What viewers really notice and care about is HDR. HD HDR offers 95% of the 4K experience without the complexity and cost of transmitting 4K.
We shooters, and the broadcasters that hire us, have long had a noble and worthy goal: that what we see with our naked eye should match what we see on our TV at home. This goal may have seemed elusive or impossible in the past but today, given the advances in technology, especially HDR, the dream of 1:1 capture and display is not only realistic but is already here.
Some shooters may remember the implementation of HDR from a decade ago that utilized bracketing and blending of multiple exposures. Despite the obvious blurring and reduced resolution, it seemed to work okay for still photography - the iPhone being an excellent case in point – but it worked horribly for motion video and television applications. This was especially apparent in night scenes like fireworks displays. The merged data sets offset slightly in time looked horrible, and only got worse after compression downstream over IP or cable.
The NTSC and PAL television systems devised decades ago have long been constrained by the limitations of CRT displays, which offered a mere 6.8 stops of latitude and a 100:1 dynamic range. For shooters, the ability to produce smooth, flattering flesh tones is the measure of which any imaging system should be regarded, and in this respect, our traditional standard dynamic range systems have been limited to +2 stops above and -4 stops below reference grey (18%). Excess values beyond +2 stops were subject to ugly clipping, while lesser values below -4 stops sank into the murk to never be seen again.
Today’s large-sensor cameras, like the Sony PXW-FS7M2, feature 14 or more stops of latitude, and can capture sufficient highlight detail in low light to produce excellent HDR. Close attention to exposure in-camera is critical.
Early HDR technology
By the end of the 1990s, broadcast cameras were capable of capturing over ten stops of latitude with enhanced highlight control and improved tonality. The Panasonic VariCam, for one, introduced in 2004, employed a low-contrast FILM_REC curve offering 10.8 stops of latitude and a 1500:1 dynamic range.
Originating on film offered some respite from the tyranny of television’s constrained dynamic range, at least, for image capture. With a contrast range of 4000:1 and 12 stops of latitude, we reveled in the five-stop advantage over video origination. We still had to face the music later, however, understanding the ultimate output to 709 gamma CRT displays.
Example low-contrast image produced with the FILM_REC curve.
In the telecine suite, we spent endless hours peering into the film’s response curve, squishing and shrinking the highlights into the knee of the output signal. The advent of a knee control in professional cameras helped in this respect, reducing the risk of clipping in the highlights during capture. However, the anemic low dynamic range CRT displays of barely 100 nits continued to be the overarching problem.
The benefit of HDR is apparent to even the most unsophisticated viewer. Many current HDR TVs can only display 10 - 11 stops of dynamic range so some clipping of highlights may still be evident in the latest generation sets. Image courtesy of Samsung. Click to enlarge.
By 2007, home displays had grown much brighter. Plasma sets running at full blast could output 200-250 nits, and LCDs fitted with potent backlights could output 250 nits or more. Therefore, while displays grew brighter, and cameras were able to capture more dynamic range than ever, the 709 gamma-based CRTs in viewers’ homes were still limited to the scant 6.8 stops.
Today’s HDR technology
Today, most professional cameras can capture 14 stops without clipping, and home LED and OLED displays may exceed 1000 nits. Without HDR, the brighter displays, however, only make things worse, as the shortcomings of 709 gamma appear even more egregious on the more luminous screens. Clearly, the once-venerable 709 signal, introduced in 1953 as a byproduct of NTSC and CRT displays, needed to go away, and go away quickly.
Television and film professionals understand that capturing audio without sufficient headroom can lead to clipped, unacceptable recordings. Video recording acts in the same way, with nothing much useful registered above 100% brightness, or 100 nits peak white in a traditional TV system. The implications of this are profound if the goal is to produce vibrant, life-like images. The specular highlights reflected from a chrome surface, for example, or a bright light source inside the frame, could not be properly displayed above the normal white point. Today’s cameras and displays are capable of doing so, but only if broadcasters implement the wider-range HDR signal.
