From: Remaley, Dan
Date: Tue, Oct 16, 2001, 10:04 AM
RE: [colortheory]
Hi, I have just joined this forum on color. I work in the Process Control group at GATF. In my 'previous' life I worked for 14 years at a high end color trade shop in Pittsburgh where I was "Dirty Harry", having to justify to printers why they couldn't match my scans and proofs! Most recently I have presented seminars for printers and PIA (Printing Industries of America) members across North America.
I hope I can bring some of my process control experience to this group. Color lithography is difficult to control without some form of standardization like SWOP or GRACoL.My personal belief is that the use of GCR near 70% (on a high end scanner) is a good choice - understanding that the control of gray balance, dot gain (TVI), solid ink density, and Print Contrast are very important. Color bars on proof and press that measure 50% dot gain and gray balance are needed. There are many advantages of GCR -
#1-press deviation - the presses standard movement is (+/- 2% at the 50%). With GCR we can increase this window to +/- 4% WITHOUT changing the color.
#2- colors are more 'pure' because the 'graying' component (cyan in the example of a Red color) are replaced with black (no color cast).
#3-print with less yellow, mag, cyan inks - more Black ink is used. If you believe that color correction takes place at press - GCR is not for you! You cannot change the color at press with high GCR. In fact, you can't 'color correct' at press - but few people believe this. I have a great little handout showing the differences using GCR and gray balance issues, I'll be happy to send you a copy if you drop me your address.
The 'secret' to color printing - gray balance and GCR!Dan Remaley
Process Control Tech.
p:412.741.6860 x450
f:412.741.2311
From: Dan Margulis,
Date: Wed, Oct 17, 2001, 7:51 AM
RE: [colortheory] Re: GCR
Dan Remaley writes,>>My personal belief is that the use of GCR near 70% (on a high end scanner) is a good choice - understanding that the control of gray balance, dot gain (TVI), solid ink density, and Print Contrast are very important.>>
"GCR near 70%" has no meaning without further information. It's as if I told you that the street number of my house is #9, and asked you to find it without telling you what street and city it was located in.
Percentage of black generation will vary depending on darkness. To understand what kind of GCR you are advocating, we need to know roughly how much black you might generate if the CMY-only values were, for example, 6c4m4y, 20c15m15y, 40c33m33y, 60c50m50y, 80c70m70y, and 95c85m85y. Any sane method is going to replace with a higher percentage of black as the CMY gets higher.
AFAIK the expression of GCR as a "percentage" was picked out of a hat by somebody at Crosfield in the early 80s, just as its expression in Photoshop as "Light" "Medium" or "Heavy" was picked out of a hat by Thomas Knoll about ten years ago. Neither one means anything in isolation. However, we all use Photoshop, whereas few of us use drum scanners. Therefore, I'd suggest that saying "GCR in the general style of Photoshop's Heavy setting"--which is, I think, what your "70% GCR" means--is more informative.
>>There are many advantages of GCR ->>
There are, but it also carries substantial baggage, which explains why its rate of adoption is so limited.
>>#1-press deviation - the presses standard movement is (+/- 2% at the 50%). With GCR we can increase this window to +/- 4% WITHOUT changing the color.>>
If extreme color fidelity is important, AND the output process is fully understood and well controlled, heavier GCR is in order. However, the flip side is that there may be major changes in lightness/darkness. And it's well known that viewers are a lot more willing to forgive minor color variation than they are an inappropriate luminosity.
>>#2- colors are more 'pure' because the 'graying' component (cyan in the example of a Red color) are replaced with black (no color cast).>>
This is in fact the Achilles heel of GCR. If the calculations are done properly and process control is good, there should be no difference in the result. If not, too much cyan may create a color that's too purple, but too much black may create a neutral mud. The impact of incorrect printing of black is roughly four times greater than incorrect cyan, and that's before we take into account that black is much more likely to be printed in an unexpected way than cyan is.
We live in a real world, not one of theory. In the real world, few printers have the commitment to process control that eliminates most of the dangers of GCR. We don't want to hand these children loaded guns nor do we want to hand them subtle colors loaded with an ink that can kill us if they use too much of it.
