I work a booth at NAB so I didn’t see a lot this year, but I saw enough. Here are a few things that stuck out during my week in Las Vegas.
LED LIGHTS, AND THE QUALITY THEREOF
Years ago I spent a lot of time writing about infrared filters and camera deficiencies. I shot a lot of tests and posted a lot of articles, and I became known as the “IR guy.” More recently I’ve written articles about color science and camera color, so now I’m known as the “color guy.” I’m much happier being the color guy, especially as it has landed me a gig working at the DSC Laboratories booth at NAB.
DSC Labs makes the best video test charts in the world, and in return for flying me to Vegas and putting me up I spend my mornings teaching lessons in how to use charts to examine and set up cameras. In conjunction with Adam Wilt, my fellow PVC writer who works the booth in the afternoons, we strive to blow people’s minds with the inside scoop on cameras and their innards. Thanks to Sony, who loaned us the amazingly versatile FS7 camera and dazzling PVM-A170 OLED monitor, and Tektronix, who loaned us a waveform/vector scope, we were able to do this very well, and with pictures.
We don’t always bring lights to Vegas to light our demo charts, and we were asked early on if we’d like to use someone’s new LED lighting product to illuminate our charts. The lights were delivered, and as I was not familiar with them I immediately powered them on for evaluation. I was immediately suspicious as the tungsten-balanced light looked very warm to me, and excess warmth in LED light is usually due to an excess of green. (Poor quality tungsten-balanced LED lights often exhibit too much green, with a yellow cast, or too little green, with a magenta cast.)
Adam being Adam, he opened up his bag and pulled out his iPhone and a Luxi lumisphere. He then fired up his Cine Meter II app, which turns the iPhone’s camera into a color meter and offers gel correction recommendations. It confirmed that this lamp was too green and required 1/4 plus green to correct, although simple gel corrections almost never work on LED lights as the rolling spectral cut of a gel rarely matches the sharp spectral peaks of a bad LED light.
After that, Adam—once again, being Adam—fired up his handy Lighting Passport Spectrometer and measured the light’s spectrum. It was severely compromised, with a huge red spike, a huge yellow spike, very little blue and lots of gaps. It didn’t seem like the right light for us, given that a significant portion of DSC’s product line is color calibration charts, so we set out to look for something else.
NAB is a funny place: one tends to run into the same people every year, and when you do it’s always a pleasure to catch up and see what they’re up to. I’d met Tama Berkeljon of Outsight at an NAB show years before, and when he stopped by our booth for a chat on the Saturday before NAB opened we asked if we could check out one of his latest CreamSource units. He happily loaned us a CreamSource Mini, which Adam immediately checked out with Cine Meter II app and his Lighting Passport spectrometer.
Not only did the CreamSource unit look good by eye (its tungsten output indeed looked “creamy”—very neutral and not yellow/green) but Cine Meter II read it as dead-on 3200K with no green/magenta shift. Additionally, the Lighting Passport showed it had a very broad spectrum. It did particularly well in the red portion of the spectrum, where it got a very high grade on the R9 CRI patch. (R9 is a saturated red CRI test target. Flesh tone requires a bit of this hue in order to appear alive and healthy but cheap LEDs have a hard time generating enough of it to do a good job.)
The CreamSource Mini performed so well at generating both tungsten and daylight that we used it the entire week to light our color charts for camera demos. I’ve long been wary of using anything other than remote phosphor LEDs for critical color but CreamSource has convinced me to be a bit more open minded.
LED and fluorescent lights can be very hard to assess. We can roughly measure their color (although the rule of thumb with color meters is to use about half the green/magenta correction that they recommend) and their spectra, but neither can accurately predict how that light will appear to the eye. Spectral graphs are particularly difficult to interpret: we can determine whether a light has gaps in its spectra and if its spectrum is well rounded, but ultimately we have to judge the results by eye as it is impossible (at least for me) to look at a spectral graph and predict how well a light will function. I can look after the fact and see why it might have failed, but I can’t visualize results from the graph.
