Abstract-The generalized forward-backward and novel spectral acceleration (GFB/NSA) method is applied to capacitance extraction problems of finite planar periodic structures. In the GFB method, the interaction within a unit cell can be calculated and stored beforehand. The interactions between relatively far-separated unit cells are however calculated by the GFB/NSA method to further accelerate the calculation speed. The contributions to a receiving element on finite planar periodic structures are separated into weak and strong source contributions by an appropriate separation index, which is conveniently specified by an amount of unit cells rather than a distance. The strong source contribution is performed by the standard matrix-vector multiplication in the GFB method, while the weak source contribution is computed using the NSA algorithm. Numerical examples show comparisons of the GFB/NSA method with a commercial software, including the efficiency of the method. With the array increment in one direction, the GFB/NSA method shows O(N ) in the calculation time per iteration, while its memory requirement for a very large problem also tends to be O(N ), where N is the number of unknowns.