Tracking RFI: Interference localization using a CRPA

A controlled radiation-pattern antenna can preserve GNSS positioning while providing at least an azimuth angle towards an interference source. If integrated with an attitude and heading reference system (AHRS), only a few lines of position pointing towards the RFI source could provide a fast indication of the probable ground location.

By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

GNSS is an essential enabler for many aviation applications that rely on either accurate position or time synchronization. While the idea of “sole means” GNSS is disappearing, it remains challenging to match the performance and coverage of GNSS with terrestrial systems. This is why aviation is working on Alternate Positioning, Navigation and Time (A-PNT) to cope with the potential for a wide-area GNSS outage. Current navigation aids are clearly part of this approach in the short term. We will continue to need a terrestrial capability for some time, but we don’t expect that it will support the same level of performance as GNSS. Even if we have back-up, we must be able to resolve GNSS outages efficiently.

Among principal GNSS vulnerabilities — constellation performance issues, space/solar weather and radio-frequency interference (RFI) — RFI is the one where observability on the ground is often limited. While the protection of radio services from interference is a state responsibility typically assigned to a telecommunications or other government agency, it is in the interest of an air navigation service provider (ANSP) to be able to request help and enforcement action from the telecommunications regulator in an efficient manner.

As a part of its contribution to Single European Sky ATM Research (SESAR, a collaborative project to improve European airspace and its air traffic management), Eurocontrol has developed an RFI Mitigation Plan as a guidance framework with the objective to maintain risks to GNSS and the associated operations at tolerable levels. The document will be published by ICAO in its GNSS Manual in the new 2017 edition.

MITIGATION PLAN

RFI can be a security issue. Consequently, a commonly used philosophy in the security domain was used in the mitigation plan: there are many potential threats, but not necessarily all of them translate into operationally relevant risks. Threats are thus sort of dormant risks, which, if left to develop unmitigated, could develop into risks to aviation. The mitigation process monitors threats, assesses risks, and then implements suitable mitigation to stop threats from developing into risks. Three successive stages have been identified where such barriers can be applied:

• Prevent transmission of RFI, mostly through radio regulatory actions and coordination;
• Prevent interruption of positioning and navigation capabilities in the presence of RFI. This is achieved at the avionics level by making sure receivers can tolerate some RFI as well as redundant capabilities;
• If interruption cannot be avoided, ensure that other communication, navigation and surveillance capabilities provide continued safety while being able to detect, locate and eliminate an RFI source efficiently.

This third barrier is where flight inspection or other aerial work platforms can play a significant role. However, this role is not limited to risk mitigation. Aerial measurement capabilities can also play a role in threat monitoring by getting data on RFI emissions that are too weak to pose operational risks, and facilitate risk assessment by providing a reliable reference of the impact of such signals on an aircraft in flight.

FLIGHT INSPECTION

Similar to the subject of flight validation, airborne GNSS signal-in-space testing must not necessarily rely on traditional flight inspection capabilities. Other aerial work capabilities can be used, and it is hoped that, over time, data from regular aircraft operations and event recording systems can be used at least for threat-monitoring purposes. However, as soon as a significant RFI occurs, purpose-built aerial detection and localization capabilities are hard to beat. Given that aviation is carrying the risks related to RFI, and telecom regulators are unlikely to have such capabilities, this naturally points to the experience and resources of flight inspection aircraft and their crews.

Even if a significant amount of ground-based RFI sensors are available, local building shadowing can make it difficult to impossible to detect and locate an RFI emitter. Aircraft-provided data can be superior to ground data, and a rough aircraft-based localization can greatly increase efficiency of ground-based localization and source elimination efforts. Aerial RFI localization capabilities offer unique strengths in an overall cooperative process.

EVOLVING SIGNALS

GNSS manifests the transition from analog signals of conventional navigation aids to digital ones. A common characteristic of digital signals is their better use of a frequency channel by spreading the carrier energy such that distinct carrier or subcarrier tones become difficult to observe. Unfortunately, RFI sources have kept up with this, and now most commonly employ swept CW signals, easy to produce but still looking essentially like broadband signals. Many unintentional RFI sources also look like broadband.

