...and deep too

[DrXparaMental]

Got a few questions.

With respect to a noticeable audio quality improvement, does it really do any good to open a 16bit file as a 32bit file with respect to remastering that file or just working with the file in terms of editing?

For instance: Lets say I rip a tune off a CD (16bit) and then open it in Adobe Audition as a 32bit file. Is there an advantage to doing this?

[pixelbox]

No benefit. 16 bit is 16 bit. If you open a 16 bit file with a 32 bit application, you still only have 16 bit information.

[DrXparaMental]

No benefit. 16 bit is 16 bit. If you open a 16 bit file with a 32 bit application, you still only have 16 bit information.

That's what I expected to a certain degree, but I honestly wonder why they make that conversion option available if it's of no real use. Know what I mean?

[Lovin Teris]

For further processing. Keep it 32 bit until the end, then dither and export to whatever format you like.

[ciw]

if you're going to add effects of any sort, even just plain old eq, it will sound better to process your 16 bit file as 32 bit. then maybe dither.

[DrXparaMental]

makes sense. Twice the bit depth equals twice the detail with respect to effects processing.

What has always "thrown" me is the fact that you have to down convert to 16bit in order to burn to disc. I guess it's what takes place while the file is actually 32bit deep that really counts though.

I can definitely hear the difference with respect to recording 32bit processed and recorded live clips into a 24bit converter with 32bit processing and then into a 32bit editor (on a different machine). Much warmer/dynamic and clean. More defined space between the tracks so to speak.

I want to take the chance while I am thinking about it to recommend Adobe Audition 3 to Ableton Live users. It's a REALLY nice editor that I don't think can be beat for the money. Although it's a little CPU intensive, it's very friendly with a shitload of capability.

Thanks again for the input!

[Lovin Teris]

Twice the bit depth equals twice the detail with respect to effects processing.

Actually, no. What you get with more bit depth is headroom, not detail. That's why it's called depth and not e.g. resolution.

So recording in 24 bit versus 16 allows you to keep peak level at a safe distance from 0dBfs without having to worry about system noise.

But the big difference with 32 bit is that it's floating point, so the headroom is, for all intents and purposes, infinite. (Or at least extremely large.) Which is nice because you can, say, combine to signals to a level higher than 0dBfs internally, pull the master fader down, and still have a noise-free output (the output will be 24 or 16 bit, depending on your interface and settings).

[DrXparaMental]

Twice the bit depth equals twice the detail with respect to effects processing.

Actually, no. What you get with more bit depth is headroom, not detail. That's why it's called depth and not e.g. resolution.

So recording in 24 bit versus 16 allows you to keep peak level at a safe distance from 0dBfs without having to worry about system noise.

But the big difference with 32 bit is that it's floating point, so the headroom is, for all intents and purposes, infinite. (Or at least extremely large.) Which is nice because you can, say, combine to signals to a level higher than 0dBfs internally, pull the master fader down, and still have a noise-free output (the output will be 24 or 16 bit, depending on your interface and settings).

God I wish I understood what you just said!

I kind of do but I know I don't fully grasp it. If I even in the slightest bit (no pun intended) understand, 16bit files react react more "carelessly" to applied effects. They peak more because the information being processed is less data and more useless digital bulk ????

but then you say less signal to noise...I am lost. It's not your fault. It's just my inability to understand what you're saying because of a lack of education with respect to digital audio science.

[Lovin Teris]

God I wish I understood what you just said!

So would I ...

16 vs 24 bit: These systems are equal apart from the number of bits. The bit depth determines the dynamic range of the system, that is the difference between the noise floor and peak level (0dBfs or full scale). Both systems, when properly constructed, are just as precise in the available range, but the range of 24 bits is larger.

In other words, there is no need to push 24 bit recordings close to the top, recommended practice is to aim for something like -12dBfs and peak at -6. That way you don't have to worry about sudden peaks clipping, and you still have more dynamic range than if you push a 16 bit system to its max.

That being said, in most home studios the background mechanical noise will be way above the system noise. But there's no reason to make matters worse, esp. if you do lots of mixing and processing.

32 bit: A slightly different beast due to being floating-point. I personally record to 24 bit, but any modern DAW will mix and process in 32 bit or more. Which is good.

The forums and articles at Sound on Sound will explain this and related issues way better than me.

