Advanced noncoherent demodulation and decoding

• Main Author: Goh, J. G.
• Published: University of Cambridge 1995
• Subjects: 621.382
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.599459

This dissertation is concerned with some advanced noncoherent communication techniques. In many practical cases, phase coherent communication is not desired or impractical due to its disadvantages of long acquisition time, data loss in phase estimation overhead, loss of lock associated with a phase estimation circuitry and high performance loss resulting from phase estimation error. The dissertation is divided into three main parts. In the first part, a new noncoherent detection scheme called the Noncoherent Viterbi algorithm (NVA) for constant amplitude signals is presented. In this algorithm, a complex path is derived for each state of the trellis, and the complex paths are then used to compute the path metrics. The algorithm is applied in Trellis-coded multilevel PSK and uncoded continuous phase modulation (CPM). It is shown analytically that the performance of NVA approaches that of the coherent detection, in a constant phase channel. A forgetting factor is introduced in order for the algorithm to dynamically track the channel. The bit error probabilities and lower bounds which are derived in terms of the forgetting factor provide an insight of the sensitivity of a scheme with respect to the channel phase noise. In the presence of random walk phase noise, the NVA is superior to the decision directed phase-locked loop (DDPLL). The second part of the dissertation presents the application of parallel concatenated codes with a recently discovered iterative decoding, namely 'Turbo-codes', to various types of binary noncoherent channels in AWGN as well as Rayleigh fading environments. The metrics of the binary noncoherent schemes are derived for the two popular iterative decoding algorithms: maximum <I>a posteriori </I>(MAP) algorithm and soft-output Viterbi algorithm (SOVA). Performance of iterative decoding in coherent and noncoherent detections are compared. As in the coherent channel, Turbo-codes converge well in the noncoherent channels investigated and outperform the conventional convolutional codes by a large margin. In the third part, the multiple-access capabilities of two types of noncoherent frequency-hopped code-division multiple-access channels have been studied and compared.