>>#3-print with less yello, mag, cyan inks - more Black ink is used.>>
True, and the black ink costs less, too. In this economy we'd like to do whatever possible to help the printer keep his costs down, but until the printer starts offering discounts for GCR seps, most of us will probably think of quality first and the cost of ink later.
>>If you believe that color correction takes place at press - GCR is not for you! You cannot change the color at press with high GCR.>>
That's all true.
>>In fact, you can't 'color correct' at press - but few people believe this.>>
I don't, for one.
Dan Margulis
From: "Remaley, Dan", dremaley@gatf.org
Date: Wed, Oct 17, 2001, 8:52 AM
RE: RE: [colortheory] Re: GCR
Dan Margulis writes,
> "GCR near 70%" has no meaning without further information. It's as if I
> told you that the street number of my house is #9, and asked you to find it
> without telling you what street and city it was located in.
>
> Percentage of black generation will vary depending on darkness. To
> understand what kind of GCR you are advocating, we need to know roughly
> how
> much black you might generate if the CMY-only values were, for example,
> 6c4m4y, 20c15m15y, 40c33m33y, 60c50m50y, 80c70m70y, and 95c85m85y. Any
> sane
> method is going to replace with a higher percentage of black as the CMY
> gets higher.
>
> AFAIK the expression of GCR as a "percentage" was picked out of a hat by
> somebody at Crosfield in the early 80s, just as its expression in
> Photoshop
> as "Light" "Medium" or "Heavy" was picked out of a hat by Thomas Knoll
> about ten years ago. Neither one means anything in isolation. However, we
> all use Photoshop, whereas few of us use drum scanners. Therefore, I'd
> suggest that saying "GCR in the general style of Photoshop's Heavy
> setting"--which is, I think, what your "70% GCR" means--is more
> informative.
>> The percentage # (ie 70% GCR) means a replacement of Black = to
70% of the gray component (cyan in RED for example- cyan - being the graying
component). You can use 100% GCR removing ALL of the graying component and
inserting Black.Dan M.
"There are, but it also carries substantial baggage, which explains
why its
> rate of adoption is so limited."
>> The reason its not used is because it isn't understood!
> >>#1-press deviation - the presses standard movement is (+/- 2% at the
> 50%).
> With GCR we can increase this window to +/- 4% WITHOUT changing the
> color.>>
Dan M.
"If extreme color fidelity is important, AND the output process is
fully
> understood and well controlled, heavier GCR is in order."
>> Absolutely true.
Dan M.
> "However, the flip
> side is that there may be major changes in lightness/darkness. And it's
> well known that viewers are a lot more willing to forgive minor color
> variation than they are an inappropriate luminosity."
>> Absolutely NOT true. The truth is, that we as an observer, will
tolerate lightness & darkness NOT color casts. Example: go to your monitor
and create a box with 50C 40M 40Y. Create new boxes with a change of +2% in
Cyan only, another with +2% Magenta only and a +2% yellow only. Note the hue
shift in each box. Now do the same thing - only increase all colors 2% or
decrease by -2%. This is a 'contrast' change and much more acceptable to the
human eye. Agree?
> >>#2- colors are more 'pure' because the 'graying' component (cyan in the
> example of a Red color) are replaced with black (no color cast).>>
Dan M.
"This is in fact the Achilles heel of GCR. If the calculations are
done
> properly and process control is good, there should be no difference in the
> result. If not, too much cyan may create a color that's too purple, but
> too
> much black may create a neutral mud. The impact of incorrect printing of
> black is roughly four times greater than incorrect cyan, and that's before
> we take into account that black is much more likely to be printed in an
> unexpected way than cyan is."
>
> We live in a real world, not one of theory. In the real world, few
> printers
> have the commitment to process control that eliminates most of the dangers
> of GCR. We don't want to hand these children loaded guns nor do we want to
> hand them subtle colors loaded with an ink that can kill us if they use
> too
> much of it.
