The PQ curve, short for Perceptual Quantizer, is the transfer function that tells an HDR display how bright each digital code value should be. Published by SMPTE as ST 2084, it replaces the gamma curve used in SDR, and its 0 to 1 signal range represents luminance from 0 to 10,000 cd/m² (nits). It was developed by Dolby, standardized by SMPTE in 2014, and adopted by the ITU in 2016 in Rec. 2100. If you watch HDR10, HDR10+, or Dolby Vision content, you are looking at a PQ signal.

Gamma vs. PQ: the core difference

An SDR display follows BT.1886, a power-law "gamma" of roughly 2.4. It is relative: signal 100% means "as bright as this display goes," whether that is 100 nits or 400 nits. The curve was inherited from the physics of CRT phosphors, not from vision science.

PQ is absolute. Every code value maps to a specific luminance, and the same 10-bit value should produce the same nits on a mastering monitor in Burbank and a TV in Albuquerque. Where gamma represents the response of a physical CRT and only coincidentally resembles human vision, PQ imitates the true response of the human visual system using a relatively simple functional model.

The shape comes from psychophysics. Dolby modeled PQ on Barten's 1999 contrast sensitivity function, which predicts the smallest brightness step a viewer can detect at any given luminance. The curve is built so that each code step sits just below that detection threshold across the entire range. In 12-bit PQ, one code value at 1 cd/m² is a step of about 0.0048 cd/m², while at 1,000 cd/m² one code value is a step of about 2.24 cd/m². That is why PQ spends most of its codes on shadows and midtones, where the eye is sensitive, and fewer on extreme highlights, where it is not.

The math, briefly

The EOTF is defined in ITU-R BT.2100 as:

Y = 10,000 × [ max(E^1/m2 − c1, 0) / (c2 − c3 × E^1/m2) ]^1/m1

with m1 = 0.1593, m2 = 78.84, c1 = 0.8359, c2 = 18.85, c3 = 18.69, where E is the normalized signal (0 to 1) and Y is luminance in nits. In practice the useful anchor points on a 10-bit scale are:

• 100 nits (SDR peak white) at roughly 51% signal
• 203 nits (HDR reference white per BT.2408) at roughly 58%
• 1,000 nits at roughly 75%
• 10,000 nits at 100%

Half the code range covers everything up to SDR white, which is exactly the point.

Why 10,000 nits when no consumer display reaches it

PQ is a container, not a target. Mastering monitors sit at 1,000 to 4,000 nits and most content is graded to those levels. Consumer panels, from a 400-nit LCD to a 1,400-nit tandem OLED, cannot reproduce the full range, so they apply tone mapping: tracking PQ exactly up to a knee point, then rolling highlights off toward the panel's real peak. ITU-R BT.2390 documents display mapping for panels with a limited brightness range, and manufacturers layer their own dynamic tone mapping on top. This is the behavior that gaming initiatives such as HGiG ask the TV to switch off so the source can map once instead.

PQ vs. HLG

BT.2100 defines two HDR systems. PQ achieves a very wide brightness range for a given bit depth using a non-linear function finely tuned to human vision, while HLG (Hybrid Log-Gamma) offers a degree of compatibility with legacy displays by more closely matching established television curves. PQ is display-referred (absolute nits, needs metadata, ideal for graded film and streaming); HLG is scene-referred (relative, no metadata, ideal for live broadcast). PQ is not backward compatible with the BT.1886 gamma curve, whereas HLG is.

Engineering perspective

Three things I check on any HDR panel:

EOTF tracking. Measured luminance is plotted against the PQ target at each gray step. A panel that tracks within a few percent up to its knee, then rolls off smoothly, reproduces the grader's intent. Panels that run bright in the 20 to 60% range look "punchy" in a showroom but crush the tonal separation the curve was designed to preserve.

Bit depth. PQ produces no visible banding at 12 bits. HDR10 ships at 10 bits, which is generally adequate but can show contouring in smooth gradients, especially near black where PQ packs codes tightly and panel driver nonlinearity is worst. Dolby Vision's 12-bit path exists for this reason.

Absolute means absolute. Because PQ defines nits rather than percentages, room brightness and the panel's automatic brightness limiter both distort the intended image. A 1,000-nit window measurement on an OLED can be far above what the same panel delivers full-screen, so full-field tracking matters as much as the 10% window.

Bottom line

The PQ curve is the vision-science-based EOTF that makes HDR possible: it assigns absolute luminance to every code value, allocates bits where the eye can see them, and spans a 10,000-nit container that displays tone map down to their real capabilities. It is defined in SMPTE ST 2084 and ITU-R BT.2100, with implementation background in Report BT.2390.