Quested Provides Perspective on ITU-2100 HDR Spec

The chairman of the ITU-R subcommittee responsible for the recent release of IUT-2100 is Andy Quested, who is also an employee of the BBC. We had a chance to speak with him recently to gain his perspective on the new document.

Andy Quested

We caught up with him while he was attending the London-based HPA Retreat event (We had a reporter at that event, so for more coverage see our report for subscribers – Man. Ed.). He noted that this was the first time the SMPTE-backed event had been produced in Europe. While he found the content stimulating, it was quite Hollywood centric. He believes it will become more European focused in the future though.

The new document from the International Telecommunication Union (ITU) is officially called ITU-R BT.2100. In a way, it is a follow-on recommendation to BT.2020, which has been widely discussed and adopted in the professional community. The full BT.2100 document can be viewed at: http://www.itu.int/rec/R-REC-BT.2100.BT.2100 deals with the capture, production and international program exchange of HDR/WCG content – not the distribution of it.

We asked Quested what the sticking points were in developing consensus on BT.2100. First, he explained that ITU is under the United Nations and these are “recommendations” with no formal voting. He said they are developed by consensus and consensus is reached when there are no more objections to the draft language in the document.

Continuing, he said there were three main sticking points:

  • Understanding what dynamic range is
  • Avoiding large luminance changes that may have human factor issues
  • Why two options were important

While the documents of the ITU are drafted by technical experts, not all participating in the committee are experts, so there was a lot of time spent making sure the language was right to explain what high dynamic range is and is not. It is a hard concept to grasp, he said, but he believes members now understand this is not brighter pictures but an “expansion” of them (although the definition of what range counts as HDR is still not defined).

The second point has to do with abrupt changes in the average luminance levels either within a program or when going from a program to a commercial. For example, if a viewer is dark adapted and watching a dark movie, then is hit with a very bright commercial, they may not find that very pleasing. BT.2100 does not address operational issues but the ITU will start to look at this now.

BT.2020 featured only HD or Standard Dynamic Range curve (EOTF), but BT.2100 adds replaces this with the PQ (Perceptual Quantization) and the HLG (Hybrid Log Gamma) options. Many think that PQ (Perceptual Quantizer) EOTF, standardized as SMPTE ST-2084, is best suited for file-based HDR workflows and that HLG is better for real time workflows. Quested didn’t quite agree with this characterization stating that he thinks the two approaches just come at the problem from two different directions and cultures. But each meet the needs of these groups allowing a transition from SDR to HDR in the way that suits each best, which is why the committee agreed that both transfer curves were needed.

As a result, BT.2100 defines all the formulas for converting from linear light to non-linear light and vise versa using the PQ or HLG transforms, and formulas for converting between the two  These details are expanded in the ITU Report ITU BT.2390 (https://www.itu.int/pub/R-REP-BT.2390). This is important as HDR content may be captured for a live broadcast using HLG, but it might be played back on a channel that uses the PQ curve support.

So how does the TV know which transfer curve the content was mastered in?  Quested says this could be carried in several ways but a simple Auto Format Descriptor (AFD) may be the easiest way. This is not really metadata but it kind of acts like it. It carries some basic information about the format of the signal (4:3 or 16:9, stereo or DTX, etc.) that is mainly used to help ensure smooth format changes when changing channels. How this will be carried and acted on by the TV or set top box still needs to be worked out, which is one of the projects Quested will be working on next.

Note that the recommendations of BT.2100 do not include the carriage of metadata within the production facility for either the PQ or HLD based workflows. Such metadata or AFD data can be added at the point of distribution.

BT.2100 copies much of the basics from BT.2020, but it now adds support for HDR, WCG and high frame rate to the 1080p format as well. Clearly, members are seeing a value proposition for adding so called “better pixels” to 1080p signals.

To explain why this makes sense, Quested summarized some research the BBC recently did on the distance that viewers sit from their TVs in the UK. He started by noting that they did a study 12 years ago just before HD took off, with a very small sample and found the average viewing distance was 2.7 meters from the TV (~8.9 feet). (http://www.bbc.co.uk/rd/publications/whitepaper090) A more rigorous survey was done a year ago with exactly the same distance result! (http://www.bbc.co.uk/rd/publications/whitepaper287)

Certainly the TV sizes 25 years ago were much smaller, and today, the average is somewhere between 42” and 55”, he thinks. The point is, with these screen sizes and viewing distance, you can’t see the increased resolution of a UHD TV compared to a HD TV. As a result, adding HDR to 1080p seems like a good value proposition and will likely co-exist with native UHD services during the transition from HD to UHD.

BT.2100 also lists a number of frame rates that can be used. Quested sees the HDR, WCG, HFR and resolution parameters as options that content creators can configure to serve various market needs and segments. Sports may very well benefit for higher frame rate and HDR, but operators may choose roll out 1080p versions to optimize bandwidth, for example.

BT.2100 Table 1

We also asked Quested about the RGB primaries in BT.2100, which remain the same monochromatic wavelengths as in BT.2020 (630, 532 and 467nm). Many believe producing a display with these primaries requires lasers which can introduce speckle and metameric issues (colors can be perceived differently be different people when the primaries are very narrow). Quested said there are no tolerances on this spec as it is really an aspirational spec. It may take 5 or 10 years to create acceptable and fully compliant displays, he noted, but they wanted a goal for display makers to try to achieve – a good standard should not be out of date the day it’s published!!

The new standard also specifies a “reference viewing environment for critical viewing of HDR program material.”  This includes recommendations for background and surround color temperature and luminance, viewing distances based on resolution, as well as the peak (>= 1000 cd/m²) and minimum luminance (<= 0.005 cd/m²) levels. These levels may only be achievable in a very few expensive professional monitors today, so they are also a bit aspirational.

Finally, there is also a new color difference option available in BT.2100 referred to as Constant Intensity color difference signal calculation. This is an alternative to the conventional way of creating color difference signals in the YCbCr format.

Constant Intensity is based on the idea of trying to preserve the hue of a pixel when going from an RGB to a color difference format like YCbCr (less data for distribution) and back to RGB at the display. Quested said that hue shifts have been apparent in the YCbCr system for some time, but they are subtle. The Constant Intensity system called ICtCp would minimize these effects.

ICtCp has also been proposed by Dolby to perform color space conversions in the TV or set top box when moving from a 2020 color space to 709 or whatever color capabilities the display may have. Current conversions often introduce shifts in the hue, which again, would be minimized by using ICtCp.

BT.2100 is an important new document. It does not obsolete BT.2020 – it builds upon it. But there is more to do.