Dan Margulis Applied Color Theory
A Prepress Question: Conversion to B/W
Date: Sat, 15 Apr 2006 15:39:40 -0700
From: Leila Joiner
Subject: Prepress question
Hi -- I'm new to this list, which was recommended to me
by a photographer friend. This may or may not be the place to find the
answer to my problem. If not, maybe one of you can suggest where to look.
I'm an independent publisher and also do book and cover
design. Two books that I'm currently preparing for publication contain
grayscale illustrations -- one has about 50 B&W photos, the other 12
pencil sketches.
For both of these books, I have to use Lightning Source
Inc. (LSI) for printing, and they are notorious for lousy grayscale output.
The only information I've been able to get from them is that they use an 85
linescreen, and their endpoint limits are 13% white, 93% black.
I use Photoshop 6, and am trying to figure out how to
adjust these images to get the best possible output from LSI. I've read a
bunch of stuff on Levels and Curves, but I'm still not sure how much to
adjust the Input data, and how much I need to adjust the Output data
(targeting). Or what combination of the two I need to use.
Some of the images have a lot of white space that needs
to remain white. Targeting to the printer's specs compresses the data and
grays everything out. In some cases, I can compensate for this by limiting
the Input data first. OTOH, some images have both necessary white space and
necessary light gray detail, and I have trouble maintaining both.
Also, I don't know how to ensure that what I see on my
end is anywhere close to what I'll get from LSI.
Any suggestions appreciated.
Leila Joiner
Imago Press
Tucson AZ
___________________________________________________________________________
Date: Sat, 15 Apr 2006 17:24:21 -0700
From: Marco Ugolini
Subject: Re: Prepress question
In a message dated 4/15/06 3:39 PM, Leila Joiner wrote:
I'm an independent publisher and also do book and cover
design. Two books
that I'm currently preparing for publication contain
grayscale
illustrations -- one has about 50 B&W photos, the
other 12 pencil sketches.
For both of these books, I have to use Lightning Source
Inc. (LSI) for
printing, and they are notorious for lousy grayscale
output. The only
information I've been able to get from them is that
they use an 85
linescreen, and their endpoint limits are 13% white,
93% black.
Hi Leila.
Wow...that *is* lousy quality! Sounds like specs for
printing on newspaper. Is that the sort of output you are expecting?
I use Photoshop 6, and am trying to figure out how to
adjust these images
to get the best possible output from LSI. I've read a
bunch of stuff on
Levels and Curves, but I'm still not sure how much to
adjust the Input
data, and how much I need to adjust the Output data
(targeting). Or what
combination of the two I need to use.
Some of the images have a lot of white space that needs
to remain white.
Targeting to the printer's specs compresses the data
and grays everything
out.
Did they provide precise dot gain curve point data,
besides the white and black point aims? Without anything more detailed, the
output will remain a crapshoot from your point of view.
In some cases, I can compensate for this by limiting
the Input data
first. OTOH, some images have both necessary white
space and necessary
light gray detail, and I have trouble maintaining both.
Also, you have no information on how the high and low
quartertones will play themselves out.
Also, I don't know how to ensure that what I see on my
end is anywhere
close to what I'll get from LSI.
In your situation, I would see if it is possible to
find a vendor that (a) offers better than newspaper-like quality, and (b)
cooperates with you and gives you the means to control and preview the
results.
Break a leg!
--------------
Marco Ugolini
Mill Valley, CA
___________________________________________________________________________
Date: Sun, 16 Apr 2006 00:42:34 -0000
From:Stephen Marsh
Subject: Re: Prepress question
Leila Joiner wrote:
For both of these books, I have to use Lightning Source
Inc. (LSI) for
printing, and they are notorious for lousy grayscale
output. The only
information I've been able to get from them is that
they use an 85
linescreen, and their endpoint limits are 13% white,
93% black.
Leila, this sounds similar to newsprint specs, but the
highlight min. dot value sounds way too high to me. I would expect a 5% min
in such conditions and would question the 13% min. dot quoted.
I use Photoshop 6, and am trying to figure out how to
adjust these
images to get the best possible output from LSI.
I would use the colour sampler tool (shift+i) to place
info palette readings on the highlight and shadow areas. The sampler would
be a 3x3 reading or 5x5. I would adjust via the info palette, so that
the lightest printing tone where detail is needed meets the min. dot value
- and same for the shadows where detail is needed.
