Noncoherent and Coded OFDM Using Decision-Feedback Demodulation and Diversity

Multi–carrier techniquesare an attractive alternative to single–carrier systems when transmitting over frequency– selective channels. For example, o rthogonal frequency–d ivision multiplexing (OFDM) for digital audio broadcasting [1], dig ital video broadcasting [2], and HIPERLAN [3] are successful ly proposed and standardized multi–carrier transmission sys tems in Europe. The American National Standardization Institut e has selected discrete multitone (DMT) transmission for asymmetric digital subscriber lines [4]. Currently, OFDM is discussed as modulation scheme for power–line communications providin g high–speed network access, e.g. [5]. The main advantage of multi–carrier transmission compared to single–carrier systems is the low complexity of channel e qualization, which is even more manifest if differential encodi ng and noncoherent detection are applied. Then, explicit channel estimation, which usually is based on pilot symbols, and equaliz ation can be circumvented. Generally, the price to be paid for robust and low complex noncoherent detection is a performan ce loss against coherent receivers, which strongly depends on the characteristics of the underlying transmission channel. In this paper, we concentrate on OFDM schemes, where OFDM symbols consecutive in time are supposed to be separatelydetected. This is for example necessary in multi–user systems applying time–division multiple access. To utilize diversity channel coding in combination with bit–interleaving acros s the OFDM subcarriers is regarded. A good trade–off between coding gain and complexity is given by convolutional codes and maximum–likelihood decoding with the Viterbi algorithm. W e will consider the possibility of using a further degree of di versity by transmitting the coded information over several indepen dent fading channels, which in a practical system can be achieved e.g. by frequency hopping or multiple–antenna receivers. T his becomes necessary if unfavorable short–time fading condit i s lead to severe performance degradation. Since for the descr ib d situation coherent detection requires pilot symbols for ch annel 1Corresponding Author estimation ineachOFDM symbol, the overhead for coherent receivers can be regarded as prohibitive. Hence, reception wi thout channel state information is recommendable. For the scenario explained above we adopt a simple receiver structure proposed in [6], [7]. There, power efficiency of noncoherent M–ary differentially encoded p hase–s hift keying (MDPSK) transmission over flat Rayleigh fading channels is considerably increased by enlarging the observation inter val of noncoherent reception to N > 2. Specifically,hard decisions of the Viterbi algorithm are fed back in an iterative decodin g procedure. This d ecision–f eedback d ifferential demodulation (DF–DM) can be regarded as analogous to d ecision–f eedback differential detection (DF–DD) for uncoded transmission [8], [9]. Though hard decision feedback is clearly suboptimum, i t has shown sufficient rate of convergence and substantial per formance gains with only a very moderate increase in complexity compared to conventional differential detection with N = 2 [6], [7]. Noteworthy, there are a number of other known iterative decoding schemes for MDPSK transmission, e.g. [10], [11], [12 ], [13]. There, however, computational complexity increases exponentially withN . In this contribution, we extend the approach of [6], [7] to OFDM over frequency–selective fading channels. Whereas in [6], [7] high diversity has been achieved by sufficient bit – interleaving, for OFDM this strategy is only partly suitabl e. Instead, we quantify the influence of different degrees of dive rs ty on system performance. First, the transmission model is int roduced. Then, the metric calculation for DF–DM and the iterat ive decoding algorithm are given. For evaluation of the propose d noncoherent receiver simulation results of OFDM with param eters according to established systems [1], [2], [3] are disc us ed.