Digital Server Technology and the Digital Cinema Revolution

Dr. James Clark
Grass Valley Group

While film will not disappear as a medium anytime soon, the technology to support digital cinema is rapidly falling into place—for the first time since the birth of the movie industry, an alternative path for movie distribution and playout is emerging. Once the studios and technology groups have agreed upon a standard for encryption of the transport datastreams, digital cinema systems will emerge—systems that will provide consistently better picture quality, reduce overall distribution costs, and enable content creators and exhibitors to build more lucrative business models.

The drive toward fully capable digital cinema systems is accelerating quickly, fueled by the continued expansion of disk drive capability, the increased sophistication of compression algorithms, and improvements to other, key digital technologies. One of those technologies is the video server.

The video server has been the cornerstone of worldwide broadcast industry's transition to providing high-quality, digital video content instead of traditional, analog-based video content. A leader in the broadcast industry and in defining digital compression standards such as MPEG2, the Grass Valley Group defined the video server product category for the broadcast market more than five years ago—and worked diligently to help create the MPEG2@HL standard to ensure that the video server output was of a high enough quality to meet the demands of broadcasters and their viewers.

As the market share and technology leader for video servers worldwide, the Grass Valley Group is using its Profile^® XP Media Platform to define new opportunities for content providers, from Internet streaming to Interactive TV. And it is using this same core digital video technology to hasten the development of digital cinema.

Today's digital technology such as that found in video servers provided by the Grass Valley Group, among others, can easily support the compression and image quality requirements of digital cinema. Modern disk technologies have the capacities and throughput capabilities to store several multi-hour movies on one unit. With the maturation of these and other technologies, the emergence of digital-cinema systems that offer the picture quality, lower overall cost of distribution, and provide income-generating opportunities necessary to make them attractive to film studios and theatre chains is all but certain.

Clearly, picture quality must come first. Basic quality issues, such as adequate resolution, bit rates, adequate frame rate, acceptable minimum contrast ratio, and acceptable color palette range, must be defined. Achieving these standards, however, will be more a matter of evolution than revolution.

Resolution

The present MPEG definition of 1920x1080 resolution is a very acceptable resolution for the initial rollout of digital cinema. This image standard defines 1920 pixels per horizontal line with 1080 horizontal lines per frame (a pixel is defined as a single dot on a screen representing any color or intensity in the color palette capability).

An image with the kind of line resolution that the 1920x1080 standard provides approaches the photographic quality of 35mm film. On the other hand, while a top-quality 35mm film print (e.g., an answer print) can actually resolve more lines than 1920x1080, the limitations of mechanical film transport seriously mitigate its effective resolution. Horizontal and vertical jitter of the film frame in the projector gate limits film's resolution to well below that of which it is fundamentally capable. Variations in color timing during processing, variations in film chemistry along its length, and aging all contribute to visible "flutter" in film brightness. Lastly, accumulated scratches and debris compromise the stability and quality of the image of the average film print shown in a movie theatre.

Digital image projection has none of these limitations.

Due to the electronic scanning nature of digital-image projectors, there is absolutely no jitter or flicker from mechanical movement—the projectors simply have no such mechanisms. Similarly, there is no variation in color accuracy over time because digitally stored images are not subject to chemical dye variances or media aging. In fact, once a person observes the high-quality digital image projection of a movie, the jitter, flutter, and noise experienced in viewing a 35mm film become distracting.

Keeping in mind that most of the demonstrations of digital cinema to date have been at a resolution of 1280x1024, some would say even this lower level is good enough to begin digital cinema implementation in some venues. This is reasonable for early implementations in that the advantages listed above largely outweigh the limitations of the lower resolution, which for the most part amounts to rough edges in subtitles and "pixelization" which can be visible in the first few rows of an average movie theatre.

Eventually, digital cinema projection will evolve beyond demonstration-quality images to resolutions as high as 4096x2048, or IMAX-class quality. While the average human eye cannot resolve more than 1500 horizontal lines in an image, the emergence of screen resolutions that are higher than the human limit for a single frame will allow a person to shift his or her attention to different sections of a screen without detecting artifacts or blurred images.

