19-10-2012, 01:08 PM
Clutter Filtering and Spectral Moment Estimation for Doppler Weather Radars Using Staggered Pulse Repetition Time (PRT)
Clutter Filtering and Spectral.pdf (Size: 159.84 KB / Downloads: 58)
ABSTRACT
In this paper, a new algorithm for the estimation of spectral parameters from the signal time series, collected
using the staggered pulse repetition time (PRT) transmission in a Doppler weather radar, is presented. The
algorithm uses the Fourier transform and a magnitude deconvolution procedure to reconstruct the signal spectrum,
and then the spectral parameters are estimated from the reconstructed spectrum. There is a significant improvement
in the variance of the spectral parameter estimates compared to previously published methods of processing
staggered PRT sequences. Further, a novel spectral domain clutter filtering procedure allows 1) accurate velocity
estimation even if the clutter-to-signal power ratio is as high as 40 dB and 2) does not incur the loss of velocity
information in certain Doppler bands experienced by other clutter filtering techniques. With this algorithm, the
staggered PRT technique becomes a practical contender for implementation on Doppler weather radars in the
quest to increase the unambiguous velocity and range.
Introduction
One of the long-standing problems affecting Doppler
weather radar performance has been the velocity and
range ambiguity. For a uniform pulse repetition time
(PRT) transmission, the unambiguous range ra and velocity
y a, are governed by the equation y ara 5 cl/8,
where c is the speed of light and l is the wavelength.
Further, in Doppler weather radars, velocity ambiguities
can occur in the presence of overlaid echoes. In such
cases, the overlaid signals must be separated prior to
determining the proper spectral moments of each signal.
Several schemes to ameliorate the ambiguity problem
have been proposed. Some of the important ones are (a)
polarization coding (Doviak and Sirmans 1973), (b) random
phase technique (Zrnic´ 1979; Laird 1981; Siggia
1983; Zrnic´ and Mahapatra 1985), © systematic phase
coding (Sachidananda and Zrnic´ 1997), (d) staggered
PRT technique (Zrnic´ and Mahapatra 1985), and (e) the
method using two sampling rates (Sirmans et al. 1976).
Processing procedure
In the proposed new approach we seek to reconstruct
the spectrum of the weather signal from the staggered
time series samples, that is, generate a time series with
a uniform sampling period of Tu, and then estimate the
spectral parameters from this reconstructed spectrum.
This procedure allows estimation of the spectral moments
with a much lower variance than the earlier methods.
Further, a novel scheme for clutter filtering in the
spectral domain is proposed that can achieve clutter suppression
in excess of 40 dB and almost complete elimination
of all spurious rejection bands in the 6y a interval.
The result is a nearly linear phase response of
the clutter filter, which has been eluding researchers so
far. This is the most important feature of the clutter
suppression scheme; it makes practical the use of staggered
PRT in weather radars.
Ground clutter filtering
The clutter filtering is carried out in the spectral domain
before the signal spectrum is reconstructed using
the deconvolution procedure described in section 3a. To
prevent clutter spreading in the frequency domain and
thus achieve better clutter rejection, the time series data
must be multiplied by a suitable window function (von
Hann window is used here). The spectrum of the ground
clutter is assumed to have narrow width and is centered
on zero Doppler velocity. Because the spectrum of the
derived time series Vk is the convolution of the code
spectrum Ck and the signal spectrum Ek , it will have
weighted replicas of the clutter spectrum centered on
each of the nonzero spectral lines of the code spectrum.
The weights are the spectral coefficients of the code
spectrum. An example of the convolved spectrum is in
Fig. 4c for k 5 T1/T2 5 Å. The method (to be described)
is not confined to k 5 Å; this stagger ratio is used only
as an example to illustrate the clutter-filtering procedure.
Because the code spectrum in this case has five nonzero
spectral coefficients (see Fig. 1b), the power from each
of the clutter spectral coefficients is spread over five
spectral coefficients separated by N/5 coefficients in the
convolved spectrum.
Conclusions
A new spectral domain processing technique is presented
for filtering the clutter and estimating the spectral
parameters of a weather signal; the scheme is meant for
Doppler weather radars that transmit a staggered PRT
sequence to mitigate velocity and range ambiguities.
Although we have demonstrated it on a staggered PRT
sequence, our new procedure of both spectral reconstruction
and clutter filtering is suitable for any variable
PRT scheme. It starts by representing the variable PRT
sequence as a product of a uniformly sampled signal
with a high PRT code that has unit elements that coincide
with the positions of samples in the variable PRT
sequence and zeroes otherwise. Then, clutter filtering
in the spectral domain and magnitude deconvolution are
used to restore the spectrum in an extended unambiguous
interval.