A novel method to perform adaptive memoryless polynomial digital predistortion

摘要

This paper presents a novel procedure to dynamically calculate the polynomial coefficients for memoryless digital predistortion (DPD) of power amplifiers (PAs) in mobile handsets. The basic idea is to calculate the adjacent channel interference during the operation time and to adapt the DPD coefficients in order to minimize the spectral regrowth. This is different to state of the art approaches, which use indirect learning methods to estimate the inverse PA behavior without taking into account the signal statistics or the spectral interference. This paper shows the details of the new concept and the comparison with state of the art DPD when both are applied to a CMOS PA. Measurement results of high crest factor (CF) signals show the drawbacks of the indirect learning approach at power levels close to compression, whereas the new method is capable to achieve significantly better performance over the whole power range. Using in both cases a 3^rd order memoryless polynomial as predistorter, it is possible to obtain an improved linear output power of 3.8 dB, while only 1.9 dB are achieved with the state-of-the-art approach.