ERAD฀฀ ERAD฀฀2006 2006 Proceedings฀of Proceedings฀of Estimation of Doppler Spectrum Parameters: Comparison between FFT-based Processing and Adaptive Filtering Processing J. Figueras i Ventura 1 , M. Pinsky 2 , A. Sterkin 3 , A. Khain 2 , H.W.J. Russchenberg 1 1 International Research Centre for Telecommunications and Radar (IRCTR)-Delft University of Technology, Delft (The Netherlands). 2 Institute of Earth Sciences-The Hebrew University of Jerusalem, Jerusalem (Israel). 3 Weizmann Institute of Science, Rechovot (Israel). 1 Introduction A new method to estimate the Doppler spectrum parameters of atmospheric signals based on adaptive filtering is presented. The new method is compared with the traditional FFT-based method. In order to do so both methods have been applied to process the same atmospheric data set obtained by the FM-CW polarimetric Doppler radar TARA (Transportable Atmospheric Radar) developed by IRCTR. In section 2 the adaptive filtering processing is described. The details of the FFT-based algorithm will not be described. The reader is referred to Doviak and Zrnic (1993) for an extensive discussion. Section 3 describes the practical implementation of both algorithms in the radar. In section 4 the results of both algorithms are discussed and compared. Section 5 outlines the advantages and disadvantages of both methods. 2 Adaptive Filtering Processing 2.1 Signal and noise models We assume that the backscattered signal from atmospheric objects placed at a particular range has the form, after discretization, of a complex autoregressive series of first order Z t , with an unknown complex coefficient α: 1 1 * 2 t t i j ij Z Z ε Where t is discrete time, ε t is a normal white random series with variance σ ε 2 , α is a complex parameter characterizing the signal, * denotes conjugation, means averaging over the number of samples and δ ij is the Kronecker symbol. The spectrum of signal (1) is symmetric relative to the average Doppler frequency and can be represented by the following formula, Yaglom (2004): ( ) 2 2 1 z i F e ε ω σ ω α − = − (2) 2 2 2 1 z ε σ σ α = − (3) Where σ z 2 is the power (variance) of the signal Z t and ω is an angular frequency which lies within the range [-π, π]. A frequency equal to –π or π of the discrete signal corresponds to the Nyquist frequency of the continuous signal. When /α/→1, being less than 1, the signal can be approximated by a quasi-sinusoidal random series. As shown in Appendix A, the mean frequency of the spectrum, ω coincides with the frequency of the maximum. The standard deviation of the spectrum width depends only on /α/: t α ε ε ε σδ + = + = + (1) ( ) ( ) ( ) 1 2 2 2 1 arg 1 4 3 k k k k ω α π ω ω α + ∞ = =Λ= − Δ = (4) −Λ = − ∑ In order to estimate the mean Doppler frequency and the Doppler spectrum width only the estimation of α is required. Correspondence to: J. Figueras i Ventura. j.figueras@irctr.tudelft.nl We assume now that the atmospheric signal Z t is measured mixed with uncorrelated additive normal complex white