August 15, 1997

DVD: breathtaking sight and sound, significant challenges

Maury Wright, Technical Editor

Long-awaited DVD technology stands poised and ready to enter our living rooms and our PCs and promises to converge the entertainment and computer environments. Designers simply have to figure out how to decode the compressed video and audio, mix DVD and graphics, handle security and encryption, and do it all cheaply.

Offering almost a tenfold boost in capacity compared with CD-ROM media, DVD-ROM media can easily store an entire digitally encoded and compressed movie or a more complex computer game than the entertainment world has yet conceived. It's precisely the potential of that mammoth ca-pacity that makes DVD a different animal from CD. To effectively integrate DVD into a system, you also have to decode MPEG-2 video and Dolby AC-3 audio. The pristine video quality has Hollywood worried about unauthorized duplication of titles, so you have to design in hefty decryption capabilities and get a license. In a computer-centric design, you must also mix these new-media streams with standard 2- and 3-D graphics. Surprisingly, however, IC vendors are already shipping DVD chip sets, and continuing µP performance gains will enable soft DVD not too far down the road.

The successor to CD technology, DVD has taken seemingly forever to arrive for those individuals who have followed its development. Bickering camps of consumer-electronic giants almost announced and shipped incompatible DVD-like products several years ago. Fortunately, IBM coerced the two camps into developing a unified standard and forgoing a VHS-vs-Betalike battle in the market. Even today, however, DVD proponents can't agree on what DVD stands for--digital video disc, digital versatile disc, or nothing. (See box, "DVD basics." For more detailed information on the history and current state of DVD technology, go to the Internet and find the DVD FAQ in the alt.video.dvd newsgroup. The FAQ is also mirrored at several Web sites, including www.videodiscovery.com. Toshiba's Web site also includes an informative DVD section.)

DVD-ROM media stores 4.7 Gbytes of data per side and will find extensive use in both the consumer-entertainment and computer industries. Vendors of consumer players have a clear DVD vision. The technology offers far superior video and audio quality than do VCRs and marginally better video and audio quality than do laser disc players. Compared with laser discs, DVD also offers greater capacity; lower cost; and more "trick" features, such as multiple camera angles and more audio tracks. DVD players arriving at consumer-electronic superstores have prices of $500 to $2000.

DVD use in the computer industry will evolve more slowly than it has in the consumer-entertainment industry. You can use DVDs to play movies on a PC, but doing so hardly seems like an everyday application unless your PC is in your living room. The added capacity that DVDs afford will allow educational titles, such as multimedia encyclopedias, to proffer more data and particularly more lengthy and compelling video and audio clips. Without doubt, over time DVD technology will also yield more compelling games that mix video with 3-D graphics.

You can buy DVD add-on kits for PCs today for around $500. The kits from companies such as Creative Labs (Milpitas, CA) include an ATA (AT-bus-attachment)-interface DVD-ROM drive (see box, "Evaluating DVD drives") and a PCI-bus board that handles MPEG-2 video and Dolby AC-3 audio decoding. Note that the DVD-ROM drive can read CD-ROMs and can thus simply replace the CD-ROM in a PC. In fact, the dropping price of DVD-ROM drives ensures that by the end of 1998, the CD-ROM will effectively disappear. High-end systems will include DVD-ROM drives as standard features by the end of this year.

DVD has broad appeal

The allure of DVD should have all system designers thinking about how to integrate the technology into their designs. DVD will certainly proliferate quickly in PCs and workstations. Moreover, embedded applications, such as flight or driving simulators, should find use for the technology, and DVD-RAM (DVD erase and record many times) can significantly extend the capacity and capabilities of data-acquisition systems.

Table 1 summarizes the chip sets that are available and that squarely target the DVD market. It's doubtful that you could discern any differences in audio or video quality should you experience a DVD demonstration played via reference designs using these chip sets. All of the offerings produce better video and audio than you are likely accustomed to. Ultimately, you must evaluate the offerings based on the MPEG-2 decoder architecture, security and encryption implementation, audio quality, video/graphics support, level of integration and road map to single-chip DVD, and system costs.

The MPEG-2 decoder block easily qualifies as the most complex part of the overall DVD subsystem, as the C-Cube Ziva architecture shows (Figure 1). You can't differentiate one MPEG-2 decoder from the next just by looking at picture quality. The MPEG-2 encoding that produces a DVD title ultimately limits the output quality achievable in the decoder. You still may want to consider, however, just how the vendors implement the decoder.

For example, Toshiba, SGS-Thomson, and LSI Logic offer hard-wired MPEG-2 decoders that ultimately can handle higher incoming bit rates without fear of ever dropping frames. On the other hand, all the other decoders that are essentially implemented in firmware for DSP or RISC processors should handle any DVD stream with no problem. The firmware approach could conceivably allow vendors to tweak their MPEG-2 decoders to optimize quality or to accommodate changes in the MPEG standard. Such capability proved valuable during the development of the MPEG-2 spec, but today the spec is fairly concrete.

