PaperlayerView exact revision on arXiv

Source-linked AI summary

Full-Duplex OFDM Radar With LTE and 5G NR Waveforms: Challenges, Solutions, and Measurements

Carlos Baquero Barneto, Taneli Riihonen, Matias Turunen, Lauri Anttila, Marko Fleischer, Kari Stadius, Jussi Ryynänen, Mikko Valkama

arXiv:1908.03418v1eess.SP

TL;DR

The paper addresses missing LTE/NR subcarriers and limited TX–RX isolation in OFDM radar sensing. It combines interpolated frequency-domain processing with tailored RF and digital self-interference cancellation, showing high-quality sensing and approximately 100 dB measured isolation.

  • Problem

    LTE and NR radar processing must handle unused passband subcarriers and severe TX–RX self-interference, especially in shared-antenna base stations.

  • Method

    The paper uses known LTE/NR downlink resource grids, time–frequency interpolation, and OFDM-specific RF and digital cancellation for frequency-domain radar processing.

  • Results

    Simulations and RF measurements demonstrate target detection and range–velocity estimation for static and moving targets, with measured TX–RX isolation of approximately 100 dB.

  • Takeaways & Limitations

    5G NR bandwidths and configurable subcarrier spacing provide good radar performance, while cancellation is particularly important for static and slowly moving targets.

Abstract

from arXiv · show

This paper studies the processing principles, implementation challenges, and performance of OFDM-based radars, with particular focus on the fourth-generation Long-Term Evolution (LTE) and fifth-generation (5G) New Radio (NR) mobile networks' base stations and their utilization for radar/sensing purposes. First, we address the problem stemming from the unused subcarriers within the LTE and NR transmit signal passbands, and their impact on frequency-domain radar processing. Particularly, we formulate and adopt a computationally efficient interpolation approach to mitigate the effects of such empty subcarriers in the radar processing. We evaluate the target detection and the corresponding range and velocity estimation performance through computer simulations, and show that high-quality target detection as well as high-precision range and velocity estimation can be achieved. Especially 5G NR waveforms, through their impressive channel bandwidths and configurable subcarrier spacing, are shown to provide very good radar/sensing performance. Then, a fundamental implementation challenge of transmitter-receiver (TX-RX) isolation in OFDM radars is addressed, with specific emphasis on shared-antenna cases, where the TX-RX isolation challenges are the largest. It is confirmed that from the OFDM radar processing perspective, limited TX-RX isolation is primarily a concern in detection of static targets while moving targets are inherently more robust to transmitter self-interference. Properly tailored analog/RF and digital self-interference cancellation solutions for OFDM radars are also described and implemented, and shown through RF measurements to be key technical ingredients for practical deployments, particularly from static and slowly moving targets' point of view.

I. INTRODUCTION

The paper investigates using standard LTE and 5G NR downlink waveforms for joint communications and radar sensing, focusing on missing subcarriers and TX–RX self-interference. It develops interpolation and cancellation approaches and evaluates their sensing performance.

  • RF convergence combines communications and radio-based sensing on shared frequency bands and potentially shared hardware.
  • The proposed radar uses standard-compliant LTE and 5G NR downlink signals from TDD base stations as monostatic sensing waveforms.
  • Time–frequency interpolation reduces radar-processing effects caused by null or unused subcarriers in LTE and NR transmit grids.
  • Limited TX–RX isolation is especially challenging in shared-antenna monostatic base stations because transmitter self-interference can mask target reflections.
  • Tailored RF and digital cancellation suppress self-interference while preserving target reflections, with measurements demonstrating static and moving-target sensing.

II. OFDM RADAR:

The radar processes known LTE or NR transmit-grid samples in the frequency domain to detect targets and estimate their ranges and relative velocities. It compares matched-filter and channel-estimation-like processing for forming radar data.

  • System Model and Basic Subcarrier Processing: The LTE/NR frequency-domain resource grid X contains known samples from the composite downlink waveform across S active subcarriers and R OFDM symbols.
  • System Model and Basic Subcarrier Processing: Target propagation delays create subcarrier-dependent phase shifts for range estimation, while Doppler shifts create symbol-dependent phase shifts for velocity estimation.
  • System Model and Basic Subcarrier Processing: The received grid Y is processed either by matched filtering, GMF = Y ⊙ X*, or element-wise channel estimation, GCH = Y ⊘ X.
  • System Model and Basic Subcarrier Processing: Channel-estimation-like processing produces a data-independent noiseless component, whereas matched filtering optimizes individual-target SNR but can have data-dependent profiles and sidelobes.

B. Interpolation, Target Detection and Range–Velocity Estimation

Unused subcarriers in LTE and NR grids disrupt the exponential structure used for radar estimation. The paper fills these gaps through linear interpolation across OFDM symbols before constructing range–Doppler profiles.

