Low Power Pre-Distorter Design For 5G Radio Using Machine Learning

Sammanfattning: A Power Amplifier (PA) is an essential electronic component in all microwave and millimeter-wave applications and, more specifically, in any transmitting system where the level of input power signal needs amplification to the desired level.Linearity and high efficiency are of utmost importance in PAs. However, high-efficiency PAs tend to be non-linear, and PAs working in the linear region might have low efficiency. Hence, there is always a trade-off between efficiency and linearity while designing a PA. For an efficient system design, the efficiency of the PA gets prioritized by the designers, and for linearity, an additional linearization technique can be deployed. Designers have been considering many linearization methods. Among those, digital predistorter tends to be the most popular one as it can provide a right amalgamation between linearity performance and implementation complexity. However, the computation process used to obtain an inverse PA behavior inside a digital predistorter consumes significant power. In this thesis, the main target is to find a power-efficient way to enhance the current algorithm for the digital pre-distorter (look-up-table based) and evaluate the power results along with Adjacent Channel Power Ratio (ACPR)