The 8-bit 709 gamma curve, based on a 1953 standard, can display less than 7 stops of dynamic range. This limitation persists even today on non-HDR displays.
CRTs have always required significant gamma correction, but with today’s displays, such crippling correction is no longer required. Current HDR implementations allow 1000% more headroom compared to the old 709 gamma curve. Based on human response studies pioneered by Dolby, HDR today can produce a truly impressive viewing experience on consumer displays that are approaching, and in some cases, exceeding 1000 nits.
HDR utilizes anElectrical Optical Transfer Function(EOTF) to determine how the video signal is ultimately displayed. Dolby’sPersonal Quantizing (PQ)system, aka SMPTE 2084, accurately reflects the human visual response, with a very precise assignment of values for grey scale and luminance. Dolby PQ’s curve is capable of producing very flattering images and is ideal for mastering feature films, but PQ also poses some serious challenges for broadcasters whose live productions require precise tweaking and matching of multiple cameras.
Most critically for broadcasters, the PQ/SMPTE 2084 standard is incompatible with current 709 gamma-based processing, and while PQ can accommodate displays up to 10,000 nits, this level of intensity on a home theater screen would be most painful to viewers, akin to peering into a welding arc. Obviously, this is not the ‘enhanced’ viewing experience that most broadcasters are seeking in a new television system, HDR or otherwise.
Dolby’s PQ system, derived from human visual response studies, is very accurate, providing much greater screen brightness up to 10,000 nits, and wider dynamic range than conventional 709 gamma displays. Hybrid Log Gamma (HLG), (represented in the graph above), unlike PQ, is compatible with existing 709 TV systems, and is thus preferred by broadcasters like the BBC, NHK, and others. Click to enlarge.
The Hybrid Log Gamma (HLG)system is preferred by broadcasters because it is compatible with conventional 709 displays. A log curve, implemented above 50%, accommodates the log recordings captured increasingly in today’s broadcast cameras. By setting a scene white reference at 50%, HLG provides more than sufficient contrast and dynamic range for satisfactory viewing of programs on a traditional SDR display. The 709 non-HDR set will simply display the compressed, slightly yellow highlights we’ve grown accustomed to, since the advent of knee processing.
On an HDR HLG display, the upper log portion of the signal is expanded to accommodate the greater contrast and increased dynamic range. With peak levels on some top-end displays in the range of 600-800 nits, the displayed image has a much more realistic look with three additional stops of dynamic range in the highlights. For shooters, given the compromises inherent to 709 displays, HLG’s increased bandwidth in the highlights is truly a gift from Heaven.
While compatibility with non-HDR sets is a major advantage, HLG has a significant drawback. The less-efficient hybrid system complicates the transcoding to other HDR profiles like PQ. Thus, the strength and appeal of HLG is almost entirely in the broadcast realm where some degree of closed production is possible. For the major studios and feature film producers with greater resources and time available for post-production, the more versatile and precise PQ flavor of HDR is preferred.
Even though HDR10, a 10-bit system, specifies a screen brightness of 1000 nits, most viewers prefer a more comfortable 300-400 nits of for general viewing. While higher-end HDR TVs feature a maximum screen brightness of 500-600 nits, most inexpensive non-HDR sets today average around 200-250 nits.
In the near future, it seems likely that camera manufacturers will offer an HLG menu option to avoid the hassle and inefficiency downstream of HDR transcoding. More versatile HDR-friendly cameras can be expected, as broadcasters come to grips with HDR and the unique challenges it represents to the industry.
4K—a fool’s gold?
