Head-Fi.org › Forums › Equipment Forums › Sound Science › Why would 24 bit / 192 khz flac sound any better than 16 bit / 44.1 khz flac if both are lossless (if at all)?
New Posts  All Forums:Forum Nav:

Why would 24 bit / 192 khz flac sound any better than 16 bit / 44.1 khz flac if both are lossless... - Page 14

post #196 of 251
Kudos to you esisude. Can I get a "Alleluia!"
post #197 of 251
Quote:
Originally Posted by castleofargh View Post
 

yup, that's why I said "and as long as the sample is big enough to get the right wave, should it matter?" having a few less "dots" on the wave wouldn't change the wave unless there was really too few reference values. but I don't really know how many is enough as music is not juste 1 sine wave.

...

 

So long as there are at least two "dots" for the highest frequency contained in the music, you have enough.

 

(To be pedantic, that should be "significant frequency". Some instruments produce frequencies (harmonics) that can be higher than you can hear. Some people argue that you can hear their presence or absence and the sampling rate should be chosen to capture them all. Others argue that since you can't hear them, there is no need to capture them.) 

post #198 of 251

thank you.

so in the end back to the "if it can't be heard, some will hear it" situation. ^_^

post #199 of 251
Quote:
Originally Posted by castleofargh View Post
 

thank you.

so in the end back to the "if it can't be heard, some will hear it" situation. ^_^

The ability to hear frequencies, sounds, distortions, etc. that the all humans have been conclusively proven not to be able to hear is directly proportional to the amount of money which can be made by claiming to hear such frequencies, sounds, distortions, etc.. So for example, if you one claims to be able the pico-second length jitter distortions present in non-asynchronous USB (which has been claimed) then as a solution to this very serious "jitter problem" one can sell asynchronous USB based DACs to replace all those horrible sounding and jitter filled non-asynchronous USB DACs.

 

Perhaps the best example of the FUD principle (FUD = fear, uncertainty and doubt) is the Beatles catalog - first released as remastered standard resolution CDs (16bit/44.1kHz), then released as 24bit/44.1kHz "high resolution" files with the as yet to be released "super high resolution" (24bit/96kHz) files waiting in the wings.

 

Seems to me the software side as the edge in the FUD war since they get the poor audiophile to buy everything over and over and over again.

post #200 of 251
Quote:
Originally Posted by Don Hills View Post
 
Quote:
Originally Posted by castleofargh View Post
 

yup, that's why I said "and as long as the sample is big enough to get the right wave, should it matter?" having a few less "dots" on the wave wouldn't change the wave unless there was really too few reference values. but I don't really know how many is enough as music is not just 1 sine wave.

...

 

So long as there are at least two "dots" for the highest frequency contained in the music, you have enough.

 

 

(To be pedantic, that should be "significant frequency". Some instruments produce frequencies (harmonics) that can be higher than you can hear. Some people argue that you can hear their presence or absence and the sampling rate should be chosen to capture them all. Others argue that since you can't hear them, there is no need to capture them.) 


 I've looked at http://people.xiph.org/~xiphmont/demo/neil-young.html not getting half of it, that lead me to this http://en.wikipedia.org/wiki/Sampling_theorem and then in-between all the stuff I really don't get, I've seen that sentence "Nyquist's result that equi-spaced data, with two or more points per cycle of highest frequency, allows reconstruction of band-limited functions." and I remembered you telling me exactly that, but at the time it didn't compute for some reason.

 

so if I'm not lost in space, the reason why we have 44khz (or 48khz) isn't random luck but "simply" to have at least 2 sample values of the highest audible frequency, for us 20khz. \o/ because yeah I'm a genius 2*20=40 . I'm self amazed by my math talent.

for lower frequencies we obviously get more reference points as the wave is bigger in time, so it's of no concern at all, at least precision wise.

 

and restricting the frequency range (cutting, filtering.. whatever) is just a condition to make that rule stay true?

sorry for the ranting, but such a simple thing is making me understand tonnes of stuff I've seen and accepted, but without knowing why.

