Polynomial memory pre distortion based on non uniform delays for wireless applications

With increasing importance of spectral efficiency an RF power amplifier (PA) needs to be highly linear for the communication system to meet stringent spectral emission requirements. There has been extensive research in memory nonlinear modelling and memory pre-distortion algorithms. All known solutions represent the equivalent discrete baseband PA memory model with a finite impulse response filter (FIR). This model is limited to modelling the long term memory effect as it requires a large number of taps, resulting in slow convergence of the memory pre-distortion algorithm. We suggest a new approach using sparse delay taps in the pre-distortion model. This solution optimally represents the memory effect of the inverse transfer function of the transmission path using a limited number of delay taps. In comparison to solutions where all adjacent delay taps are taken, this solution finds a select subset of the adjacent delay taps, performing digital pre-distortion with significantly less resources, rendering optimal performance over all subsets with the given number of delay taps. Less resources are used as each added delay tap used to represent the memory effect requires additional HW processing, die size and power consumption.