Digital Linearization of a Receiver Optimized for Radar

Abstract

Linearity is an important property of radio frequency amplifiers that has implications for performance in a wide range of applications, from communications to radar. As digital hardware grows ever cheaper and more power efficient, digital techniques for improving the linearity of amplifiers have emerged as an effective replacement for older more power-hungry analogue methods. This thesis seeks to investigate the effectiveness of digital linearization techniques for post-distortion on the receiver side, and in the context of digital radar systems. The investigation has been performed by testing linearization algorithms on recorded signals using hardware typical of modern radio receiver systems. Triangular chirp signals and combined two-tone signals using frequency pairs based on coprime integers are shown to be effective calibration signals. Post-distortion techniques using either memoryless polynomials, memory polynomials or generalized memory polynomials are shown to suppress intermodulation distortion by up to 20dB, and to remain stable for a temperature drift of about 10C. The coefficient estimation algorithm is shown to find inverse models that can improve linearity using only a small number of samples, indicating the possibility of implementations using a low amount of digital resources and model complexities. Ways of compensating for temperature drift have been investigated but to inconclusive results.