Head-Fi.org › Forums › Equipment Forums › Sound Science › 24bit vs 16bit, the myth exploded!
New Posts  All Forums:Forum Nav:

24bit vs 16bit, the myth exploded! - Page 109

post #1621 of 1923

for frequency needs (for humans) 44khz is all we will ever need. people saying otherwise are mathematically wrong.
now 48khz allows for better work on the low pass filter, without being uselessly big in data. so it does sound like the very reasonable best choice. not that I can really complain about 44.1.


24bit has no defaults except size. worst case scenario,it doesn't do anything. so when possible it can actually be a cool thing. but playing 16bit as 24bit in foobar already brings most of the benefits of 24bit.
between playing a 16bit in 24bit or having 24bit files:
- digital volume control get the same margin
-quantization noise from the dac will also be the same

the only real difference is that the original file will have higher noise if in 16bit from the ADC right?  but we're talking about stuff done in pro studios, so it's reasonable to say that noise is really well under -96db and means nothing to us.


everything I've seen pushes me to be a fan of 16/48 but I have no problem with 24/48(except size).

wouldn't 48khz be better than 44.1 when the dac does the upsample/oversample into 96khz or 192khz as it would be a simple division by 2 or 4 operation?

post #1622 of 1923
Imagine uploading a raw image file for use of your profile picture on Facebook, that is how I feel about these high res nonsenses. What can we do about it. Life goes on.
post #1623 of 1923
Quote:
Originally Posted by esldude View Post
 


I think 24 bit /48 khz would be a more appropriate standard than 16/44 actually

++

post #1624 of 1923
Quote:
Originally Posted by bigshot View Post
 

Whenever you establish a threshold where you say "anything beyond this doesn't matter", there's always some joker in the crowd who insists that the line needs to be just a few feet further along... "just to be safe... just in case..." This happens over and over again with people worrying about "that last five percent" until the line gets pushed way over into La La Land.

 

Redbook standards were based on human hearing thresholds. If it performs to redbook spec, it's good enough for human ears. Period. Anyone who thinks they need more are either super humans or not in touch with reality. Take a guess which group I think they belong to!


People living in super human reality obviously!

post #1625 of 1923
Quote:
Originally Posted by castleofargh View Post
 

wouldn't 48khz be better than 44.1 when the dac does the upsample/oversample into 96khz or 192khz as it would be a simple division by 2 or 4 operation?

 

Not exactly that sample. Still need to apply anti-aliasing filter prior to downsampling.

post #1626 of 1923
Quote:
Originally Posted by bigshot View Post
 

In mixing, higher bit rates tax the processor more. This means that if you add too many RT filters, frames will start to drop in monitoring. If you have a very complex mix, like I often do in TV work, with multiple music, dialogue and effects tracks all with different RT filtering applied, super high bit rates can grind the processor to a halt. The more data you can push through a processor, the more data wants to be pushed through. It pays to be efficient and not waste resources on things that just don't make a difference.

 

I see it as one argument for native 32bit or 64bit word size, actually. 24bits processed on a 32bit machine is still going to eat up same cycle. There is really no advantage to trying to take advantage of a "fractional" cycle. The 8 bits just get thrown away.

post #1627 of 1923
Quote:
Originally Posted by Digitalchkn View Post
 

 

I see it as one argument for native 32bit or 64bit word size, actually. 24bits processed on a 32bit machine is still going to eat up same cycle. There is really no advantage to trying to take advantage of a "fractional" cycle. The 8 bits just get thrown away.


Well, a good many software DAWs end up doing 24 bit files in a 32 bit float format.  Allows processing that for the most part doesn't intrude on your 24 bits from rounding or truncation issues. Some software even uses 64 bits.

post #1628 of 1923
Quote:
Originally Posted by Digitalchkn View Post
 

 

Not exactly that sample. Still need to apply anti-aliasing filter prior to downsampling.

 

Which for playback is handled by DACs that upsample to apply the anti aliasing filter, then back to redbook. You don't need to use up sampling native to get around that problem.

 

None of this discussion matters, because it just plain ain't audible.


Edited by bigshot - 6/9/14 at 9:30pm
post #1629 of 1923
Quote:
Originally Posted by bigshot View Post
 

 

Which for playback is handled by DACs that upsample to apply the anti aliasing filter, then back to redbook.

 

Going from 48K to redbook still needs anti-aliasing applied somewhere, upsampled first or direct downsampling. Upsampling should apply an interpolation filter, technically. So it's not a simple e.g. x2 or /2 operation.

post #1630 of 1923

antialiasing is transparent. The problem is the brick wall.

post #1631 of 1923

It depends on the filter you apply. You sort of want a sharp filter (brickwall) with narrowest possible stopband bandwidth and highest stopband rejection. All this while having the minimum passband ripple and closest to linear phase response.  The more processing horsepower you have available the higher the filter order can start getting you closer to ideal. But no filter is perfect of course. That's one reason for having headroom in your source's sampling rate when converting and converting to 48K rather than redbook.  It's just you can get slightly sloppier converting to 48K since your <20KHz would be less affected by a less ideal filter.

post #1632 of 1923
Quote:
Originally Posted by Digitalchkn View Post
 
Quote:
Originally Posted by castleofargh View Post
 

wouldn't 48khz be better than 44.1 when the dac does the upsample/oversample into 96khz or 192khz as it would be a simple division by 2 or 4 operation?

