Adaptive Linearization of Power Amplifiers in Digital Radio Systems

High-frequency power amplifiers operate most efficiently at saturation, i.e., in the nonlinear range of their input/output characteristics. This phenomenon has traditionally dictated the use of constant envelope modulation methods for data transmission, resulting in circular signal constellations. This approach has inherently limited the admissible data rates in digital radio. In this paper we present a method for solving this problem without sacrificing amplifier power efficiency. We describe and analyze an adaptive linearizer that can automatically compensate for amplifier nonlinearity and thus make it possible to transmit multilevel quadrature amplitude modulated signals without incurring intolerable constellation distortions. The linearizer utilizes a real-time, data-directed, recursive algorithm for predistorting the signal constellation. Our analysis and computer simulations indicate that the algorithm is robust and converges rapidly from a blind start. Furthermore, the signal constellation and the average transmitted power can both be changed through software.