Fast pipelined CORDIC-based adaptive lattice predictor: algorithms and architecture

The authors present a novel CORDIC-based adaptive algorithm and a pipelined architecture for unnormalised lattice prediction filter. Previously, they have presented a CORDIC-based adaptive lattice filtering (CALF) algorithm for normalised lattice filters which features a sign-sign direct (rotation) angle updating scheme (Hu and Liao, 1992). The authors consider a delayed CALF (DeCALF) algorithm in which the rotation angle is updated based on `delayed´ prediction errors. In doing so, they are able to develop a fully pipelined implementation of DeCALF which achieves B-fold throughput rate increase where B is the number of CORDIC iterations (stages). This is accomplished at insignificant hardware overhead and minor parameter tracking performance degradation