>> This is the 'catch', if you understand GCR and monitor
the correct printing of the black plate, through process control (dot gain,
density, print contrast). GCR seps print "better" than non-GCR. With GCR
scans the Black makes it or 'breaks it'! Without GCR ALL the colors need to
be process controlled Mag, cyan, yellow, (black is ghosted-no big deal).
Whenever I see a 'ghosted' black plate - traditional scan -
I know that all the color is made up of Y-M-C and I know there will be
problems matchiing color.
Also, remember, all color corrections are done prior to the
GCR conversion - in fact you can't color correct a GCR scan. (color is
saturated, black is detailed).> >>If you believe that color correction takes place at press - GCR is not
> for you! You cannot change the color at press with high GCR.>>
Dan M.
> That's all true.
>
> >>In fact, you can't 'color correct' at press - but few people believe
> this.>>
>
> I don't, for one.
>
> Dan Margulis
>> See if this makes any sense- You scan for gray balance and then
make 'selective color correction'.
This is based on a proof of some sort - measurable I hope. Now if
you changed the gray balance by let's say +4% in Magenta. What happens to
our reds, greens, blues etc. They ALL change.
You are now at press and you want more magenta in an area - (unless
it's the only image in that key zone) you are going to change the same
red,greens, blues as you would in scanning! Any image behind this one would
be affected also. Therefore you can't color correct at press.I have printed samples of this subject which I'll be happy
to send please forward me your mailing address.
Dan Remaley
From: Dan Margulis,
Date: Wed, Oct 17, 2001, 11:29 AM
RE: RE: [colortheory] Re: GCR
Dan Remaley writes,>> The percentage # (ie 70% GCR) means a replacement of Black = to 70% of the gray component (cyan in RED for example- cyan - being the graying component). You can use 100% GCR removing ALL of the graying component and inserting Black.>>
Without saying at what point of darkness this 70% would take place, the number is meaningless, because no sane method of GCR is going to take the same percentage of CMY out of a light color as it will of a dark.
40c100m100y is a red. Contrary to what you indicate, cyan is *not* the gray component; the gray component is a *combination* of approximately equal amounts of CMY which can be removed in favor of black. But if you're saying, add an amount of black such that 70% of the cyan is removed, fine, I understand that, you'll wind up with 10c85m85y30k or something along those lines.
If that 70% replacement were true everywhere, then 4c10m10y, which is also a red, would have to be treated in the same fashion, knocking cyan down to one and adding three points of black. Nobody in their right mind would do that.
Even in the first example, realistically nobody wants black as high as 30%. Traditional separations will put the black anywhere between 5 and 15%. 20% would, in my view, be a heavy GCR, and I would not recommend it in most cases. As we have noted earlier, though, there are certain advantages to having a black that high. But black is a very powerful ink, and 20% is going to give all the benefits one would possibly need in terms of aiding registration and preventing color shift. Going to 30% is a pointless gamble. All it does is give the pressman an opportunity to wreck the job.
>> Absolutely NOT true. The truth is, that we as an observer, will tolerate lightness & darkness NOT color casts. Example: go to your monitor and create a box with 50C 40M 40Y. Create new boxes with a change of +2% in Cyan only, another with +2% Magenta only and a +2% yellow only. Note the hue shift in each box. Now do the same thing - only increase all colors 2% or decrease by -2%. This is a 'contrast' change and much more acceptable to the human eye. Agree?>>
No. I am concerned with the evaluation of images, not of boxes. And serious observers of color have known for 500 years that correct luminosity is more important than color fidelity. Da Vinci commented on this, as did Chevreul and Goethe. And Ogden Rood, who IMHO is the most perceptive writer on color generally, wrote more than 100 years ago:
"We forgive, then, a partial denial of the truths of colour more easily than those of light and shade, which probably is a result of the nature of the optical education of the race. For the human race, thus far, light and shade has been the all-important element in the recognition of external objects; colour has played only a subordinate part, and has been rather a source of pleasure than of positive utility."
Dan Margulis
From: Martin Orpen
Date: Wed, Oct 17, 2001, 4:27 PM
RE: Re: [colortheory]
Dan RYou're going to have to adapt your posting style, or your mail software.
