As the Ground Based Augmentation Systems (GBAS) have to often operate in close proximity to high trafic roads, parking garages, and other transportation infrastructure, the chances of being afected by unintentional Radio Frequency Interference (RFI) from low-cost jamming devices and malfunctioning equipment are high. Multiple recent studies have indicated that interference levels in such environments are frequently exceeding the tolerable limits for safety-critical system/infrastructure operations. It is therefore important for GBAS to have efective RFI detection and mitigation mechanisms. This article presents sample results and analysis of GNSS receiver responses to diferent RFI signal types generated based on real life observations made as part of multi-dimensional parameter spaces extracted from a multi-site, multi-year monitoring campaign carried out in Europe and Scandinavia. The obtained results are discussed in light of RFI monitor design in support of the dual frequency multi constellation (DFMC) GBAS ground system providing initial recommendations for the potential monitoring scheme with respect to which observables should be used and how they interrelate over the various RFI signal types.
While GNSS signals used in aviation are located in a protected Aeronautical Radio-Navigation Signal
(ARNS) frequency band, jamming afecting this band is occurring frequently as it has been shown by
multiple projects including the EU H2020 STRIKE3 [
To support the in-lab testing carried out in this work, data captured by the ARFIDAAS GNSS RFI
monitoring network was used as the basis. The network comprised 14 sites in Europe and Scandinavia,
where each site is equipped with a low-cost monitoring unit supporting all navigation bands transmitted
by GPS (L1, L2 and L5), Galileo (E1, E5a, E5b and E6), GLONASS (G1, G2 and G3) and Beidou (B1, B2
and B3). For more details about the monitoring unit design as well as event statistics and analysis see
[
As presented in [
The event classification results accumulated across sites and over time allows one to study the observed
jamming signal characteristics, define key parameter sets for each of the main signal categories, as
well as define parameter ranges based on real life observations. Tables 1-3 show parameter ranges for
the three of the most frequently observed categories [
Event shown in Figure 2 (a) for example was classified as multilevel chirp, and while there are some features in it resembling one, it is dificult to say why such a signal structure would be used to jam GNSS as it might not be as eficient as a less complex signal modulation. For this reason, and based on the fact that the center of the RFI is located several MHz above the band center of GLONASS G2, it is believed that this signal is unintentionally generated. The example shown in Figure 2 (b) is a special case as it contains three distinct perturbing signals at power levels low enough that the main lobes of the GPS and Galileo L1/E1 signals are still visible. While the weak CW signal is of no note, the simultaneous occurrence of two distinct time modulated signals is. In addition to the traditional linear chirp signal, an exponentially swept signal is present.
This exponentially swept form was not observed during the first two years of the ARFIDAAS network operation and therefore was not defined as a separate signal type, however, by the end of the fourth year of operation this signal type was present with some regularity at multiple sites. While such a scenario containing multiple diferent RFI types plus a novel modulation which the classifier was not designed to isolate represents a significant challenge to classification, the classifier successfully categorized this event as a form of time-modulated interference. The identified parameter ranges as shown in Tables 1-3 have been used to define the limits of the simulation space. In this paper, we focus on a selected subset of results from the category of time-modulated (single linear, single exponential, and linear in combination with exponential chirp simultaneously as shown in Figure 2 (right)). The motivation for focusing on the dual signal case comes from a concern that this newly observed RFI combination may have been selected by malicious actors due to particularly disruptive impacts to GNSS signal tracking. In order to establish a baseline for comparison, it was decided to also test the traditional linear chirp impacts over their observed parameter space as linear chirp is by far the most frequently observed time-modulated RFI type in the wild. A particular note is that these evaluations focus on threshold impact power levels to stimulate disrupted tracking behavior in the receivers while allowing them to maintain signal lock so as to still provide observables/measurements for analysis. Since the GNSS receiver devices under test (DUT) must be run in real-time and require both initial acquisition time as well as a reset interval to ensure independence between test cycles, the number of simulation cycles and therefore the granularity of the 3D simulation space had to be constrained. With each simulation cycle desired to produce 60 seconds of RFI impacted output data, a single simulation cycle requires 5 minutes.
Given a simulation grid of 13 bandwidth steps, 17 repeat rate steps, and 8 diferent power levels the aggregate simulation cycle reaches 2.5 days for a single scenario such as linear chirp, exponential chirp or the combined dual chirp. The amount of time is tripled in the case when the performance is evaluated on two frequency bands individually and combined (i.e. L1, L5, and L1 + L5).
