Iterative DPD based Multi-Objective Optimization of RF Power Amplifier for Increased Linearity and Output Power

The viability of any communication link hinges critically upon the performance of High Power Amplifiers (HPAs). The inherent nonlinearity of HPA causes in-band distortion (i.e. elevates Error Vector Magnitude (EVM)) and out-of-band distortion (i.e. increases Adjacent Channel Power Ratio (ACPR)). To mitigate these effects, HPAs are undesirably operated at a 3 dB back-off. To this end, iterative Digital Pre-Distortion (DPD) techniques are employed to compensate for amplifier's nonlinearities and allow it to operate with either minimum distortion or maximum RF output power. Previous literature reports optimization of any one these variables at a time. It also reports simultaneous optimization of maximum output power and minimum ACPR for a fixed or manual variation of degree of nonlinearity and memory polynomial. Auto-tune techniques to optimize the efficiency while complying with a certain linearity constraint are also reported (i.e. for a fixed order and memory length). In this work, we introduce a multi-objective optimization algorithm that adaptively fine-tunes iterative DPD by modifying RF output power, degree of nonlinearity and memory polynomial. This process allows for simultaneous optimization of RF power, linearity and EVM. Simulation results demonstrate that, for a 100MHz 64-QAM input signal to a MATLAB white-box PA model, iterative DPD alone achieves an RF power of 37 dBm, ACPR of -37 dBc and RMS EVM of 1.2%. Our algorithm, by simultaneously fine tuning the adaptive DPD model (varying order and memory length), improves ACPR to -42.34 dBc and RMS EVM (1%) while maintaining the RF output power at the predefined 37.5 dBm level.