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 59

post #871 of 1923
Quote:
Originally Posted by stv014 View Post

 

It is different enough to significantly outweigh the small differences between 16 bit vs. 24 bit, and 44.1 kHz vs. 96 kHz.

 

 

Normalizing the peak amplitude does not guarantee equal loudness, especially with the above mentioned randomization that also adds randomness to the peak level. You would need to use something like ReplayGain for more accurate matching.

 

However, the best approach is to generate only one file with the software synthesizer, at the highest possible quality (i.e. 192 kHz/32-bit), leaving a couple dB of headroom so that the sample rate conversions will not result in clipping, and converting that file to all the other (lower quality) formats to be tested. Finally, to make sure that the DAC does not introduce any differences, convert the files back to 192 kHz/32-bit, and compare those with the foobar2000 ABX comparator. If the software you use for the conversions is well written, you should not need to apply any level matching or synchronization.

Ok, I went ahead and generated a new file at the optimized settings and converted it to 16/44.1 and back again. I also exported multiple files from the same Live project as I did before. I also ABX'ed. I have determined the following:

 

-I cannot tell the difference between any of the derivate files of the original 192/32 master file. You were correct.

-When using midi, none of the audio has been captured prior to exportation. The initial settings for exporting the master file greatly affect the audio quality. Exporting at 192khz vs 96khz and so on produces a marked difference in quality, which likewise applies to the bit rate. These different export settings can ABX'ed at 100% accuracy. This has been the source of my confusion. However, interconverting the files post-export does not seem to result in any noticeable difference.


Edited by ctoth666 - 9/24/12 at 11:46am
post #872 of 1923

"When 44.1 and 96kHz are compared it gets real subjective"

popcorn.gif

post #873 of 1923

I thought that what mattered was the sample rate, and that the bit depth was only to allow for higher sample rates.

post #874 of 1923
Quote:
Originally Posted by Benjamin6264 View Post

I thought that what mattered was the sample rate, and that the bit depth was only to allow for higher sample rates.

 

These are two separate things.  As for whether or not sample rate past a certain point matters, practical implementations, ultrasonics, etc., there is plenty of information (also a lot of misinformation) on that.  See the last page or two here, for instance.

 

Anyway, the bit depth is the number of bits (commonly 16, 24... also 32) used to represent each sample.  The higher the bit depth, the more values (amplitude levels) can be represented.  There are only a finite number of possible values:  2^16, 2^24, etc., depending on the bit depth.  Note that the difference between two consecutive steps at 24-bit depth or higher is much smaller than the electrical noise in any real-world DAC, and that most studio setups / recordings / etc. (also listening environments) have higher noise than two consecutive steps for 16-bit depth.  The sample rate is how many samples there are per second.  With a higher sample rate, more frequencies can be represented—up to half the sampling rate.

 

Two tracks (stereo) of 16-bit depth at 44.1 kHz sampling rate is 2 x 16 bits/sample x 44100 samples/second = 1411200 bits/second = 1411.2 kbps.  Increasing the sample rate and increasing the bit depth both increase the bitrate (number of bits / time).  Increasing the bit depth is not necessary to increase the sample rate.


Edited by mikeaj - 9/24/12 at 8:50pm
post #875 of 1923
Quote:
Originally Posted by ctoth666 View Post

Yes there is randomization in the algorithm, but the generated sound is still the same across all of the files is it not? I used the same Live project, and it was merely exported with different settings. Also how is it that they aren't level-matched? All of the files were normalized. I just don't know what any of this means tbh. I'm out of my depth. All I know is that I tweaked the export settings on Ableton Live project, and all of the files sound slightly different. That's all I know. And if I am to assume that there should be no audible difference between bit and/or sample rates, then pardon my daftness but that confuses the heck out of me. Anyway, I'm going to muck around with this a bit more and see what potentially erroneous conclusions I can come to.

Ableton is not very transparent with its downsampling algorithm, use SoX libraries for that job. Or if you are not well verse in programming(I'm pretty dumb myself), Adobe Audition 5.5/6 has very transparent sampling algorithms.

 

Anyone would like to have a 96/88.2khz downsampling test, I will downsample one of the Master's of Their Day tracks and upload them later.


Edited by firev1 - 9/24/12 at 9:48pm
post #876 of 1923
Quote:
Originally Posted by xnor View Post

Yes all three a pretty bad for what should be a clean 20 kHz sine (not triangle, square or something in between them) wave. After all, you can get 128x oversampling from < $1 chips and also clean results like this from onboard audio:

 

 

 

Sorry, I should have added that the scope (an Oscium thing that attaches to the iPad) has errors which show up whatever I measure, whether it be an iPod, my MacBook Pro or a DAC. The point I wanted to make wasn't in regards to the linearity of the DACs in particular, as I don't have the proper equipment to measure that. The point I've explained in the posts I made. If you want to find some excuse to invalidate my comments, please address where you believe I'm wrong in what I'm commenting on, which was to illustrate a real example of the behaviour of a NOS DAC and how it behaves similarly to an oversampling DAC with high-res files. smile.gif

post #877 of 1923
Quote:
Originally Posted by Currawong View Post

Sorry, I should have added that the scope (an Oscium thing that attaches to the iPad) has errors which show up whatever I measure, whether it be an iPod, my MacBook Pro or a DAC. The point I wanted to make wasn't in regards to the linearity of the DACs in particular, as I don't have the proper equipment to measure that. The point I've explained in the posts I made. If you want to find some excuse to invalidate my comments, please address where you believe I'm wrong in what I'm commenting on, which was to illustrate a real example of the behaviour of a NOS DAC and how it behaves similarly to an oversampling DAC with high-res files. smile.gif

No no I don't want to invalidate your comments. I was just commenting on the performance.

