RADIO FREQUENCY INTERFERENCE SUPPRESSION APPLIED TO SYNTHETIC APERTURE RADAR DATA

This paper addresses techniques of Radio Frequency Interference (RFI) suppression applied to Synthetic Aperture Radar (SAR) data. These techniques could also be used to protect radio-astronomical observations from harmful interference. Of the RFI suppression techniques investigated, the notch filter and the Least Mean Squares (LMS) adaptive filter have been implemented and applied to real P-band data obtained from the E-SAR system of the German Aerospace Center (DLR), Oberpfaffenhofen, and to real VHF-band data obtained from the South African SAR (SASAR) system. Both methods significantly suppressed the RFI in the real images investigated. INTRODUCTION RFI is a major problem for low-frequency SAR systems operating in the VHF/UHF-band, because the spectrum is already used extensively by other services such as television, mobile communications, radio and cellular phones. Experience with the South African SAR (SASAR) system [1] has shown that even in remote locations the interference power often exceeds receiver noise by many dB, becoming the limiting factor on system sensitivity and severely degrading the image quality. RFI is also a major problem for radio astronomy [2, 3], where the influence of RFI ranges from the complete loss of data due to receiver saturation, to very subtle distortions of the data. In radio astronomy, the task of RFI suppression is especially difficult, because the signals of interest, namely the radio frequency emissions from celestial sources, are extremely weak. APPROACHES TO RFI SUPPRESSION A number of interference suppression algorithms have been described in the literature [4, 5, 6, 7, 8], many of which require a great amount of computation. Suppressing radio interference from a received signal essentially involves three steps [4]: 1. Finding a model to establish the parameters of the interfering signals; 2. Estimating the parameters of the interfering signals using the measured data; 3. Using the estimated parameters to suppress the interference in the data. Modelling the interference environment would ideally include information such as the statistics about the density of the interference emitters, the identity (type) of such emitters, and their effective radiated power, modulation bandwidth, duty factor and temporal dependence. The most direct way of achieving this is to make use of “sniffer” pulses or “listening beforehand” schemes [4]. This method is useful for many signal processing methods, but its effectiveness depends significantly on how long the RFI remains coherent. Most approaches model the RFI as a superposition of single sinusoidal “tones”, and regard the wideband signal-plus-system noise as white noise. RFI suppression methodologies can be grouped into three main categories [2]: 1. Rejection in the temporal domain (most effective when dealing with strong and spiked bursts of RFI); 2. Rejection in the frequency domain (for weak and long-lasting RFI signals); 3. Spatial filtering (using the difference between the direction-of-arrival of the signal of interest and the RFI). This paper summarises the work done by the author on RFI suppression algorithms that have been applied to synthetic aperture radar (SAR) data. The notch filter is briefly described, whereas more emphasis has been placed on the LMS adaptive filter, which has been used very successfully to suppress RFI [4, 9, 10, 11]. THE NOTCH FILTER Since RFI is (usually) narrowband when compared with the transmitted pulse bandwidth, it shows up as narrow spikes in the frequency domain. These spikes are usually many dB stronger than the surrounding signal level. In order to estimate the interference, it is very useful to average a number of magnitude range spectra, because the SAR signal, due to its random nature, will average out, whereas the RFI, due to its often fairly constant nature, will be enhanced. A median filter can be applied to the averaged magnitude spectra, in order to isolate the RFI spikes to be notched out. THE LMS ADAPTIVE FILTER Fig. 1 shows a schematic of the LMS adaptive filter as it is used for RFI suppression. It requires a primary input d (containing RFI) and a reference input x, which is obtained by delaying the primary input for some time-delay ∆. The adaptive linear combiner weighs and sums a set of input signals to form an adaptive output y. This output is an estimate of the RFI. The error signal e, which is the desired cleaned radar signal, is obtained by subtracting the RFI estimate y from the primary input d. The LMS adaptive algorithm minimises the mean-square error e by recursively altering the weight vector W at each sampling instant according to the Widrow-Hoff algorithm [11], yielding * 1 ) ( 2 j j j j e X W W μ + = + (1) where j X is the complex conjugate of the reference signal vector at time j, e(j) is the error signal at time j, and μ is a constant convergence factor controlling stability and rate of adaptation. RFI TRANSFER FUNCTION The equivalent transfer function H(ω) of the LMS adaptive filter may be obtained once the filter tap weights have converged and are kept constant, making it unnecessary to feed the error signal e back into the adaptive filter. Fig. 2 shows the transfer function block diagram of the LMS adaptive filter with the weights kept constant. In the frequency domain, the output E(ω) is given by [ ]. ) ( ) ( 1 ) ( ) ( ) ( ) ( ω ω ω ω ω ω G F D Y D E − = − = (2) Therefore the equivalent transfer function H(ω) of the RFI suppression stage is given by ) ( ) ( 1 ) ( ) ( ) ( ω ω ω ω ω G F D E H − = = (3) where G(ω) is the transfer function of the time-delay ∆, given by ∆ − = ω ω j e G ) ( (4) and F(ω) is the Fourier Transform of the time-reversed weight vector W . Since we are dealing with digitised data, W must be zero-padded to the same length as the range compression transfer function described below. Fig. 1. RFI suppression using an LMS adaptive filter. Fig. 2. Block diagram of LMS adaptive filter with tap weights kept constant. x y d e Primary Input