Super Audio CD: The Rich Report Manufacturer's Comment
Editor: It has recently come to our attention that some questions have been raised about the viability of the Direct Stream Digital (DSD) encoding process, particularly in comparison with multi-bit PCM.
This subject was covered in a November 2000 Stereophile article by David Rich, as well as in a paper presented by Stanley Lipshitz and John Vanderkooy at the 109th AES Convention in Los Angeles. Both Sony and Philips will also present some papers on this subject at the spring 2001 AES Convention in Amsterdam. In addition, James Angus, formerly of the University of York, is preparing a presentation that responds to many of the points raised by the Lipshitz paper, a summary of which follows.
Specifically, Angus asks how triangularly dithered PCM can be "perfect," since any form of quantization will degrade the signal, whether it be linear PCM or an oversampled noiseshaping encoder. Moreover, any form of requantization will produce additional noise, which will occur if there is rounding or truncation during subsequent signal processing.
Clearly, triangularly dithered PCM is no more perfect than a one-bit sigma-delta modulation system like DSD is when operated at an infinite sample rate.
Angus also addresses how a signal is affected via digital decimation when providing a multi-bit, linear PCM output at the desired sample rate. This is because direct conversion to and from PCM requires components with tolerances at least equal to the least significant bit. This implies a minimum tolerance of 0.0015% for 16-bit and 0.000006% for 24-bit (high-resolution) audio. However, one-bit DSD sigma-delta conversion is guaranteed to be monotonic, since it has only one quantization threshold and thus is very linear.
Finally, Angus states that when linear PCM is used directly in a recording system, the analog signal will have passed through a minimum of two steps of decimation and interpolation before it arrives at the listener's ears. In practice, even more of these steps occur due to the widespread use of mixers and sampling-rate converters.
In contrast, since both interpolation and decimation can be eliminated with the DSD sigma-delta format, this signal-processing does not suffer degradations caused by rounding, aliasing, and frequency/time-domain errors.
In summary, such noted researchers as James Angus—as well as leading converter designers like Ed Meitner of EMM Labs and Mike Story of dCS—have confirmed the overall advantages of the Direct Stream Digital process vs PCM, especially as it relates to "real-world" music reproduction.—David Kawakami, Director, SACD Business Center, Sony USA