Doctoral Thesis Complex Baseband Modeling and Digital Predistortion for Wideband RF Power Amplifiers

Modern modulation methods as used in 3rd generation mobile communications (UMTS) generate strongly fluctuating transmission signal envelopes with high peak-to-average power ratios. These properties result in significant distortion due to the nonlinear behavior of the radio-frequency power amplifier (RF PA). We propose different nonlinear model structures for such amplifiers, based on memory polynomials and frequency-domain Volterra kernel expansion, where we can reduce the number of free parameters by 80% compared to traditional Volterra series approaches. Because these nonlinear models incorporate memory, we are able to model the nonlinear distortion of RF PAs with sufficient accuracy (e.g., −30 dB relative modeling error ), including the wideband case (bandwidth B = 20 MHz as needed for four-carrier WCDMA). Furthermore, we propose a method to construct RF PA models from frequency-dependent AM/AM and AM/PM conversions. For the compensation of the nonlinearities, we analyze and simulate different digital predistorter structures in terms of complexity and linearization performance. As a result, memorypolynomial predistorters (7th order) can achieve a linearization performance comparable to the full Volterra predistorter, while the number of parameters is reduced from 42 to 11. Finally we propose a new predistortion scheme with low-rate system identification and Volterra kernel interpolation which allows a drastic cost reduction for the feedback ADC.