[Tone Deft]

Twice the bit depth equals twice the detail

Actually, no. What you get with more bit depth is headroom, not detail. That's why it's called depth and not e.g. resolution.

no, Dr. X is right in that case. or I'm misunderstanding you because you got the rest of it right and explained it well.

16 bit has ~65,000 steps
24 bit has ~16,000,000 steps

if the system is set so that 0dB is at -10dB from clipping, it's proportionally the same amount of headroom. dB is a ratio, multiplication, not addition, it scales over the dynamic range. more bits does mean more resolution, more detail, not more headroom.

you make sense with the rest.

[Lovin Teris]

Well, I must admit I'm not the greatest expert. But reading Sound on Sound for a few years is a great exercise, well worth recommending!

From what I understand, this might almost be a play with words. But here's a way of seeing it:
The steps you refer to mean that when the AD-converter samples at a certain point in time, it must choose between two adjacent steps if the amplitude is not exactly at one particular step. This resulting difference comes out as noise, and it's level is determined by the number of steps (many steps = small differences), hence the noise floor is lower for higher bit depth systems. So when you say clarity, I say less noise.

There are many factors to complicate this simple image, like dither, noise shaping etc., my understanding of which equals zero. Also, you must account for both the AD and the DA part of the process to understand it.

One important thing to notice is that any comparison to the pixels of an image is wrong. You can see a pixel as a coloured dot. But you can not hear a sample in digital audio. It's the change between samples you hear, the fluctuation in sound pressure is the waveform.

Have look at this article for starters.

Regarding headroom: Yes dB is a logarithmic scale. The difference in volume between -10 and 0 in 16 or 24 bits is the same. But the distance from the noise floor is larger in 24 bits. So you can work with the same safety margin (headroom) and still have plenty of dynamic range above the noise floor, or even work with a larger headroom. There is no reason to push recording levels close to 0 in 24 bits, even if it was in 16 bits at the time. (Mastering is a different matter, if you want your song to be loud.)

Phew...

[DrXparaMental]

Well, I must admit I'm not the greatest expert. But reading Sound on Sound for a few years is a great exercise, well worth recommending!

From what I understand, this might almost be a play with words. But here's a way of seeing it:
The steps you refer to mean that when the AD-converter samples at a certain point in time, it must choose between two adjacent steps if the amplitude is not exactly at one particular step. This resulting difference comes out as noise, and it's level is determined by the number of steps (many steps = small differences), hence the noise floor is lower for higher bit depth systems. So when you say clarity, I say less noise.

There are many factors to complicate this simple image, like dither, noise shaping etc., my understanding of which equals zero. Also, you must account for both the AD and the DA part of the process to understand it.

One important thing to notice is that any comparison to the pixels of an image is wrong. You can see a pixel as a coloured dot. But you can not hear a sample in digital audio. It's the change between samples you hear, the fluctuation in sound pressure is the waveform.

Have look at this article for starters.

Regarding headroom: Yes dB is a logarithmic scale. The difference in volume between -10 and 0 in 16 or 24 bits is the same. But the distance from the noise floor is larger in 24 bits. So you can work with the same safety margin (headroom) and still have plenty of dynamic range above the noise floor, or even work with a larger headroom. There is no reason to push recording levels close to 0 in 24 bits, even if it was in 16 bits at the time. (Mastering is a different matter, if you want your song to be loud.)

Phew...

GREAT article LT. I think SOS is going to be an excellent addition to my ongoing "learning the science of sound" effort. The type of visionary intelligence that Reeves had is stunning to say the least. It's hard for me to imagine being that far in front of the curve so to speak. I got the book Mastering Audio by Katz. That is also a BIG help.

[DrXparaMental]

oops! double post

[Lovin Teris]

SoS is great.

[spkey]

I can definitely hear the difference with respect to recording 32bit processed and recorded live clips into a 24bit converter with 32bit processing and then into a 32bit editor (on a different machine). Much warmer/dynamic and clean. More defined space between the tracks so to speak.

Can you really? You should not be able to hear the difference. Unless your audio interface introduces some kind of distortion at 32bits in contrast to 24bits which happen to sound pleasing to your ears.

Also, when you have a 16bit audio sample and convert it to 24bits in order to process you are actually upsampling (i.e. adding more steps to the digital waveform) so you should also take into consideration how good your A/D converters are for this process.