You should always dither as the last process when reducing the wordlength of an audio dataset, e.g. when generating a 16-bit CD image from a 24-bit stereo master. Always mix at the highest wordlength available (24-bit, 32-bit, 32-bit FP etc), and never dither before the final 2-track reduction to a lower-wordlength format. Note that dithering is completely independent of sampling rate.

There many different dithering algorithms to choose from, and ideally you should try several of them on your material. What you would be looking for is one that does not introduce any audible artifacts into your mix, and this will vary according to the amplitude and frequency distribution in your source. [[url=http://[/URL]="http://www.users.qw…"]Here's[/]="http://www.users.qw…"]Here's[/] an article you may find helpful. There are also numerous explanations of dithering on the Wiki and other sites, although these are not constrained to audio as it is often easier for them to demonstrate effects optically rather than sonically.

Dithering ads different colors. Essentially, it is a set of randomized numbers, frequently audibly resembling white, pink, brown noise. This added noise is randomly created to superimpose noise on top of extremely low-level signals that would be lost in a land of too few samples for anything other than horrible crunchy sound. In a sense it's sort of an addition of analog tape noise to create continuity in the bitstream making it contiguous without any holes creating horrific sounding audio on low-level sound sources. So sonic items such as reverbs will trail out smoothly instead of crunching into oblivion, audibly. And the different types of dithering will make these reverb trails all sound slightly different by how they interact with the decaying signal.

A distorted viewpoint with some clarity
Mx. Remy Ann David

As Remy indicated, the place to listen for differences in the algorithm is in "silent" passages like reverb tails. I've never been able to tell the differences between the algorithms (or in most cases detect that the algorithm is being used) in a dense mix. As said above, the on line articles explain the basic concept well. Bob Katz's mastering book has a good chapter on the different algorithms.

Thanks for the help!

Follow up questions:

I get the explanation from Boswell's article regarding how the rectangular and triangular algorhythms work. Any quick info on how the pow-r ones work?

What are the basics I need to know about the Nyquist frequency?

Thanks!

I don't know what you need by way of "basics". In this sort of work, the Nyquist frequency is half the sampling rate.

Try [[url=http://[/URL]="http://www.playgrou…"]this[/]="http://www.playgrou…"]this[/] for more information on the various proprietary POW-R dither algorithms. There is little information around on the actual mechanics of the method.

I'll put this into my version of Lamens terms.

Many debate as to whether our ears can actually hear and identify when and where dither is used. This is all subject and relative of course, but from my experience and that of some of my mentors, 99% of the time, dither isn't audible. However, it is necessary in the situation of fading out an audio file. For example, occasionally when fading out audio that has a plug-in on it, you can physically hear the moment when the effect cuts out before the audio file actually ends. This is because digital effects use step-like, jagged algorithms based upon the given digital signal, which can create a margin of error for computers. Dither smooths things out and "fills in the blanks" in these margin of errors so the effect of the plug-in can be heard until the end of the file.

Someone feel free correct me if I've been misinformed.

Another analogy to dither would be like the difference between, in the analog days, utilizing Scotch, Ampex or Agfa recording tape. I mostly preferred utilizing Scotch because I like its sound better than the Ampex or Agfa equivalents. So again, totally subjective. And I think the smell of the Ampex tape did not excite my olfactory. So that's a subjective decision based only upon its smell. But basically I preferred the clarity of Scotch, blended or single malt.

I don't do bourbon
Mx. Remy Ann David