I agreed with the mastering point you made, but for a non-oversampling DAC to behave similarly to an oversampling DAC you'd need files sampled at a much higher rate. But since there's a compromise between speed and accuracy there are no AD converters that work at such high sampling rates. So you'd have to oversample on your computer or in a pre-filtering stage which kinda defeats the purpose of non-oversampling DACs. There are many more problems but I think I've made my point pretty clear.


Edited by xnor - 9/25/12 at 9:00am
post #878 of 1923

So in a nutshell, those settings should be perfectly fine for me:

 

Audio-GD NFB 12.1 -> OS: 4, Filter: 2 (altough this means 4x oversampling?)

Windows sound properties: 16 bit, 48kHz

foobar2000 + WASAPI (16 bit output) + resampler (48kHz)

 

While using Sennheiser HD600.

 

Although it seems like 24 bit won't do any harm ?

post #879 of 1923
Quote:
Originally Posted by xnor View Post

Quote:
Originally Posted by Currawong View Post

Sorry, I should have added that the scope (an Oscium thing that attaches to the iPad) has errors which show up whatever I measure, whether it be an iPod, my MacBook Pro or a DAC. The point I wanted to make wasn't in regards to the linearity of the DACs in particular, as I don't have the proper equipment to measure that. The point I've explained in the posts I made. If you want to find some excuse to invalidate my comments, please address where you believe I'm wrong in what I'm commenting on, which was to illustrate a real example of the behaviour of a NOS DAC and how it behaves similarly to an oversampling DAC with high-res files. smile.gif

No no I don't want to invalidate your comments. I was just commenting on the performance.

I agreed with the mastering point you made, but for a non-oversampling DAC to behave similarly to an oversampling DAC you'd need files sampled at a much higher rate. But since there's a compromise between speed and accuracy there are no AD converters that work at such high sampling rates. So you'd have to oversample on your computer or in a pre-filtering stage which kinda defeats the purpose of non-oversampling DACs. There are many more problems but I think I've made my point pretty clear.

 

Many DACs use 2x or 4x oversampling (though most AFAIK use 8x for 44.1kHz data), which is in the range of available high-res files (88.2 or 176.4 kHz). I think DSD and DXD recording covers the point you make, though I don't know what equipment is available to make such recordings. Whether there are actual benefits, of course, is another matter altogether. It's an interesting topic, anyway.

post #880 of 1923
Quote:
Originally Posted by Currawong View Post

 

Many DACs use 2x or 4x oversampling (though most AFAIK use 8x for 44.1kHz data), which is in the range of available high-res files (88.2 or 176.4 kHz). I think DSD and DXD recording covers the point you make, though I don't know what equipment is available to make such recordings. Whether there are actual benefits, of course, is another matter altogether. It's an interesting topic, anyway.

Nope, DACs using delta-sigma modulation typically work at 128 times oversampling / the sampling rate. 1-bit DACs wouldn't even work at just 4x or 8x oversampling. The minimum should be somewhere around 32x iirc.

 

DSD also is a single bit signal sampled at 64 times the CD audio sampling rate. DXD has 24 bit resolution and is sampled at 8 times the CD audio sampling rate, but does your NOS DAC accept such a signal?

post #881 of 1923

https://www.dropbox.com/s/n8tad28hnb1pgef/A.aif

https://www.dropbox.com/s/fmcqtmjknj6kz8i/B.aif

 

These are from Masters From Their Day episode one. Let me know what you guys think, whether A or B is the resampled one. I expect the tracks to be played on Foobar with A/BX plugin and replaygain on. 

post #882 of 1923

Wow nice read, I do need that WASAPI plug in, and although I did not read the entire thread [only pages 1 and this one]

 

While technically there may not be a differance, [and I have 24 and 16bit recording of which there is minimal or no differeance] but the placebo of "extra depth" does make your music sound better... why because we humans are emotional, and the joy of finding that 24bit flac of your favorite song just makes it all the better <3

 

Still though good read and it makes a lot of sense, good stuff to know!

post #883 of 1923

In theory nothing really matters. In reality however, I've never heard two different albums sound exactly the same. I've never heard two speakers sound exactly the same. People want to screw around with hardware to try to tweak the best sound, its understandable. I even use EQs! Yipes!

post #884 of 1923

Also can anyone explain how or why so many companies choose to rely on a 1-bit DAC?

post #885 of 1923
Quote:
Originally Posted by MrMateoHead View Post

Also can anyone explain how or why so many companies choose to rely on a 1-bit DAC?

High linearity and low cost. Btw, sigma-delta modulation can be mixed with multi-bit DACs.

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!