An Optimal Method for Evaluating a Class of Convolution Integrals with Related Kernels.

摘要

A FORTRAN 4algorithm is described for optimally evaluating certain classes of convolution integrals containing related kernel functions. Specifically,Fourier (sine,cosine) or Hankel (J sub 0), (J sub 1) transforms with related input kernels are evaluated efficiently,and without loss of accuracy,using an extension of a digital filtering algorithm previously published by the author. The optimal method is accomplished by use of special written external subprograms,and therefore,does not require any modifications to the existing filter algorithms. This method is similar to lagged-convolution algorithms,and may be used as an alternative for problems in which lagged-convolution is not directly applicable. A simple FORTRAN coding example is given illustrating the essence of the new algorithm.