  • B. Interpolation, Target Detection and Range–Velocity Estimation: Unused LTE and NR subcarriers interrupt the samples used to form sums of exponentials for target-distance and velocity estimation.
  • B. Interpolation, Target Detection and Range–Velocity Estimation: The method linearly interpolates an unused subcarrier across OFDM symbols when that subcarrier is active at bounding symbols q1 and q2.
  • B. Interpolation, Target Detection and Range–Velocity Estimation: Target detection and range–velocity estimation use the range–Doppler profile calculated from the interpolated channel-estimation grid.

( ¯GCH

The range–Doppler image is computed with configurable transforms, windows, and a threshold detector, after which detected peaks provide range and velocity estimates. Search regions can be restricted to feasible operating bounds.

  • Inner IFFTs produce OFDM-symbol range profiles, while outer FFTs produce velocity profiles from the processed grid.
  • Window weights control sidelobe levels, and transform searches can be limited to chosen maximum distances and velocities.
  • The detector declares a target when periodogram value A(s, r) exceeds threshold Tth, with CFAR setting the threshold for a chosen total false-alarm probability.
  • For a detected single target, the periodogram peak location is the corresponding maximum-likelihood range and velocity estimator.
  • Peak-refinement interpolation can improve resolution beyond the basic range–Doppler map pixel size.

C. Reference Performance with LTE/NR Carriers

Simulations evaluate LTE and NR waveform configurations for single-target detection and range/velocity estimation, showing that larger NR bandwidths support high-precision radar performance.

  • Waveform configurations: The simulations compare LTE and NR waveforms using 20 MHz, 40 MHz, and 100 MHz carrier bandwidths with 15 or 30 kHz subcarrier spacing.A 10 ms downlink radio frame is transmitted and processed at a 3.5 GHz center frequency.
  • Evaluation setup: The reference evaluations randomly vary a single target over 20–200 m and −40 to 40 m/s, with a total false-alarm probability of 10%.The feasible distance and velocity spaces are bounded by cyclic-prefix and subcarrier-spacing considerations.
  • Detection performance: Larger bandwidths provide greater subcarrier-based processing gain against thermal noise, improving detection at a given receiver input SNR.The detection comparison is shown in Fig. 2(a).
  • Estimation performance: At SNR values with high detection probability, distance and velocity estimation errors are well behaved and converge toward resolution-limited RMSE values as detection approaches 100%.The limiting behavior reflects uniform error within the range–Doppler pixel width.
  • NR capability: NR’s large carrier bandwidths support good target estimation accuracy and precision, particularly for distance, while higher center frequencies can further improve distance and velocity resolution.NR deployments at 24–40 GHz can provide continuous carrier bandwidths up to 400 MHz in 3GPP Release 15.

A. TX–RX Isolation and Self-interference Problem

TDD-based LTE and NR radar operation requires simultaneous same-frequency transmission and reception, making TX–RX isolation and self-interference especially difficult in shared-antenna base stations.

  • Simultaneous operation: Because downlink transmission and reception overlap in the radar concept, the receiver must operate simultaneously with the transmitter at the same carrier frequency.The minimum downlink allocation contains seven OFDM symbols, or one 0.5 ms slot with 15 kHz subcarrier spacing.
  • Isolation hardware: This operation requires more elaborate circulator or electrical balance duplexer circuitry instead of ordinary TDD RF switching to provide TX–RX isolation.The challenge is particularly pronounced when the eNB/gNB shares one antenna system between transmission and reception.
  • Self-interference severity: A macro base station can transmit at +46 dBm while the receiver noise floor is −97 dBm, creating a 143 dB power difference before accounting for other difficulties.High transmit power and waveform PAPR create fundamental saturation risks for the LNA and receiver.
  • Processing implications: In radar images, direct self-interference receives the same processing gain as target echoes, so noise processing gain alone cannot resolve the isolation problem.More than 100 dB of total self-interference suppression is required.
  • Cancellation requirements: Active RF and digital cancellation methods are needed alongside passive isolation and radar-domain digital suppression, particularly for static and slowly moving targets.The requirement also serves to prevent receiver saturation.

B. RF and Digital Cancellation Solutions

The paper combines active RF cancellation with time-domain digital cancellation, preserving target echoes while suppressing direct transmitter self-interference. A nonlinear digital canceller and carefully bounded memory are used to address RF nonlinearities and avoid cancelling nearby target reflections.

  • The cancellation architecture suppresses direct self-interference while preserving echoes from true targets, complementing radar-domain interference-suppression methods.The RF and digital stages are agnostic to the specific radar processing approach and can be combined with radar-domain methods.
  • The RF canceller uses PA-output references, parallel RF delays, vector modulators, and adaptive amplitude-phase control to estimate and suppress direct self-interference.The implemented digital control tracks complex tap coefficients using sampled I/Q observations.
  • A three-tap RF canceller models frequency-selective direct coupling while avoiding cancellation of echoes from targets more than 1.5 m away.Its maximum delay is 10 ns, corresponding to approximately 3 m equivalent distance.
  • The nonlinear digital canceller uses memory-polynomial basis functions and adjustable precursor/postcursor taps to suppress residual direct self-interference, including RF nonlinearities.Basis functions use odd orders through P, and parameters can be estimated with least-squares or adaptive filtering methods.
  • A self-orthogonalizing learning rule avoids explicit basis-function orthogonalization, greatly reducing the computational complexity of the main digital cancellation path.The rule uses a precomputable correlation matrix while adapting the canceller coefficients.
  • The digital canceller uses M1 = 5 precursor and M2 = 5 postcursor taps, whose postcursor memory must be selected to avoid suppressing reflections within the detectable radar range.The amount of postcursor memory is directly related to detectable radar range.

