GNSS RFI Detection: Finding the Needle in the Haystack

Transcription

1 GNSS RFI Detection: Finding the Needle in the Haystack M. Scaramuzza, H. Wipf, M. Troller, skyguide, Swiss Air Navigation Services Ltd. H. Leibundgut, REGA S. Rämi, Swiss Air Force R. Wittwer, Armasuisse BIOGRAPHY Maurizio F. Scaramuzza received his Diploma in Geomatics in 1995 at the Swiss Federal Institute of Technology (ETH) Zürich. He joined in 1995 the Institute of Geodesy and Photogrammetry at the ETH Zurich, where he received in 1998 the doctorate in technical sciences in the field of satellite based flight approaches and landings. In 1999 he joined skyguide and successively built up and led the GNSS team. Since 2006 he is head of the expert group on Communication, Navigation and Surveillance. and yaw angles of the helicopters, the GPS signals are normalized with use of an empirical derived GPS antenna pattern. The normalized GPS signals are finally statistically assessed for the determination of potential RFI. Only GPS L1 C/A data is taken into account within this paper, but with few modifications this method can be adapted to any GNSS providing pseudo range services. Over 6000 hours of recorded flight data were assessed so far. The empirical derived antenna pattern for GPS signal normalization is shown. Horizontal and vertical distribution of the flights are depicted. First results on potential GPS RFI are presented and discussed. ABSTRACT What does a needle in the haystack have in common with GNSS radio frequency interference (RFI)? According to the Cambridge Dictionary this expression means that it is impossible or extremely difficult to find something, especially because an area to search is simply too large. This is exactly the situation for GNSS RFI detection in space and time. GNSS RFI is increasingly becoming important. Safety critical applications using GNSS, which are exposed to RFI, might lead to unacceptable performance degradations. Aviation, with GNSS based flight procedures conducted under Instrument Meteorological Conditions (IMC), are particularly concerned about this threat. Therefore it is crucial to develop the capability to assess the GNSS RFI situation over a large region and during a sufficiently long period. A method based on [1] is presented on how the RFI situation can be assessed over the whole area of Switzerland. It consists in installing mini quick access recorders (mqar) on board of two dozen helicopters operated by Rega, the main Swiss Helicopter Emergency and Medical Service (HEMS), and by the Swiss Air Force, and collecting data during a period of three years. Daily missions of the two operators are used to record data. In this way, large parts of Switzerland are randomly covered. The low flight altitude is common to all helicopter missions. Therefore it is expected, that the probability having the aerial vehicles exposed to ground based RFI is higher compared to commercial fixed wing operations. Based on recorded C/No (Carrier to Noise Ratio), GPS satellite azimuth and elevation angles as well as roll, pitch INTRODUCTION Switzerland is currently implementing ICAO's Performance Based Navigation (PBN) concept within the Swiss airspace. This undertaking is managed through the Swiss-wide Implementation Programme for SESARoriented objectives (CHIPS), which was initiated back in In the frame of CHIPS a number of applied research and development efforts are conducted in order to solve specific problems related to the peculiarities of Swiss air space like the mountainous topography. One of them is called Helicopter Recording Random Flights (HRRF). Quick access recorders are installed on board of helicopters operated by Rega, the main Swiss HEMS, and by the Swiss Air Force. The objectives of this study are manifold, among others RFI detection, assessment of GNSS performance within a topographic challenging environment, assessing the potential of reduced Required Navigation Performance (RNP) values to name a few. A wide number of parameters from the onboard GPS receiver used for navigation as well as helicopter attitude and FMS data are recorded. This data set allows to identify possible GPS RFI. Data are recorded for a period of at least 3 years. Since most flights are carried out under Visual Meteorological Conditions (VMC) with a variety of mission types, it is expected, that the lower part of the Swiss airspace will be randomly sampled as a hypothesis. TECHNICAL SOLUTION AND RECORDED DATA The entire fleet of helicopters equipped with recording units consists of 6 EC145 (Figure 1) and 11 AW109SP operated by Rega, and 18 EC635 (Figure 2) operated by

