Improving sound field reproduction based on higher-order ambisonics in a small room with a 157-loudspeaker array
Takuma Okamoto, Densil Cabrera, Markus Noisternig, Brian FG Katz, Yukio Iwaya, and Yôiti Suzuki
This article presents a case study of higher-order Ambisonics (HOA) for real-time sound field reproduction in a small room with a 157-loudspeaker array. It addresses a number of specific questions and practical issues on the system design and implementation, such as the reproduction room's acoustic, loudspeaker positioning and radiation patterns, distributed computing and audio channel synchronization, and in more general the achievable accuracy of sound field reproduction. In the current configuration of the system Ambisonics up to order n = 6 is applied and the decoders are rendered in parallel on a cluster of four computers. For this reason, synchronization and communication between the different computers becomes a challenging task for achieving a good system performance. The overall system latency and the inter-channel synchronicity have been measured using time-stretched pulse (TSP) signals. The measurement results have shown a maximum (unsigned) latency of 51 samples, which corresponds to t = 1.1 ms. It is obvious that the acoustic of the reproduction room has a strong effect on the accuracy of the Ambisonics sound field reproduction. To achieve semi-anechoic conditions sound absorption materials have been installed in the room. Finally, spatial filters have been applied to each individual loudspeaker to correct for different orientations with reference to the sweet spot. These filters have been derived from radiation pattern measurements in an anechoic chamber.