This analysis reveals why Steve Yedlin conflates Rec.2100 with PQ ST 2084, how he overlooks a crucial requirement of WCG, re-examines his “data hog” fallacy, and dismantles his proposed replacement for Rec.2100.
The Conflation
≠ ITU-R BT.2100. It is a perceptual quantizer (PQ).
“It [PQ] is standardized in Rec. 2100 and also as SMPTE ST 2084.” – Wikipedia
“ITU-R BT.2100… (Also known as SMPTE ST 2084).” – Steve Yedlin
As the slide demonstrates, Yedlin’s misreading leads directly to a fatal error: claiming ITU-R BT.2100 = “SMPTE ST 2084”.
He reinforces this verbally:
“We’ve got Rec.2100… it’s pretty much the basis for all HDR that we’re actually using. And this one is also known as SMPTE ST2084 [points to SMPTE ST2084 with cursor]. And that’s because two different standards bodies have codified the exact same color space. This [pointing cursor at ITU-R BT.2100] is the International Telecommunications Union. And this [pointing cursor at SMPTE ST2084] is the Society of Motion Picture and Television Engineers.” – Steve Yedlin, Debunking “HDR”
Whether the specific Wikipedia entry above was the source is less important than the nature of the error itself: a superficial engagement with the standards, possibly fueled by cognitive bias. It also allows him to play mind games, like dismissing HLG—an HDR format that uses the HLG transfer function and BT.2020 color primaries—as “just yet another SDR colorspace.”
“My HDR demo that’s nominally about HDR kind of almost secretly is a color space demo more than an HDR demo in a way.” – Steve Yedlin
The “Data Hog” Fallacy: 8-Bit PQ
“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. So that means at ANY data rate, the SDR curve is gonna be better.” – Steve Yedlin, ASC
Yedlin attempts to prove that PQ ST 2084 is inefficient by invalidly testing it at 8 bits. ITU-R BT.2100 explicitly specifies the use of 10-bit or 12-bit precision. Yedlin intentionally cripples PQ by testing it in an 8-bit container, while testing BT.1886 in its native SDR range, practically guaranteeing banding.
His conclusion is an inversion of reality. Yedlin condemns PQ as a “data hog,” but ITU-R BT.2390 proves that BT.1886 would need >12-bits for HDR to avoid banding. PQ delivers artifact-free HDR at 10/12-bits.
Yedlin’s Crucial Oversight
Yedlin insists “Wider gamut is not a distinction between HDR and SDR.”
While BT.2020 does technically define a path to SDR/WCG, it’s effectively unavailable to consumers via streaming services, which exclusively deliver WCG content through HDR formats.
The reason the industry never implemented a mainstream SDR/WCG system is in all likelihood because it would have required a 10-bit delivery and display standard without the “marketing sizzle” of HDR’s highlights. It was never cost-effective.
I think there was a brief period in the mid-2010s when it [SDR BT.2020] was considered a viable stepping stone from SDR NCG to HDR WCG, but once TV manufacturers jumped on the HDR bandwagon there just wasn’t much value left in SDR WCG to justify investing in it. – Alex Zambelli, Sr. Platform Manager, Dolby Laboratories
A key point overlooked in the presentation is that the ITU-R BT.2020 standard explicitly specifies 10/12 bits. This is a requirement for any system using the Rec.2020 color primaries, whether paired with the BT.1886 EOTF (for SDR/WCG) or the PQ EOTF (for HDR/WCG).
The Why
The PQ EOTF was designed using the Barten Ramp, which is only applicable to monochromatic images.
Dolby’s researchers [1] realized this wasn’t sufficient, which is why they had to conduct a separate study using the “JND Cross” test and CIEDE2000 to ensure that color images were similarly artifact-free.
Their conclusion: “For noise free EDR images such as animation, CGI or graphics, using 12-bit PQ ensures that both monochromatic and colour banding artefacts are below the visible threshold. For captured images that contain some noise, using a 10-bit PQ representation is adequate, but care needs to be taken, especially with graphics.”
Therefore, when Yedlin states: “It’s a data hog,” he reveals a profound ignorance of the very standards he’s criticizing.
He’s mocking a fundamental engineering requirement as if it were a design flaw. The bits are not a “trick“; they’re a necessity for fidelity in a wide-gamut system.
Yedlin’s Folly
“I don’t know why they came up with this PQ curve.” – Steve Yedlin
Yedlin fantasizes about returning to the halcyon CRT era, proposing “going back to a relative system with a gamma-style transfer function [and] keeping a wide gamut but only going as wide as is actually used in practice (which is more like P3 than rec2020).”
However, his scheme is physically impossible. A pure gamma function would require substantially more than 12 bits to avoid banding in HDR, as illustrated in figure 13 of ITU-R BT.2390, a document Yedlin appears not to have consulted.
As Charles Poynton [2] noted a decade ago:
“Today’s BT.1886 is not capable of HDR; in order to accommodate HDR content in the transmission chain, we’ll need an HDR-capable quantizer.”
Furthermore, there would be no benefit to creating a new HDR standard with a gamut smaller than P3. Doing so would be a step backward, not a step forward for HDR technology.
- D. G. Brooks, “The Art of Better Pixels, IBC Conference Paper, 2014.
- Charles Poynton, Stessen and Nijland, “Deploying Wide Color Gamut and High Dynamic Range in HD and UHD,” SMPТЕ Motion Imaging Journal, April 2015.
Sources: Debunking HDR, ITU-R BT.2100; ITU-R BT.2390; ITU-R BT.2020; ITU-R BT.1886; SMPTE ST 2084; [yedlin.net]
The Daejeon Chronicles 
Leave a Reply