SINTEF RFI ACTIVITIES June 2019, Helsinki Aiden Morrison, Nadia Sokolova
Outline • Brief SINTEF Introduction • Our motivations for pursuing RFI monitoring • Past projects covering GNSS interference • Lessons learned, and knowledge gaps • Ongoing activities, and upcoming monitoring hardware 2
3
Navigation Focused Research Areas RFI Monitoring GBAS Indoor Navigation Custom Navigation HW Autonomous system System Design Navigation GNSS for ITS 4 Image sources: [3]
Our motivation for pursuing RFI monitoring • While all GNSS applications are vulnerable to RFI, some have higher potential losses • If my phone loses GNSS due to RFI, I might be slightly inconvenienced • If our drone loses GNSS due to RFI, we might lose a very expensive drone or crash in to important infrastructure • Example of an economic impact • Some applications have higher risks 5
CAT III GBAS-based Automatic Landing GBAS - Ground Based Augmentation System Concept of operation: • Similar to differential GNSS but with extras • Corrections and safety flags are carried from the ground to the aircraft via a dedicated VHF link called VDB • There are typically four ground stations which allow for much more sensitive monitoring for outliers and errors • But the system is very sensitive to signal degradation from RFI 6
Precision Approach Requirements GBAS Precision Operation CAT l CAT ll CAT lll Accuracy [m] Horizontal 16.0 6.9 6.1 Vertical 7.7 2.0 2.0 95 % Time-to-Alert 2 3 2 [s] H: 40 H: 17.3 H: 15.5 Integrity Alert Limit [m] V: 10-15 V: 5.3 V: 5.3 2x10 -5 2x10 -9 2x10 -9 P HMI / approach 10 -7 / 15 sec 5x10 -5 / 5x10 -6 / 15 sec Continuity Failure Rate approach Availability 0.99 – 0.99999 0.99 – 0.99999 0.99 – 0.99999 • Category type identifies system capability, indicating the minimum approach height that can be achieved. • Hitting these performance metrics is hard under ideal conditions. • If weak RFI is present on even one of the four antennas it can reduce availability and continuity • Stronger RFI is an immediate problem – there are several recent examples 7
GBAS and other aircraft events (plus a fancy boat) 8
What we need to know • From the point of view of GBAS, we need to know a few pieces of information about RFI • 1) The occurrence rate • 2) The relative occurrence rate • Future GBAS systems are intended to use both the L1/E1 and L5/E5a bands on GPS and Galileo • In the event of RFI on a single frequency it may be possible to fall back to operation using only the other • 3) Power levels • 4) Spectral occupancy • Unintentionally generated tones are mitigated by the signals themselves • It may be possible to identify individual emitters over differing events and positions • Unfortunately real-time analysis at this level can be computationally expensive • Instead, we take an alternative approach 9
Event detection - 1 • RFI detection concept – how to detect RFI with minimal processing power? • It’s not feasible to analyze all signals in real time in terms of C/N0 – that would require a full multifrequency GNSS receiver and wouldn’t provide us a bitstream for our SDR • Instead we look for in band power level deviation • Must allow for site to site variation and slow variability due to thermal effects etc. But rapid increases in power are strong indicators of RFI • Weak events visible • Strong events up to -10dBm 10 • Example of car-borne jammer events
Event detection - 2 • Mobile Jammer • Cigarette lighter style • 'Stepped CW' signal • Car drove within 10 m • Test conducted with NKOM and FFI • Other monitors deployed • Indra Navia, DFS, NLR, ESTEC • The monitor at ESTEC triggered frequently on adjacent band pulsed power sources (See Event 000) • Also caught a delivery truck or taxi with a jammer • Event 008 is a jamming event • GNSS satellites cannot be tracked in L1 band for duration • There are Drawbacks to this approach 11
Lessons learned & Knowledge gaps • Three major factors motivated the design of a second generation monitoring system • The Manual result collection and analysis required a large amount of user intervention • Far too many "false alarms" • E5b, L2, G2, E6, E1PRS, and G1 not collected/analyzed • These signals all have slightly different characteristics making them more or less suitable to different use cases • Let’s see what this coverage looks like on a signal map 12