Chromatic's Mpact media processor is a special case. When used strictly as a DVD player, an MPact1 IC has no problem handling the various decoding tasks. Chromatic developed Mpact, however, as a multipurpose processor that can handle 2-D graphics, modem, audio, and other functions in a PC. Should your design use an Mpact IC, you have to realize that the processor could stumble on DVD tasks if it were simultaneously executing the modem, 2-D graphics, or other functions. On the other hand, the board vendors currently shipping Mpact1-based DVD boards are dedicating the processor to the DVD task.

Chromatic's architecture also proves to be the most flexible of the bunch. The company licenses functions such as DVD or modems via firmware modules it calls Mediaware. IC-vendor partners Toshiba, SGS-Thomson, and LG Semicon manufacture and sell the Mpact ICs that Chromatic designs. Chromatic claims that its Mediaware approach allows it to tailor the back end of the MPEG-2 decoding process and optimize quality for specific video/graphics-accelerator ICs. For example, Chromatic offers DVD Mediaware with video postprocessing optimizations for ATI, S3, Brooktree (San Diego), and Tseng Labs (Newtown, PA) graphics accelerators.

Copy protecting MPEG-2 streams

Security and encryption provide other examples of features that either work or don't work, yet there's more to consider before you can sell a DVD system. Every vendor of DVD ICs, boards, or systems must get a CSS (content- scrambling system) license before selling its product. Panasonic/Matsushita currently administers CSS licensing, but this licensing will soon fall under the umbrella of the DVD-Forum that's under formation. The CSS process ensures that designs meet security guidelines that have been hammered out with the major Hollywood movie studios. The studios fear that the MPEG-2 digital-video technology might allow hackers to make perfect copies of movies. The CSS process essentially ensures that a decrypted but still compressed MPEG-2 data stream can't be captured.

All of the ICs in Table 1 have CSS licenses, yet you still must get an additional board- or system-level license to sell a product based on the ICs. Moreover, should you sell a board- or system-level product that a hacker subsequently cracks, you must take responsibility for fixing the security leak. Presumably, the IC vendor will also engage in solving the problem, but it's a problem that most companies want to avoid. Therefore, you should closely evaluate security and decryption implementations.

Of the vendors in Table 1, SGS-Thomson and C-Cube make the strongest argument for hacker-proof decryption schemes. Both companies tightly integrate decryption and MPEG-2 functions on the same IC and never expose the decrypted data stream outside the ICs. Toshiba uses a hardwired decryption IC, but it's currently a stand-alone chip that feeds a separate MPEG-2 decoder, thus potentially exposing the decrypted data stream on circuit-board traces. Most of the other vendors use firmware to handle the decryption, and that approach requires that data is written to a local DRAM buffer. IBM goes a step further and handles decryption on the host processor, thereby placing sensitive data in system DRAM.

All the approaches certainly avoid placing decrypted data streams on an easily accessible interface, such as the PCI bus. Moreover, Chromatic, IBM, Toshiba, and ESS Technology have all taken additional proprietary-encryption precautions for all data written into memory. The companies don't say much about how their schemes work, but you as a potential customer should be able to get enough details to evaluate the schemes.

Judging audio quality

The companies selling DVD ICs also take significantly different approaches to decoding the Dolby AC-3 5.1-channel audio stream. Here again, no one but a devout audiophile likely hears the difference, but you should scrutinize the implementations for other reasons as well. IBM, for example, again turns to the Pentium host µP to handle the AC-3 audio decoding. The company claims that its AC-3 algorithm requires 20 to 25% of the MIPS available in a 200- to 233-MHz Pentium µP. Presumably, the host-based audio saves silicon costs, and the drain on the CPU is unimportant for systems playing a DVD movie.

The IBM approach would certainly have generated substantial savings in the days when DVD boards used a discrete AC-3 decoder IC that might cost $10 to $15. Today, however, several companies integrate the AC-3 function into an almost-single-chip DVD player, so some of the cost advantage has disappeared.

The approaches clearly generate theoretical quality differences. In fact, Dolby Laboratories (San Francisco), which certifies and licenses all AC-3 decoders, has developed Class A-, B-, and C-level certifications, with Class A being the highest quality. Dolby has tried not to publicize the existence of these classes and, in fact, would prefer not to have such levels of quality. Many industry insiders believe that Intel used its considerable influence to have the classes established so that host-based decoders could be AC-3-certified.

Host-based algorithms, such as IBM's, are limited to 16-bit sampling of the audio stream. Going to 32 bits requires too much memory and CPU overhead, and even moving to 20 bits requires a 32-bit implementation. Generally, 16-bit algorithms can achieve only Class C certification. Conversely, SGS-Thomson's hard-wired ap-proach yields a Class A certification. Most other vendors are reluctant to reveal details of their Dolby certification because Dolby discourages such disclosure. Class A certification requires 20-bit or greater sampling, and the best quality implementations will likely use 24-bit sampling.

