Sunday, May 12, 11:00 13:30 h, Room K1
W11 Digital Audio Signal Processing The Good, the Bad, and the Ugly (Presentations and Roundtable)
Chair: Stanley Lipshitz, University of Waterloo, Waterloo, Canada
High-quality digital audio signal processing is not as straightforward as it appears at first sight; and, as we move into the era of high-resolution digital audio, the quality of the associated signal processing must increase commensurately. The oft-repeated refrain that, once a signal is in the digital domain, it is robust and impervious to degradation from further processing, is certainly untrue! Almost every digital operation has an associated penalty in terms of noise and/or distortion. Traps abound, and some of the more esoteric types of processing errors may not be revealed without careful listening tests or special test signals. An alternative title for this Workshop could be: Digital Audio Signal Processing The Right Way and the Wrong Way. The presenters are all experts in various aspects of digital audio signal processing, and the plan is to have each of them address a signal-processing area which, in their experience, is subject to some of these pitfalls, and to explain both the problem and its solution. Although this Workshop will assume that the audience has some familiarity with digital signal processing theory and practice, the presentations will be such that even those without much background will benefit from attending.
Presenters and Presentations:
Stanley Lipshitz (Chair), University of Waterloo, Waterloo, Ontario, Canada
Introductory Survey of Quantization, Dither, and Noise Shaping
Abstract: Good digital signal processing is just as difficult as good analogue signal processing! Almost any operation on digital data results in correlated round-off errors, and these can be pernicious if not treated properly. This talk will review the nature of these quantization errors, show how they can be made innocuous by proper dithering, and discuss the need for dither also in noise shapers and sigma-delta modulators. Examples will be played to illustrate these features.
Peter Craven, Algol Applications Steyning, West Sussex, UK
Abstract: It is by no means true that professional digital audio processing is
free of ugliness. Two misconceptions that have potential to cause ugliness will be
discussed: (a) FFT processing gives perfect control in the frequency domain; (b) 32-bit floating-point arithmetic avoids quantization problems.
Derk Reefman, Philips Research Laboratories, Eindhoven, The Netherlands
The Right and the Wrong Way to Process 1-Bit Signals
Abstract: An overview is given of various ways to process DSD signals. Attention will be paid to some pitfalls, and ways to either avoid them or to cover them up. Also some interesting different possibilities for the creation of bitstreams will be addressed.
Rhonda Wilson, Meridian Audio, Huntingdon, UK
DSP for Active Loudspeakers
Abstract: The processing power of DSPs has steadily increased over the last ten to fifteen years. The physics and acoustics of loudspeakers has remained essentially unchanged. This talk will discuss various ways in which the power available in a DSP can be used in the design of high-performance active loudspeakers. Particular attention will be paid to: (a) the balance between the mechanical, acoustical, analogue electronics, and DSP designs; (b) the advantages of high-resolution processing; (c) the options available to the DSP engineer which are not available to the analogue engineer.