Mixing Colors Digitally - Searching for Darks

The process of ‘mixing colors’ from pigments is much too complex for painting software to emulate for many reasons. For instance, artists achieve quite dark mixes from modern staining pigments. You can’t get those by mixing colors digitally.

An earlier post, “Mixing Colors Digitally—Introduction,” explained that mixing colors with physical pigments basically involves two processes: a physical process that’s subtractive, and a psychological process of viewers additively averaging the colors they see. Painting software doesn’t try to mimic either process; instead, it mixes colors by weight averaging their red, green, and blue components. In “Mixing Colors Digitally—Searching for Greens,” I shared two experiments that show how different the mixed greens produced by these disparate processes can be.

Those experiments involved non-staining pigments. In this post I’ll report experiments with staining pigments—ones that bind quickly with and resist lifting from the paper. The experiments confirm that the dark colors mixed from staining pigments depart dramatically, but in a different way, from the colors we can mix digitally.

Displaying Mixed Colors Graphically

Once again, I’ll display the results of the experiments on a radial graph. The original colors and mixed colors will be converted to HSV (hue, saturation, value) color space. Then I’ll plot each color as a dot on the graph—the direction of the dot from the center indicating the color’s hue, the distance from the center showing its saturation, and the size of the dot showing its value.

For instance, the first experiment starts with colors that resemble Daniel Smith watercolors Quinacridone Gold and Phthalo Green (Blue Shade). The first color is (189, 123, 55) or #BD7B37; in HSV color space, its hue angle is 30, saturation is 181, and value is 189. The green color is (60, 103, 73) or #3C6749; its HSV hue angle is 138, saturation is 106, and value is 103. Here’s how I display those colors graphically:

Dots for Colors Resembling Quinacridone Gold and Phthalo Green (Blue Shade)

What does the graph show? First, positions of the dots show these colors are approximately triadic hues, about one-third the way around the hue circle from each other Second, the dots being away from the edge shows the colors are somewhat unsaturated, especially the green. Finally, the larger size of the orange dot shows that color is considerably lighter than the green.

Experiment 3: Quinacridone Gold and Phthalo Green (Blue Shade)

In this experiment, only the second paint is staining. Daniel Smith’s Quinacridone Gold watercolor is a convenience mixture of PO 48 and PY 150 pigments, which the company markets as a “low staining” replacement for Raw Sienna. Their Phthalo Green (Blue Shade) watercolor is pigment PG 7, which is “super staining.” 

I sampled mixtures of these pigments from the mixing charts by Australian watercolorist Jane Blundell, who is an avid explorer of color mixing. Here are the colors she mixes from the pigments: 

How does this compare to mixtures created by painting software? The next graph plots some mixtures (with 9:1, 8:2, 7:3, and so on, ratios of the yellow to green) that can be created digitally:

The software’s weighted-averaging mixing path is more uniform than Blundell’s mixing path. Furthermore, the software path passes closer to the middle of the circle, which is achromatic gray: in other words, its mixed colors are less saturated than those in Blundell’s path.

How can we explain the outward swerve (increased saturation) in Blundell’s mixes? Part of the reason is the subtractive process that’s prominent in color-mixing with pigments. We can see this by comparing the paths above to this purely subtractive mixing path created in software:

(Technically, the software can’t mix colors subtractively on a single layer. Rather, it blends them when you paint the colors on separate layers and set the top layer to Multiply blend mode. This emulates shining full-spectrum light through two colored gels. You can vary the blend by choosing which color is on the top layer in Multiply mode and adjusting the opacity of that layer.)

The mixed colors on the purely subtractive mixing path are quite saturated (the path swings very wide toward the edge of the circle), but very dark (the dots are small). Thus, the subtractive process in pigment mixing can explain why Blundell’s pigment-mixed colors are saturated, but not why they are light in value. I suspect the pigments mix to relatively light colors because we additive-average the various colors reflected from them, a psychological process that produces lighter unsaturated tones.

Experiment 4: Anthraquinoid Red and Phthalo Blue (Red Shade)

In experiment 3, only one of the pigments was staining. Here both pigments are staining, and they produce some very dark mixtures. 

The first color resembles Daniel Smith’s Anthraquinoid Red watercolor, made from PR 177, which the company says is a “super-staining vat pigment.”  This red is (173, 59, 85) or #AD3B55; its hue angle is 346, saturation is 168, and value is 173. The second color resembles the Daniel Smith’s Phthalo Blue (red shade) watercolor, which is made from the “high staining” pigment PB 15:6. This blue is (58, 92, 156) or #3A5C9C; its hue angle is 219, saturation is 160, and value is 156. 

Notice the unusual path of colors that Jane Blundell mixes from these pigments:

Once again, the pigment mixing path is a ragged. The mixes are quite dark (the dots are remarkably small) and very unsaturated (the path bends inward toward the center of the circle). 

The software mixing path, by contrast, is predictably straight and the mixed colors aren’t so dark:

In the first three experiments I could explain the curve in the pigment mixing paths, in part, by the curve of the purely subtractive mixing path. (Recall that color mixing with pigments is largely a subtractive physical process). But that doesn’t work in this case. The purely subtractive path swings outward, which is opposite from the inward swing of Blundell’s pigment path:

The subtractive process in pigment mixing helps explain the clustering of mixed colors toward either end of the path and the darkness of the mixes. But it can’t explain the unsaturation of those mixes (that is, the inward swing of the pigment path). 

The mixing paths of two staining pigments typically have an inward swing. I don’t know why that happens, but it’s not totally unexpected. Staining paints are notorious for making ‘dead’ (that is, very unsaturated) dark colors when one is glazed over another. Apparently, something similar happens when they are mixed.

Conclusions

Mixing colors with physical pigments is complex and quirky; it’s largely dominated by a subtractive physical process but also involves a psychological process of viewers additively averaging the colors they see. Mixing colors digitally, on the other hand, is a simple process of the software weight-averaging their RGB components. 

I ran two experiments that compare the software’s mixing process with Jane Blundell’s mixes of staining pigments. The experiments suggest:

1. Mixing colors digitally has a more predictable mixing path than mixing colors with pigments. 
2. Typically, digitally mixed colors are less saturated than those mixed from pigments. But when two staining pigments are involved, the result is reversed: the digitally mixed colors are more saturated than those mixed from pigments.
3. Typically, colors blended subtractively in software are much more saturated and darker than those mixed from pigments. But when two staining pigments are involved, the pigments are equally dark.
4. Mixing colors digitally can’t match the dark colors derived from mixing staining pigments.

***

Thanks for reading!

I hope that you enjoyed this post and that it inspires you to enjoy digital painting. If you find this post helpful, please share it with your friends. And please send me your insights on digital painting and suggestions for Digital Paint Spot.

Bob Kruschwitz