On the performance of soft LLR-based decoding in time-frequency interleaved coded GFDM systems

Generalized Frequency-Division Multiplexing (GFDM) is a candidate modulation for future wireless cellular networks. The main reason stands in the flexibility of its structure that could fulfill ideas and challenges of forthcoming network scenarios. A known issue of GFDM is its worse performance compared to orthogonal frequency-division multiplexing, which is due to the interference among transmitted symbols. A mathematical model of such an interference has been proposed in a recent paper by exploiting the parallelism that exists between GFDM and discrete Gabor transform. The model allows for the design of different types of linear and non-linear equalizers. With the goal of increasing the transmission reliability, in this paper the introduction of channel coding is considered together with an appropriate interleaving. The computation of the Log-Likelihood Ratio (LLR) is described, which allows for soft decoding in the case of maximum likelihood and linear minimum mean squared error detection. The gain in performance achieved with channel coding and time-frequency interleaving is demonstrated by means of Monte Carlo simulations for the standard 64-state rate-1/2 convolutional code. A comparison with an approach for soft decoding proposed in the literature for GFDM is also reported.