QTR-Quadtone RIP

Hi Eric,

I did not see your mail in my box before today : this mail box is not 
the one I use most, I just set it when I open a Yahoo account.

I now understand where the name "Eric" comes from...

I am close to be able to use QTR as I expected now thanks to Roy and 
Steve Kale patience and cooperation : I've been harassing them with 
really basic questions and lots of confusion, but they were both kind 
enough to help setting the whole thing.

I glad you were able to get QTR to work without a densi-
spectrophotometer : I would not have been able to... at all. I have 
been using in the past a profiler based on a flatbed scanner but 
whave been always idsappointed with the results. This is why I got 
the spectro : for a amateur (e.g. a beginner) it is a piece of 
equipment you need. I can't rely on myself I need the hardware 
support to get accepatble results, but it's damn costful.

Meanwhile I think it's worth having the stepwdedge lienarised with 
the spectro readings so is it for the ICC. I guess you must find in 
the US some profiling companies that will do the job for you at an 
acceptable price or ask someone with the instrumlent to help you : it 
takes 10mn, no one would dare refusing.

I read in some posts that the best way to increase Dmax is 1) to set 
boost at Ink limit +10-15%, 2) to add a bit of dark grey to 
strenghten the black. Both advises I have found in other litteratures 
and I feel they are relevant. Doing this I get on EEM with Piezotone 
a 1.62-1.63 Dmax which to my taste is acceptable. On Epson Smooth 
Fine Art I expect something better, but I'll process it only when 
totally set with QTR (it's to dear to play with it).

ICC and Color Management is definitely a "guru" thing (and quite new, 
a V4 release has just come out and there tends to still be lots of 
issues to adress...)or fully accessible to highly competent people. 
Despite lots of existing documentation, a non-pro (like me) gets 
really confused. The whole idea behind the scene is to 1) describe 
colors a device can produce, 2) get 2 devices to interpret-reproduce 
at set of given colors in the best same possible way they can achieve 
so that these colors look alike on both (not the same but alike or 
predictable). When it's done by a commercial profiler all go well, 
you just follow the instructions (like for any OS for instance : 
which end-user really cares the way it works as long as it works and 
you don't go into deep computing). When you move one inch out of the 
path, you get into the Unknown... In QTR case both Steve and Roy did 
their best to get me understand. Here is what I could catch :

Basically when you linearise the printer output you get a smooth 
grayscale from Dmin (paper white color basically) to Dmax. Say the 
paper L* is 95 (L* goes straight from à pure black to 100 pure white) 
because no paper is pure white, and printed Black is L*17: this is 
physically what your system can produce, nothing better. Meanwhile 
your image file since it is digital can go from pure white (L*100 or 
in PS RVB = 255) to pure Black (L*0 or in PS RVB = 0), because 
digital is not the "real" word but just datas, or a "virtual" world 
in which perfect datas are achievable. So the issue is to produce a 
print which has a limited reproduction capability (L*17-95) with a 
perfect digital image file (L*0-100). Now it gets nasty (and I could 
not conceptually get it straight ahead) : when you linearise the 
printer from L*17 (of say about Dmin 1.62 in case if you prefer 
density) to L*96 (density 0.03) and print just doing it you do not 
properly reproduce the image file L*0-100 : you are just making even 
steps from 17 to 95. To understand this open a PS grey image and make 
a curve layer you can keep the curve datas in RVB (no needs to turn 
it into %). Now, you can calculate L*17 in RVB is equal to 17x
(255/100) = 43 so you move the bottom left point so that input is 0 
and output is 43, your black is not pure black. L*95 equals 242, you 
move the upper right point so that input is 255, output is 242, your 
white is not pure white. If you look at the curve dialog box from RVB 
43-242 you have a straight line (linear) but your image is flatened 
(less contrasty) and lighter (the mid point input 127 has moved 
somewhere to ouput 140) KEEP IT OPEN FOR THE TIME BEING. This is a 
representation of what QTR does when it linearises a printer output 
of the stepwedge L*17-95 and this does not properly reflect what your 
image file is and looks like. 

