Digital BW, The Print

Steve Kale wrote:

>Here's my understanding...Ernst please correct me if I am wrong.
>
>You can of course still profile the printer for colour in the normal fashion
>(print a test target, measure the results and then compute the ICC profile).
>Typically this is done with No Colour Adjustment set in the printer, ie the
>printer simply renders the RGB file numbers without adjustment.  The
>controls in the driver (eg the sliders) effectively "calibrate" or tweak the
>printer.  One can create a profile which is specific for some set of
>driver/printer calibration, eg a profile specific for a particular set of
>calibration controls set in the driver.   This profile would only be good if
>those controls are set at the same as when the profile was made.
>  
>

I wouldn't mess with sliders in making a profile. The UC models are BTW 
the only models that allow some gamma adjustment (darker lighter) on top 
of CM. As standard you select the best paper setting in the driver for 
the paper you intend to use and make a target without a printer profile 
interfering. That's done with the same settings (in the driver and PS) 
you will use later on with the custom profile but without a profile in 
PS selected (so same as source). With that target the RGB printer 
profile is made. The paper setting takes care of the (hidden) GCR or UCR 
and ink limits + the switchpoints between C/LC and M/LM and K/LK.

>This is fine for colour work.  But I am not so sure about the B&W controls.
>The B&W controls take a colour RGB image and convert it to B&W, or take an
>existing greyscale image, and then allow the user to set certain printer
>calibration features (eg tint) to render the image.  It sounds like it may
>also use a different set of lookup tables than for colour.  If one wanted to
>get an accurate soft proof of a particular set of calibration controls one
>would normally, as described above, print a test chart with those controls
>set, measure the results and generate the profile.  Unless I am mistaken, I
>think most profiler software will error out in trying to do this unless some
>accommodation is made via a new profiler software release and test chart.
>For example, think of the process of using Eye One Photo to profile an Epson
>printer.  You print the test chart and measure the results.  If the results
>differ too much from expected then the software errors.  The test chart is a
>set of colour patches.  The output will be in B&W mode a set of greyscale
>patches.  I think it will bomb out.
>  
>

Color profile creators will normally try to give you a wide color gamut. 
It is against the nature of the beast to ask it for neutralising any 
color data. There should be new B&W functions added to profile creators 
and an extra B&W rendering added to color profiles. That would be the 
most elegant solution for any printer driver. Custom color profiles have 
the basic data for that task in them already but to get them working is 
another thing. The best possible now is to unsaturate the color profile 
in all renderings. But that creates a separate profile.

>Linearization is a different (but related) issue.  It is rather rare (as far
>as most people are concerned) that a profiling package will test for
>linearization first and then construct a new test chart which takes the
>printer's linearization profile into account.  I do not expect Epson to
>allow us to tinker with the printer's linearization.  Sometimes a lack of
>linearity is better than perfect linearization when it allows the gamut to
>be extended.  Non-linearity is taken account of in profiling.  I suspect
>that Epson decides what it believes the optimum linearity/non-linearity is
>and sets it accordingly.  The references in the reviews to linearity suggest
>just two things: they have decided that more linearity is better (making it
>easier for profiling) and the pro models are individually tuned to a
>particular mix as they come off the production line whereas the 2400 is not.
>  
>
CMYK printer profile creators for RIPs assume that the printer-RIPdriver 
is linearised already and that linearisation is the first thing you do 
with RIPs before you start making a profile. Like it is done in QTR but 
with the difference that in today's QTR you get two fixed printer 
profiles with the package and you have only to make the underlying 
linearisation (without the profiles first).  That can only be done 
because the task of the 2 profiles in B&W printing is much simpler than 
it is in color printing. The advantage of a separate linearisation step 
is that you can linearise the printer every month and still use all the 
profiles you made a year ago.

Some profile creators have linearisation functions as well in the 
package. Sometimes because they do not exist in the RIP used or are not 
good enough in the RIP used.

In RGB printer profiling the profile creator has to assume that the 
printer-driver is more or less linear (as there is normally no 
possibility to do that in the driver) and if it is not it will take care 
of that function as far as it can. If the printer is linear then the 
profile creator will only add the perceptual curve, if it is not linear 
it will make a hybrid curve that linearises and creates the perceptual 
curve at the same time. It is obscure to the profile creator though as 
it only knows that the greys should get that ideal perceptual range. 
When the printer isn't consistent in a month time you can start making 
all your custom profiles again. That's the disadvantage of Epson drivers 
and RGB printer profiling.

The Wide Formats always had more factory linearisation than the desktops 
of the same generation. Proof Printing makes it a necessity. Several 
factors contributed to that and the bigger but more consistently sized 
minimum droplet is one of them. Heat problems in bigger heads can be 
taken care off more easily than in the small desktop models.  Especially 
the first UC Wide Formats improved on that better linearity and so the 
Atkinson profiles could be used quite well on any 9600 or 7600. In time 
they all get less consistent but for example the 9000 generation was far 
less linear right from the factory than a 9600 is. What they now have 
added is a linearisation test per printer that is coded in the firmware. 
So if you can get the service manual of the new models and the right 
tools it may be possible to linearise the printer later on when it isn't 
as linear any more as it was when it left the factory. It should be in 
the service manual I guess as with any head change the service man has 
to recalibrate the printer (wonder whether he waits 24 hours top let the 
target dry). It could as well be hype and just a new name for the codes 
that you have to type in the firmware when you change heads, that has 
been normal since the Epson 3000 appeared. I have no idea whether the 
same is required for desktop models but I think that they have to rely 
on production tolerances in head fabrication.

Your "Sometimes a lack of linearity is better than perfect linearization 
when it allows the gamut to be extended" is something I have thought 
about in the past. There are physical limitations on the maximum of ink 
per channel you can use, there's no gain in color saturation or Dmax at 
some point and any more ink then creates more problems than it gives 
advantages. Especially the less transparant pigments are subject to that 
law. Linearisation between zero ink and that ideal maximum doesn't 
decrease the gamut of the total ink mix. There are however two issues 
that may decrease the gamut after linearisation and because of that 
linearisation: The perceptual curves may fit better within unlinearised 
channels than within linearised channels, the opposite is possible too. 
And in time the printer may give less ink at the max inklimits selected. 
The last more or less undermines the advantage of profile once and 
linearise every month but one can change inklimits at the same time 
based on the Dmax of the targets.  With QTR I did select a slightly 
higher inklimit for the black than Dmax told me, to compensate for both 
issues.

It is too early to know what can be done with the new Epsons on B&W 
calibration. There's a thread going on in the Colrsync list and that one 
isn't telling us much either and has my speculations as well :-)

Ernst