Piecewise companding transform is a flexible and efficient way to solve the high peak-to-average power ratio (PAPR) problem for orthogonal frequency division multiplexing (OFDM) systems. A novel threshold-based piecewise companding transform is proposed in this paper. Based on the statistical characteristics of amplitudes, OFDM signals are classified into three groups (i.e., small, average and large signals). Different from conventional approaches, two dedicated designed thresholds are set to amplify the small signals and compress the large signals, respectively. Simulation results verify the improvement in PAPR reduction of the proposed scheme. Moreover, a lower bit error rate (BER) performance loss can be obtained by introducing the iterative detection with a moderate increase in complexity. Piecewise companding transform is a flexible and efficient way to solve the high peak-to-average power ratio (PAPR) problem for orthogonal frequency division multiplexing (OFDM) systems. A novel threshold-based piecewise companding transform is proposed in this paper. Based on the statistical characteristics of amplitudes, OFDM signals are classified into three groups (ie, small, average and large signals). Different from conventional approaches, two dedicated designed thresholds are set to amplify the small signals and compress the large signals, respectively. verify the improvement in PAPR reduction of the proposed scheme. Moreover, a lower bit error rate (BER) performance loss can be obtained by introducing the iterative detection with a moderate increase in complexity.