KNOWLEDGE MAP · REFERENCE

Keep the numbers. Keep their context.

Eight reference tables and a nominal-level calculator. Each value stays connected to its component, bit depth, transform direction and scope.

Keep the definition, the test and its limitations together.

CODE DESK / 03Nominal digital levels
Bit depth
INTEGER CODE502round(876 × E′ + 64)
64940
Decoded component0.500000
Rounding error0
Error bound±0.0005708
Formula and interpretation +D = round(876 × E′ + 64) E′ = (D − 64) / 876

Quantisation uses rounding to the nearest integer. This calculator covers nominal levels only, not the reserved-code rules or a complete codec. Full chroma uses a different centre/endpoint contract and is deliberately not offered.

Computed codes, not an HDR image or display measurement. 12-bit Video levels follow the corresponding BT.2100 scaling; 8-bit illustrates the BT.709 case.

COLOR SCIENCE / 03 / REFERENCE

A reference, not an automatic diagnosis.

On this page ↓

01 / REFERENCE TABLES

A reference you can verify.

Read the row together with its units, direction and note. Tables scroll independently on small screens.

RGB spaces and working systems

Basis, white and encoding are separate. A standard number is not a delivery specification.

RGB spaces and working systems
SystemPrimariesWhite pointEncodingContext
Rec.709Rec.709D65BT.709 OETF; display response is separate, including BT.1886HDTV / SDR, controlled path
sRGBSame as Rec.709D65sRGB piecewise functionGraphics with a known profile
Display P3P3D65Same function as sRGBWide-gamut display graphics
DCI-P3P3x=0.3140, y=0.3510Power 2.6 for this variantDisplay description; not the DCP file format
Rec.2020Rec.2020D65BT.2020 for SDR; BT.2100 specifies PQ/HLGUHD; not automatically HDR
DaVinci Wide GamutDWGD65Commonly DaVinci Intermediate, a separate functionResolve working representation
ACEScgAP1x=0.32168, y=0.33767LinearScene-referred CG / VFX
ACEScctAP1x=0.32168, y=0.33767Log section + linear toeGrading; requires separate output

Conditions. ACES white is D60-like, not exact CIE D60. P3-D65 with power 2.6 is not Display P3.

Transfer functions: direction and conditions

A simple power law and a specified reference EOTF are shown separately.

Transfer functions: direction and conditions
FunctionInput → outputScaleCheck
Gamma 2.4Normalised code → relative light in a simple modelL ∝ V²·⁴Not full BT.1886 with nonzero black
BT.1886Normalised signal → display luminanceDepends on Lb and LwPower 2.4 with black/white parameters
sRGBLinear components ↔ codesRelativePiecewise, not exactly power 2.2
Camera LogScene quantity ↔ codeCurve-dependentGeneration, gamut and precision from the specification
DaVinci IntermediateWorking linear values ↔ codeRelativeNot an ordinary display response
PQ / ST 2084Code → reference display luminanceAbsolute; up to 10,000 cd/m²Master and physical display peaks are separate
HLGScene → relative signal; display includes OOTFResult depends on display systemOETF, system gamma, black and white

Conditions. These functions do not replace primary conversion, codec processing or Video/Full interpretation.

SDR / HDR reference anchors

Every value has a context: white reference is not content peak.

SDR / HDR reference anchors
PropertySDRPQHLG
Signal meaningKnown SDR pathAbsolute reference EOTFRelative OETF + OOTF
WhiteAgreed target and conditions; not a universal screen luminance203 cd/m² is a nominal HDR anchor under specified conditions75% ≈203 cd/m² on a 1000-cd/m² display, γ=1.2, reference black model
PeakDepends on display target10,000 is the function maximum, not a required master peakDisplay peak enters the rendering model
PrimariesOften Rec.709; another agreed gamut is possibleBT.2100 uses Rec.2020BT.2100 uses Rec.2020

Conditions. Physical luminance needs measurement. A browser picture or pixel code is not a nit measurement.

Y′CbCr: NCL coefficients

For the ordinary NCL model: Y′ = KrR′ + KgG′ + KbB′. Primes mark nonlinear components.

Y′CbCr: NCL coefficients
StandardKrKgKbCheck
BT.6010.29900.58700.1140Do not select from frame size alone
BT.7090.21260.71520.0722Check metadata against the actual recording
BT.2020 NCL0.26270.67800.0593Do not confuse with BT.2020 CL or ICtCp

Conditions. For neutral R′=G′=B′ the weights sum to the same Y′. A grey-ramp-only test may miss a matrix mismatch.

Bit depth and subsampling

Check value precision and spatial sampling density independently.

Bit depth and subsampling
ParameterMeaningCheck and limitation
8-bit256 integer bit patterns per componentNominal range may not use every code
10-bit1024 bit patternsNot universally sufficient for every log curve
12-bit4096 bit patternsLogC4 requirement is not a rule for every camera; file size also depends on codec
4:4:4Full relative chroma densityDoes not guarantee no compression
4:2:2Half horizontal chroma densitySample position and filters matter
4:2:0For ordinary progressive signals, half density on each axisDoes not mean Cr or Cb is absent

Conditions. Float requires its format definition, not only a bit count. Converting 8 bits to 10 does not recover source detail.