 

Not exactly that sample. Still need to apply anti-aliasing filter prior to downsampling.


again I realize I write stuff as they come in my head without a care in the world for making sens. it looks like I wrote "take (sampleA+sampleB)/2 to generate the middle sample" as if it was a straight line..

but you answered me anyway (you guys are really strong at getting into my brain even when I'm writing nonsense).

yeah I was thinking about the all process, if it comes from a master in 24/96, going 48khz for our albums would be best so that we would just have to make an interpolation to get the exact downsampled values back. but yeah I didn't think aliasing, so it can't be that simple and precise. and thus I don't know if 44.1 or 48 change anything for that particular matter.

I'm still a 48khz fan for everything else, but we would need cds directly done in 48khz, no point turning a perfectly ok redbook to 48 on our own.

post #1633 of 1923

So people think its okay to pay more for the sake of "just in case". To me that makes no sense at all. True storage is quite cheap these day's but i think people underestimate the need for space with 24bit files. Storage is not the only problem, battery life is also a problem. Everybody knows that files in FLAC eat much more battery life than mp3 files.

Until they invent new form of storing energy, new battery. I don't think its viable to even consider 24bit at least not for portable use. Currently the only option would be to increase battery size.

That would affect the price of portable devices.

 

On a side note, it has always amused me how people have to constantly load smart phones because the battery wont last long.

Personally i consider that unacceptable. My current phone i load only once a month and it lasts even when i use it every day.

post #1634 of 1923

in theory 48khz allows for a 24khz bandwidth of which they say humans only have about 20khz range of hearing.  so the extra 4k would be for good measure but the reality is all of this makes a number of assumptions that are not true in a real system.

 

recognizing nyquist theory is one thing but if you think for a second about how that sampled waveform will look you see that 2 samples per cycle is really not enough even at a minimum.  take a 1hz sine wave and sample it 2 times per cycle.  it is entirely possible that both of your samples fall at EXACTLY the two zero crossings that occur in every cycle once on the way up and again on the way down and your reconstructed output becomes a flat line. by this it becomes pretty clear to see that 2 samples per cycle really is not enough at all to reconstruct your sine wave.  nyquist theory assumes a number of ideal conditions, none of which hold in a real sampling system.  it merely represents an absolute bottom line minimum spec requirement if you want to represent ANY of the information at a certain frequency but not necessarily all of it in all circumstances.  infact given 2 samples per wave taken at any point in the cycle i can draw for you an infinite number of possible near sine waves of different amplitudes that pass through those points.  lets assume the amplitude of the signal matters.  by raising the sample rate beyond the nyquist rate you quickly begin to remove the singularities and the multitude of waveforms that fit the samples for a given frequency and you allow breathing room for the realities of the limitations of a real sampling system, such as non-ideal pre-ADC bandpass and post-DAC reconstruction filtering.

 

if we assume 20khz is a reasonable limitation of human hearing without regard for the fact that our amp, speakers, etc further attenuate the response at these edge of spec frequencies

then 48khz is really still an absolute minimum. in my opinion 96khz gives plenty of breathing room to counter all of the additive realities of a true system from end to end. 


Edited by groovyd - 6/10/14 at 7:18am
post #1635 of 1923
Quote:
Originally Posted by groovyd View Post
 

in theory 48khz allows for a 24khz bandwidth of which they say humans only have about 20khz range of hearing.  so the extra 4k would be for good measure but the reality is all of this makes a number of assumptions that are not true in a real system.

 

recognizing nyquist theory is one thing but if you think for a second about how that sampled waveform will look you see that 2 samples per cycle is really not enough even at a minimum.  take a 1hz sine wave and sample it 2 times per cycle.  it is entirely possible that both of your samples fall at EXACTLY the two zero crossings that occur in every cycle once on the way up and again on the way down and your reconstructed output becomes a flat line. by this it becomes pretty clear to see that 2 samples per cycle really is not enough at all to reconstruct your sine wave.  nyquist theory assumes a number of ideal conditions, none of which hold in a real sampling system.  it merely represents an absolute bottom line minimum spec requirement if you want to represent ANY of the information at a certain frequency but not necessarily all of it in all circumstances.  infact given 2 samples per wave taken at any point in the cycle i can draw for you an infinite number of possible near sine waves of different amplitudes that pass through those points.  lets assume the amplitude of the signal matters.  by raising the sample rate beyond the nyquist rate you quickly begin to remove the singularities and the multitude of waveforms that fit the samples for a given frequency and you allow breathing room for the realities of the limitations of a real sampling system, such as non-ideal pre-ADC bandpass and post-DAC reconstruction filtering.

 

if we assume 20khz is a reasonable limitation of human hearing without regard for the fact that our amp, speakers, etc further attenuate the response at these edge of spec frequencies

then 48khz is really still an absolute minimum. in my opinion 96khz gives plenty of breathing room to counter all of the additive realities of a true system from end to end.

 

As you point out Nyquist sampling is valid from DC to *less* than 1/2 sampling frequency. Perfect reconstruction of the original waveform requires a sinc interpolation filter, a mathematically perfect filter that can't be implemented in real life using physical components.  That's one reason 44.1K was chosen for redbook (and not something close to 40K sampling rate) as a tradeoff.

 

There is plenty of signal energy beyond 20KHz in majority of recordings. Regardless of whether majority of humans can directly hear ultrasonics or not,  brickwall filtering has immediate impact on time domain waveform response. With modern processing and storage capability limiting to 44.1K makes less sense than it once did.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › 24bit vs 16bit, the myth exploded!