Bandlimited Digital Predistortion of Wideband RF Power Amplifiers

The increase in the demand for high data rates has led to the deployment of wider bandwidths and complex waveforms in wireless communication systems. Multicarrier waveforms such as orthogonal frequency division multiplexing (OFDM) employed in modern systems are very sensitive to the transmitter chain nonidealities due to their high peak-to-average-power-ratio (PAPR) characteristic. They are therefore affected by nonlinear transmitter components particularly the power amplifier (PA). Moreover, to enhance power efficiency, PAs typically operate near saturation region and hence become more nonlinear. Power efficiency is highly desirable especially in battery powered and portable devices as well as in base stations. Hence there is a clear need for efficient linearization algorthms which improve power efficiency while maintaining high spectral efficiency. Digital predistortion (DPD) has been recognized as one of the most effective methods in mitigating PA nonlinear distortions. The method involves the application of inverse PA nonlinear function upstream of the PA such that the overall system output has a linear amplification. The computation of the nonlinearity profile and the inversion of the PA function are particularly difficult and complicated especially when involving wideband radio access waveforms, and therefore memory effects, which are being employed in modern communication systems, such as in Long Term Evolution/Advanced (LTE/LTE-A). In the recent technical literature, different approaches which focus on the linearization of specific frequency bands or sub-bands only have been developed to alleviate this problem, thereby reducing the complexity of DPD. In this thesis, we focus on the development and characterization of a bandlimited DPD solution specifically tailored towards the linearization at and around the main carrier(s) in single carrier deployment or contiguous carrier aggregation of two or more component carriers. In terms of parameter identification, the solution is based on the reduced-complexity closed-loop decorrelation-based parameter learning principle, which is also able to track time-varying changes in the transmitter components adaptively. The proposed bandlimited solution is designed to linearize the inband and out-of-band (OOB) distortions in the immediate vicinity of the main carrier(s) while assuming the distortions more far away in the spectrum are suppressed by transmit or duplex filters. This is implemented using FIR filters to limit the bandwidth expansion during basis functions generation and to restrain the bandwidth of the feedback observation signal, thus reducing the DPD sample rates in both the main path processing and the parameter learning. The performance of the proposed bandlimited DPD solution is evaluated using comprehensive simulations involving memoryless and memory-based PA models, as well as true RF measurements using commercial LTE-A base station and mobile device PAs. The achieved results validate and demonstrate efficient suppression of inband and OOB distortions in real-world application scenarios. Furthermore, the bandlimited DPD consistently outperforms the conventional DPD solutions in the memory-based PA model and practical PA scenarios in suppressing the OOB distortion in the immediate vicinity of the main carrier(s) by approximately 1 - 2 dB. The results provide sufficient grounds for the application of the bandlimited DPD solution in the classical single carrier deployment or in contiguous carrier aggregation of two or more component carriers where conventional DPD solutions would otherwise be highly complex