X-MCFX: Comparison Of Partitioning Schemes in a Non-Equal Partitioned Multi-Channel Convolver
Luca Battisti, Maria Cristina Tommasino · 2025
Convolution is a widely used signal processing operation with many applications in digital signal processing. In audio processing, for example, it is used to impose spectral and/or temporal structures on signals. The acoustic footprint of real or virtual environments, stored as impulse responses (room IR), can be transferred entirely to another sound signal. Convolution also plays a central role in acoustic measurements, where impulse responses are obtained using the sine sweep technique to enable precise characterisation of rooms and devices. In a multichannel scenario, convolution has an even wider range of applications. One example is the use of 3D sound techniques such as Ambisonics or Wave Field Synthesis, which both require multi-channel mathematical computations that can be performed as a convolution operation. A similar concept can be applied during the audio post-production mixing phase, where directional audio objects are converted to Ambisonics for reproduction in similar speaker setups. This paper analyses an existing algorithm related to multichannel convolver software. Its efficiency has been evaluated, along with optimisation of performance in terms of computational cost. The results demonstrate that different partitioning schemes for the convolution filter can significantly impact computational cost in real-time, multichannel scenarios.