Angular speckle correlation is a method of surface-roughness measurement about amplitude parameter independent on spacing characteristic of the surface roughness profile. The mathematic model of this method is usually based on ensemble average. Through simulating the speckle intensity distribution in the far field of a random rough surface, we compute the degree of angular speckle correlation by substituting spatial average in speckle pattern for conventional ensemble average, and determinate the surface roughness by the mathematic model of ensemble average. The study proves the applicability of this angular speckle-correlation method using spatial average to roughness measurement. Enough measuring precision can be got, as long as a few parts of area on the same surface are measured and the results are averaged. For surfaces with a mean root square error Rq larger than 2 μm, the relative measuring error is less than 15%. According to the best measuring criterion, this method fits the surface with large roughness.