This is from Reddit somewhere. > If you can measure your shoe or mop a floor, you can understand color spaces. First things first: _color_ is a _subjective_ characteristic of light that can be measured _objectively_, and is _perceived_ by humans but can be _created_ by devices like prisms, plants or Panasonic flat panel displays. Certain colors are so extremely different from each other that they can’t be effectively described subjectively; these are called _primary colors_ or _primaries._ Forget transfer functions/gamma curves for now, they’re not important. > > _Color space_ is a complex, intimidating concept related to limits and calculus, which aren’t subjective at all. Think of it as _the set consisting of every color that can be created through relative combinations of intensities of each defined primary extreme._ Scary, right? Nah. It’s just a bucket of light. > > Three points plotted on an x-y axis can describe the physical characteristics of the primaries in most practical color models (more on that concept in a moment). Every point within the triangle whose corners are those points _partially_ describes a color created by a unique combination or _permutation_ of those primaries. Every possible (or in some cases, every _practical)_ intensity of these primary permutations can be represented as a set of points on a _line segment._ Testing has indicated that the intensity of a color is independent of the relative combination of its primaries. Where the color’s point is on the x-y axis doesn’t matter in terms of brightness. It’s impossible to plot all the intensity values on that same 2D x-y axis all of the primary extremes share. You have to plot in a third dimension and extend your curve on a non-coplanar axis, which ends up describing a _volume_ or _space._ That’s why you see pedants talking about color volumes sometimes. Now, most color professionals are used to seeing a color space represented as a triangle on an x-y graph, but that’s only a cross-section at a given intensity; a full graph would look more like a tapered bucket full of light. Rec. 709, P3D65 and ACES are all color spaces. > > A _color model_ is _how primaries are specified mathematically._ If you were going to design a bucket full of light, how would you describe its shape and size? A cylinder defined by a function involving length, width, height and pi? A solid or conic section defined by bottom radius, pi, height and top radius? A conic section defined by bottom radius, pi, height and angle to top vertex from bottom edge? Are you measuring in centimeters, millimeters, nanometers or size-6 shoes? > > It sounds silly, but precise and accurate measurements of the mathematical relationships between the same primaries and intensities, _no matter how those primaries and intensities are defined,_ can yield the same tapered bucket of light we call a color space. RGB, YCrCb and LMS are all color models, and they are all interconvertible given enough precision and accuracy in the measurements of their color values. (32-bit floating-point values offer enough precision to craft accurate conversion equations for all _practical_ purposes... which leads us to consider gamut.) > > Color gamut is _the subset of all the possible values of a color space as limited by the physical precision of the implementation of any given color model._ A bucket full of light made with millimeter tolerances will be much closer to the Platonic bucket envisioned by the designer than a bucket with centimeter tolerances, or with size-6 shoes tolerances. > > Because we live in a universe subject to chaos, entropy and decay, a device’s color reproduction is limited by the physical characteristics and imperfections of the primary emitters, reflectors and refractors. When it comes to negative timing or video grading displays, color gamut can be thought of as the density of primary permutations and intensities that are _practically_ possible for a given device. For example, at the extreme minimum of gamut, consider a traffic light, with only four colors (“off” can be a color, as anyone who’s been in a session with a jerk director can attest). > > More familiar are the typical 8-bit sRGB Dell or Asus computer monitors most of us first used to explore coloring video. Modern LCD flatscreens can typically display 256 separate values of each primary for a gamut of about 16 million distinct colors. But we all drool over the thought of the 10-bit+ multi-laminate LED panels of the Sony BVMX310, which can display at least 1024 separate values of each primary in sRGB for a total of over a billion distinct colors. Same color space, same color model, different color gamuts. > > But the practical part is what’s important with gamut. Even the legendary X310 is no match for physics—try to use it on the CRT transfer function setting and a set of built-in fans will kick into high gear. Any intensity value produced by a signal and a transfer function that could result in observed screen brightness over 1000 nits will be clipped internally prior to display instead of being allowed to pull enough power to trip a circuit breaker, or indeed melt the screen and display cabinet.