RF signal background - 3 of 3 • A chart is helpful: • Ignoring the S-band signals, this chart shows the L-band • Signal plans evolve over time • Uncertain if the GLONASS CDMA plans are still accurate • Most of these signals are now turned on and ’healthy’ • The rise of RFI • As GNSS has risen to solve more and more problems, it has become a larger target • People are right to be paranoid about being tracked, but jamming GNSS is hazardous • If usage based road tolling depends on GNSS it will become a target • This is why we need a radically different front-end • From 2x24 MHz to 4 x 55 MHz • This is a nearly 5x increase in the amount of data • For reasons we will cover later we also want to go from 1 bit to 4 bit • 4x again so ~20x the data relative to the first system • A higher dynamic range is also desirable 13
Quantization concepts and requirements - 3 of 3 • Our design needs to tolerate 6 or 7 orders of magnitude • (Images from wikipedia, Nuand, and Maxim) • 60-70dB+ dynamic range • Each bit of ADC range represents 3dB of amplitude dynamic range, 6dB of power • Each bit of ADC range takes up more data whether or not the signal is using these bits • Therefore – it’s inefficient to throw bits at the problem outside of processing applications that need or benefit from having a wide dynamic range signal • E.g. Novatel 7 series receivers have 8 bit quantization but only turn it on when requested for RFI mitigation, otherwise they use 4 bits to save energy • Instead we need to employ what is called ’gain control’ to feed back from the digitized signal to the analog gain levels • Dynamic range considerations map back to every stage of the design • IP3 etc. Of first gain stage • Maximum power handling of components (SAWs and LNA) • Representation range of the ADCs • Gain Control and feedback to use a VGA • The MAX2120 mixer chip includes a VGA with a 70 dB analog linear range • Also 15dB of digital control range at the baseband, for a combined 85dB • Probably not feasible to use the full range, but still a good start 14 • System block diagram
Design Evolution - Concept • System automation, and updated hardware capabilities Indicator LEDs, and Buttons User configuration options VGA, Keyboard/Mouse Mixer, OCXO, Ethernet Cloud Storage, (USB2) LPF 1 SAW, LNA, Archiving and Retrieval Clocking E1/L1/G1 Splitter GNSS Antenna Ethernet Notifications Raw and Monitor (HDMI) RGMII VGA, (e.g. Email) Interface processed data Mixer, i.MX8 RF Power LPF 2 SOM Meter A Hardware: Embedded LNA, USB3 8x ADC FPGA USB3 computer module (COTS) Splitter FIFO RF via USB3 - RF Power VGA, Hardware: Reconfigurable Meter B Notifications Mixer, USB3 Software: Online multiband front-end. Detection, (e.g. Email) LPF 3 monitoring, analysis and and pre- notification. Periodic SAW, LNA, PCIe to SATA2 processing Other Low Level PCIe SATA2 archiving and reporting. E5/L2/E6 Splitters SATA2 IC SSD VGA, software Interfaces Other Low Level Mixer, Interfaces to FPGA LPF 4 Power In (<90W) 15
Design Evolution - Hardware • First version powered up and ran code • Now in the process of implementing firmware • Why not just buy an off-the-shelf front-end? • A few small reasons • 1) None have integrated power level measurement for environment characterization and triggering • 2) Onboard oscillator and clock generators need to be low phase noise components for GNSS signals • 3) GNSS Pre-filtering /Sensitivity to adjacent band signals: • Front-ends like HackRF, bladeRF, and LimeSDR focus on wide bandwidth • This makes sensitive pre-triggering difficult and pushes up false alarm rates • 4) Many off the shelf solutions use fractional-N PLLs in their mixers • Can introduce large residual code-carrier divergence 16 • Deployment plans...
Deployment plans • Multiple locations in Norway and Europe (e.g. research facilities at ESTEC and NLR). • Several sites to be operated by the Norwegian Public Roads Administration. SINTEF monitors will be co- located with monitoring equipment deployed by Nkom. Patterødkrysset E8 Kilpisjärvi-Skibotn (Busiest intersection in Norway) Road used for autonomous testing 17
Future activities • Starting next year we hope to have phased array based RFI detectors which provide bearing information to the host system • These will be tested for drones based detection and tracking of GNSS RFI sources • I want the drones to be able to ’remove’ the source, but unlikely to get HMS approval 18
Policy outlook • If GNSS based road pricing is adoped there will very likely be a large spike in RFI events • It produces a financial incentive • It will be necessary to have a comittment to both prevention and response • Prevent jamming equipment from entering the country • The credible ability to rapidly detect and localize jamming 19
Teknologi for et bedre samfunn
Recommend
More recommend