Digital Predistortion for Power Amplifier Linearization in Radar-Communication Shared Apertures

Abstract

Shared radar and communication apertures are commonly optimized with radar performance as the primary design objective. As a result, transmitter operating conditions such as output power and amplifier efficiency are often chosen in a way that is sub-optimal for communication performance. In particular, operating a power amplifier close to compression improves efficiency and radar signal-to-noise ratio, but introduces nonlinear distortion that degrades communication metrics. This thesis investigates the use of digital predistortion as a software-based linearization technique for communication signals transmitted through commonly utilized radar power amplifiers. Two amplifier models are considered: a memoryless Saleh model representing a traveling-wave tube amplifier (TWTA), and a generalized memory polynomial (GMP) model representing a solid-state power amplifier (SSPA) with nonlinear memory effects. Four communication waveforms are evaluated: two single carrier waveforms and two orthogonal frequency-division multiplexing (OFDM) waveforms. Digital predistortion models are trained using an indirect learning architecture and assessed in terms of linearization performance, model complexity, robustness to noise, and sensitivity to amplifier mismatch. The results show that digital predistortion can significantly reduce nonlinear distortion without requiring changes to the radio frequency hardware. For the memoryless TWTA model, a low-complexity memoryless polynomial predistorter provides sufficient linearization performance. For the memory-dependent SSPA model, memory depth is the dominant factor affecting performance, especially for OFDM-based waveforms. Polynomial order and cross terms provide additional improvements, but with diminishing returns. Further analysis also shows that the predistortion models are robust to moderate noise and amplifier variations, although the SSPA model case is more sensitive to nonlinear memory mismatch. Overall, the thesis demonstrates that digital predistortion is a promising approach for improving communication performance in shared radar-communication transmitter systems.