Additionally, the rest of the tonal range should be
known too. What is the tonal value increase for midtones and other tones.
At a minimum, it is nice to know enpoints and midtones, if not more.
Some of the images have a lot of white space that needs
to remain
white. Targeting to the printer's specs compresses the
data and grays
everything out. In some cases, I can compensate
for this by limiting the Input
data first. OTOH, some images have both necessary
white space and necessary
light gray detail, and I have trouble maintaining both.
Also, I don't know how to ensure that what I see on my
end is anywhere
close to what I'll get from LSI.
Where you need no dot, then pure white 0% is required.
Where you need the lightest, first printing dot is where you would set your
highlight point- whether this is 3%, 5% or 13% or some other number is up
for debate.
It seems doubtful that you can get a profile from these
people for either softproofing, but you never know your luck. I would ask
for some previous printed samples of others mono work and a copy of the
final press ready data that you can open in Photoshop to make evaluations
from. As offset print varies, it is good to have samples from different
times/jobs or more than one sample from the same print run if you only have
one sample. A test print or proof would also be another option, but like
the profile, probably unrealistic.
With what is known so far, I would guess that 5-90%
range and 30% dot gain would be expected. I would also add more contrast to
the images and make them a little lighter in the midtone.
Stephen Marsh.
___________________________________________________________________________
Date: Sat, 15 Apr 2006 18:48:31 -0700
From: Leila Joiner
Subject: Re: Prepress question
Hi Marco -- I only have the info I gave you, which was
passed on to me from the tech department through my client services rep. I
asked to talk directly to someone in the tech department, but they
sidestepped that. I have to use this printer for these two books because
they're connected to the biggest US distributor, and it's the only way I
can get that connection. It's sort of a bite the bullet situation.
I do have two other printers I use, one of whom has
great grayscale output,
the other adequate. Just can't use them here.
--Leila Joiner
___________________________________________________________________________
Date: Sat, 15 Apr 2006 19:01:10 -0700
From: Leila Joiner
Subject: Re: Re: Prepress question
Hi Stephen -- Yeah, I thought the 13% was pretty
radical, but that's what I was told. Wouldn't hurt to doublecheck that,
though.
More contrast definitely improves the output. One thing
I'm not sure of,
though, is how to use targeting. Say I make changes
using the input sliders, then do I have to set the output sliders to their
specs before sending to them? And if I do that, does that then give me a
close approximation onscreen of what the final output will be (ignoring
further
damage from the 85 linescreen)?
Thanks for all your good suggestions!
--Leila Joiner
___________________________________________________________________________
Date: Sat, 15 Apr 2006 23:40:11 -0400
From: Lee Clawson
Subject: Re: Prepress question
on 4/15/06 6:39 PM, Leila Joiner wrote:
For both of these books, I have to use Lightning Source
Inc. (LSI) for
printing, and they are notorious for lousy grayscale
output. The only
information I've been able to get from them is that
they use an 85
linescreen, and their endpoint limits are 13% white,
93% black.
From what I heard they are digital on-demand-book
printers using IBM equipment. But the output info you have received reminds
me of low end newspaper set-ups from 15 years ago on newsprint paper. And
maybe closer to the output you'd get with an ordinary copy machine. If
possible, I'd get someone else.
....I've read a bunch of stuff on Levels and Curves,
but I'm still not sure
how much to adjust the Input data, and how much I need
to adjust the Output
data (targeting). Or what combination of the two I need
to use.
If you go ahead and follow their info set up the 1st
printing tones close to their 13% and shadow areas with any detail close to
93%. The rest of the shadows will be black. And if I had to guess the
shadows from 88% to 100 might look so close to one another that they for
all purposes might have been all the same tone. The highlights below 8 to
10% will likely go white.
Use either levels or curves and watch the readouts with
the eyedropper. In general you'll be making the existing images flatter.
Some of the images have a lot of white space that needs
to remain white.
Targeting to the printer's specs compresses the data
and grays everything
out. In some cases, I can compensate for this by
limiting the Input data
first. OTOH, some images have both necessary white
space and necessary
light gray detail, and I have trouble maintaining both.
I don't know what your light gray detail is in %. Nor
do I know how many light tones you need to hold. In addition I don't know
for sure if you're on a digital press, high speed copy or offset. So read
the rest as a guess.
If the "necessary light gray detail" need to
hold its far enough from 10%, like 18 to 25% then you might be OK. But if
you're trying to hold details in very light tones that could be more
difficult.
