A HIGH-PERFORMANCE REDUCED-COMPLEXITY GMSK DEMODULATOR

A four-state adaptive maximum-likelihood sequence estimation (MLSE) Ga ussian minimum-shift keying (GMSK) demodulator for modulation (BT = 0. 3) on AWGN channels is analyzed and simulated. This demodulator uses t he linear representation of GMSK signals and achieves near-optimum BER performance. The channel-impulse response used in the MLSE: demodulat or is initialized to the highest energy component in the linear repres entation, and then adapted in a decision directed mode to offset any p erformance losses incurred by initially ignoring other lower energy (a nd time-varying) components in the linear representation. The number o f MLSE states is reduced to two, at about 0.1-dB performance loss, by implementing a whitening matched filter which concentrates most of the GMSK pulse energy in its two leading samples.