Digital BW, The Print

Hi Keith

My brain is starting to hurt on this stuff and I am about to ask Tyler if I
can borrow some of his Tylenol <g>

But I thought I would add a point which I made to Paul Roark offline that
has to do with step wedges, mid grey and differences of tonal range.  This
may have some relevance to your exercise.

The way step wedges are printed at the moment, we put dMax at one end and
dMin at the other and look for tonal change immediately.  (We also kind of
improperly label these by putting a 100 next to the black and 0 next to the
white.  This is because we are thinking in terms of ink rather than the
values of the space we are working in. In LAB 0 is pure black and 100 is
pure white.  But that is a different story.)

Assume for a minute that the printer perfectly replicates LAB's defined
change in density FROM DMIN TO DMAX.  If my printer can print from L=16 to
L=96 (dMax =1.68 to dMin =0.04) and I want to put this range down on paper
in 21 uniform increments of LAB then I have to ask it to print a step wedge
with L values of 16, 20, 24, 28, 32, 36, 40, 44, 48, 52, 56, 60, 64, 68, 72,
76, 80, 84, 88, 92, 96.  Note the mid point is not L=50, but rather L=56.
If I am using say Epson Premium SemiGloss and can reproduce, say, from L=6
to L=96 and I want to print in equal steps of LAB then I have to print a
different series of L figures (6, 10.5, 15 etc) and the mid point here is
51. L=50 still prints at a density of 0.73 but this isn't necessarily the
mid point of the printer range.

Even if the printer RIP automatically automatically spaces LAB values from
dMin to dMax, ie 0 gets mapped to 16, 5 to 20 etc, you still end up with the
same result: the mid-point shifts.

(By the way, take a look at this sequence of co-ordinates, and their density
equivalents, and look how all print values are shifted.  Plot the step vs
density figures of the two papers and overlay LAB.  Even if the RIP doesn't
make this linear, as I believe was suggested, but curved in equal increments
of LAB, each paper has a different gamma and non have the same value or
curvature as LAB. Only as paper white moves closer to perfect white and ink
dMax approaches perfect black do the curves begin to converge and have the
same gamma.)

This reflects no other fact than I get can get better dMax on EPSG than EEM
or that the I have a more equal reach into black as into white.

So if you do curves or suchlike so that your step wedge always prints with
the 50 step print at L=50 then you are distorting one end of the scale.
What's more this curve will be different across papers. EPSG is more
balanced around L=50 than EEM, for example, and requires less distortion.

This whole setup gets awkward very fast.  You either have a shifting mid
point and a greater or lesser error from LAB or you have a fixed mid and
varying contortion to fix the mid.

With the setup I have proposed there is no error from LAB (or whatever you
want to calibrate to) over the dynamic range of the printer, L=50 always
sits in the middle of the step wedge, and to the left or right of it is LAB
until you hit the limit of the printer's dynamic range.  The corollary to
this is that just prior to printing I need to actively recognize the tonal
range of my printer vs the image.  If the image is all in range then I need
do nothing.  But most likely it was already, or manipulated such that, it
used the full tonal range of my workspace.  I can¹t print this tonal range.
I must now decide how to alter this range to fit within my printer range.
One simple curve.  If I don¹t shift the mid point, any part of the image
which I exposed to L=50=Kodak grey (and did not shift later) prints Kodak
grey.

I hope this helps.

Cheers

Steve

> From: Keith Douglas <kdouglas@...>
> Reply-To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Date: Tue, 07 Dec 2004 15:24:17 -0800
> To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Subject: Re: [Digital BW] Tonal range and linearization
> 
> 
> I'm on windows, so I don't have the luxury of using Roy's curve creation
> tools.  I've started from scratch with Excel and Matlab.  Here's an
> overview of what I've done so far (using UT1 on a 1280 btw):
> 
> Print K, C and LC step-wedges with (mathematically) linear transfer
> function for QTR.
> Measure dmax and gamma for each, create 1/gamma curves.  This roughly
> linearizes the output.
> Print K,C and LC step-wedges with my new 1/gamma curves.
> Measure each step on the step-wedge.  Fine-tune curve to give linear
> brightness from dmin to dmax for each ink.
> Partition these three inks to give overall linear brightness.
> 
> Here's where I've stopped before.  The output is very linear (except I'm
> still having trouble with the transitions between inks).  Middle grey is
> the exact middle between paper-white and ink-black.
> 
> Now what I'm thinking is adding an overall curve to first map (128,128,128)
> to Kodak middle-grey.  Then I'll adjust the slope of the curve to make the
> on-screen contrast the same as the
> printed-contrast.  What-I-see-is-what-I-get.
> 
> -Keith
> 
>