QTR-Quadtone RIP

Hi Leping,

You've got a lot of inter-related issues going on and I think they tend 
to
confuse the situation.

In QTR there are two similar but different features.   First, there's 
the
Linearization in the QTR profiles and second, there's the ICC profiling
with either generic QTR Gray Matte/Photo Paper profiles or custom
Create-ICC profiles.

On Thursday, May 3, 2007, at 01:02  PM, Leping Zha wrote:

> Again, I found I am answering some of my own
> questions posted to this list.
>
> 1. Convert or Not Convert?
>
> In Mac OS X QTR Printing Tutorial ("Tutorial.pdf"),
> the instruction is clear that we should let Photoshop
> manage the color and convert the source image profile
> to either "QTR Gray Matte Paper" or "QTR Gray Photo
> Paper", Roy's standard QTR device profiles.  There is
> also a separate instruction sheet explaining the
> switching from using no such conversion ("No Color
> Management" or "Same As Source" in Photoshop) to
> using the standard profiles, a change made not long
> ago, to address the different visual characteristics
> of matte and photo paper, and for other purposes.
>
> The QTR profiles come with a Gamma close to 2.5, so

QTR profiles don't use a gamma function which is a pure exponential.
QTR linearization uses the Lab L* function.   The idea being to
make a straight-line in L measurements.  The line always goes
from dMin to dMax and the goal is a simple straight-line.
Gamma functions are a different shape.

> unless you work in the same color space in your
> source images the numbers will change during the
> conversion process, stretching the lower midtone and
> shadows little bit and compress the highlight
> slightly when the source image profile has a Gamma
> of 2.2, much more so when it is 1.8.  You can easily
> watch such changes in your histograms.
>
> If such conversion is an intrinsic necessary step in
> the QTR printing workflow now, we would be logically
> thinking that all the linearizations needs to be done
> also with the same conversion in place.
>
> But, on top of Page two, Roy instructed us not to do
> so when linearize: "Print targets with QTR and all
> the selections you want to profile or linearize. The
> intended resolution is very important. The targets
> are Untagged and should NOT be color managed in any
> way. Do not Convert an image. Always print with No
> Color Management or Same As Source."

This is specifically for printing targets that will be used for
linearization or ICC profiles.  You need to get "raw" density
measurements so that a curve can be fitted to correct/linearize
the output.

>
> After letting targets to dry for 48 hours, my
> Eye-One measurements last night confirmed that,
> the Roy's default QTR curves, at least for Epson
> 3800 for the papers I am dealing, were linearized
> with no conversions.  When we print either the 21-
> or the 51-step target with the default curves and
> no conversions to the Gray Matte/Photo Paper
> profile, the resulting linearity were nearly ideal
> (ruler straight lines), while when the conversion is
> inserted to the workflow, the resulting linearization
> is off with reduced shadow separations (the lines
> bend down at lower left, the high density region).
>
> So, the conclusion is, we should NOT perform the
> conversion described in the Tutorial, when the
> source image Gamma is 2.2 and the default QTR curves
> are used.  If you work on Gamma 1.8 you should first
> convert the images to Gamma 2.2.  Thus, I believe the
> Tutorial needs to be updated.

It sounds like what you are doing is measuring a stepwedge for
Lab linearity.  Since Lab linearity is what the basic QTR profile
goal is then naturally you'll get the best linearity with just the
basic QTR profile and No Color Management conversions.

Color Management and ICC profiles however have a different
purpose and goal.  It's much harder to measure the precise result
because the goal is matching of our perception not a specific
linearity.

When you print on matte paper there's a mapping of dMin's and dMax's
from the file to the paper.  In general you lose more at the shadow/dMax
end.  So with straight QTR profiles this will usually mean a lighter 
print
because the straight-line pushes the whole range lighter.  The idea
of Color Management is to do a mapping that better (but not perfectly)
matches our perception.  This sacrifices some of the shadow separation
in favor of midrange density.

In addition to this issue you also have to consider the source embedded
profile (or working space for untagged files).  Say you have a stepwedge
and look at K=95, it's really the corresponding L value that determines
how dark it is.  So for Gamma 2.2 there is very little separation 
between
K=100 and K=95.  Look at the eye-dropper in Photoshop, the difference
is only L=0 to L=1 -- much less than the separation between any other
steps.

The bottom line with all color management issues is how good the print
and the screen match with our perception not measurements with a
densitometer.