Because GNSS is a multi-modal system not uniquely used by aviation, a new type of RFI threat is becoming more common: intentional RFI, which is not directed at aviation, but may nonetheless have an impact. Because there is no direct intent to harm aviation, the nature of these signals and RFI scenarios can become diverse and unpredictable. Furthermore, given the prevalent and ubiquitous nature of GNSS, the number of potential RFI threats is more significant and will evolve more dynamically than aviation capabilities.

A recent effort collecting GPS outage data reported by pilots revealed that a small but surprising number of outages that could potentially be linked to RFI occur on a regular basis, even during en-route operations in some limited regions of the world. For flight inspection, this implies it would be useful to increase the sensitivity of RFI source detection commensurate with the digital nature of GNSS and consistent with the power levels that can impact receivers.

Another particular challenge comes from the specification of an interference mask for GNSS. Other navigation systems do not have such a mask, or any kind of minimum signal-to-noise ratio standard. The mask represents a realistically achievable interference environment. It has been adopted as a global benchmark where receivers experiencing signals above the mask may not produce misleading information, but may stop operating.

However, in practice, little is known about by how much typical receivers exceed the minimum masks. Some tests have reported a margin as significant as 23 dB to CW and 10 dB to broadband signals. This means that an RFI which may not bother one type of receiver at all could be a significant problem for another, limiting the possibility to rely on observed receiver performance. It also implies that signal-in-space effects should be detectable at the low levels of the ICAO receiver RFI mask.

CRPA LOCALIZATION

For civil aviation as opposed to military operations, a CRPA could make sense provided that it outperforms current RFI localization methods at a reasonable price. In military applications, the exact location of the RFI source may be of a secondary nature, as long as desired signal tracking can be maintained.

However, by steering a null (negative gain) towards the angle of arrival of an undesired signal source, a line or sector of possible source positions can be obtained. In this case, the main objective would not be to null a deliberate interferer or jammer, but to obtain a bearing on the type of the interferer. The main scenario we worry about that leads to low-power events are those where aviation is not the desired target, such as a PPD. Unintentional cases can be a mix of high- or low-power cases. The use of a GNSS-specific antenna is expected to provide the required sensitivity, while being able to profit from the military off-the-shelf development. When further integrated with standard flight-inspection sensors such as an attitude and heading reference system (AHRS) and additional geolocation software, this approach has the potential to increase the reliability, accuracy and speed of geolocation while reducing operator effort and flying time. An additional potential benefit is the preservation of ownship position when flying into an area of significant RFI.

The suggested use of military technology brings with it the question on how such use could be authorized. CRPA antennas and associated antenna electronics manufactured in the United States fall under the International Traffic in Arms Regulation (ITAR). While this is a solvable but, nonetheless, cumbersome issue, the approach taken by this project was first to evaluate possible benefits from using a CRPA before worrying about the ITAR issue.

This study was conducted by Eurocontrol in the frame of a SESAR Project on GNSS, including a contract with Rockwell Collins for a feasibility study of the CRPA RFI localization concept. The French (DSNA/DTI) and U.S. FAA Flight Inspection service supported the project with expertise and in-kind contributions. The FAA conducted an overflight with a direction-finding-equipped aircraft for direct comparison between the CRPA approach and other, non-GNSS specific, commercial solutions.

TECHNOLOGY OPTIONS

Current, common GNSS CRPAs come in either 4- or 7-element variants. CRPAs always require antenna electronics for further processing of the RF inputs, and perform either nulling (steering negative gain towards RFI sources) or beamforming (steering positive gain towards GNSS satellites), or both. The most performant system is a 7-element CRPA in combination with digital beam-former antenna electronics. The 7-element CRPA has a diameter of 36 cm (14 inches), which is of some concern for installation on a typical flight-inspection aircraft such as the Beech King Air. But for a feasibility study, it makes sense to first evaluate the most-performing option. If there is unnecessary margin, the solution can be simplified afterwards.

A top-mounted solution on the airplane fuselage was retained due to experience with military anti-jam performance suggesting that RFI localization performance would be sufficient while retaining the benefit of stable ownship position. A key element of the assessment focused on how to best use aircraft banking to facilitate geo-localization.

