Sony European Imaging Ambassador Alister Chapman uses his platform to promote categorically false claims about HDR. We dismantle them one-by-one using the standards themselves.
The Data Efficiency Argument
QUOTE:
“PQ is very wasteful of data as it’s designed around theoretical 10,000 Nit displays whereas BT1886 is designed for displays that actually exist.” — Alister Chapman
CLAIM: “PQ is wasteful for targeting a 10,000-nit range.”
REALITY: Only ~7% of PQ’s code space is allocated to the 5,000-10,000 nit range, providing future headroom. The curve is perceptually uniform, prioritizing more codes for lower luminances where the eye is more sensitive. Source: HPA Tech Retreat 2014 – Day 5 report, Q&A on PQ efficiency.
REALITY CHECK
The graph below plots quantization error against the thresholds of human vision (Barten, Schreiber).
The result: 12-bit PQ, even for a 10,000-nit display, stays below the visibility threshold. 12-bit BT.1886 for a 1000-nit display does not. PQ is perceptually efficient; BT.1886 is not suited to HDR.
The Numbers Game
QUOTE:
“PQ is poorer for gradients than BT-1886 (Rec-709). If you are grading to 1000 Nits then you are only using around 500 code values for the entire mid range and maximum 700 out of a possible 970 code values, where as BT1886 will usually be using 730 for the mids with 970 used in total.” — Alister Chapman
CLAIM: “PQ has “poorer” gradients than BT.1886.”
REALITY CHECK
Chapman’s math isn’t mathing:
- Consumer 8-bit SDR code value range: 0 to 255
- 10-bit PQ 1K nits code value range: 0 to 769
The Yedlin camp sneers at PQ’s 10-bit requirement, but substitutes 10-bit SDR when it suits them:
“So, the SDR is totally smooth at 8-bit, and the HDR isn’t. You should be able to see how much more efficient SDR is at using the bit depth than HDR. When they say it’s 10-bit, they mean ‘it NEEDS 10 bits’ [snickering]. It’s a data hog, is what it means.” Steve Yedlin, “Debunking HDR”
Nevertheless, Chapman’s comparison is invalid no matter how you slice it. He incorrectly treats BT.1886 as having a fixed allocation of code values, when by its very nature as a relative standard, this allocation varies according to the display’s brightness setting and capability.
Conversely, the PQ EOTF is an absolute standard. A specific code value always corresponds to a specific brightness level in nits.
The Color & Saturation Argument
CLAIM: “saturation is not different between HDR and SDR.” — Alister Chapman
REALITY: Chapman’s claim that “saturation is not different between HDR and SDR” is directly contradicted by ITU-R BT.2408.
The document states that different formats have different “native looks” due to their different OOTFs, and that these differences “can be substantially characterized as saturation differences.”
The ITU analysis ranks the average saturation levels among formats: HLG < HLG traditional colour < PQ < BT.709 < BT.2020.
It notes that “the HLG format, by design, has the lowest saturation of all formats because it preserves the chromaticity of the scene as imaged by the camera; all other formats increase saturation compared to the scene.”
These differences are inherent to the technical formats themselves, arising from their OOTFs, and are present before any artistic “saturation adjustment” is applied.
CONCLUSION: BT.2408 shows that HDR and SDR formats have different saturation by design.
The Shadow & Contrast Argument
QUOTES: “there is absolutely no difference in the shadow range or contrast between BT1886 (Rec709) and HDR.” “Shadow range and detail is identical in HDR and SDR…” — Alister Chapman
CLAIM: “Shadow range and detail is identical”
FACT CHECK: False. ITU-R BT.2408 Annex 6: “The HDR formats (particularly HLG) preserve a higher luminance near black than the SDR formats, so the HDR formats show/preserve more detail in the dark.”
CLAIM: “No difference in shadow range or contrast”
FACT CHECK: ITU-R BT.2408 Clause 7.10: “the subjective mid-tone contrast of a BT.709 image shown on a reference display in the reference viewing environment is greater than that of either PQ or HLG.”
CONCLUSION
Both of Chapman’s claims are contradicted by ITU-R BT.2408.
Terminology Errors
QUOTES:
“In most real world scenarios there aren’t many hues that BT-1886 (rec-709) can’t already accurately reproduce…”
“There is very little difference between BT1886 and Rec2020 in red and blue other than lines of purple which almost never actually exist in the real world.” — Alister Chapman
REALITY CHECK: The statements conflate color spaces and transfer functions. BT.1886 is an EOTF, not a color gamut. The hues are defined by Rec.709.
The Standard
The Scope of ITU-R BT.1886 defines only a reference EOTF.
As defined by ISO 22028-1, a color space requires three components:
1. RGB Primaries
2. A White Point
3. A Transfer Function
BT.1886 provides only (#3). It contains zero specifications for (#1) primaries or (#2) a white point. These are exclusively defined by ITU-R BT.709.
CONCLUSION: Conflating color spaces and transfer functions is pedagogical malpractice.
Sources:
- ITU-R BT.2390
- Scott Miller, Mahdi Nezamabadi, Scott Daly. “Perceptual Signal Coding for More Efficient Usage of Bit Codes”. 2012 SMPTE Annual Technical Conference & Exhibition.
- ITU-R BT.2408
- ITU-R BT.1886
- ISO 22028-1
The Daejeon Chronicles 