Roy

>
> Then the question remains: what happens if we keep
> the conversion in the printing process, but to build
> costume QTR curves to fit the workflow?  I did more
> experiments along this line, and found it not a good
> idea.
>
> First, for the Hahnemuhle Photo Rag 308 paper I am
> mainly interested in with the Epson UC K3 matte ink,
> and with the QTR Gray Matte Photo profile, it is
> impossible to linearize in the K=95-100 region.
> Looking through the Mac OS X's ColorSync utility,
> the QTR Gray LAB and the QTR Gray Photo Paper
> profiles contain gray transfer curves that start
> from the origin (X=Y=0) while the QTR Gray Matte
> Paper profile has a small intercept (X=0, Y=2 or 3).
>
> With the conversion to the QTR Gray Matte profile,
> this small gap in the transfer curve compresses all
> the K values below 2 or 3, or RGB values below
> around 5 or 6, to 0.  This is again easily seen with
> the Photoshop histograms.  Such crushing of deep
> shadow values may help some matte QTR print's looking,
> but analytically makes linearization in the dark
> shadow region impossible.  The pre-linearization
> readings at K=95 and K=100 are always identical with
> the curve (naturally), and as a matter of fact the
> pre-linearization curves are almost vertical in the
> K=85-100 region, making the inversion of the curves
> difficult or impossible.
>
> The real world results with the small intercept in
> the profile is (A) the resulting linearization is
> not mathematically stable, (B) the QTR curve making
> program often fail in making the curves (.quad files)
> from the curve description text source with the
> measured linearization values in place, for a stated
> reason in the error message that the linearization
> values do not decrease monotonously (in the dark
> shadow region), and (C) even when the linearized
> curves are successfully made, the test results with
> these curves are not good - still no tonal separation
> between K=95 and K=100, since the QTR Gray Matte
> Paper curve always brings them (nearly)together!
>
> Actually when the (B) happens, the linearization
> data is monotonically decreasing, but obviously the
> differentials in the deep dark region are too small
> so some kind of fitting routine failed inside the
> program.  Also, when the conversion is in the
> workflow, the resulting linearized curves produce
> higher density at K=50 (around 0.61 against the
> 0.58 with the linearized curves without the
> conversion, which is closer to the ideal value of
> 0.568), indicating Gamma value errors.
>
> This re-confirmed that, the QTR curves should be
> always used WITHOUT converting the source image to
> the QTR's GRAY paper profiles, just keeping the
> source image Gamma at 2.2 and print in Photoshop
> with "No Color Management" to QTR driver, when
> printing images, AND when making new curves, AND
> when performing linearization.
>
> 2. Ink Limits, Black Boost, and D-Max
>
> I also explored heavily on the effects of the QTR
> curve's K Ink Limit and Black Boost settings.
>
> The general result is that, we can gain a bit
> D-Max from raising the K Ink Limit of the default
> curves by 10 (45 to 55, 50 to 60, and 55 to 65)
> and maximizing the Black Boost values to 100.
>
> As Roy pointed out, the D-Max actually decreases,
> when the ink limits are set too high, such as
> typically 75 and higher for K.  Also, the overlap
> of the inks does not raise the D-Max but only mess
> up other things.
>
> Here are the typical D-Max values I am getting from
> the Eye-One measurements (the "Warm" curves tend to
> produce the highest D-Max, since it uses no color
> inks).  "Boost" means both increasing the K Ink Limit
> by 10 from the original values, and setting the
> Black Boost to 100.
>
>        Paper         Ink   Driver    Boost     D-Max
> ------------------  -----  ------  ----------  -----
> Hah PhotoRag 308    Matte   ABW        -       1.59
>                             QTR        NO      1.55
>                             QTR        YES     1.61
>
> Museo Silver Rag    Gloss   ABW        -       2.21
>                             QTR        NO      1.94
>                             QTR        YES     2.47
>
> Hah FineArt Pearl   Gloss   ABW        -       2.04
>                             QTR        NO      1.92
>                             QTR        YES     2.15
>
> Ilford Smooth Pearl Gloss   ABW        -       2.23
>                             QTR        NO      2.05
>                             QTR        YES     2.44
>
> Epson PLPP          Gloss   ABW        -       2.25
>                             QTR        NO      2.08
>                             QTR        YES     2.47
>
> Remember every 0.3 increase of the D-Max is one
> stop darker (double the density).
>
> Thanks,
> Leping
>
-
Roy Harrington
roy@...
Black & White Photo Gallery
http://www.harrington.com