The least squares(LS) minimization problem constitutes the core of many real-time signal processing problems. A square root free scaled Givens rotations algorithm and its systolic architecture for the optimal RLS residual evaluation are presented in this paper. Upper bounds of the dynamic range of processing cells and the internal parameters are analyzed. Thus the wordlength can be obtained to prevent overflow and to ensure correct operations. Simulation results confirm the theoretical conclusions and the stability of the algorithm.
McWhirter J G. Recursive least-squares minimization using a systolic array, Proc. of SPIE, Vo1.431,[2]Real Time Signal Processing VI, 1983, 105-112.[3]Heish S F, Liu K J R, Yao K. A unified square-root-free approach for QRD-based recursive least squares estimation. IEEE Trans. on SP., 1993, SP-41(3): 1405-1409.[4]Barlow J L, Ipsen I C F. Scaled Givens rotations for the solution of linear least squares problems on systolic arrays[J].SIAM J. Sci Stat. Comput.1987, 8(5):716-733[5]Proudler I K, McWhirter J G, Shepherd T J. The IaRD-based least squares lattice algorithm: Some computer simulations using finite wordlength, Proc. IEEE ISCAS, New Orleans: May 1990, 258-261.[6]Liu K J R, Heish S F, Yao K, et al. Dynamic range, stability and fault-tolerant capability of finite-[7]precision RLS Systolic array based on Givens rotations. IEEE Trans. on Cir. Sys,1991, CAS-38(6): 625-636.[8]Frantzeslmkis E N, Liu K J R. A cla9s of square root and division free algorithms and architetures for QRD-based adaptive singnal processing. IEEE Trans. on SP., 1994, SP-42(9): 2455-2469.