the analysis

technique which has been masking the differences.

fmak

Link to comment

But they also found this, which isn't dependent on downsampling:

 

Furthermore, for the orchestral excerpt, they were able to discriminate between files recorded at 88.2kHz and files recorded at 44.1kHz.

 

 

 

I did read that paper and it seems like the authors might have messed up their down-sampling of the files, based on some of their odd results.

 

"Furthermore, our findings show that listeners were

more sensitive to differences between files recorded

at 88.2 kHz and their 44.1 kHz down-sampled

version, than to differences between files recorded at

different sample rates. As we down-sampled the files

through a single software program, further

investigation of down-sampling algorithms is

required to draw conclusions regarding the impact of

down sampling vs. recording at 44.1 kHz. "

 

With that said, these days you can't even guarantee that the hi-rez version of a recording even has the same mastering as the cd-rez version. A dirty trick indeed.

Main listening (small home office):

Main setup: Surge protector +>Isol-8 Mini sub Axis Power Strip/Isolation>QuietPC Low Noise Server>Roon (Audiolense DRC)>Stack Audio Link II>Kii Control>Kii Three (on their own electric circuit) >GIK Room Treatments.

Secondary Path: Server with Audiolense RC>RPi4 or analog>Cayin iDAC6 MKII (tube mode) (XLR)>Kii Three .

Bedroom: SBTouch to Cambridge Soundworks Desktop Setup.
Living Room/Kitchen: Ropieee (RPi3b+ with touchscreen) + Schiit Modi3E to a pair of Morel Hogtalare. 

All absolute statements about audio are false :)

Link to comment
... You really need 3 samples for the first half of the wave, and 3 samples for the second half of the wave, and then when you curve fit the 6 points, you will have a reasonable recreation of the wave form

 

If you sample at 192 KHz you will be capturing frequency content below 96 KHz, but with extremely poor fidelity if you look at the wave form on a scope. You can see the lack on fidelity on a scope, and well trained ears can hear the difference as well. In the trumpet example there will be about 4 data points from which you need to recreate the full wave form, that's not enough to do a decent job

...

 

OMG, Nyquist's theorem was wrong...

"People hear what they see." - Doris Day

The forum would be a much better place if everyone were less convinced of how right they were.

Link to comment
I can honestly say I don't expect my 63 year old ears to be able to hear the difference between a 24/192 file properly down converted to 16/192.

 

However that does not mean the content industry should not be mastering at 24 bits and even better DSD. Any kind of mixing needs more bit depth to minimize the sound degradation. Some of latest algorithms even go as far as 64 bits ! So with data storage and transmission so cheap why throw away bits 17 to 24 ?

 

what is far more important for PCM recording is the sample rate. I am becoming more and more convinced that to do a good job of recording in PCM format the sample rate needs to be 352 / 384. 192 does not get the job done satisfactorily.

 

i think there is a trap that people fall into that because we can not hear sine waves at much more than 12 KHz in my case, that high frequency content, and the recording and transmission of that content is unimportant.

 

A key part of creating the pleasureable listening experience, is recreating the wave form faithfully. There are some instruments such as trumpets and cymbals that have significant energy at frequencies over 50kHz. To capture a trumpet wave form faithfully you then need a sample rate of more than 6 times that ie 300 KHz.

 

You really need 3 samples for the first half of the wave, and 3 samples for the second half of the wave, and then when you curve fit the 6 points, you will have a reasonable recreation of the wave form

 

If you sample at 192 KHz you will be capturing frequency content below 96 KHz, but with extremely poor fidelity if you look at the wave form on a scope. You can see the lack on fidelity on a scope, and well trained ears can hear the difference as well. In the trumpet example there will be about 4 data points from which you need to recreate the full wave form, that's not enough to do a decent job

 

it also explains to me why DSD 64 recordings do such a good job of capturing the waveform, the sampling frequency is above 2 MHz, never mind the coding scheme is stupidly primitive ;-)

 

sorry to be off thread, but being doing a lot of reading recently and getting flamed over at Pink Fish media

 

.......and you would have some validity to your point IF there were microphones capable of accurately capturing those frequencies at flat levels. Apparently microphone mfgrs don't share your viewpoint though, so I wouldn't hold my breath for your proposed sample rates.

 

As for speakers and faithfully reproducing the waveform, there are a very small amount of HF devices that can do to 30khz........NONE of them accurately. Another industry that needs to update their engineering to meet your proposed standards.