2 the Swiss Air Force. The parameters needed for potential GPS RFI detection are available on the EC145 and EC635 and therefore further discussion is limited to these two types of vehicle Figure 1: EC145 of the HEMS operator REGA. The GPS antenna is mounted on the top of the fin in front of the strobe light (Courtesy REGA). Because of the prolonged period of three years for data gathering it was decided to have fixed recording installations. The technical solution is a mini Quick Access Recorder (mqar) connected to the aircraft's ARINC bus. The mqar is a small size and light weight equipment. Figure 3 depicts an installed mqar, indicated by the red arrow. A large amount of data parameters are available and recorded onboard, basically from GPS, Attitude and Heading Reference System (AHRS) and Flight Management System (FMS). GPS data consist on GPS positions, satellite vehicle positions, pseudo ranges and pseudo range rates, horizontal and vertical integrity limits and figure of merits, carrier to noise ratio and different status parameters. AHRS data consist on roll, pitch and heading information. Finally the flight plan as well as the selected waypoints are available from the FMS. Sampling interval on GPS and AHRS is 1Hz. OUTLINE OF POTENTIAL RFI DETECTION The basic idea for the detection of any potential GPS RFI is to assess the behavior of the C/No of each tracked GPS satellite. An RFI occurrence would decrease the C/No by a constant value at each single epoch because the entire GPS receiving antenna is affected by the same interference level. To do so it is necessary to know the GPS antenna pattern which affects the GPS C/No depending on the satellite position referred to the GPS antenna. This allows to normalize the C/No and minimize any signal attenuation not related to RFI. Finally it is possible to determine whether a constant decrease of all C/No is present, indicating a potential RFI, or not. Following flow chart (Figure 4) depicts all steps in order to detect any potential RFI. Figure 2: EC635 of the Swiss Air Force. The GPS antenna is mounted analogous to the EC145 (Courtesy VBS). Figure 3: Installed Avionica mqar (red arrow). Figure 4: Flow chart of all required steps in order to detect potential GPS RFI.

3 The synopsis in Figure 5 shows the context in which the man-made RFI is located. It is understood, that RFI is not the only electromagnetic type of interference to have an adverse impact on the quality of the GNSS signals. Only the red marked boxes are subject of RFI detection within this project. Figure 5: Context of man-made RFI within the discussion on GPS RFI. Only the red marked boxes are discussed. DETERMINATION OF GPS ANTENNA PATTERN Under dynamic conditions the positions of the satellites in reference to the antenna have an impact on the C/No. Changes of attitude (roll, pitch, yaw and heading) of the aerial vehicle affect the C/No values. This is due to the non-isotropic nature of the antenna pattern. This effect is depicted in the Figures 6 and 7. In Figure 6, a GPS antenna is mounted on the top of a helicopter fin. The colored hemisphere indicates the estimated expected C/No values in polar coordinates. The antenna coordinate system is aligned with the topocentric coordinate system, i.e. the roll axis is oriented towards north, the pitch axis towards east and the yaw axis towards the nadir. The red arrow indicates the direction towards an arbitrary GPS satellite. According to the color scale, a C/No of roughly 51 db-hz should be expected. In Figure 7 the helicopter attitude is altered (yaw angle -30, roll angle -15 and pitch angle -5 ). Due to the fact, that the GPS satellite remains fixed in the topocentric coordinate system, the red arrow is rotated within the antenna coordinate system which is denoted through the black arrow. The expected C/No has now decreased to roughly 45 db-hz. This effect has to be taken into account when analyzing the changes on C/No for the detection of any potential RFI. Figure 6: GPS antenna pattern on top of a helicopter fin. The red arrow indicates the direction towards an arbitrary GPS satellite. The helicopter is aligned with the topocentric coordinate system. Figure 7: GPS antenna pattern and direction of GPS satellite for a helicopter attitude of yaw angle 30, roll angle 15 and pitch angle 5. Therefore it is necessary to first derive the antenna pattern for each helicopter as shown in Figures 6 and 7. To do this the hemisphere above the helicopter is subdivided into regular bins of the size of = 1 by = 1. Based

