Working Papers • July/August 2016
Acquisition output of the Galileo E5 signal when no interference mitigation is used to cope with DME interference of 3,000 pps density.
Interference Mitigation in the E5a Galileo Band Using an OpenSource SimulatorResearchers describe the result of investigations into three narrowband interference mitigation methods based on time and frequency processing and compare them in terms of signal acquisition and tracking performance.
Share via: Slashdot Technorati Twitter Facebook Four global navigation satellite systems are scheduled to be fully operational orbiting Earth in the coming years: the NAVSTAR Global Positioning System (GPS) from the United States, the GLObal NAvigation Satellite System (GLONASS) from Russia, the Compass/BeiDou2 System (BDS) from China, and Galileo from Europe. A considerably high number of signals, coming from the satellites of those constellations, will share the radio electric spectrum. Moreover, some aeronautical radio navigation systems (ARNS) operate in the E5 Galileo band. For example, distance measuring equipment (DME) and tactical air navigation (TACAN) systems (both in the ARNS category) broadcast strong pulsed ranging signals that interfere with Galileo E5a and GPS L5 signals. As analyzed in the work by F. Bastide et alia, listed in the Additional Resources section near the end of this article, DME/TACAN interferences can severely degrade the receiver performance if left unmitigated. Galileo receiver simulators are a powerful way to investigate the initial performance of Galileo receivers without the need of heavy measurement campaigns. Applications of opensource Galileo simulators, especially regarding the E5 band, are still hard to find in the current literature. This article presents the development of an opensource 64bit Galileo simulator, including the acquisition and tracking parts and the interference mitigation blocks for continuous wave interference (CWI) and DME. The simulator is available on demand and upon agreeing to its opensource conditions (Details listed in the Manufacturers section at the end of this article). This article thoroughly analyzes three narrowband interference mitigation methods explained in the next sections (notch filtering, zeroing, and pulse blanking) with Galileo E5a signals based on the opensource simulator created in our group (Signal Processing for wireless positioning group at Tampere University of Technology). The performance studies are done with both the benchmark CWI and the DME interferences. The novelty of our work comes from analyzing jointly these three techniques with a practical Galileo simulator and from selecting the best method according to the interference type. We show that zeroing methods are best used for robustness and with strong narrowband CWI while pulse blanking methods are better than notch filtering methods for strong DME interferers. We also show that interferers with up to 10–15 decibels stronger power than the E5a signal power can be tolerated relatively well and that all considered approaches have relatively similar performance for medium strength interferers.
GNSS Interferences Some interference can be mitigated well using time or frequency processing methods. However, when dealing with wideband interference, the performance of these methods degrades, and additional processing has to be carried out, such as spacebased processing methods (i.e., antenna array–based methods). Minimum variance distortionless response (MVDR) and minimum power distortionless response (MPDR) beamformers are some examples. These spatial approaches are not assessed in this article, however, and we focused our research on narrowband interference — those whose bandwidth is much lower than the bandwidth of the GNSS signal of interest. Narrowband interference can be created, for example, by TV harmonics, intermodulation products or signals from very high frequency (VHF) and ultra high frequency (UHF) stations, or signals generated by systems such as DME or TACAN. Figure 1 illustrates the different types of interference in Galileo bands. Another criterion can be the intentionality. Within the unintentional interference group, we can emphasize: DME/TACAN, amateur radio, TV, surveillance radars, or wind profiler radars. Under the name of intentional interference (see Figure 2 and Figure 3), three different interference signals can be distinguished: jamming signals, which deliberately block or interfere with authorized wireless communications through illegal devices decreasing the signaltointerferenceplusnoise ratio (SINR); spoofing signals, which falsely imitate the signalinspace (SIS) and may hack a targeted GNSS receiver; and meaconing signals, which are the interception and delayedrebroadcast of actual GNSS signals. In this article, we have simulated and studied two interference signals: CWI and pulsed signals such as those generated by the DME or TACAN systems. A CWI signal can be modelled as
Equation (1) (see inset photo, above right, for all equations) where Δf_{cwi} is the frequency offset with respect to the GNSS carrier, A is the CWI amplitude, and ϕ_{0} is the CWI signal initial phase. Signals from air radionavigation systems, such as DME or TACAN, consist of Gaussian RF paired pulses. Pulse separation is 12 microseconds with each pulse lasting 3.5 microseconds. The maximum repetition rate is about 3,000 pulse pairs per second (pps). DME systems are designed to provide service for 100 planes simultaneously and the transmitted power may vary from 50 watts to 2 kilowatts. A DME signal is typically modeled as: Equation (2) where α = 4.5 • 10^{11} s^{–2} is a parameter controlling the pulse width and Δt = 12 • 10^{–16} s is a parameter controlling the gap between paired pulses. The DME system operates between 960 and 1215 MHz; hence, it overlaps the Galileo E5 band. Figure 4 shows an example of a DME signal in the time domain, its envelope, and its frequency spectrum.