 

so if I got it all right, increasing sample rate cannot add any precision unless we use a recording that has cues above 20khz, on a player that will not cut them out, for the benefit of friendly animals able to ear those sounds?

post #201 of 251
Yes. The amount of samples used to reproduce the audible spectrum in regular CD quality sound is the same as the number used for the same spectrum in high sampling rates. The additional samples in high sampling rates just extend frequencies beyond the range of human hearing.
post #202 of 251
Quote:
Originally Posted by castleofargh View Post
 

 

so if I got it all right, increasing sample rate cannot add any precision unless we use a recording that has cues above 20khz, on a player that will not cut them out, for the benefit of friendly animals able to ear those sounds?

And the benefit of the many different people who profit by selling a high resolution versions of a recordings.

post #203 of 251
Quote:
Originally Posted by castleofargh View Post
 

 

and restricting the frequency range (cutting, filtering.. whatever) is just a condition to make that rule stay true?

 


The rule stays true no matter what. You cut the top frequencies to avoid aliasing.

The maths here can seem more complicated than it really is, so put more simply:

 

Imagine you write a computer program that records outside temperatures through the year, but by chance you neglect to allow for the possibility of negative temperatures.
So when the temperature gradually creeps down below zero degrees, what does the program record? It might just get completely confused and return error messages, or it might faithfully record the data the only way it knows how. As far as the program knows negatives do not exist, so it just discards the minus signs and returns 3, 2, 1, 0, 1, 2,… mirroring the negative numbers back up as positive, making a right mess of your data.

Much the same happens when you try to sample a signal that is more than half the sampling frequency. Using a sampling frequency of 48kHz, a 24kHz signal remains as 24kHz, but 26 becomes 22, 28 becomes 20 and 48 becomes 0. Everything above 24kHz gets folded symmetrically back down, just like the temperature recorder folded data below zero back up.

In both cases you can mend things by either increasing the sample space, or if you want to keep things as they are, just decide that it's better to return nulls than non-sense, and filter out the impossible values before you feed them on.


Edited by limpidglitch - 4/1/14 at 2:57am
post #204 of 251
Quote:
Originally Posted by castleofargh View Post
 ... so if I'm not lost in space, the reason why we have 44khz (or 48khz) isn't random luck but "simply" to have at least 2 sample values of the highest audible frequency, for us 20khz. \o/ because yeah I'm a genius 2*20=40 . I'm self amazed by my math talent.

 ... and restricting the frequency range (cutting, filtering.. whatever) is just a condition to make that rule stay true?

 ...so if I got it all right, increasing sample rate cannot add any precision unless we use a recording that has cues above 20khz, on a player that will not cut them out, for the benefit of friendly animals able to ear those sounds?

 

As others have pointed out, yes, that's basically it.

There are those that argue that the small gap between 20 KHz and 22.05 KHz is too small to apply an effective enough filter to make sure aliasing doesn't occur, and that the filter can cause audible side effects. 48 KHz is better, and of course even higher sampling rates allow using filters less likely to cause audible effects. 

There are also those that argue that the frequencies above 20 KHz matter. I suggest doing the equivalent of sitting back with some popcorn and watch the fight.

post #205 of 251
Quote:
Originally Posted by Don Hills View Post
 

 

As others have pointed out, yes, that's basically it.

There are those that argue that the small gap between 20 KHz and 22.05 KHz is too small to apply an effective enough filter to make sure aliasing doesn't occur, and that the filter can cause audible side effects. 

 

It's often been repeated that a too steep cut-off will cause ringing, but I've never really understood why this is (though I suspect some parallels to ringing in band limited square waves). 

Anyone have a simple explanation, maybe including how to decide on an optimal cut-off rate?

 

 

Quote:

Originally Posted by Don Hills View Post
 

 

There are also those that argue that the frequencies above 20 KHz matter. I suggest doing the equivalent of sitting back with some popcorn and watch the fight.

 

 

There are also those that argue that super sonic frequencies will do more harm than good (oscillations, IM distortion and whatnot)
I think I'll join you with that popcorn.


