This paper tackles the sum-rate maximization in FBMC systems. It is well-known that multipath fading destroys the orthogonality of the subcarriers, which brings a non-convex problem that can be interpreted as the maximization of a difference of concave functions. This reformulation enables using DC programming to find the global optimal but the complexity becomes unaffordable in the FBMC context. Aiming at reducing the complexity we have devised a new suboptimal algorithm. The idea consists in smartly partitioning the band into three subsets so that subcarrier signals belonging to the same subset do not overlap in the frequency domain. If subsets maximize sequentially their own rate the original problem is turned into three simpler problems. Simulation-based results show that the sum-rate is in the same order as the water-filling used in OFDM and a suboptimal solution based on DC programming. However, the addressed technique enables saving up to 25% of the power at SNR=22 dB.

Peer Reviewed