Audio Engineering Society

Chicago Section

Meeting Review, January 15th 2002

other meeting reports

view pictures from this meeting

1/15/02 Meeting Highlights
by Kelly Kay

1-bit sigma-delta modulators.

On Tuesday January 15 2002 the Chicago Section of the AES had the honor receiving an extended presentation on 1-bit sigma-delta modulators from AES fellow, Silver Medal recipient (for his research contributions to digital audio), and University of Waterloo Professor, Dr. John Vanderkooy.

The recent introduction of the Super-Audio CD as a new high-resolution consumer format has prompted Dr. Vanderkooy’s study of 1-bit systems. Unlike the current 16-bit/44.1kHz Linear PCM format used in today’s Compact Disk format and the newer formats such as 24-bit/96kHz Linear PCM (used in the DVD-Audio format) the 1-bit 2.82MHz Direct Stream Digital (DSD) format used by Sony and Philips can not be dithered.

By applying “proper” TPDF dither to the traditional and newer Linear PCM formats it is possible to remove the non-linearity and noise modulation that results from truncation. With DSD this is not possible, and the proper use of noise shaping becomes critical. Effectively residual noise is moved up into higher frequencies to reduce the noise level at lower frequencies. This combined with DSD’s inherent low pass filtering is said to provide DSD with a bandwidth of 100kHz.

After explaining the above, Dr. Vanderkooy then proceeded to show us the results of his largely simulation based research. Using a coherent averaging method developed for this task, some of the underlying nonlinear mechanisms in the 1-bit modulator were clearly shown. The noise in DSD is not random, it is in fact fully deterministic noise, resulting in individual spurious tones in the noise floor which can be heard if one “cranks up” this noise floor. Dr. Vanderkooy has found there are two distortion mechanisms involved: sweeping of the idle tone and a saturation tone which adds odd harmonics.

The conclusion was that 1-bit modulators are not self dithered, and that chaos is not as good as real dither. Such modulators are capable of good audio, but they are wasteful of storage requirements and do not have the perfectable characteristics of wider bit width linear PCM. At the end Dr. Vanderkooy left us with the question: Do we really need 1-bit PCM (given the conclusions) when Linear PCM is so much easier?