A Question of Bits Page 3
Two of the five proposals for high-definition TV currently being tested by the FCC employ AC-2 coding for stereo sound. At the NAB convention, Tom Holman of Lucasfilm went a step farther. He reviewed the psychoacoustic factors that affect the perception of timbre and imaging, concluding that the optimum number of audio channels is 5.1: five full-bandwidth channels (left, center, right, left-surround, right-surround), plus a narrow-band 100Hz sub-bass channel. A SMPTE Digital Sound on Film committee reached a similar conclusion, as did a recent study in Europe. The emerging consensus is called, for simplicity, the 5.1-channel concept. With fewer than five full-range channels, imaging accuracy and spatial realism are impaired. Increasing the number of channels beyond five raises both cost and complexity while providing only slight, often negligible, audible benefits. [Holman has more recently decided that 7.1 or even 10.1 will offer audible benfits.Ed.]
The 5.1 concept is embodied in Dolby's SR-D digital system for movie soundtracks. Digital codes for the 5.1 channels are recorded on the film between the sprocket holes. In December 1991 three US theaters (in San Francisco, LA, and NY) used SR-D playback during the entire run of Star Trek VI, with no public announcement. SR-D provides greater dynamic range than analog optical soundtracks can, with better clarity in loud or complex passages. Compared to conventional Dolby Surround, SR-D's discrete left/right surrounds provide better screen/surround separation and clearer localization of off-screen sound effects. After 200 plays of each film print the digital soundtrack was still playing fine, despite visible signs of wear in the picture. During the spring SR-D equipment was installed in ten theaters in major US cities for this summer's showings of Batman Returns. In the months to come, SR-D will be used with other films and in a larger number of theaters.
SR-D is based on another Dolby digital audio compression system called AC-3. AC-3 employs the same principles as the AC-2 family; but while AC-2 processes audio channels individually, AC-3 was designed to process audio channels in groups. That makes it ideal for surround-sound where, as with two-channel stereo, there usually is a high degree of correlation between the channels. (On the rare occasions where it is not, as when a discrete off-screen sound is located in a surround channel, the remaining channels usually are relatively quiet during that moment.) Taking advantage of these channel relationships, AC-3 achieves amazing coding efficiency, squeezing 5.1 channels of surround-sound into a total bit-rate of only 320 kilobits/s—an average of only 64 kilobits per second per channel—while providing audio quality similar to what AC-2 achieves with 192kb/s per individual channel.
Dolby has been showing a demonstration version of the AC-3 system to manufacturers in Europe and Japan, hoping to encourage its adoption in all new digital media. These include HDTV broadcasts, HDTV VCRs and laserdiscs, digital radio, and multimedia formats such as CD-I. In most of these, adoption of a standard now would be premature. But Dolby is encouraging developers to make sure, wherever practical, to leave 320kb/s of capacity in any new digital medium, space that can be allocated to AC-3 audio coding when the time comes. An inexpensive one-chip AC-3 decoder will be available to manufacturers next year. It is a reasonable bet that within a few years digital VCRs will contain AC-3 decoders delivering the same discrete-surround sound in the home video theater that SR-D provides on film.
Selectable dynamic range is one of the possible enhancements that could be included in future versions of the AC-3 system. Listeners who want to watch Terminator 2 or listen to a Mahler symphony at midnight without waking the neighbors would not need to impose brute-force compression on the analog audio signal. Instead, recording engineers could include codes in the bitstream that would tame playback dynamics on request while preserving the producer's artistic intent.
In principle there is room enough in the DCC and MiniDisc systems to accommodate AC-3 coding, providing 5.1 channels of sound instead of 2. But in practical terms it is too late to change those formats; their coding systems have already been standardized. Such a change might be pointless anyway; 5.1-channel sound is far more practical at home than in portable playback.
Historically, a new audio medium emerges every ten years or so, offering either a new level of sound quality or convenience or both. Electrical recording arrived in 1926, magnetic tape recording in 1936 (in Germany, delayed elsewhere by World War II), the vinyl LP in 1948, the stereo LP in 1958, the Dolby-B cassette in 1970, the CD in 1982, digitally compressed DCC and MD in 1992. The next major audio format will be due around the millennium. By then a large population of listeners will be accustomed to the realism that surround-sound can provide for a broad range of music formats and TV programming as well as movies. Since any two-channel stereo medium can carry analog Dolby Surround encoding, the next few years will bring greatly increased use of surround encoding in CDs, FM, tapes, digital radio, and multimedia.
So when the next new medium replaces the CD ten years hence, it is likely to be a multichannel format using something like AC-3 coding to combine the compact size and convenience of the MiniDisc with 5.1 channels of wide-range discrete surround-sound. That, at long last, will be the first true "high-fidelity" audio format, bringing to home listeners a genuinely lifelike you-are-there impression. Especially with large-scale music, a five- or six-channel surround-sound system provides a far more lifelike impression than any high-end two-channel system you ever dreamed of.
The next time I run into Jack Renner of Telarc I will urge him to stop mixing his recordings on-site to a two-channel recorder. Telarc's miking, with left/center/right mikes across the stage plus two hall mikes in the rear, is a natural for AC-3 coding and 5-channel home playback. Telarc (and others) should begin right now to record those channels discretely on a multitrack machine, so that today's recordings can be re-released in real surround-sound a decade hence.