At 09:36 AM 9/17/00 -0400, you wrote:
> But the fact is, there is really no such thing as linear...especially in
>A/D converters, but no need to go there in this discussion.
>
A/D converters can be as linear as you want to take the trouble to make them.
>> The ratio of DMin:DMax is constrained by this, giving a maximum value of
>> OD of which the scanner is theoretically capable which is determined by
>> the bit depth. Otherwise we'd all be perfectly happy with 8 bit scanners,
>> possibly even less.
>
Using 8 bits does not necessarily give you more density range. It is
possible to do scaling before the A/D so that the DMin:DMax ratio of the
film you are scanning is scaled to the A/D range, whether it is 8 bits or
more. More bits gives you more resolution of precisely what the density or
color is. In other words, it introduces less error in the quantizing of
the value, therefore less noise, per the following.
>With 8 bits, you only get about 60-70db SNR (signal to noise ratio), and
>using a higher number of bits gives you a better SNR. As I'm sure you
>know, A/D converter systems don't give you ALL the bits perfectly, the LSB
>is always off, and possibly the two or more LSBs just aren't accurate, so
>to get truly 8 usable bits, you would need more than 8 bits...and you drop
>n LSBs to get 8 (or use 14 to get 12...etc.).
>
I don't know about scanner A/D converters, but I know that with these
converters in general, it is not true that the LSB is always off. It can
easily be located between the bits on either side, which is what is
important in scanning. I believe that since you are likely to apply gamma
correction anyway, being exactly linear is not as important as making sure
each step from level to level is roughly constant.
>> Drum scanners and CCD scanners which use analogue shaping of the voltage
>> to a non-linear gamma function upstream of the ADC aren't constrained
>like this, as much wider ratios can be compressed to fit within the bit
depth.
Exactly!
>
>This can be done in the digital domain (after the ADC). I can't imagine
>why anyone would want to do that in the analog domain. Who does that?
It is too late to apply these corrections if you wait until you are in the
digital domain. At that point you are limited to the compromises in the
scaling and digitizing that have already been made. To scale the film
range to the A/D converter range, you need to do this analog. Otherwise
you are either throwing away data from the film (clipping) or losing S/N
because you did not use the full range of the A/D.
>
>====================================================================
>The filmscanners mailing list is hosted by http://www.halftone.co.uk
>To resign, <mailto:[EMAIL PROTECTED]> with UNSUBSCRIBE
FILMSCANNERS in the title, or UNSUBSCRIBE FILMSCANNERS_DIGEST if you are
reading the Digest.
>
====================================================================
The filmscanners mailing list is hosted by http://www.halftone.co.uk
To resign, <mailto:[EMAIL PROTECTED]> with UNSUBSCRIBE FILMSCANNERS in the
title, or UNSUBSCRIBE FILMSCANNERS_DIGEST if you are reading the Digest.
