Optical micro-structure array, including microlens array and pyramid array, has the function of integral imaging or diffraction beam-splitting. Careful measurement of the 3D profile of the array is a basic approach for insuring its quality. However, due to the limited numerical aperture of microscopy, when the surface is too steep, typically larger than 45 degrees, little light will be reflected or scattered back to the measurement equipment. The signal-to-noise-ratio will drop below the measurable threshold and information will be lost during measurement. In our case, the dihedral of the sample surface is 90 degrees. Intuitively, the reflected rays should be parallel to the incident rays after twice reflection and can be picked up by the detector. Nevertheless, the white-light interference microscope still showed no information on the 45- degree-inclined surface. In this paper, we study the twice-reflection of the dihedral angle of 90 degrees. We put it in the test beam of a spherical interferometer to simulate the situation in microscope. Simulation and real experiments suggest that the twice-reflection beam is of low spatial coherence and may act as the background intensity in white-light interferogram. This result cannot lead to a novel testing approach directly but points out the problem. We will sprout new idea based on it.