From what they are telling you everything from 0 to
about 8% is going to be white (as white as your paper) and tone will build
from there with 1st gray highlight details visible at about 13%.
Also, I don't know how to ensure that what I see on my
end is anywhere
close to what I'll get from LSI.
Without any other info from the printer I can't see how
you're going to be predict anything. Maybe you can make prints on a B&W
laser printer and then make copies on a B&W copy machine to get closer
to the poor results.
Any suggestions appreciated.
In order of preference; (1)--Please get someone else,
(2)--Send a grayscale that shows tones in 2 or 3% increments and get a
proof sample to use as a guide for image edits on the paper you use and
(3)--Spec the whitest, brightest, smoothest paper they offer and (4)--Get
samples of their B&W work.
Lee Clawson
___________________________________________________________________________
Date: Sun, 16 Apr 2006 10:04:33 -0400
From: Jim Rich
Subject: Re: Prepress question
Leila,
You need to request a prepress contract proof. If who
you work for thinks that is too much money then it is clear this process is
not about quality. And if the prepress proof is not available then
you are at the mercy of someone else and there is not too much you
can do.
Jim Rich
___________________________________________________________________________
Date: Sun, 16 Apr 2006 16:01:33 -0700
From: "J Walton"
Subject: Re: Re: Prepress question
On 4/15/06, Leila Joiner wrote:
More contrast definitely improves the output.
How do you know? Sometimes more contrast (and by that I
mean an "S" curve) causes problems, *especially* if the bottom
end is going to drop off that dramatically. You need to see the output to
know if more contrast is better. If it *definitely* improves output then
your images are likely WAY too flat.
One thing I'm not sure of,
though, is how to use targeting. Say I make changes
using the input
sliders, then do I have to set the output sliders to
their specs before
sending to them?
This sounds suspiciously like Levels, which IMO would
be a good thing to avoid as a practice. You will most certainly not be able
to move 1/4tone or 3/4tone contrast using Levels, and if anyone here says
they can then they have the expertise to do it in Curves to begin with.
At any rate, if you want to use Levels to set the
endpoints that's fine. But rather than "targeting" using the
output sliders just examine the image to look for the lightest area that
isn't white and slide the highlight over until you get it to (I say) 5-7%.
Then look for the darkest area that shouldn't be black and get it to (I
say) 85%.
And if I do that, does that then give me a close
approximation onscreen of what the final output will be
(ignoring further
damage from the 85 linescreen)?
No. A close approximation onscreen could be had with a
proper profile from them and a calibrated monitor. A better close
approximation could be had by getting a proof from them. Nothing you do in
Levels (or hopefully Curves) should make the image look like what they are
going to do to it based on their specs.
J Walton
___________________________________________________________________________
Date: Mon, 17 Apr 2006 00:20:56 -0000
From: "George Machen"
Subject: Re: Prepress question
Some of the images have a lot of white space that needs
to remain
white. Targeting to the printer's specs compresses the
data and grays
everything out. In some cases, I can compensate for
this by limiting
the Input data first. OTOH, some images have both
necessary white
space and necessary light gray detail, and I have
trouble maintaining
both.
To solve this, don't use Levels; use Curves.
Moving the endpoints horizontally toward the center in
Curves is the same as moving the Input sliders in Levels. (Moving the 50%
point left or right is the same as moving the Input Gamma slider.)
Similarly, moving the endpoints vertically is the same as moving the Output
sliders. (Moving the 50% point up or down is a poor man's adjustment of dot
gain compensation.)
After setting your targeting, in your case by moving
the 0% input to 13% output and 100% to 93%, then *if* you can afford to
clip either or both your highlights & shadows, move the 13% targeted
highlight (or both) horizontally by the same amount(s), respectively. Then
click the little freehand draw icon at the bottom of the Curves dialog and
very carefully draw freeform a straight line leftward on the x-axis at
bottom from directly under your 13% point to the zero origin.
Now everything below 13% suddenly will drop out to
paper white, which is what you want.
If you chose to pull both your highlight & shadow
points horizontally centerward, it will preserve the original contrast of
the image before targeting between 13% and 93%, but, to repeat, at the
expense of clipping the shadow & highlight.
George Machen
___________________________________________________________________________
Date: Sun, 16 Apr 2006 22:04:04 -0700
From: "J Walton"
Subject: Re: Re: Prepress question
On 4/16/06, George Machen wrote:
To solve this, don't use Levels; use Curves.
