Battle of the Discs

By Richard Elen
August 2001

In case you haven’t noticed, the high-quality audio disc marketplace is in the middle of a format war, reminiscent of the days of VHS vs. Betamax. But unlike that combat, when the clearly superior technology lost to more powerful marketing and the undeniable power of porn, this time it is much less clear which technology is superior, and even less easy to guess who will win. But while the war goes on, we consumers are the losers.

Several years ago, there was talk about the need for a new, high-quality audio disc (HQAD) format that would overcome the limitations of the CD. An obvious physical basis for the format is the Digital Versatile Disc – DVD – which has the high-density storage capability required to get super-quality audio on a disc with a respectable playing time, along the same lines as a conventional CD. Two technologies emerged from that discussion: DVD-Audio and Super-Audio CD.

DVD-Audio, like the compact disc before it, is based on Pulse Code Modulation (PCM) digital audio technology. In this process, incoming analog audio is fed to an analog-to-digital converter that essentially measures ("samples") the instantaneous value of the signal and stores the result as a binary number: a "digital word." This is done tens of thousands of times a second – 44,100 samples per second in the case of the compact disc, and up to 96,000 or even 192,000 times per second for DVD-Audio. The more samples per second (the higher the "sample rate"), the better the quality – up to a point. In fact, we can’t really hear any significant difference above about 64,000 samples per second (a "sample rate" of 64 kHz), but the DVD-A sample rates of 88.2 and 96 kHz are simple multiples of the 44.1 and 48 kHz rates used in CD and DVD-Video, which means that converting between them is easier and sounds better. And MLP (Meridian Lossless Packing, the lossless compression system used in DVD-Audio) recovers much of the extra storage space needed by higher sample rates. Even so, it can be argued that sampling at 192 kHz is a complete waste of time.

PCM stores each sample as a digital word a certain number of binary digits ("bits") in length. Up to a point, the number of bits determines the noise floor and dynamic range of the digital system, and in some cases the amount of detail you can capture. The compact disc format uses 16-bit words to give a noise floor at about –96 dB. DVD-Video generally uses 20-bit words. And DVD-A can use up to 24-bit words. Above about 22-24 bits, there is no point in adding bits, as the noise floor of the system is below the threshold of thermal noise and other effects. A true 24-bit converter, for example, would give a noise floor of –144 dB, and apart from the fact that you can’t realistically do anything with a –144 dB noise floor, no one can make components that quiet anyway: most 24-bit converters are hard pressed to reach –120dB. So 24 bits is more than enough.

This all means that the general sample rate and word length used in DVD-Audio surround recordings today – 24 bits, 96 kHz sampling – is just about right for high quality PCM.

However, there is a different way of doing digital recording, which has been developed by Sony and Philips over the past several years, and has been waiting in the wings for the appropriate moment to emerge. This technology, called "Direct Stream Digital" or DSD, differs fundamentally from other formats in the way that it stores digital information. Instead of storing the digital sample values as "words" with up to 24 or so bits, a stream of individual bits is stored. This on the face of it sounds impossible, but in essence, all each bit has to do is to indicate whether the sample is louder or quieter than the one before. In fact, the vast majority of converters used for regular PCM until a decade or so ago actually worked by generating a bitstream. However, the bitstream was then "decimated" to create the familiar multi-bit digital words used in PCM. DSD proponents claim that what they are doing is simpler than PCM, because they just leave the decimation stage off and record the live bitstream.

The sample rate used in DSD is also enormous: 2.82 MHz, or 64 times the conventional sample rate of CD, although "oversampling" is also commonly used in conventional PCM. This theoretically offers an extended high-frequency capability. In addition, the D/A converter in a DSD system is essentially nothing more than a very special low-pass filter. A cheaper filter would suffice for portable players, for example, while a more highly specified filter can be used in hi-fi gear – and both would play the same source material.

DSD is the technology used in the Super Audio CD (SACD) system released by Sony, Philips and their affiliates in competition with DVD-A. SACD proponents say that it sounds significantly superior to PCM and DVD-A, with a smoother high-frequency performance and added detail and clarity. Like DVD-A, it also includes a lossless compression system, so quality is not compromised by lossy compression and the playing time is similar to that of a CD or DVD-A disc.

