Kang Guixia

A Stable Recursive Algorithm for Memory Polynomial Predistorter

The memory polynomial model has been used for predistorter design, which will result in numerical instabilities when higher order terms are included. In this paper, basing on least square estimation, a recursive algorithm for indirect learning structure predistorter is introduced. Simulation results show that higher order (higher than 7th-order) memory polynomial model can not bring significant improvement on predistortion effect and modeling accuracy, but brings much more computational burden. So it is unnecessary to choose a higher order polynomial predistorter, and the algorithm will always be stable. Results show that even 8th-order memory polynomial predistorter is used, the algorithm is convergent within 10 iterations, and can improve outband spectrum of 20 MHz bandwidth signal by 30 dB, with a 2×1010 matrix condition number.