Digital BW, The Print

--- In DigitalBlackandWhiteThePrint@yahoogroups.com, "mike_nunan"
<mike_nunan@h...> wrote:

snip

> I would expect at least 16-bit precision in all internal calculations
> and data tables in the driver, probably 32-bit in fact. If you're
> talking about a RIP such as IP which works directly from the TIF
> file, then by using high-bit files you should also get 16-bit
> precision for the input values.

It still depends. THat may be the case for IP, but though it and some
other drivers/RIPs can accept 16 bit input, whether or not it's
actually utilized remains uncertain. I've seen very few definitive
statements from any of these people that they actually do. There have
been indications or hints that these drivers/RIPs do some manipulation
of input data to their needs. Certainly some kind of ppi remapping is
done, and bit downsampling could be happening as well.

> I'm more suspicious regarding the
> precision used to pass values from PS via the OS to the Epson driver,
> especially since Adobe have been so tardy about implementing high-bit
> support even in the places where it's immediately obvious.

It has more to do with the Epson driver than Adobe, and may be OS
related as you suggest (but I know little about that). Some time ago
Photoshop let you hit the print command with files that were high bit
(it didn't used to) but all it was doing was changing to 8 bit on the
fly as a convenience since that's what the driver accepts.

> It's this
> lack of precision that could make it impossible to compensate for non-
> linearity upstream of the driver, whether by curves or profiling,
> without introducing banding. I appreciate that the grey balance is an
> issue too, but I'm taking the long view that it *can* be fixed via
> profiling provided you have enough precision to tweak the values
> accurately enough. I do realise that this is non-trivial to do (an
> understatement!) but the other problem is worse still.

This all gets into a huge issue about where the best place for massive
correction is best done. You can send 8 bit pure color gradients (not
neutral) through the epson driver with no profile conversion and see
good gradients on paper. In my experience most color banding comes
from converting to profiles, and doing so in high bit did not help.
C.D.Tobie once explained it well to me but I've lost the email. It has
to do with large remapping of lab values, and best guesses. When the
driver is so out of balance with massive crossovers etc., remapping
color based on LUTs becomes problematic. You can wind up jerking close
colors in multiple directions. So it still makes sense to have the
driver as linear as possible in the first place, then profiles can do
their jobs somewhat coherently. Others with more day to day experience
with profiling in different environments/drivers/RIPs etc. would have
more valuable opinions about how much practical difference it makes.

> I think I'll wait to see what becomes of GimpPrint then. I'm
> reluctant to spend big money on a commercial RIP at this stage, and
> I'm not in a rush.

The fact is, despite this somewhat techie discussion, people make
beautiful prints every day within these limitations. Profiles are like
darkroom materials, you get the best one you can, learn what it can
and can't do, and make it work for you. Softproofing makes this much
easier. But it's like type C paper, neither Kodak nor Fuji was "right"
but people decided which one they liked and made great prints based on
it's strengths and weaknesses. Sometimes I think the high degree of
technical discussion around digital raises our expectations too high,
we think technical perfection may be attainable, but it ain't so.
Which brings us back to art, thankfully.
Tyler