Hi!

I loaded a 24bit wav file to FL Studio 10 (it's a 32bit DAW).
I edited the wav and I want to render it also to a 24bit wav file.
Should I render it with or without dithering?

FL Studio 10 processes audio in 32bits, so when I want to render this file to 24bit, do I need to use dithering?


thanks!

 

I don't know if FL has 32 bit audio engine, but never mess up x86 and 32 bit audio processing, because different things. 32 bit version of izotope ozone has 64 bit double precision or so, or psp master q 2 either x86 (32 bit plugin) or x64 (64 bit plugin) has 80 bit precision. So audio engine has nothing to do with binary, and vice versa.
So, if you use Microsoft Word x64, it doesn't mean fonts there are so round, analog looking and twice more pixels per letter))

So, you must dither. Because you are lowering bit depth. Truncation is worse than dithering. Details saved in noise are better than distortions and harshness and spoiled detaility.
For further processing, 24 bit dithering should be tpdf or very light noise shaping.
From float point there is no dithering. So, rendering 32 bit float from 64 bit float must be without any dither. Because it will have wrong calculations or you will get just dithering as a noise and nothing more.
So I always process, edit, etc with saving as 32 bit float/int (fixed). 64 bit rendered from reaper is not working in RX because supports up to 32 bit float / int. And 64 bit format is useless as a render. 24 bit is not enough for work for me. If we have 32 bit options for saving, use it. You loose less, and may achieve more with it. Float saves audio overclipped above 0 dB.

So, everytime use dither when lowering bit depth and for fixed bit depth only. 16 and 24.

But final downsampling must be always before final dithering and better before limiters. So, src -> limiting -> dithering -> render.

 

Last edited: May 3, 2017

Without.

Dither is for 16 bit rendering? And then only if the source is higher ie 24bit+

[It doesn't matter whether you use 32bit or 64bit DAW]

Post says "wants to render to 24 bit". So, not reducing bit depth (of audio)?


OTOH: http://productionadvice.co.uk/when-to-dither/

LOL take your pick.

 

Last edited: May 3, 2017

And what is the question? What's the problem here?
Want 24 bit file? Use 24 bit dithering, because you render from higher bit depth (plugins with 32 bit precision and higher, DAW mixing engine is 32 bit or higher). You reduce bit depth to 24 bit, it means you truncate, so dithering saves more details and no distortion you get.

Want 16 bit file? Same here. But here is less dynamic range, more truncation you achieve.

 

It doesn't matter whether you use 32bit or 64bit DAW, always apply dither when bit depth goes down (32 to 16, 32 to 24, 24 to 16). 24 and 16 bits are fixed so you must apply dither.

 

i asked this question years ago and there's no mutual agreement
then i decided to find it out by a/b comparison & comparing audio files in hex. editor

the result is
it is more related to volume adjustment & "bit depth loudness boundary" rather than just bit depth
reducing the volume of any audio file requires dithering if the lowest volume exceed the loudness boundary

maybe a few examples will help a lot
16bit source + reduce the volume + render to 16bit = dithering is required (very likely it will < -96db)
16bit source + increase the volume + render to 16bit = dithering is NOT required
16bit source + reduce the volume + render to 24bit = dithering is NOT required if the lowest volume >144dB
16bit source + increase the volume + render to 24bit = dithering is NOT required

hope that helps

 

I guess by "loudness boundary" you mean the threshold of audible quantization noise level?

 

If the question was once or not - if you render to 16 bit from dithered 24 bit, in any case you should use 16 bit dithering. But it is better to render to 16 bit from source project, that is from your maximum bit depth.
Bob Ohlsson dithers to 24 bit always in the end, before it hits DAC (for further analog processing and monitoring).

 

I wrote this in first message, reread this please.
First sentences there are about that x86 and x64 has nothing with audio engine at all.

 

↑
- So, you must dither. Because you are lowering bit depth.
- Post says "wants to render to 24 bit". So, not reducing bit depth (of audio)?

