Causes, Identification and Compensation of Short-, Medium- and Long-Term Memory Effects in Power Amplifiers

Power amplifiers are continuously being pushed to their maximum operation capability in terms of efficiency, bandwidth and output power in modern base transceiver stations. Maintaining the output linearity in a power amplifier operated in these conditions is then addressed with predistortion, where the digital variant, DPD, has been the most employed solution. One of the challenges in DPD is to properly model and compensate the different types of memory effects present in power amplifiers. Therefore, the proposed talk will start by classifying the different types of memory effects, according to their time-scale, as short- medium- and long-term memory; covering the effects of the PA matching networks and such phenomena as electro-thermal and trapping, present in GaN HEMTs. The talk will then go over identification procedures that allow a clear observation of the presented effects and an evaluation of their impact in any particular power amplifier. Finally, this talk covers compensation strategies to address these different memory phenomena showing how a physical approach was essential to overcome the huge amount of coefficients required by the traditional – purely empirical – approach.