4 on the helicopter's position, its attitude and the position of the GPS satellites, the measured C/No can be attributed to the corresponding bins. Finally the C/No of each bin is averaged. The described derivation of the averaged C/No is subject to different perturbing effects, which might alter the result: 1. multipath by the airframe 2. signal attenuation caused by the air frame (mainly shadowing) 3. antenna gain pattern 4. multipath from environment (non-airframe) 5. atmospheric effects, mainly ionosphere and troposphere 6. signal attenuation caused by terrain and obstacles (mainly shadowing) 7. signal attenuation due to RFI 8. Noise The first three effects are the main reasons for antenna pattern forming. The fourth effect is minimized by eliminating all C/No measurements where the helicopter falls below a minimum velocity. This allows having a rapidly altering geometry of the GPS satellites, the reflectors and the GPS receiving antenna leading to multipath effects appearing only during short periods. The fifth effect could be modeled through standard models. These attenuations are nevertheless small and therefore negligible [2], [3] and [4]. The sixth effect usually occurs when GPS satellites are located close to the radio horizon. The signal attenuation can be of higher magnitudes and is statistically characterized as an outlier. Therefore these measurements can be eliminated by taking only values within a certain quantile range into account. This approach is acceptable due to the large amount of available data. The seventh effect is the target of the experiment. It has nevertheless to be filtered because it would affect the determination of the antenna pattern. Filtering can be achieved analogously to the sixth effect. The eight effect, noise, remains present. Figure 8 shows the derived antenna pattern for the antennas on two different EC635 and one on the EC145. The two diagrams describing the antennas on the EC635 (Figure 8 top and middle) are expectedly almost the same due to the identical environment of the antennas and the minimization of the perturbing effects on the C/No. The typical decrease of the signal gain for lower elevations on patch antennas is clearly visible. The low C/No in the azimuth region around 180 at low elevations is most probably due to installations in the antenna near field. The interferometric like pattern in the front area of the antenna isn't assessed yet, but might be caused by the rotor blades. The antenna environment on the EC145 and EC635 is different (see Figure 1 and 2) and the corresponding antenna pattern differs clearly (Figure 8 bottom). Figure 8: Derived antenna pattern of two different EC635 (top and middle) and EC145 (bottom). The quality of the derived antenna pattern can be assessed through the estimation of the standard deviations of the C/No for each bin, which is shown in Figure 9 for an EC635. For almost the entire antenna area the value is below 2 db-hz. At some smaller spots, mainly in direction of the rotor blades, it increases up to 4 db-hz. Most problematic is the area at elevations below 5, where values up to 10 db-hz appears. It is therefore reasonable to eliminate C/No measurements with low elevation angles when using these measurements for RFI detections. Figure 9: Standard deviation of the C/No for each bin for an EC635. NORMALIZATION OF RECORDED C/No The next step in the detection of potential RFI is to normalize C/No measurements at each epoch, i.e. to make the measured C/No independent from the location of the GPS satellites within the antenna coordinate system. Thus, for all measured C/No the corresponding GPS position within the antenna coordinate system is derived. Then the corresponding bin and averaged C/No value within the antenna pattern is determined. Finally the averaged C/No is subtracted from the measured C/No. A value of zero is expected under non-interfering conditions. Figure 10 shows the normalized C/No (black lines) and the averaged C/No (red line) for a noninterfered signal on a flight over a period of roughly 900 seconds. In this case the average C/No is calculated