As shown in Figure 1, the CRPA is connected to the Digital Integrated GPS Anti-Jam Receiver (DIGAR). As there is one RF cable per CRPA element, it is useful to install the DIGAR as close as possible to the CRPA. The standard military-production DIGAR contains not only the antenna electronics but also the receiver including baseband processing. For civil purposes, either a civil receiver would need to be integrated into the DIGAR or, alternatively, a single RF output is available to connect a standard civil GPS receiver. The DIGAR will also feed angle-of-arrival information into a direction-finder software.

Figure 1. System configuration. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

The software provides angle-of-arrival information with respect to the antenna/aircraft reference frame. To provide a geolocation capability, this must be combined with ownship position and aircraft attitude. As most flight inspection aircraft are equipped with an AHRS, this is not expected to be a problem. Project resources did not permit full integration, so testing was done using the direction-finder display only. The AHRS would need to provide 10–50 Hz updates with an error of not more than ±2 degrees.

Figure 2 shows an example of the direction-finder output. Lighter areas show where the antenna electronics produce negative gain, while darker areas represent stronger positive gain. The red dot indicates a potential interferer has been identified. Source location is at about 280 degrees of azimuth with respect to aircraft nose.

Figure 2. Excerpt from direction finder polar display of RFI signal angle of arrival. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

Correct detection probability will depend on the sensitivity threshold and associated false detection probability being considered acceptable. A visual localization may still be possible at carrier-to-noise density ratios (C/N0) below those needed to produce the red dot here, especially if the visible ambiguity can be removed through some aircraft maneuvering. It can be inferred from the system description that once the full integration is accomplished, the provision of a direct output using only a few lines of position to find a probable RFI source location in terms of approximate lat/long coordinates should be straightforward.

SIMULATOR TESTING

A well-calibrated simulator capable of feeding the seven RF inputs was used to assess detection performance for different flight patterns near an RFI source. The tested patterns include a rectangular, a circular and an oscillating, S-shaped trigger-and-hunt trajectory. A variety of different encounter scenarios in terms of power levels and free space path loss were tested. Power levels were adjusted to produce a 1-dB reduction in the C/N0. Both a continuous wave (CW) interferer at the L1 center frequency and a broadband (BB) interferer were simulated (using a 20-MHz-wide PSK signal). Figure 3 shows an example of achieved detection accuracies in both azimuth and elevation angle.

Figure 3. Example result of angular detection performance. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

While there is a strong peak within ±10 degrees of azimuth, there are also significant outliers. For the elevation (note the normalized scale), however, the main peak is thinner with even stronger sidelobes. Due to the installation of the antenna on top of the aircraft fuselage, the simulation results indicate that the elevation angle output is not very useful for detection. The time series result for the azimuth is given in Figure 4, where it can be seen that there are many good detection matches but also some “sympathetic nulls” that move in the opposite direction of the ground track truth reference (circled in grey). It is expected that with additional software processing, these sympathetic nulls can be filtered out.

Figure 4. Azimuth Time Series Result Corresponding to Figure 3. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

For all tested scenarios (assuming additional filtering), azimuth detection capability was better than ±10 degrees (one standard deviation), and in some cases as accurate as ±2 degrees. There was no significant difference between CW and BB results. As could be expected, simulated aircraft banking significantly improved detection capability. Consequently, the use of orbits seems to be the best search strategy. The simulator testing used a figure-eight pattern with one of the orbits passing over the interference source.

LIVE-SKY VAN TESTING

Rockwell Collins has an authorization to broadcast RFI test signals at the GNSS L2 frequency. Previous work showed that the results at L2 can be applied equally to L1. Figure 5 shows the test area, including a –100-dBm signal level boundary. The interferer was installed on a tripod and fed by a signal generator using a normal GPS fixed radiation pattern antenna (FRPA).

Figure 5. Live-sky test area. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

Locations B and C were used to both calibrate the RFI level and as check points for the van trajectory. The test van included a fixture that allowed a tilting of the CRPA by 30 degrees from zenith to either side. Figure 6 shows a schematic of the tilt fixture. It can be seen that this set up creates a realistic RFI path that arrives with an elevation slightly below the horizon at the unit under test. Two sets of tests were performed: one where the van drove straight into or out of the area of interference to determine overall equipment sensitivity, and varied paths to quantify angular detection performance. Again, both CW and BB RFI signals were evaluated.

