A high dynamic-range projector from Christie and Dolby promises something special in commercial cinemas, but how does it work?
Last December, I posted an item about Dolby's announcement that 2015 would see the inauguration of Dolby Cinema, a bold plan to take commercial cinema to the next level with Dolby Atmos immersive sound and Dolby Vision high dynamic-range (HDR) laser-illuminated projectors co-developed with Christie. One of the first public showings of this projection technology was presented last week at CinemaCon, the convention of the National Association of Theatre Owners.
Ever since I first heard about these HDR cinema projectors, I've been very curious about how they manage to achieve high dynamic range. I've asked several people from Christie and Dolby to explain it, but they have all declined. Fortunately, AVS member CinemaAndy pointed me to the patent application submitted by Christie for an HDR projection system
, which must be the basis of Dolby Vision in commercial cinemas. And since that information is in the public domain, I can share it with the AVS community.
The patent application starts with a description of current digital-projector technology, which is illustrated in the following diagram:
A conventional digital projector sends white light from a lamp (1) through a lens (2), which reflects from a mirror (7), travels through a transparent integrator rod (5), reflects from another mirror (7), passes through more lenses (6), reflects from another mirror (7), passes through yet more lenses (6), and ends up in the imaging engine (8), where it is split into red, green, and blue components. Each component illuminates a corresponding spatial modulator or "imager" that forms the image for that color in an array of pixels, after which the red, green, and blue light is recombined and projected through the main lens (9) to the screen. The imagers can be DLP DMDs (Digital Micromirror Devices), LCD panels, or LCoS panels. And the white-light lamp can be replaced with red, green, and blue lasers that illuminate the corresponding imagers directly or a hybrid design with blue lasers and a yellow phosphor wheel whose light is split into red and green.
According to the patent application, the dynamic range of a digital projection system is determined by the capabilities of the imager. In the case of 4K DMDs, the application claims the dynamic range is roughly 12 bits at 24 frames per second, and less at higher rates.
The architecture of the proposed HDR projector is much the same, with two critical differences—another spatial modulator and an added array of integrator rods:
In the proposed HDR architecture, the mirror in the upper left of the previous diagram is replaced with another spatial modulator (15) that is divided into an arbitrary number of zones. Each zone is controlled to send more or less light through an array of integrator rods (16), depending on the brightness of the final image in each zone. In the case of an RGB laser-illuminated projector, each laser would have its own integrator rod (5), zonal modulator (15), and array of integrator rods (16) as well as its own imaging modulator that creates the final image for that color.
For those who might not be familiar with an optical integrator rod, it's a transparent rod whose surface is internally reflective—that is, when light enters one end, it is reflected multiple times by the internal surface of the rod. This "homogenizes" the light, converting round or irregular patterns of illumination into a uniform, rectangular pattern. The cross-sectional shape of the integrator rod is typically the same aspect ratio as that of the imagers.
In this conceptual example, the zonal modulator (15) is divided into only four zones (20a-d); in practice, there would be many more zones. In the case of a DMD, the micromirrors in each zone are oriented in a pattern that reflects more or less light, depending on the brightness of that part of the final image. Virtually no light is reflected when all the mirrors in a zone are in their "off" position, while the maximum amount of light is reflected when all mirrors are in their "on" position (20b), and half that amount of light is reflected when the mirrors are in a checkerboard pattern (20d). A complete grayscale is generated by orienting the mirrors so more or less light is reflected from a given zone; for example, zone 20a in this diagram reflects less than half the maximum amount of light, and zone 20c reflects more than half the maximum light.
Interestingly, the mirrors in each zone are held statically in their positions during each entire frame of the video. Why not create a grayscale by alternating all the micromirrors in each zone between on and off many times per frame as an imaging DMD does, varying the percentage of time they spend in the on and off positions (a technique called pulse-width modulation)? Because it is exceedingly difficult to control the PWM frequency so that it's exactly identical for the zonal and imaging modulators, and any difference can result in visible artifacts. Instead, the mirrors in each zone are oriented in a spatial dithering pattern that reflects the desired amount of light as uniformly as possible.
Even so, the light from each zone must be highly uniform, so it is sent through another optical integrator rod. In fact, each zone has its own integrator rod and these rods are arranged in an array that corresponds to the array of zones.
In this diagram from the patent application, light from a laser (1) enters the first integration rod (5) and hits the zonal modulator (15), which reflects the light into an array of integrator rods (16; each rod is labeled 18 in this diagram). The light from the integrator-rod array passes through a single, hollow integrator (19) to blur the seams between the individual rods in the array. The light then hits the imaging modulator (13) at a non-right angle, causing keystone distortion, which can be compensated for with well-known optical techniques and/or image-processing algorithms.
This approach seems very similar to full-array LED backlighting with local dimming in LCD flat-panel TVs, in which the LEDs behind the LCD panel are divided into a number of zones that are dimmed and brightened according to the overall brightness of the image in each zone. This works well to increase the apparent contrast of the image, but it's not without problems of its own, such as halos around very small bright objects on a dark background within a single zone. Obviously, the more zones there are, the less haloing there will be, so I hope the Christie/Dolby Vision projectors have lots of zones.
The patent application does not specify the number of zones, nor does it specify exactly how much the dynamic range is increased over that of conventional digital-cinema projectors, other than to day say it's increased by several orders of magnitude. The basic technology achieves this by lowering the overall black level—which is annoyingly high in most commercial cinemas—and possibly increasing the peak brightness by cranking up the lasers.