Considering high-resolution, multi-channel audio, it is worth re-examining the concept of psychoacoustically optimized noise shaping to ensure that signal quality is preserved when word-lengths are reduced. In this paper, approaches in static (time-invariant) and signal-adaptive (time-variant) noise shaping are discussed. We identify problems occurring when equal-loudness level contours such as the inverse F-weighting curve are used as a model for noise shaping. As a remedy, we present two alternative time-invariant filter designs. Regarding adaptive requantization, we introduce a noise shaper based on work by Verhelst and De Koning with an improved design of the time-variant filter. The paper concludes with a comparative evaluation based on listening tests with different transducers.
展开▼