I did a lot of LED light eyeballing at NAB and it was interesting, to say the least, to see how few LED lights, operating at standard color temperatures, matched each other in hue. I’m sure there were some amazing new LED lights at NAB that I didn’t get a chance to see, but I couldn’t tell which ones they were simply by looking at the color they cast on the inside of a Chimera’s diffusion. Everywhere I looked I saw LED-lit diffusion materials that were a little too warm, or too cool, or too green or magenta.
Some lights were clearly awful. I don’t know how to describe it, but some lights just feel “flat” or “electric white” (if daylight) or too yellow (if tungsten). One lighting company boasted a lamp that was rated a 99 on the TLCI scale, but by eye it looked ghastly. (I go into a lot of detail in this article about why I think the TLCI scale is usable only under very limited conditions, and I had several conversations at NAB where I learned I was not alone in this assessment.)
I later learned that my friends from Cineo Lighting were at NAB, and I know I can always get a light from them, but in this case I was very happy to use a new lamp that seemed to do nearly as good a job. (I’m a big fan of remote phosphor lights, and Cineo’s lights in particular.)
THE FEW CAMERAS I SAW
The Arri Alexa 65 was a joy to play with, although I learned that its rental list price approaches $9,000 a day. I think I’ll save it for a special occasion.
Last year RED stirred up a bit of controversy by building a basement set complete with cadaver, organs in jars and a presumed serial killer who puttered about on occasion, but this year they did something much smarter and opened up their booth, and their cameras, to vendors who were allowed deep internal access to REDs hardware and software.
I was particularly impressed by KipperTie’s collection of custom-made low con and diffusion OLPF filter packs that are built into RED’s own hardware, with their blessing, and FoolControl, a family of apps that allows full control over any wifi-enabled RED camera. Mostly, though, I was impressed that RED is allowing vendors to work with customers to build the camera they want.
I didn’t get a chance to critically examine the Canon C300 Mk2, although I’ve seen one at a local rental house and it looks very nice. It has the traditional Canon “everything looks like it was shot at sunset” look, and while I know that’s popular I prefer more neutral color. I’m looking for an opportunity to take it for a test drive and create a more neutral look that retains Canon’s gorgeous flesh tones.
I saw a little of Blackmagic’s booth, and I need to look a little harder at their cameras. I’ve shot some projects with their original 4K digital cinema camera and they turned out pretty well: the camera was noisy but the color looked fairly good. The booth footage, though, showed flesh tones that were consistently a flat warm hue (often orange or pink), which makes me wonder if they have issues with hue discrimination. In all the Blackmagic demo footage I’ve seen, flesh tones tend to be primarily one hue and lack the depth and texture that human skin normally shows.
One early demo shows two men sitting in the window of a cafe having a chat, and the camera, which is inside the cafe, frames them against the window. The interior and exterior retain detail, which is an impressive demonstration of dynamic range, but there are two red cars parked across the street and they are exactly the same hue of red. That never happens. My guess is that the cars are subtly different in hue—say, blueish red or orangeish red, which tend to be common car hues—and the sensor can’t discern the difference, making both appear to be the same hue of red.
Still, Blackmagic shows plenty of footage that looks great, so it’s clearly possible to do good work with Blackmagic cameras. Their dynamic range has clearly improved. At the same time, though, we still get what we pay for, and there are always compromises (even in high-cost cameras).
Panasonic showed off the Varicam LT, a new lower-priced single-sensor Varicam that boasts the same ISO 5000 capabilities as its predecessor, the Varicam 35. Both cameras feel a little too noisy at ISO 800, but they show the same ISO 800 noise at ISO 5000 which is very impressive. Combined with Varicam’s trademark color they should be formidable competition for the other manufacturers, but there’s some concern that Panasonic has come out with too little too late, and for too much. Recent price decreases should help spur adoption, particularly in regard to their proprietary media which started out at a jaw-dropping $2000 but has now stabilized at a slightly more affordable $1200.
