Volker Pauli

Design and Analysis of Low-Complexity Noncoherent Detection Schemes

Entwurf und Analyse komplexitätsreduzierter inkohärenter Detektionsverfahren

Band 16, Erlanger Berichte aus Informations- und Kommunikationstechnik, Herausgeber: A. Kaup, W. Koch, J. Huber. Shaker Verlag, Aachen, 2007.


Signal detection without the need for channel state information at the receiver, so-called noncoherent detection, constitutes an interesting alternative to the widely-used concatenated scheme of channel estimation and subsequent detection, so-called coherent detection, in adverse fading channel environments. However, existing approaches to noncoherent detection are either too complex or fail to achieve satisfactory power efficiencies under general fading conditions. This thesis deals with the design and analysis of power-efficient, yet low-complexity noncoherent detection schemes for point-to-point multiple-input/multiple-output (MIMO) communication systems. The starting point of this work is multiple-symbol differential detection (MSDD), which simultaneously processes blocks of N>2 received samples to estimate the transmitted data. While MSDD is known to be capable of achieving power efficiencies close to that of coherent detection with perfect channel state information if N is large, it quickly becomes computationally intractable, as the candidate-signal space is (N-1)-dimensional, i.e. the number of possible transmit sequences grows exponentially in N. The application of tree-search algorithms, that have attracted considerable attention in the recent communications literature, to overcome the complexity limitations of MSDD is investigated. Furthermore, a nested MSDD structure consisting of an outer and a number of inner tree-search decoders is developed, which renders MSDD feasible over wide ranges of system parameters. A second approach to low-complexity MSDD based on methods from combinatorial geometry is proposed for the interesting special cases of differential phase-shift keying (DPSK). This approach is particularly appealing due to the fact, that its complexity is practically constant, whereas tree-search based methods may have very high instantaneous complexities. Inspired by decision-feedback differential detection (DFDD) and the observation that decisions in the different positions of the MSDD observation window are not equally reliable, a new noncoherent detection scheme, referred to as decision-feedback MSDD (DF-MSDD) is devised. DF-MSDD achieves power efficiencies comparable to those of MSDD while the dimension of the candidate-signal space and thereby the decoder complexity is reduced significantly. Here, the tree-search methods developed for conventional MSDD are still applicable, such that a computationally highly efficient noncoherent detector results. Following the development of the detection methods based on a generic MIMO channel model, their application to transmission over time-selective and frequency-nonselective MIMO channels and to transmission using orthogonal frequency-division multiplexing (OFDM) over time- and frequency-selective channels is studied. While well-known differential space-time modulation (DSTM) is applied for transmission over frequency-nonselective MIMO channels, a new signal allocation scheme for OFDM-based transmission over frequency-selective channels is devised, which makes use of both spatial and/or spectral (multipath) diversity and is particularly apt for power-efficient, low--delay MSDD. For this transmission scenario the application of a two-dimensional observation window to MSDD to exploit channel correlations in both time and frequency direction is investigated. These practical aspects of this work are complemented by analytical studies regarding the achievable power efficiency and computational complexity of the different detection schemes. These investigations provide interesting insights into the connections between the performances of the different detection schemes and their dependence on system and channel parameters and into the behavior of the decoder complexity as a function of the system and channel parameters, respectively. In consequence, they provide valuable guidelines for quick decoder design and make time-consuming system simulations largely expendable. In summary, this work shows how power efficiencies very close to that of idealized coherent detection assuming perfect channel state information can be achieved by means of noncoherent detection with moderate computational complexity, even in adverse fading channel scenarios.