Digital signal
A decomposition of digital video into independent layers: primaries, transfer, RGB↔YCbCr matrix, full/narrow range, bit depth, chroma sampling, codec and metadata, with diagnosis of common Resolve errors.
The phrase '10-bit 4:2:2 Rec.709' describes several independent properties at once. None replaces the others. Correct decoding requires reconstructing the complete signal state rather than guessing one dropdown.
Quick reference
Independent interpretation layers.
Spatial chroma resolution, not bit depth/gamut.
VIDEO SIGNAL STATE
A useful model is: colorimetry/primaries → transfer → representation/matrix → numeric range → bit depth → chroma sampling → codec/container metadata. For example, 10-bit describes quantisation precision, 4:2:2 describes spatial chroma sampling, Rec.709 describes colorimetry/transfer context, and ProRes/H.265 describes the codec. These are different axes.
Video can be stored as RGB/R′G′B′ or Y′CbCr. Y′CbCr is built from nonlinear R′G′B′ using a standard-specific matrix and colour-difference components. With correct 4:4:4 sampling and sufficient precision, both representations can describe the same visual signal without one name being magically superior.
BT.709 uses Y′=0.2126R′+0.7152G′+0.0722B′; BT.2020 NCL uses 0.2627/0.6780/0.0593; SD BT.601 uses 0.299/0.587/0.114. If a decoder applies the wrong matrix, luma/chroma relationships shift and systematic hue/luminance errors appear, easily mistaken for a bad grade.
In common 8-bit narrow video, nominal luma black/white are 16/235 and chroma nominal extremes are 16/240 with neutral at 128. In 10-bit they are 64/940 for Y′ and 64/960 for Cb/Cr with neutral at 512. Full range uses 0…2^n−1. Range does not change primaries or gamma; it only defines how the normalized signal is packed into integers.
If a full-range signal is mistakenly read as video/narrow, black/white can clip and contrast becomes excessive. If narrow range is read as full, black lifts, white falls and the picture looks washed out. This is a numeric remapping issue, not a colour-space transform. Resolve exposes Data Levels interpretation/export settings for this reason.
8/10/12-bit provide 256/1024/4096 possible integer values per component before range restrictions. More bits reduce quantisation step size and give transforms more numerical headroom. But 10-bit does not mean HDR, wide gamut or 4:2:2; each of those attributes is chosen separately.
4:4:4 keeps chroma on the full sample grid, 4:2:2 halves chroma resolution horizontally, and 4:2:0 halves it horizontally and vertically. The luma grid remains full. Subsampling is therefore most visible on fine coloured detail, key edges, UI/text and repeated encode/decode cycles.
A sampling ratio describes how many chroma samples exist, but not always their exact spatial phase/location relative to luma. Codec/VUI metadata can signal chroma sample location. Incorrect resampling/siting can shift coloured edges by fractions of a pixel or a pixel, which is especially visible on graphics and keying.
ProRes, DNxHR, H.264/H.265/AV1 define compression and coding tools, but a codec can have different profiles/pixel formats. The word H.265 alone does not tell you 8/10-bit, 4:2:0/4:4:4, PQ/HLG or full/video range without inspecting metadata and actual stream parameters.
Modern bitstreams/container tags can carry colour_primaries, transfer_characteristics, matrix_coefficients, video_full_range_flag and chroma location. These tags do not change pixels, but they change interpretation. Wrong or missing tags are a classic cause of 'it looks fine in Resolve but different after upload/player'.
First identify the media pixel format/bit depth/chroma, then Data Levels, then input colour space/gamma/transfer. If washed/clipped blacks are fixed only by changing Data Levels, do not add a CST. If skin/hues shift because of a matrix mismatch, do not repair them with Lift/Gamma/Gain. Technical diagnosis should match the layer where the error exists.
After export, inspect stream metadata with an external inspector/ffprobe-like tool: pixel format, bit depth, colour primaries, transfer, matrix and range. An encoder/container can constrain the profile or rewrite tags. Final verification should be performed on the actual render artifact delivered to the client/platform.
The most useful habit when a file has a colour problem is to write down `primaries / transfer / matrix / range / bit depth / chroma`. If one of those is unknown, the signal state is incomplete.
Technical definitions are anchored to standards bodies and primary specifications wherever that distinction matters.
BT.709 Y′ coefficients and digital quantisation.
BT.2020 matrix coefficients, 10/12-bit digital representation and chroma formats.
Standardised colour primaries, transfer, matrix and video-range signalling.
Data Levels, media interpretation, Colour Management and Deliver metadata.
