“…One of the most common iterative methods is the conjugate gradient method (CGM) [Van der Berg, 1984] that has guaranteed convergence in a finite number of steps when applied to continuous operators. The inclu-sion of discretization processes, the intrinsic ill conditioning of convolution operators (worsened by the fact that CGM squares the condition number of the operator), the round-off errors, and some other reasons make the CGM in the practice converge slowly or even diverge when the problem is difficult enough (i.e., in some cases of coaxial-fed phased arrays [Cuevas and Sierra, 1989] Berg, 1984Berg, , 1989]. For these two choices the operator LA is self-adjoint regardless of L and the GCGM may be used.…”