Pipelined QR decomposition based multi-channel least square lattice adaptive filter architectures

QR decomposition based multi-channel least square lattice (QRDMLSL) algorithm possesses good numerical property and regularity which are attractive for VLSI implementation. But due to the presence of local recursive loop in its implementation, the algorithm’s speed is limited to the computation time of each computation cell. In this paper, a novel approach for pipelining QRDMLSL adaptive filtering algorithm is developed. The proposed architecture is pipelined at fine-grain level, thus it can be operated at arbitrarily high speed. Furthermore, it provides exact least square solution, so there is no performance degradation. Computer simulations are presented to demonstrate the functionality of the proposed pipelining methodology.