Colorimeters vs. Spectroradiometers: Why Calibration Needs Both

A tristimulus colorimeter sees the way the eye does: three filtered photodiodes whose responses approximate the CIE color-matching functions, giving XYZ directly in one fast, low-noise read. A spectroradiometer sees the way physics does: it disperses the light through a grating and samples power at hundreds of wavelengths, a full SPD, then computes XYZ (or anything else) from it. One is fast and eye-shaped; the other is slow, expensive, and complete. Serious display calibration uses both, because each covers exactly the other's weakness.

Its strengths are exactly what calibration's inner loop needs: it reads dark patches with low noise, measures fast enough for hundreds of points and progress bars, and it's affordable enough to leave mounted on the panel. That's why it does the actual grayscale and gamut runs. Its weakness is built into the glass: real filters only approximate the color-matching functions. Where the true curve and the filter curve disagree, the instrument reports a slightly wrong XYZ, an error that stays hidden on broad spectra and explodes on narrow ones.

Because it measures actual power per wavelength, typically across and beyond the visible, from the ultraviolet through to the near-infrared, it doesn't care what shape the spectrum is: a laser spike and a broad phosphor are both just data. Compute XYZ against any observer, CCT, CRI, or spectral metric from the same capture. The cost is real: it's slower, noisier in deep shadow, more expensive, and more sensitive to stray light and calibration drift. You don't want it running a 200-point grayscale sweep; you want it establishing what the truth is, so a colorimeter can chase it quickly.

A filter mismatch matters in proportion to how much light lands where the filter is wrong. A broadband white LED smears energy across the spectrum, so local filter errors average out. A wide-gamut display (laser, RGB LED, quantum-dot) concentrates its primaries into narrow spikes that land on the steep slopes of the filter curves, where a few-nanometer mismatch becomes a large XYZ error. Switch the display to QD-OLED or RGB LASER above with the colorimeter selected: the raw error jumps, worst in blue and deep red where MacAdam tolerances are tightest. This is why a meter that was perfect on a Rec. 709 CRT can read visibly wrong on a modern WCG panel.

The fix pairs the two instruments. On the exact display you'll calibrate, the spectroradiometer measures the true chromaticity of the red, green, blue, and white; the colorimeter measures the same four. From the difference, a 3×3 correction matrix (the Four-Color Matrix Method, developed by Ohno and Hardis at NIST) is solved and loaded into the colorimeter's software. Now the fast colorimeter inherits the spectro's accuracy, for that display type. Toggle FCMM above and watch the raw error collapse toward the reference. The catch, and it's the whole point: a profile is only valid for the spectral family it was made on. Profile on a WLED and measure a QD panel and it can make things worse; this is why the spectro has to come back out whenever the display technology changes.

The color-matching functions come in two standard flavors. The CIE 1931 2° observer was measured on the fovea, a ~2° field: small, distant sources. The CIE 1964 10° observer used a larger field and is recommended for fields beyond ~4°; it's the norm in textiles, paint, and large uniform patches. For display calibration the convention, and what the standards bake in, is the 2° observer: Rec. 709, Rec. 2020, and DCI define their white and primaries against 1931 2°, so calibrating to those targets means measuring in 2° or your numbers won't match the spec. The 10° option exists for wide-field viewing research and some large-format work; toggle it above and the reference white shifts slightly, proof that "which observer" is part of the measurement, not a detail. (A colorimeter measures raw XYZ; the observer is applied in software, so the same read can be reported either way.)

In practice with Calman or equivalent: bring the spectroradiometer to the actual panel, profile the colorimeter to it with the four-color method, then let the fast colorimeter run the calibration and verification passes (grayscale, gamma/EOTF, gamut, saturation sweeps) reading confidently into the shadows. Re-profile whenever the display technology changes, and periodically re-verify against the spectro. The deliverable quotes the instruments, the observer (2°), the target standard, and the residual ΔE. That is what "probe-verified" actually means: two instruments, each doing what only it can, checking each other. Reserva una calibración →