Re: Laminate profile - Abstract profile?
Re: Laminate profile - Abstract profile?
- Subject: Re: Laminate profile - Abstract profile?
- From: Hanno Hoffstadt <email@hidden>
- Date: Sat, 8 Aug 2009 15:43:05 +0200
Dear Nils,
thanks for your questions. Now be prepared for the story of my life...
(just kidding)
Am 08.08.2009 um 13:21 schrieb Nils Johansson:
Dear Hanno,
You are quite right about the change in hue angle! It's NOT
negligible, we totally missed that it was expressed in radians in
our report, but we should certainly change this to degrees. Now it
also matches your results, with as a high difference as 6 degrees in
the patches you pointed out. Thank you for making us aware of this!
Have you performed any visual appraisals of the changes as well? We
only performed measurements.
When I started my investigations, I worked for a book printer where
most of the color printing was paperback covers or hardcover dust
jackets. They always received some kind of surface finish, either
lamination or varnish. The printers were aware of color shifts and
relied on their experience, not quite adjusting to the customer-
supplied proof but keeping everything more open, starting with plate
curves, sometimes using adhesive tape to check. Dispersion and UV
varnish was inline and could therefore be judged during the print.
Lamination was offline and still sometimes produced surprising results
(and reclaims).
We always kept both unlaminated and laminated samples for each job, so
we had quite a lot of visual examples for testing. But those were
usually not printed to standard, but to customer proofs which often
had a horrible quality (in the days before standard profiles and the
Ugra/Fogra media wedge), requiring big adjustments at the ink zones.
To sum up, the printer tried to compensate for the changes so that the
final product matched the customer proof (which is what typical
customers expect - before education).
One of the biggest issues seems to be to find the best conversion-
technique between the individual colours in the unlaminated and the
laminated gamuts... Different L*a*b*-values change differently.
Using traditional conversion techniques, it should be hard to
convert less saturate colours to a more saturated value if the
target profile has a larger gamut than the source profile. This is
why we suggested the method of assigning the profiles (using the
common CMYK-values as an intermediate step). But as you pointed out,
the use of standardized profile is very often more preferable than
custom-made ones. Have you tried this assign-method as well?
For separated CMYK data, compensation has its limits. It's impossible
to reach the density of a paper type 1 proof on a matte laminated
print. And you are not paying the brilliant, glossy finish just to
have the potential dynamic range crippled down to the unlaminated
result. So, over time, some printers have decided not to compensate
anymore (to some extent backed up by the offset process standard).
This is also the "official statement" of the ECI working group on
surface finishing.
Therefore I'm not going to suggest how to convert CMYK data to retain
their appearance after lamination...
What one can do is to use laminated profiles for separation (which
produces the optimal result for the available gamut), proof with a
post-lam profile for the print buyer, proof with a pre-lam profile for
the printer at the machine and tell him to really print the wacky
colors. The "assigning method", yes, and we used this quite a lot
internally, but also with selected (and properly educated) customers.
There are still some problems in practice because you cannot find the
correct output intent for a PDF-X. If you keep RGB data and use post-
lam as output intent, this will correctly separate the data and proof
the final result, but then you cannot create a PDF-X compliant pre-lam
proof, since on one hand you need the post-lam separation, on the
other hand you must manually override the reference profile to
simulate, avoiding any more conversions. This still has to find its
way into today's proofing software, or the PDF-X evolution process.
Which software did you use when making your Abstract profiles?
That was a little piece of handwritten software, and I remember now
that I dropped this soon, because even the *same* Lab value will
change differently depending on its black generation, because the
optical effects interact with a different dot structure (think a gray
from pure K versus CMY). This also applies to spot colors versus their
CMYK translation. Since solids do not have additional dot gain, there
is a big difference between a pastel solid spot color and its CMYK
alternative, the former hardly changes, but the CMYK quartertones are
heavily affected by lamination! (Yes, I've seen that, and alas, it
once happened in a technical book series where the publishing house
suddenly decided to print the 4th book in CMYK to save some money.
Naturally, we had to reprint in CMYK+spot.)
By the way, I later used CMYK-CMYK device link profiles (post-lam to
pre-lam, preserving black generation), but not for conversion
purposes, but to study the additional dot gain in overprints. At the
Fogra Symposium I (over)simplified things to a +10% rule of thumb,
which means that all tone values increase by 10%, even the 10% ones.
