Fig. 7.

Reconstruction accuracy [with error labeled $d$ as defined in Eq. (1)] comparisons between different inversion methods with different noise models. (a) The noise-free simulations where the optimized iterative method performs better. (b) The performance of each algorithm with added noise similar to the all-digital system. (c) is identical to (b) except noise similar to the network analyzer system was used. The LUT method (green) used a $500\times 500$ point table shows slightly higher error on average than the iterative method. For single frequencies, we also show the error when an HD-LUT consisting of $2000\times 2000$ points was used. For a single modulation frequency, the inversion speed of the HD-LUT is $\sim 5\times$ faster than the default iterative method.