5 through the arithmetic mean. The resulting mean value of the normalized C/No is db-hz whereas the standard deviation is 0.27 db-hz. Any interfering signal would affect these statistical parameters. Figure 11: Mean value of all normalized C/No (red line) and of all normalized C/No, which don't differ more than 3 db-hz from the mean value (blue line). Figure 10: Normalized C/No (black lines) and average of normalized C/No (red line). FILTERING OF GPS SIGNALS AFFECTED BY TERRAIN AND OBSTACLES The list of major perturbing effects for a helicopter flying above a minimum velocity is now reduced to: 1. signal attenuation caused by terrain and obstacles outside the airframe (mainly shadowing) 2. signal attenuation due to RFI 3. Noise The main difference between the first and the second effect is that under circumstances of signal attenuation due to terrain and obstacles outside the airframe only single GPS satellites and their C/No are affected. Therefore the mean value is affected. However, an RFI affects all C/No by the same attenuation resulting in an altered mean value too. Therefore the first effect has to be filtered in order to distinguish an altered mean value caused by signal obstructions from that caused by RFI. This is achieved by eliminating normalized C/No values differing more than certain value from the mean value. Figure 11 shows for the same time span as in Figure 10 the resulting mean value of all normalized C/No (red line) and the mean value after eliminating in this case C/No which differ more than 3 db-hz from the mean value (blue line). The improvement of the averaged normalized C/No is depicted by the black arrows in Figure 11. With application of this method, the mean value improves to db-hz with a standard deviation of 0.22 db-hz. Figure 12 shows a case where loss of track of one GPS satellite (black bold line) is due to terrain shadowing. The thin black lines represents the normalized C/No of all other tracked satellites. The red and blue lines have the same significance as in Figure 11. The total number of tracked satellites are represented by the black dots. It is clearly visible, that without the elimination of the attenuated satellite signal the mean value would decrease by 3 db-hz although no RFI is present (red line). After filtering the GPS signal affected by the terrain, the mean value of the normalized C/No is barely not altered (blue line). Loss of tracked satellites due to terrain is very common for the helicopters of the Swiss Air Force and REGA due to the low flight altitudes within mountainous regions. Figure 12: Situation of one satellite being shadowed by the terrain (black bold line). The red line shows the mean value of all normalized C/No and is decreased by 3 db-hz due to the terrain effect. The blue line shows the mean value after filtering the shadowed satellite.

6 DETECTION OF POTENTIAL GPS RFI A potential RFI is now detectable by assessing the mean values of the normalized C/No while eliminating outliers. A typical event is shown in Figure 13. The colors correspond to those in Figure 12. The time span is 90 seconds. During this period, the mean value falls by up to 10 db while the standard deviation of the normalized C/No remains similar to the situation when no RFI is present. This is a strong indication that all GPS signals are affected by the same amount of perturbation. So this effect could be explained by a potential RFI. permanently active or not. In a first step the time of recorded data for different regions is determined. To do so, the entire area of interest is subdivided into regular horizontal squares of 250m side length and the flight time of all helicopters is summed for each square resulting in a flight distribution. Figure 14 shows this distribution for an area of 90km by 70 km. The colors indicates the time the helicopters were within the square. Figure 14: Horizontal flight distribution. The colors indicates the total time the helicopters were within a square of 250m side length. Note that the scale ends at 240s, but the time can be significantly higher in some areas. Figure 13: Situation where a potential RFI might be present. The mean value of the normalized C/No decreases by up to 10 db-hz. The standard deviation remains similar compared to situations without RFI. FLIGHT DISTRIBUTION Classification of the detected potential RFI are of importance in order to gain insight into the RFI situation. Vehicle's exposure to RFI depends on the one hand on the RFI transmitter characteristics and its environment: intentional / non-intentional natural / manmade location: stationary / moving temporal: permanent / sporadic emitted power level transmitter antenna modulation (CW, chirp, wide band, etc.) environment (topography, buildings, vegetation, etc.) On the other hand the RFI exposure highly depends on vehicle's position referred to the RFI transmitter vehicle's attitude referred to the RFI transmitter GPS receiver's characteristics Figure 14 would obviously not pass a test of complete randomness as set out in the initial hypothesis. The initial argument of following missions under visual flight rule by two operators having somewhat diverse mission profiles like HEMS and military air transport nevertheless led to a certain uneven distribution in aircraft positions. The pragmatic approach to that fact is, to realize, that there is a higher probability of RFI detection where the aerial vehicle's increased likelihood of its position is. Implying this likelihood is also not too different when operating under instrument flight rules, because the flight routes are governed in both flight rules by the existing infrastructure like helipads, airports or helicopter base. The helicopters fly very often at low altitudes above ground. Figure 15 displays the height distribution where the tendency of low flight altitudes above ground is clearly visible and with it also the increased danger of exposure to ground based RFI transmitters compared to commercial fixed wing operations. These height values are determined by subtracting the terrain altitude derived from a Digital Elevation Model (DEM) at the position of the helicopter from the recorded altitude above mean sea level. Taking into account all of these parameters is a complicated task. Therefore an approach is chosen which allows to estimate the probability of exposure and to determine whether the ground RFI transmitter is

