Digital BW, The Print

Hi Kevin,

> > That is true of the consumer scanners you and I use.  There are some
> > scanners, and I've designed one, that have analog gain control between
> > the
> > CCD and the A/D, and has, basically, hardware setpoints.
>
>
> That would be add a whole new meaning to the auto-exposure function!!

I am surprised none of the current "high end" consumer scanners do it...it
just isn't that hard, or costly.

> >> Cool, thanks for the primer.  Wouldn't the software use some algorithm
> >> to interpolate values for the gaps instead of leaving them empty?
> >
> > That is a good guess, but no...remember, every pixel has a value
> > associated
> > with it.  What are you going to interpolate?  If you did somehow
> > interpolate, you are giving data that was the same, different values,
> > or
> > creating NEW data that didn't exist before...  The gaps really don't
> > mean
> > anything, as you are taking the N bit data and converting it to
> > CONTINUOUS 8
> > bit data...in other words, when you go to print, you DO have every data
> > value from 0-255 (or most every one).

BTW, did you understand that?  It isn't an easy thing to understand...so if
you didn't, I can try to give a usable example.

> Just let me make sure I have understood what you are saying and that we
> do in fact agree.  In the same hypothetical the N-1 development
> (density range of .7), the N development (density range of 1) and the
> N+1 development( density range of 1.5) all produce virtually the same
> scan after set points are applied.

Hum.  I don't know what you mean by "same".  The data for the .7 density
range will be "narrower", meaning, say values from 100-240, vs 1, which
would give you 100-440 vs 1.5, which would give you values from 100-600 (I
am making up the numbers, they just illustrate that you get more different
data values, as in a wider range of data values, the more density you are
scanning.

> With an extremely shallow exposure/density curve you may have too few
> discernible points from the scanner to fill the 256 data points
> (assuming 8bit).

Correct!

> As a principle more data points (ie maximum density
> range) is better.

Yes.

> Except in situations with an extreme brightness
> range adding development time to your 'normal' time (or the
> manufacturers recommended time if you haven't done any testing at all)
> will increase the density range. And be a good thing.

Yes, but what's the "except" for?  This statement, and the previous one ("As
a principle...") are the same, are they not?

Regards,

Austin