Intuitively, the point spread function (PSF) of a computed tomographic (CT) system can be determined by the reconstruction of a microphantom (diam<1 mm), such as a microbead for 3-D PSF or a thin wire for 2-D PSF (or line spread function). In fact, this direct method has its own limitations and drawbacks. First, the microphantom (not necessarily infinitesimal) is difficult to be precisely manufactured. Second, the reconstructed volume (containing the micron phantom image) is prone to the uncertainty in digital representation due to the aperture effect associated with grid cells (pixels for 2-D grid or voxels for 3-D grid). Third, the reconstructed digital volume of a microphantom is both material dependent and size dependent, and so is the PSF; this is not convenient for CT system evaluation. We reveal the pitfalls associated with the microphantom-based cone-beam CT PSF measurement by theoretical explanation and experimental demonstration. To avoid the pitfalls, we are in favor of the edge-based techniques for PSF measurement through the use of a macroedge phantom (diam>1 mm). In conclusion, the CT PSF can be efficiently determined by an edge-based technique, such as the iterative edge-blurring algorithm.