Quantization and truncation effects in the design of adaptive digital filters

This paper presents an implementation of an adaptive digital filter. The hardware is outlined and the limitations of the design discussed. The effect of rounding on the filter characteristics is described. A method of improving filter performance, when using low adaptive gains, by means of the addition of a dither signal to the coefficient updates is described. Experimental results from the implementation are presented. These results concur with trends predicted by computer simulations, results of which are also included. Potential applications are suggested as are methods for improving the design.