Radio Interference Cancellation Breakthrough

[2512.18427] On the Limits of Coherent Time-Domain Cancellation of Radio Frequency Interference

On the Limits of Coherent Time-Domain Cancellation of Radio Frequency Interference

Xinrui Li, R. Michael Buehrer

In many sensing (viz., radio astronomy) and radar applications, the received signal of interest (SOI) exhibits a significantly wider bandwidth or weaker power than the interference signal, rendering it indistinguishable from the background noise. Such scenarios arise frequently in applications such as passive radar, cognitive radio, low-probability-of-intercept (LPI) radar, and planetary radar for radio astronomy, where canceling the radio frequency interference (RFI) is critical for uncovering the SOI. In this work, we examine the Demodulation-Remodulation (Demod-Remod) based interference cancellation framework for the RFI. This approach demodulates the unknown interference, creates a noise-free interference replica, and coherently subtracts it from the received signal. To evaluate the performance limits, we employ the performance metric termed \textit{interference rejection ratio} (IRR), which quantifies the interference canceled. We derive the analytical expressions of IRR as a function of the optimal estimation variances of the signal parameters. Simulation results confirm the accuracy of the analytical expression for both single-carrier and multi-carrier interference signals and demonstrate that the method can substantially suppress the interference at a sufficient interference-to-noise ratio (INR), enabling enhanced detection and extraction of the SOI. We further extend the analysis to the scenario where the SOI is above the noise floor, and confirm the validity of the theoretical IRR expression in this scenario. Lastly, we compare the Demod-Remod technique to other time-domain cancellation methods. The result of the comparison identifies the conditions under which each method is preferred, offering practical guidelines for interference mitigation under different scenarios.

 

Analysis of Figure 4: Time-Frequency Comparison of Interference Cancellation Methods

Overview

Figure 4 (shown above) presents time-frequency spectrograms comparing the performance of three different radio frequency interference (RFI) cancellation techniques on a linearly frequency-modulated (LFM) radar signal. The figure demonstrates how each method affects both the interference and the underlying signal of interest.

Signal Configuration

Test Scenario:

• Signal of Interest (SOI): LFM chirp radar signal (the diagonal yellow/orange stripe)
• Interference: Strong narrowband signal centered at approximately 0 MHz (the vertical bright feature)
• Signal-to-Interference Ratio: SOI power is 6 dB lower than interference power
• Interference-to-Noise Ratio (INR): 20 dB
• Reference Signal INR (for filtering): 50 dB

The LFM chirp sweeps linearly across frequency over time, creating the characteristic diagonal signature in the time-frequency plot. This type of signal is fundamental to many radar systems.

Panel-by-Panel Analysis

Panel 1 (Far Left): Original Spectrum - STSA, IRR = 23.03dB

This shows the spectrum before any interference cancellation. Two distinct features are visible:

1. Vertical bright band (center): The strong narrowband interference dominating the spectrum
2. Diagonal yellow stripe: The LFM chirp signal sweeping from approximately -0.4 MHz to +0.6 MHz over the 15 ms observation period

The interference completely obscures portions of the chirp signal where their frequencies overlap.

Panel 2 (Middle Left): STSA Method - IRR = 23.03dB

The Short-Time Sinusoidal Analysis (STSA) approach achieves 23.03 dB of interference rejection. Key observations:

Strengths:

• Substantial suppression of the narrowband interference
• The vertical interference feature is significantly reduced
• The LFM chirp becomes more visible

Weaknesses:

• "Scooping" effect: Visible blue/darker regions within the interference bandwidth (near 0 MHz), indicating overcancellation that removes signal energy along with interference
• Spectral discontinuities: Minor artifacts visible as horizontal banding patterns, caused by discontinuities between adjacent estimation windows
• Out-of-band distortion: Some spectral artifacts affect frequencies outside the interference bandwidth

The STSA method works by estimating amplitude, frequency, and phase of sinusoidal components in short time windows. When the observation window is shorter than the inverse of the interference bandwidth, the interference appears as a simple sinusoid. However, this approach occasionally detects spurious sinusoidal components in noise, leading to the scooping effect.

