Abstract: Marple algorithm is one of the most effective methods for the autoregressive spectrum analysis It yields AR spectra with no apparent line splitting reduced spectral peak frequency estimation biases and higher resolution But the criterion for AR order determination still has not been resolved.
Two ill-conditions and two tolerance factors proposed by Marple for his new recursive algorithm have been discussed from the view of experimental in this paper. The resolution in AR spectra versus AR order has been studied with various noise levels. The problem of spurious spectral peaks was also discussed.

Key words: Autoregressive spectrum analysis, LS algorithm, Resolution