A curve fitting and pole extraction algorithm has been developed and applied to exact frequency domain data for the surface fields on a perfectly conducting sphere. The data are fitted extremely closely, and for at least a handful of the lowest order SEM poles, the extracted poles and their residues are in good agreement with their known exact values. Unfortunately, this is not true if the frequency response is degraded in accuracy. In particular, noise effects are explored, and it is found that for noise levels typical of the best frequency domain experimental data it is no longer possible to locate more than (at most) the dominant SEM pole to a reasonable degree of accuracy.