Bit Depth (8/10/12-bit) Color Bit Depth (Bits Per Component)

What bit depth is

Bit depth, also written bits per component or bpc, counts the bits spent on each of the red, green, and blue values of a pixel. Each extra bit doubles the number of levels in that channel. Eight bits give 256 levels per channel and 2^24 = 16,777,216 colors. Ten bits give 1,024 levels and 2^30 = 1,073,741,824 colors, which is past the one billion colors that mark deep color. Twelve bits give 4,096 levels and 68,719,476,736 colors. CTA-861 defines the same idea as component depth: "the number of bits used to represent a Color Component Sample."

HDMI carried deep color from version 1.3 onward at 30, 36, and 48 bits per pixel, which is 10, 12, and 16 bits per component. The HDR formats in use today sit on that ladder. HDR10 is a 10-bit format, Dolby Vision can reconstruct 12 bits, and the HDR10+ specification tops out at 16 bits while streaming services deliver every HDR title at 10 bits.

How bit depth works

Real video does not use every code value. In ITU-R BT.2100 narrow-range signaling, 10-bit black is code 64 and nominal peak is 940, with a legal data range of 4 to 1019. For 12 bits the numbers are 256 black, 3,760 peak, and a data range of 16 to 4,079. Full range uses 0 to 1,023 and 0 to 4,095, which is the 1,024 and 4,096 levels per channel counted above, and the standard's 4 to 1,019 limit means that narrow-range signals may stray slightly below black or above peak white without being cut off. Narrow range therefore leaves 876 steps between black and white at 10 bits, and 3,504 at 12 bits. ITU-R BT.2020 sets the same coding format of 10 or 12 bits per component, with red, green, and blue primaries at x,y of 0.708,0.292, 0.170,0.797, and 0.131,0.046.

The visible effect is banding: where a gradient should be smooth it turns into steps. A worked example: a sky gradient that spans only 32 of the 256 levels of an 8-bit channel across a 3,840-pixel-wide frame has 120-pixel-wide bands. The same stretch at 10 bits spans 128 levels and the bands shrink to 30 pixels, below what most viewers resolve. For more on how banding appears, see color banding.

Bit depth vs HDR

HDR stretches the brightness range, so each step has to cover more luminance. The PQ curve in SMPTE ST 2084 maps code values to luminance from 0 to 10,000 nits using a curve built around human sensitivity to banding. Wikipedia reports that PQ "is able to produce no visible banding in 12 bits," and that a power function stretched to 10,000 cd/m2 would have needed 15 bits. That is why HDR can get by with 10 or 12 bits instead. ITU-R BT.2100 itself notes that PQ "achieves a very wide range of brightness levels for a given bit depth."

On disc, Dolby's UHD Blu-ray workflow shows how the numbers stack up. The Dolby Vision Profile 7 base layer is a 10-bit HDR10 bitstream and the enhancement layer is also a 10-bit stream. Together a decoder can reconstruct a 12-bit master that may be graded as high as 4,000 nits, though Dolby cautions that 12-bit performance depends on base layer resolution, content type, and enhancement layer bit rate.

Bit depth vs bandwidth

More bits per component cost more data. Active-pixel math for a 3840 x 2160 signal at 120 Hz with full 4:4:4 color is 3,840 x 2,160 x 120 x bits-per-pixel. At 8 bits (24 bits per pixel) that is 23.9 Gbit/s, at 10 bits (30 bits per pixel) it is 29.9 Gbit/s, and at 12 bits (36 bits per pixel) it is 35.8 Gbit/s. HDMI 2.1 FRL signaling runs at 48 Gbit/s with a data rate of about 42.0 Gbit/s, so even 12-bit 4K at 120 Hz sits under that figure before blanking overhead is counted, while HDMI 2.0 class links, which top out near 18 Gbit/s, cannot carry any of these. Chroma subsampling is the usual release valve; see chroma subsampling, HDMI 2.1 bandwidth tiers, and the HDMI cable tool.

Where each depth lives

VESA DisplayHDR requires a 10-bit video input signal per channel at every tier, and lets the driver chip be 8-bit plus 2-bit dithering at minimum. Ultra HD Blu-ray and Dolby Vision use 10-bit layers as shown above; streamers deliver 10-bit HDR. In these sources 12 bits appears as reconstruction precision and as a signal capability, so treat a "12-bit" claim on a TV box as a signal spec, not proof of a 12-bit panel.

Common confusions

A 10-bit input does not guarantee a band-free picture. TFTCentral argues the matrix bit depth is not the main cause of banding, which more often comes from inaccurate monitor processing of color temperature, gamma, and contrast, and that monitors with 10-bit or 12-bit internal processing and higher-bit lookup tables help. Panel bit depth also differs from signal bit depth: many panels are 8-bit with temporal dithering and are sold as 10-bit, covered under dithering and FRC. Finally, bit depth is separate from color gamut. Bit depth sets the step size and gamut sets the extent of the colors, as when HDR pairs 10-bit PQ coding with BT.2020 primaries.