Disc Structure and Options

SACD is an extension of the CD format, not an entirely new system created from scratch. An SACD disc is identical to a CD in size and thickness, but can store far more information. By increasing the amount of information representing the musical signal, SACD can provide higher fidelity than CD.

A key provision of SACD is the potential for a dual-layer "hybrid" disc, shown in Figure 1. A hybrid disc has two information layers; the bottom layer is identical to that of a CD, and contains 16-bit/44.1kHz digital audio (the so-called "Red Book" layer). The second "high-density" layer uses much smaller information-carrying pits to store more data. This high-density layer is semi-transmissive, meaning it reflects light from the SACD playback laser, but is transparent to a CD player’s laser.

A hybrid SACD can present CD-quality sound to any of the world’s 600 million CD players, and also deliver high-resolution digital audio when played on an SACD machine. SACD’s backward compatibility with CD has several important implications. The first is that you can play a hybrid SACD in your car or portable CD player, then move the disc inside for SACD playback in your listening room. Second, there’s no dual inventory for record companies and retailers; they carry a single disc for both CD and SACD customers. Third, CD consumers will buy into the SACD format without making a conscious decision to stop buying music in one format and start buying it in a new, incompatible format. After several years of buying what they think are CDs, the average consumer will theoretically have amassed a library of SACD titles.

The hybrid disc is only one of three possible disc options. Record companies could also choose to release music only in high-resolution form, with a single high-density layer and no backward-compatible CD layer. Or the third option, a disc with two high-density layers for about 150 minutes of high-resolution playing time.

The high-density layer contains not only a two-channel stereo version of the music, but can also store a six-channel mix (Figure 2). Artists and record companies will decide whether to release two-channel or multi-channel discs. SACD thus works equally well for two-channel and surround-sound listeners.

Packing eight channels (a two-channel plus a six-channel mix) of high-resolution digital audio in the same space that contained two channels of relatively low-resolution CD-quality digital audio is made possible by three techniques. First, the information-carrying pits in the high-density layer are about half the size of conventional CD pits. Second, the distance between tracks is about half that of CD. Third, the digital audio data are compressed with a lossless coding system that more efficiently encodes that data. For example, rather than store a stream of eight consecutive zeros, the code 8X0 may be recorded instead. This is a crude oversimplification of a very complex technique. What’s important, however, is that this data compression system is perfectly lossless – that is, the same ones and zeros that go in are the same ones and zeros recovered from the disc. Consequently, there’s no reduction in sound quality. SACD’s lossless compression is contrasted with lossy compression systems such as Dolby Digital that discard musical information and degrade fidelity.

Direct Stream Digital (DSD) Encoding

The compact disc and most professional digital audio systems rely on Pulse Code Modulation (PCM) to convert analog audio into digital information. In PCM encoding, the audio signal is sampled at precise intervals. At each sample point, a binary number is assigned to represent the audio signal’s amplitude at the time the sample was taken. Assigning this number is called "quantization." Sampling is akin to taking snapshots of the audio waveform; quantization is "weighing" the audio signal contained in the snap-shot. A movie camera is a form of sampler; the images aren’t continuous, but are taken at discrete time intervals.

Sony and Philips believed that PCM encoding had been so improved that any future advancement would be incremental. Consequently, they devised a radically different method of encoding digital audio called Direct Stream Digital (DSD). In DSD encoding, the analog audio waveform is sampled at a lightning fast 2.8224 million times per second. Each sample, however, generates only one bit of information. The musical information is encoded in the width of the 1-bit-high pulses.

The data rate from DSD is exactly four times that of 16-bit/44.1kHz digital audio. (CD produces an audio datastream of 705,600 bits per second per channel.) This four-fold increase in the number of bits representing the music allows DSD to encode more musical information over a wider audio band-width compared with CD.

Using a single bit to encode the musical signal’s amplitude results in a very high noise floor. But because DSD’s 2.8224MHz sampling frequency produces a theoretical bandwidth of 1.4MHz (according to the Nyquist Theorem, the audio bandwidth is half the sampling frequency), that noise floor is spread over this very wide frequency range. This noise floor can also be shifted from one frequency range to another, a technique called noise shaping. In DSD, noise in the audio band is shifted to a higher frequency where it becomes inaudible (Figure 3). Note that the total noise power in the audio signal doesn’t change; noise shaping merely shifts its frequency away from the audioband.

This technique of quantizing the audio signal with one bit is far from new; the earliest digital audio systems used this so-called "delta-sigma modulation." One advantage of delta-sigma modulation is that it is a much simpler process than PCM encoding and decoding. DSD requires no decimation and interpolation filtering as does PCM, processes that introduce noise and errors into the signal. In fact, the DSD bitstream looks remarkably like the analog signal it represents (Figure 4). In theory, this DSD bitstream could be converted to analog with a low-pass filter made from a single capacitor.

