16bit/24bit/36bit when saving to .wav

[Blue Wind Project]

I use FL Studio, and when I save a project as a .wav file, the default is 16bit int

I assume that's the default, but I also notice that there is a 24bit int and
36bit float.

What's the difference and which is best to use ?

[adam_firegate]

For master archive purposes 24 bit is the standard. Also got way more headroom than 16 bit. I cant notice any difference in 24 bit vs 32 bit.

[Biome_Digital]

you will get all different answers and arguments about dithering etc... but I will say choose 24 bit.

[Tricky-Loops]

Isn't floating point better than integer, i. e. more accurate?

I've never heard of 36 bit but I'd prefer 24 or 32 bit...

[Blue Wind Project]

Cool...thanks everyone. Will use 24bit for now on.

[Blue Wind Project]

Isn't floating point better than integer, i. e. more accurate?

I've never heard of 36 bit but I'd prefer 24 or 32 bit...

Opps...yes, 32 bit. Just had surgery on my right hand and the pain meds got my vision a
bit blurry.....the Evan Williams is doing it's fair share of blurring the vision as well

[Tarekith]

32bit files are usually floating point, and really all it buys you is more headroom so it's almost impossible to clip your Mixdown. Other than that, no difference between 32bit and 24bit in terms of sound quality. Like everyone else, I recommend 24bit as long as you're sure you're not clipping.

32bit is always the safest option, but like I said if you're not clipping anyway (any why would you be?) just use 24.

[aciddose]

24 bit will have higher noise than float, as float actually has 25 bits of precision.

In addition to that, the host is using float to store the audio in the first place and any conversion back and forth from 24 bit will add noise. You might as well completely skip this step.

24-bit ints are 3/4 the size of floats. That is their only advantage.

If you have enough disk space for 1000 tracks in float, you'll get 1250 in 24 bit.

Also, there are two formats for "24 bit". One is unaligned "three bytes per sample", one is aligned "24 bit packed into 32 bits". This can lead to issues when you move files between hosts and plugins.

The aligned files will be able to be decoded far more quickly than the unaligned. The unaligned will be 3/4 the size, but this is unlikely to make any significant difference.

There is only one format for float.

You're far better off using float unless you have a particular issue that you can measure. In which case, do some benchmarking and use whatever works best. Otherwise just use float.

[JCJR]

I agree with aciddose, use 32 bit float unless you are hurting for disk space. For streaming off disk, I'd wager that float might be a measurably faster load than packed 24bit, because modern processors are not optimized for odd addresses (which is what you have to read when loading packed 24 bit data from disk).

For sharing files with non-musicians, 16 bit is probably best. I don't think it impossible that a random granny or geezer, might not have trouble clicking on a 24 bit file and getting it to play. Maybe freeware thangs like mp3 encoders and such might choke on non-16 bit files as well, dunno. There are so many freeware audio programs, some old and cranky but still useful on occasion.

[itsclipping]

Yep agree totally with aciddose and JCJR. My preferences I'd recommend say:
32bit (for future editing potential) or,
16 bit (for final renders of tracks/general listening)

24 bit is something I only use when necessary. (like say, for using with out-of-date and neglected proprietary programs like Re-Cycle that won't support 32bit, but are one of the only programs that exports .rx2 loops for example.)

It's bigger than a 16 bit, doesn't sound any better on playback (to our ears, which 16 bit is good enough for basically, that's why it's used as the CD specification for example), and lacks some of the future ability for editing that 32 offers. Why bother?

[metamorphosis]

Agree with aciddose in the case, as in my testing, there are minor differences when downsampling from 32-bit as opposed to 24-bit. This may be the mere fact that 32-bit float is (typically) the internal format that audio hosts deal with, so when bouncing down to 16-bit from 24-bit you're actually getting the result of two format conversions rather than one, or it could be the increased headroom giving minor differences in conversion.

[Samplecraze]

Is your project in 24 bit? If it's in 16 bit then you will add nothing but 'size' to the file. If the project is in 24 bit then render as 24 bit so the ME can then dither. If not then stick to the project bit depth.

[fluffy_little_something]

There is an article on audio quality on the Guardian today:

http://www.theguardian.com/technology/2 ... sic-hi-res