A high-performance multi-purpose DSP architecture for signal processing research

This paper presents a powerful and flexible digital signal processing (DSP) architecture based on the Texas Instruments TMS320VC33 DSP and high speed PCI bus. The DSP board provides a convenient, flexible means to test signal processing algorithms in real-time hardware. Algorithms implemented for several research projects include normalized least mean square (NLMS) adaptive filter, recurrent neural network (RNN), Viterbi decoding, and adaptive beamforming. The low-cost, reconfigurable system is presently being used in various research projects such as multiple channel sampling and filtering MEMS-based acoustic arrays, wireless LAN hardware implementation, and neural net classification of primate EEG waveforms. The paper provides a detailed description of the DSP board, the theory behind its selection of components, and how it is being used in the earlier mentioned research projects.