A method of enhancing the performance of a numerical FM discriminator in specific receive signal conditions, using frequency feedback techniques, is developed and evaluated. The receive signal conditions of interest are: (1) frequency deviations that are large relative to the bandwidth of the message signal; (2) a low-uniform (complex) sample rate commensurate with the bandwidth; and (3) noise strengths that yield values of input carrier-to-noise ratio of the order of 10 dB or more. Through computer simulation, observed phase cycle-slip counts and phase noise variance estimation are used to compare the enhanced demodulator to the performance of a numerical FM discriminator baseline. Based upon the theoretical analyses and simulation results, the presented method of numerical FM demodulation is shown to be a viable technique for enhancing the performance of numerical FM discriminators under the given modulation conditions.