Wider coordinates. Not necessarily HDR.

BT.2020 describes a UHD system for production and international exchange. It covers wider primaries as well as picture dimensions, frame rates, signal structure and quantization.

This chapter uses BT.2020-2. For grading, distinguish the system from one of its components: RGB gamut. UHD resolution does not automatically imply BT.2020, and selecting BT.2020 does not automatically make a file HDR.

BT.2020 primaries are R (0.708, 0.292), G (0.170, 0.797), B (0.131, 0.046), with D65 white. This basis differs from P3 and Rec.709, so identical linear RGB generally produces different XYZ.

The laboratory converts a Rec.2020 colour through XYZ into Rec.709 or P3. It retains negative components and values above one: these indicate that the bounded destination RGB model cannot reproduce the stimulus unchanged.

Rec.709, Display P3 and BT.2020 share reference D65. A purely colorimetric conversion between their linear coordinates with matching normalization needs no separate white adaptation, but it does need an RGB basis change.

Decode transfer encoding before applying the matrix and encode afterwards as required. Gamut conversion, transfer conversion and white adaptation are different steps even when an application combines them in one control.

BT.2020-2 includes its own relative transfer characteristic. HDR television is addressed by BT.2100, which uses PQ and HLG with wide primaries. These are not two names for the same specification.

A primary tag alone cannot tell you whether the signal is SDR, PQ or HLG. Correct interpretation also requires transfer, levels and signal-representation information. Changing gamut cannot add dynamic range to the source.

The relative OETF has a 4.5E linear branch and an αE^0.45 − (α − 1) branch. BT.2020 Table 4 gives smooth-join conditions and practical rounded constants: α = 1.099, β = 0.018 for 10 bits; α = 1.0993, β = 0.0181 for 12 bits.

Do not confuse this function with PQ or HLG. Do not silently substitute constants from another document either: slightly different rounding can affect numerical tests near the breakpoint even when curves look similar.

BT.2020 specifies 10- and 12-bit component coding. Bit depth determines the available digital steps; primaries determine the basis in which those steps are interpreted.

Higher precision helps gradients and transforms, but does not expand gamut by itself. Converting an already damaged 8-bit gradient to 10 bits cannot recover missing original values. The stage at which rounding occurred also matters.

Nominal 10-bit RGB/Y′ black and white are 64 and 940; for 12 bits they are 256 and 3760. Neutral chroma codes are 512 and 2048 respectively; chroma bounds are considered separately.

Range maps normalized quantities to integers. It does not change primary coordinates. Do not stretch levels by appearance when the problem is the decoder’s incorrect range interpretation.

For conventional nonlinear BT.2020 component coding, luma is Y′ = 0.2627R′ + 0.6780G′ + 0.0593B′. These are not the Rec.709 coefficients. The matrix must match the chosen signal representation.

An incorrect matrix may be less evident on neutral ramps than on saturated colours. Test both grey and coloured samples. Adjusting Saturation in a grade cannot systematically repair incorrect matrix interpretation.

BT.2020 defines conventional non-constant-luminance and separate constant-luminance variants. The former derives luma from already nonlinear RGB; the latter treats linear luminance before its corresponding encoding.

The branches cannot be exchanged by renaming YCbCr. Establish which variant the encoder produced and the decoder supports. BT.2020 without matrix information leaves that unresolved. This chapter’s lab uses linear RGB and does not simulate a component codec.

Chroma subsampling describes the spatial grid of colour-difference samples. 4:2:0 and 4:2:2 are not gamut names. Bit depth, sampling structure and colorimetry answer different questions.

Check fine coloured graphics and edges for reduced chroma resolution, saturated colours for gamut limits, and smooth gradients for quantization. A single face shot cannot replace these distinct tests.

A signal coordinate space may exceed the colours actually used. Material limited to P3 during mastering can be carried in a BT.2020 representation. That is not a reason to stretch saturation to every vertex of the larger triangle.

The lab first calculates a coordinate conversion, then separately shows hard clipping of destination RGB to 0…1. Its post-clipping XYZ difference is a numerical model error, not perceptual ΔE. Production gamut mapping may behave differently and needs separate evaluation.

Accepting a file and physically reproducing every encoded chromaticity are different capabilities. A display may correctly recognize BT.2020 and map it into a narrower panel gamut.

A two-dimensional diagram omits luminance-dependent limits of saturated colours. Neither input compatibility nor one coverage percentage verifies an entire HDR result. The browser preview is also limited to its stated sRGB path and does not measure your display.

Define input and output with gamut, transfer, image state, range and format. In managed projects, establish where technical transforms already occur.

For a narrower Rec.709 output, choose appropriate tone and gamut rendering, then inspect saturated objects and highlights in motion. Do not merely change an export tag to Rec.709 after working in BT.2020. Metadata describes a result; it does not perform missing processing.