Edited by limpidglitch - 4/1/14 at 5:28am
post #206 of 251
Quote:
Originally Posted by Don Hills View Post

As others have pointed out, yes, that's basically it.
There are those that argue that the small gap between 20 KHz and 22.05 KHz is too small to apply an effective enough filter to make sure aliasing doesn't occur, and that the filter can cause audible side effects. 48 KHz is better, and of course even higher sampling rates allow using filters less likely to cause audible effects. 
There are also those that argue that the frequencies above 20 KHz matter. I suggest doing the equivalent of sitting back with some popcorn and watch the fight.

Modern DACs upsample to apply the rolloff clleanly.

Controlled testing has shown that frequencies above the range of human hearing add nothing to music. In fact, above 14kHz, there isn't much of anything to hear.
post #207 of 251
Quote:
Originally Posted by bigshot View Post


In fact, above 14kHz, there isn't much of anything to hear.

That makes me feel so kHz deprived. (Please, say it ain't so!)
post #208 of 251
In response to the original set of questions, I can say, anecdotally, that I hear a very distinct difference between 16/44.1 and HDTracks titles that I convert to 24/48. Sometimes it's a small improvement, sometimes quite significant--but so far, with every title I've downloaded, there's been an improvement. I'm not going to speculate on the cause. But, considering I was, for a time, a professional classical musician and have a highly trained ear (thanks to well over a decade of private lessons, teaching, etc.) I think it's fair to rule out a placebo effect. I can't hear above 16.5 kHz (hell, why would I want to?), nor do I believe frequencies higher than that have an unconscious or subconscious effect (it'd be like saying pinching a paraplegic in the leg produces subconscious pain. It doesn't). For whatever reason, 24/48 always sounds better to me than 16/44.1. But, frankly, I think the difference--while often significant to me--would be non-existent to an untrained ear. This is not an ad for HDTracks: I think their customer service is consistently wretched--far, far worse than any other company I buy from--and I can't wait for competing companies to emerge.
post #209 of 251
Quote:
Originally Posted by Chesterfield View Post

In response to the original set of questions, I can say, anecdotally, that I hear a very distinct difference between 16/44.1 and HDTracks titles that I convert to 24/48. Sometimes it's a small improvement, sometimes quite significant--but so far, with every title I've downloaded, there's been an improvement. I'm not going to speculate on the cause. But, considering I was, for a time, a professional classical musician and have a highly trained ear (thanks to well over a decade of private lessons, teaching, etc.) I think it's fair to rule out a placebo effect. I can't hear above 16.5 kHz (hell, why would I want to?), nor do I believe frequencies higher than that have an unconscious or subconscious effect (it'd be like saying pinching a paraplegic in the leg produces subconscious pain. It doesn't). For whatever reason, 24/48 always sounds better to me than 16/44.1. But, frankly, I think the difference--while often significant to me--would be non-existent to an untrained ear. This is not an ad for HDTracks: I think their customer service is consistently wretched--far, far worse than any other company I buy from--and I can't wait for competing companies to emerge.


It is never fair to rule out placebo.  It is something that effects all humans.  Even with your background you would be susceptible. 

 

Unless they are the same master (and they won't be from HDtracks), then they can sound different to varying degrees for reasons having nothing to do with sample rate or bit depth.  Convert the HDtracks to 44.1 khz, and back to original rate (96, 192khz whatever).  Then put the two files in Foobar and see if you can successfully ABX them. 

post #210 of 251
I'm well aware of the power of suggestion, and of the placebo effect. I have a Ph.D. in a field affiliated with psychology. I wasn't trying to suggest that I'm immune to the placebo effect, but I do resist having all my training and expertise in music so casually dismissed, which is why I mentioned it.

Yes, it's entirely possible that of the 40+ titles I've bought from HDTracks, every one uses a better master than their CD counterparts.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Why would 24 bit / 192 khz flac sound any better than 16 bit / 44.1 khz flac if both are lossless (if at all)?