In a computer-centric DVD design, you must also mix the decoded video and audio streams into graphics and audio subsystems. Almost all graphics-accelerator ICs can accept video data, arbitrarily scale the size of the video, and play the video in a window or full screen. Still, no industry standard provides a way to route video to the graphics IC. In a PC environment, you have three choices for how to route the video. You could transfer it via PCI, but, because the MPEG stream is already decoded, it would likely saturate the bus. You could take the VGA output and loop it into the DVD decoder card. You could then control the video/ graphics multiplexing, but you would need extra circuitry, such as video scalers, that already resides on graphics accelerators. Sigma Designs (Fremont, CA) uses this approach on its REALmagic Hollywood DVD card, but Sigma has a distinct advantage in that it specializes in video-reproduction ICs.

Most designers opt to use one of the digital-video interfaces developed by the graphics-IC companies. There's an alphabet soup of such proprietary interfaces that all sprang from the work of the VESA (Video Electronics Standards Association) and its VMC (VESA Media Channel) interface. To the DVD designer, the lack of standards means supporting multiple video interfaces and choosing which graphics accelerators a DVD design will support.

In its DVD chip set, for example, Toshiba supports ATI, S3 (Santa Clara, CA), and Cirrus Logic (Fremont, CA) interfaces. These interfaces are generally the most popular because those companies lead the graphics-IC market. SGS-Thomson has made an effort to make its VIP interface an industry standard, but the jury is still out on whether more companies will use the interface. SGS-Thomson has a potential winner with its 128-bit Riva graphics accelerator that it codeveloped with Nvidia (Sunnyvale, CA) and just began shipping. Success of that chip could help drive the VIP effort. On the other hand, the reliance on VIP in its DVD ICs could limit SGS-Thomson's market in the short term.

Integration and costs

It's clear that DVD technology has arrived, but it's equally clear that prices must drop before consumers widely adopt the technology. DVD chip sets cost $25 to $70 in the multithousand-unit quantities typical of the PC market. To sell products by the end of the year, chip-set vendors have to be at the low end of that range. Increasing the level of integration always proves to be the best way to lower prices, and single-chip implementations are arriving.

Keep a few things in mind when you evaluate the cost of DVD chip sets: The IC cost and the system cost can be quite different. For example, the SGS-Thomson Sti3560 includes a PCI interface, but other "single-chip" DVD implementations require an external PCI-bus interface. Make sure you evaluate everything required, including DACs, memory, and external microcontrollers.

Ultimately, DVD must integrate directly onto the graphics board for the technology to proliferate in the computer market. Expect some companies to have combo DVD and graphics boards this year. ATI, for example, is working on a design based on IBM's MPEG-2 decoder IC.

When you think of single-board implementations, some factors may change relative to how you evaluate the various chip sets. For example, Chromatic and its partners offer the new Mpact2 IC, which includes a hardware-based 3-D accelerator and easily handles DVD along with graphics, telephony, and audio duties. Toshiba will ship samples of the IC for $60 this month with production scheduled for October. SGS-Thomson will also have the Mpact2, and can alternatively combine its DVD chips with its Riva accelerator. You may find that companies holding multiple technologies have an advantage in providing single-board implementations.

The lowest cost implementations will likely be those that rely the most on the host CPU. IBM is clearly heading in that direction, but down the road the host will take on some or all of the MPEG-2 decoder's chores. In fact, between the end of this year and mid-1999, increases in µP performance and enhancements to graphics chips will obsolete DVD ICs in the PC market.

In the short term, even the Pentium II µP can't single-handedly implement a DVD player. Already, however, ATI has added features to its graphics accelerator that enable host-based DVD. Specifically, the 3D Rage Pro that just began shipping includes hardware support for motion compensation, which is the final stage of MPEG-2 decoding. With the assistance of ATI's new IC, Zoran has demonstrated its SoftDVD player running on a 233-MHz Pentium II µP. Moreover, ATI plans to support the SoftDVD technology and offer hardware-based designs.

Multipurpose devices

Some entrepreneurs will also find opportunities to integrate divergent but symbiotic technologies into one product. For example, ESS Technology targets its new ES3308 primarily at the consumer-player market but developed it in such a way that the same IC can handle set-top-box duties for digital-satellite and terrestrial-cable transmissions. You could use the IC to create a PC-based product that handles DVD and multiple set-top-box functions.

More realistically in the short term, expect a market for products similar to Gateway's (Sioux City, SD) Destination Entertainment PC. The bundled DVD technology and dual-purpose TV/monitor somewhat seamlessly merge computing and entertainment devices.

The building blocks for such products are readily available. For example, ATI's $299 All-In-Wonder card can drive a TV as well as a computer monitor. The card also includes a TV tuner, and you can use it to capture video clips from a VCR or camcorder. Interlink Electronics is shipping its VersaPoint wireless keyboard with an integrated touchpad. Simply plug an infrared receiver into mouse and keyboard ports, and you can roam the living room with your keyboard or Web-surf from your easy chair. The keyboard sells for around $100 at discount prices and for even less to OEMs.

Still, you have to address other issues in designing living-room equipment. For example, the fan in a PC actually makes a discernible noise that's bothersome while watching TV--albeit it's no problem when watching Batman in Surround Sound. PCs can also emit RF energy that can cause interference with standard analog audio and video signals. Despite these obstacles, it's difficult to imagine a modern living-room entertainment rack without a PC two to three years hence.

| EDN Access | Feedback | Table of Contents |