This is good for single ink wedges, and also for, say, a M wedge on
100 Y, but there are ink combinations where a second ink has a
decreasing effect on the additional dot gain of the other ink. In
terms of optical paths, the clear pattern of a single ink halftone
becomes "polluted" by the presence of other ink dots, there are less
edge effects etc. This is what has been empirically determined and
accounted for in the new beta profiles I mentioned earlier.
I saw that some have suggested a conversion between the unlaminated
profile and the laminated profile, but I don't know which rendering
intent best simulates the actual L*a*b* changes... Has there been
any evaluation of the results from different rendering intents
concerning this approach?
No (more) comments on conversion :-)
It would certainly be very interesting to compare the colour change
for a spot colour and its corresponding CMYK-translation, having
identical L*a*b*-values – as you said. Whether they behave the same
or not.
They don't as I said above...
As said, it could be the dots, not the actual L*a*b*-values, which
cause the different changes in colour... The lpi appeared to affect
the results when reading your report, and the solid patches did not
change that much for glossy laminate?
Yes, the dots. A very coarse screen has no additional dot gain
(because dot edges are small against dot area) and very fine screens
have low additional dot gain (because the additional pathlength of
light due to total reflection is large compared to the screen ruling,
and the probability to hit dot or non-dot does not depend anymore
where the light came from). It is just the intermediate screen ruling
from 120 to 180 lpi which has the highest effect. Also, blurry dot
edges (e.g. inkjet on some substrates) have almost no effect.
Therefore, better not try to laminate a standard proof (not even an
inkjet dot proof) and hope that the same happens to the proof with
happens to the printed sheet.
Solids: see below.
In our report we reached the conclusion that the measured dot gain
post lamination is not the *sole* reason for the colour changes, as
seen in diagrams 9 to 14, were the solid line represents the L*a*b*-
values for a CMY tone scale from 0 to 100 %, and the dots represents
the L*a*b*-values for each step after lamination. If the colour
changes follow the dot gain, then the solid line and the dots should
superimpose, but they don't. Thus the dot gain is not the sole
reason for the changes... Have you reached any other conclusions
about this after publishing your report?
I didn't say that the additional dot gain explains everything :-) Not
even in single ink wedges. Especially not for the solids which change
despite having no additional gain.
But did you try to add or subtract stray light from the measurements,
which I did in my report? A matte finish will cause more diffuse light
to be captured in an 45/0 geometry, so you can use the "model":
XYZ(measured on matte) = mixture of XYZ(measured on unlaminated) and
XYZ(D50). D50 and not, say, "lamp spectrum" A, since all the detected,
reflected light (with the A bias) is first corrected to a spectral
reflectance factor (independent of illumination for nonfluorescent
samples), then scaled to D50 illuminant before Lab calculation.
The other way round for gloss (not in the report, but in the beta
profiles): assume that you still have some stray light in your 45/0
measurement of the unlaminated sample, and less so with a very glossy
finish. So to predict the gloss effect, you have to subtract D50 stray
light from your XYZ data which makes colors more intense. This is
linear in XYZ, but nonlinear in Lab, and maybe that's what causes the
discrepancy (mainly in the magenta trace in diagram 9)?
(I later learned that a CRA paper by Dalal and Natale-Hoffman, "the
effect of gloss on color", in 1999 described similar ideas to compare
45/0 geometry and 0/d specular included/excluded.)
Hanno
--
Dr. Hanno Hoffstadt
Color Scientist
------------------------------------------------------------------------
Kirchstrasse 5
D-89081 Ulm
Germany
phone: +49(0)731 38900957
GMG GmbH & Co. KG
Moempelgarder Weg 10
72072 Tuebingen
Germany
Phone: +49 7071 93874-0
Fax: +49 7071 93874-22
mail to: email@hidden
www.gmgcolor.com
------------------------------------------------------------------------
GMG is a leading developer of high-end color management solutions
with more than 9,000 systems in use across the industry spectrum by
ad agencies, prepress houses, offset printers and international
gravure printers.
GMG GmbH & Co. KG. Registered Office: Tuebingen.
Court of Registration: Stuttgart Local Court, No. HRA 381463
General Partner: GMG Verwaltung GmbH. Registered Office: Tuebingen.
Court of Registration: Stuttgart Local Court, No. HRB 382505
Managing Directors: Joerg Weihing, Robert Weihing, Paul Willems.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden