The Color Key
"The color Key" - an article by CHROMiX President Steve Upton
The color Key
If you work long enough in or near the print industry you start to take CMYK for granted. Especially K.
In looking over my previous articles I noticed that I had yet to cover the topic of K so it seemed like it was time.
The K in CMYK stands for "Key", NOT black as many might have you believe. The Key plate, in traditional color separations, is the plate that holds the detail in the image. In CMYK this is usually done with black ink.
In the modern color-managed workflow, an RGB image has an associated profile so each RGB number combination can be converted to a defined Lab color. This is fairly straight-forward and repeatable. When creating a CMYK combination to represent that color on output, things get considerably more complicated.
Lets talk a bit about how color is created using CMYK. If you apply yellow ink to paper, your color range starts at paper white and then becomes more yellow and saturated with the more ink you apply. But once you get to 100% yellow there's nothing more you can do without adding other inks. If you are looking for a medium-dark yellow you now have a whole host of choices to get it. First, you can add cyan and magenta ink. They are both required in order to offset each ink's tendency to move the color toward green or red. The addition of cyan and magenta does darken the yellow but they are also, together, blue - which is anti-yellow. So this addition of blue desaturates the yellow ink quickly, limiting the range of dark yellow colors available. A second choice is to add black ink. As black is added, the yellow darkens but is not desaturated nearly as quickly. This can result in a greater gamut of dark yellows.
When CMYK colors are created in normal workflows, either or both of these techniques are used. In fact, for a single original RGB color, many different combinations of CMYK can be used to (theoretically) create the same color on press.
So how do we decide which one to use? How much black should be used instead of CMY? The answer, as you should already guess, is the classic color management answer: "It depends".
Lets cover a bit more technique and terminology first.
Cyan, magenta and yellow inks used for offset printing are not pure enough colorants to be mixable in equal amounts for gray. The SWOP standard expects 50% cyan and 40% each magenta and yellow inks to produce a neutral gray. So let's say we had a CMY color of 60, 50, 40. In theory, if we removed the components of the color that produced gray (50,40,40) we would then be left with (10,10,0). If we then add enough black ink to bring us down to the same darkness (about 54% K), we have the CMYK combination of 10,10,0,54 that should appear very close to the original color yet is composed of VERY different amounts of ink! The total ink used drops from 150% to 74% AND changes from the more expensive color inks to black ink.
This kind of color replacement is called Gray Component Replacement (GCR). If the color range affected by this replacement is limited to dark, near-neutral colors, then it is called Under Color Removal (UCR). GCR, on the other hand, can be applied to neutral and non-neutral colors that are either light or dark. That brings us back to the "how much?" and "when?" questions.
It's probably best to cover this next section in point form:
Black start / max / width
In ICC profiles, the black generation method and amount is chosen at the time the profile is calculated and "baked" into the profile. For flexibility, we suggest calculating several profiles from the same measurements; each with different black generation settings (this is what we do with our ColorValet press profiling service). Black Start is the amount of cyan ink where black starts replacing other inks. For instance, a black start of 10 means that when cyan ink is less than 10%, CMY will be the only inks used to create colors, but for any colors where cyan is greater than 10%, K will replace some component of the color. Black Max is how high you want the K level to get in the resulting color. If you find your shadows are plugging, reducing Max K can help. Finally, black width is how far "out" into the saturated colors you want black to be substituted. A low black width will limit K substitution to near-neutral colors (similar to UCR). A high black width will allow substitution much farther out into the saturated colors. If you find your saturated colors look muddy, try lowering your black width.
Dot Gain vs color stability
Here are a few basic facts about printing (unsubstantiated, I admit, there's just not enough space here). Color casts, especially in neutrals, are the single most common (color) reason customers refuse print jobs. Dot gain varies due to press fluctuation the most in the mid tones of each channel - ranging from 40-60% - and the variance can differ from channel to channel. This means that when neutral colors have CMY inks that are in the range of 40-60%, they have a fairly good chance of shifting on press. Our example color of 60,50,40 has all three inks in the "danger zone" and so could be quite unstable on press. The converted color of 10,10,0,54 however, has only the black ink in the danger zone and changes in black dot gain won't create a color shift. (Thanks to Dan Remaley of the PIA/GATF for turning the light bulb on for me with this issue.)
Print job stability on press VS adjustment on press
With the above explanation in mind, you might think that GCR should be used heavily in all images, and yet many printers recommend light GCR or UCR. Why? Well, the very same reason you would use GCR - stability of color on press - makes it very difficult to change color on press. Many printers don't trust their customers' color management or separations and so they want the ability to adjust the job on press. GCR IS a very good thing. Just remember that if you use it, ensure that the color in the file is good (verified with an effective soft / hard proofing system) and that the printer can print according to the specs that defined your separation. If they are nowhere near SWOP, or your files are not well color managed, sending them high GCR files is going to cause problems. Or, more correctly, will limit their ability to correct problems. (of course, exchange the "you"s and "them"s if you are the printer)
In offset printing, the line screens for each channel tend to be the same or very close. When black is used in highlight colors, the dot size and shape is similar to the other inks and doesn't typically cause any problems. Inkjet printers using light cyan and light magenta are a different story. Highlight colors on inkjets are often built from CMY inks only. Inkjets that don't use Epson's UltraChrome inkset have no light black (gray) ink. If GCR is used in the highlight colors, the dark black ink exposes the true resolution of the printer, creating speckles that people often mistake for lowering of the printer's resolution. On these systems, use a black start value of 40-50% to keep K out of the highlights. This means that the careful balance of CMY is required for light colors - beware that color balance failure (often mistakenly called metamerism) might cause problems. Also, your profiling target had better effectively sample CMY-only near-neutral colors or your profiling software has little hope in finding how to print neutral highlights. I bet you've seen this problem as well.
Ink costs & drying time
This one is pretty simple. Less ink = less money AND less dry time. GCR = less ink so it saves both time and money.
If you think back to the dark yellow example above, recall that to build dark colors we could either add cyan / magenta to yellow OR add black. The difference to the size of the shadow gamut (the number and saturation of dark colors available) is significant. Black slowly desaturates yellow as it darkens it. In comparison, the blueish cyan/magenta mixture darkens and quickly desaturates yellow because it cancels it out. The result is many more saturated dark yellows when black is used rather than cyan/magenta. If you haven't realized it yet, dark-yellow is brown, and having full saturation in browns means wood, leather, and hair or flesh tones as they fall into shadow look MUCH better. If you are missing your browns, the profile will desaturate them into becoming more gray and you will be disappointed in your wood and leather and often see sharp transitions as faces and hair fall into shadow... sound familiar?
ColorThink Pro has a new graphing feature, Constrain Channels, which features sliders that allow you to "pull back" color channels individually. Pulling the K channel back to 0 is an interesting demonstration of how important black is to the gamut of a press. The gamut size is reduced by 20 to 30%.
Muddying saturated colors
The use of too much black in saturated colors can "muddy them up" on some printing systems. See Black Width above.
UCR vs GCR
Restricting black replacement to neutrals is what UCR is all about. GCR came along later and extended this technique beyond the neutrals and into the colors. If done correctly, GCR can be very effective and improve image quality. Early GCR didn't always work as expected and many fell back to UCR for safety. GCR is now at the point where it is reliable and effective, and UCR is falling out of use. As the function of UCR is a special case of GCR (dark neutrals only), I expect the term UCR will fall out of use and we will refer to all black replacement as GCR.
100% K is not typically the darkest color you can get from a printer or press. When Cyan and other inks are added, the darkness can often drop noticeably (from an L* value of 19 down to 9 in SWOP TR001 for instance). The combined color can either be neutral or have an intended hint of color. Blue blacks are popular as well as warm blacks and tend to be used differently in different cultures and parts of the world. Some profiling packages allow you to select the combination of inks for "max black".
Standard vs Custom separation profiles
As I mentioned above, the black generation method and amounts are set at the time of profile generation. They are not changeable on-the-fly as many would like. This is a limitation of the current ICC profile architecture. The profiles that ship with Adobe publishing products are high-quality and some are based on published standards, such as US Sheetfed Coated v2, but there is only one profile for each printing condition. This means Adobe chose a middle-of-the-road group of GCR settings when the profile was originally created. Color images vary widely and different levels of black generation should be used for each image type. Images that are predominantly neutral should typically have more GCR than those with many saturated colors. Flesh tones benefit from GCR but too much makes them appear muddy on press. Because of the baked-in limitation of ICC profiles it's best to keep a group of press profiles on hand so you can vary your black generation optimally.
As you can see, the subject of black is large and I haven't really scratched the surface. Suffice to say that "that extra channel" adds a lot of depth to color printing both in the color sense and in the number of choices we have and the number of topics there are to study. I hope I've at least opened your eyes to some of the effects of black and black generation. As always, I'm open to comments and more information if you have it.
Thanks for reading,
Steve Upton October 18, 2005