To generate the jamming signal types observed live, the equipment setup illustrated in Figure 3 was used. A HW GNSS Spirent GSS800 simulator was supplemented by two Ettus Research universal software radio peripheral (USRP) B200 units for jamming signal generation. The USRP radios used have an adjustable transmit power gain between 0 and 89 dB, where a gain setting of 15 dB corresponds to an increase in total received power in band of 1 dB at each of the connected receivers. The supported signal bandwidth can go up to 50/55 MHz. When two jamming signals are simulated in the same frequency band, this setup allows for digital adjustment of the power level between signal components sent by the same radio. While this test bench equipment setup allows for the evaluation of four receivers simultaneously, the results in this paper focus on observations made from two geodetic grade receivers (NovAtel OEM7 and Septentrio mosaic-T). To confirm the capability to produce complex combined signal events prior to running a large number of simulations, the equipment setup was thoroughly tested to ensure that the interference signals were generated correctly. For this purpose, the ARFIDAAS front-end was used to capture and analyze the combined signal outputs.
While this test bench equipment setup allows for the evaluation of four receivers simultaneously, the results in this paper focus on observations made from two geodetic grade receivers (NovAtel OEM7 and Septentrio mosaic-T). To confirm the capability to produce complex combined signal events prior to running a large number of simulations, the equipment setup was thoroughly tested to ensure that the interference signals were generated correctly. For this purpose, the ARFIDAAS front-end was used to capture and analyze the combined signal outputs.
GBAS development must be carried out in accordance with rigorous standards therefore system evolution process is complex and time demanding with dual frequency architecture still in development phase. Design of an appropriate RFI monitoring scheme for a dual frequency system is therefore still a topic for investigation. While there is no dedicated RFI monitor, interference detection in GBAS ground reference receivers is typically carried out by monitoring the variation of the Carrier-to-Noise density ratio (C/N0) or various metrics derived from it in the form of the required low signal power monitor. While other signal/measurement quality monitors might be triggered by RFI, their purpose is to ensure integrity of service in the presence of other specific threats, so while the faulty measurements/ranging sources will be excluded ensuring required performance, this information is typically not used for flagging the presence of RFI. This misattribution could lead to incorrect assumptions about the environment at given sites and otherwise lead to overlooking mitigation options such as berms and fences which would potentially resolve the RFI. Although the use of the Automatic Gain Control (AGC) outputs and histograms of analogue-to-digital converter (ADC) bins for RFI monitoring have been demonstrated to be very efective and sensitive, their availability is not guaranteed. While any receiver that utilizes multi-bit sample quantization would be expected to contain an AGC if not a histogram engine, these measurements are not always output by the receiver/visible to the user. With specific consideration of GBAS, currently there is no certified dual frequency GBAS ground receiver available, while the available certified single frequency GBAS receivers do not output either of these metrics. In this work we have therefore focused on the known available/supported measurements. It is also noted that while additional RFI suppression measures are desirable (e.g. use of the Controlled Reception Pattern Antennas), their use in GBAS may introduce calibration errors that will degrade system integrity and availability even if these measures will reduce the system’s vulnerability against RFI.
As mentioned above, the results presented in this article focus on a sub-set of interference signal types, namely single linear, single exponential, and linear in combination with exponential chirp simultaneously. In the case of a single signal, the identified parameter set space is covered for the considered signal type. The dual signal case covered in this article is represented by the combination of a variable exponential chirp signal, meaning that for this signal type in each test a diferent sub-set of parameters is used, while the modulation parameter set for the linear chirp signal is kept the same at 200 kHz repeat rate and 20 MHz bandwidth. The power level of both signals is varied with constant relative ofset. Since the primary interest of this study is the relative impact of the various jamming signal modulations over their parameter space, the results are presented to emphasize trends and not the absolute performance. High level and expected results include the observations that diferent receivers respond diferently to the generated RFI types and regions of the parameter space. Due to reasonably expected diferences in the C/N0 estimators and the tracking loop designs and parameters implemented in the receivers, their responses to various parameter combinations were notably diferent. In GBAS, this factor is taken into account by characterizing the performance of a specific certified receiver model when designing fault monitoring algorithms. One observation of note is that the impact of certain RFI modulation and parameter combinations result in diferentially observable efects between the C/N0 degradation and the carrier phase uncertainty amplification. This is illustrated in Figures 4 and 5in the case of a single linear chirp signal where the NovAtel OEM7 receiver reports a near random but low loss of signal power for all RFI parameters for both GPS L1 C/A and Galileo E1 signals, but high noise in the carrier phase measurements in response to lower sweep bandwidths and lower repeat rates for GPS L1 C/A, while Galileo E1 is sensitive to bandwidth, but insensitive to the repeat rate. This scenario has the interference signal power level injected which raises the received total power in band by 1 dB. This is a very low power RFI signal, so the observation is of note in that even a very weak RFI signal causes both observable degradation as well as noticeable diferential observable behavior. The same behavior trend has been observed when higher linear chirp signal power was considered though with a more frequent and longer lasting loss of observables starting with frequent cycle slipping and progressing to complete loss of lock.
