Digital BW, The Print

Steve Kale wrote:

>I don't disagree with the goal of opening up linearization to all.  But many
>people don't want to go down that route.  But even if they don't there is a
>lot to be gained from profiling whatever greyscale they've got.  For
>example, black only printing has been very popular for many.  I would argue
>that the principal reason for this is its simplicity.  But the greyscale
>output of black only is anything but linear.  If one could profile it,
>though, then those users could get the benefits of a CM workflow (or
>luminance managed workflow) even if the baseline isn't able to be
>linearised.  The CMM has to do more work but at least it can do some work.
>  
>
>The greyscale on my 4800 with Epson Adv B&W is really very linear.  I don't
>see the need for QTR with it for photo paper (assuming I stay with the
>standard inks).  For matte paper I would like to try increasing the ink
>limits to see if I can get a better dMax.  So with the 4800 and Epson inks I
>am not going to be using QTR for its linearization.  Nonetheless, I would
>like to be able to profile the luminance axis of its output and use CM to do
>the dynamic range compression.
>  
>
We have used custom RGB printer profiles on all kinds of drivers for 
color printing, there are advantages in cheaper profiling software and 
less skill is needed. But we are building those profiles on Epson paper 
setting with unknown GCR, ink limitations and most likely basic 
linearisation.At the same time we have argued that for third party 
papers and inksets there's a limitation in that approach. So far the 
printers have not been stable enough to keep those RGB profiles running 
for ever. So new profiles are needed while in a RIP solution one could 
linearise for a whole set of profiles more frequently and keep them 
unchanged.

>Regarding the number of observations, it is interesting to note that Bruce
>Fraser mentioned to me he was going to be doing "a whole lot more" than 51
>for the Adv B&W soft proofs.
>  
>
For the soft proofs ?   Based on the hidden, hard coded linearisation of 
the Epson I can understand that you need more points to correct that 
linearisation and fit the perceptual curves for printing, if it was just 
for fitting the perceptual curve you would need much less I expect. For 
soft proofing the translation of the Epson irregularities may ask for a 
number like that. But again if the printer is perfectly linearised I 
think a perceptual profile needs very little points to fit.

>As for why Jon went for QTR I am not sure but I would suggest that he needs
>a RIP to drive his ink set and QTR works well and is cheap for his potential
>users.  They can then provide "ink curves" for their ink set with just a
>small additional cost to the user (Roy's $50).  They can customise the
>"linearization" to what they think is appropriate by adjusting the variables
>that Roy has provided (or tweaking the .quad files).  That is, the cost of
>adoption is low.  But as I noted earlier in this post, the final greyscale
>will still have a particular profile (black point, white point and gamma)
>and I would argue that even Jon would do well to go this extra step and
>provide a greyscale ICC profile that depicts this compressed dynamic range.
>Typically, in Curves or Levels in PS, we adjust the black point and white
>point of our image file to be (near) perfect black or (near) perfect white.
>These two points will never correspond to the print on paper points.  For
>these reasons alone it would be nice to use CM to do the necessary
>compression.  I do not see Jon's choice of gamma as reducing the desire for
>this.  
>  
>
When Epson announced the new range of printers with the advanced B&W 
mode + the hard coded linearisation I expected a spin off from the CM in 
the printer for Advanced B&W printing too. With all the work and 
knowledge gone into the paper settings and related color profiles it 
shouldn't be difficult to take relevant data from that part and add it 
to the B&W settings. It has not been done as far as I know (asked here 
and on the colorsync list). Maybe they thought that the hard coded 
linearisation and an additional perceptual curve per paper setting was 
enough.

>So I do see the ICC profiling as being independent of greyscale
>linearization.  What we really need is some input from a colour engineer
>with blinkers on.  By "blinkers on" I mean we need to convince him/her, and
>this is very difficult to do, that we don't care about hue - just luminance
>and that we want CM to just manage luminance.  Now tell us how to set the
>tags and tell us how the different intents can be/should be employed.
>Unfortunately this is a very foreign world for them.  It is quite hard to
>get clear answers to questions like "if I create a kTRC tag which has this
>shape and select the primary intent to Perceptual, what actually goes on?"
>or "why are my highlights getting clipped when I have the media white point
>set correctly and the kTRC correctly portrays the stimulus/response
>behaviour of the printer?  Shouldn't they be compressed but not clipped
>because the intent calls for it as part of its definition?" - especially
>when they are not actively involved in what you are trying to achieve.
>
I have advocated a B&W rendering added to normal ICC profiles. It must 
be possible to use the measuring data in the opposite way they were 
collected for. To take out any cast as good as possible and achieve the 
smallest gamut possible for a given inkset. That real quad B&W printing 
needs as a base at least 3 greys, a long black generation and total ink 
limitation on the CM and less on the cmY ink channels is unavoidable but 
the same B&W rendering approach should work on those choices too.

For some RIPs you can predict that the manufacturers will rather add 
special (color) paper settings for B&W and (color) profile that 
combination then add a separate Advanced B&W mode. I'm also curious how 
they will split EPS and PDF documents in B&W and color. Will it be BO 
for greyscale data like it has been or shall we get an extra split for 
quad quality aka Advanced B&W?  I bet that the advice will be to use 
neutral RGB for semi quad (= good neutral profiled color printing), 
greyscale for BO and the rest color. With good profiling the result will 
be better than it used to be, enough for a majority of RIP users

There's a thread on the colorsync list how the different renderings and 
ink limitation interfere (and shouldn't) so it isn't as easy as we think 
it should be.

Ernst