A non-collinear descent algorithm for controlling the first loop in a feedforward power amplifier

An algorithm for controlling the first loop of a feedforward amplifier is discussed. The ratio of the error and input powers is used to measure the residual of the carrier cancellation and the loop alignment is adjusted to minimize this residual. Three successive non-collinear measurements are used to choose the descent direction, which is close to the gradient direction of the cancellation contours without being collinear with the two most recent measurements. Quantization of the descent direction simplifies the implementation. The step size of the descent is decoupled from the gradient magnitude to ease implementation and improve stability.