A. Measurement Setup

The measurements use a shared-antenna 5G NR transceiver, RF canceller, digital control hardware, and directive antenna system to evaluate TX–RX isolation. The experiments compare cancellation stages and digital-canceller configurations under a 40 MHz NR waveform.

  • The RF measurement setup comprises a vector signal transceiver, 42 dB external power amplifier, developed RF canceller, digital control boards, and a shared TX/RX horn antenna.The antenna and circulator provide approximately 25 dB of passive TX–RX isolation at the 2.4 GHz ISM band with about +20 dBm transmit power.
  • The experiments use a 5G NR waveform with 30 kHz subcarrier spacing, 40 MHz bandwidth, 2048-point FFT/IFFT, and a 240 MHz digital-front-end sample rate.The digital canceller operates at the 240 MHz rate, with linear and nonlinear orders evaluated.
  • The measured cancellation stages include RF cancellation, linear digital cancellation, and nonlinear digital cancellation, with representative output powers of −60.1 dBm, −78.7 dBm, and −84.7 dBm, respectively.The corresponding transmit signal and RF-canceller input powers are 15 dBm and −9.74 dBm.
  • Fig. 6 compares isolation for 5 precursor and 5 postcursor taps against varying total digital-canceller tap counts, while contrasting nonlinear order P = 11 with linear order P = 1.The comparison is conducted with a 40 MHz NR waveform at 2.44 GHz.
  • The RF canceller provides more than 50 dB of direct self-interference suppression before additional digital cancellation increases overall isolation.The measurements are performed without actual targets in an anechoic chamber with the antenna connected.
  • Increasing digital-canceller complexity beyond 5+5 taps provides only marginal performance improvement, so the following experiments adopt the 5+5 configuration.The linear canceller is limited by RF nonlinearities, whereas the nonlinear canceller also suppresses nonlinear leakage.

C. Sensing Static Targets under SI

Outdoor measurements show that RF and digital cancellation improve static-drone sensing and enable detection of moving vehicles despite strong self-interference, clutter, and building reflections.

  • Static drone measurements: Without cancellation, transmitter self-interference appears as a strong static target near zero range and velocity, with sidelobes extending across large distances.This behavior is observed in the measured drone scenario and matches synthetic radar images.
  • Static drone measurements: RF or combined RF and digital cancellation reveals the static drone at approximately 40 m, while propeller rotation produces non-zero-velocity micro-Doppler peaks.The drone remains fixed in position, so the additional velocity peaks arise from rotating propellers.
  • Detection assessment: The measured ROC curves compare drone detection using an RF canceller alone against using both RF and digital cancellers.The supplied caption identifies the two cancellation configurations but does not state the ROC outcomes.
  • Dynamic range: RF and digital cancellers increase measured dynamic range, exposing reflections from high-rise buildings at approximately 120 m and 180 m.A nearby reflection around 10 m is attributed most likely to antenna sidelobes and a nearby building surface.
  • Vehicular measurements: The proposed system detects all three moving vehicles despite strong building reflections and static clutter, using a 20 ms NR processing interval.Estimated distances are 51.1 m, 66.8 m, and 102.1 m, with relative speeds of 12 m/s, -9 m/s, and -9 m/s for vehicles A, B, and C.
  • Vehicular measurements: Vehicle velocity estimates represent projections along the radar-to-vehicle line rather than full trajectory speeds.The stated approximately 20° geometry gives absolute vehicle speeds between 30 km/h and 40 km/h.

V. CONCLUSION

The paper combines LTE and 5G NR downlink radar processing with interpolation for null subcarriers and cancellation for shared-antenna self-interference. It reports strong simulated sensing performance and RF-measured operation for static and moving targets, while also examining simultaneous inband uplink reception.

  • V. CONCLUSION: LTE and 5G NR downlink waveforms are processed for radar sensing using frequency-domain methods and interpolation for missing null-subcarrier samples.The interpolation addresses missing samples within the transmit waveform passband.
  • V. CONCLUSION: With 100 MHz channel bandwidth, distance estimation accuracy on the order of 1 m and target detection probability exceeding 90% are feasible at SNRs below −30 dB.These results are reported for the considered method under ideal TX–RX isolation.
  • V. CONCLUSION: For monostatic shared-antenna deployments, tailored RF and digital cancellation reduce receiver saturation and direct-leakage sidelobe masking while handling frequency-selective TX–RX coupling.The paper also studies simultaneous inband uplink reception during downlink-reflection sensing.
Loading 1908.03418v1…