A modified golden search algorithm for second loop control in a pilot-assisted feedforward power amplifier

A coordinate descent algorithm that is a modification of the multi-dimensional golden section search is proposed for controlling the second loop alignment of a pilot- assisted adaptive feedforward compensated power amplifier (FFPA). Symmetry of the cost function derived from log-power measurements of the residal pilot is exploited to improve convergence and dampen the search trajectory compared to a standard golden section search. A mathematical proof is provided that shows an improvement in convergence and dampening of 4.2 and 30.0 percent, respectively. Simulations show comparable results with improvements in convergence and dampening of 4.5 and 28.2 percent. Experimental results obtained using a commercially available feedforward power amplifier demonstrate the convergence of the algorithm.