Combined coding and modulation in frequency-selective mobile communications.

Abstract

Due to constraints on spectrum availability and transmitter power, both bandwidth and power efficient communication techniques are desirable for mobile radio. Continuous phase modulated (CPM) signals have gained attention because of their attractive power spectra (Steele, 1992). It has been shown that the trellis coded modulation (TCM) schemes could provide better bit error rate performances compared to the uncoded schemes (Ungerboeck, 1982). Therefore, the combination of TCM which improves error probability and CPM signals which yield low spectral occupancy is expected to provide good coding and modulation over bandwidth and power limited channels such as the one encountered in mobile radio communications.In this research, a Trellis Coded Modulation (TCM) scheme, which combines convolutional coding and partial response Continuous Phase Modulation (CPM) such as Gaussian Minimum Shift Keying (GMSK), is investigated. Also, this study concentrates on the use of rate-half convolutional codes, and GMSK (B(subscript)0T=0.3). The latter has been adopted in the Global System for Mobile Communications (GSM) system.Appropriate codes are selected assuming Maximum Likelihood Sequence Detection (MLSD) based on the Viterbi algorithm using an extensive computer search. The bit-error-rate (BER) performances of the selected trellis coded GMSK schemes are theoretically evaluated in the presence of additive white Gaussian noise (AWGN) and frequency-flat fading. In the case of fading, the analysis is simplified to assume only amplitude-fading, and without considering the effect of fading on the phase of the received signal.Computer simulations are used to evaluate the BER performances of the proposed trellis coded GMSK schemes in the presence of AWGN and practical impairments, such as sample timing offset and carrier phase errors. Coding gains of up to 2.2dB at a BER of 10(subscript)-3 are obtained under ideal sample timing and carrier recovery conditions. This has been achieved without increasing the receiver complexity based on the number of states in the Viterbi decoder, compared to the uncoded GMSK scheme. Furthermore, these coded schemes are more tolerant to sample timing and carrier phase impairments.Also, the BER performances of the proposed trellis coded GMSK schemes have been extensively investigated by computer simulations for frequency-flat and frequency-selective fading channels. In the case of frequency-selective fading, the Viterbi decoding is made adaptive to cater for the channel impulse response variations with time. With this adaptive receiver, the irreducible BERs of the coded scheme is found to be lower than that of the uncoded. Performance improvements are obtained with a trellis coded GMSK scheme using a constraint length 2 code with a Viterbi decoder of 16 states compared to the 128 states required for the uncoded scheme. Further, the coded scheme has shown less sensitivity to carrier phase errors, compared to the uncoded.