Bit Rates

When some people hear the term digital video, the first thing that comes to mind is digital TV and/or the visual artifacts of high-definition (HD) broadcasting. Just as it is important not to confuse MPEG technology with HD quality, so too is it important to distinguish digital video technology from mere HD broadcast quality, which is simply not good enough for theatre-level presentations.

HD images have a resolution of 1920x1080. The various artifacts of HD transmissions are a direct result of the 19.4 Megabits per second (Mbps) bandwidth limitation imposed by terrestrial broadcasting. There is no compression technology available now—or that will be available in the near future—that can compress a 35mm-quality image with acceptable quality at this bit rate. At best, transmitting such images at this kind of low bit rate results in fuzzy images and "patchwork quilt" artifacts.

The 1920x1080 images of HD transmissions are also compromised by their interlaced frame structure, which causes many visual artifacts. These artifacts look like line-fringe interference patterns, flicker, and beat-noting. In fact, this structure has its roots in bandwidth-limited environments, as interlaced frames have long been used to minimize flicker on cathode ray tube displays due to the low display frame rates of 30 frames per second.

Lastly, the bit rate limitations of an HD transmission restrict the color accuracy of the images it carries, both in terms of contrast and color saturation. This restriction is due to the limited number of bits that can be used to encode the primary colors of an image.

These same limitations need not apply to digital cinema. Video server data rates can be high enough to support contrast, color, and resolution significantly better than the HD standard. Digital cinema servers exist today that are easily capable of 100 Mbps decoding of compressed images. Modern disk technology can support this data rate for multi-hour movie lengths.

Frame Rate

Now that our workday world is filled with computer screens, we have heard and sometimes experienced (consciously or subconsciously) the term 'flicker' associated with frame rate. In this context, frame rate applies to the rate at which the CRT phosphor is 'refreshed' or redrawn. This is purely a display technology issue and should not be confused with the information frame rate, which applies to the human perception of motion in a movie image. Film projectors run at an information frame rate of 24 frames per second (fps) and a display frame rate of 48 fps due to the double flashing of the shutter.

The necessary frame rate that a digital server should provide must be based on the information frame rate, and that rate will be determined by the projection technology used. For instance, the Texas Instruments Digital (DLP) technology inherently does not require any screen refresh to sustain an image and therefore presents an incoming image at the 24 fps rate without any flicker artifacts whatsoever. Other projector technologies can receive an information frame rate of 24 fps and digitally triple the refresh rate of the display to 72 fps, which is beyond the flicker perception of most people.

Contrast Ratio

Contrast ratio and color saturation are important considerations that also have a direct impact on the data bandwidth required to sustain image quality. There are two main physical parameters that determine this bandwidth impact: bits per pixel and color sampling.

The number of bits per pixel directly impacts the bit-rate requirement. Other than gamma compensation, there is no way to compress this data, unlike scene and motion compression. A contrast ratio of 1000:1, which is generally considered the minimum for digital cinema, requires a 10-bit per pixel data rate to the image device. A contrast ratio of 2000:1 would increase the bit depth to 11-bits per pixel. Some experts in the industry feel the contrast ratio should be higher than 2000:1. Keeping in mind that these extra bits are added to each pixel in a frame, it is apparent that the impact is significant. Video server technology exists today that can meet—or can be enhanced to meet—the contrast ratio needs of digital cinema.

Color Palette

Color sampling is a less familiar term to those in the movie industry because it has no counterpart in film. The HD MPEG2 video standard does not sample each of the RGB primary colors in every pixel. This was deemed necessary to meet the 19.4 Mbps bandwidth limitations of terrestrial broadcasting. It is generally agreed, however, that in the world of 47-foot movie screens this is not an acceptable compromise. Video server technology exists today that can be enhanced to meet full RGB dynamic range for each pixel.

What Does It Take To Support Digital Cinema Quality Images?

Considering all the elements discussed above, it is reasonable to set an initial target for digital cinema rollout of 1920x1080 at 24 fps (progressive) with full color sampling.

