Is it worth exporting to 24bit...

[Goldilox]

...if the tracking session has been done in 16bit?

Have been happily recording away with the standard record session set at 16bit wav, not realising that the DAW had reverted to this default setting at some stage. I was just wondering whether there's any advantage to be had from exporting to a 24bit file from 16bit material or if the damage is already done?

There's no midi elements to the songs in question - only audio tracks.

 

[shortedaman]

as far as i know most mastering houses ask you to export in the same bit as you tracked. but i could be wrong.

 

[Chibi Nappa]

I imagine most DAWs do their mix and plugin calculations at 52 bit. So... maybe.


A metaphor to guess at how it could help: Instead of bits I'll use significant figures.

My "tracks" are 1 significant figure each: 2 and 5
In the first case I divide the tracks keeping 1 significant figure
(metaphor for the DAW calculating at 16 bit): 5/2 = 3
In that case, no it would not be beneficial to export at more than 1 significant figure.

In the second case, I upsample my numbers to two significant figures and then divide
(metaphor for the DAW with 16 bit data calculating at 52 bit): 5.0/2.0 = 2.5
In that case, yes it would be beneficial to export at 2 significant figures.


This is all a total guess, but it would not surprise me in the least if the 24 bit export was in fact more detailed. The question is this: Does the upsample from 2 to 2.0 and 5 to 5.0 introduce more or less error than performing the math operation at a lower "resolution"?

In the case with 1 sig fig, the dividend "3" could in reality be anywhere from 2.5 to 3.499999999. In the case of 2 sig figs, the dividend "2.5" could in reality be anywhere from 2.45 to 2.549999999.

If our numbers 2 and 5 are actually 1.500 and 5.400, then the true answer to the division is 3.600. Neither the 1 sig-fig calculation nor the 2 sig-fig upsampled calculation allow for that and the 1 sig-fig calculation seems closer. But these numbers are very coarse and this is only one calculation. I'm not going to sit down and see how it shakes out for hundreds of sig-figs over multiple calculations.

Again, the answer is: Maybe it could help.

 

[mattr]

Yes and no.

(For reference, most DAW audio engines operate with 32-bit float math)

Yes in the sense that, whilst the individual tracks are 16-bit, the resultant mix may have a wider dynamic range. As soon as you drop the level of one of your 16-bit recordings below unity, its going to start eating into dynamic range that wouldn't be there in the 16-bit realm. If you were to import a 16-bit track into your DAW, drop the level by 15db then bounce it to a 16-bit file, you would have lost some resolution. Apply any processing or plugins? Chances are the output of that reverb VST is 32-bit float, so you're prematurely reducing the resolution of that nice reverb tail if you bounce to 16-bit.

But no in the sense that, realistically speaking, its probably not worth fussing about if the mix is already done and dusted... in any normal situation you're highly unlikely to hear any difference at all. At the end of the day the final master is almost always going to be in 16-bit and those extra bits are going to be truncated and dithering applied, but its all about when along the line you decide to throw them away (and the best time and place for that is with an experienced mastering engineer). The 24-bit file will give the ME a better source to work from - greater resolution / dynamic range - and also as a side effect probably discourage you from doing damage to the mix (in the form of mix bus processing) when bouncing because you can feel comfortable leaving the mix a lot quieter.

 

[Massive Master]

Always export in 24-bit. Even if your source was *8* bit, the math is being thrown around in 32FP (or higher) and the resolution is exponentially higher.

Mastering engineers ask for the same sample rate as the project was recorded --- Fixed word lengths are ("should be") almost universally 24-bit from concept through completion, up until the actual replication master is produced.

 

[TripleM]

Always export in 24-bit. Even if your source was *8* bit, the math is being thrown around in 32FP (or higher) and the resolution is exponentially higher.

Mastering engineers ask for the same sample rate as the project was recorded --- Fixed word lengths are ("should be") almost universally 24-bit from concept through completion, up until the actual replication master is produced.

Hey John.

If someone gave you 16 bit files, couldn't you just convert them to 24 bit yourself? And in the end wouldn't that be equivalent to them exporting a 16 bit mix to 24 bit? In other words, does it matter where the word length expansion takes place?

Just curious.

 

[Massive Master]

Once lower resolution is achieved, the benefit to the higher resolution is gone.

Summing 16-bit files into a 24-bit file is going to have the higher (theoretically, 256x higher than 16-bit) resolution intact with what was done during summing.

16-bit is a fine delivery format. Heck, a good 320kbps MP3 is a reasonable delivery format in many cases. But it's not something you'd want to work with. Add the extra dither noise into the equation also.

You can make a nice 4x6 photo from a 3MP camera. But if you want a poster of that same photo, it's going to look freakishly better if shot with a 12MP camera. The 12MP photo will look fine in the 4x6 print also. The 3MP photo will not make a nice poster.

Always start and stay in high resolution with obscene amounts of clean headroom at every single possible step in the game. Once any of that is gone, it's gone forever.

 

[Farview]

Hey John.

If someone gave you 16 bit files, couldn't you just convert them to 24 bit yourself? And in the end wouldn't that be equivalent to them exporting a 16 bit mix to 24 bit? In other words, does it matter where the word length expansion takes place?

Just curious.

There is a difference between a 16 bit mix and mixing a bunch of 16 bit files together.

If you are mixing 16 bit files together, you are going to wind up with a mix that has more than 16 bits of resolution, so you would render it at 24 bit.

If you simply render it at 16 bit, all the extra information is lost, so converting that file into a 24 bit file will accomplish nothing other than eating up more hard drive space. You will just end up with a 24 bit file with only 16 bits of information in it.