Here's some stuff that I just compiled.
Dither
Dithering is the answer to low-level quantization error. An uncorrelated white noise signal with an amplitude slightly less than one quantization level is added to the original signal before sampling. The noise signal is called the dither.
http://www-personal.engin.umich.edu/~jglettle/section/digita laudio/intro.html
Less than one quantization level, not sure what that means. It would imply that it's added below the 16th bit, if they are calling one quantization level as losing the 8 bits. Still confused.
Dither
Dither is the process of adding low-level analog noise to a signal, to randomize or "confuse" the quantizer's small-signal behavior. Dither specifically aims to address two problems in quantization. The first of which is that a reverberating, decaying signal can fall below the lower limit of the system resolution. That is to say that an attempt to encode a signal below the LSB results in nothing getting encoded. Clearly,
information is lost. The second, as discussed in the previous section, is that system distortion increases as a percent of a decreasing input signal. It is important to note
that not only does dither remove some quantization error from the signal, it effectively removes it.
http://www.tc.umn.edu/~erick205/Papers/paper.html#ditherIt's been my understanding that dither spreads the quantization errors out, so that they become less audible, not remove them completely.
Dither
Literally, dither is noise added intentionally to a digital recording. Low level signals are difficult for digital gear to record; the sampling machine simply has difficulty deciding whether the necessary bits should be turned on or off, creating "quantization noise." By adding a small amount of very controlled noise to the original signal, the bits can be made to positively switch on or off, improving low level sound resolution.
http://www.mus1171.neu.edu/glossary/TERMS/DITHER.HTMThis is the closest that I could find to dither being added to more than one bit, but than they say that it allows their system to encode amplitudes smaller than the least significan bit, there's that LSB talk again that has led to my confusion, they reference Pohlman at the bottom.
Dither
A well-design digitization system must be able to suppress
any audible effects of its quantization error. One could
increase the number of bits in the quantizing word, which
would result in a decrease in error amplitude of 6 dB per
each additional bit. However, this is a rather uneconomical
solution, as many bits are necessary to reduce effectively
the audibility of the quantization error. Additionally, the
error will always be significant with low-level signals,
which are quite common. Dither seems to offer a more
elegant solution. Dither is a small amount of noise added to
the audio signal before sampling. This causes the audio
signal to shift with respect to quantization levels.
Quantization error is thus decorelated from the signal and
the effects of the quantization error become negligible.
Dither does not prevent the quantization error; instead, it
allows the system to encode amplitudes smaller than the
least significant bit.
www.africandl.org/research/audio_digitization.pdfBit of a conflict from above where they say that it removes quantization errors.
Here's one from Berklee School of Music that I ran across where he's saying to not dither to 16 bit after the signal is dithered at 24 bit and it remains at 24 after the mastering plug-in.
The Last Step
To finish, you need to bring your mix back down to 16-bit format to burn it to an audio CD through the process called dithering. If you were to convert it back to 16 bit without dithering, you would simply be truncating the lowest eight bits and introducing quantization errors. Dithering corrects for these errors before converting from 24 to 16 bit, allowing the song to retain the sound quality you've been hearing all along. Your maximizer plug-in most likely does the dithering, though the file itself
remains at 24 bit after processing. When you save it from your audio editor, do so in 16-bit format without applying any further dither. After the maximizing and dithering steps, you're done. Nothing further should be done to the file other than burning it to audio CD. If you need to make changes go back to your original unmastered file, call up the plug-ins and presets you've saved, tweak and reprocess.
http://www.berklee.edu/bt/162/mastering.htmlSo far I haven't been able to locate any information that specifically mentions adding noise at the 16th bit, the other hits all say pretty much the same thing, very low random noise, but no mention of level or where it's applied. The hits went from audio dither to graphics dither pretty fast. There were some conflicting comments even on the edu sites. No surprise that there is so much conflicting information on the recording newsgroups.
Here's one where Benchmark Media Systems say that their dither is a whopping 14dB better than TPDF 16 bit. I've been using TPDF, I think Brad mentioned that he uses TPDF and thought that you guys might like to see it.
http://www.benchmarkmedia.com/appnotes-d/wrl.htmlThe reason why you probably thought that I read Pohlman, DC is because I quoted some of his information that I pulled off the net, on a discussion that we had about bit depth a long while back. I'll definitely read his full book though, it's way past time that I did.