azslow, your files are the equivalent of 10 bit audio.

drewfx1
azslow, your files are the equivalent of 10 bit audio.

In that direction... But that is just an illustration for the effect. Exaggerated illustration, to allow understand it on the phone with $1 headphones. In better conditions the (same) effect is audible with higher signal levels.
The noise level from not dithering stay the same, with lower level we obviously lower SNR, down to what is demonstrated in the files.

Take a step back - the reason you only used 10 bits is because if you used all 16 bits no one would hear it.  If the only way one can hear it is to "exaggerate" the effect, that's another way of saying it's inaudible under normal conditions.

drewfx1
Take a step back - the reason you only used 10 bits is because if you used all 16 bits no one would hear it.  If the only way one can hear it is to "exaggerate" the effect, that's another way of saying it's inaudible under normal conditions.

"Not everyone" would hear it. The purpose was to demonstrate the effect for everyone. People can decide do they need it or not for what they do. It can be hard decision without understanding what it is at first place, what exactly it produce and when.
 
I have just mentioned that without dithering this bit is ALWAYS garbage. Original DAW output (especially synthetic) ALMOST ALWAYS has correct information for this bit. Dithering can preserve it. That are mathematical facts.
I repeat, if someone do not want or do not need one bit from total 16, I have no problem with that!
 
There is a "big clan" which say "I can hear the change from golden to silver wires of my speakers", some "pro" with "it is very important to do proper dithering at 24 bit" (Samplitude creator...) and yet "who needs the bit 16, lets use 15".
I do not care in what people believe.
Also I do not judge what they can hear or not, till that is technically impossible.
 

You guys are why I love this forum. Please QA my "Explain Like I'm 5", explanation:
 
In my limited understanding, bit depth represents dynamic range (absolute silence to the "11" on the volume scale.....ha!).  Right?  And at 24-bit you can take each sample of audio in your ADC and convert it in to potentially millions of slices (low volume to high volume) for each portion of sampled data (16,777,216 slices (of volume)). That's more than enough. 
 
To me, the other aspect, sample RATE, is far more crucial.
 
A sample rate of 44,100 will yield 441 samples for EACH sine wav for a tone of 100hz (bass drum, e.g).
 
A sample rate of 44,100 will yield 100.2 samples for EACH sine wav for a tone of 440hz (A note on a keyboard).
 
Mapping out each sine wave of a 440hz tone with 100 samples, and 16.7 million "volume resolution" (24-bit) will render a very good resolution of the wave form.  It is really quite good.
 
At the upper end of the sampling spectrum, 14,700hz for example, the 44,100 sample rate will only be three samples per sine wave.  Imagine a sine wav and placing 3 equidistant points on the wave form and you can see the resolution go to hell real fast.  Thankfully there are great dithering algorithms out there, plus our ears, most of us anyway, suck at hearing freqs that high.
 
Boosting sample rates to 48,00 are meh (to me), and 96,000 more than doubles the 44.1 sample resolution (great for higher freqs, I suppose).....
 
Ultimately all of this fretting is lost on the user who is walking the noise downtown streets listening to your tune on their complete **** ear buds, with their ears ringing after a night out on the town in a noisy club (not to mention the fact that soundcloud is going to take your amazing creation and crap it out at a lovely 128kbps (basically tossing out 94% of the clarity of your 24bit (2116kbps / sec) original wav file, not to mention the fact that their 128K stream is being generated from your [hopefully] higher resolution mp3 submission.
 
I usually upload 320K mp3 files to soundcloud, and Who knows what sort of algorithm they run it through before dumping them out on the web at 128K...... yikes!!
 
Am I misinformed?
 
Thank you.  I yield the floor.

azslow3

drewfx1
Take a step back - the reason you only used 10 bits is because if you used all 16 bits no one would hear it.  If the only way one can hear it is to "exaggerate" the effect, that's another way of saying it's inaudible under normal conditions.

"Not everyone" would hear it. The purpose was to demonstrate the effect for everyone. 

But you didn't exaggerate by a little. For each bit you reduce the bit depth you double the amplitude of the quantization error + dither (QE + dither). You didn't increase the level by 2x or 4x (4x = 14 bit) - you increased it by a factor of 64x!
 
So why such a huge increase? Hmmm....

berlymahn
You guys are why I love this forum. Please QA my "Explain Like I'm 5", explanation:
 
In my limited understanding, bit depth represents dynamic range (absolute silence to the "11" on the volume scale.....ha!).  Right?  And at 24-bit you can take each sample of audio in your ADC and convert it in to potentially millions of slices (low volume to high volume) for each portion of sampled data (16,777,216 slices (of volume)). That's more than enough. 

 
I would say that for properly dithered audio, higher bit depth simply means less noise.
 
And once this noise is either below our absolute threshold of hearing (ATH) or buried under other noise (in the listening environment or in the signal itself) increasing bit depth accomplished nothing. You can't increase resolution by adding more bits when the bit depth isn't the thing limiting the resolution (fix that instead).
 