 

......or in other words, the entire industry on both ends from recording to reproduction do not agree with your proposed wave form theory.

Link to comment

I think I can reliably hear the difference between 24/96 and 16/96 when I try to downsample on the fly with JRiver. But the methodology of the study seems quite sound. So assuming the results are genuinely true, the main difference between my home test (and I presume most people's) and the study is that when we listen to 16/96, we don't put the source back into a 24/96 container as the study had to do to ensure proper blinding. So maybe the difference I'm hearing has nothing to do with whether my ears can discern 24/96 vs 16/96 data. Maybe what I'm hearing is the difference between how my DAC handles a 16-bit file and a 24-bit file? I guess one can argue that the extra on-the-fly processing might affect the sound too but somehow I would think that would be an even less important contributor. I'm now more unsure than ever.

 

That said, if the problem is with how my DACs handles 16-bit file vs 24-bit file, obviously, I would still buy the highest quality files because I'm not going to buy 16/44 files that are downconverted from 24/44 and then put the file in a container to 24/44 just to get better sound.

Link to comment

You do NOT need a sampling frequency that is 6 times higher than the highest frequency to reproduce. Mathematically, a good 2 times is enough and practically 2,5 times is enough to avoid most of the undesired effects of real-world low pass filters.

 

In other words, 48KHz is ok for audible frequencies and 192KHz is more than enough for up to 50KHz signals. 44.1KHz is a bit tight, often leading to compromises in the filter design.

 

Regarding 16 vs 24 bits per sample, really good DACs can use approximately 20 bits.

 

In my opinion, 24 bits / 48 KHz is great for reproducing the audible frequencies as faithfully as possible, while 24 bits / 192 KHz gives ample headroom to include ultrasonic harmonics and other higher frequency content. Personally, I'm fine with faithful reproduction up to 16 KHz - don't hear much beyond anyway ;-)

Home: Apple Macbook Pro 17" --Mini-Toslink--> Cambridge Audio DacMagic --XLR--> 2x Genelec 8020B

Work: Apple Macbook Pro 15" --USB--> Focusrite Scarlett 2i2 --1/4\"--> Superlux HD668B / 2x Genelec 6010A

Link to comment
I have done the criticizing of the testing methods quite a few times previously, and I am not about to keep repeating them ad- nauseam for every vocal newbie that comes along, especially members of Gasbag Audio..

 

My stuffy academic brain is unable to comprehend your normal person Latin.

Link to comment

 

I did read that paper and it seems like the authors might have messed up their down-sampling of the files, based on some of their odd results.

 

 

I posted to this effect when he first posted the 'survey'.

 

He was dismissive and so I decided to ignore the exercise.

 

So should anyone who cares about sound quality.

fmak

Link to comment

 

You do NOT need a sampling frequency that is 6 times higher than the highest frequency to reproduce. Mathematically, a good 2 times is enough and practically 2,5 times is enough to avoid most of the undesired effects of real-world low pass filters.

 

 

Who cares about Nyguist, who is often used to justify an invalid argument on audio perception.

 

The brick wall filter and any analog filter that follows 2X destroy phase coherence and simply

produces the digitis SQ of PCM replay.

 

6X is much better; 8X better still

fmak

Link to comment

This is a good discussion, and I think a common theme here is that the audible difference is generally quite subtle. If you find you're listening to a very specific audible clue (e.g., the way a drum thwack dissipates into the sound field), it's not really a "night and day" difference ;)

 

Like others have said, I do not intend to change my buying habits - I'll continue to buy recordings I like, at the resolution / format that suits my fancy (I'm currently taken with DSD) - but I like the idea that people are doing real research into what is / is not important / audible. There is room for both objective and subjective approaches, and neither should be seen as an "attack" on the other.

 

IMO.

John Walker - IT Executive

Headphone - SonicTransporter i9 running Roon Server > Netgear Orbi > Blue Jeans Cable Ethernet > mRendu Roon endpoint > Topping D90 > Topping A90d > Dan Clark Expanse / HiFiMan H6SE v2 / HiFiman Arya Stealth

Home Theater / Music -SonicTransporter i9 running Roon Server > Netgear Orbi > Blue Jeans Cable HDMI > Denon X3700h > Anthem Amp for front channels > Revel F208-based 5.2.4 Atmos speaker system

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...