Digital BW, The Print

The linearization of a printer is separate from profiling the printer.
Typically as part of a RIP you have the ability to linearise the output - in
effect, in our case, smooth the greyscale ramp so that there is a constant
increase in L* as we go from dMax (pixel value 0) to dMin (pixel value 255).
In the case of the Epson driver you do not have the ability to relinearize
the output.

Once linearization is complete and, in the case of a RIP, embedded in the
curve that tells the printer what inks to fire and when, you can profile
that output with an ICC profile.  A greyscale ICC profile only profiles the
luminance axis (not hue).


> From: odesmais <odesmais@...>
> Reply-To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Date: Wed, 03 Aug 2005 16:48:16 -0000
> To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Subject: [Digital BW] Re: QTR ICC Profile (MAINLY) + R2400 ABW (MUCH LESS)
> 
> I have had and still have the same trouble understanding the QTR ICC
> effect. In essence my understanding of a traditional icm is that it
> reflects/includes both the printer-paper-ink caracterization and the
> conversion needed to map from one space to another thus including the
> needed linearisation calculation. In the case of QTR caracterization
> and linearization are done in the QTR curve creation process

You tell QTR what inks to use in the greyscale, to what limits and in what
order etc.  You then linearize this output by including the linearize
function as you mention so that there is a smooth ramp in L*.  I would not
call this "characterization" - certainly not in the profile sense.  It is
also confusing that the term profile has been introduced here.  We used to
refer to this process as the process of building ink "curves".  They are
still called curves in the QTR driver.  They are not profiles in the ICC
profile sense.


>and 
> curves are called from the printing interface (this is why curves are
> refered to as profiles in the manual and the soft). So while at this
> first stage you can't proof you do have a linear output.

At this stage you have nothing to soft proof - you have not yet generated an
ICC profile of any sort.  You could go on from here and do one of two
things: you could follow the process to use Eye One Match to generate a
"preserve color numbers" ICC profile which would allow you to proof for hue
and luminance.  This profile merely portrays the tone change and compression
- it does NOT help you manage it.  Alternatively, you could go on to use QTR
Create ICC to accurately PROFILE the luminance axis of the greyscale created
by your QTR Curve (or Curve mix).  (It is not necessary to use gray-lab
space at all - think of this as something that was developed as we fumbled
our way to the QTR Create ICC process.)
> 
> ICC generator adds accurate soft proofing without a doubt :
> Well,to me at least, and it seems to be the feature Roy wanted to
> add, this would, to me again, explains why before the Generator he
> created a gray-lab space and generic mat/photo grey icc in the hope
> one would go for a first basic soft proofing ability, but I tend to
> believe BP and WP were then only generic and not describing the real
> ones while he probably wanted true soft proofing with actual BP/WP.

Here is where you are getting confused.  When Roy first pottered around with
profiling the luminance output with an ICC profile he made two GENERIC ICC
profiles - one for an average matte paper and one for an average photo
paper.  (There were a few things wrong with these and ideally you would not
use them anymore.)  In the course of developing on this idea, Roy enhanced
QTR Create ICC to read actual data rather than hypothetical data.  This
allows us to create an ICC profile of our ACTUAL output (ie our actual QTR
Curve, ink, printer, paper combination - or in fact any combo that generates
a greyscale).  So QTR Create ICC is a refinement or development of the
generic to actual data.


> 
> Back to the new ICC, it can also serve a conversion in PS from RGB to
> Grayscale. 

I would not recommmend using these profiles as a means of converting from
colour to B&W.  That is niot their intent.  They are intended to profile the
luminance of the greyscale produced by a particular workflow.

>Besides, from memory, using this Grey ICC as a printing
> space profil also produces a B&W image from a RVB colored file in the
> OEM driver (though I did not print, I just visualised the preview).
> So I would tend to conclude it is a true ICC generator that can be
> used like any other profile except that instead of mapping source
> space colors to destination space "available" colors it maps s.s.c.
> to destination space grey values, probably only retaining the L*
> values to do so. 

You need to understand colour management a little more.  At the end of the
day a printer just gets numbers and knows what to do for each number.
Colour management allows us to know what is meant by a particular number and
so translate between numbers.  When Same as Source is used the numbers are
sent unadjusted.  When we use PS to do the colour management the numbers are
altered by PS before being sent to the printer.  When we ask the Printer to
do the Colour Management the raw numbers are sent along with the profile so
that the printer driver alters the numbers.  In the recommended QTR workflow
the file numbers are altered by PS before being sent to the QTR driver.

>And Roy recently informed he wants to provide full
> icc support including colors to reflect toning and white paper color
> if I'm not wrong.

I would say this is a very long way off.  At this point you cross into the
realms of the full colour world.  We all know how poorly ICC colour managed
workflows perform for B&W.  They do not yet allow the compressed gamut
functionality that we want so dearly.  But using colour management to help
manage luminance compression is a huge step forward.  We can get the
luminance axis right and manage hue the old way - in altering the amounts of
the inks selected to make up the ink curve.
> 
> Now since the read target is an already linearised stepwedge (this is
> how I understood you have to create the ICC not from the non-linear
> setpwedge), the profile just becomes a description of the printer-ink-
> paper BP and WP with evenly distributed grey values.