I’ll be curious to see if this camera catches on in smaller markets like mine. My guess is that media prices will prevent smaller rental houses from investing short of significant interest, as most of their cameras use interchangeable media. I’m told that Varicams are seeing a lot of success in episodic television so maybe that will drive up sales numbers outside of LA. If nothing else, their color is to die for.
What really blew me away, though, was Panasonic’s HDR demo. More on that in a moment. Let me set the stage first.
WHAT I LEARNED ABOUT HDR
HDR, or high dynamic range television, was the buzzword at this NAB. Now that 3D is effectively dead, HDR’s enhanced realism seems to be the next TV-selling frontier. When done well, it’s stunning. Sadly it’s not being done terribly well yet.
The basics are simple. Normal TV sets are designed to deliver peak brightness at 100 nits. HDR sets can deliver anywhere from 500 to 4000 nits of brightness, but not across the entire screen. Rather than blind viewers with their own TVs, these high brightness levels are employed only in small areas of the screen to enhance specular highlights. This sounds simplistic, but the result is startling: when a shot includes the sun, the sun is BRIGHT. Specular highlights on water are dazzling. Neon signs in night exterior shots are stunning.
HDR isn’t just about increased brightness but increased color gamut as well. While most sets at NAB were described as “Rec 2020” the reality is that they didn’t extend much beyond P3, but that is still a significant enhancement. One demo showed a scene from The Revenant that included a partly cloudy sky, and the combination of white, detailed clouds and baby-blue sky (a hue one doesn’t see in Rec 709) showed me what I’ve been missing. It didn’t look like a TV image so much as an oil painting.
There were, however, a number of HDR failures on the show floor.
Nearly everyone had a Sony BVM-X300 $30,000 HDR monitor in their booth. This is surely the gold standard in HDR viewing, but in many cases it showed what I was missing elsewhere. What I wanted to see more than anything else was what the consumer displays looked like, and my first stop was the Sony booth where they showed a comparison between two 4K consumer displays illustrating two of the current HDR standards, SMPTE-2084 (based on Dolby’s PQ log curve) and HLG (hybrid log gamma, developed by NHK and the BBC). I didn’t see any obvious differences between the two displays, but that could have been because their footage had been horribly compromised. It appeared that the footage had been severely over compressed: shadows showed terrible macro blocking that looked like the worst noise ever, and highlights and saturated colors were terribly clipped and appeared as smudges rather than richly-textured clouds and mountainsides. Bright whites that should have rolled off gently instead appeared as if they’d been shot on early F900s with the knee engaged and overexposed. It was a hugely embarrassing display.
Nearby another display area at the same booth did a slightly better job. A single camera shot a small set containing a model of a wooden pirate ship, lots of colored glass baubles, a neon sign and some random hits of “sunlight” meant to push highlights into the brighter tonal ranges of HDR. The video feed then fed several HDR monitors: one displayed 4K HLG, another displayed 4K SMPTE-2084, and a third showed a quad split with Rec 709 and both HDR images. The quad split made a great sales piece for HDR as Rec 709 looked sad by comparison. I didn’t see any obvious gamma differences between the two other HDR processes, but I did notice that the 2084 image had more green to it. By eye the pirate ship’s sides were a slightly warm brown hue, which the HLG-setup monitor rendered accurately, but the 2084 monitor made them appear a little yellow. No one I asked, either at the Sony booth or elsewhere, understood why that was happening.
Unfortunately the camera used to shoot this scene was a prism camera, which was both a huge mistake and highly educational. Bright highlights showed prism artifacts, such as magenta and cyan fringing, and the enhanced color space exaggerated these defects dramatically. Bright areas naturally draw the eye, and it’s clear that these areas need to be carefully monitored and controlled as the eye will immediately find and detect flaws in these areas.
