There was once a time when 4K resolution, or at least the claim of 4K, was special. Sony was possibly the first to produce a projector, in their SXRD range, while Dalsa‘s 4K Origin camera was probably the first just-about-fieldable camera that significantly exceeded the resolution of HD video. With the F65, Sony finally has a camera to match their projector, but it’s no longer a sparsely-populated market. JVC recently showed both its 4K camera and projector products at a demo at the Soho Theatre in London. The combination has a lot of resolution and the projectors are frankly rather nice, but they’re an odd coupling, at opposite ends of the price spectrum.
While the F65 was announced some time ago, JVC’s first releases of information regarding a 4K camera probably predate the Sony offering. JVC’s GY-HMQ10, is of course a much smaller, lower-end and cheaper device, even given the competitive pricing Sony have mooted for F65. Much as it’s very likely to be the cheapest 4K camera in the world on release, though, this does create some challenges for JVC. Lens quality is a concern given that mere HD cameras, at a quarter of the output pixel count, often struggle to land enough resolution on their sensors with consumer-priced glass. There are issues of storage, and of processing power given that JVC’s camera is a single-chip Bayer device like more or less all of the other 2K-plus cameras that exist.
None of this is really news, though, so let’s think for a moment about what 4K is actually for. It’s a number cited a lot by people interested in matching or exceeding 35mm film, but it’s a sad fact that neither the photochemical nor digital-intermediate processes were ever capable of putting more than about 2K on screen, after all the duplication. The very best, slowest, 35mm negative has been measured at around 6K in laboratory conditions, but whatever you think of digital cinematography, it is unavoidable that D-cinema systems lose far less in the journey from the camera to the screen. As such, digital cinema probably only strictly needs 2K acquisition to equal what film has ever done.
With regard to glass, a lot of people are shooting on the relatively low-cost 4K cameras that are now available, but using lenses decades old. Much of this glass isn’t as sharp as the sensors with which it is paired, and while in some cases that is something that’s desired and looked for, it does question the idea that ever more pixel count is always useful. There’s also the argument as to whether a single Bayer sensor that’s 4000 pixels wide is really a 4K imager at all. It certainly doesn’t have the resolution of a device that would have been called a 4K film scanner or a 2K high-definition video camera with three image sensors and a color splitter block. Any single-chip camera has an RGB output that’s 67% interpolated guesswork, unless, like the F65 or Canon’s C300, the sensor oversamples the image by something like twice.
Even if you work around all this, though – if you shoot F65 with the best lenses, project it with Sony’s SXRD projector, and view it from one screen height away – how short-sighted do you have to be before it all becomes pointless? Not very. The average small-town projectionist has historically struggled to get the most out of a 35mm print, and I also fear the viewing public, who often can’t tell the difference between the ads and trailers that are in 3D and those that aren’t, will not be willing to pay more for 4K. Certainly not on the basis that it’s visibly any better than the 2K DCPs they’re watching at the moment, although I think that the situation with 3D proves that a proportion of people can be programmed to want something regardless of its actual merit. 4K as an origination format may make more sense, inasmuch as there should no longer be a concern about acquired resolution and the flexibility in post is a plus. Anyway, in the long term, until F65 there wasn’t really a 4K camera out there that you couldn’t argue about on some level, and nobody outside a dailies screening room, on a feature shooting very slow 35mm stock on excellent lenses, had ever actually seen a 4K image.
JVC’s GY-HMQ10 is another “4K” device that has a single 4K image sensor, and it is a consumer-targeted camera with a fixed lens that produces visible chromatic aberration that’s many pixels wide at the corners of the 4K output. It’s also a ½ inch sensor, tiny for one that has so many photosites on it, so other characteristics of the image are compromised – dynamic range is far from outstanding. All of this was made extremely obvious at the demo simply because the output of this poor little camera was being blown up by an excellent projector to an image twelve feet wide which I could stand a foot from.
All of this has been true of small prosumer cameras for ages, though. None of it stopped them from being used for higher-end work than perhaps would have been expected. Whether that really applies to the new JVC camera, where the resolution that makes it special is compromised by lens problems, remains to be seen.
There are interesting things being done, though. JVC’s camera deals with 4K as quad-HD, 3840 by 2160, for the very simple reason that the camera records the image – or will when released – on four SD cards in parallel, each using the 35Mbps codec that’s an off-the-shelf component for JVC. They intend to ship software to stitch these four streams back together to produce a single file for postproduction. This slightly Rube-Goldberg approach is presumably a side-effect of the desire to get the thing shipping quickly, but it does have the interesting implication that the 4K output has an aggregate bitrate of 140Mbps. This is a lot for HD if not 4K, and scaling the result to HD should make for a pleasantly artifact-free image. It might even mitigate noise sufficiently to allow more cautious exposure, making that tiny, clippy sensor look much better. These measures might take the thing from being a fairly mediocre 4K camera to being a pretty reasonable HD one, for its size and cost.
In the prototype I saw, there were four HDMI outputs, each transporting one quarter of the image to the DLA-RS4000U projector, continuing the quad-HD approach of the storage subsystem (although the projector offers a slightly more generous 4096×2400 resolution). Much as the camera is clearly targeted at the lower end of the market, the projector is a much higher-end piece of gear. Some 2K projectors use the fantastically named “wobulation” technique to achieve a slightly sharper image with better fill factor, but this thing is the real deal. There are three of JVC’s large 30mm D-ILA chips – the component which actually imposes the picture on the beam of light – and a proper 4K RGB output. The image is bright as a result of the Xenon lamp – wattage not given, although the entire device is rated at 825W – and the large effective F-number given by those big chips.
D-ILA, along with Sony’s SXRD and others, is a term for a liquid-crystal-on-silicon device, effectively a modulated mirror that places the LCD element on a polished metal surface as a reflective rather than transmissive controller of light. As such the liquid crystal device is bound directly to a comparatively large lump of metal that is easy to heatsink. LCoS can therefore tolerate an impingement of light that would cause a transmissive LCD to burst into flames, and can therefore use higher-density polarizing layers and produce higher-contrast images. The performance is not quite as good as DLP, but it is cheaper and can be easier to manufacture at high resolution. A few years ago JVC’s DLA-HD2K was the first consumer display that I saw successfully evaluated by Filmlight’s extremely fastidious Truelight calibration system, which would completely refuse to calibrate conventional TFT devices on the basis that the black level was too poor. It is not necessarily unreasonable for JVC to promote its D-ILA range as grading reference displays.
The demo I’ve seen of the LCoS projector and quad-HD camera combination are unscientific, and it is difficult to divorce my concerns about dynamic range from the knowledge that Japanese manufacturers actually seem to like their cameras like this. The Canon DSLRs, as both still and moving image devices, suffer a default colorimetry setup that tends to clip highlights harshly. The prototype camera that JVC is currently showing is far from complete, and I have no idea what, if any, control will be available over its contrast and brightness curves, or what may be intrinsic characteristics of the sensor. Other ambiguities are created by the combination of a low-end camera and a high-end projector, so it’s difficult to be unequivocal about the performance of either. What’s certainly true is that from F65 at the top end of the market to JVC at, well, not the top end, getting something that you can at least claim is a 4K camera is about to get a lot cheaper.
I’m just not sure whether that’s really necessary.