Okay, I just tried reducing the max K value to 95% and I got totally crazy
curves. I don't think messing about with the K limit will help in this
situation. This was the command used for the profile linked to in the
previous post:
xicclu -L 95 -g -kp 0 0.33 0.95 1 0.5 -l350 -fif -ip
KHLIndigo10000-Nevia-limit400-CalGlossyDefaults-31Aug18-SpectralM0-averaged-kh-l350-s20-ca27-cg15.icm
And here is a screenshot of the xicclu calculated curves:
https://drive.google.com/file/d/18QO0tRcIHZ-PATmzBqjHIwEhmrIziSjK/view?usp=sharing
Regards,
Samuel
On Mon, 10 Dec 2018 at 10:59, Samuel Chia <inksandpaper@xxxxxxxxx> wrote:
Hi Graeme,
Thanks for the tips. Here is the profile used for generating the black
curves: https://drive.google.com/open?id=1UsGoZMv5FKlZdRN8Eus0g35YvLsN-6UI
Here is a layered TIFF file with each layer named according to the black
curve command variations:
https://drive.google.com/open?id=1Bt2BjGNaYqz2ElEUQaqJzSDvXSv2lNiz
Note the layer named "-kp 0 0.33 1 1 0.5", and it will make it easier to
see where the kink is in " -kp 0 0.33 0.95 1 0.5 ", with only the enpo
changed slightly.
This kink is always at the 90% point on the x-axis, where the curve (seems
to) always wants to be convex between 90% and 100% even if the main body of
the curve is concave from the "shape" command being <1.0. If the main body
curve shape is >1.0, (convex throughout) absolutely no kinks are visible.
If the main body curve is mildly concave (~0.8), then the kink is not as
visible. High stle and stpo points with a concave shape tend to make the
kink at 90% more obvious. If there is no forced turning point at 90% and
that point is allowed to freely move to prevent kinks, the final curve
shape could be smoother, and the corresponding CMY curves are smoother as
well (compare "-kp 0 0.33 0.95 1 0.8" with "-kp 0 0.33 0.95 1 0.5")
so it's a challenge to provide
controls that would be easy to use.
I actually have no issues with the current set of controls. Stle, stpo,
enpo, enle and even shape are quite intuitive to me, I approach them like
how I would with parametric curves in image editors.
I have yet to test the profile on a printed result and anyway the Indigo
in question has far bigger issues than tonal smoothness (it can't even
reliably put ink on paper, the main job of a printer). Regardless, smoother
curves cannot hurt?
Because that is where there is most choice about ink composition. There is
no choice
at white.
Looking at the collection of different curves in my layered TIFF, I see
that the curve at the white end is happy to be concave or convex to
smoothly transition into the main curve body shape (compare the concave
stpo of "-kp 0 0.33 0.95 1 0.5" with the convex stpo of "-kp 0 0 1 1
1.5".). However the black end always wants to be convex. I certainly don't
know anything useful about splines or whatever to be able to make a
constructive suggestion beyond this, though I can present my observations
from the curve shapes that xicclu is showing and I hope it is somewhat
helpful to arrive at the issue. It would seem possible to carry over the
behaviour from the white end to the black end of the K curve?
I am also struggling with a different issue which I had hoped you'll have
the time to look into. For some reason, I am getting strong tone reversals
and discontinuities with my CMYK profiles built using the
lightness-preservation perceptual command. I later realised that by
lowering the total ink limit from 400% to 350% greatly reduced the
discontinuities (I would like to know why that is), though it is not 100%
free of it like my other RGB printer profiles are. Also, increasing the -S
compression helps to reduce the discontinuities, though it always resulted
in far worse gamut mapping (too much compression by the time the profile is
smooth enough). I posted about it earlier:
https://www.freelists.org/post/argyllcms/Tone-Reversals-and-discontinuities-in-shadows-of-CMYK-profile
In the meantime I have resorted to not using a highly concave curve shape
to avoid this issue. I will also try your suggestion of reducing the
maximum K value. Hopefully, the behaviour can be changed for the better in
a future update. I really appreciate your time looking into these issues.
Thanks Graeme!
Regards,
Samuel
On Mon, 10 Dec 2018 at 09:32, Graeme Gill <graeme@xxxxxxxxxxxxx> wrote:
Samuel Chia wrote:
Hi,
I've noticed that doing more severe concave curves tends to result in akink at the black
end point. This kink is always at the 90% point. Here is a link to thescreenshot of the
curve preview generated by xicclu.https://drive.google.com/file/d/1UE-DaCallwUTQjVS-1qrV53NEyM3TuTa/view?usp=sharing
yes, the range of control is not idea, but the interactions of black
level and
resulting ink combinations is not that intuitive, so it's a challenge to
provide
controls that would be easy to use.
The graph you show looks pretty good as such things go. Are there
actually any
visible artifacts in the printed result ?
Interestingly, the white end of the curve never develops any kinksregardless of the stle
and stpo value for any shape parameter between 0.0 and 2.0. It only andalways affects the
black end.
Because that is where there is most choice about ink composition. There
is no choice
at white.
When reducing the concave-ness of the shape to ~0.8, the kink isvirtually gone. Perhaps
this curve's spline turning point is hardcoded and I'm wonderingwhether it might be
possible to get a smooth gradual transition to dmax whilst enforcing amore concave shape.
The curve is as plotted in the documentation. The interaction depends on
the
characteristics of the printer though. Without having the profile to play
with,
it's hard to make any suggestions. One thing you might try though, is
slightly
reducing the maximum K value (-L parameter - i.e. try something between
90 and 100%).
Cheers,
Graeme Gill.
