Neural DPD

Technical Explanation

Radio frequency power amplifiers are most energy-efficient when driven close to their maximum saturation power, but this introduces severe non-linear distortion that splatters energy into adjacent channels and distorts constellations. Neural Digital Pre-Distortion (DPD) models the amplifier's non-linear memory effects using deep neural networks and pre-distorts the input signal in reverse, canceling amplifier distortion and boosting power amplifier energy efficiency by 30-50%.

Key Functions

• Deep neural network modeling complex non-linear power amplifier memory effects
• Pre-inverting signal distortion allowing power amplifiers to operate near peak saturation
• Improving power amplifier energy efficiency by 30-50% in mobile base stations
• Suppressing adjacent channel leakage ratio (ACLR) to meet strict regulatory emissions masks
• Real-time adaptive retraining compensating for amplifier thermal heating and aging