Adaptive multiple symbol decision feedback for non-coherent detection.
Govender, Nishkar Balakrishna.
MetadataShow full item record
Non-coherent detection is a simple form of signal detection and demodulation for digital communications. The main drawback of this detection method is the performance penalty incurred, since the channel state information is not known at the receiver. Multiple symbol detection (MSD) is a technique employed to close the gap between coherent and non-coherent detection schemes. Differentially encoded JW-ary phase shift keying (DM-PSK) is the classic modulation technique that is favourable for non-coherent detection. The main drawback for standard differential detection (SDD) has been the error floor incurred for frequency flat fading channels. Recently a decision feedback differential detection (DFDD) scheme, which uses the concept of MSD was proposed and offered significant performance gain over the SDD in the mobile flat fading channel, almost eliminating the error floor. This dissertation investigates multiple symbol decision feedback detection schemes, and proposes alternate adaptive strategies for non-coherent detection. An adaptive algorithm utilizing the numerically stable QR decomposition that does not require training symbols is proposed, named QR-DFDD. The QR-DFDD is modified to use a simpler QR decomposition method which incorporates sliding windows: QRSW-DFDD. This structure offers good tracking performance in flat fading conditions, while achieving near optimal DFDD performance. A bit interleaved coded decision feedback differential demodulation (DFDM) scheme, which takes advantage of the decision feedback concept and iterative decoding, was introduced by Lampe in 2001. This low complexity iterative demodulator relied on accurate channel statistics for optimal performance. In this dissertation an alternate adaptive DFDM is introduced using the recursive least squares (RLS) algorithm. The alternate iterative decoding procedure makes use of the convergence properties of the RLS algorithm that is more stable and achieves superior performance compared to the DFDM.