A computer application was developed to simulate the process of microphone positioning in sound recording applications. A dense, regular grid of impulse responses pre-recorded on the region of the room under study allowed the sound captured by a virtual microphone to be auralised through real-time convolution with an anechoic stream representing the sound source. Convolution was performed using a block-based variation on the overlap-add method where the summation of many small subconvolutions produced each block of output data samples. As the applied RIR filter varied on successive audio output blocks, a short cross fade was applied to avoid glitches in the audio. The maximum possible length of impulse response applied was governed by the size of audio processing block (hence latency) employed by the program. Larger blocks allowed a lower processing time per sample. At 23.2ms latency (1024 samples at 44.1kHz), it was possible to apply 9 second impulse responses on a standard laptop computer.

Download