Moving the endpoints horizontally toward the
center in Curves is the
same as moving the Input sliders in Levels.
Stop. You had me at Curves. :-)
After setting your targeting, in your case by
moving the 0% input to
13% output and 100% to 93%, then *if* you can
afford to clip either or
both your highlights & shadows, move the 13%
targeted highlight (or
both) horizontally by the same amount(s),
respectively. Then click the
little freehand draw icon at the bottom of the
Curves dialog and very
carefully draw freeform a straight line leftward
on the x-axis at
bottom from directly under your 13% point to the
zero origin.
You are beginning to lose me....
Now everything below 13% suddenly will drop out
to paper white, which
is what you want.
Say what??? Why would you want to blow out everything
below 13%? First of all, it's a bogus spec. They may as well have said 31%
- there's no way you ruin a file based on a number like that. Remember,
this came from a CSR who wouldn't let their client talk to a "real
person."
Second of all, even if they meant 13%, why do you want
to destroy detail? If for some amazing reason everything below 13% is gone
mangling the image won't do anything constructive. Better to assume better
conditions than to create worse ones.
I may be misunderstanding the technique but this makes
no sense to me at all.
-----
J Walton
___________________________________________________________________________
Date: Tue, 18 Apr 2006 13:10:38 -0000
From:Stephen Marsh
Subject: Re: Prepress question
Leila Joiner wrote:
Hi Stephen -- Yeah, I thought the 13% was pretty
radical, but that's
what I was told. Wouldn't hurt to doublecheck that,
though.
That's what I would do, clarify the info. Print has a
restrictive tonal range as it is, let alone taking out a further 10pts of
range than expected. A toner based copier that is poorly maintained I would
expect to have problems reproducing anything correct under 10-15% in any
CMY or perhaps K. But on a press or some other beast that has output
similar, I would think that 3-5% to 90% is a good range.
As previously suggested, knowing what the midtone and
or other gain is going to be needed as well as the endpoint range. It would
seem safe to assume 25-30% dot gain. This is all 'unlucky expert' advice
though, it is not ideal to second guess.
More contrast definitely improves the output.
My comment on putting in a little extra contrast was
for two reasons:
i) Experience with coarse screens on poor or better
stock makes me add in more contrast (without plugging shadows or blowing
out highlights) - but that is personal preference on my part, to be taken
with a grain of salt.
ii) If the minimum reproducible dot is indeed 13% and
this is not a mistake, then one is starting with less contrast than what is
ideal, so the rest of the image may need the tonal levels adjusted to suit.
I would still ask that monotone output samples be
provided (with density bars if possible, before trimming), and the
Photoshop images that appear on the same test print. More than one print
sample is good too, from different times in the same print run. You are not
after the best, or a really poor example - just what is average. More than
one job would be nice as reference, with the same 'key' as your images.
One thing I'm not sure of,
though, is how to use targeting. Say I make changes
using the input
sliders, then do I have to set the output sliders to
their specs before
sending to them? And if I do that, does that then give
me a close
approximation onscreen of what the final output will be
(ignoring
further damage from the 85 linescreen)?
I will attempt to be brief, but to expand upon my
earlier post that may have presumed too much reading between the words.
I would not use levels.
I would use curves.
Let's presume a grayscale image, with a full range of
gray tones, 0% specular highlights all the way to 100% solid areas of no
importance.
This image does have highlight detail, in the 5% to say
15% tonal range. It also has shadow detail. We will presume for the sake of
argument that the printer has (correctly) told us that they can hold a 5%
dot and a 90% dot at each end of the tonal range.
My first step would be to use the fixed colour sampler
tool (shift+i) set to 3x3 or 5x5 radius. I would place a sample point on
the lightest significant highlight area that needs a minimum dot. I would
do the same for the shadows too. The info palette would expand to show
these two fixed colour samplers, giving you live readings of the edits in
the curves dialog box.
Then I would open a curves adjustment layer (or just
curve). I would not pay attention to the input/output fields. I would pay
attention to the monitor preview, the info palette fixed samplers and the
info palette while moving the cursor around the image while editing.
It is not the input/output fields that are critical,
but the actual image tones indicated by the info palette readings.
I would then adjust the endpoints and or place further
control points for editing the curve (cmd/ctrl click on the image tones in
the main window while in the curves command to auto place a tonal edit
control point). When editing, when the info palette shows that the tone
that was say 9% is now 5% and that the shadow is now 90% when it was
79% - then things are close.
