QTR-Quadtone RIP

Thanks Ferdinand.  It's interesting to see how many different
transformationsshow up.  Unfortunately solving it all seems to be elusive.
 The CMS of Qimage
is done by yet another person, so it becomes very difficult to pin down the
problems.
In fact even to agree on the existence of something wrong is hard.

The difference between my generic profiles and now custom profiles is mostly
an
accident of the evolution of my understanding ICC profiles.  There are two
major formats
for curves in ICCs -- one uses a "transfer function" and the other uses
"lookup tables".
The transfer functions are simpler and what I did first -- they have one set
of curve values
and the same curve is used for both directions of a profile -- i.e. K to L
and L to K with the
CMS doing the inversion.  I used this first and that's what the generics
use.  Later on
I used the lookup tables which allow lots of curves for different things
---each direction,
each intent and full color as well.  This allowed softproofing with the hue
of the print.

So the custom ICCs use the tables and I have separate K to L curve and L to
K curve.
L=0 in your file maps to K=100 to the driver, but in softproofing K=100 maps
back to L=dMax
so you can see "Simulate Black Ink".  I think somehow the softproofing curve
is getting
used during the print conversion but I haven't any idea where or why. (maybe
BPC isn't
even the real issue).

You can actually see some of this and play with conversions using Photoshop
and softproofing.
I think there are some free programs that you can download to graph the
innards of ICC profiles.

   *  Ernst's point about the PS CM never being fully off is well taken,
      but these are large changes.  All programs than can display the
      histogram (e.g.  BreezeBrowser Pro), display the same dMax loss.
I think you are safe if you look at histograms -- the info palette is ok
ONLY for the native type
of the file.  I.e. Don't look at RGB values for a grayscale, only the K is
valid for a grayscale.

Roy

On Dec 24, 2007 4:43 AM, ferdinand_paris <ferdinand_paris@...> wrote:

> CLARIFICATIONS:
>
> *  All conversions were perceptual (but see below).  CM_OFF was done
> in PS.  Others were done in Qimage.
> *  All conversions were to my newly created custom RGB profile.
> *  Ernst's point about the PS CM never being fully off is well taken,
> but these are large changes.  All programs than can display the
> histogram (e.g.  BreezeBrowser Pro), display the same dMax loss.
> *  I was using Qimage 2008.104 but have since upgraded to .109 (latest
> - see below).
> *  I tested Ernst's idea of using 8-bit input to Qimage to no avail -
> same problem.
> *  Since the initial QTR-related problems with the CMS in Qimage were
> discovered in early 2006, I have generally followed Ernst's workflow:
>  i.e. create a copy and do the CM in PS, and do all other printing
> manipulations in Qimage with CM off.  But this partly defeats one
> purpose in using Qimage, the ability to print from master files and
> not have to create special versions just for printing.  It works for
> colour, but not it seems for B&W.
>
> FURTHER RESULTS
>
> I upgraded to 2008.109 and one histogram changed.  The one with
> profile conversion done in Qimage but no BPC ("CM On, BPC off in Q")
> now looks exactly the same as the one done in Qimage *with* BPC ("CM
> on, BPC on in Q").  Now in both cases the black point shifts from 1 to
> 27 with an associated loss in dMax.
>
> This tests my sanity.  I am fairly sure that I used all perceptual
> conversion when I produced those histograms.  But I can't now
> replicate the middle histogram ("CM On, BPC off in Q") using *any* of
> CM options in Qimage.  This suggests that something has change in
> Qimage, but the revision log says nothing about any changes to the CM.
>
> I decided to push the tests a step further.  I also produced
> histograms for conversions in Qimage to generic profiles.  Results are at:
>
>  http://www.ferdinand-paris.com/BPC-Qimage_V2.png
>
> CONCLUSIONS
>
> *  BPC is not the problem per se when I convert to a QTR custom
> profile, the reading of the custom profile by the whole Qimage CMS is.
> *  There is something about the custom profiles that Qimage can't
> handle, but can with the generic profiles.  I don't know if this has
> been discovered and commented on before.
>
> @Roy:  I still don't understand the differing treatment of BPC between
> custom profiles and the generic profiles that you mentioned in post
> 5728.  What practical difference would I expect to see (other than the
> fairly obvious ones in Qimage)?
>
> F_P
>
>


[Non-text portions of this message have been removed]