24bit vs 16bit, the myth exploded!

[gregorio]

So without dither 24bit would sound superior to 16bit?

In theory it might under certain circumstances. Without dither we’re left with quantisation error, in virtually all cases that would be inaudible at reasonable listening levels. However, it would be possible to create a signal and create a type of quantisation error called “truncation error” that would be borderline audible given a very loud (just about “reasonable”) listening level and an excellent listening environment. The chances of that type of signal existing in a music recording and those other conditions all aligning are probably somewhere between none at all and extremely remote.

However, your question is hypothetical. In almost all cases where quantisation error would occur, dither is applied automatically and cannot be disabled (in ADCs and DACs for example). The only place it could occur is when manually reducing bit depth but then of course you should manually apply dither when doing so.

G

 

[71 dB]

It's generally accepted you can't hear a difference between 24 and 16 bit but going from 24 to 16bit introduces noise from the quantization so you need dither to smooth the sound and make it sound less harsh. So without dither 24bit would sound superior to 16bit?

Not using dither causes distortion due to quantization error, but in case of 16 bit the distortion level is so low it is practically inaudible. Using dither when truncating the bit depth is however the "correct" way to go, because it keeps the signal distortion free and only raises the noise floor.

 

[71 dB]

That is a very good point.

I did the exact same process again, converting 24bit to 16bit disabling all dither and changing from best quality (slowest) to worst quality) fastest and surprisingly, the result was lower.

With using dither and highest quality, the difference was -90db white noise below 20kHz and -85db white noise above 20kHz

Without dither and lowest quality, the difference is -110db white noise equally spread over the whole spectrum. So the quality is actually higher without dithering...

The noise without dither is quantization error which is much worse than dither noise. In this case it doesn't matter, because the noise is inaudible in both cases, but if you go 8 bit, it matter a lot, because quantization error sounds much nastier than dither noise. Dither noise allow the signal decay "below" the noise floor (you can hear the signal fading into the noise and becoming gradually inaudible, because the noise masks it more and more. Quantization error doesn't do that. When the signal becomes quieter than the least significant bit, the signal disappears. Dither adds the amount of noise, but transforms the noise into something much less annoying, because the signal has become distortion-free.

 

[gregorio]

Without dither and lowest quality, the difference is -110db white noise equally spread over the whole spectrum. So the quality is actually higher without dithering...

It’s not entirely clear from your post but it appears you’re defining “quality” simply by the level of noise. Dither results in more noise (a higher noise floor) than not using dither but not using dither results in signal correlated distortion. The former being preferable as far as “quality” is concerned but it appears you haven’t measured the latter.

With using dither and highest quality, the difference was -90db white noise below 20kHz and -85db white noise above 20kHz

That’s expected. It’s possible to apply 1 bit (or less) of dither and thereby have a lower noise floor but not fully eliminate the signal correlated distortion. Higher quality would apply more than 1 bit of dither, result in a higher noise floor but no signal correlated distortion at all.

The actual solution to the potential issue you describe was published in 1989 and has been standard practice for 25 years or more. Which is not to apply a higher or lower dither “quality” but to apply noise-shaped dither! This would result in both no correlated signal distortion and a digital noise floor at roughly the -110dB to -120dB level (within the critical hearing band).

G

 

[Ghoostknight]

Without dither and lowest quality, the difference is -110db white noise equally spread over the whole spectrum. So the quality is actually higher without dithering...

you think all studio guys use dither for fun? its not about the noise, the added noise is a compromise because it gets rid of quantization artefacts which ARE audible (also for 24 bit to 16 bit conversion)... with dither it should sound smoother

these quantization errors actually sit on top of the main tone (for a sine wave for example), it distorts the sinewave from "perfect" to something less perfect, dither, even sitting below -90db gets rid of the distortion on the main tone

 

[gregorio]

its not about the noise, the added noise is a compromise because it gets rid of quantization artefacts which ARE audible...

Dither is not really a compromise as such, it’s a required function (like a low pass filter to limit the bandwidth) in order for digital audio to fulfil its purpose. Even in the early development years, before chips were able/available to generate dither functions, dither was still applied by injecting an equivalent analogue signal into the conversion process.

Also, your consistent (false) claims of audibility are getting tiresome! In the context of this thread (16bit vs 24bit) quantisation errors could only be audible in extreme cases which probably never exist for consumers and even if they did, then still they would only be borderline audible at best!

Either start providing some reliable evidence of audibility or stop repeating the same false audibility claims!

