Dynamic Diffuse Signal Processing for Sound Reinforcement and Reproduction

High inter-channel coherence between signals emitted from multiple loudspeakers can cause undesirable acoustic and psychoacoustic effects. Examples include position-dependent low-frequency magnitude response variation, where comb-filtering leads to the attenuation of certain frequencies dependent on path length differences between multiple coherent sources, lack of apparent source width in multi-channel reproduction and lack of externalization in headphone reproduction. This work examines a time-variant, real-time decorrelation algorithm for the reduction of coherence between sources as well as between direct sound and early reflections, with a focus on minimization of low-frequency magnitude response variation. The algorithm is applicable to a wide range of sound reinforcement and reproduction applications, including those requiring full-band decorrelation. Key variables which control the balance between decorrelation and processing artifacts such as transient smearing are described and evaluated using a MUSHRA test. Variable values which render the processing transparent whilst still providing decorrelation are discussed. Additionally, the benefit of transient preservation is investigated and is shown to increase transparency.