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 130

post #1936 of 1938
Quote:
Originally Posted by audiosampling View Post
 


Yes, this is my website. And yes, I am 100% sure: there is no dithering applied. Actually, I did apply dithering first, but the fade out (at the end of the extract) would then fade out into audible noise (with or without shaping). So, it would have been very easy for people to find out which was the 8bit version, by simply listening to the fade out. That's what I wanted to avoid and why I removed the dithering.

 

I am very aware of what dithering does and how it works. BTW, I wrote the following page, providing audio examples and audible tests:

 

http://www.audiocheck.net/audiotests_dithering.php

 

It shows how dithering, applied to an 8-bit file, increases the dynamic range by 18dB!


Thanks for confirming. I also took a look at the other page. Yeah, fading into audible noise would be more obvious i guess, I was just suprised by how inaudible the quantization noise was in the tracks.

It's an interesting website so thanks, I'm curious to know how many people can tell pass the abx test using those tracks though.

post #1937 of 1938

Ethan Winer covers dithering in his Audio Myths video. See my sig for a link.

post #1938 of 1938

Check out these couple tests. One is from Reddit to which I replied to elaborate on and summarize:

 

From:

https://www.reddit.com/r/audiophile/comments/2n5zwe/beating_the_24bit_horse/

Quote:
 Do you think you can hear the difference between 24- and 16- bit audio? In that case, I have prepared a torture test for you.

Here is a zip file (stored with -0, so no compression) with set of two 24-bit wav files (16MB).

In it are two 24-bit audio named x[six-digit number].wav that contain the following:

  1. One of the files is a 30-second excerpt of "Solstice" by Versa & Rowl
  2. The other file is also an excerpt of "Solstice", with a twist. Into the lowest nine bits of the track, I have mixed in a 30-second excerpt of "Off to the Races" by Lana Del Rey.

Now:

  1. Open the files in your favorite ABX tool.
  2. Make a test run with 10 tests. How many correct guesses did you make? (If you have an ABX log, feel free to add it)
  3. If you know of means to cheat, don't - that takes the fun out of it.

What is this test testing?

By mixing in a signal that is uncorrelated to the original, I'm making sort of a worst-case test here - the signal I'm mixing in is twice as loud than the LSB of a 16-bit file, and I've chosen a quiet passage of the Versa & Rowl track and the loudest passage of the already loud Lana Del Rey track.

If you can't hear the difference between the two, it's highly unlikely that you'll see any sort of benefit from 24-bit audio files

 

 

This is incredible. Let me explain what was done to make this test with another file you can use to test the theory even more.

Here is a 24-bit flac, where the data has been attenuated so it roughly fits into the 9 bits. This file was posted by Arve earlier.

This file has a song with the music volume turned so low that there is only data in the least significant 9 out of 24 bits of the sound file. If it was a 16-bit file at the same volume, we would only have the last bit of audio data present which would encode basically the presence of what would sound like a tick or a pop - the quietest possible sound that could be in the file. Any lower is pure silence. In the 24-bit file though, the last 9 bits allow an actual recognizable song to be hidden in there instead of just ticks, so we can use it to check if we are able to actually hear a real song buried deeper than what's possible in a 16 bit file.

Normally I run my receiver at level 50 (I am not sure if this corresponds to db) for medium-loud near-field listening (I sit about 1 meter from the speakers). Often I use it at about 43-45 for medium volume listening. The receiver is turned up so high because it gets a quiet input due to an EQ that turns the total level down by approx 15db.

When I play the file at my normal loudish volume, I hear nothing since it is so quiet. If I turn the receiver up to max which is level 74, AND disable the DSP (so it is total 40-45db higher than normal level), and all of this with the computer output volume as high as possible, I at first hear nothing from the speakers except noise from the receiver being turned to max. And my quiet fridge humming away 7 meters away across the room. This is absolutely as loud as this 2x80W system goes.

If I put my ear up to the speakers mid/tweeter, I can hear the song through the speaker hiss. But when I sit at a normal position, 1 meter away from the speakers, I can't hear anything but the noise that's a side-effect of my room and having the speakers so loud.

There's another file with the music in the 8 least significant bits, with the music half as loud, as expected. I could barely pick the song out of the hiss with my ear right up to it. Mind you my receiver is a cheap 8 year old and hisses at max volume like a running water tap would sound from down the hall. Some other receivers have lower noise floor and in a silent room with the volume maxed, you'd hear the track more clearly. But, you would never run your receiver at max!

What Arve has done here in the post is take this barely audible file and put it into and under another track that is at a normal level. The only way you can hear the quiet sound is to play the regular-volume song to ear-destroying levels.

Basically what this test shows is that in order to have not only the 8 extra bits that 24-bit gives you audible, but even the least significant bit of the 16 bit file audible, I have to have my stereo so loud that my ears would literally burst, if the speakers wouldn't go first.

If our ears and speakers could handle it, you couldn't make out the barely audible least significant 9 bits through the hurricaine of noise anyway, due to multiple mechanisms:

  1. Your brain very likely wouldn't be sensitive to it with all the much louder sound present. But if it was;
  2. There would be mechanical noise generated on the eardrum from the extremely high SPL that would be louder than the sound present in the 9 least significant bits, completely overpowering it. But, even if we had perfect eardrums;
  3. The ear has a mechanism that engages with loud sounds where it adjusts the bones to dampen incoming sound, possibly to the point where we then wouldn't hear the quietest fraction even if was present on it's own without all the other loud sound.

There is no question that unless you have a special application that requires enormous dynamic range, there is no hearing the extra quiet details granted in a 24 bit track. An example application where it would be necessary, would be to have the ability to encode explosions or flashbang grenades at lifelike levels, along with quality quiet whispers, in the same track. For example, you could use the least significant 16 bits of the 24 bit stream to encode a regular movie soundtrack, and you use the loudest 8 bits as headroom to be able to actually literally blow the audience's ears out, or to encode bass so loud so as to represent an earthquake or the earth's atmosphere splitting in half.

When is 24-bit necessary?

The use of 24-bit by engineers or musicians for recording and working with audio is explained elsewhere but here are other applications for why we may want 24 or higher:

Recordings of underwater or space phenomena, for example, where there can actually be sounds thousands or millions of times quieter than other sounds, and we want all of them to be in high quality. With fewer bits you either cut out the loudest stuff, or cut out the quietest stuff. If you want it all, you need 32, 64, 128 bits or more. Your microphone or measuring instrument and electronics better be that accurate though, limiting the application to the highest of scientific instruments/research.

This also opens the door to new technology. Assuming the microphones used were sensitive enough, the 24-bit file can contain sounds so quiet, such as the musician's breathing, their heart beating, the engineer turning a knob in the other room, or things in the room rattling due to the instruments. In the future with new processing algorithms, we may be able to extract these details to create more realism in the experience, or figure out more information about recordings.

If we develop a brain-audio interface to send the audio data directly into our neuropsychology, our brain can then potentially perceive the data to any level of detail, and with any amount of dynamic range with no potential for hearing damage. It is closer to "awareness" data about every frequency or sound and its level sent into our perception. In this case, the higher quality, the better, whether it's bit depth or sample rate. More digital data is directly proportional to more thought information. So, obtaining/storing files at 24 bit may be useful later even if we can't hear it now, but by then our current storage and delivery methods may be obsolete anyway.


Edited by k00zk0 - 11/26/14 at 12:50am
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!