A Master of Science thesis in Electrical Engineering by Moustafa Farag Abdelnaby entitled, “Behavioral Modeling of RF Power Amplifiers Using Reduced Sampling Rate”, submitted in November 2018. Thesis advisor is Dr. Oualid Hammi. Soft and hard copy available.
The increasing demands for high data rates on wireless communication systems necessitate the enhancement of those systems in terms of software and hardware requirements. This creates real complications in designing power amplifiers (PA) that simultaneously meet the efficiency and linearity specifications of the modern wireless communication systems. Consequently, digital predistortion (DPD) is often used to compensate for the nonlinearity of the PA when operated at high efficiency. However, the increasing growth of the data rates and bandwidth requirements of the transmitter’s architectures are setting new burdens on the use of some of its hardware components, such as the analog-to-digital converters (ADCs) of the feedback path that is used to capture the PA’s output signal needed to derive the predistortion function. This thesis explores a technique suitable for extending the correction bandwidth of DPD systems for a given hardware specification. This approach is based on DPD model extraction from narrow-band measurements using under-sampling ADCs. The DPD is implemented using a two-box architecture which consists of the cascade of dynamic nonlinearity function followed by a static nonlinearity. The proposed method is validated in the presence of a 20-MHz long term evolution-advanced (LTE-A), which requires an ADC with a sampling rate of 100-Msps to capture the output of the PA due to the spectral regrowth behavior of the nonlinear amplifier. The output signal in this work is captured using an ADC with several under-sampling speeds that are reduced up to 75% of the full-rate requirements (i.e. 50-Msps, 40-Msps, 30-Msps, 25-Msps). Furthermore, the proposed approach also aims to reduce the software computational complexity of the typical DPD systems that involve high resolution delay alignment to be able to successfully extract the DPD coefficients. This is done by extracting the dynamic nonlinearity function in the two-box model using various delay mis-alignment conditions, that vary from -1 to +1 samples in steps of 0.1. Successful PA linearizations were achieved using the under-sampled signals from the ADC and dynamic nonlinearity predistorters that are extracted under delay mis-alignment conditions.