We Need a Definition for a High Dynamic Range Display

I am at the SMPTE Fall Conference in Hollywood, where I am learning all about the ecosystem for High Dynamic Range (HDR), Wide Color Gamut (WCG) and High Frame Rate (HFR), among other topics. That means ways to capture, process and produce, encode, deliver and display such content. This of course, is a hot topic and there are rapid developments happening almost on a daily basis. But lots of definitions and standards are still evolving. Let’s just focus on the definition of an HDR display, for example.

Here is a tentative partial definition, which mostly follows the definition put forward by CEA, but is it enough?.

  • An HDR display must support a new gamma curve that can provide better gray scale rendering for a wider range of luminance values. That will likely mean support for the SMPTE ST 2084 HDR Electro-Optic Transfer Functions (EOTF), the so-called PQ gamma curve at a minimum, with other EOTFs, like the BBC/NHK Hybrid Log Gamma (HLG) EOTF, implemented as an option. It should also support SMPTE ST 2086 Mastering Display Metadata. CEA 861.3 provides for such support.
  • An HDR display must support signaling to allow the TV to identify HDR content and to allow the source to know the capabilities of the HDR display. That means the display EDID data block has details on the max and min luminance and other parameters that can help optimize the HDR rendering. It also means that the display should support HDMI 2.0a, which allows for delivery of HDR signaling and 4K content at 60 fps.

So far, so good. The problems start to crop up when you want to specify color, contrast or luminance ranges.

For example, LCD HDR sets need dynamic backlights to increase the contrast. 1D adjustable backlights are possible, but 2D direct backlights with hundreds of adjustable zoned are much better. OLEDs are emissive so can be turned way down for very deep blacks. Some display makers say this dynamic contrast is infinite as there is no light if you turn off the LEDs or OLED pixels. That’s not really true as there is usually some leakage and measurable output even with a zero video signal. But infinity doesn’t really tell you anything, so should there be a minimum contrast level to be an HDR display?

Measuring contrast is also much trickier in an HDR display. A full white/full black measurement is one way, but others are suggesting a black background with 5 small full white areas – a sort of modified ANSI pattern as being more representative of what the capabilities are. Maybe you should display a gray scale step ramp and call the contrast the points at which you can just discern a difference in luminance in the black and white ends of the ramp.

Sony has suggested going even further with this approach and doing the same thing with red, green and blue ramps. This will help determine the color tracking capabilities over the dynamic range of the display.

Should a definition include a peak luminance level in addition to the contrast? Or maybe a black level floor and minimum contrast is sufficient? LCD sets will be better with a peak luminance spec, but OLEDs will excel on black levels.

But even the definition of peak luminance is not so easy to pin down. For example, Sony says that on its HDR LCD TV with direct backlights, the peak luminance is variable. That means if part of an image is quite dark but there is a very bright area in another part of the image, the power that could go to the LEDs in dark area can be diverted to boost the peak luminance in the bright area. So, the TV might have a peak luminance of say 800 cd/m² for full white, but can support small areas of peak luminance that can go to say 1500 cd/m². Both numbers tell you something about the performance.

Color performance is not so easy either. The UHD specification is for the BT-2020 color gamut, but currently there are no tolerances on the primaries so no display can claim full compliance. This will have to change. So what is acceptable? OLEDs are close to the P3 color gamut, while LCDs can vary from a little bigger than 709 to 90+% of the 2020 color gamut using quantum dots (Nanosys says 97% with the right filters – Man. Ed.). Is just “bigger than 709” acceptable, should it be P3 as a minimum or full 2020 compliance once we know what that is?

The color capabilities of the display are a volume, not the simple 2D CIE diagram, so should we be specifying a minimum color volume that includes a color standard such as P3? Displaying a blue, cyan or red color at 100 cd/m² may be possible by the display, but can it show the same color at 500 or 800 cd/m²? This is important information that a color volume representation might help illustrate.

Study Group On High Dynamic Range HDR Ecosystem

SMPTE issued a new report on HDR at the conference this week, which I have not had time to read, but can be downloaded here. I glanced at the table of contents and see lots of good topics covered, but I did not see a clear definition of an HDR display.

I presume this definition is something the UHD Alliance is currently debating, but I have not contacted them to ask (They told me this was one of the current hot topics at IBC – Man. Ed.). If you have an opinion on this topic, please feel free to comment below. (CC)