QTR-Quadtone RIP

Again Roy is far more experienced than me to reply but if of any 
interest see below

 If I  
> understand clearly what I've read, then there have been two main  
> workflows ending with QTR.

To my understanding it's a single workflow : step 1 linearization so 
you get evenly spaced density and control the amount of ink laid on 
the print together with the nozzle firing top achieve a smooth ramp. 
step 2 you profile this output so that perceptive rendering is 
achieving meaning you re-map values from source to destination in a 
way 1)you can softproof, 2) you achieve a proper distribution of L to 
human perception.

> 
> The first (and apparently outdated?) process, completely bypassed 
any  
> color management system and sent the file straight to QTR sans  
> conversion, as per the printing instructions in the tutorial. So, 
the  
> calibration & characterization/profiling of any ink/paper set were  
> being combined in the QTR curves. And, any particular bumps in the  
> output were taken care of by further manipulating the curves or  
> through correction layers or some similar tactic.

No profiling was not done before the Create ICC droplet exists. Yes 
you could tune the output with a layer, but profiling is far more 
accurate. All in all you just take a file value of say 128 (in RVB or 
grayscale) to map it in a way it will be close to L*50 when you read 
your print and so on and so forth from 0-255 to L 0-100 (in fact 
often L 2à to L95 for instance).
> 
> Now however, it appears that the linearization is only meant to  
> calibrate the output and any characterization/profiling is being 
done  
> via QTR-Create-ICC, essentially leading to the same type of 
process  
> as a color managed workflow.

Yes except there's no color, so you only handle L value. 
> 
> While I can't find the quote, I recently read something regarding  
> color management that stated, ideally calibration and 
linearization  
> would be all that was needed to color manage a device, but since  
> there are still irregularities in the output, profiling is 
necessary.

In fact you can do without linearization and only do profiling : 
linearization just make sure you have a stable devise thus 
facilitating the profiling in terms of accuracy.
> 
> If the above statement is true, then am I right in coming to the  
> conclusion that using profiles in addition to the QTR curves is  
> optional, as long as one is willing to tweak their linearization  
> curves or use corrective layers in order to get their desired 
output?

Yes but again layers and curve correction is a hassle as compared to 
profiling.
> 
> Also, if working with profiles, since the calibration is largely  
> dependent on the particular paper being used, would one have to  
> create a set of QTR curves for each paper as well as create a 
profile  
> or in the case of differing tonal output, multiple profiles per 
paper?

Yes : 1 linearization (density varies), 1 profile (gamut varies) per 
combination of printer/ink/paper.
> 
> What exactly is the target performance that is desired, when  
> linearizing your output, re: the "ideal" densities? Am I correct 
in  
> assuming that the densities are meant to correspond to the human  
> visual response to changes in value, or are they more an arbitrary  
> division of densities based on mathematical/mechanical efficiency?

I am not sure I understand : there's no ideal density. One only tries 
to maximise Dmax, then get a smooth ramp in the linearization 
process.  Then the profiling process is a bit complex. It's based on 
CIE works and used "models" that describes human vision independently 
to the devices. So when you profile, you get a value from a file 
source that are known and reading from the destination source. This 
is process thru an independent space (CIELab or XYZ) so that 
correspondance between source/destination is achievable independently 
to the device to profile and ends up in a conversion matrice or 
table. In our case you want a 128 RVB value to produce a mid tone 
gray on your paper given the ink/paper capabilities and printer 
irregularities similarly, quarter tones needs to visibly look like 
quarter tones ect... As said your file can go from RVB 0 to 256 (or 
gray 0% to 100%) while you can only achieve L 20 (you can not lay 
down a total black patch) to L 95 (the paper is never fully white). 
Still you want to compress and remap you 0-100% file to this L20-95 
destination space : this is what the profile will achieve for you so 
that when you look your print you feel like you have got the best 
values distribution and it 1) ressembles your screen to the best of 
the paper/ink ability (in fact very close) 2) does correspond to what 
a normal person would expect as a properly produced BW image (mid 
grey at L50, with a feeling a 21stepwedge is properly distributed). 
CIELab and other similar models are based on the fact that Delta is 
the smallest possible difference a human can perceive between 
different colors (or gray in or case). But human perception is not 
linear : 2kg in your hand do not weight twice a kg when you take 
them, 2 sugar spoons in your coffee de not make it twice as sweet 
when you taste it and the same for our vision (this is why you can 
perceive such a large dynamic).

This is why just linearisation looks black : you need to increase L 
much more than twice to perceive a print twice as brighter.

Roy will correct me if I'm wrong somewhere. I just inteneded to share 
my little knowledge of it since I remember I had a hard and desoerate 
time trying to get it.

Olivier
> 
> Thanks again in advance,
> 
> Eric
> www.ericashworth.net
> 
>