24 bit is a reducing from 32 or 64 bit. You need 24 bit dither.

 

if you go from one bit depth to the other, you get holes in your samples , basically Dither adds noise to the file to fill those holes.

 

16bit source + reduce the volume + render to 24bit = 24 bit dithering is NOT required if the lowest volume "-144dB or louder".
If you reduce the volume more than 48 dB, you need 24 bit dither. Because lowest volume of audio goes under 24 bit threshold, then truncation appears.
24 bit gives 48 dB more range. Means 96 + 48 = 144. So, if you reduce 48+ db volume, you need dither before render. If you reduce your 16 bit file with 20 dB, and save as 24 bit, you don't need dither, you have 28 db room. If more than 48 volume reduction, dither! Clear?

 

Last edited: May 5, 2017

To be clear, Dithering prevents truncation and distortion, not masks them. And dithering is before reducing bit depth, not after. If you dither post render, you just add useless noise to truncated audio. So, you get noise upon distortion. So, better truncation, than truncation with added dithering noise. The best is dithering noise without truncation.
Dither -> render
Not vice versa

 

Last edited: May 3, 2017

My personal belief is that dithering should virtually never be used. The only scenario I believe it should be used is if you have 24 bit+ and 96kHz+ samples in the project and you are wanting to export it to a fairly low quality format, i.e. 16 bit or less and 44.1kHz or less. Other than these extreme changes, I wouldn't bother with dithering personally.

 

I'm with Ian Shepherd as per the link posted by G String above. Unless you're staying in 32 bit floating (or your native DAW's engine) then always dither (just select the right bit depth to your desired output... BUT... as has been mentioned, dither only once. So when rendering/printing a mix, unless your printing at the same bit depth as your DAW's engine - use dither. Even if your source file is 24bit and you are processing it using any plugins and your output is 24bit; use 24 bit dither, but ONLY once i.e. only as the very last plugin on your master out/mix bus. Then if this goes onto to be mastered down to 16 bit, dither again!

The simple thing to do is listen and check but you can also use a dither meter to see if each bit is being used or if you'll be truncating. Again, in my opinion, as stated elsewhere, dither noise is generally preferable to quantization distortion. You can use auto black (or equivalent) if you have quiet passages which is effectively a gate for the dither noise during low passages (or silent).

 

DAW bits does not matter.
If you will keep the same bit resolution of your audio, so you do not need to add dither.

 

But dither noise does sound like quantization error, and according to my personal tests, dither is louder than quantization noise.

http://wiki.hydrogenaud.io/index.php?title=Dither

http://ethanwiner.com/dither.html

16bit has a DR of ~130dB, if you're quoting 96db, you're measuring it incorrectly.

So unless you're rendering to 8bit, you don't need dither.

If you're paranoid, Don't take what I say, instead, DO TESTS YOURSELF at 16bit.

 

I don't think I mentioned anything about measuring other than using one's own ears. Try it with and without... use whichever is preferable to your own taste. I personally prefer dither noise with the music I produce rather than quantization distortion (hence "generally preferable"). It's certainly not illegal to do one or the other. And another thing to note is there is a significant difference on the shape and the quality of the tool to generate the dither... Horses for courses.

As I said, Ian Shepherd made a very compelling argument and demonstration for me. I'm happy to be the only one to use this approach. So the OP has a wealth of information to read and make an informed choice.

 

Academic rule is : You ALWAYS apply dither when reducing bith depth (and dont mistake this with the 32-64 bits of a DAW/System it's not the same, FL is a 64 bit DAW that can work in 32-bits floating point, like any serious DAW). But after some posts on this thread and on AS, I'm not that much sure. Personnal tests I did aren't that much conclusive. So I tend to follow the "academic" rule I got told through my sound engineering grade.

 

if its 24 you can't add bit data make it sound 32. changing it would only add 0's [or nothing's] to allow it to be "labeled" 32, which would in turn increase the file size for no increased audio quality. so id only dither if going down to 16,8,4,2,1,0.5 lol