Yes

> 
> This is where I am confused and additional testings need to be run :
> 1. convert a RGB file with the grey profile

As I noted above there are far better methodologies for getting the image
into greyscale in the first instance

>and send it as such to
> the Epson driver or print with "print space" profile being the grey
> icc to measure linearity. If unlinear (and it must be) the profile
> simply act as a PS grey mode action in this direction.
> 2. Create a profile from the non-linear stepwedge and use it in the
> destination space profile from an RVB file : though it will
> demonstrate nothing but the linearisation is achieved by the profile
> and could serve in ABW mode for instance.
> 3. print with QTR with and w/o converting to the profile (oh yes I
> run Windows as an OS) adn mesure the stepwedge.

You have lost me here.
> 
> I understood that without QTR, the EOM driver will interpret whatever
> comes gray into a CMYK output thus blending the color ink in a non-
> linear way (the same with a quadtone setting). So the whole QTR
> profile generation would turn to be mainly a soft proofing feature :
> whether you convert or not a gray file before sending it to QTR will
> have no effect (as long as you create the profile from the linear
> stepwedge).

No.  You can use QTR Create ICC for Black Only printing for example.  BO
printing generates a greyscale which likely differs from printer to printer.
It is also unlikely to be particularly linear and you have no way of
linearizing it - the Epson driver doesn't allow you to.  But you can profile
it with QTR Create ICC.  Print a step wedge using BO, read it into QTR
Create ICC to generate a profile of that output.  You can then, rather than
printing Same as Source, convert to this profile on the fly at printing in
PS.  The file values going to the printer will be altered to reflect the
print space.  This has nothing to do with soft proofing but you can also
create a soft proof with this ICC profile to see how the image will look
when converted to this profile at printing.

The situation is a little confused by the fact that Roy is doing the black
point compensation in the profile (in a similar way to how he is required to
do the white point compensation - required by the ICC spec).  When you do a
soft proof using an ICC profile generated by QTR Create ICC at the moment PS
sees that it has perfect black, ie that no black point compensation is
necessary (because Roy has altered the kTRC to do BPC there and so it goes
from full white to full black).  Remember this is just a beta version and
things are subject to change!!

> 
> If not wrong in the above, the "better" B&W experienced using a QTR
> ICC would come from teh fact that actual BP and WP would have been
> measured and grey values re-distributed accordingly, proving QTR ICC
> provides a true B&W profiling feature.
> 
> But still they are questions :
> 1. Do you need to(or should you) use this profile when printing with
> QTR for anything else than softproofing ?

Yes.  The idea is to convert to the profile from your workspace at printing.
Think of it as exactly analogous to doing colour managed colour printing -
except the QTR ICC profile only cares about luminance.

> 2. Would you consider converting a RVB file into a Grayscale one in
> PS using this profile ?

No.  If you mean a straight conversion to this profile then NO.  First there
are better ways of making the conversion to B&W.  Secondly remember that
this ICC profile is not a workspace but rather a printer output profile - it
describes a very narrow situation.

> 3. Would you use it in an ABW workflow with a 2400/4800... and if so
> do you create from a linear stepwedge or non-linear one in the case
> of the 2400 ?

Yes.  With ABW you can only print a step wedge.  The linearization is
already done and embedded in the driver.  SO print a step wedge and profile
it.  Convert to this profile on the fly from PS so that file values are
altered by PS before being sent to the ABW driver.


> 
> I'm sorry all this sounds confused and probably repeating part of
> what Steve wrote, but to me... it is very confusing.

Again just think of a colour colour-managed workflow.  You work in ProPhoto
or Adobe RGB and you files are tagged as such.  When you come to print you
select the printer profile for you printer, paper, ink combination.  You
might first do a soft proof.  When you print you are likely to have PS do
the colour management (the new terms in CS2 really are much more
informative) and hence PS converts (read: alters the file numbers so that
colors look the same in the print space and manage out of gamut colors
according to the rendering intent selected) the image file to the print
space before sending it to the driver where you have selected no colour
management (you don't want to convert the file twice!!)
> 
> Now one question to Steve :
> 
> 
>> In the current
>> incarnation of the Grey ICC profile generator, Roy is doing the
> black point
>> compensation rather than leaving it to PS (this is why you can't
> proof for
>> the weaker print black at the moment).
> 
> This I don't understand : "can't proof the weaker print black" ? In
> the (little) testing I did I would have said the proofing also adjust
> the BP. This would also be demonstrated by Roy's instruction
> mentionning you need to soft-proof using perceptual and BPC (the BP
> being thus embeded in the grey ICC). I'm sure I wrongly understood
> the sentence.


Read up above.  Black point compensation was designed by Adobe to plug a gap
in the ICC spec.  When Roy was looking at this ICC stuff he decided to code
the BPC into the profile generator (ie into the profile) rather than leaving
it to PS - "it has to be done anyway" right?  In essence the curve that
describes the printer output is scaled so that it runs all the way down from
the measured black point to perfect black.  (If you open one of the ICC
profiles in Colorsync and look at the kTRC tag you can see that the curve
starts at the bottom left corner 0,0 instead of say 0,10.)  So when PS looks
at the profile when you check Black Point Compensation it sees that no
compensation is necessary ie the profile says that the printer can print
perfect black.  And so the black you see on screen (which is a file perfect
black displayed as the best black your display can produce according to the
intent you select in your colour management preferences) is not altered to
reflect the weaker black of the printer.  Checking and unchecking BPC makes
no difference to the image.  Roy and I have chatted about this and it is
something he will take another look at whenever he gets a chance.  I'm sure
he has a lot of other things to do other than just push forward the B&W fine
art printing world!

I hope this helps somewhat.


Steve