Another touted benefit of the SACD system is that discs can be dual-layer, with a standard Red Book CD-compatible layer on top of the high-resolution DSD information, allowing SACDs to play on conventional CD players, albeit at lower resolution (this is, incidentally, also possible on DVDt). However, the truth is that being able to play an SACD on a regular CD player is only a benefit if the new disc sells at the same price as a compact disc – which is not the case, thereby locking out the majority of today’s price-sensitive record buyers. This is probably why very few SACD titles on the market include the Red Book layer

Not everyone agrees about the theoretical technical superiority of DSD and the Super Audio CD. There have been many learned papers presented at successive Audio Engineering Society conventions recently in which the pros and cons have been hotly debated. One complaint is that the DSD process introduces a very high level of ultrasonic noise. Demonstrations of SACD often make special mention of the audio system’s wide frequency range, including, for example, tweeters that will handle up to 100 kHz. It is, of course, doubtful whether there is any "audio" information at such high frequencies (and if there was, it would be submerged under the noise, with a dynamic range of only 40 dB). The DSD process may produce such a high level of noise at such frequencies that if the replay system cannot handle it, it may fail. You are therefore left with two choices: either use more robust equipment that will handle absurdly high levels at absurdly high frequencies, or filter the input signal and thereby remove one of the main alleged benefits of DSD, namely an extended HF response, by "turning it down" to 96 kHz PCM levels.

There are also challenges with DSD when it comes to digital signal processing. To handle DSP operations – necessary in the process of making most records, where mixing, level changes, EQ and compression often take place in the digital domain – the DSD signal must be decimated (turned into something very much like PCM), processed and then returned to the bitstream format. There is arguably little point in using DSD if you are going to turn it into something like PCM on the way to the master.

It has also recently emerged that in the studio environment, DSD does not employ a "bitstream" at all, but rather a form of quasi-PCM, albeit at a very high sample rate, probably in order to allow DSP processing as described above. Some people have considered this a form of cheating.

As a result of all this, it has been suggested that the primary application for "pure" DSD is for recordings in which the technique is essentially to capture a live performance and then edit it without the use of digital signal processing. This may be particularly possible in the classical field, and for other types of music where there is an actual performance to capture, but this only applies to about five percent of recorded music. The rest is created in the studio, multitracked, and reliant on a great deal of mixing, processing and other digital operations. It may be that for these types of recordings – i.e. most of the recordings that are carried out – PCM is more appropriate. However, it may also have an application where analog recordings are being reissued.

Despite the controversy and the apparent challenges of DSD and Super Audio CD, supporters of the latter are very vocal in their enthusiasm. These supporters include some of the top names in the classical field, such as Michael Bishop at Telarc, who has an astounding reputation. If he says it sounds better, few will doubt him. There is no arguing with the fact that DSD and SACD offer audible benefits for some kinds of material and recording techniques.

As a result, the question is whether or not these small, specialized areas of the recording industry can support their own recording system and disc format, which may very well be superior to DVD-A and PCM for certain types of largely "purist" material and styles of recording. The answer at this time has to be a resounding "maybe."

It is difficult to find out how sales are going for either DVD-A or SACD. Most record stores carry a selection of both, and there are more SACD titles than DVD-A titles available at the time of writing. But, and this is a very big but, almost all the SACD titles are stereo only. Sony originally launched SACD as a system for stereo audiophiles, and the fact is that they are a dying breed. A look at the audio magazines will instantly tell you that there are hardly any "stereo" magazines left. They have all either folded or metamorphosed into home theater titles. Surround Sound is the medium of the future, and by treating it so lightly, the proponents of SACD have done themselves a disservice. At last, surround SACD players are now emerging, and so are a limited number of discs. However, at present, there is far more surround DVD music available than there is on SACD. Meanwhile, DVD-A is hobbling itself by silliness over watermarking.

What is to be done? It is quite certain that the existence of two systems is hurting the sales of high-quality audio discs by making both buyers and manufacturers uncertain about entering the market. Very soon, all DVD players will be able to play DVD-Audio discs, and they are already very affordable, with prices starting as low as $299. SACD players are still expensive and difficult to find, especially if you want surround.

At one time, Sony and Philips were involved in the Working Group that derived the DVD-Audio specification, and the DVD-Audio specifications theoretically include the ability to release bitstream audio discs, but in fact, the way things have developed, DVD-A players won’t play SACDs and SACD players won’t play DVD-A discs, although often both will play DVD-Video discs.

This isn’t good enough. All high-quality audio disc players should be able to play both types of disc – DVD-A and SACD. The technologies are very close as far as getting data off the disc is concerned, and although there are no chip sets today available that will decode both types of data, they would become available almost at once, given the demand. Then the consumer would not have to choose which system to buy, and the producers could choose the recording technology that would provide the best results for their material and their recording techniques, safe in the knowledge that consumers could play them back.

If such an agreement to produce truly "universal" players does not emerge soon – if the two camps cannot put away their greed and focus on the common good and on the consumer’s needs – the future of both systems, and of high-quality audio discs in general, may be severely limited. Without universal players, both formats may die a lingering, expensive and embarrassing death, which would be a major disappointment to those of us who enjoy the higher audio quality offered by these technologies.

It does not have to be that way. All that’s needed is a little cooperation.