Latency of spatial audio plugins: a comparative study

The use of spatial audio plugins (SAPs) with Ambisonics processing and binaural rendering has become widespread in the last decade, thanks to their increased accessibility and usability. SAPs are particularly relevant in scenarios involving real-time music playing with headphones, such as networked music performance and individual recreational music-making using backing tracks. However, a crucial issue that has been largely overlooked thus far is the measurement of the processing latency introduced by currently available SAPs. Identifying which SAPs are the fastest is essential to enable designers, musicians, and researchers to create time-sensitive applications involving 3D audio. To bridge this gap, we compared nine systems formed by different SAPs that enable 3D audio management. We measured the latency of each system throughout the third-order Ambisonics plugins pipeline: encoding, room simulation, sound scene rotation, and binaural decoding. In particular, the measurements were performed utilizing different buffer sizes. Results showed that to achieve a minimization of the latency, it is necessary to use a combination of different SAPs from different systems. Based on our measurements, we propose two spatial audio systems that mix different SAPs. Considering a sampling rate of 48 kHz, a Dell Alienware x15 R2 laptop running the Windows 10 operating system, and an RME Fireface UFX sound card, the two systems achieved an overall latency of 0.33 ms and 0.94 ms respectively.