Figure 6. CRPA with tilt fixture. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

Not surprisingly, elevation angle results turned out not to be very reliable given the below horizon signal path. But azimuth errors were slightly greater than obtained during the wavefront simulator testing (±12 degrees, one sigma). This can be attributed to both multipath and a less accurate heading truth reference. Taking these additional factors into account, the results are very consistent. Tilting the antenna by 30 degrees towards the RFI source significantly improves azimuth resolution (to about ±8 degrees) while also reducing sympathetic nulls. When the tilted antenna points away from the RFI source path, azimuth accuracy will decrease, which is considered helpful in avoiding false detections.

Summary. Even if a good bit of integration work remains necessary to produce a production-ready system for flight inspection or other similar aircraft, the approach shows promise. Further testing, especially using an actual aircraft installation, is recommended. Installation of a 7-element CRPA will be challenging on a typical Beech King Air, but possible. Antenna calibration requirements are expected to be manageable with a standard network analyzer. To avoid further complications with export regulations, the use of a separate civil GNSS receiver is recommended. The overall system is, at this stage, still on the costly side.

While a 4-element CRPA could be used, this was estimated to double or triple angular azimuth detection errors and reduce the detection distance, and consequently not likely to be worth the additional cost. While smaller 7-element CRPAs than the one used are available, their performance would need to be assessed.

For a top-mounted CRPA, aircraft banking is essential to ensure good performance. This could increase the amount of airspace required for detection and lead to operational complications. Furthermore, since the aim is to increase detection sensitivity to geo-locate weak power sources such as personal privacy devices, maintaining ownship position is not that critical, as it can be managed by maintaining an appropriate distance from the RFI source if needed. Consequently, both DSNA and FAA recommend using a bottom-mounted CRPA. In addition to adding 10 dB of detection sensitivity on average and reducing the need for maneuvering, it may restore the utility of the elevation output, thereby potentially further reducing search time. Either way, it will be useful for equipped aircraft to have alternate positioning capabilities to GNSS both for aircraft guidance and truth reference systems.

The system required a 15-dB stronger signal to transition from detection to localization. However, this is dependent on the accepted false-alarm rate. A tunable procedure can be envisaged where the software accepts a higher false-alarm rate at first to maximize search capability and moving to a lower alarm rate to confirm suspected RFI source locations later. Both the potential of the additional filtering software and any human-machine interface aspects would need to be further evaluated.

GENERIC CAPABILITIES

The two common options for in-flight detection of RFI sources in any relevant frequency band are the use of either a spectrum analyzer or, if available, a direction finder. The spectrum analyzer approach depends on connection to a suitable antenna, preferably with some directionality. In this way, the aircraft can be maneuvered to point the antenna either towards or away from the RFI source. Normally there is very little directivity, making this a challenging search. A direction finder is a significant improvement. Figure 7 shows the L-band antenna array used by a DF-4400 as installed on the bottom of the aircraft.

Figure 7. CRPA with tilt fixture. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

Newer generation spectrum analyzers with a good GNSS-specific pre-amplifier, using digital sampling with a fast A/D converter, could provide useful capability. However, the subject is beyond the scope of this discussion, and we focus here on comparing the CRPA approach with a standard direction finder.

The FAA Flight Inspection service conducted complementary flights during the Rockwell Collins live-sky van testing. The flights included orbits and a direct overflight of the RFI source. This was complemented by additional laboratory calibration to ensure that results could be compared. The sensitivity results of the CRPA approach are more meaningful in comparison with a generic direction-finder capability. Since test data is only available for a top-mounted CRPA, the comparisons here are made for the preferred bottom-mounted CRPA using engineering estimation.

The key finding was that while direction-finding capability was quite comparable between the CRPA- system and the DF-4400 for CW, the CRPA-system outperforms the DF-4400 by a significant margin when encountering broadband signals. This is considered to be a significant improvement given the expected nature of RFI sources. During the FAA overflight, the aircraft did not manage to detect the broadband signal. Consequently, the values given here are reconstructed from laboratory analysis. Table 1 compares the estimated achievable sensitivities.

