Wide-band Linear Characteristic Compensation for DPD Systems with Direct Learning

Direct learning with ideal PA function assumption is an efficient method for adaptive DPD because the information of PA is not needed. However, high efficiency PA design results in worse amplitude response and group delay characteristic within the linearization bandwidth. Then underlying assumption may be not hold, and the direct learning is divergent. In the paper, we analyze condition for convergence using the direct learning with ideal PA assumption. Four schemes are proposed to compensate wide-band linear characteristic when condition is violated. Simulation results and experimental tests confirm the analysis and the effectiveness of proposed algorithms.