By solving Poisson's equation it is shown that the reason for the failure of Evjen's method for evaluating crystal potentials is the neglect of a surface integral. A simple method for evaluating this surface integral is devised, and Evjen's method is extended to any lattice, in which the dipole moment of the unit cell vanishes because of symmetry. The method is applied to the NaCl, CsCl and Cu 2O structures, and Madelung potentials for Cu 2O are reported.