A code is not a brightness percentage.

PQ specifies an EOTF: normalised signal is converted to display light. With known normalisation, the same N has the same reference value regardless of whether a particular screen can reproduce it. Code 0.5 corresponds to approximately 92.25 cd/m².

“Absolute” does not mean identical appearance on every device. Black level, peak and colour volume limit the display, which may apply additional mapping. Surroundings affect perception too. The lab verifies the function, not the screen.

ST 2084 operates on normalised N, not directly on 0…1023 integers. Ten-bit nominal Video RGB or Y′ uses boundaries 64…940; Full uses 0…1023. This allocation is separate from the PQ curve.

Disable rounding to compare the function with 10/12-bit storage. The lab reports reconstructed N, luminance error and the next-code light increment. These bounds describe RGB/Y′; Cb/Cr allocation has different limits and a centre code.

PQ value 1 corresponds to 10,000 cd/m² in the reference function. This does not require stretching every master to the top, or assigning code 1 to all white graphics. Mastering requirements and level allocation belong to the project.

Unused headroom is not necessarily lost contrast. A dark scene may remain dark with highlights far below the encoding ceiling. Distinguish code-space capacity, actual image values and reference-display capability.

PQ allocates codes with visual sensitivity in mind rather than equal cd/m² increments. The light difference between neighbouring integer codes therefore changes across the scale. “1024 values over 10,000 nits” is an inadequate description.

More bits do not guarantee band-free results in every pipeline. Earlier quantisation, compression, processing and output also matter. This lab measures one idealised quantisation step, not codec quality or perceptual visibility for a particular observer.

203 cd/m² encodes to about N=0.5807 in PQ, often rounded to 58% in production tables. 100 cd/m² is about 0.5081. These are different light targets, not evidence that one application darkens the image.

Reference white aligns a workflow; it does not turn every white object into a test patch. For SDR graphics, identify the input encoding and intended white before converting. A corrective Gain after arbitrary interpretation is not a substitute.

A transfer function specifies a component’s light relationship, not RGB geometry. Television PQ is commonly used within BT.2100 with BT.2020 primaries. ST 2084 alone does not specify the primaries or white point.

Coloured pixels require all three channels and their colour representation. This page uses neutral R′=G′=B′ samples. An arbitrary Y′ from Y′CbCr cannot automatically be interpreted as a direct photometric Y measurement.

The inverse PQ EOTF encodes chosen display light. It does not decide where camera-recorded faces, grey objects or emitters belong on a screen. Rendering establishes contrast, colour and output constraints.

“Log → inverse log → inverse PQ” without defined normalisation and rendering is not a universal HDR workflow. First identify what the linear values mean. ACES output processing similarly separates rendering from display encoding.

A signal may contain light above the screen’s peak. Clipping, global scaling and soft compression fit it differently while the PQ reference function remains unchanged.

ITU operational guidance includes adaptation to display black and white. This is additional mapping, not a redefinition of PQ constants. Device capability and content information can influence the result; review the master and its intended playback.

Computing PQ from normalised N does not require movie-specific information. Delivery and playback may nevertheless use colour signalling, mastering parameters and content information. The mathematical function and delivery-format requirements are different layers.

Bad metadata can misdirect playback; correct tags cannot repair double-transformed pixels. Verify signal, description and receiver behaviour. This lab neither writes nor analyses metadata.

Identify whether a scope shows integer levels, normalised nonlinear signal or luminance interpreted through PQ. Its scale and signal context matter more than a familiar trace shape. A percentage of nonlinear code is not a percentage of peak light.

Check known neutral steps and coloured highlights. A luminance scope does not replace RGB-channel and colour-volume checks. Near black, consider storage precision and scale magnification rather than the number of visible lines alone.

PQ output in Resolve must match the current working representation. Camera log and a finished PQ master need different input handling. In managed projects, verify automatic identification and explicit overrides.

Encoding, tone mapping and gamut fitting are connected but distinct tasks. Do not duplicate project-owned processing in a CST. An Output Gamma setting alone does not validate the rest of the chain.

Check that the selected format carries the required depth, signalling and range. Delivery specifications matter more than a generic “HDR settings” recipe. A short representative render exposes systemic mistakes faster than a full programme.

Include gradients, shadows, bright graphics and transitions in average brightness. Record numerical checks separately from viewing observations: matching codes and appropriate display review complement each other.