Re: New to Linearize Piezo K7 inks with Eye One 2 X-Rite software

Roy Harrington

Hi All,

Hopefully I can add some more to this without creating more confusion.  I think its worth a bit of background.

In QTR Linearization is a simple correction curve that  results in a straight-line of L values when
plotted against a straight-line of grayscale values.  This has been the principle right from the start.  The idea
is to have consistent behavior over a wide range of curves.  The correction curve is incorporated into the .quad values.
L-values (Luminosity from LAB) are based on human vision, evenly spaced L-values should be perceived as evenly spaced. There are mathematical formulas for this but its based on human studies not a math theory.

There are a couple of things to note about this.  First the idea is to be able to edit images in say Photoshop such
that you know what the print will look like in the end.  Back in the darkroom days you took notes so that you
could do the same thing when you made another print from the same negative.  The beauty of a digital workflow is
that it's at least conceivable that you can actually see (actually a bit of imagining) what you are going to get 
before you print the image.  So screen-to-print matching is always the ultimate goal -- within the limitations of
reflective paper versus a glowing screen.

When you edit/view an image in Photoshop or any other program, file have an embedded ICC profile that maps
the actual numbers in the file to an L-value that is displayed.  The Color Management System is the big black box that
converts file numbers to screen numbers or to printer numbers in such as way to match L-values as much as possible.
With the screen this always going on all the time -- there's no disable.  This is easy on a screen since you only
have to profile the screen once and all programs can use the same thing.  This is all built into the OS (both PC & Mac).

With printers though there are lots of variables so the one-sizes-fits-all concept does not work much of the time.
So how does this all relate to printing B&W with QTR driver (or any other way)?

There are two main issues.  The usual  embedded profiles for editing -- both RGB and gray -- are all "idealized"
profiles, meaning they theoretically range from absolute white to absolute black neither of which are attainable.
Second, they all have various curves defining the match between gray value (K) and luminosity (L).
So there are always issues of compressing into dMin through dMax and matching the curve in between.

So that's the basic problem and there have been many varied ways to deal with it.

While Linearizing L-values makes a lot of sense, just NoCM mapping say GG2.2 to QTR driver linear on
matte paper results in noticeably lighter prints than you'd think or want.  -- because dMax is pretty low that end of
scale is farther from the ideal pure black.  See the pink line in Brian's post -- the whole line is significantly
higher (i.e. lighter) than the 0-to-100 diagonal would be.  Additionally the GG2.2 L-values are not linear
either so what you based editing the image on is different.  I made a linear L-value profile Gray Lab that helped
a little but still the mapping is not the best.

Since Jon Cone started making his own .quad profiles he's reportedly made them to better match GG2.2 with
NoCM.  I'm not sure of the exact mapping so its hard to comment for sure.

Others such as Paul Roark have just designed their own PS curves that can be applied before printing.
Both of these are simply color management hand-done and can work just fine.

Color managed ICC profiles with Black Point Compensation are the industry standard way to do this mapping
in the most flexible way.  Most everything CM talks about color but B&W fits into this just fine too.
On a PC this would always require a manually applied Convert-To-Profile so its a bit of a hassle. On a Mac
this worked quite well up through CS3 but then that was messed up from CS4 on and later OSX versions.
Eventually I wrote Print-Tool for the Mac so I could get back to the old working method.
If you read the BPC doc that Ernst posted there's lots of confusing technical info but the bottom line is the
mapping in XYZ space (for B&W really only Y) makes for perceptually better matching of screen-to-print.
Note that nothing is perfect -- see Brian's second image for the curved black line -- i.e. the BPC mapping.
Throughout most of the range its closer to diagonal but curves at the bottom compressing the shadows a bit.
So, yes, it doesn't work miracles but its the "best" compromise.  If you want more open shadows you've
got to edit your file to do that -- I think you have the best control doing this.  Soft-proofing with simulate
ink-black can help visualize if you like.  -- another gotcha here is use the RGB-Matte-Paper because the
Apple CMS used by Print-Tool doesn't support this in grayscale (argh!).  (I think you can pretty much get over
needing the soft-proof and just visualize/imagine how to edit)


Finally to come full circle on the discussion.  K7 .quad files are not re-linearizable by QTR driver because
there is no .txt/.qidf file.  I think they are all pretty good to start with so usually this is not an issue.
But adding an ICC profile built on the K7 curve you have gets you the full ICC screen-print matching
regardless of how "linear" the K7 works on a different paper or different printer.  This is by far the most
straightforward method on a Mac with Print-Tool but a bit more hassle on a PC.


On Sun, Feb 8, 2015 at 9:56 AM, richard@... [QuadtoneRIP] <QuadtoneRIP@...> wrote:

Like I said earlier in the thread, I have been kept awake with this problem off and on for the last few months/years and it all started clicking earlier last week. partly due to the conversation about the problems with QTRs linear "curve" and the reason for needing ICC profiles to match the print space to the working space. I still think using the ICC profile with QTR severely blocks up the shadows, so I was attempting to map the .quad values to a "modified gamma" curve and print without color management—kinda going in-between where there is an increased smooth transition to the dark shadows but not the hard dip that results in using the ICC profiles. The natural next step is to be able to map any existing .quad profile to different targets—either a straight linear line or "modified gamma". There is some problem if the original profile is "too far off" but what I have been testing is a way to get away from the QTR linearize function, and only use QTR to build the raw profile using the ink limits and gray overlaps. I just did a three print test series with K6 inks and that approach worked better than the standard QTR profile creation workflow of linearizing a raw curve with the 21 step target and then creating an ICC profile (or printing with just the straight line increase).

All this came about because I figured out that my 1430 I use for testing at home is so far out of whack that no profiles other people have created work for this printer. I am looking into getting it replaced but since I need to build profiles from scratch, i developed some tools and methods for making it easier. I then take what I am doing to the studio and test on a properly functioning 3800 with Cone inks, and a 9900 with Epson inks. 

Anyway. I am still testing and refining all this so I don't want to get too far ahead of myself here, but I am working on something that could be very interesting, and could solve a lot of problems linearizing wonky printers, or just make the standard QTR workflow less prone to error. I am not sure how it would translate to digital negatives yet, but that might just be the next step. 

Richard Boutwell


Join to automatically receive all group messages.