A computer model using finite ray tracing methods was developed to simulate a videokeratoscope analysing an average cornea. Different faceplate designs were tested using five points in the faceplate subtending angles between 15° and 75° in 15° intervals at the corneal vertex. Image quality was assessed by adding the geometrical blurs of the 5 image points; the best focal plane was considered the one that gave the smallest sum of the blurs. Surfaces were divided into five categories according to their blur sums and a representative of each group selected. Differences (error) between accurate sagittal radius of curvature and radius of curvature calculated by the van Saarloos algorithm were calculated for the selected surfaces at the same corneal points. The calculations were repeated for the tangential radius of curvature. Differences equal or bigger than 0.02 mm were regarded as clinically significant. The surface that provided the sharpest image for an average cornea was a cylinder with the base 120 mm away from the corneal vertex and a diameter of 26 mm. Changing the faceplate design results in clinically significant differences for an average cornea. Focusing errors have more influence than aberrations (controlled by instrument pupil aperture).