I found it practically impossible to read your responses to Dan M's post.
Trim the original post so that only the relevant text remains and then leave a line space (or two) before you begin your reply.
I'm sure that we would all benefit from your press knowledge -- if only we could read it!
Regards
--
Martin
Idea Digital Imaging Ltd - the "image" specialists
http://www.idea-digital.com
From: Stephen Marsh
Date: Sat, Oct 20, 2001, 9:04 AM
RE: [colortheory] Some mixed thoughts on separation techniques...(was GCR)
--- Dan Margulis wrote:> This whole topic has caused so many people so much
difficulty that I've
> devoted my current (October) Makeready column to it, where I
spend five
> pages discussing an image of a black cat.
>Dan, thanks for the reminder - I had not checked the EP site since the dust settled on the whole false profile thread.
Great article too - just at the right time. Cheers.
Now some comments...
Taken from your article for reference:
> "Furthermore, although the image itself is full of noise (look at what's happening in the yellow channel) the black isn't especially bad, and even if it were it wouldn't be a tragedy. We can still apply unsharp masking. With the black covered up by so much amorphous cyan, magenta, and yellow, we can afford to have it bloodcurdlingly harsh.
After applying the curve at top left and a touch of unsharp masking, the final black, right, seems too harsh. The formless CMY plates, however, soften it. Bottom left, a selective color move reduces the quantity of CMY inks in black areas."<
When using a prosumer Umax Powerlook scanner with bundled Binuscan automated 'colour enhancement' - I first came across this harsh black plate issue.
It seems that some separation software vendors approach to general separation is to separate with a short black then enhance and oversharpen it - knowing that the CMY plates will disguise the fact but still gaining from the exercise.
For example, Binuscan used this as a key quality feature for it's separation method in their software.
Not just in the case of black objects of primary importance - but as a general rule for all seps. Perhaps this is some 'Euro' approach?
-->
"Noise in scans, speckling in shadow areas, vertical streaking in images: We often receive comments that the separations produced by Binuscan often appear on film to be extremely noisy, especially on the black plate in shadow areas - this is typical of a Binuscan separation, and is very rarely visible when the image is actually printed [this type of separation gives far superior shadow detail and print contrast].
Binuscan deliberately searches for and emphasises areas of highlight and shadow detail that will normally be lost separating an image "conventionally", and it adjusts these areas using contrast and sharpening so that they are more clearly visible in the final image. One minor problem is that "detail" can include dust, reflections, marks on the glass or tranny and faint scanner noise that is normally invisible"
http://support.umax.co.uk/binuscan/technotes/iq02.htm
<--
From experimentation - if using a UCR or Lt GCR approach with good dot gain settings etc, this approach seems to work more often than not. Although it may not be safe to use as a default method, one can trot it out more often than not.
======
Here is a link to some basic advice on UCR and GCR from Binuscan, with some diagrams to help:
http://support.umax.co.uk/binuscan/technotes/nbi02.htm
======
Binuscan also use a method of histogram smoothing 'RECO' - where missing original data is interpolated.
Some form of 'mung & blur' perhaps?
Any thoughts?
support.umax.co.uk/binuscan/whitepapers/nb01.htm#RECO%2
0Technology%20-%20Lossless%20Correction======
Dan has quoted from Ogden Rood many times:
"We forgive, then, a partial denial of the truths of colour more easily than those of light and shade, which probably is a result of the nature of the optical education of the race. For the human race, thus far, light and shade has been the all-important element in the recognition of external objects; colour has played only a subordinate part, and has been rather a source of leasure than of positive utility. Ogden Rood"
To see what Ogden is talking about - take some common images, go to LAB and remove the data from the L channel via curves or whatever.
Then imagine attempting to drive or walk around with only chroma and hue data, with no luminosity. I would guess that the role of colour played little or no part in the evolution/creation of the eagles eye, for example.
You find this out when you start exploring LAB retouching.
I recently helped a user of another list with stitching three outdoor shots taken with auto exposure into a 'panorama' (an exercise in futility given the outdoor conditions and auto exposure). LAB seems the best correction/retouching option, if this 'less than ideal' panoramic task is to be attempted.