At first glance, it would seem like a 100 Mbps data rate of today's digital video servers would provide five times the image quality of HD broadcasting. But the relationship is not one-to-one. Tests over the years have shown that image artifact quality improves sharply between 20 Mbps and 30 Mbps. As the bit rate increases to 40 to 60 Mbps, the color accuracy and depth become competitive with the best quality print of a film.

Presently available compression technologies can encode a 1920x1080 image at 24 fps with a 64 Mbps average image data rate. But keep in mind that unlike DVDs (12 Mbps) and HD broadcasts (19 Mbps), there is no need to fix the data rate of an encoded image. Modern decoding technology allows the decoders to track the incoming data rate of encoded streams. The concept of a standard film speed or incoming broadcast bit rate does not apply to digital cinema.

Disk drive technology today is capable of supplying sustained data rates of 104 Mbps from a single disk drive. The capacity of a state-of-the-art single disk drive can support an 80 Mbps datastream for 2.0 hours. Clearly it is reasonable to expect that digital cinema can rollout with 64 Mbps as an average data rate and that 1920x1080 resolution at the standard 24 fps is a perfectly reasonable goal.

The role of a digital server in the modern theatre will be to absorb input files from various sources and decode these files into real-time movie images. But note that the input data rate does not have to be real time. Obviously, this is the case for files delivered as physical media. Less obviously, it is not necessary for satellite transmission or telecom trunk feeds. For a satellite feed bandwidth of 45 Mbps, the ratio is only 1.4x real time to ingest an image encoded at a data rate of 64 Mbps. Keeping in mind that a modern server can automatically accept the incoming feed unattended, this feed can come at a time of day when data rate usage is discounted. The same is true for telecom lines.

So the digital server for a modern theatre must be able to accept incoming datastreams automatically from a variety of sources, do so unattended, and with high levels of reliability.

What Kind Of Equipment Will It Take To Support These Data Rates?

In order to support a theatre complex reliably, redundant disk storage will be necessary—especially as modern disk drive technology pushes disks closer and closer to the physical limits of magnetic media. Optical disk technologies do not yet offer any advantage over magnetic technology for high-bandwidth server applications.

The two technologies that provide redundant storage are mirrored disk systems and RAID (Redundant Array of Independent Disks) disk systems. Mirrored disk drive technology is pretty much what the name implies: an alternate set of disks contains a duplicate image so that there is always an instantaneous backup file available. Modern RAID III technology is more complicated but is also more efficient in its use of disk space. The RAID III controller in current Grass Valley Group systems, for example, uses one error-correcting disk for every four data drives. Thus disk overhead for redundancy protection is only 25 percent instead of 100 percent.

The complex systems that provide this level of redundancy won't be practical in modern theatres unless they can be managed day to day by unskilled staff. The RAID III system is a good candidate: not only can it self-diagnose and repair errors and failures, a failed disk unit can be replaced while the system is operating. This process is simply a matter of pulling out the drive unit from the front panel with the handle built in to each drive.

In the Grass Valley Group systems, diagnostic monitoring occurs automatically over regular telephone channels. Replacement disk units can be ordered and shipped automatically from a central location. This is, of course, also true of any other part of the system. Depending on just how critical a customer considers downtime failures, Grass Valley Group systems can be delivered with other redundant components such as power supplies and operating system disk drives for quick replacement on site.

Delivery Of A Film To The Theatre

For decades, the theatre industry has been using physical media (film) for movie delivery. So most initial digital cinema systems will probably continue the use of physical media for distribution. This is a business model everybody is used to. This is a security model everybody is used to. This is a procedural model that everybody is used to.

Even at the outset, the distribution of digital films on physical media will have the advantage of not requiring any skill to load a movie and no attendance required beyond the simple process of inserting disks or cartridges into a digital-cinema systems and pushing the 'load' button. Gone is the time and skill necessary to splice together and supervise the loading of multiple reels of film.

With current technologies, a typical movie can be delivered on eight double-sided DVD-ROM disks. Weighing just several ounces, this package set represents a very significant savings in shipping costs compared with several reels of film—and the use of DVDs drops manufacturing cost from thousands of dollars to under one hundred. Today's commercial technology can provide real-time loading with multiple drives. For DVD input, there already exist multiple-drive systems that can handle all eight disks automatically.

