Digital BW, The Print

Michael,

> 1.) almost all high-end scanners today scan INTERNALLY at 10, 12
> or 14-bits (only
> the cheapest consumer scannersnow  scan at 8 bits internally.
> Almost NO scanners
> actually scan at 16 bits))

The new Nikon in fact does have a 16 bit A/D, but the sensor/AFE (Analog
Front End) simply won't provide 16 *clean* bits...so yes, it does scan at 16
bits, but they aren't 16 *good* bits.

> 2.) assuming that the SCAN curve is optimal for the negarive in
> question (i.e. the
> instructions to the scanner, including white point, black point,
> and shape of curve)
> there will be ABSOLUTELY NO DIFFERENCE between an optimal "8 bit"
> scan and a raw
> "16 bit" scan that has been manipulated to equal the "look"
> obtained by the above
> optimized curve -- regardless of the number of bits.
>
> It can best be diagrammed as such:
>
> scanner (12bit) + optimal curve --> optimal file (8bit) --> printer (8bit)
> scanner (12bit) --> expanded file (16bit) --> manipulate (16bit)
> --> optimal file
> (any bit) --> printer (8bit)

Except that not all scanner software has a decent setpoint/curve tool, so
scanning in high bit mode, and setting the setpoints and making the tonal
adjustments in PS is a better workflow for that circumstance.  My scanner
software happens to have a great setpoint, curve tool, and histogram...so I
do the do all that in the scanner driver and get 8 bit data.

> All things being equal, the above two workflows will produce the
> EXACT SAME IMAGE
> at the printer (i.e. there will no perceptible difference at all)...

Of course, because the "optimal curve" step you outline above is being done
to the high bit data in the scanner.

> ...It is NOT TRUE AT
> ALL that if you
> perform extreme manipulations in 8 bits, then you are destined
> for posterization and
> enhanced noise.

If you are talking about B&W, it may be true that you aren't "destined for
posterization, but any substantial (and sometimes not so substantial) tonal
manipulations done to 8 bit grayscale data *can* certainly result in
posterization...and certainly sometimes quite readily.  It very much depends
on where along the curve these manipulations take place, and the image.

> - As others have noted, printing in 16 bits is a total waste.

Well, that is debatable...for B&W that is.  It's true that the human eye can
only discern less than 200 graytones given any singular lighting
situation...but...in order to not distinguish "adjacent" tones (which is
what you want to do to get smooth gradations, with no two tones immediately
perceptible next to each other), you can introduce intermediate tones...and
it improves the smoothness of the gradation.  As usual, this is very image
dependant.  So, having a few extra bits, say 12 bit grayscale data, or at
least the ability of the dithering process in the print driver to
accommodate the transitions, may give *some* benefit in *some*
circumstances.

Regards,

Austin