The AB are pretty much easy for correction - in attempting to match the sky and ground, it's the L channel and careful toning which makes or breaks the task.
Regards,
Stephen Marsh.
From: Dan Margulis
Date: Sun, Oct 21, 2001, 10:04 AM
RE: [colortheory] Some mixed thoughts on separation techniques...(was GCR)
Stephen writes,>>It seems that some separation software vendors approach to general separation is to separate with a short black then enhance and oversharpen it - knowing that the CMY plates will disguise the fact but still gaining from the exercise.>>
Outside of Binuscan I haven't seen this, but it does make sense. We don't
sell black channels, we sell composite images. If the "oversharpening" isn't detectable in the composite, then it isn't oversharpening.>>Binuscan also use a method of histogram smoothing 'RECO' - where missing original data is interpolated. Some form of 'mung & blur' perhaps?>>
I have seen this and my impression was that it was just a marketing effort to capitalize on those who think that a smooth histogram is necessary. There are an infinite number of ways of creating a smooth histogram out of a rough one in ways that don't damage the image discernibly. Which one Binuscan employs, I neither know nor care.
Dan Margulis
From: Dave Badger,
Date: Wed, Oct 24, 2001, 8:03 AM
Re: [colortheory] Black Sep
Throughout this thread, I've seen people talk of black shadow endpoints of anywhere between 70%-95%.- Is there a "standard" black endpoint for various TIC's (280%, 300%, 320%, etc.)
-And does a 95% black provide more tonal steps (i.e. detail) or just a richer looking shadow? (I know, it depends on what's going on with the three color, but I always thought those three plates didn't hold much shadow detail.)
-My scanner software has the GCR curve ending at 64% and the Conventional curve ending at 95%. Does this mean from 64% on up in the black sep, only black generated by the Conventional curves is being added in? Thanks
Dave Badger
From: Stephen Marsh
Date: Wed, Oct 24, 2001, 8:41 PM
RE: [colortheory] Re: Black Sep
--- In colortheory@y..., Dave Badger wrote:
> Throughout this thread, I've seen people talk of black shadow endpoints of
> anywhere between 70%-95%.True, it depends on many factors.
Photoshop's built in custom CMYK settings default to 100% K, and can be changed to any figure.
Pre built profiles may have any figure - a common one I use for tough images has a 90% as the pre defined limit.
Even with the specified max black limit - your sep may get less K than expected, depending on how dark the original RGB or LAB shadow is.
I have not found any transform to CMYK that does not need the highlight or shadow endpoints - specifically the K plate of shadows 'massaged'.
> - Is there a "standard" black endpoint for various TIC's (280%, 300%, 320%,
> etc.)Not IMHO - this is image and output specific.
> -And does a 95% black provide more tonal steps (i.e. detail) or just a
> richer looking shadow? (I know, it depends on what's going on with the three
> color, but I always thought those three plates didn't hold much shadow
> detail.)Both of the above would be true.
It depends on three factors - the start point where black is generated, the maximum point of black density _and_ the slope of the curve which defines how the black channel is generated. Most are not linear, and take a 1.8-2.2 gamma drop.
The start/end point for black generation will define the dynamic range of the plate, while the curve of the black generation will affect the midtones and other tones to a lesser extent.
Photoshop has these parameters built in, unlike other software with more control of the separation process (unless you want to explore the custom GCR option of Photoshop, which seems poor compared to all the other curve functions).
> -My scanner software has the GCR curve ending at 64% and the Conventional
> curve ending at 95%. Does this mean from 64% on up in the black sep, only
> black generated by the Conventional curves is being added in?Not sure what you are saying here.
You are using GCR and not UCR, correct?
Max black could commonly be 64 or 95%?
What is the start point for K generation (based on the grey ramp), what amount % of GCR is in use etc?
Am I correct that you are attempting to understand how your scanner separates and makes the K plate? If you can compare the grey ramps of both software, you might be able to get a better understanding of what is going on.
Regards,
Stephen Marsh.
