Digital BW, The Print

Hi Bruce


> From: bruce greene <bagreene@...>
> Reply-To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Date: Sat, 4 Dec 2004 22:45:41 -0800
> To: <DigitalBlackandWhiteThePrint@yahoogroups.com>
> Subject: [Digital BW] Re:tonal range
> 
> 
> Hi Steve,
> 
> I've been following your interesting discussion. Haven't always
> understood it, but it got me to thinking.
> 
> I disagree with your statement that the two tonal spaces can be made
> the same. In color or in b&w.

Colour has a degree of complexity that massively exceeds B&W - note that
this discussion intentionally divorced itself from hue and stuck with a
discussion of dead neutral density range (defined as tonal range, ignoring
hue).
> 
> The same frustration you are having trying to get the same "impact" in
> a b&w print (on matte paper) as you see on your monitor also occurs
> when printing color on matte paper.

I like the prints I make with QTR. I like all on this list continue to hope
that we can overcome the challenges of glossy papers so that tonal range can
be extended.

> 
> I think you are quite right that to keep the matte print from looking
> "flat", the mid tone gamma should match the gamma of the monitor image.

This is not really true per se.  Arguably it could look better in print than
the screen version (in compressed space) BECAUSE the printer uses a
different gamma but then hopefully we would have recognised this with the
screen image and made the adjustment already.  As a workflow, though, it
would certainly be easier if the printer gamma (and tonal range) was the
same as the workspace so that we got WYSIWYG.

> To accomplish this we have two bad choices:
> 
> 1. clip the highlights and the shadows from the image to keep the gamma
> the same as the display thus loosing details in the highlights and
> shadows.
> 
> 2. compromise by using an "s" curve to keep the midtone print gamma as
> close as possible to the display and compressing the highlight and
> shadow gamma (flatter curve) so we don't completely through away the
> shadow and highlight detail, but this detail is now less noticeable or
> easy to see.
> 
> A combination of 1 and 2 above may be the best choice for many images
> printed on matte paper with pigment inks.

Or in fact glossy despite it's greater tonal range - because it is less than
the monitor a compromise is always necessary.  My point is simply one of
workflow - knowing that the adjustment curve (your "s" curve) embodies the
complete adjustment taking place at the print stage.  Run through the
example I mentioned to Paul.  Take a look at the curve (better yet go
through the process of aligning the other points on the step wedge (I just
did the end points and mid point).  It is very informative.  You can see
dMin and dMax, where mid, and all other points are going to be mapped versus
the Gray Gamma (or Adobe RGB) workspace.  You can see if you are clipping or
not.
> 
> The best approach is to extend the tonal range of the print, which
> means printing on glossy paper.

Or continue the quest for better Dmax and better paper white on ANY paper
surface. ;-)

> Otherwise, compromise is in order
> whether one has colorsync to make the conversion from screen to print
> or not.

Yes
> 
> I think what I'm saying is that a "linearized" printing curve on matte
> paper will show all image detail from the display, but in a way that
> often looks too flat, lacking impact and realism.

Here I think you need to check what you mean when you say linearized.
Linearized doesn't mean the print is flat.  It means that the curve plotting
log10(luninance) against log10(pixel value) is a straight line or linear.
Even with your screen this is a linear function.  Take a look at this site
(about 1/3 of the way down):

http://www.normankoren.com/makingfineprints1A.html

> 
> Today I was building qtr curves for German Etching paper with FS inks
> on my 1160. I made a good curve that showed all the steps of the step
> wedge in the same relation that they appear on the monitor. But prints
> using this curve lacked realism though they successfully displayed all
> the tones in the photograph. I then made a qtr curve that incorporated
> an "s" curve from photoshop (it's cool that qtr lets you do this,
> though it took me a while to figure out the cryptic directions!). The
> general print looks better, though it's harder to see details in the
> extreme shadows and highlights as expected.

Ah!  You demonstrate my point exactly!  When you made the curve (as I do
mine) QTR linearizes from dMin to dMax over the range of pixel (or image
file) black to pixel white. The reason why when you print your first image
(without the S curve) it looks flat is because you sent a file with full
tonal range to a printer with limited tonal range where the points in
between were linearised across the full potential values rather than for the
set of values it could reproduce. That is, it had a different rate of change
of density going from 0 to 1 than the screen.  You didn't get what you
expected.  (You can replicate what you got on paper on screen by a curve
which shifts 0/0 to say 0/48 and 255/255 to say 255/243 - with a straight
line in between - depending on the dMin and dMax of the ink/paper - see the
example I went through in calculating the end points for a density range of
0.04 to 1.56.)  So obviously this first print is unsatisfactory and you know
it is because the printer has a much more limited tonal range than the
screen.  You prefer not to clip so you create an S curve.  But in making the
S curve you are trying to adjust for two affects - the curve itself and the
different gamma and tonal range of the printer.  I am simply suggesting that
this can be made easier and more intuitive.
> 
> I think the reason I've rambled on here is that my point is  trying to
> match gamma from monitor to matte paper and keep a linearized scale is
> impossible. It's like trying to fit a 3 inch diameter pipe into a 2 7/8
> inch pipe. It won't fit unless you grind down the outside of the 3 inch
> pipe. And of course, it's no longer a 3 inch pipe.