StateoftheArt Narrowband Interference Mitigation Frequencydomain approaches are those based on signal alterations in the frequency domain. The article by A. Rusu and E. S. Lohan listed in the Additional Resources section near the end of this article presents a filtering method that exploits the cyclostationarity property using the spectral correlation function (SCF) and, therefore, can suppress additive white Gaussian noise (AWGN). Another, even simpler method is called zeroing, which is an excisionbased method that we will explain in the next section. The literature also presents various transformed domain mitigations that are worth mentioning briefly. One is the wavelet transform which is a timescale representation technique that overcomes the common limit of fast Fourier transform (FFT) transformations using the short time Fourier transform (STFT), and another is the Gabor transform. Both of these methods separate useful signal and interference, removing the coefficients with high energy before the inverse transform. (These methods are described in articles by E. Anyaegbu et alia and K. Ohno and T. Ikegami, respectively, cited in the Additional Resources section. Studied Mitigations. We selected the methods explained in this section based on the tradeoff between performance in acquisition and tracking and the method’s complexity, which in turn is directly proportional to the amount of computational load. The pulse blanking and notch filtering methods are timebased approaches, while the zeroing method is a frequencybased one in which the simulated signal is grouped into blocks that become suitable for FFT processing. Pulse Blanking. This method is simple to implement: it blanks incoming signals that exceed a certain threshold, as illustrated in Figure 5. The threshold can be chosen, for example, as a factor of the mean value of the absolute value of the received signal, i.e., γ = k • E(s(t)) with k optimized according to the interference. In our simulations, we used, for example, k = 3.5, chosen empirically. Figure 6 shows an example of pulse blanking performance in the frequency domain in the presence of a DME interferer. Notch Filtering. Another timedomain method is notch filtering. A second order infinite impulse response (IIR) notch filter to mitigate the narrowband interference has been proposed, for example, by C. YingRen et alia (see Additional Resources), based on the following transfer function: Equation (3) is the 3 dB filter bandwidth, and f_{I} is the frequency of the interferer that must be canceled. The interfering frequencies are searched in a recursive manner, based on a threshold, as illustrated in Figure 7. As an example, Figure 8 shows the spectrum of a GNSS signal affected by DME interference, with and without notch filteringbased mitigation. Zeroing. The discrete Fourier transform of a sample GNSS signal s(n) can be written as: Equation (4) Narrowband interferences can be rejected just by zeroing the spectral samples above a certain threshold. This time, the threshold γFFT is obtained according to the mean and the variance of the absolute value: Equation (5) where ε is a parameter adjusting the threshold (in our simulations ε = 0.5). Figure 9 presents an example of the zeroing method (in the frequency domain).
Qualitative Comparison Among Narrowband Mitigation Techniques
OpenSource Simulator Our team at Tampere University of Technology developed a simulator with which to analyze Galileo signals; Figure 11 illustrates an overview of this development. The simulator was initially started within the European Union’s Galileo Ready Advanced Mass MArket Receiver (GRAMMAR) project and is now offered via free licensing for research purposes. The simulator implements the transmitted signal based on an AltBOC(15,10) modulation with a constant envelope signal, according to the Galileo Open Service SIS Interface Control Document (SISICD). The signal is sent over a multipath channel with up to five Rayleigh fading paths; noise and interference are added inside the channel block. Due to computing capacity, the signal is transmitted at an intermediate frequency (IF) of 20 megahertz. The downsampling factor is applied before the channel is employed to reduce the simulation time. Because the processing of the E5a band is only carried out at the receiver, a lower bandwidth is needed. The E5a sampling rate in our simulator is 31.5 megahertz, while the transmitter sampling rate is four times higher. The interference generation block, included inside the channel simulation, is detailed in Figure 12. The receiver includes the interference mitigation block, the acquisition, and the tracking unit. Figure 13 illustrates the interference mitigation block. The acquisition block estimates the time and frequency initial values that are then fed into a tracking block. Figure 14 and Figure 15 show, respectively, examples of the timefrequency acquisition mesh without and with interference mitigation, in the case of a CWI interferer at 1176.45 MHz, i.e., an E5a carrier frequency. The acquired signal is passed through a narrow correlator tracking block, including a delay lock loop (DLL) and a joint frequency lock Loop (FLL) – phase lock loop (PLL). Figure 16 presents the tracking unit block diagram.
Performance Comparison In order to achieve a high detection rate, the blanking method for DME pulses and zeroing method for CWI are the most effective techniques among those studied. Regarding the tracking results, it is worth mentioning how large the tracking error can become if no mitigation is taken into account to deal with the interference. The acquisition threshold is selected based on the highest peak of the timefrequency mesh. (For further discussion of this point, see the article by E. Pajala et alia in Additional Resources.) Due to some type of interference, for instance DME pulses, large fluctuations can appear at some point along this mesh, and as a result the initial values can be extremely large as Figure 19 shows. The computed position error could even be on the order of kilometers, due to the fact that the acquisition stage would feed an erroneous estimate into the tracking. However, as might be expected, the studied mitigations are able to keep this error within reasonable values as shown in Figure 20.
Conclusions
Acknowledgments
Additional Resources ManufacturersThe opensource 64bit Galileo simulator described in this article was developed with the aid of Matlab Simulink graphical programming environment from The Mathworks, Inc., Natick, Massachusetts USA, including the acquisition and tracking portions and the interference mitigation blocks for continuous wave interference and DMEs. The Simulink Galileo E1 and E5a baseband transmitterreceiver chain is available on demand and upon agreeing to its opensource conditions. It can be found here. The Simulinkbased simulator referenced in Figure 11 was initially started within the EU Galileo Ready Advanced Mass MArket Receiver (GRAMMAR) project and is now offered via free licensing for research purposes here. Copyright © 2017 Gibbons Media & Research LLC, all rights reserved. 