Nominal digital levels

These are reference codes, not project presets or a complete list of interface-permitted values.

Nominal digital levels
RepresentationLower anchorUpper anchorCb/Cr neutralNote
8-bit Y′ narrow16235—Black / white
8-bit Cb/Cr narrow16240128−0.5 / +0.5; these are not black/white
10-bit Y′ narrow64940—Black / white
10-bit Cb/Cr narrow64960512Zero colour difference at centre
12-bit Y′ narrow2563760—BT.2100 reference anchors
12-bit Cb/Cr narrow25638402048BT.2100 reference anchors
8-bit Y′ full0255—Full integer-range model
10-bit Y′ full01023—Full: agree recipient support
12-bit Y′ full04095—Full: agree recipient support

Conditions. Sub-black/super-white may exist outside nominal narrow anchors within the permitted data range. Full-chroma and reserved interface codes require separate rules; the calculator does not certify a master.

Checks along the signal path

Find the first mismatch; the same symptom can have several causes.

Checks along the signal path
StageRecordCompare
1 · SourceCamera, curve, gamut, versionActual recording versus assigned Input
2 · DecodeRAW settings, matrix, levelsSame frame under known settings
3 · WorkingTool-input representationOperation order and LUT expectations
4 · RenderingWho prepares the image for displayMissing or duplicate output stage
5 · FileCodes, pixel format, tags, rangeRe-import with verified interpretation
6 · ViewApp, OS, display, modeSame file and frame in a controlled path

Conditions. One screenshot or tag does not establish whole-pipeline correctness. Preserve the original during analysis.

CST, RCM and ACES: stage ownership

An architecture choice, not an image-quality ranking or a mandatory node recipe.

CST, RCM and ACES: stage ownership
ApproachWhere it is definedWhat to verify
CSTExplicit transforms at chosen locationsInputs, outputs, order and additional mapping settings
RCMProject and clip-level managementInput assignment, working system, automatic stages and Output
ACESSpecified system version and transform setInput Transform, working encoding, Output Transform and interchange

Conditions. Mixed workflows require deliberate stage ownership. Matching output names do not prove identical rendering.

TOPIC ILLUSTRATION

THE ESSENTIAL IDEA

Cheat sheets and tables

A table shortens the search, not the list of required conditions. Establish the signal representation and intended output first; then choose the row that actually applies.

  • Check Y′ and Cb/Cr ranges separately.
  • Do not transfer conditions between standards because their names look similar.
  • Use a number as a check, not a finished grade.
A topic illustration, not a measurement diagram or an interface screenshot.
01

A reference is not a preset

A tabulated value requires the input, output and measured quantity. Gamma 2.4 does not describe all of BT.1886 at arbitrary black levels, and 203 cd/m² is not the required luminance of every white HDR object.

Use a row as a conditional assertion: under these conditions, expect this result. Establish missing conditions first. Delivery requirements take precedence over a value copied from a tutorial; standard and implementation versions matter too.

What this means in practice

For an SDR delivery, do not import 203 cd/m² from an HDR table merely because the row is labelled Reference White.

02

Assemble the whole signal description

A practical minimum includes primaries, white point, transfer function, component representation, range and bit depth. Subsampled Y′CbCr also needs chroma location and reconstruction assumptions. A codec name cannot replace these fields.

The space table is separate from workflow role. Primaries alone do not say whether an image is before or after rendering. Add the intended purpose: source, working interchange, finished master or viewing copy.

What this means in practice

Display P3 and DCI P3 are not interchangeable just because both contain P3: check white point, transfer function and the output contract.

03

A working basis is not a delivery format

DWG and ACES AP1 define useful computational bases, but their coordinates are not automatically ready for every monitor. Known encoding, target-display processing and correct output description are still required.

A wide intermediate representation can contain negative components or values above one. This does not justify clamping before every node. Target-gamut clipping and compression are separate operations with colour consequences.

What this means in practice

For DWG / Intermediate, record two fields: DWG for primaries/white; Intermediate for the nonlinear function. Do not label a delivery DWG instead of its actual output space.

04

HDR anchors belong to a model

PQ provides an absolute luminance relationship; its 10,000 cd/m² upper boundary is a function limit, not a required content peak or a claim about your screen. HLG depends on a display model: 75% corresponds to about 203 cd/m² under reference conditions, not on every display.

Keep nominal production white, mastering-display peak and content statistics separate. MaxCLL and MaxFALL describe different characteristics; a metadata field alone does not prove a correct content analysis.

What this means in practice

The 75% HLG row keeps its conditions: 1000 cd/m² and system gamma 1.2 for the reference calculation. When the display model changes, recompute instead of copying the value.

05

Identify the transform owner first

The workflow table does not rank RCM, ACES and explicit CSTs. It identifies ownership of input, working representation, rendering and output. Each technical operation needs a clear owner; extra nodes do not inherently add accuracy.

Test a short segment before applying a grade across the project. Record system version, input profile, expected LUT input and output. For multiple masters, ensure it is the output branch that changes while the working source remains available.

