A method for the numerical inversion of Laplace transforms (Q791306)

The authors describe Durbin's method with consists in approximating in an interval [0,2T] the unknown function by a trigonometric polynomial (with period 2T). They discuss the problem of proper balancing the discretization error (stemming from T being finite) and the truncation error (stemming from the trigonometric polynomial being a finite sum). By combining an asymptotic method of correction with techniques of convergence acceleration they find a way out of this dilemma. The paper contains a FORTRAN subroutine to treat the problem and numerical case studies with impressive results.