 
 
To me, the other aspect, sample RATE, is far more crucial.
 
A sample rate of 44,100 will yield 441 samples for EACH sine wav for a tone of 100hz (bass drum, e.g).
 
A sample rate of 44,100 will yield 100.2 samples for EACH sine wav for a tone of 440hz (A note on a keyboard).
 
Mapping out each sine wave of a 440hz tone with 100 samples, and 16.7 million "volume resolution" (24-bit) will render a very good resolution of the wave form.  It is really quite good.
 
At the upper end of the sampling spectrum, 14,700hz for example, the 44,100 sample rate will only be three samples per sine wave.  Imagine a sine wav and placing 3 equidistant points on the wave form and you can see the resolution go to hell real fast.  

No. What the sampling theorem says is that a signal that contains nothing greater than or equal to 1/2 the sampling rate (aka the Nyquist frequency) can be reconstructed perfectly, including the part between the samples. IOW, adding more samples per cycle doesn't increase resolution because all of the information about the part between the samples is already stored there in a long string of samples. And in fact if we know it's a pure sine wave, only 3 successive samples are enough to mathematically reconstruct the sine wave - the frequency, amplitude and phase of the sine wave can be calculated from only those 3 samples.
 
So increasing sample rate really only allows for higher frequencies in our signal, not higher resolution, at least not in and of itself.
 
In the real world perfect reconstruction is impossible, but modern converters are extraordinarily good up to ~10% below the Nyquist frequency.

drewfx1
So why such a huge increase? Hmmm....

Common. Make an example which YOU can hear. I am sure you can do this with at least 12-13bits, probably even all 15. Put please do not follow cymbal advise.... take at least clean guitar or FM synth.
I have just demonstrated that on ANY device, for ANY person there is NOT ZERO number of bits, when dithering effect can be observed.
 

berlymahn
Am I misinformed?

A bit
 

In my limited understanding, bit depth represents dynamic range (absolute silence to the "11" on the volume scale.....ha!).  Right?  And at 24-bit you can take each sample of audio in your ADC and convert it in to potentially millions of slices (low volume to high volume) for each portion of sampled data (16,777,216 slices (of volume)). That's more than enough. 

Almost... existing AD conversion does not work with 24 bits precision, best of them work with ~20. Once they manager to do all 24 bit, they will immediately claim >140dB SNR. You can easily check in interfaces specification that they mention 115-120dB at the moment (with some variations about RMS / dBA )
 

To me, the other aspect, sample RATE, is far more crucial.
 ...
At the upper end of the sampling spectrum, 14,700hz for example, the 44,100 sample rate will only be three samples per sine wave.  Imagine a sine wav and placing 3 equidistant points on the wave form and you can see the resolution go to hell real fast.
  Thankfully there are great dithering algorithms out there, plus our ears, most of us anyway, suck at hearing freqs that high.

Do not thank dithering nor ears for that. There was people which you can thank:
https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem
At "I'm 5 years" level: they have proved it is possible EXACTLY reproduce original wave with even less (!) then 3 points
Hard to believe at first. On "fingers": when you look far away on the road, do you need to recognize wheels or other fine details before you are sure that you see a car? I mean not a cow, not a house, not a bike. And it is on the road, not in garage and it does not fly.
Signal processing knows that in case there is SOMETHING (dot), that IS a wave. And it just need a tiny bit of extra information to identify WHICH one. If this wave is not 14,700hz but 14,701hz, the "dot" will be at different place (as long as you do not cross "magic border", that is why LPF is required during down-sampling...).
 
EDIT. I was slow, as usual . I am AZ Slow

JohnEgan
Thanks, I guess what I’m only asking is if I should maintain the same record bit depth setting in preferences I had set when I recorded (record bit depth 32),  from mix to master, (until final export to 16 bit MP3) or does it matter, as its already only actually 24 bit depth resolution/accuracy anyways?

Short answer: no need to over-think it; set your render depth to 32 and forget about it.
 
Export to...
    32 bits if you're going to be using another program for mastering or file conversion
    16 bits for writing to a CD
    24 bits if you want to import it to another DAW or a media player that can't handle 32 bits
    8 bits if you're composing for a video game in 1985
 
And BTW, it sounds worse in your car because it's in your car. Not because it's 16 bits.

azslow3

drewfx1
So why such a huge increase? Hmmm....

Common. Make an example which YOU can hear. I am sure you can do this with at least 12-13bits, probably even all 15. Put please do not follow cymbal advise.... take at least clean guitar or FM synth.
I have just demonstrated that on ANY device, for ANY person there is NOT ZERO number of bits, when dithering effect can be observed.

Yes. Your example demonstrates that dither is desirable in a 10 bit scenario. I don't think anyone will find that controversial.
 
Had you mentioned in the first place that it was an example of dithering to a 10 bit level then I wouldn't have commented. But you neglected to say that and thus people may have mistakenly believe your example (and any conclusions based on it) were based on 16 bit resolution rather than 10 bit resolution.