Rainbow and Smooth Gray Chart

Koz's Calibration Corner

Rainbow and Smooth Gray Chart

256/100 (2.9K Picture)
Smooth Grayscale

527/325 (30K Picture)

Rainbow.jpg and Gray256b.jpg are the "torture chamber" graphics...for equipment and operators. Between them, they will test your graphics system color accuracy as well as possibly your patience.

The original intent was to provide two perfect test patterns and then samples of how they would look on different systems. I couldn't do it. It's just too crazy. I'm just preparing the two patterns and writing descriptions of generally what happens to them on different systems.

Not enough colors

Until relatively recently, all the colors of the rainbow (17 million of them) was a painful and expensive thing to achieve on your computer. Most computers make do with less than that. The variations are:

  • 17M Truecolor or 16,777,216 different colors, 8-bits per color, r/g/b. (24-bit color)
    In Mac land, this is "Millions". This system, properly done, produces photorealistic displays that may rival commercial TV reception and wet process photography. Unless you're on a high-end SGI workstation with 10 bits per color or higher, this is a good as it gets.

  • 65K HiColor with 65,536 different colors; 5-bits red, 6-bits green, and 5-bits blue.
    In Mac land, this is "Thousands."

  • 32K HiColor with 32,768 different colors; 5-bits per color, r/g/b.

  • Enhanced VGA or 256 different colors; 8-bits total.
    In Mac land, "Hundreds." There are a lot of general purpose machines that ship with this display method. It's cheap, fast and 256 colors can be mapped and dithered in MS Windows for an "OK" display. Usually a lot better than you think it's going to look.

  • VGA, plain and simple, with 16 colors; 4-bits total.
    This number of colors must be mapped and dithered strenuously to produce only dreadful results...always.

    Different systems cope with restricted colors differently. MS Windows simulates all the colors of the rainbow by "dithering" whatever is available. A looking-through-a-screen-door "Bayer" dither will be recognized immediately by anyone doing graphics with a 16 or 256 color restriction on a small screen in Windows. If you throw enough pure red and blue dots on the screen in an alternating pattern and then step back from the monitor far enough, you see purple. It's the best compromise.

    Other systems are not so talented. I know of a very old color system that looks at the graphic and picks the 16 most popular colors. It then takes all the rest of the colors and converts them into whichever of the 16 is closest. The "slash and smash" method. This gives you three people's faces with one single fleshtone. If the faces weren't very large, you might not even get that! This particular machine is now used for bookkeeping.

    There can be unusual restrictions in your color system. There are some web browsers out there that will download and save complex graphics in full color, but will only display them in 256 colors. To get the best results, you have to display the saved graphic using a program outside of the browser.

    Then, there's human error. When I started this project, I noticed that the gray pattern looked far worse than I thought it should. I went into Windows setup and discovered my display system was running 32K HiColor instead of 17 million Truecolor. Go figure.

    The patterns

    Both graphics were created in Photoshop 3 for Windows on a system with 17 million colors the whole way. Both patterns were saved as gracefully compressed JPEG files with the "lossy", as much as possible, turned off.

    Gray256b.jpg is exactly 256 pixels wide and has 256 linear levels of gray. You can tell which gray you're on by which pixel you're viewing. (remember, the count goes from zero to 255) It's 100 pixels tall. There is no top-to-bottom shading.

    Rainbow.jpg is not a perfect rainbow. It is based on a 6-point, fully saturated, chromaticity model. The R,M,B,C,G, and Y points are accurate, but the points in between, while arithmetically centered, are probably not chromatically accurate. This model results in a very subtle bright band down the middle of all three color stripes. I'm doing more research and will probably post a better rainbow later. Until then, this one is painful enough.

    Rainbow.jpg has 1530 different colors not counting the gray (127 r/g/b) border. There are two identical perfect blues; far upper left, and far lower right. Similarly, there are two reds, and two greens. All other colored pixels are distinct, different colors. Each stripe is 100 pixels tall. There is no top-to-bottom shading.


    17M Truecolor systems will display each test pattern perfectly smoothly withno hint of scratching, striping, or banding, even under high magnification.

    65K HiColor systems will be very, very good, but under magnification will start to fall apart. The top band of the rainbow, about half-way between blue and magenta may show faint vertical stripes. The gray pattern will definitely show vertical banding and under magnification, will show subtle shifting between purple and green which betrays HiColor's reliance on that extra bit in the green channel.

    32K HiColor systems will have more pronounced banding, but will not have the green/purple thing in the gray pattern.

    256 color Enhanced VGA systems will, depending on the system, look gritty, screen-door-ish, approximately OK, or garish, pasty, and cartoonie awful.

    16 color VGA will produce recognizable graphics only if someone told you what they were supposed to look like ahead of time.


    The truely compulsive/masachostic will immediately try to print these patterns. "The results", as a software designer once said, "will probably be significant."