Panel 3 (Middle Right): Demod-Remod Method - IRR = 31.65dB

The Demodulation-Remodulation approach achieves 31.65 dB of interference rejection—approximately 8.6 dB better than STSA. Key observations:

Strengths:

• Superior interference suppression: Better IRR than STSA
• Minimal spectral distortion: The LFM chirp signal remains clean and undistorted
• No scooping effect: The interference bandwidth shows residual interference rather than overcancellation
• Excellent out-of-band preservation: Frequencies outside the interference bandwidth are essentially unchanged

Characteristics:

• Incomplete cancellation: More residual interference remains visible compared to STSA
• Better signal fidelity: The trade-off between interference removal and signal preservation favors signal integrity

The Demod-Remod method leverages partial knowledge of the interference signal structure (pulse shape and symbol rate) to demodulate the interference, reconstruct a clean replica, and subtract it coherently. This knowledge-aided approach provides better performance than the blind STSA method.

Panel 4 (Far Right): Filtering Method - IRR = 49.05dB

The reference-antenna-based filtering technique achieves 49.05 dB of interference rejection—the best performance of all three methods. Key observations:

Strengths:

• Near-complete interference removal: The vertical interference feature is almost entirely eliminated
• Minimal distortion to SOI: The LFM chirp appears clean and continuous
• Best overall performance: Highest IRR by a substantial margin (17+ dB better than Demod-Remod)

Requirements:

• Requires a reference antenna providing a high-quality copy of the interference
• Reference signal has INR = 50 dB in this example
• Additional hardware complexity (second antenna and receiver chain)

This method uses adaptive filtering to exploit the correlation between the interference observed at the main antenna and the reference antenna, effectively subtracting the interference while preserving the SOI that is uncorrelated with the reference.

Key Technical Insights

Performance vs. Complexity Trade-offs

1. Filtering (49.05 dB IRR): Best performance but requires reference antenna with high-quality interference sample
2. Demod-Remod (31.65 dB IRR): Good performance with single antenna, requires partial signal knowledge
3. STSA (23.03 dB IRR): Moderate performance, no signal knowledge required, but introduces spectral distortion

Signal Preservation Characteristics

The figure clearly demonstrates that interference rejection ratio alone does not tell the complete story. While STSA removes substantial interference energy, it introduces spectral distortion that can corrupt the signal of interest. The Demod-Remod approach achieves the critical balance:

• Higher interference suppression than STSA
• Superior preservation of SOI spectral integrity
• Minimal introduction of artifacts

For wideband sensing applications like radar, where preserving the entire signal bandwidth is essential for accurate target detection and parameter estimation, the Demod-Remod approach offers significant advantages over STSA despite leaving more residual interference.

Practical Implications

This figure illustrates why the research team emphasizes that "no single technique is universally optimal." The choice depends on:

• Availability of reference signals: Filtering is superior when high-quality reference signals exist
• Prior knowledge: Demod-Remod excels when signal structure is partially known
• SOI bandwidth: Wider SOI bandwidths favor methods with better spectral preservation
• Application requirements: Some applications tolerate residual interference better than spectral distortion

Conclusion

Figure 4 provides compelling visual evidence that the Demod-Remod technique offers an attractive middle ground for single-antenna interference cancellation. It substantially outperforms knowledge-free methods while avoiding the hardware complexity of multi-antenna approaches, making it particularly suitable for spectrum-congested environments where radio astronomy, passive radar, and cognitive radio systems must operate.

 

New Algorithms Enable Detection of Buried Signals

BLUF: Researchers at Virginia Tech have developed and validated theoretical performance limits for demodulation-remodulation interference cancellation techniques, enabling radar and radio astronomy systems to detect signals of interest buried up to 70 dB below interference levels—a critical advancement for spectrum-congested applications including passive radar, cognitive radio, and planetary observation.