So what you want is your output at L*17-95 to look like your image at 
L*0-100. Let me insist on this : it should look like it e.g. to the 
best ability of the physical system. It will not be like it, but 
ressemble it.
This is where ICC comes in. There is some litterature you can find 
browsing the web for a more in depth explanation, I will only 
describe what I understood for QTR. It is only practical and not 
theoretical nor mathematical. Again the idea is to get a ressembling 
to the screen results : this ressembling in the case of QTR is called 
perceptual. In the ICC you have four rendering intents, the 
perceptual one is the one meant to produce an output that for human 
eyes is uniform (take a round balloon, compress it in the same round 
form, imagine air molecules in it, you could say they kept a uniform 
spacing : this is more or less what perceptual does with colors) 
since we best perceive variations between tones instead of absolute 
tones this is the ratio between tones that is preserved when you 
convert from one color space to another one. You could also take a 
colored patchwork fabric for instance make it (reasonably) fade, if 
all colors fade in equal proportion your eyes will be fooled. What 
perceptual intent algorithm does is to compress the image file values 
so that they fill in what the printer (destination color space) can 
physically print and this is precisely what we want to make a image 
with datas ranking from L*0-100 to fit on a paper that can only stand 
L*17-95.You may know that no human senses and perceptions are non-
linear meaning if you double the input you don't get the feeling it 
has doubled : put one spoon of suggar in your coffee tastes it, now 
put two, does it feel like it's twice as sweet? Put one kilo in your 
hand, now put two, does it feel like you doubled the weight ? Well : 
no. For the eyes it's the same : double the light you will not feel 
like it. So since vision is not linear the output should not be 
either. To feel the same "contrast" (in BW let's say this is what you 
feel : darker or lighter) you simply need to make black blacker and 
white whiter. This is the ultimate purpose of QTR perceptual ICC. 

To visualise this go back to where you left PS e.g. a light stepwedge 
as compared to the original file. Now plot those points ! 1) input 35 
output 45, 2) input 127 output 127, 3) input 215 output 235. Does 
it "ressemble" more to your initial stepwedge 0-255 as compared to 
the 43-242 one : it should, it's not the same but it is more alike.

This is my understanding of QTR ICC with the invaluable support of 
Roy and Steve.

As for the 6 inks of Piezotone, I was too very disappointed with the 
initial banding. I actually manage to now have the 6 heads firing 
without banding : you need to first set the first ink (in 6 inks 2 
inks are the same say LC=M) to whatever relative density and ink 
limit 50 and do "copy from" and input this ink for the same other 
ink. so M would be density 12 ink limit 50 and LC would simply 
be "copy from M". I have posted my curve to Roy describing the file. 
I can also send it as a attachment if useful. Here it is :

PRINTER=Quad1290
CURVE_NAME=EEM2
GRAPH_CURVE=YES
N_OF_INKS=6
DEFAULT_INK_LIMIT=65
BOOST_K=75
LIMIT_K=
LIMIT_C=
LIMIT_M=50
LIMIT_Y=50
LIMIT_LC=
LIMIT_LM=
N_OF_GRAY_PARTS=4
GRAY_INK_1=K
GRAY_VAL_1=100
GRAY_INK_2=C
GRAY_VAL_2=39
GRAY_INK_3=M
GRAY_VAL_3=12
GRAY_INK_4=Y
GRAY_VAL_4=5
GRAY_HIGHLIGHT=4
GRAY_SHADOW=8
GRAY_OVERLAP=5
GRAY_GAMMA=1
GRAY_CURVE=
N_OF_TONER_PARTS=0
TONER_HIGHLIGHT=10
TONER_SHADOW=10
TONER_GAMMA=1
TONER_CURVE=
N_OF_TONER_2_PARTS=0
TONER_2_HIGHLIGHT=10
TONER_2_SHADOW=10
TONER_2_GAMMA=1
TONER_2_CURVE=
N_OF_UNUSED=0
COPY_CURVE_LC=M
COPY_CURVE_LM=Y
LINEARIZE="96,37 81,17 70,98 62,61 54,44 48,72 45,2 41,69 38,03 34,46
32,23 30,35 29,03 27,75 25,28 22,65 20,87 19,72 18,61 18,08 17,75 "

I am not too sure of the 6 heads firing since you can read gray part 
= 4. However n_of_inks reads 6 and at the end of the file you can 
read copy_curve_LC=M and LM=Y : so I assume this is it.

I don't know much about other ink sets : I chose Piezotone based on 
1) my willingness to print only BW without toning : if needed I use 
my color printer, I believe it is well profiled enough to produce 
acceptable to my taste toned BWs. 2) Various beliefs in France 
favoring Piezotone, when you don't know where to go you follow the 
crowd... there are too many debates to really single out a system.

At that point, the only thing I'm sure of is I am far above 
expectations on QTR results and extremely glad with it : this is a 
soft that has changed my miserable printing. With a (tiny) bit of 
work and appropriate willingness to progress there's a lot to take 
out of it. Sadly enough, I have not (yet) found any one in France 
using it...

I post on the group list so that if I'm wrong in some saying, I don't 
fool you and it gets corrected by more competent people than me.

Enjoy your printing.

Olivier