7 Figure 17: Probability of exposure to potential GPS RFI. Figure 15: Helicopters height distribution above ground level. Note that the time scale is logarithmic. EXPOSURE TO POTENTIAL GPS RFI A first estimation to determine the probability of an helicopter being exposed to possible RFI can be done. An area is selected, where potential RFI is detected repeatedly. Figure 16 shows a small part of this area with two longer and one short potential RFI for which a mean normalized CNo below -3dB is measured (colored dots). The gray squares indicates the total amount of time the all helicopters were present at each square of 250m side length. It has to be denoted, that this area is located within a small city. If the potential RFI transmitter is located close to buildings, then shadowing effects might appear and the probability would be lower than compared to an area without obstructions. Possible shadowing effects of the potential ground RFI transmitter are observed for another area, where a large amount of flights are recorded (Figure 18). This area is located in proximity of an aerodrome and the flights exposed to potential RFI are part of VFR departing and arrival routes at low altitudes. The three red arrows indicates a possible shadowing effect, most probably due to terrain. It is noted, that the possible shadowing effect is highly depending on the vehicles altitude at these locations. Further it is visible, that the GPS signal attenuation is larger towards the center of this area. These properties will be used in following works in order to confine the area, where the potential RFI might be located. Figure 16: Three cases where helicopters were exposed to potential GPS RFI. The probability of exposure to potential RFI for this area is derived by determining the ratio of total epochs per square with a mean normalized RFI lower than -3 db and the total time of flight within a square. This ratio is depicted in Figure 17. The maximum probability of exposure within this area is estimated to be 5%. Figure 18: Possible shadowing effects (red arrows) of the potential ground RFI transmitter.

8 CONCLUSIONS A method is presented which enables to detect potential RFI based on measurements of C/No, aerial vehicle position and attitude. C/No attenuation due to GPS antenna pattern and terrain is taken into account. Potential RFI affecting the C/No by only a few db can be detected with this method. Finally, first results of potential GPS RFI based on an assessment of over 6000 hours of recorded data is presented. ACKNOWLEDGMENTS This study has been financed by Skyguide in the frame of the applied research and development programme to implement future navigation applications (CHIPS). REFERENCES [1] M. Scaramuzza, H. Wipf, M. Troller, H. Leibundgut, R. Wittwer, and S. Rämi, GNSS RFI Detection in Switzerland based on Helicopter Recording Random Flights. IFIS 2014, Oklahoma City, USA, Proceedings. [2] Curry, Richard, 2012, Radar Essentials, Scitech Publishing Inc. [3] ITU, 2013, Attenuation by atmospheric gases, Recommendation ITU-R P , P Series, Radiowave propagation, 09/2013. [4] Christie, Jock et al., 1996, The Effects of the Ionosphere and C/A Frequency on GPS Signal Shape: Considerations for GNSS-2, Proceedings of the 9th International Technical Meeting of the Satellite Division of The Institute of Navigation, ION GPS 1996.