Advanced Signal Processing Tackles Spectrum Congestion

The electromagnetic spectrum has become increasingly contested territory, forcing sensing systems to operate in environments where interference from external sources can dominate received signals by several orders of magnitude. A new research paper from Virginia Tech's Wireless research group provides the first comprehensive theoretical framework for evaluating coherent time-domain cancellation of radio frequency interference (RFI) using demodulation-remodulation techniques.

The work, led by doctoral candidate Xinrui Li and IEEE Fellow R. Michael Buehrer, addresses a critical gap in interference mitigation for applications where traditional approaches fail. "In many sensing and radar applications, the received signal of interest exhibits significantly wider bandwidth or weaker power than the interference signal, rendering it indistinguishable from background noise," the researchers explain in their paper published in IEEE Transactions on Aerospace and Electronic Systems.

Theoretical Performance Bounds Established

The research team derived analytical expressions for the interference rejection ratio (IRR)—a metric quantifying how effectively interference is removed—as a function of optimal estimation variances of signal parameters. Their analysis reveals that the demodulation-remodulation approach can achieve interference suppression exceeding 60 dB at sufficient interference-to-noise ratios, substantially outperforming alternative methods under specific conditions.

The technique works by demodulating unknown interference, creating a noise-free interference replica, and coherently subtracting it from the received signal. Key parameters estimated include amplitude, carrier frequency offset, phase offset, symbol timing offset, modulation type, and transmitted symbol values.

"We demonstrate that the IRR depends on the accuracy of the estimated parameters used during the demodulation process," Li and Buehrer note. Their theoretical framework accounts for both single-carrier and multi-carrier OFDM interference signals, with validation using synthesized signals and real-world Iridium satellite data.

Practical Applications and Comparative Performance

The research has immediate applications across multiple domains. In passive radar systems, which exploit non-cooperative illuminators of opportunity such as FM, DVB-T, or cellular signals for target tracking, strong direct-path interference can mask reflected signals from targets of interest. Similarly, radio astronomy observations frequently contend with satellite communications interference that can overwhelm faint astronomical signals.

Performance comparisons reveal important operational distinctions between available techniques. The demodulation-remodulation approach requires partial knowledge of the interference signal structure—specifically pulse shape and symbol rate—but needs no reference antenna. This contrasts with filtering-based methods requiring high-quality reference signals that are highly correlated with the interference observed at the main antenna.

The research team's analysis shows that filtering techniques outperform demodulation-remodulation when reference signals are available with interference-to-noise ratios exceeding 50 dB. However, when reference signals are weak, delayed, or unavailable, the demodulation-remodulation approach demonstrates superior performance. The method also significantly outperforms short-time sinusoidal analysis (STSA) techniques in preserving the spectrum of the signal of interest, introducing minimal distortion to wideband signals.

Signal Preservation Critical for Wideband Applications

A key finding involves spectral integrity preservation. Testing with linearly frequency-modulated (LFM) chirp signals—common in radar applications—showed that STSA methods can introduce "scooping" effects within the interference bandwidth and spectral distortion outside it. These artifacts arise from spurious correlations with noise and discontinuities between adjacent estimation windows.

"The Demod-Remod algorithm does not fully cancel the interference but introduces significantly less distortion to the SOI," the researchers found. In time-frequency analyses of real Iridium satellite interference affecting radar signals, the demodulation-remodulation approach maintained superior spectral integrity compared to STSA, particularly important for wideband sensing applications where preserving the entire signal bandwidth is critical.

The research extended previous work on interference rejection ratio metrics, originally developed for filtering-based methods, to the demodulation-remodulation framework. The team also developed an alternative metric, IRRc, specifically designed for real-world collected data where noise-free interference signals are unavailable—addressing a significant practical limitation in field evaluations.

Implementation Considerations and Future Directions

The Virginia Tech team implemented their algorithms using established parameter estimation techniques. Frequency offset and symbol timing for single-carrier signals employ cyclostationary approaches, while OFDM signals use maximum likelihood estimators. Modulation classification relies on Kolmogorov-Smirnov testing to distinguish among BPSK, QPSK, 8-PSK, 16-QAM, and 64-QAM signals.

