Focality of sound source placement by higher (ninth) order ambisonics and perceptual effects of spectral reproduction errors

Higher-order ambisonic rendering is an increasingly common soundscape reproduction technique that, in theory, enables presentation of virtual sounds at nearly any location in three-dimensional (3D) space, relatively unconstrained by the veridical locations of loudspeakers. We evaluated whether 3D sound reproduction through a ninth-order ambisonic loudspeaker array was indeed sufficiently accurate to probe the limits of human spatial perception. We first estimated minimum audible angles for human listeners for a variety of reference points on the horizontal plane. We demonstrated that the system can reproduce sounds with a spatial resolution that is equal or superior to the limits of human acuity, at least on the horizontal plane at the front. Importantly, the resolution of ambisonic reproduction appeared equivalent for regions of the system with high and low loudspeaker density. We also estimated localization cues at the same locations and showed that although localization cues for low-frequency components were well preserved, they were somewhat distorted for components above 4000 Hz. Finally, we provide evidence that these high-frequency distortions can serve elevation cues by human listeners. In summary, we showed that a ninth-order ambisonic system is able to render highly focal sound sources, and that high-frequency reproduction distortions may introduce unwanted localization cues.