On real-time implementation of the BNLMS algorithm using the SHARC ADSP-21489

This paper reports on a real-time implementation of a noise canceller using the BNLMS adaptive filter algorithm on a floating-point Digital Signal Processor (DSP), more precisely, the high performance SHARC® ADSP-21489 from Analog Devices. Before an actual hardware implementation, extensive analysis, usually carried out with the help of a simulation environment, is required in order to choose optimal parameters of the adaptive algorithm. The hardware implementation poses additional challenges as real-time constraints, such as execution time, must be satisfied. The implementation described in this work has been found to achieve effective results in noise cancelling, with suitable power consumption and meeting the necessary requirements of a real-time system, precise timing, and reliability. Relevant implementation details are presented, followed by a discussion of the results based on objective as well as on subjective measurements.