One signal. Different display conditions.

HLG was designed with television production and relative signal transmission in mind. It combines encoding and display behaviour rather than merely changing a file label. Receiver support and a coherent chain remain necessary.

Start with relative scene light E, nonlinear signal E′ and display light. The OETF connects the first two; final display behaviour also includes an OOTF. Omitting that stage gives a different tonal response.

The lower HLG OETF branch is √(3E); the upper branch is logarithmic. They meet at E=1/12 and E′=0.5. Direction matters: decoding signal back to relative light uses the inverse expressions.

The lab applies the inverse OETF before forming display light. Its curves and values belong to this chain. At zero display black, zero code yields zero, and maximum neutral code reaches the selected reference peak.

The Hybrid Log-Gamma name does not mean “apply 1/2.2 everywhere”. A power approximation may match some points while deviating in shadows or highlights; the error then propagates through display rendering.

The hybrid OETF and system gamma serve different purposes. One encodes relative light; the other participates in scene-to-display rendering. The lab uses the defined piecewise function and its analytic inverse, not a CSS approximation.

HLG relative encoding does not assign one unique cd/m² value to every code without display assumptions. Changing peak in the lab keeps code fixed while changing system gamma and display light. This is intentional, not quantisation error.

“Relative” does not mean arbitrary playback. BT.2100 defines a reference system and adjustment rules. The lab restricts peak to 400…2000 cd/m², the domain of the simple formula used here, rather than extrapolating it.

The reference HLG OOTF derives one multiplier from scene luminance Ys and applies it to all three components. For a neutral, this simplifies to a power relationship; raising coloured R, G and B independently to γ gives a different colour.

BT.2100 also specifies black-level adjustment through β. The lab exposes this effect explicitly. Surrounding conditions remain reference assumptions: this is not a simulation of visual adaptation or a particular television.

With a 1000 cd/m² reference peak, system gamma 1.2 and zero black, neutral code 0.75 yields approximately 203 cd/m². These assumptions explain the production anchor; they do not make that code equally bright on every screen.

The lab reports the 75% neutral reference separately from the current coloured sample. Changing to blue does not redefine blue-pixel luminance as white. Update expectations when peak or black changes.

BT.2100 needs primaries, white point and component representation as well as HLG. The lab’s Ys weights belong to linear BT.2020 RGB; do not reuse them for an arbitrary RGB basis.

Y′CbCr input requires the correct matrix and normalisation. Treating luma alone as a neutral step is not equivalent to reconstructing a coloured RGB pixel. At export, verify transfer, primaries and matrix coefficients together.

The HLG reference system does not require a mastering-display description for each programme, which suits a consistent television chain. Receivers must still recognise HLG and the stream’s agreed parameters. Image appearance alone cannot reliably signal HDR.

Live production particularly needs stable reference levels and predictable input handling. Switching among cameras, graphics and SDR archive should not introduce brightness jumps through interpretation errors. This is a production-wide technical and operational task.

HLG was designed with some viewing compatibility in mind, not identical results on every SDR device. Colourimetry, receiver mode, levels and contrast matter. Arbitrary playback of BT.2020 HLG as Rec.709 SDR can alter both tone and colour.

For an SDR master, perform a defined conversion and review it separately. A tolerable fallback is not an authored SDR deliverable. BT.2446 describes HDR/SDR conversion approaches with different trade-offs.

SDR graphics codes cannot simply be interpreted as HLG. Identify their space, decode appropriately and place them into the target system. Finished PQ also has its own rendering and needs a coordinated conversion.

Decide where compositing occurs relative to output processing. Linear-light blending differs from code-value blending, and alpha handling must agree. Title white should follow intent and production references rather than an accidental channel maximum.

HLG signal percentage and cd/m² are different scales. If a scope reports luminance, establish its reference peak, black and system assumptions. Coloured signals also require the correct RGB-to-luminance relationship.

Compare a known neutral step and coloured samples. Together they help separate OOTF mismatch from matrix errors. Software values alone cannot measure the actual monitor or perceived brightness.

In Resolve, distinguish finished HLG input from scene-referred footage awaiting HLG output. Input assignment describes existing samples; output conversion forms a target representation. They are not interchangeable.

Inspect project management, explicit CSTs and monitoring. Record reference assumptions when comparing HLG with PQ. Unmatched viewing systems can explain differences without implying an incorrect source curve.