Réduction de moitié du volume mémoire et de la durée de calcul dans l'usage des algorithmes de transformation de Fourier rapide (FFT)

Abstract

After a brief review of the crystallographic Fourier transform problem, this paper compares three methods for its computer implementation with reference to core-storage capacity and computation time: the classic method, the simple Fast Fourier Transform algorithms and a new method. The first method needs, especially for a macromolecular three-dimensional structure, too many operations and therefore too long a computation time and poor accuracy. The remarkable FFT algorithm, on the other hand, leads to large savings in computation time with a better accuracy, but using no crystallographic symmetry wastes too much storage capacity. The third method, using FFT but taking into account Friedel's law by means of a simple in situ process, approximately halves the storage capacity as well as the computation time.