Table 1. Comparison of direction-finding sensitivity. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

In view of the limitations of the data analysis performed, these values must be interpreted with caution. In general, we can conclude that the direction-finding sensitivity of the CRPA system is relatively insensitive to the encountered modulation of the RFI signal, and that the bottom-mounted CRPA system outperforms the DF-4400 system by a small margin in the CW case and by a large margin in the broadband case. How many additional dBs can be gained by both approaches through further optimizations is for future analysis. The performance improvement of the CRPA system does come at a cost, as could be expected.

DETECTION

Before the search for an RFI source can begin, it must be detected. Normally it should be easier to detect an RFI source than to locate it, since direction-finding requires a certain signal strength to obtain bearing information. However, given the directionality of DF arrays, this may not necessarily be true. Another potential factor is the reliance on a spectrum analyzer to detect RFI, which may not achieve the corresponding noise floor, especially when using a broad scan across a wide frequency range. The direction-finder system needs about a 15-dB difference between detection and localization ability.

Figure 8 shows the detection ranges for the top-mounted CRPA system for a given ground-based emitter while the aircraft altitude is assumed at 2000-ft AGL. The bottom mounted system would improve the minimum detection threshold further. Given that 15 dB can translate into a significant difference in free space path loss distance, concepts for efficient direction finding once an RFI source is detected deserve further attention.

Figure 8. Detection ranges for top-mounted CRPA system. Source By Gerhard E. Berz, Pascal Barret, Brent Disselkoen, Michael Richard, Vincent Rocchia, Florence Jacolot, Todd Bigham and Okko F. Bleeker

HUMAN FACTORS

During the FAA overflight, the broadband RFI couldn’t be detected by either the spectrum analyzer in use or the DF-4400. Part of the challenge was using the right equipment settings. For the DF-4400, it was found that best performance could be obtained for detecting broadband RFI when using the FM wide mode of demodulation. Similar findings were obtained for the use of the spectrum analyzer, where specific skills are necessary to use the equipment to its fullest capability. Similar issues are expected when having to interpret the display of a CRPA-based system. This means that regardless of the RFI source geo-location approach used, specific training should ensure that aircraft operators have the greatest chance of success in finding RFI sources.

CONCLUSIONS

An approach using a CRPA antenna, electronics and processing software proved superior to current, generic direction-finding capabilities, especially with respect to broadband signals. Maintaining ownship position in the presence of RFI is a secondary objective when looking for the expected weak signal sources, and the use of a bottom-mounted CRPA system is preferred. Additional filtering to eliminate sympathetic nulls and other issues require further investigation.

Significant benefit derives from employing aerial work aircraft in cooperation with ground-based capabilities. We recommend that equipment manufacturers further study all aspects of GNSS RFI geo-location and improve their capabilities. Such capabilities are expected to limit the exposure time to RFI cases and allow a more efficient deployment of ground-based spectrum enforcement resources. These studies should include the improvement of detection and localization equipment, and the development of corresponding operational procedures for flight crews.

ACKNOWLEDGMENTS

The Eurocontrol-funded contract with Rockwell Collins is part of the Eurocontrol contribution to SESAR Project 15.3.4, GNSS Baseline and the GNSS RFI Vulnerability Mitigation Task.

Rockwell Collins provided the DIGAR and Direction Finder Software.

This article is based on a paper presented at ION-GNSS+ 2016.

Disclaimer. This article does not contain any official Eurocontrol, SESAR, FAA or DSNA position or policy. It does not constitute any endorsement of a particular product, or a statement of any kind relating to any future procurement activity.

GERHARD BERZ and PASCAL BARRET work at Eurocontrol, Belgium; VINCENT ROCCHIA and FLORENCE JACOLOT with Direction des Services de la Navigation Aerienne, France; BRENT DISSELKOEN and MICHAEL RICHARD at Rockwell Colins, U.S; Okko F. Bleeker with OFBConsult System Engineering, the Netherlands; and TODD BINGHAM with the U.S. Federal Aviation Administration.