16 bit vs 24 bit

[Bitfiend]

I know we've already kicked this one around a bit (no pun intended) but since i've been reading and learning more a couple questions have arisen that i don't think we covered before.

Since the noise floor of an average 24 bit file is -139 dBFS and the noise floor of the average 16 bit file is -91 dBFS and since you'd have to lower the 24 bit file by 48 dB to yield an effective 16 bit file; wouldn't it be better to record or work ITB at 24 bits?

Additionally, when people talk about Quantinization Distortion, is that refering to quantinization from the grid?

[admin]

I don't understand your logic here. Why would it be better to work in 24 bit based of that observation?

Quantization distortion is related to the level of precision a samples' bit depth. Distortion comes in two types, correlated and non-correlated. Correlated means as the nature of the audio changes (for instance in tone) the distortion changes. In other words different harmonics are generated via the distortion. Non correlated is another name for noise. In this case the distortion (noise) does not change according to the audio, it's just there as a constant source.

When using dither the quantization distortion results in noise, without dither its correlated distortion (like a guitar amp or the effect of using a bit reducer plugin).

Don't think of audio as on a grid though, that is a misconception. We're using Cubase here not Photoshop ;-)

[Bitfiend]

I don't understand your logic here. Why would it be better to work in 24 bit based of that observation?

Well if i'm understanding what i'm reading (and that my friend is a VERY big if);
Because of all the additional headroom you'd get at 24bit, you'd a have much better dynamic range and a cleaner recording as distortions can accumulate. Plus when applying a dither, you'd get more of the useful sound in the audible range than with a 16bit file. Plus with a lower noise floor and distortions, noise, ect. would be way down at the bottom.
So am i understanding this stuff correctly? And in actuality, does this crap really matter to the average ear? I mean on a good recording. Does anyone actually hear this stuff besides audiophiles and trained pro's like ME's, and such?

Quantization distortion is related to the level of precision a samples' bit depth. Distortion comes in two types, correlated and non-correlated. Correlated means as the nature of the audio changes (for instance in tone) the distortion changes.

So this would be like a guitar amp with distortion, or a saturation plug-in on a channel?

In other words different harmonics are generated via the distortion. Non correlated is another name for noise. In this case the distortion (noise) does not change according to the audio, it's just there as a constant source.

The dither inserts a noise level at a certain range to push the audible freq's up a bit, correct? would this also be noise from an analogue device or console?

Don't think of audio as on a grid though, that is a misconception. We're using Cubase here not Photoshop ;-)

Oh i wasn't, i was thinking quantinization as in moving parts around on the grid in cubase. but it's a different quantinization which is what i didn't know.

[admin]

You have to remember that your mixing environment is 32bit so any calculations after you record are done in that precision. Prior to that all you need to worry about is making sure the noise floor inherient in the system is below the noise floor of your recorded source. Most sources are above 93db. Yes they acumulate but it still doesn't add up to anything. Too many people forget about the scale and actual perception of these errors. On paper it might seem like it means something but it doesn't.

"So this would be like a guitar amp with distortion, or a saturation plug-in on a channel?"

Yes I just meant that this is the type of distortion you think of when you hear the word 'distortion'
as opposed to noise which I'd think most people just think of as noise, not a type of distortion.

"The dither inserts a noise level at a certain range to push the audible freq's up a bit, correct? would this also be noise from an analogue device or console?"

Not exactly. What dither does is 'jiggle' the lowest bits so they toggle randomly and in no way correlated to the signal. This means there's no harmonic distortion going on (grungyness/nastyness etc..). You just get noise...

The term dither comes from the word 'didder' which means to tremble/vibrate. The concept was originally in the form of a physical movement. Bomber guidance systems in WWII had a tendacy to stick and not be as accurate as needed. Someone found that by vibrating the mechanics of the guidance system that it would keep it from getting unstuck as much (since it was always in a bit of motion).

Dither does the same thing, it stops the lowest bits from getting stuck in a manner that is related to the signal itself. What this does is completely linearize the signal so there is no correlated distortions. You have a completely continous signal just like in analog, only drawback is a bit of noise added. I will post a little lesson on this soon to help you guys see and hear the effects so it's more clear.

[Rambunkcious]

Thats some interesting info on didder and the history of the word, thx.

[Bitfiend]

Prior to that all you need to worry about is making sure the noise floor inherent in the system is below the noise floor of your recorded source. Most sources are above 93db.

How does this apply when everything is ITB? Like using a vst synth or vst plug-ins on either samples or vst synths? How does it apply to one hit samples like the House Definition collection?

Yes they accumulate but it still doesn't add up to anything. Too many people forget about the scale and actual perception of these errors. On paper it might seem like it means something but it doesn't.

So, except in I guess extreme cases, this really isn't audible for most people?

I will post a little lesson on this soon to help you guys see and hear the effects so it's more clear.

that would be most helpful and appreciated. But dithering shouldn't be done until the mastering phase correct?

Actually just write a damn book. The explanations you give are much easier to understand then the "textbook-esque" stuff I’ve been reading. It’s really hard to put it into practical terms and even harder trying to figure out how it applies to electronica and working ITB only.

[admin]

The rule for dithering is you do it anytime you are reducing bit depth. Inside your DAW this is happening in certain stages of the mix path automatically.

So even if you are sending a song to be mastered technically you should dither as the output of most DAWs now is 32bit. Most people send masters over at 16 or 24, which is a reduction so in this case 'technically' you should apply it.

But getting to your other question no unless you really abuse the system it's not going to be noticeable. Again some examples will help illustrate this.

Well that's good to hear. You have the benefit of me having gone through all this stuff for many years now, at points obsessively for months on end researching and testing this type of info.

There is a new Medway Studios site on the way which is designed for a lot more content and user interaction. Once that's up then I'll be posting a lot more stuff. I think it's going to be really cool...

[Bitfiend]

.....Again some examples will help illustrate this.
Well that's good to hear. You have the benefit of me having gone through all this stuff for many years now, at points obsessively for months on end researching and testing this type of info.
There is a new Medway Studios site on the way which is designed for a lot more content and user interaction. Once that's up then I'll be posting a lot more stuff. I think it's going to be really cool...

Yes i think examples, best and worst case scenario would REALLY help out!

[admin]

Yup I think it will be a lot clearer when I do that. This is one of the harder things in audio to get your head around, especially after hearing so much info that's wrong regarding it.