Length-Variable Prototype Filter Design for the PAPR-Reduced FBMC System

To flexibly design prototype filters for the PAPR-reduced FBMC system, i.e., the pruned DFT spread FBMC system, this paper proposes a weighted cost function considering perfect reconstruction and out-of-band energy. For minimization, the original fourth-order function is transformed into a second-order function, which is then optimized by an iterative least-square (ILS) algorithm. In this proposed method, the filter length is variable and no more limited as in OQAM-FBMC. Simulation results demonstrate the effectiveness of the method. Performance tradeoffs can be aroused by changing the weighting factor and the filter length. Consequently, the proposed method can be used to design filters with comparable performances to its competitors. It is also possible to design longer filters with better BER or shorter filters that perform better in PAPR, system latency and complexity. In our designs with filter length no more than the subcarrier number plus three, the system BER and PAPR are improved. Meanwhile, the system complexity and latency caused by the prototype filter are reduced. To a certain extent, due to the variable filter length and weighting factor, the proposed method gives the flexibility to design filters for different performance needs.