Digital BW, The Print

Paul Roark wrote:
> Ernst,
> 
> I think what you're saying relates to the difference between what I've heard
> referred to (by Roy, I think) as "hardware" v. "software" linearization.  In
> a properly written rip or driver, the different shades are generated by
> different dot or dither patterns.  As such the file values are not altered
> and the number of different levels of gray is not reduced.  However, if the
> adjustments to get to the final ramp are made prior to the generation of the
> dot pattern, the file values that are sent to that stage are altered and, if
> all file values were being used, some grayscale steps and information are
> lost.
> 
> The loss of information is minimized if the curve shape is mild, and, of
> course, high bit depth minimizes the problem.
> 
> However, to the extent possible, to preserve as much information as
> possible, we'd like as much of the work done with the dot/dither pattern as
> possible.  For example, I recommend in the "EZ" approaches that the driver
> controls first be use to achieve the best ramp.  Only when that is not quite
> right do I recommend a curve, transfer function, or maybe ICC (if the ramp
> toe problems can be fixed).  I think Roy recommends that the QTR curves be
> as accurate as possible before one relies on an ICC linearization step.
> 
> Paul
> www.PaulRoark.com 

Paul,

The Colorbase discussion, Jon Cone's somewhat stealthy replies
on the K7 curves, your modeling to the eye of the RGB curves
in Photoshop and Steve's questions brought me back to how I
got quads printed with the Wasatch RIP.

At that time I used linearisation of the CMYK channels as the
base and with extra adjustment curves I fitted the partioning
to get a final result that would be near linear. I could check
but not apply that end linearity with the SpectroCam again. On
top of that a "perceptual" S curve was placed. This was a
complicated approach, partly defined by the limitation of that
RIP and my lack of knowledge. I asked Wasatch for a top
linearisation for B&W purposes but the answers were not
positive. This straightening, bending, straightening, bending
of the data is the worst example of what could happen in a
quad RIP. My philosophy was that I could at least linearise
the CMYK channels from time to time and keep the card house
above that stable.

Your RGB curves for Photoshop lacked the linearisation and
fixed the ink channel choice much more than my approach but on
the other hand you were able to do (a mind bending) one curve
step to get to the perceptually nice greyscale. It would
depend on the Epson's driver Dmax and the black box RGB-CMYK
conversion of the driver but nevertheless. I made some printer
profiles with a profile editor based on your curves.

Roy's QTR partitioning curve creation and linearisation over
the combined partitions were a good starting point. The
addition of the perceptual curve and the use of the ICC model
for that in printing, in the greyscale space and for soft
proofing where more or less brainstormed in this list.
Possibly inspired by Jon Cone's quad profiles of that time but
quite similar to how RIPs handle this for color printing. In
fact the greyscale space profile is most likely the most
innovative part of that development.  (I know that gamma 2.2
will do as well)  After that the profile creator was made.

The Colorbase discussion on whether it is a linearisation or a
calibration utility did question the underlying linear base
where the Epson driver builds on. My curiosity wasn't so much
directed to how profiling fits on the
calibration/linearisation like Steve now looks for (related to
his using of QTR's profiling tools on the Advanced B&W part of
Epson's new driver) but whether Epson uses calibration that
isn't linear but fits the native character of the printer
better and at the same time follows Steve's suggestions on how
the rest should fit on top. If this all fits like ying yang
you need milder shifts by the created profiles and by that the
total gets more stable and smoother. This concept is abstract
and I have no numbers to show. Roy is quite capable to push
the pin in the balloons I make so I will not write more on
this than I have done already.

The strange thing with the Wasatch SoftRip was that the curves
after linearisation shouldn't be seen as representing the
actual non linearity of the printer but should be seen as
special correction curves of that non linearity. The manual
said so. If that implied the creation of a non linear base to
build the rest on is something I'm now asking myself again.

                    --
           Ernst Dinkla


www.pigment-print.com
(         unvollendet         )