Testing with Iridium simplex downlink signals—QPSK modulated at 25 kbps in L-band between 1626.0-1626.5 MHz—demonstrated the approach's effectiveness with real-world satellite communications interference. The researchers restricted parameter estimation windows to approximately 3 milliseconds (6,000 samples at 2.048 MHz sampling rate) to account for time-varying satellite signal characteristics induced by Doppler shifts.

Performance degradation at lower interference-to-noise ratios stems primarily from larger parameter estimation errors and higher symbol error rates. The research shows that at interference-to-noise ratios below -5 dB, both symbol error rates and parameter estimation accuracy significantly impact performance. Above this threshold, parameter estimation accuracy becomes the dominant limiting factor.

The work was supported by the National Science Foundation under Grant ECCS-2029948. The researchers acknowledge Dr. Steven W. Ellingson for data collection support and previous foundational work in coherent time-domain interference cancellation for radio astronomy applications.

Operational Guidelines for System Designers

The research team provides practical selection criteria for interference mitigation techniques based on specific operational scenarios. Filtering with reference antennas remains optimal when high-quality reference signals are available with interference-to-noise ratios exceeding 50 dB, or when multiple geographically distributed antennas can be employed.

For scenarios without reference antennas, STSA methods are appropriate when no prior knowledge of signal structure exists, offering modest performance with relatively high computational complexity. The demodulation-remodulation approach is recommended when partial knowledge of interference signal structure is available—specifically pulse shape and symbol rate—and computational resources permit more sophisticated parameter estimation.

The framework also addresses scenarios where signals of interest are not buried beneath the noise floor. By approximating signal-of-interest power within the interference bandwidth as additive Gaussian noise, the theoretical expressions remain valid with appropriately adjusted effective interference-to-noise ratios. This extension significantly broadens applicability to radar and communications systems where signal-of-interest power may be comparable to interference levels, provided the signal-of-interest bandwidth substantially exceeds interference bandwidth.

Future research directions include extending the analysis to accommodate fading channel conditions beyond the additive white Gaussian noise model analyzed in this work, and investigating hybrid approaches combining multiple cancellation techniques to optimize performance across varying operational conditions.

---

Verified Sources

1. Li, X., & Buehrer, R.M. (2025). "On the Limits of Coherent Time-Domain Cancellation of Radio Frequency Interference," IEEE Transactions on Aerospace and Electronic Systems, Vol. XX, No. XX, arXiv:2512.18427v1 [eess.SP] https://arxiv.org/abs/2512.18427

2. Ellingson, S.W., & Buehrer, R.M. (2022). "Coherent time-domain canceling of interference for radio astronomy," Publications of the Astronomical Society of the Pacific, Vol. 134, No. 1041, p. 114505 https://iopscience.iop.org/article/10.1088/1538-3873/ac9ab4

3. Li, X., Buehrer, R.M., & Ellingson, S.W. (2025). "Parametric methods for coherent time-domain canceling of radio frequency interference," Publications of the Astronomical Society of the Pacific, Vol. 137, No. 4, p. 044502 https://iopscience.iop.org/article/10.1088/1538-3873/ad93db

4. Palmer, J.E., Harms, H.A., Searle, S.J., & Davis, L. (2013). "DVB-T passive radar signal processing," IEEE Transactions on Signal Processing, Vol. 61, No. 8, pp. 2116-2126 https://ieeexplore.ieee.org/document/6461833

5. Colone, F., O'Hagan, D., Lombardo, P., & Baker, C. (2009). "A multistage processing algorithm for disturbance removal and target detection in passive bistatic radar," IEEE Transactions on Aerospace and Electronic Systems, Vol. 45, No. 2, pp. 698-722 https://ieeexplore.ieee.org/document/5089387

6. Virginia Tech Wireless@VT Research Group https://wireless.vt.edu

7. National Science Foundation Award ECCS-2029948 https://www.nsf.gov/awardsearch/showAward?AWD_ID=2029948