Estimates of Large Spectrum Width from Autocovariances

Abstract The authors demonstrate that there are maximum measurable (saturation) spectrum widths for standard autocovariance techniques, the 0,1-lag autocovariance estimator and the 1,2-lag estimator. The maximal mean measurable spectrum widths from the two estimators depend on the number of samples and are substantially lower than the Nyquist velocity. Furthermore the maximal mean spectrum width of the 1,2-lag algorithm is approximately 2 times smaller than the maximum mean width of the 0,1-lag estimator. Simulated signals, solar noise, and weather signals are processed to verify theoretical predictions.