From: Dan Margulis,
Date: Fri, Oct 26, 2001, 8:08 AM
RE: Re: [colortheory] Black Sep
Dave Badger writes,
>>- Is there a "standard" black endpoint for various TIC's (280%, 300%, 320%, etc.)>>
No, although if the TIC (total ink limit) value is low (as for a newspaper) you'd certainly have to have a black ink limit of 80++ as otherwise it would be impossible to achieve a dark enough shadow and still stay within the ink limit.
>>-And does a 95% black provide more tonal steps (i.e. detail) or just a richer looking shadow? (I know, it depends on what's going on with the three color, but I always thought those three plates didn't hold much shadow detail.)>>
It depends on what's going on with the other three.
A reasonable rule might be the following.
Where there is detail in the shadows or 3/4tones but it isn't especially important to the image, use 70-75% black.
Where the blacks and 3/4tones have no detail of any importance, use 75-85%.
Where the whole image depends on detail in very dark shadows (e.g. a picture of a black cat) use an incredibly harsh-looking black channel that ends at 95%.
Dan Margulis
From: Gordon Pritchard
Date: Fri, Oct 26, 2001, 2:03 PM
RE: RE: [colortheory] Black Sep
RE: Dan Margulis wrote:
> No, although if the TIC (total ink limit) value is low (as for a newspaper)
> you'd certainly have to have a black ink limit of 80++ as otherwise it
> would be impossible to achieve a dark enough shadow and still stay within
> the ink limit.
>
> Where there is detail in the shadows or 3/4tones but it isn't especially
> important to the image, use 70-75% black.
>
> Where the blacks and 3/4tones have no detail of any importance, use
> 75-85%.
>
> Where the whole image depends on detail in very dark shadows (e.g. a
> picture of a black cat) use an incredibly harsh-looking black channel that
> ends at 95%.
>
=========================================================Please explain this in more detail. Why would you restrict the black at all. It would seem that you would have more color stability in your shadows by using black than by using three strong process colors. I.e. it would be better to replace, with black, as much of the process colors in the shadows as you could while still maintaining TAC. (Total area Coverage)
thx, gordo
Gordon Pritchard
Commercial Print Specialist
CreoScitex
Vancouver Canada
T: 604.451.2700 ext 2870
C: 604.351.2437
http://www.creoscitex.com
From: Dan Margulis,
Date: Sat, Oct 27, 2001, 9:17 AM
RE: RE: [colortheory] Black Sep
Gordo writes,>>Please explain this in more detail. Why would you restrict the black at all. It would seem that you would have more color stability in your shadows by using black than by using three strong process colors.>>
Color stability is much less important in shadows than elsewhere in the image because our color perception decreases as colors get darker. Luminosity variation becomes much more important in the shadows because instead of making objects merely too dark, it can wipe out detail in them altogether. So, if there is a soft gradation from three-quartertone to shadow, running a heavy black is dangerous.
Also, dot gain is less in the shadows than in the midtones, making an unexpected shift in color less likely. The typical example that causes trouble for advocates of greater black is a woman with long dark brown hair. Any color variation is likely to be acceptable *except* if the hair turns black. That is the risk that a heavier black plate would take unnecessarily.
This follows closely a recent exchange I had with Dan Remaley. The advantage of printing with more black generally is that it reduces variation in color. The disadvantage is that it risks greater variation in luminosity. The bigger disadvantage is that if there isn't good process control and good color management, many more things can go wrong with a heavier black than without one.
For these reasons, I generally prefer a skeleton black, but I agree that there are certain types of images where the advantages of the heavier black outweigh the disadvantages. However, the advantages largely evaporate in the shadows.
>>I.e. it would be better to replace, with black, as much of the process colors in the shadows as you could while still maintaining TAC. (Total area Coverage)>>
There's a case for doing this where the shadows aren't important either for color or for detail. And, when the treatment of the shadow detail is absolutely critical to the success of the image, there's no alternative to a super-contrasty black ending up in the 90s coupled with suppression of the CMY. But for the common, garden variety of shadow, more black equals more problems.
Dan Margulis
Adobe Photoshop training classes are taught in the US by Sterling Ledet & Associates, Inc.