Predistortion of OFDM Waveforms using Guard-band Subcarriers

Digital predistortion (DPD) is an important technique that is commonly used in wireless transmitters to reduce the out-of-band emissions caused by power amplifier (PA) nonlinearities. This frequency-domain goal is generally achieved by learning a baseband equivalent time-domain inverse transfer function of the PA and applying it to the transmitted digital baseband signal. In this work, we take advantage of the frequency-domain nature of orthogonal frequency-division multiplexing (OFDM) signals by injecting cancellation tones into the guard-band subcarriers to perform DPD. We experimentally evaluate our OFDM-based DPD (ODPD) method using a Doherty PA and show that, when compared to a standard generalized memory polynomial solution, the ODPD can achieve better suppression of out-of-band emissions with lower complexity. Moreover, when combined with a neural network model of the PA, our proposed ODPD method only requires oversampling the transmitted signal by a factor of two, which has important implications for the analog transceiver front-end.