[Jason]

In article , says...


Oh, go to 32 bit floats and you have enormous room to move.


There's a parallel here with the emerging HDR ("High Dynamic Range")
capabilities in pixel editors like Photoshop. They convert the 8-, 12-
or 16-bit/primary samples into the floating-point realm to achieve a
vastly larger "sample space."

I agree the OP's analogies are mostly metaphorical, but there is another
parallel with image editing: the order in which transformations are
applied to the data. A cardinal rule processing photos is to perform any
required sharpening as the final step before printing. Doing it before,
say, adjusting exposure, contrast, color balance, etc. will lead to
weird visual results because of the "damage" shapening does to the data
by increasing the contrast around edges in the image. I think an
analagous situation in the audio domain might be whether to apply EQ
before or after noise reduction. I'm sure there are many other decisions
like that.

Jason

--
reverse my name in email address