KNOWLEDGE MAP · SPACES & STANDARDS
Wide working coordinates. Explicit encoding.
Separate DaVinci Wide Gamut from Intermediate. Calculate codes, decoded light values and near-zero behaviour, then organize input, working operations and output without duplicate CSTs.
Read the numbers. Notice the assumptions. Then apply the idea to your footage.
Light → Intermediate → light.
0.1800000.3360430.180000|ΔL| = 0Blackmagic v1.1 transfer equations; L is relative, not nits. Negative and above-one results are not silently clipped. The optional ACEScct curve compares encoding of one scalar, not full DWG/AP1 colour conversion.
TOPIC ILLUSTRATIONTHE ESSENTIAL IDEA
DaVinci Wide Gamut / Intermediate
DWG defines RGB coordinates; Intermediate defines encoding. Together they support wide working workflows, without replacing source identification or display preparation.
- Understand the DWG matrix and DI curve as separate operations.
- Test negative and bright values without hidden clipping.
- Do not confuse working representation with camera input.
DWG/Intermediate is two specifications working together
DaVinci Wide Gamut and DaVinci Intermediate describe separate properties. DWG defines primaries and D65 white; Intermediate defines how linear values become nonlinear codes. They commonly work together, but are not one concept.
The pair is intended as a wide working representation, not a finished display standard. Do not assign it to an arbitrary file instead of identifying the actual source. Camera material needs a correct conversion from its known input.
What this means in practice
Distinguish “DWG / Linear” from “DWG / Intermediate”: matching gamut does not imply matching numbers.
Official DWG primaries and white point
The published DWG coordinates are R = (0.8000, 0.3130), G = (0.1682, 0.9877), B = (0.0790, −0.1155), with D65 = (0.3127, 0.3290). Calculations here use the corrected v1.1 specification.
The blue primary’s negative y is a mathematical basis property, not negative luminance from a blue lamp. Working gamuts may use unrealizable primaries. An ordinary monitor is not expected to physically reproduce this entire basis.
What this means in practice
Do not “fix” the negative coordinate in the table: that creates a different matrix.
Why the working gamut is wider than Rec.2020 and AP1
A wide working representation avoids immediately reducing every source to one narrow delivery target. Grading can precede final display preparation. Gamut size itself does not add information to a source.
A Rec.709 file gains no new saturated colours merely by conversion into DWG. Conversion changes coordinates; later clipping, inappropriate operations or limited storage can still lose information. Working width is a capability, not an automatic quality guarantee.
What this means in practice
A correct no-grade Rec.709 → DWG → Rec.709 round trip should preserve the signal, not boost saturation.
DaVinci Intermediate encodes relative scene-linear light
Intermediate has a linear toe around zero and a logarithmic branch above it. In specification v1.1, the linear-input threshold is 0.00262409. Above it: V = (log₂(L + 0.0075) + 7) × 0.07329248. Below it: V = L × 10.44426855.
Negative L uses the linear branch; this formula itself does not clip it. Later tools and storage formats may not preserve it. The lab separately displays the code, inverse result and numerical residual.
What this means in practice
In “Near zero”, choose a negative input and check its sign after decoding.
18% grey and headroom in Intermediate
In this teaching model, 18% relative grey means L = 0.18, encoded to approximately 0.336043 in Intermediate. L = 1 maps to about 0.513837; L = 100 maps to about 1. These are relative-value correspondences, not automatic monitor-nit assignments.
Exposure is a 2^EV multiplier relative to the selected reference. A nonlinear code range of 0…1 is not a universal floating-point ceiling. Check the actual tool and output format separately.
What this means in practice
The same working value can produce different display luminances under different output transforms.
Why DWG/Intermediate is called a scene-referred working context
Nonlinear Intermediate does not mean the image is already display-rendered. Coordinate encoding and image state are different layers. A system may represent scene-referred data nonlinearly for grading convenience.
Consider source decoding, working operations, output rendering and output encoding in order. Values within 0…1 do not prove “finished SDR”. The lab encodes one positive L with DI and ACEScct; differing codes are not differing luminances of two monitors.
What this means in practice
The hero’s DI/ACEScct comparison isolates transfer functions; it does not replace a complete DWG/AP1 conversion.
RCM: automatic conversion of inputs into DWG/Intermediate
DaVinci YRGB Color Managed links identified inputs, working processing and a chosen output. DWG/Intermediate can be the working pair in that process. Verify automatic input identification, particularly for files with incomplete or incorrect metadata.
Settings vary with Resolve version and management mode. Do not copy a menu screenshot without understanding the chain. Record where transforms occur and what a node receives. An extra CST is not needed just because another tutorial adds one.
What this means in practice
Check each camera input separately; a shared working gamut does not prove correct identification.
A manual CST workflow makes the same logic explicit
An explicit chain converts a known input into the selected working representation, performs the grade, then prepares the required output. CST has its own controls; coordinate conversion, white adaptation, tone rendering and gamut limiting do not always reduce to one matrix.
Test no-grade mathematical reversibility separately from creative output rendering. Tone mapping may intentionally change values, so that difference is not automatically a matrix error. Define what should be preserved before diagnosing a result.
What this means in practice
Do not stack manual input/output CSTs over equivalent transforms already performed by RCM.
