Ground-Based DWL Laboratory: SNR Studies

In processing the single shot data, VADS were used to identify a sequence of range gates where only noise existed. Shot-by-shot average amplitudes of the signals are computed for range gates 17 through 27. The average noise threshold is around 5.25 in arbitrary units. A velocity azimuth display (VAD) displays the single shot data filtered along each shot by the average noise found in range gates 17-27. Using a VAD sine fitting model, vertical soundings were computed.

Single shot wind velocity estimates were analyzed per range gate for all azimuth angles. In this particular scenario, gates 8 through 15 were observed to have a well defined, coherent single shot returns. Thus, a least squares sine fit routine was applied to the poly-pulse pair data in each of these eight gates, and the resulting sine fit was used as the representation of the true wind velocity in each gate. In this way, the difference between each single shot wind estimate in gates 8 through 15 and the corresponding poly-pulse pair sine fit was calculated in order to analye the accuracy of single shot estimates. The SNR value corresponding to all single shot estimates was also determined. From this information, the standard deviation of single shot estimates from poly-pulse pair "truth" was obtained for a range of SNRN values. We computed the percentage of total single shot wind velocity estimates per SNRN value which were less than the threshold values of 5 m/s, 4 m/s, 3 m/s, 2 m/s and 1 m/s. The trend was as expected, i.e., better estimates were obtained at higher SNRN. Additionally, obtaining accurate single shot estimates to within 1 m/s 50% of the time an SNRN value of about 2.5 dB is necessary. Likewise, to obtain accurate estimates to within 5 m/s 50% of the time an SNRN value of approximately -2 dB is needed. This verified generally held expectations that an SNR (narrow band) of about 0 dB is the limit in obtaining "good" single shot estimates about 50% of the time. Finally, the "rule of thumb" that confidence in single shot wind velocity estimates rapidly decreases for SNRN values below 5 dB is verified.