The pirate ship boasted white sails, and in HDR they were bright crisp white. The problem, though, is that they were also featureless. I realized that bright white highlights aren’t enough: to be truly impressive large highlight areas need to show some texture. I have no need to see featureless clipped highlights at 1000 nits; they’re bad enough at 100 nits!
I’d seen similar things on the consumer displays, where one of the demos showed a little girl running through a field of bright yellow flowers with the sun in the background. The sun produced lens flares, and while these might have been a subtle pale blue in Rec 709 they were bright electric blue in HDR. The compression errors rendered an otherwise sharp lens flare as a highly-saturated bright blue blob, and it was hugely distracting.
The X300 HDR grading monitor is stunning, and Sony’s OLEDs are the industry standard for a reason (the PVM-A170 is currently the San Francisco Bay Area on-set standard, and I won’t use anything else without a fight), but Sony sure didn’t know how to promote their other HDR products. Even the demo footage, while perfectly competent and in some cases beautiful, didn’t hold a candle to another display manufacturer who beat everyone else’s HDR demos hands down.
I didn’t spend a lot of time at the Canon booth, and I probably should have spent more. Their 8K monitor demo was stunning, and not just because they provided a magnifying glass and dared us to find the pixels. The combination of high resolution and extended color gamut resulted in an unforgettable viewing experience, although it also demonstrated that the farther we push image capture technology the more careful we have to be when shooting it. A slow crane down/tilt up move was enough to make me feel a bit seasick. I’m not saying that’s a bad thing, only that camera moves that work in a less realistic medium might be less successful when the experience is viscerally realistic.
Canon’s 8K camera demo with live HDR was disappointing. Their 8K camera was impressive as hell: they’d stripped the guts out of a C300 body and replaced them with an 8K sensor and processing, and the recording system consisted of four Convergent Design Odysseys stacked on the back. Their setup consisted of a woman in a white linen dress sitting on a couch in front of a (fake) window, through which she was backlit by (fake) sunlight. The sunlight on her shoulders caused the textured white dress to bloom, and HDR enhanced that effect significantly. Unfortunately the dynamic range of the camera couldn’t retain enough detail in the linen while maintaining decent exposure on the model’s face, so what should have been a bright-yet-textured highlight was instead a bright clipped blob. This is the video look that we all try to avoid when shooting digitally, and trust me when I say that it doesn’t look any better when the resulting highlight on your screen is 3-4 stops brighter than normal. (For the record, I couldn’t find a T-stop on the lens that held the model’s face while also holding texture in the dress, and the gentleman supervising the camera wouldn’t let me delve into the menus to find a kinder gamma curve.)
It’s clear that clipped highlights don’t work in HDR unless they are miniscule. Short of that, large bright areas should contain enough detail to be interesting to the eye without simply becoming another light source in a consumer’s living room.
It doesn’t help, though, that Sony, Canon and other manufacturers seem to be hiring classically-trained video lighting directors to light their booths. The power of 4K HDR is not just in displaying brighter and more saturated and textured highlights, but also showing us gradations and subtleties of hue, tone and texture that we can never experience in Rec 709. With this in mind, I would have lit their sets with soft-yet-contrasty top and backlight, while using hard light judiciously to bring out textures and fine detail. Instead nearly every set was lit with a ring of fresnels, often undiffused, and aimed downward at very steep angles, resulting in flatly-lit sets with many, many hard shadows. It seemed to me that this was exactly NOT the way to show off a high dynamic range, high saturation capture medium. When details matter, why blast them away? It felt as if many vendors were lighting their sets using the visual equivalent of a fire hose. (Know your audience!)
