A Multiple-Symbol Turbo Detector for Coded M-DPSK Over Time-Varying Rayleigh Flat Fading Channels

Abstract

Abstract-In the past few years, a number of publications concentrated on developing soft-in-soft-out algorithms for coded noncoherent detection, thereby enabling iterative processing and avoiding the need of channel estimation. This paper focuses on iterative noncoherent detection of convolutionally coded M -DPSK signals for time-variant Rayleigh flat fading channels without the receiver having channel state information (CSI). Specifically we show that a noncoherent soft demodulator for minimum-shift keying, which was previously reported in the literature, lends itself quite naturally for soft demodulation of DPSK. We present an extension of this receiver in terms of persurvivor processing and examine the influence of different symbol labelings.I. INTRODUCTION DPSK is well-known as a robust means to transmit data because it renders the need of carrier phase tracking unnecessary and eludes the problem of phase ambiguities, thus allowing for a simple receiver design. To overcome the entailed penalty in shape of a loss in SNR (approx. 2-3dB for QDPSK against QPSK), multiple symbol differential detection (MSDD) schemes were introduced. Numerous publications center around the idea to extend the observation interval of two symbols for conventional detection of DPSK to larger intervals. Thus e.g. for the AWGN channel, it is possible to attain the performance of differentially encoded and coherently detected M -PSK. In [1] it is shown, how MSDD is implemented for AWGN as well as Rayleigh fading channels. There, a number of maximum-likelihood metrics are derived from the multivariate Gaussian distribution. Several publications adopted the idea of MSDD to implement iterative strategies, in order to combine the benefits of coding and multiple symbol observations. In[2] a noncoherent BCJRtype algorithm is derived which introduces a MSDD-like transition metric by truncating the observations for a transition to a few previous symbols. Hence, the generated APPs are only approximate, which is made up for by feeding back extrinsic information from the channel decoder. Only AWGN is considered in [2]. A similar approach is taken in [3], where the focus lies on turbo-decoding. Unlike [2], the alphabet is not restricted to PSK, but also only AWGN is considered. [4] advances the scheme of [3] by incorporating block fading. Although minimum-shift keying is considered, the described transition metric is generally applicable.