I disagree.  Let's change your 3 inch pipe to a flat piece of metal in the
shape of a wedge 5 inches long.  At one end it is a sharp point, at the
other it is 3 inches wide (a right angle in the bottom right corner).  Now
the printer.  Chop a slither of each end so that instead of a point oon the
left there is a blunt 1/2in straight line and the right end is now just 2
7/8 high.  Put the two on top of each other and line them up.  The slope of
the top edge hasn't changed.  QTR keeps the wedge the same length but lifts
the left point up and the top right point down - the line between the two is
straight but the slope has changed.
> 
> And this leads me to conclude that linearizing print output by
> measurement is kind of a waste of time. A compromise needs to be made
> that only the skilled eye of a printmaker can shape for most pleasing
> results. There is no translation tool automatic solution.

You are confusing two things here.  When you do a QTR curve and print the
step wedge and then add at the bottom of your .txt curve file the LINEARIZE=
information you are telling QTR the readings it needs to linearize with its
linearization method.  This is important.  Without it you won't get a
sensible movement from black to white.  QTR, I believe, linearizes to LAB
with this info.  This is the linearization process we are talking about  -
the slope and the range over which this slope is applied (on a log basis -
the change is not linear when plotted alone - see the link above).  So you
do linearize.  And then you remap the image file with a curve - changing the
file values - to get the aesthetic result you desire with a curve.

> The photoshop 
> "soft proof" using icc profiles is a way to visually accomplish this by
> limiting the output of the monitor to match the range of the print. The
> monitor can be changed to match the print, but not the other way
> around. This is the only way to " make the two tonal spaces the same
> and not have the issue to begin with!"

On this you are wrong I am afraid.  A soft-proof visualises the impact of a
transfer curve (the mapping of pixel values from A to B) without altering
the underlying file.  When you put up a soft proof from your QTR work note
how massive the change is - it is worse for a narrow range paper like a
matte.  This is because the two spaces are so different.  (Do a softproof
with a general colour photo and to a paper like Epson Premium Semi-gloss.
You don't typically see such a massive transformation because the gamut of
the printer is much closer to that of the image on screen.)  A soft proof is
nothing more than the curve I went through in my last response to Paul
(except that it is just visual and no change is made to the image file).
For colour it is enormously complicated.  If we are just concerned with
tonal or density range (not hue) in a B&W world, it becomes quite easy to
replicate in a simple curve.

Probably the best way to think about it is this.  Pull up your favourite
image so that it looks great on screen (no "print s curve").  Soft proof to
your favourite paper with the relevant soft proof file you made by printing
out the step wedge with QTR and reading the results a la the instructions.
The change you see reflects the remapping conducted within the printer RIP.
Note that all parts of the image (all image file values) are affected - EVEN
THOSE POINTS (incl the mid) WHICH ARE WELL WITHIN THE CAPABILITY OF THE
PRINTER.   Why?  Because when you print you are changing spaces (radically)
WITHOUT the benefit of a sophisticated system (colorsync) for managing the
change.  You simply dump the pixel values on the printer.  Then you are left
trying to create a curve (a mini colorsync if you like) to manually get the
image back to "balance".

Trying to align devices in a color world would be impossible.  Imagine if
every printer had to print the same and every user had to make sure they
worked in the same space as everyone else.  Colorsync was developed to
overcome this issue by creating a sophisticated system for profiling
devices, converting file values from one space to another (transfer
functions or curves) and algorithms for dealing with out of gamut colour
(which include clipping).  Unfortunately these systems have not generally
been refined for the vagaries of B&W - metamerism.  As a result, at the
moment we use a Same as Source workflow with colorsync off.  Hence we end up
fiddling with curves to do what colorsync would otherwise do for us.  That
is, we "deal with it".  Yet the B&W space is very narrow and a lot less
complicated than colour if we ignore hue in file preparation (we do this
anyway because we work in grey scale space not a colour one).  It would not
be that hard to align the workspace and the printer space.  In QTR terms,
this would not entail anymore work than what is done currently to write and
linearize a curve.

Cheers

Steve