A modified simultaneous perturbation stochastic optimization algorithm for digital predistortion model extraction

In this paper, a novel stochastic approximation technique is presented as a low-complexity alternative to conventional least squares-based digital predistortion model extraction solutions. The proposed technique is based on the simultaneous perturbation stochastic approximation (SPSA) algorithm. It avoids the hardware-intensive matrix operations associated with least squares by using an iterative procedure to converge on a set of optimized predistortion model coefficients. The technique has been shown to achieve acceptable predistortion accuracy with greatly reduced hardware resource requirements. Linearization performance is evaluated in a scenario featuring a wideband LDMOS power amplifier (PA) excited by a long term evolution (LTE) signal.