Simplified Tone Reservation-Based Techniques for Peak-to-Average Power Ratio Reduction of Orthogonal Frequency Division Multiplexing Signals

Abstract

Orthogonal Frequency Division Multiplexing (OFDM) is one of the preferred modulation techniques for modern wireless communications networks, due to its high spectral efficiency and immunity to frequency selective channels. However, OFDM signals are known to suffer from a large peak-to-average power ratio (PAPR). OFDM signals with high PAPR values will inevitably be clipped by the power amplifiers (PA), causing signal distortion and out-of-band radiation, that would lead to the deterioration of bit error rate performance. This thesis focuses on a class of PAPR reduction techniques called tone reservation (TR) techniques, which possesses three desirable features, namely high PAPR reduction gain, no side information required at the receiver, and no in-band distortion. Clipping Control Tone Reservation (CC-TR) is an iterative TR-based technique that can achieve high PAPR reduction gain but at the cost of expensive computational time requirements. Therefore, this thesis aims to provide a novel TR-based technique that has reduced computational time requirements while maintaining a PAPR reduction performance that is the closest to the CC-TR. The proposed technique uses the Particle Swarm Optimization (PSO) to determine a predefined efficient set of 8 proper canceling signals for the TR technique, which significantly improves PAPR reduction gain and result in approximately 0.5 dB loss of PAPR reduction gain when compared to the conventional CC-TR. There are different generic classifiers available for selecting proper peak canceling tones and classifying high and low PAPR OFDM signal classes. We select the ANN for both tasks. First, the binary class ANN is applied to classify the input to low and high PAPR input. Then, the multiclass ANN is applied to select the canceling signal for the high PAPR input. As a result, this ANN model reduces the computational time of the proposed TR-PSO PAPR reduction technique further while maintaining the same PAPR reduction performance. Numerical results show that the proposed TR-based PSO with binary and multiclass ANN classifier can achieve the average accuracy of 98% and 95%, with its binary and multiclass ANN classifier modules respectively, while significantly declining the computational time by 98% for 60 data subcarriers.