Looking toward the future, once the industry has worked out the business relationships, security issues, and physical infrastructure for satellite and telecom interconnections, wide area network technologies will eventually dominate digital-cinema distribution. This sophisticated approach will allow completely automated delivery of digital films to theatres.

What About Trailers, Advertisements, Etc.?

Using digital technologies, the process of bringing into theatres movies, trailers, sports events, and other materials becomes a simple matter of merging and deleting files. It is important that a digital cinema server be sophisticated enough to offer an easy-to-use interface for this process. Using a familiar interface, such as that of the Microsoft^® Windows^® operating system, would be preferable. The Grass Valley Group digital video server does offer such an interface.

Merging files together, as in splicing films together, must be frame-accurate. So a digital cinema server must be capable of merging clips together seamlessly. Both Wavelet and MPEG data files can be handled in this way. The Grass Valley Group already has this capability standard in its MPEG servers.

What About Security?

Due to the very high quality that digital cinema images carry, and the ability to make perfect copies of digital files, it is clear that the various files and datastreams will have to be encoded in all exposed spaces. The most obvious and dangerous exposure is during delivery from studio-approved reproduction facilities to individual theatres. This is true for physical media such as DVD-ROM disks and even more critical for satellite and telecom links where the exposure to sophisticated snooping is undetectable. The use of physical media offers at least a first layer of security.

The current delivery standard is the Digital Transmission Content Protection (DTCP) system jointed developed by five major companies, including Sony and Intel. The studios have given approval to this system for D-VHS delivery of HD quality files, although the management of encryption keys has not yet been agreed upon. This system makes the most sense for the initial rollout of digital cinema due to the significant effort that the electronics industry has already invested in both software and silicon hardware development.

While stored in the digital server(s) within the theatre, a film's files remain encrypted. They are decrypted in real-time as they are fed to the decompression engine.

The link between the digital server's decompression engine and the projector interface represents another space in which the signal could be copied. However, as with physical media, there is already a first level of security that can be used to control this exposure. Thus, the encryption technology at this interface point does not have to be as sophisticated as for the delivery path to a theatre.

Again, a viable candidate for this path during the initial rollout of digital cinema is already in development by Intel. It is called Digital Visual Interface Content Protection (DVICP). There is less agreement within the industry about the standard for this interface. Some in the industry believe that it may be necessary to keep the datastream both compressed and encrypted until the signal has passed into a projector's electronics. This is a complex issue that will take a year or two to work into any kind of accepted standard.

Will Digital Servers Be Economically Viable Anytime Soon?

The savings that will be realized in digital film distribution are readily apparent and the studios can see an economic motivation to begin the transition when the equipment becomes cost effective. Less obvious, however, is that there is also considerable economic benefit for exhibitors. The ability to reallocate screens without the time delays and additional costs of print production and delivery will allow theatres to respond rapidly to a film's popularity. Because distribution costs are significantly lower, there will be more opportunities to bring in additional content, such as foreign films, documentaries, etc.

Although this paper has been careful to point out that HDTV should not be used in place of 35mm film, it is important to note that HDTV live event coverage is completely compatible with digital-cinema projection. Keep in mind, however, that this compatibility will only exist if the compression standard used is MPEG-based.

As we enter the new millennium, prototype digital servers and projectors are becoming available to begin testing various parts of the transition to digital cinema. Over the next year, we will see the arrival of economically viable hardware to begin the transition to a digital cinema system in the United States and around the world. Digital servers will likely be able to meet a target price of $20,000 per screen, dropping to about $15,000 as volume builds. Implementations that do not put a premium on reliability will be available at lower prices. This is especially true with MPEG-based servers, where the cost of the disk storage dominates manufacturing costs.

The technology to support digital cinema is rapidly falling into place. Once the studios and technology groups have agreed upon a standard for encryption of the transport datastreams, the worldwide adoption of digital cinema systems will accelerate, providing consistently better picture quality, lowering overall costs of distribution, and providing new ways for theatres to improve margins and generate additional income.