More on curves and editing curves 'by the numbers' here:
www.gurusnetwork.com/tutorial/curves/
www.ledet.com/margulis/PP7_Ch02_ByTheNumbers.pdf
www.adobeevangelists.com/pdfs/photoshop/tipsandtricks/CorrectByNumbers.pdf
www.curvemeister.com/ (created by a list member)
Hope this helps,
Stephen Marsh.
___________________________________________________________________________
Date: Tue, 18 Apr 2006 15:48:48 -0000
From: "George Machen"
Subject: Re: Prepress question
Say what??? Why would you want to blow out everything
below 13%? First
of all, it's a bogus spec. They may as well have said
31% - there's no
way you ruin a file based on a number like that.
Remember, this came
from a CSR who wouldn't let their client talk to a
"real person."
Second of all, even if they meant 13%, why do you want
to destroy
detail? If for some amazing reason everything below 13%
is gone
mangling the image won't do anything constructive.
Better to assume
better conditions than to create worse ones.
Yeppir, you're right. Setting the minimum highlight dot
*with significant detail* to his 13% (irrespective of targeting) and then
blowing everything out to paper white below that would be butchery. And
that's precisely what my procedure as written would do. But I only was
trying to describe the inevitable tradeoff between targeting and contrast,
and what would happen if you wanted to completely hold original contrast.
(And the latter only could be maintained between the targeting points.)
Sometimes clarity doesn't always come out in email - I
muddled it; I apologize.
Actually, I think your numbers are far more realistic:
After targeting by moving the endpoints vertically, in his case from 0% to
13% and from 100% to 93%, he should *not* move them horizontaly by the same
respective amounts in the name of preserving original contrast (between 13%
& 93%); he should before targeting pick your points of 5-7% and 85% for
highlight/shadow points (of important detail), then move the endpoints
centerward by the percentages he determines are in areas of specular
catchlights or at least insignificant highlights he's prepared to
sacrifice, then draw the freeform line leftward on the x-axis from directly
under that point to the zero origin. Then his specular highlights will blow
out to paper white, which is what he wanted, instead of being muddy 13%
gray. (And oppositely on the shadow end.)
His contrast then would improve (in the range between
whatever horizontal percentages he finally chooses) compared to the
flattening inevitably imposed by the targeting, but not completely, as in
my first example. Again, I admit to confusing things by throwing together
two operations at the same time, and I came out sounding like advocating an
outrageous setting of highlight/shadow in the name of symmetry with
contrast for the sake of the tradeoff description.
George Machen
___________________________________________________________________________
Date: Wed, 19 Apr 2006 00:09:39 -0700
From: Leila Joiner
Subject: Re: Re: Prepress question
Thanks to all, this is great. At least it tells me I've
been somewhat on the right track. I was leaning toward using curves rather
than levels. But you've given me lots more information to work with than I
had before.
I'll hang around and lurk, if you don't mind, and see
what I can learn from other people's problems. And I'll be sure to let you
know if I manage to get better than average results from Lightning Source.
Leila Joiner
Imago Press
Tucson AZ
___________________________________________________________________________
Date: Wed, 19 Apr 2006 09:35:03 -0400
From: Lee Clawson
Subject: Re: Re: Prepress question
on 4/19/06 3:09 AM, Leila Joiner wrote:
I'll hang around and lurk, if you don't mind, and see
what I can learn from
other people's problems. And I'll be sure to let you
know if I manage to
get better than average results from Lightning Source.
Leila,
As soon as you get more info I'd like to know a what
kind of press they're using.....offset, digital or high speed copy ???
Thanks
Lee Clawson
___________________________________________________________________________
Date: Wed, 19 Apr 2006 12:14:53 -0700
From: Leila Joiner
Subject: Re: Re: Prepress question
Lee, it's digital POD. Sometimes it's the only solution
for small pubs like me who can't afford to print and warehouse a several
thousand books.
But digital's not the problem. I use a short run
digital printer who has excellent output on photos and illustrations, but
he doesn't print from PDFs and he prints all images at 600 dpi.
--Leila
___________________________________________________________________________
Date: Wed, 19 Apr 2006 22:39:12 -0400
From: Lee Clawson
Subject: Re: Re: Prepress question
Leila,
I thought is was digital. I still can't fathom why you
received such a compressed tone range. Using PDF work-flow should not have
any effect.
Good luck with this..........
Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________