DVD-A vs SACD

"This is offensive!" muttered usually mild-mannered Malcolm Hawksford, who was sitting next to me. "I'm leaving." The good professor was right. One thousand or so attendees at the 103rd Audio Engineering Society Convention, held at the end of September in New York, were being subjected to truly terrible sound. The irony was that the sound was that of 2- and 5-channel recordings made with 24-bit resolution and a 96kHz sampling rate, being played over a colored PA system to demonstrate the future of audio, in the form of DVD-Audio.

Of course, it wouldn't be right to draw any inferences from this. It was wrongheaded on the part of the DVD Forum member companies to attempt a public demonstration under such adverse circumstances. And smaller, more sensibly arranged demonstrations for showgoers were to follow. But a shiver did run down my back: If such a major mishap can occur on such an important occasion, what else might have been overlooked by those involved in designing the music carrier to take us into the first decade of the next century? Particularly considering that the DVD Forum's Working Group 4 (WG-4) has been working hard to allow for all possible eventualities in their recommendations for DVD-Audio.

Since its inception in December 1995, WG-4 has been working with the three organizations that represent the worldwide record industry—RIAA, RIAJ, and IFPI—to develop the DVD-Audio specifications. These were unveiled at the same convention (footnote 1) and include: support for pure audio, with additional-value content such as text, video, and graphics data; high-quality audio and multichannel capability; compatibility with DVD-ROM; maximum data rate equal to or less than 9.216 megabits/second, with the possibility of lossless compression allowed for; "navigation" features that would allow the directory information to tell the hardware the contents of the disc (192kHz sampling PCM and Sony DSD encoding are included among the definitions); and copy protection.

Of particular interest to audiophiles are the audio parameters: support for 44.1/88.2kHz and 48kHz/96kHz sample rates; quantization of 16, 20, and 24 bits; and up to six channels, with the possibility for reduced resolution and/or sample rate on some channels as long as neither parameter changes for any one data stream. A two-layer, 8.54GB DVD-Audio disc, for example, could offer three channels of 96/24 and two of 48/20 and play for 78 minutes. And while a separate stereo mix is envisaged, if the producers wish to omit this from a DVD-Audio, information will be included so that a player could automatically derive the stereo mix from the multichannel data.

Also coincident with the New York AES Convention, Sony and Philips gave demonstrations of their proposal for the future of audio: "Super Audio CD." This 1.2mm "hybrid disc" meets the WG-4 requirements but features a "Red Book" data layer that, to an existing CD player, makes the disc appear to be a regular CD. But a second, semitransparent, DVD-like layer can be read by a shorter-wavelength laser. This layer has the capacity to carry multichannel, high-quality sound data losslessly packed in Sony's DSD format, which is a 2.82MHz one-bit datastream (see previous issues of Stereophile). The beauty of this proposal is that just one disc caters to past and future: the disc is both backward- and forward-compatible, which will enormously increase its chances for commercial success. (Why would anyone introduce a whole new medium just to satisfy the demands of maybe one million audiophiles worldwide?)

So, has anything been left out of these two proposals? It appears not, but a paper presented in New York by Bob Stuart of Meridian, who is also Chairman of the lobbying group the Acoustic Renaissance for Audio (ARA), gives pause for thought. Rather than assume anything about existing technology and practice, Bob developed from first principles a set of specifications for a medium that would be capable of, in his words, "for the first time, delivering every sound to which a human ear can respond: sufficient dynamic range; sufficient bandwidth; sufficient linearity; suitable channel coding; and sufficient channels" (footnote 2).

Among Bob's conclusions are that an 88.2kHz or 96kHz sampling rate is higher than necessary to capture the full bandwidth of perceptible sound. A rate of 58kHz would be sufficient, he calculated. Regarding dynamic range, assuming a peak playback level of 120dB spl means that 20-bit linear-PCM words should be sufficient, provided that no further operations were performed on the data. And if noise-shaping and pre-emphasis were used, taking advantage of the greater ultrasonic bandwidth in which to move audio-band noise, 14-bit words would be sufficient to provide full resolution. Given that lossless compression or "packing" is now feasible, the data rate that can be inferred from the above conclusions means that a DVD-Audio disc could be capable of storing six channels of information, each of which would have sufficient bandwidth and resolution to be audibly transparent to all listeners at all times with all types of music. (In an AES workshop on the audio implications of DVD, Bob Stuart emphatically made the point that he does "not think there is currently a lossy compression scheme to which we can entrust our music.")

Judging the preliminary DVD-Audio and Sony/Philips Super Audio CD proposals by Bob's criteria, the first has the capability of meeting his requirements with data space to spare. But as commercially attractive as Super Audio CD is—particularly regarding the single-inventory implications for the record retailer—and as well-presented as Sony's demonstrations were in New York, the data rate does not appear high enough to guarantee audible transparency for multiple channels under all circumstances to all listeners.

So which, if either, will take audiophiles into the next century—DVD-Audio or Super Audio CD? All I can offer is that most useless of answers: "It all depends..."



Footnote 1: This information is based on my notes taken at the presentations. The DVD-Audio specification is expected to be finalized by December 31, 1997.

Footnote 2: See J.R. Stuart, "Coding Methods for High Resolution Recording Systems," Preprint 4639, 103rd Audio Engineering Society Convention, New York, September 1997. Available from the AES, 60 East 42nd Street, New York, NY 10165-2520. Tel: (212) 661-8528. Fax: (212) 682-0477. Website.

X