Canon did show a still life in HDR in another section of their booth that was beautifully lit, and I suspect that was because Curtis Clark, ASC was scheduled to give a talk in that very spot. Here Canon showed how an image that was properly exposed in log, specifically for HDR display, appeared compared to the final result. One image showed an original Canon Log 3 image that was extremely bright and bordered on being washed out, and the other showed the same image in HDR as a beautiful, high dynamic range scene with subtle dark tones against gorgeous specular highlights on shiny surfaces. It’s becoming clear to me that we’re going to have to pay a lot of attention to log waveform values when shooting for HDR as most log curves place diffuse white at around 60%, and all the HDR magic occurs above that point.
While Canon’s live HDR demo with the 8K camera was less than stellar, they did a great job in the dedicated HDR section of their booth.
Panasonic’s HDR demo was, by far, the best. It appeared adjacent to a ColorFront color grading station, where a colorist gave HDR grading demos. The footage consisted of Panasonic’s original Varicam demo reel re-rendered in HDR, as well as some new night exterior and interior footage that took place both in city streets and in a boxing ring.
It was beautiful. Really, amazingly beautiful. Part of this is Panasonic’s color, which appears to discriminate between subtle hues better than most other cameras around. Part of this is due to Panasonic actually spending the money and making the effort to make sure their images were shot to be cinematic instead of simply eye catching. Sony showed nature footage of badly compressed fire eaters, flowers, mountains and moose on their displays, while Panasonic showed shots that could have been pulled straight out of a feature film or a high-end car commercial.
Bright red, blue and green primaries are any camera’s most basic requirement, but such hues appeal primarily to children and people who are not visually sophisticated. The hues in between are harder to master and hold more appeal for artists and high-end consumers. Panasonic’s color palette consisted of a lot of secondary and tertiary hues that I don’t see properly reproduced by other cameras, and the HDR process was used tastefully and not excessively. The running footage of a white car driving down a neon-lit street was particularly appealing as the white highlights of the car popped off the screen and made it feel very clean and crisp. A crisp, white car driving down a brightly-colored street at night, in HDR, is more dazzling than one might expect—certainly more so than a moose in broad daylight shot at 60fps.
I have no doubt that Sony and Canon are capable of pulling together the resources necessary to show off their HDR capabilities. Canon succeeded in their HDR demo, although they missed the mark on their 8K live demo. Sony missed the mark completely with footage that was okay at best before it was badly compressed. (They did show aerial footage of Las Vegas in HDR, but as the bright neon hues were clipped the HDR effect was lost.) Panasonic showed footage that made my jaw drop… but my biggest complaint is that between Sony, Canon and Panasonic my jaw should have dropped at least three times.
My last stop was Dolby, who showed HDR-graded clips from several films including Man of Steel. This was also an impressive display, but it made me worry a bit about how far we are from supplying viable HDR displays to the consumer market.
I stepped into a darkened room in which two monitors showed clips from expensive feature films. The one on the left was clearly HDR, and the other appeared to be a standard edge-lit LED display. The Man of Steel clips were excellent HDR material: the images on the leftmost display were dark but richly textured, with HDR highlights limited to distant explosions and textured pools of light. I asked a nearby Dolby employee to explain what I was seeing, and she told me the monitor on the left was a Dolby display with localized lighting: this means that, instead of the screen being edge lit or broadly backlit, the screen was broken up into small sections and each of those had its own very bright LED. To my eye this looked very similar to what I’d expect from an OLED display, and it looked phenomenal.
What floored me, though, was what she said next: the other monitor was an edge-lit consumer display, and it was also supposed to be HDR!
It DID NOT look like HDR. Not at all. If that’s the best edge-lit HDR set around, then that technology simply doesn’t work well enough in HDR to be usable.
This is why I think HDR is a few years away yet, at least for consumers. I’m not sure that anyone is making consumer-based HDR TVs that are significantly different from what they have at home already. Until that advancement takes place I think we’re going to live in a world of future proofing for HDR, rather than turning it out for immediate viewing. That will be the subject of a future article.
Disclosure: I have done consulting work for Sony, Canon, Arri, DSC Labs and Cineo Lighting.