Optimization of a Blind Adaptive Spatial Filter for Interference Mitigation in GNSS Receivers

E. Tasdemir, L. Kurz, T.G. Noll

Abstract:	In recent years, the vulnerability of Global Navigation Satellite Systems (GNSS) to jamming has become a concerning issue. Nowadays, very cheap portable jamming devices, radiating different types of interference signals in the L1-frequency band, can easily be purchased by individuals [1]. Since available commercial receivers are blocked by any kind of interference overpowering the satellite signals, counter measures have to be taken on the receiver side. In case of a single-antenna receiver, a narrow-band interferer can be mitigated in the frequency domain or a pulsed interferer in the time domain. A wide-band interferer, however, can efficiently be detected and mitigated in the spatial domain using multi-antenna receivers. In [2], a 2x2-multi-antenna receiver architecture using subspace-based methods for interference mitigation in spatial domain is described. In the absence of interference, the gap between a coarse quantization of the antenna signal and the infinite precision case in terms of code delay estimation performance is relatively small [3]. Accordingly, many commercial receivers use one bit analog to digital converters (ADC), in order to reduce the hardware complexity of the ADC and the correlator channels. In the presence of a strong interferer, however, a higher resolution is required. Thus, in [2] antenna signals are initially digitized with a 14-Bit ADC. An adapted spatial filter (projector) is applied immediately to the digital antenna signals at pre-correlation stage, projecting the signals onto the interference-free subspace. After projection, the signal resolution is reduced to a conventional word-length (re-quantization), in order to keep down the complexity of the correlator channels. Coefficients of the spatial filter are derived from the eigenvalues and -vectors of the covariance matrix determined from the high resolution antenna signals in a dedicated building block. For the Eigendecomposition of the covariance matrix in software, Jacobi-Algorithm is proposed, which computes the eigenvectors and eigenvalues by diagonalizing the covariance matrix by means of Givens-rotations. On the Xilinx xc4fsx55 FPGA the covariance matrix computation utilizes 15.5% of all available slices and 9.4% of the DPS48-blocks (3800 slices and 48 DSP48-blocks). For comparison, one tracking channel corresponds to 9.2% of the slices (2200 slices). It is noticeable that the presented filter approach does not feature any feedback, i.e. in a straightforward manner the filter coefficients are recalculated at the end of each episode based on the high resolution input data without making use of the results from the previous episode. Thus, Jacobi-Algorithm repeatedly suffers from a “cold start” resulting in a relatively large number of iterations (sweeps). Furthermore, computation of the covariance matrix before re-quantization makes the hardware unnecessarily complex due to the high input word-length. In order to remove the downsides of the filter approach in [2], a new adaptive spatial filtering approach has been developed by restructuring it into a “sub-space tracking loop”: Instead of subtracting the interference subspace immediately, antenna signals are projected onto orthogonal subspaces using eigenvectors and each subspace is re-quantized after this stage. The covariance matrix computation is based on the decomposed signals and is shifted to the low precision domain. Outputs of the Jacobi-Algorithm from the previous episode are fed back to the next episode. The eigenvalues are required for a conditioning of the covariance matrix for the Jacobi-algorithm. This conditioning contains most notably the computation of the square roots of the eigenvalues. Jacobi-algorithm is then initialized with the last eigenvectors and the covariance matrix. Since this new covariance matrix is already nearly diagonal, the Jacobi-algorithm requires less iterations compared to the approach in [2]. Spatial filtering is completed with a final processing (matrix-vector-multiplication) at post-correlation stage. At the final processing stage, the proposed approach can easily switch between a realization of the projection given in [2] (projection mode) or the classical prewhitening (prewhitening mode). In the first mode complete subspaces are eliminated, that are identified as interference subspace by a comparison of the corresponding eigenvalue to a threshold. In contrast, prewhitening suppresses the power in the interference subspace to the noise level. Simulations show that the steady state performance of the proposed approach in projection mode for stationary scenarios is equal to the one in [2], as expected theoretically. For realistic dynamic scenarios, no practical loss in code delay estimation performance was observed as well. In case of a weak but still detectable interference coming from a similar direction as the satellite signal, i.e. in case of an overlap of the signal and the interference subspace, the proposed approach in prewhitening mode is superior with respect to the SNR. In case of a 2x2 antenna-array and a word-length of 14 bits before and 2 bits after re-quantization like in [2], the hardware complexity of the covariance matrix computation is reduced by 80% due to the low input word-length. The typical number of complete sweeps required by the Jacobi-algorithm is reduced from 6 to 2. As a result, complexity of computing the filter coefficients in software including Jacobi-algorithm is reduced by 60% effectively, taking the required conditioning of the covariance matrix into account. Increasing the number of antenna elements increases the aforementioned hardware and software complexity gain of the proposed approach over the previous approach. [1] R. H. Mitch, R. C. Dougherty, M. L. Psiaki, S. P. Powell, B. W. O'Hanlon, J. A. Bhatti, and T. E. Humphreys, “Signal characteristics of civil GPS jammers”, Proceedings of ION GNSS 2011, the 24th International Technical Meeting of The Satellite Division of the Institute of Navigation, Portland, Oregon, 19–23 Sep. 2011 [2] L. Kurz, E. Tasdemir, D. Bornkessel, G. Kappen, F. Antreich, M. Sgammini, “An Architecture for Embedded Antenna-Array Digital GNSS Receiver Using Subspace-Based Methods for Spatial Filtering”, Satellite Navigation Technologies and European Workshop on GNSS Signals and Signal Processing, (NAVITEC), 2012 6th ESA Workshop, 5-7 Dec. 2012 [3] A. Mezghani, F. Antreich, J. A. Nossek, “Multiple Parameter Estimation With Quantized Channel Output”, Smart Antennas (WSA), 2010 International ITG Workshop, 23-24 Feb. 2010
Published in:	Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013) September 16 - 20, 2013 Nashville Convention Center, Nashville, Tennessee Nashville, TN
Pages:	3424 - 3432
Cite this article:	Tasdemir, E., Kurz, L., Noll, T.G., "Optimization of a Blind Adaptive Spatial Filter for Interference Mitigation in GNSS Receivers," Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013), Nashville, TN, September 2013, pp. 3424-3432.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In