Dual-State Systolic Architectures for Adaptive Filtering Using Up/Downdating RLS

Abstract

We propose a dual-state systolic structure to perform joint up/down-dating operations encountered in windowed recursive least squares (RLS) estimation problems. It is derived by successively performing Givens rotations for updating and hyperbolic rotations for down-dating. Due to the data independency, a series of Givens and hyperbolic rotations can be interleaved and parallel processing can be achieved by alternatively performing updating and downdating both in time and space. This flip-flop nature of up/down-dating characterizes the feature of dual-state systolic triarray. To further reduce the complexity and increase the throughput rate, Cordic cells can be used to mimic the operations of rowbroadcasting and only one control bit is required along each row of processors. Efficient implementation to obtain optimal residuals and a transformation of the hyperbolic rotation to an algebraically equivalent orthogonal operation to provide a more stable implementation are also considered. This systolic architecture is very promising in VLSI implementation of the sliding-window recursive least squares estimations.