Digital BW, The Print

George,

it's great to "see" you here. Thanks for joining our inaugural discussions. You 
raised a whole bunch of hot issues each of which calls for a new thread. 
I'll start with the scan workflow. You wrote:

>>>
Scanning protocols are also problematical. Silverfast is the only 
third pary scanning software that allows you to make your own 
LUT's. I've fooled around and come up with N+1, N-1 and N+2 
LUT's, and they work well - in 8 bit, of course. Both Mike Kravit 
and I scan in 16, drop down into 8 bit Silverfast HDR and 
optimize the image before it even gets into Photoshop. What do 
you guys do at the scanning stage? What works consistenly?<<<

While I see the point of scanning in 16bit, I don't see that as necessary a 
priori. It seems to me a matter of whether or not you have enough control in 
the scanner interface to let the scanner  spit out an 8bit file where the tones 
are laid out as best they can  without local controls. The Crosfield at BowHaus 
in Los Angeles, only outputs 8bits, for example, but they are "great" 8bits, 
because of how it's driven. 

With less capable scanners and drivers, the 16bit is a workaround t o allow 
Photoshop to do the heavy lifting. That's all I see in it for now. All this may 
change if Adobe implements 16bits across everything.

However, considering that the upcoming release of PiezoBW software will 
allow for 16bit printing on the desktop, there will be the dilemma: If you didn't 
scan it in 16bit and _stayed_ in 16bit, what's the use? Which means that the 
brave printer must scan in 16, do all the other maneuvers needed to fake 
layers and masks and produce a printing file without ever dropping to 8. For 
some who need only one overall move, this may be the ideal and highest 
possible quality. I suspect, for most, it will not be used untill Photoshop 
embraces 16 more fully.

Here now is a workflow idea (8 or 16bits)  that I would like to post for feedback 
and comments:  

-The scan of the negative or other monochrome original is done in RGB (8 or 
16) and  saved out of the scanner software with the appropriate profile. 

- The file is opened in photoshop for tonal adjustments while still in RGB. But 
before starting, we check our color settings (this assumes PS 6). 
----Under gray we have created a curve that matches the paper/ink/printer we 
intend to use. That means that a _grayscale file_ on screen will look identical 
to a print made under those settings. 
----Under RGB we set the same color space that the file came from, so that the 
intentions of the scanner and our view in PS match. The only real critical 
number here is the gamma.

- We now do global and local adjustments as needed while still in RGB mode. 
The critical step in this workflow is to stay in RGB. At this point we have the 
option to add a channel mixer set to mono as the top layer (if in 8bits) which 
does 2 things: Helps us see a consistent gray-color image, and allows us to 
pick whatever mix of scanner channels looks best (more important for CCD 
users).

- After we are happy with the look of the image, we make a dupe of the file, 
flatten (if it was 8bits) and convert to grayscale. We notice that the file looks 
exactly the same in grayscale as it did in RGB. What in fact has happened is 
that Photoshop has internally "adapted" the RGB file to the Gray settings we 
previously set, which compensated for our paper/ink/printer. It is like putting it 
through a custom profile made for our specific output.

- We now take that grayscale and go to piezo and print with no further 
adjustments. I bet that histogram will look pretty damn good now.

We have gained several things from this simultaneous drop to gray and match 
the output device.

-We have a master file in RGB from which subsequent gray files can be made 
for different papers just as one would for color. Alternatively,  a neutral 
conversion can be produced (for other uses), based on a gray gamma that 
simply matches the RGB gamma of the file.
- We have used all the information our scanner is capable of by utilizing all its 
channels with whatever benefits they each may have in noise, sharpness etc.
- We have done no moves in the fragile 8bit gray mode. We stand a chance at 
a better histogram depending on whether we move all 3 channels at once or 
separately.

That's the general idea: work in RGB, drop to a pre-calibrated gray when done 
and print with no other tweaks.

I hope we can compare notes on the above.

Antonis Ricos