Frequency estimation variance with the Burg algorithm

The variance of the Burg frequency estimate is computed for the simplest case of a single noisy complex sinusoid through a Monte Carlo simulation over a useful range of data lengths, signal-to-noise ratios, and model orders. It is demonstrated that it is possible to provide a reasonable fit to this data by a relatively simple expression which provides considerable insight into the performance of the Burg algorithm, especially when used in conjunction with the expression for the Cramer-Rao lower bound, which is itself particularly simple in this case.<
>