Abstract
This thesis describes investigations into differences between wavefield synthesis (WFS), stereophony and natural sound sources with regard to spatial perception. One aim of the investigations was a comparison and a better understanding of the potential of wavefield synthesis and stereophony to reproduce spatial sound fields. A second aim was to find a rationale for observed perceptual differences by examining the general perception mechanisms. The two sound reproduction techniques were discussed and compared with regard to attributes of localisation, sound colour and distance perception at a fixed listening position. A natural source was considered to be the reference for all investigations. Both for localisation and sound colour, significant differences between the systems were found. Spatial aliasing had a substantial impact on the perceptual performance of WFS, and its reduction or prevention led to significant improvements. Although stereophonic phantom sources were shown not to be localised as certainly as virtual sources in WFS, the sound colour differences between adjacent phantom sources were found to be smaller. It was shown that the wave front curvature does not provide cues for distance perception in WFS for a static listener. A combination of WFS and stereophony, which is the OPSI method, was proposed to improve the performance of WFS with regard to sound colour reproduction. The results showed that by this technique, which incorporates stereophony into WFS, the sound colour perception can be optimised while the localisation properties do not change. The investigation aimed at interpreting the observed perceptual differences through a discussion of the applied perception mechanism. It is hypothesised that different mechanisms apply in the perception of stereophonic and WFS sources. The estimation of the subjective colouration grades by numerical predictors supported this idea, because systems incorporating stereophony were graded better than predicted. A full decolouration, however, as proposed by the association model of Theile could not be proven.