At the moment, television networks are highly constrained with respect to IP delivery, typically managing a scant 25Mbps of bandwidth to the consumer. Within this narrow channel, a typical broadcaster must multiplex multiple program streams. In addition to its flagship programs, a broadcaster might also transmit several low-bit rate streams such as a secondary news channel and home shopping. In this environment, broadcasters cannot afford to transmit a single 4K HDR program but can send multiple streams in HD. For broadcasters looking ahead to HDR, the choice could not be clearer: HD HDR, not 4K, is the future.
Given the seeming non-stop hype for everything 4K these days, it is ironic that most viewers at home cannot reliably distinguish 4K from 2K or even HD programming. There may be several reasons for this: our home sets are too small, large inexpensive LCD screens lack contrast, or, more simply, we sit too far away from our TVs.
HDR, on the other hand, is another kettle of pixels. Its more compelling images are readily apparent to even the most unsophisticated viewer, with a 4000:1 contrast ratio, and colors that no longer clip at 100% but at 1000%. When all is said and done, HDR delivers 95% of 4K’s wow factor, without the hassle and expense of actually transmitting 4K.
HDR—No single solution
As HDR progresses, there are a few variations to contend with and to sow confusion. Theoretically, an HDR-enabled TV reads a flag contained in the signal metadata and displays the proper flavor of HDR.
HDR10, a subset of PQ, utilizes the Rec. 2020 color space at 10-bits, and limits the screen display intensity to 1000 nits, a reasonable encoding compromise, given that Blu-ray and most feature films are graded for 1000 nits. HDR10+, from Samsung and Amazon, introduces dynamic metadata to allow scene-by-scene and frame-to-frame manipulation of brightness levels. Most new HDR-enabled sets support HDR10 and HLG by default. Dolby Vision PQ, however, despite its greater precision and efficiency, is not as widely supported due to the licensing fee imposed on set manufacturers, who prefer open-source HDR, like HLG, for this reason.
Unlike Netflix that uses a handshake to determine a receiver’s capability and then send the proper stream, broadcasters are not likely to employ a handshaking routine anytime soon. For broadcasters, it isn’t practical to deliver content to millions of viewers all demanding different bit rates. Networks operating a launch-and-leave system will likely set a delivery rate of 25Mbps and let it go at that.
HDR means new cameras
For camera folks, the advent of HDR has major implications. HDR requires a minimum of 12 stops of latitude in-camera, and fortunately these days, most professional cameras can do that or more. The problem is, most current HDR TVs can display only 10 to 11 stops (at 500-600 nits) so some clipping of highlights may still occur even in HDR displays.
The Panasonic VariCam AJ-HDC27, introduced in 2004, featured a low contrast FILM_REC curve that accommodated 10 stops of dynamic range. The extra 3-4 stops captured in-camera allowed filmmakers to create more real-looking images with smoother flesh tones during color grading. Unfortunately, broadcasters seldom had the time, resources or inclination, to achieve greater dynamic range in this way.
HDR requires a lot of light during capture, so cameras fitted with extra-sensitive pixels tend to work well. Large-sensor cameras like the Panasonic VariCam with large five-micron pixels are very efficient at capturing light and thus are particularly well suited for HDR applications. Small-format cameras, on the other hand, like Panasonic’s 1/3 type AJ-PX270 camcorder or Sony’s 2/3-inch HDC-4300 4K camera is at a disadvantage due to the inherently smaller pixels that struggle to capture sufficient light and highlight detail in poorly illuminated conditions. The tiny sensors in cell phones suffer for the same reason; their minuscule pixels are barely able to produce eight stops of dynamic range.
For broadcasters, HDR is the most significant advance in imaging since the advent of color. Forsaking 8-bit 709 processing and clipped ugly highlights, HDR in combination with OLED displays can finally deliver on the dream of presenting truly realistic images that viewers can appreciate – and, hopefully, be willing to pay for.
HDR is more than a marketing term. However, we need to get it right or we’ll be stuck with SDR and 709 processing for the next 60 years.
Barry Braverman is a cinematographer and digital media expert. He is a frequent contributor to The Broadcast Bridge.