DSD provides a dynamic range of 120dB and an audio bandwidth of more than 100kHz. Note that the 120dB dynamic range is available up to 20kHz; above that frequency the noise shaping technique described earlier reduces dynamic range. Nonetheless, DSD can encode information well above 20kHz, a first for a consumer digital audio system. DSD can also encode a 10kHz square wave, something 44.1kHz PCM cannot come close to achieving. Some designers consider this ability a fundamental prerequisite for high sound quality.

A Direct Stream Digital signal can be converted to a 16-bit/44.1kHz PCM signal for storage on an SACD’s CD-compatible layer. This processing, which Sony calls Super Bit Mapping Direct, preserves some qualities of the high-resolution DSD bitstream. Record companies can thus issue hybrid CDs from a single DSD master. This has the benefit of improving the CD layer’s sound quality.

Super Audio CD and DVD-Audio

It’s no news that a second high-resolution digital audio format, called DVD-Audio, is on the horizon. An extension of the wildly popular DVD-Video format, DVD-A also provides high-resolution multi-channel audio on a CD-sized disc. There are, however, important differences between SACD and DVD-A.

The first of these is the encoding method used to store the music. DVD-A uses PCM representation, with sampling rates anywhere from 44.1kHz up to 192kHz, and quantization word lengths from 16 bits to 24 bits. Music producers can specify the resolution for each channel separately, such as 24-bit/96kHz on the front channels and 16-bit/48kHz on the rear channels. By contrast, SACD lacks this "scalability"; it delivers the same quality on all six channels.

Another important difference is that SACD always contains a separate two-channel mix on the disc. DVD-A, by contrast, can contain a multi-channel version, from which a two-channel mix is created by the player on-the-fly according to codes buried in the bitstream. Producers and artists have voiced concerns about the integrity of the two-channel mix created by the DVD-A player. It is possible to include a separate two-channel mix on DVD-A, but at the expense of playing time and resolution of the multi-channel mix.

In addition to using different encoding schemes, SACD and DVD-A function very differently from each other. Sony likes to say what SACD is by describing what it isn’t. SACD isn’t a convergence product that integrates your computer with your music system; SACD doesn’t connect to a video monitor; SACD isn’t a home-theater product; SACD doesn’t use on-screen menus to navigate and select music titles; SACD has no provision for lossy encoding (DVD-A discs will include a Dolby Digital track for compatibility with the 3.5 million DVD-Video players already sold).

Despite SACD’s appeal (sound quality, backward compatibility with CD, ease of use), it may have a tough time in the marketplace. First, the average consumer isn’t clamoring for a better-sounding replacement for the compact disc. Indeed, in these days of severely compromised MP3 sound quality, CD sound is seen as a reference standard, not as the limited medium it is. Second, four of the five major record labels support DVD-Audio and have no plans to release SACD titles. A large software library is essential to any hardware format’s success. (Notably, however, the audiophile record labels DMP, AudioQuest Music, Telarc, Water Lily Acoustics, Delos, and Mobile Fidelity have lined up behind SACD. Further, Sony owns a massive music library, including the classic Columbia recordings.) Third, the manufacturing cost of a hybrid SACD is higher than that of CD, making music more expensive not only for the SACD customer, but for listeners with CD players. Telarc is talking about a beginning list price of $24.95 for hybrid discs, which they expect to fall to $19.95 within a year. That’s still more expensive than CD for the cost-conscious mass-market consumer.

Sony downplays the prospect of a format war between SACD and DVD-Audio by pointing out these differences between the formats. They contend that DVD-A is more of a video-oriented, mass-market format, while SACD is a pure High End music medium that emphasizes sound quality above all else. Still, it’s likely that DVD-A will dominate by virtue of the greater number of titles expected to become available, and the popularity of the DVD-Video format. These machines will play only the sonically compromised Dolby Digital track on DVD-A, not the high-resolution digital audio of which DVD-A is capable.

So which sounds better, PCM at 24-bit/96kHz on DVD-Audio, or Direct Stream Digital on Super Audio CD? That’s a tough question to answer without comparing a microphone feed (or analog tape source) to both 96/24 and DSD. From what I’ve heard, however, DSD appears to have an advantage in that it sounds more analog-like, particularly in the treble. Some good listeners in the recording industry, who have fully evaluated both 96/24 and DSD, share my view. On the other hand, designers for whom I have the greatest respect maintain that high-resolution PCM is fundamentally superior to DSD’s 1-bit representation.

I expect both formats to survive because they address different markets. Moreover, several large Japanese manufacturers have announced plans for a single machine that plays all formats: DVD-Video, DVD-Audio, SACD, and CD. The availability of such universal players no longer forces the consumer to choose sides, as they might have to do in a conventional "format war," and instead allows him to play any disc without even thinking about the format. Ultimately, that’s good news for SACD, and by extension, for music lovers long dissatisfied with the CD format.