A practical method of designing windows by usingC-spline functions is proposed. Given sidelobe fall-off controls the frequency characteristics at higher frequencies, while desired characteristics and specified error tolerance control those at lower frequencies. The optimal dimension is determined as the minimum to satisfy the tolerance of error in approximating the desired characteristics in the logarithmic scale. The effectiveness of this method is demonstrated by several design examples superior to Kaiser-Bessel, gaussian, Hanning, Blackman, Parzen or cosxxwindows, respectively.