What this means in practice

If the project already performs input and output processing, inspect those settings first. Then decide whether a separate CST is needed and what local task it performs.

06

What a table cannot validate

Tables do not measure a display, identify an unknown file’s true profile or validate a decoder. The code calculator models one nominal component scale and rounding, not an actual codec.

Permitted and reserved codes follow rules beyond nominal levels. Full Cb/Cr cannot be inferred mechanically from Full Y′: chroma has a centre and endpoint details. Here Full is offered only for Y′/R′G′B′, while Cb/Cr uses nominal Video coding.

What this means in practice

For 10-bit Cb/Cr, zero gives 512 and nominal −0.5/+0.5 gives 64/960. Video Y′ instead uses 64/940. Verify the difference in the calculator.

DO. OBSERVE. EXPLAIN.

Verify a range using known codes

Use this page’s calculator: it checks a formula, not a real video export.

  1. 01

    Check Video Y′

    Choose 10-bit and check normalised 0 and 1, then try 0.5.

    CHECK

    Endpoint codes are 64 and 940; any midpoint error is explained by rounding.

  2. 02

    Change component

    Switch to Video Cb/Cr and check −0.5, 0 and +0.5.

    CHECK

    The results are 64, 512 and 960; 940 was not used as the chroma limit.

  3. 03

    Check Full separately

    Select Full Y′/R′G′B′ and repeat 0 and 1. Do not conflate a component change with result interpretation.

    CHECK

    The 10-bit Full range is 0…1023, not 0…1024.

  4. 04

    Check your delivery contract

    Find your signal’s remaining fields in the tables and mark unknown ones.

    CHECK

    The code scale was not used as proof of HDR or a nit measurement.

Check your understanding

0 / 4

Tick only what you can explain yourself.

TEST THE EXPLANATION

A situation, not a trick question.

Answer first. Open the reasoning afterwards.

01Does 75% HLG always equal 203 cd/m²?

No. This approximate correspondence requires the stated reference conditions. The display model and system gamma remain part of the calculation.

02Is BT.1886 simply x^2.4?

The corresponding power form follows at zero black. The full reference EOTF depends on black and white luminance; the table separates those cases.

03Does 10-bit Cb/Cr end at 940?

In nominal Video coding, 940 is the upper Y′ level. Cb/Cr uses 64…960 with neutral centre 512. This is not a list of all permitted or reserved codes.

Avoid these shortcuts

  • Do not replace missing information with familiar settings.
  • Keep the conditions with the number or definition.
COLOR / DAVINCI RESOLVE

From the model to your workflow.

Record the Resolve version, colour-management mode, input interpretation and viewing conditions. This reference helps define a check, but does not replace reviewing the full shot or validating a delivery file.

Color Management ↗Scopes ↗Color Space Transform ↗

CHECK THE FOUNDATION

Sources and model limits.

Editorial review: 1 October 2026. Links identify the definitions and formulas; exercises are independent teaching examples.

  1. [4]
    ICC · sRGB ↗

    Primaries, D65 and the piecewise transfer function.

  2. [5]
    ICC · Display P3 ↗

    P3, D65 and the sRGB function.

  3. [6]
    ICC · DCI P3 ↗

    White-point variants and power 2.6; not a DCP packaging specification.

  4. [7]
    ITU-R · BT.709-6 ↗

    Sections 1, 3, 4: primaries, OETF, coefficients and 8/10-bit digital levels.

  5. [8]
    ITU-R · BT.1886 ↗

    Annex 1: EOTF depending on black and white luminance.

  6. [9]
    ITU-R · BT.2100-3 ↗

    PQ, HLG; Table 9: 10/12-bit integer representation.

  7. [10]
    ITU-R · BT.2408-9 (03/2026) ↗

    HDR reference white and conditions for 75% HLG corresponding to about 203 cd/m².

  8. [11]
    ITU-T · H.273 (07/2024) ↗

    Independent video-identification fields; metadata is not a transform.

  9. [12]
    ACES · White point ↗

    x=0.32168, y=0.33767; D60-like is not exact CIE D60.

  10. [13]
    ACES · ACEScct ↗

    AP1 and nonlinear scene encoding with a linear toe.

  11. [14]
    ACES · Output Transforms ↗

    Target-display rendering and encoding; not merely a tag change.

  12. [15]
    Blackmagic · DWG / Intermediate v1.1 ↗

    DWG and Intermediate specify distinct properties of a working representation.

  13. [16]
    Blackmagic · Color ↗

    CST, RCM and Color tools. Menu details depend on the Resolve version.

  14. [18]
    ARRI · LogC4 specification ↗

    Precision is format-specific: LogC4 specifies a 12-bit minimum; 10 bits is not universal.

  15. [19]
    FFmpeg · ffprobe ↗

    Read stream fields without modifying the file; unknown does not prove a particular colour space.

  16. [20]
    FFmpeg · Content light metadata ↗

    MaxCLL and MaxFALL are different content-metadata fields, not measurements of the current display.

  17. [21]
    OpenColorIO · Displays & Views ↗

    Working data and viewing transforms are distinct.