G

 

[Ghoostknight]

Dither is not really a compromise as such, it’s a required function (like a low pass filter to limit the bandwidth) in order for digital audio to fulfil its purpose.

true, i just wanted to say people who use dither dont care for the added noise floor because that is not what its really about

Also, your consistent (false) claims of audibility are getting tiresome! In the context of this thread (16bit vs 24bit) quantisation errors could only be audible in extreme cases which probably never exist for consumers and even if they did, then still they would only be borderline audible at best!

there are some videos on youtube comparing dither vs no dither even with youtube codec i was able to spot a slight difference on dithered 16bit vs nondithered, tho i agree, on 8bit the effect is far easier to hear

well i dont "really" care if studys show no audibility if i can hear a difference myself, this is to me more of a contradiction then being biased and curious could be

Edit: i also dont think thousands of people use dither on 24bit->16bit conversion because its "rarely audible"

 

[gregorio]

there are some videos on youtube comparing dither vs no dither even with youtube codec

The YouTube codec is not a comparison of 24bit vs 16bit dither. It is a lossy and loudness normalised signal. So again, you seem to have compared something and then made a conclusion about something else that’s substantially/entirely different to what you’ve compared!

well i dont "really" care if studys show no audibility if i can hear a difference myself

You don’t really care about studies AND you also don’t really care “if you can hear a difference yourself”! The only thing you appear to care about is whether you can perceive/imagine (rather than hear) a difference and then making up some nonsense theory (that isn’t a theory) to explain it!

If you don’t care about the proven science or even basic/simple science you can do yourself (at pretty much zero cost), then why do you even visit this subforum, let alone repeatedly post here?

G

 

[castleofargh]

It's generally accepted you can't hear a difference between 24 and 16 bit but going from 24 to 16bit introduces noise from the quantization so you need dither to smooth the sound and make it sound less harsh. So without dither 24bit would sound superior to 16bit?

Think about the level of the quantization noise. Now take your favorite music, play it as usual or even fairly loud if you want, and then lower the volume by 40dB. How loud does that feel?

Some people are talking about night and day delusions on how something yet another 40dB and then some is clearly killing the soundstage, the perceived dynamic, and the bees, all while the music is still playing at full loudness over the crazy quiet stuff of concern!

We can discuss audibility thresholds as a matter of curiosity. I quite like to learn about that myself. But those who pretend that ants are the size of moutains and that BS testing is absolute proof for generalized claims, need to be given the level of credibility they deserve.

 

[castleofargh]

well i dont "really" care if studys show no audibility if i can hear a difference myself, this is to me more of a contradiction then being biased and curious could be

Of course! If I used absolutely perfect methods and reasoning to get my incontrovertible conclusions on audibility like you always do, I would also be satisfied with my findings.
I mean, generic conclusions about filters from, which one was it, pushing a button on one DAC? And declaring that your casual impression is a fact of audibility for filters under widely different circumstances even in an EQ topic. Or right now, a claim that 16 bit truncation is audible without bringing up any testing condition for when the claim applies, just because you felt something while playing some youtube video.
Why is anybody even questioning the legitimacy of your empty claims? It doesn’t make sense.



If your plan is to become a professional reviewer, don’t change a thing. The type that can make up differences for everything and never applies any rigor to his "critical listening", has always been popular in the hobby.
For everything else, a little introspection and a lot less claim posting could go a long way.

 

[71 dB]

Here is an illustration of how dither is able to allow signal levels below least significant bit:

 

[bigshot]

It's generally accepted you can't hear a difference between 24 and 16 bit but going from 24 to 16bit introduces noise from the quantization so you need dither to smooth the sound and make it sound less harsh.

Even with 16 bit dither is a very small difference. Dither isn’t a big thing. It’s a small improvement that it many cases doesn’t make any significant difference in the real world. In my sig there’s a link to a video by Ethan Winer where he demonstrates a track with and without dither. On his website you can download the tracks and listen to them carefully yourself. It’s worth checking out for yourself so you know exactly what degree of difference it makes, but the short answer is that you probably wouldn’t notice that dithering hadn’t been applied with normal music listening. It isn’t a very big difference at all. It’s more of a low level polish than it is a make or break.

I’m not saying you don’t need to use dithering. I’m just saying it doesn’t make a huge difference.

 

[eq1849]

Here is an illustration of how dither is able to allow signal levels below least significant bit:

Let me ask you this....what do you think the purpose of dither is?

 

[bigshot]

well i dont "really" care if studys show no audibility if i can hear a difference myself,

Spoken like a true solipsist! …or by the least sharp knife in the drawer.

 

[71 dB]

Let me ask you this....what do you think the purpose of dither is?

To avoid distortion due to quantization error. Instead of distorted signal we have distortion free signal + uncorrelated noise which contain all the problems of quantization. It is a better, wiser and more pleasant way to swallow quantization errors.