Trust the verification. Not a single checkbox.

Between project pixels and emitted light, the application view transform, operating system, graphics output, video interface and display mode can all matter. Draw your own connection diagram first. Mark the output transform and the location of calibration correction.

The desktop Viewer and a dedicated video output are not automatically the same path merely because they share a computer. Blackmagic describes DeckLink and UltraStudio as professional output options. That does not eliminate the need to verify signal range, display mode and the final result. A verification patch must travel through the path used to judge the work.

Calibration adjusts device behaviour using controls or correction. Profiling characterises the behaviour in a particular state. Verification compares the result with a stated target. A software package may perform all three consecutively, but one button does not make them the same operation.

Ask three questions: what changed, what was characterised and what was checked again? A profile built yesterday is not a measurement taken today. Even a correctly loaded file may describe a different brightness setting. Store the settings, profile identifier and verification report as separate pieces of evidence.

A target should answer the production question: primaries and white point, SDR or HDR response, white luminance, surround and acceptance criteria. A gamma number without these conditions is an incomplete specification. Do not transfer print-proofing settings to a video master simply because both are called professional.

The lab initially compares against D65, white at 100 cd/m², black at 0.05 cd/m² and BT.1886. These are declared teaching conditions, not universal instructions. EBU defines several monitor grades and multiple characteristics; matching one example on this page cannot establish a monitor grade.

Before measuring, define the working position, lighting, display mode and stabilisation time specified by the device procedure. Avoid accidental changes during a series: SDR/HDR switching, energy saving, automatic brightness or unplanned dynamic processing. Preserve the intended production mode rather than indiscriminately disabling every feature.

A reflected window changes viewing conditions, but a contact measurement may not capture its effect as an observer sees it. Record the surround separately. Real footage is useful after measurement to expose practical problems; it does not replace known input values and an instrument.

A colorimeter estimates colour through spectrally weighted channels; a spectral instrument provides additional information about emitted spectra. The important question is not just whether a device is called a calibrator, but whether its response, correction and measurement limits fit the display. ArgyllCMS explicitly documents colorimeter corrections for different emission spectra.

Record the instrument, correction, geometry, integration time and near-black limitations. Extra decimal places do not create accuracy. Two instruments disagreeing does not automatically identify a faulty display: compare conditions, repeatability and spectral corrections first.

A white patch checks one endpoint, not the reproduction of mid-levels. BT.1886 depends on the specified black and white; with nonzero black it is not simply Lw·V^2.4. The lab computes its target independently of fixed synthetic report samples.

Inspect error units as well as apparently overlapping lines. A 1 cd/m² difference near white and near a dark patch has a different relative scale. This page explicitly normalises to target white rather than inventing a gamma-accuracy score. V is the normalised signal after correct digital-range interpretation, not scene exposure.

White chromaticity answers a different question from luminance. Keep x,y or u′,v′ alongside Y and the chosen observer. Correlated colour temperature alone does not specify which side of a locus the white lies on. Check several grey levels too: a correct white does not rule out a midtone cast.

The lab’s Δu′v′ is only chromaticity-coordinate distance. It is neither ΔE2000 nor a universal visibility prediction. Near black, weak signals make chromaticity estimates sensitive to noise; the black patch in this model intentionally has no reported chromaticity.

Uniformity requires a separate spatial series. In this nine-zone example, every zone receives the same nominal white and is compared with the centre. The denominator is explicit: −15% at a corner means 85 cd/m² against a 100 cd/m² centre, not a 15% error in every film pixel.

Sequential zone readings can mix spatial differences with temporal drift. A real protocol benefits from repeated centre readings and a recorded measurement order. Patch size, duration and dynamic backlight behaviour are also conditions. A successful centre measurement does not establish a uniform field.

Internal display correction, output-signal correction and an ICC-based colour transform have different roles. Identify which chain produces the picture in a particular mode. A profile appearing in system settings proves neither that an application uses it nor that the required calibration correction reaches a separate output.

The opposite mistake is applying the same correction twice. Use a known test signal and name each operation rather than randomly toggling options. Record LUT and loader state, device mode and destination profile. Recheck the stored preset after changing display inputs.

Read the report header first: device, mode, date, instrument, correction, patch set and target. Then examine neutrals, saturated samples, range extremes and repeatability separately. An average can hide a few poor samples; a maximum without identifying the sample offers little diagnosis.

Checking ICC-file structure or numerical transform round trips does not measure the current monitor. ICC lists those as distinct assessment tasks; success in one does not validate everything else. A ΔE result needs a formula and conditions. The lab’s luminance-error ramp is not a substitute for coloured-patch verification.

HDR needs verification in the corresponding signal and display mode. Peak luminance from a small patch is not the entire story: area, duration and the distribution of bright regions matter. Production-monitor requirements extend beyond an HDR-support label.

This lab is limited to a static SDR report. It does not measure automatic brightness limiting, local dimming, colour volume, latency or motion artefacts. An omitted test is not a positive result. State explicitly which behaviours were tested and which remain outside the procedure.

Control intervals should follow production requirements and observed stability, not one universal schedule. Material changes in mode, connection, firmware, driver, lighting or correction justify checking the affected path again. Saved reports help distinguish gradual drift from a sudden configuration error.

A useful log records the target, complete output path, display settings, correction and profile identifiers, instrument, conditions and control readings. Before an important delivery, a short known-patch check can reveal a lost state. It is not a substitute for full verification, but it can show that yesterday’s setup is no longer reproducible.