Performance Evaluation of Non-quadratic Algorithms in Adaptive Channel Equalisation

S.A. Jimaa, M.I. Jadah (UAE), and B.S. Sharif (UK)


Equalisation, Adaptive Signal Processing, Non-Quadraticcost-functions


This paper studies the characteristics and behaviour of two non-quadratic cost functions algorithms in term of their convergence rate and BER performances. The two algorithms are the least mean lower order (Lp) and the least mean switched error norm (LMSF) algorithms. The Lp algorithm is based on the cost function E[ekp ], where p=1.4. The LMSF algorithm consists of applying the LMF algorithm and switching to the LMS algorithm when the absolute value of error is greater than one. Performance of the two algorithms is compared with the Least Mean Square (LMS) algorithm by means of computer simulations, which indicate that the stable dynamic range of the step-size in the two algorithms is larger than that in the LMS algorithm. Accordingly, linear filters adapted with the Lp and LMSF algorithms are more stable compared with those using LMS. Moreover, the results show that the BER performance of the LMSF is better than that of the LMS and Lp algorithms.

Important Links:

Go Back