filmscanners: Back to the basics

[Alessandro Pardi <alessandro.pardi@inferentia.it>]

After the thread about 16 bits vs. 8 bits, I started thinking about another
golden rule (a few nights ago, victim of jet-lag :) The rule states that
it's much better to make the tweakings at scan phase, and to perform the
least possible afterwards. This is perfectly reasonable for my LS-30, which
internally works with 10 bits but only outputs 8 (forget Vuescan for the
sake of this argument): if you work with curves you simply select the best 8
bits out of those 10. But for the majority of scanners, which I suppose are
not as reluctant to deliver what they got from the film, I wonder, when I
apply a curve at scan phase, whether this ends in a software/firmware
post-processing of the data or it somehow changes some hardware parameter.
If the former is true, why should it be better than Photoshop? I admit I'm
completely clueless about scanners internals, but I think that you can't
change much besides exposure, when it comes to hardware: you may perhaps
have different exposures for the 3 colors, and thus change color balance,
but I can't see how to implement curves modifications.

Alex Pardi