> > Ah, but the 12-bit mode uses a perceptually-coded nonlinear compression; > > its dynamic range is still pretty much 96 dB. > > Adam (or anybody else), do you have any more information on how this works? It's in the DV spec (i.e., the Blue Book). Low levels are encoded 1:1, with higher levels being compressed a bit. The spec shows the compression as a table lookup, with re-expansion on playback. With all due respect to those saying "12 bits is 12 bits and you can't get any more dynamic range", human hearing is logarithmic in nature, and employing this sort of quasi-logarithmic compression is a valid and useful way to squeeze more dynamic range into a given number of bits than linear encoding allows. At loud volumes it takes a greater change in level (i.e., more linear bits) for the level change to be detected than at low volumes; what the 12-bit nonlinear mode sacrifices is the fineness or granularity of change at high levels -- just the place where the ear is insensetive to fine changes. True, it is NOT as perfect or clean as 16-bit uncompressed, but you are unlikely to hear the difference between them as far as quantization error goes! The difference between 32kHz 12-bit and 48kHz 16-bit modes is much more in the frequency-reproduction area than it is in the dynamic range area as far as psychophysics is concerned. Cheers, Adam Wilt
The active discussion on using color filters is proving rather inaccurate in
places! Early monochrome film as used in silent movies was Orthochromatic
(rather than Panchromatic) and had a restricted response to colors towards
the reds. This is why deep red objects (like made-up lips) came out looking
dark. To get a similar response you should shoot through a 'minus' red
filter which would be cyan.
Since DV is a component system, you already have a very effective access to
a 'minus' red filter in post so I would agree that this is probably a more
controllable way to go if you can afford the processing time.