Frequency-Selective Digital Predistortion for Unwanted Emission Reduction

In this paper, we present a novel digital predistortion (DPD) solution based on a direct learning approach, which is capable of reducing the unwanted emissions resulting from the power amplifier (PA) at any prespecified frequency located in the transmitter's out-of-band or spurious domain. The proposed scheme is based on evaluating the power spectral density (PSD) of the PA output signal and optimizing the DPD coefficients iteratively in order to minimize the output PSD around the prespecified frequency. To highlight the feasibility of the proposed implementation, the predistortion processing is kept as simple as possible, deploying quasi-memoryless polynomial models. Efficient mitigation of unwanted emissions around the target frequency is demonstrated via simulations and actual RF measurements, in both single- and dual-carrier waveform scenarios, using memoryless and memory-based PAs. The proposed DPD solution could be potentially employed in applications such as mobile devices utilizing noncontiguous multicarrier transmission, where the intermodulation spurs may overlap with the device's own receiver band, or could be potentially violating the spurious emission limits. Another target application is cognitive radio, where the PA may produce unwanted emissions that are interfering with primary-user transmissions. To the best of the authors' knowledge, there does not exist a similar technique in the open literature, and thus, the purpose of this paper is to encourage scientific discussions and technological innovations toward the creation of relatively low-complexity frequency-optimized predistortion techniques employed against selected unwanted emissions produced by the transmitter.