Numerical Validation

Since the cyclic ACF estimates can also computed using the DFT, we can compare with a completely different approach in [16] to validate our math. We used the cyclic ACF (e.g. equation 9.8) with $N=32$ and $M=3$. The results are shown in Figure 9.1. The two methods agree very closely. The differences are so small that they can be explained by differences in the stopping point of the iteration to find the saddlepoint. See software/test_quadspa.m.

Figure 9.1: Comparison of Saddlepoint Approximation from previous work (x-axis) with method of this section for cyclic ACF estimates. The largest difference was .021