[argyllcms] Re: how many patches, profiling Epson printers

  • From: Roger Breton <graxx@xxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Tue, 26 Jul 2005 13:02:55 -0400

Dear Jan-Peter,

You are 100% richtig: linearisation is paramount in printer profiling.

> If a output-device behaves not linear, the number of patches for
> profiling has to grow.

Agree, but up to a point. Not so much because of more samples will not
result in more exhaustive characterization but, read below.

> In the case of inkjet printers, the linearity depends very much on the
> limit for every CMYK-channel.


> For every combination og paper an Ink, there is a point were more Ink
> results not to more saturated color, but to huge shift in the Hue.

True again.

> The normal way to detect this, is to print steps from 0 to 100% of every
> pure color, measure it and plot the results in an a-b diagram in the
> Lab-colorspace. The point, were the hue began significantly to change,
> is the point to set the inklimit for every channel.

Yes, that is the correct analysis. And, as you point out, this presupposes
having complete control of *all* channels.

> After setting this limit, a 1-D LUT for every channel should calculated
> which results to equal steps of the pure colors and a good
> CMY-greybalance. 

Aha! I agree with the first part of your sentence but I cannot confirm the
second part. Why? Linearizing each channels to perfection, however you want
to do it, does not, in my experience results in CMY-graübalance. Ist das
deine experienz?

> For doing this step, softwares are using differnt
> models. 

Das ist der "secret sauce". I wish I could compute this myself :(

> Some are based on density, and some are based an Lab-values...


> In the last step, it is necessary to detect the max. inklimit / TAC
> (Total amounf of color) for a given combination of paper and Ink. This
> is mainly done by a visual inspection of a testchart where big patches
> with different TAC an fine details are printed. If the fine details are
> detroyed be "bleeding" og the big patches with a given TAC, this is
> point to set the max. Inklimit.

Good description again.

> After this three steps, the testchart for profiling is printed.
> In most cases a Testchart with e.g. 1485 patches like the ECI2002 gives
> good results.

Yes and no. I have an Epson 4000 here I'm profiling through PrintOpen v5,
ProfileMakerPro v5.0.3, Monaco Profiler v4.7.2 and Fuji ColourKit v4.2 and
it's never perfect. I always have to perform profile editing to get a good
match. Why? I don't know. But there is something non-linear somewhere in the
way these UC inks are mixing that defeats the profiler prediction.

The following point is *very* well taken:

> One last point to Epson Ultrachrome-printers:
> For this printers, it is very hard to get a visual nice greybalance.

How widespread is this view, Jan-Peter? No one on the ColorSync List ever
admitted to this in public. I guess it depends on a lot of factors. Not so
much the fact that the ink behavior is heavily metameric but the fact that,
I suppose, a lot of people don't really notice it or complain about it.

> I´m 
> getting significant better results by using a strong GCR starting at 0,
> so that neutral tones are mainly rendered with black and light black.

Well, this is what I've been lately experimenting with. Traditionnally, on
all the inkjet printer I ever profile, I always used UCR. But I found it
constrained me too much when editing the profile. So I started to use GCR. I
can't say, however, that I find the graü balance better?

> Such black generation can only be used, if the printer has additional
> grey ink, 

Epson 4000 have Black and Light Black.

> otherwise the strong black produces a "peppering" in
> skin-tones (big black dots...)

Yes, like Epson 3000.

> greetings from europe
> :-) Jan-peter

Greetings from Kanada!


Roger Breton  |  Laval, Canada  |  graxx@xxxxxxxxxxxx

Other related posts: