# Topic: Dithering Audio: When Where and Why?

1. ## Dithering Audio: When Where and Why?

I am a bit confused about the concept of dithering and would like some quick clarification. It is my understanding that dithering is used to help "smooth over" the digital artifacts that occur when going from 24bit to 16bit. Is this correct, or a gross oversimplification?

I ask because I'm currently working on a project using VSL Opus 1 (16 bit), Atmosphere (32bit internal?), and Stormdrum (24 bit transfered). If I wanted to bounce the files as 24 bit, do I need to dither at all? How about if I bounce as 16 bit, do I dither all Atmosphere and Stormdrum tracks but NOT the VSL stuff, as it is already 16 bit? If it helps, I'm using Logic 7.1.

Thanks in advance for any help and advice!

2. ## Re: Dithering Audio: When Where and Why?

I'd recommend bouncing everything in 24-bits without dithering. DIther to 16-bits as your final step in the process.

Dithering is just a process of adding noise to a signal to help hide rounding errors when cutting down the number of bits on your signal. It can also be used to mask the problems of too few bits in general.

Here's one example: You have a video signal that is a smooth ramp from black to white. The left of your screen is black. The right is white. The middle is gray. Now you try to represent that with too few bits. Instead of a smooth wash, you see a stair step pattern from black to white. The contrast of every two neigboring levels causes a sharp transition that is easily visible.

Now picture the value at the transition between levels 16 and 17. Just to the left of the transition, the number before truncating was 16.49. Just to the right the number is 16.5.

By dithering we add noise before truncation. That means that the pixel that had a value of 16.49 will get randomized each time we present it. 49% of the time it will be closer to 16 and 51% of the time it will be closer to 17. The 16.5 value will be truncated to either 16 or 17 50% of the time.

The result is that the transition point becomes randomized with the stuff on the left darker on average than the stuff on the right. There's no longer a clear line of transition.

Anyway, keep as many bits as possible for as long as possible. Dithering adds noise to your signal, so only do it once at the very end.

-JF

3. ## Re: Dithering Audio: When Where and Why?

Well, I found some interesting information. Basically I need to dither everything.

I guess when you work inside a 32bit program (like Logic), any sort of automation or change in the files (I guess even playing the samples) causes the need for dithering.

I'm off to do something dithering!

edit: Thanks for the post Jon (Guess I was writing at the same time you were! And we share the same name....Weird ) Your analogy was great and really helped explain the reasons for dithering! Also, your suggestion to bounce at 24bit first, and then dither it to 16 is a great idea! Thanks!

4. ## Re: Dithering Audio: When Where and Why?

Well.. a lot of sequencers will dither automatically, so it's not something you have to worry about. Check to see if your sequencer does it, too.

5. ## Re: Dithering Audio: When Where and Why?

Originally Posted by zircon_st
Well.. a lot of sequencers will dither automatically, so it's not something you have to worry about. Check to see if your sequencer does it, too.
I avoid dithering from the sampler. I always end up doing some normalization (or at least verifying the levels) and cropping of silence before the end product, so I save the dithering for a final, post-sampler, step.

-JF

6. ## Re: Dithering Audio: When Where and Why?

Originally Posted by JonFairhurst
I avoid dithering from the sampler. I always end up doing some normalization (or at least verifying the levels) and cropping of silence before the end product, so I save the dithering for a final, post-sampler, step.

-JF
I noticed you used the word "normalization". So if i've mastered something and i'm burning it to disc, and my burning program (Nero) defaults to a ticked box that reads "normalize" I should untick that box because i've already dithered? Or does that not apply if Nero "notices" the content has been dithered? Or does it mean something else altogether?

Cheers,

7. ## Re: Dithering Audio: When Where and Why?

Originally Posted by zircon_st
Well.. a lot of sequencers will dither automatically, so it's not something you have to worry about. Check to see if your sequencer does it, too.
I would avoid any "automatic" dithering. Dithering/noise shaping should be intentional processes designed into the overall production. Otherwise, you'll end up reprocessing noise, and it takes very little of this to begin degrading and dulling the integrity of the signal. It's something a trained ear can hear immediately.

There was a big debate here about whether people could "hear" the difference in bit depths once upon a time. I was able to identify the mixes in question accurately--not because of the bit depth, but because I could hear the effects of the dithering.

In the most general terms, one can get away with sloppy practices easier on pop/groove-oriented material than with orchestral material. Anything pianissimo or sparse will tend to reveal dithering, re-dithering, etc. So you want to be very careful. The best dithering products also have noise shaping, which is a very good thing at the end of the chain, but a sometimes disastrous thing if reprocessed.

Waves has some good explanations in the literature for their various end-of-chain processors like L3, et. al. But Jon gave the basics a few posts up.

8. ## Re: Dithering Audio: When Where and Why?

The two analogies I like for dither are:

1. Kids peering over a fence at a baseball game. Some of them are taller than the fence, some are shorter, so you can't see them all to get a head count. But if they jump up and down you can get a fairly accurate picture.

2. A noise gate chattering really quickly.

The rule is pretty simple: stay at 24 bits as long as you can, but always dither when you go down to 16 bits. Whether to dither from 32-bit internal processing to 24 bits is something I have no opinion about.

And much as I hate to disagree with my man Bruce, I actually think dither could be automatic whenever you go from 24 bits to 16. The only thing that can't be automated is which type of dither you use. For example, Waves has two different types, one of which is less "aggressive" but doesn't cause problems if you're dithering more than once.

I also have to say that noise-shaped dither is freaking hard to hear. You have to crank the volume all the way up to hear it being toggled on and off, at least I do (and my hearing isn't gone yet). And even then it's quiet as hell.

9. ## Re: Dithering Audio: When Where and Why?

Originally Posted by Bruce A. Richardson
The best dithering products also have noise shaping, which is a very good thing at the end of the chain, but a sometimes disastrous thing if reprocessed.

Waves has some good explanations in the literature for their various end-of-chain processors like L3, et. al. But Jon gave the basics a few posts up.
Bruce, what's your take on the noise shaping within the L3? do you use anything other than normal shaping? and if so, in what context?

Cheers,

10. ## Re: Dithering Audio: When Where and Why?

I use the "normal" in case someone messes with the mix down the line. Almost all of my work gets mixed into other productions, and I have some clients that still want a redbook CD as delivery medium, so I don't get too aggressive with it.

When I can, I deliver a 24-bit product, which leaves the responsibility with someone else. The noise threshold is so low on a 24-bit mix that you're really not dealing with the same issues, even if it is necessarily dithered down from the 32-bit native resolution. 16-bit delivery is a whole other story. You can actually find a lot of content simmering down there next to the noise floor.

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•