Timeline Colour Space affects tool behaviour
A tool’s behaviour depends on its input representation. Addition, blur, keying and LUTs may assume different meanings for values. A DWG timeline does not automatically make every third-party effect aware of that context.
For an effect designed around a specific representation, arrange the necessary transforms and test the result. Include shadows, values above one, saturated edges and alpha where relevant, not just one average frame.
What this means in practice
If a LUT behaves unexpectedly, inspect its expected input first; lowering Blend does not repair misinterpretation.
The output DRT handles tone and gamut mapping
Output must match the task: primaries, transfer, image state, target luminance and format. Moving from a wide working representation to a limited display may require tone and gamut rendering, not just RGB conversion.
Evaluate SDR and HDR versions as separate results. A common base helps, but one setting does not automatically satisfy both. Lighting, saturated details, graphics and visual emphasis need individual review.
What this means in practice
An additional HDR delivery needs more than a different tag or “Rec.2020” in the filename.
DaVinci Wide Gamut ≠ Blackmagic Wide Gamut
DaVinci Wide Gamut is a working basis. Blackmagic camera encodings and their generations describe particular sources. Similar “Wide Gamut” menu wording does not make them interchangeable.
Identify the source from camera documentation, recording mode and decoding, then select its conversion into DWG/Intermediate. An input mistake propagates through the chain; a carefully configured output does not undo it.
What this means in practice
Select the actual source in Input, not the working space you would like to use.
DWG/Intermediate is not an ordinary delivery master
Document the working pair, inputs, application version, technical-operation order and output. For intermediate handoff, state whether display rendering is baked in and which values are preserved. “Looks a bit flat” is not an interchange specification.
Inspect the render as a separate artifact: metadata interpretation, levels, appearance and motion. A self-consistent view inside the source project can hide an incorrect automatic interpretation during re-import.
What this means in practice
Provide a reference frame and encoding description alongside the intermediate file, not as substitutes for it.
A MODEL YOU CAN CHECK
Separate input, working pair and output.
Keep the meaning, the calculation and the limitation together.
V = L ≤ 0.00262409 ? 10.44426855L : (log₂(L + 0.0075) + 7) × 0.07329248Constants from Blackmagic specification v1.1. L is a relative linear value here. Published thresholds and coefficients have finite precision; round-trip checks allow for that rounding.
| Concept | What it describes | What not to conclude |
|---|---|---|
| DWG | Primaries and D65; coordinate matrix. | Not a log curve. |
| Intermediate | Piecewise encoding of relative light. | Not a standalone gamut or nit assignment. |
| RCM / explicit CSTs | Organization of input, work and output. | Not a reason to duplicate existing transforms. |
| Delivery master | A rendered representation for the intended use. | Not working DI codes under a new filename. |
DO. OBSERVE. EXPLAIN.
Move from 18% to negative input.
Begin at 0 EV relative to L = 0.18. Plots are educational; the display is not being calibrated.
- 01
Check the reference
Record L and Intermediate code at 0 EV, then compare decoded L with the 0.18 reference.
CHECKExpect a code near 0.336043, not 0.18; decoding returns the selected reference.
- 02
Raise exposure
Increase EV to +10 and inspect L, code and decoding. Return to 0 EV without changing the formula.
CHECKL can exceed one without hidden calculation clipping.
- 03
Explore zero
Enable near-zero mode and select a negative value.
CHECKThe linear toe preserves sign; EV is undefined at zero.
- 04
Record the chain
In a test project, locate actual input, DWG/DI and output, then check for duplicate CSTs.
CHECKThe working pair is not assigned to an arbitrary camera as its original encoding.
Check your understanding
TEST THE EXPLANATION
A situation, not a trick question.
Answer first. Open the reasoning afterwards.
01Does DI = 1 mean 100 nits on screen?+
No. The curve encodes relative linear values; display luminance depends on output. Do not turn encoding-table numbers into monitor calibration.
02Is negative y for DWG blue a typo?+
No: the published basis uses an unrealizable primary. These are mathematical coordinates, not a physical emitter.
03DI and ACEScct give different codes for 0.18. Which is correct?+
Both may be correct for their own curves. The lab compares encoding of one scalar; a full working-space conversion also changes primaries and white.
Avoid these shortcuts
- Equating DaVinci WG with a Blackmagic camera gamut.
- Treating every nonlinear code as display-ready.
- Clipping intermediate values without a deliberate purpose.
From the model to your workflow.
In production, identify the source, transform placement and required output separately. The lab verifies a bounded mathematical model; it does not measure your display, replace project settings or certify an exported file.
CHECK THE FOUNDATION
Sources and model limits.
Editorial review: 1 October 2026. Links identify the definitions and formulas; exercises are independent teaching examples.
- [9]ITU-R · BT.2100-3 (2025) ↗
HDR television parameters. PQ/HLG are not implied by the BT.2020 gamut label alone.
- [15]Academy · ACEScct ↗
AP1, linear toe and logarithmic branch, encoding and decoding equations.
- [20]Blackmagic Design · DWG / Intermediate v1.1 ↗
August 2021 document: DWG coordinates, matrices, DI constants and examples. Light values are treated as relative, not display luminance.
- [21]Blackmagic Design · Color ↗
Color and colour-management context. Setting names and locations depend on Resolve version and configuration; live projects were not tested here.
- [22]ITU-T · H.273 (2024) ↗
Identification of video primaries, transfer characteristics and matrix coefficients.