Abstract:

This talk discusses the integration of GNU Radio with the White Rabbit timing distribution system for creating distributed, phase-coherent Software Defined Radio (SDR) networks, primarily demonstrated through passive radar applications. Leveraging the White Rabbit framework, developed initially for high-energy physics (CERN, GSI) to achieve sub-nanosecond synchronization over long-baseline Ethernet networks, the presenter shows how multiple Ettus X310 SDRs can be synchronized for coherent signal processing. The system uses the French GRAVES space surveillance radar (143 MHz) as a passive illuminator to track aircraft by measuring Doppler shifts and direction of arrival (DoA) based on phase differences across spatially separated antennas. The importance of high-stability clocks (hydrogen maser) for resolving fine Doppler shifts is highlighted. Mathematical foundations linking spatial FFTs to DoA and matrix methods for handling sparse antenna arrays are briefly covered. Experimental results demonstrate successful synchronization over hundreds of meters, aircraft tracking (velocity, azimuth, elevation), and analysis of the GRAVES signal structure, including its beam-sweeping characteristics and limitations for high-precision frequency dissemination. The talk concludes by advocating for tighter integration of White Rabbit synchronization capabilities directly into SDR FPGAs (e.g., RF over White Rabbit concept) to simplify the creation of distributed coherent systems.

Distributed Coherent SDR with GNU Radio and White Rabbit

• 0:00 Introduction & Context: The talk focuses on using GNU Radio for radar applications, specifically leveraging White Rabbit synchronization, inspired by work presented at CERN and the unique co-location of GNU Radio and White Rabbit experts at GSI.
• 0:30 White Rabbit Synchronization: White Rabbit is a timing system developed by high-energy physics labs (CERN, GSI) to synchronize devices like SDRs over networks with sub-nanosecond precision, compensating for time-of-flight delays.
• 1:00 GSI Collaboration: Utilizing GSI-developed White Rabbit PCI Express boards to synchronize Ettus X310 SDRs for coherent measurements.
• 2:10 Narrowband Radar Techniques: While lacking range resolution, narrowband radar excels at interferometric (phase) measurements and fine Doppler shift detection for high-accuracy velocity measurements, crucial for the presenter's time/frequency lab background.
• 3:14 Passive Radar Setup: Using the French GRAVES radar (400 kW at 143 MHz, 40km away) as a non-cooperative illuminator for detecting flying objects (planes, satellites) via Doppler shifts.
• 4:21 Coherent Processing Need: Without range information, location requires Direction of Arrival (DoA) or Time of Flight (ToF) measurements using multiple, synchronized receivers (multistatic radar).
• 5:03 DoA Estimation: DoA can be determined from the phase difference of signals arriving at multiple antennas. For a Uniform Linear Array (ULA), this is analogous to a spatial Fourier Transform.
• 6:51 Sparse Arrays & Matrix Methods: The Fourier Transform approach assumes dense, periodic sampling (many antennas). A matrix multiplication formulation allows DoA estimation even with sparse, irregularly spaced antennas, provided their positions and signal timestamps are known. This introduces side lobes and potential ambiguities.
• 10:43 Experimental Synchronization: Successfully synchronized X310 SDRs across the campus (up to 600m apart) using GSI White Rabbit boards and optical fiber, achieving phase lock despite significant fiber path lengths (compensated by White Rabbit).
• 13:51 Hardware Latency: Caution: Even with identical firmware (White Rabbit core), different hardware implementations can introduce nanosecond-level timing offsets that need calibration for highest precision.
• 15:05 GNU Radio Implementation: Using the Multi-USRP source block in GNU Radio is essential for synchronizing multiple USRPs. Signal processing involves capturing the aliased GRAVES signal (143 MHz sampled at 200 Msps) and heavy decimation using cascaded FIR filters.
• 16:33 Data over White Rabbit: The White Rabbit network (Gigabit Ethernet) can also be used to stream the decimated SDR data back to a central processing point.
• 18:02 Passive Radar Results: Waterfall plots show detected aircraft tracks via their Doppler shifts. A stable reference clock (Hydrogen Maser) ensures the direct GRAVES signal is a stable line, enabling fine Doppler measurements.
• 19:36 Coherent Measurements: Phase differences between synchronized antennas show continuous variations corresponding to the changing DoA of aircraft.
• 20:48 Azimuth/Elevation Tracking: Using a square array of four antennas, both azimuth and elevation of aircraft can be estimated from the phase differences.
• 23:14 Sparse Array Ambiguity: Using widely spaced antenna pairs (e.g., 7.5m apart) increases angular resolution but introduces ambiguity (multiple parallel tracks in the DoA plot due to aliasing).
• 24:21 GRAVES Signal Analysis: The phase of the direct GRAVES signal reveals discrete jumps, indicating its electronically steered beam array (8 beams identified). This limits its use for ultra-precise time/frequency transfer.
• 25:20 Frequency Transfer Stability: Allan deviation analysis shows the GRAVES link stability is around a few parts in 10^-11, limited by the propagation path noise, not the source or receiver clocks.
• 26:38 Future Vision: Integrated Synchronization: Proposing tighter integration of White Rabbit core logic directly into SDR FPGAs (like CERN's RF over White Rabbit concept), potentially eliminating external clock synthesizers and making distributed coherent SDR systems easier to build.