A Question of Bits

Someday we may speak wistfully to our grandchildren about the "golden age" of digital audio when consumer formats (CD and DAT) contained a bitstream that was an exact bit-for-bit duplicate of the original studio master recording—not a digitally compressed, filtered, copy-resistant version whose sound is "close enough" to the original. Digitally compressed formats such as DCC and MiniDisc represent, in effect, a return to the pre-CD era when consumer-release formats were always understood to be imperfect copies of the studio original. Even the most ardent audiophile accepted the fact that LPs and mass-produced tapes did not, and could not, sound as good as the master tapes they were derived from.

This became obvious to Boston-area audiophiles in the mid-'70s when Victor Campos, now a design consultant to NAD, launched a weekly program on WGBH-FM called "Adventures in Sound." Over a period of several years Victor played Dolby-A copies of hundreds of studio master tapes, obtained from engineers at major record companies—DG, RCA, Vanguard, etc. To present the sound of these recordings at their best, he cleaned up the station's signal path and bypassed the dynamic compressor and modulation limiter that all FM stations use. The Dolby-A decoder was placed not at the output of the tape recorder but ten miles away at the transmitter, so that all of the station's electronics and its studio-to-transmitter link were included in the noise-reduction path.

Result: listeners marveled at how much better the master tapes were than the best available LP playback—smoother, airier (despite the 15kHz limit of FM stereo), with more authentically textured bass and a wider, deeper, more stable soundstage, unmarred by ticks, pops, rumble, or inner-groove distortion. In the coming era of compressed-digital consumer formats the disparity between studio originals and consumer formats will be smaller than it was in the 1970s, but the CD will remain the "high-end" choice for listeners who insist on studio-quality two-channel sound.

Today's analog satellite TV signals and 12" laserdiscs, with their literal reproduction of a movie image, may also be remembered fondly a decade hence. By the end of the 1990s most people will be buying movies on 5" CDs containing digitally compressed video images together with digitally compressed audio. Heavy TV watchers will get satellite TV from small-dish receivers that pull in 100+ channels of digitally compressed video. (Today's 12' satellite dishes will revert to their former status as a professional tool and rich-boy's toy.) The existing system for widescreen 1125-line high-definition TV, which was invented a decade ago at NHK in Japan, will be only a studio "production" format. Home-delivery HDTV formats, whether via broadcast, cable, laserdisc, or videotape, will all employ digital compression to shrink the required bandwidth from 30MHz to the 6MHz bandwidth of today's TV.

In recent months it has become clear, not as a topic for speculation but as a pattern of concrete developments, that the MiniDisc and Digital Compact Cassette—the first reduced-bit-rate formats due to arrive in stores within the next few months—represent just the tip of the camel's nose getting under the tent. Virtually all of the new signal-delivery systems that are due to arrive in the consumer market in the next few years, for both audio and video, are based on computerized bit-rate compression schemes. This was the message of many technical papers and exhibits at the Las Vegas convention of the National Association of Broadcasters (NAB) in April.

In the past, the word "compression" has meant reducing or limiting the dynamic range of a recording. That is a valid use, for example to squeeze the dynamics of a symphony so that it can be enjoyed in a noisy environment such as a car or plane. But in the world of digital recording and processing, compression means bit-rate reduction—using fewer bits per second to transmit or preserve a signal.

On the face of it, compression seems antithetical to the spirit of high-end audio. Audiophiles and videophiles are idealists in pursuit of a more satisfying aural or visual experience. Compression, on the other hand, is about practical and economic goals—making recorded signals available to consumers in formats that are more compact, portable, or affordable.

This may involve slight compromises in signal quality. No doubt the first DCC and MiniDisc systems will be criticized in these pages for audio quality that is not quite as good as the best CD sound (which in turn may not be quite as good in some respects as the best LP sound). But compression will also bring substantial benefits, to audiophiles and videophiles as well as to mass-market consumers.

In some cases the motivation for digital compression is not lower cost but sheer necessity. For example, the direct way to broadcast digital audio requires using most of a TV channel just for the digital code (without any picture). One experimental digital radio station has been broadcasting that way in Boston since 1986, on UHF Channel 44. But there simply aren't enough available TV channels to put dozens of digital radio stations on the air using that method in every town. Compression will make digital radio a practical reality.

High-definition TV transforms television into a much more satisfying and involving medium, in part by providing movie-like picture resolution, aspect ratio, and size. Besides improving the enjoyment of movies at home, HDTV also does a much better job at conveying the visual aspect of opera, ballet, rock concerts, and sports. But an HDTV video signal occupies the bandwidth of five or six standard television channels, and without digital compression we might not get it in our lifetime.

Perceptual coding
Digital compression may involve either "lossless" or "lossy" coding. Lossless coding is widely used for computer programs, where the change of a single code bit could cause the program to crash. The original signal is reconstructed exactly in playback, with no loss of information. The bit-rate is reduced by using a more efficient coding scheme and by removing redundant bits.

What is redundant? In a video signal where half of the picture is featureless blue sky or an unchanging wall behind a newscaster, dramatic bit-rate savings can be achieved by transmitting only the pixels that change from line to line or from frame to frame. In binary coding for either audio or video, long strings of ones or zeros can be replaced by shorthand: in place of 11111111, send 1*8.