Abstract

A novel realization of IIR decimation filters is proposed which is based on merged delay transformation. The transformation is derived analytically and can be applied directly to first- and second-order IIR filters. Computational efficiency is enhanced because the current output can be directly computed from Mth old output. The output data rate is decreased by M by merging M number of delay elements in the recursive path. The proposed transformation is applied to higher-order IIR filter by decomposing it into parallel first-order and second-order sections. This transformation not only gives better stability for coefficient quantization but also reduces the requirement on processing clock, for sample, rate reduction. Filtering and down sampling are performed in the same stage. Number of multiplications is reduced by 45% as compared to the conventional IIR filters where all output samples are computed.