An advanced real-time digital signal processing system for linear systems emulation, withspecial emphasis on network and acoustic response characterization

Abstract

A fast digital signal processing (DSP) system is described that can perform real-timeemulation of a wide variety of linear audio-bandwidth systems and networks,such as reverberant spaces, musical instrument bodies and very high order filternetworks. The hardware design is based upon a Motorola DSP56309 operating at 110million multiplication-accumulations per second and a dual-channel 24 bit codecwith a maximum sampling frequency of 192 kHz. High level software has beendeveloped to express complex vector frequency responses as both infinite impulseresponse (IIR) and finite impulse response (FIR) coefficients, in a form suitable forreal-time convolution by the firmware installed in the DSP system memory. Analgorithm has also been devised to express IIR filters as equivalent FIR structures,thereby obviating the potential instabilities associated with recursive equations andnegating the traditional deficiencies of FIR filters respecting equivalent analoguedesigns. The speed and dynamic range of the system is such that, when sampling at48 kHz, the frequency response can be specified to a spectral precision of 22 Hz;when sampling at 10 kHz, this resolution increases to 0.9 Hz. Moreover, it is alsopossible to control the phase of any frequency band with a theoretical precision of10−5 degrees in all cases. The system has been applied in the study of analogue filter networks,real-time Hilbert transformation, phase-shift systems and musical instrument bodyemulation, where it is providing valuable new insights into the understanding ofpsychoacoustic mechanisms.