Use is made of five sets of multibeam observations of the lower atmosphere
made by the Indian mesosphere-stratosphere-troposphere (MST) radar. Two asp
ects of signal processing which can lead to serious underestimates of the s
ignal-to-noise ratio are considered. First, a comparison is made of the eff
ects of different data weighting windows applied to the inphase and quadrat
ure components of the radar return samples prior to Fourier transformation.
The relatively high degree of spectral leakage associated with the rectang
ular and Hamming windows can give rise to overestimates of the noise levels
by up to 28 dB for the strongest signals. Use of the Hanning window is fou
nd to be the most appropriate for these particular data. Second, a techniqu
e for removing systematic de biases from the data in the time domain is com
pared with the more well-known practice of correction in the frequency doma
in. The latter technique, which is often used to remove the effects of grou
nd clutter, is shown to be particularly inappropriate for the characteristi
cally narrow spectral width signals observed by the Indian MST radar. For c
ases of near-zero Doppler shift it can remove up to 30 dB of signal informa
tion. The consequences of noise and signal level discrepancies for studies
of refractivity structures are discussed. It is shown that neither problem
has a significant effect on Doppler shift or spectral width estimates.