Shed some light on dithering?

As I work towards the final steps and of a project, I was wondering if someone could shed some light on dithering... Or point me to something that I can read up on to get more info. I've tried doing some of my own research, but most of what I found was just confusing and overly technical...


I know that dithering is used to convert 24+bit to 16 bit, but that's about all I know. Not sure exactly what settings should be used in what situations. Thanks in advance

I was wondering if someone could shed some light on dithering... Or point me to something that I can read up on to get more info.

I like to use the POW-R's for dithering. 3 i use for rock/metal music and 2 and 1 i use for other genre's Even if you in 16 bit and you have a plugin effect that processes the sound in 24bit or even 32bit float, you need to dither.


Here are some other things i wrote on dithering - http://www.audio-mastering-mixing.com/FAQ___Q_A_Page_2.html#50

Cj

1. Read the iZotope guide twice
2. Then forget about it. Which dither algorithm you choose is almost irrelevant. Just remember the one cardinal rule: only apply dither during the final export, and only if you're reducing the bit depth to 16 bits for a CD. Dither is not necessary when encoding an MP3.

BTW, despite popular misuse of the word, "dither" does not refer to converting 24- or 32-bit audio to 16 bits. Its purpose is to mask noise caused by the unavoidable rounding errors that occur during such conversions.

vaultwit


So when I, say, purchase an mp3 on iTunes, is it dithered or is it 24 bit? If its not dithered, what happens when you burn this purchased mp3 onto a CD? If I buy a CD at the store containing the same song I purchased on iTunes, it IS dithered right?

We can't know off the bat if an MP3 from iTunes was dithered or not. The point though is that one need not dither MP3's; they use floating point when created. A CD should have dithered tracks on them being that they were bit-depth reduced to 16-bit format. Straight truncation to 16-bit is never a good idea especially with more processing. 

I need to know whether I need to dither my songs or not. I will be distributing my music online in the form of mp3s, so non-dithering may work, but what if people decide to burn it onto a cd? Then there will be a problem

Which is why I was wondering if songs bought from iTunes, amazon, rhapsody, spotify, etc have dithering or not. Since I'm sure many people by songs from there and burn them onto cd's

Transcending Music

vaultwit


I need to know whether I need to dither my songs or not. I will be distributing my music online in the form of mp3s, so non-dithering may work, but what if people decide to burn it onto a cd? Then there will be a problem

Which is why I was wondering if songs bought from iTunes, amazon, rhapsody, spotify, etc have dithering or not. Since I'm sure many people by songs from there and burn them onto cd's

Why is it a problem? It's just an MP3 on a CD

Well assuming they burn an audio cd, not an mp3/data cd. Audio cd's require 16-bit (and therefore dithering), no?

CJ, Your post is very interesting to me.  Either here or in some book somewhere, I read that I should always use POWR3 if I had the computer horsepower, so I was mindlessly following that.  My music is a bit of an odd mix of Broadway, country, and rock, but definitely not metal, so maybe you are advising me to go with POWR2.  If so, I will try that—although I doubt my ears will pick up the differences yours will.
 
 

One technical point - dithering doesn't "mask" quantization error, but rather decorrelates the errors from your signal.

And a very important point from the iZotope "Dithering with Ozone" PDF that goes against what many people who don't really understand the subtleties of noise shaped dither recommend:

Last, and not least, the psychoacoustic curves are designed to be effective at the lower thresholds of hearing. If you turn up the level to an unrealistic level to "better hear" the dither, you're undoing the whole point of psychoacoustic shaping. If you convert to 12 or 8 bits, you're also turning up the level of the dither signal into a range that it wasn't designed to be effective at. What this means is that psychoacoustic shaped dither has to be evaluated at normal listening levels -- take your mix, play it at a normal listening level so the full scale part of the mix is appropriate, and then evaluate the dither during quiet parts, fade-outs, etc. without turning up the level. We recognize that it is difficult to hear at this level -- which is the whole point.

My head hurts.

Hey, if this stuff was easy everybody could do it! Plus we'd have very little to talk about.

Page 936 in the Sonar X2 Reference Guide describes all of the dithering types in Sonar.