Performance analysis of maximum likelihood spectral estimator compared with PM estimator

The effect of the wideband signal-to-noise ratio (R_SNW) and the effective power spectral width on the Doppler frequency estimation performance using maximum likelihood (ML) estimator have been studied. The performance of the ML algorithm is applied to simulation signal. Experimental data are described by the detection probability and the standard deviation (SD), and also compare with periodogram maximum (PM) estimator. The data are acquired by 1.5 μm all fiber coherent Doppler lidar in Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences. The results indicate that ML's SD can enhance 0.5 MHz higher than PM's under the same spectral width. ML is close to PM under the narrow spectral width. To satisfy the 90% detection probability, the required R_SNW is 2 dB smaller than PM's. In the real-process, ML's SD can be 1.1 MHz lower than the PM, and the detection probability is 9% lower than PM. To obtain the wind speed accuracy less than 1 m/s while the detection probability is above 80%, the required R_SNW is larger than -14 dB.