The same general observation holds when considering the exponential chirp signal as it impacts the GPS L5 tracking by the same receiver. Figure 6 shows the C/N0 degradation and the carrier phase measurement uncertainty observed in a scenario with a higher level of in band power (7 dB). Here the C/N0 degradation is sensitive to only the bandwidth of the exponential chirp, while carrier phase uncertainty is particularly sensitive to only a small region of the bandwidth and the repeat rate parameter space. While it is believed that the enhanced sensitivity of the L5 signal tracking to certain repeat rates of the exponential chirp RFI is related to the symbol rate of the signal, the same cannot be said for the L1 and E1 signal responses to the same RFI type. As shown in Figure 7 illustrating the results captured by the Septentrio mosaic-T receiver from the combined simultaneous linear and exponential chirp signals, the C/N0 loss and the carrier phase noise responses of the L1 and E1 signals match in their trends over the parameter space and the peak disruptions correspond to the parameter set where the quasi-constant frequency portion of the exponential sweep overlays a modulation main lobe. These sensitivities manifest regardless of the presence of the simultaneous linear chirp signal, meaning they are the function of the exponential chirp and not the combination of the two signal types. The same general observation holds when considering the exponential chirp signal as it impacts the GPS L5 tracking by the same receiver. Figure 6 shows the C/N0 degradation and the carrier phase measurement uncertainty observed in a scenario with a higher level of in band power (7 dB). Here the C/N0 degradation is sensitive to only the bandwidth of the exponential chirp, while carrier phase uncertainty is particularly sensitive to only a small region of the bandwidth and the repeat rate parameter space. While it is believed that the enhanced sensitivity of the L5 signal tracking to certain repeat rates of the exponential chirp RFI is related to the symbol rate of the signal, the same cannot be said for the L1 and E1 signal responses to the same RFI type. As shown in Figure 8 illustrating the results captured by the Septentrio mosaic-T receiver from the combined simultaneous linear and exponential chirp signals, the C/N0 loss and the carrier phase noise responses of the L1 and E1 signals match in their trends over the parameter space and the peak disruptions correspond to the parameter set where the quasi-constant frequency portion of the exponential sweep overlays a modulation main lobe. These sensitivities manifest regardless of the presence of the simultaneous linear chirp signal, meaning they are the function of the exponential chirp and not the combination of the two signal types.
Based on the analyzed results presented in this article, initial recommendations for the DFMC GBAS
monitoring scheme based on observables available within GBAS implementations include the need to
monitor for the impacts of RFI separately on each individual signal modulation even in cases where those
modulations are spectrally overlapped as indications are that subtle interactions of signal parameters
and tracking strategies can lead to substantial diferences in impact. Further, it is believed that the use of
existing monitors to isolate the RFI impact would imply further work due to the lack of uniformity in the
impact over the parameter space on the observables generated. This variable impact leads to triggering
of diferent monitors for diferent RFI modulations and varied parameters within a modulation. For
this reason, monitoring of RFI directly via automatic gain control feedback or other novel monitoring
is desirable though acknowledged to require new outputs be made available. While the authors still
believe that the exponential chirp signal is tuned to defeat certain types of RFI mitigations as previously
discussed in [
The authors would like to thank the European Space Agency (NAVISP program) for funding the development of the ARFIDAAS system used for live data capture, as well as the Norwegian Council of Research (project number 344275) and HEU/EUSPA EDGAR project (project number 101130407) for funding the in-lab simulation environment development and data analysis.
The author(s) have not employed any Generative AI tools.