Adaptive differential detection using linear prediction for M-ary DPSK

This paper proposes a novel adaptive differential detection scheme (adaptive DD), which can significantly reduce the irreducible bit-error rate (BER) of M-ary DPSK due to Doppler spread by the adaptive linear prediction of the reference signal. The predictor coefficient is adapted to changing channel conditions by using the recursive least-square (RLS) algorithm. A phase sequence estimation based on the M-state Viterbi algorithm (VA) and another based on the decision feedback algorithm (DFA) are presented. A theoretical BER analysis is presented for adaptive DD-DFA. BER performances of 2 and 4DPSK in Rayleigh fading channels are evaluated by computer simulations. When the RLS forgetting factor of β-1 is used, simulation results show that the irreducible BER of 4DPSK can be reduced to 7.2 x 10 5 (3.9 x 10 4) for VA (DFA) while conventional DD offers 3.9 x 10 3 when foTb (maximum Doppler frequency times bit duration) -0.01 and average Eb/No [signal energy per bit-to-additive white Gaussian noise (AWGN) power spectrum density ratio] -60 dB, where most errors are produced by Doppler spread. Adaptive DD is also effective in AWGN channels-simulations show that for the case of 4DPSK, a performance gain of 1.2 (0.7) dB is achieved over conventional DD for VA (DFA) at BER = 10 3.