A High-Precision Anti-Jamming Algorithm Based on Newton-Iteration-Enhanced Three-Spectral-Line RIFE with Real-Time Implementation Authors: Xinhua Tang, Yiming Wang Published: 2026 ABSTRACT GNSS signals are extremely weak at the Earth’s surface and are highly vulnerable to in-band interference, particularly high-dynamic linear frequency-modulated (LFM) jamming, which may lead to receiver loss of lock. Existing anti-jamming techniques struggle to balance real-time constraints with high-precision frequency estimation. This paper proposes a Newton-iteration-enhanced three-spectral-line RIFE algorithm implemented on a heterogeneous FPGA platform (Zynq-7000 SoC). The method performs coarse frequency estimation using the three-spectral-line RIFE to mitigate FFT fence effects, followed by Newton-based quadratic refinement, enabling high estimation accuracy with reduced FFT size. A fast–slow loop architecture is adopted, where the FPGA (PL) performs real-time interference suppression and the ARM (PS) handles system control and parameter updates. Experimental results show that, under static interference, the proposed method achieves a 10.9 dB improvement over direct estimation algorithms. Under chirp interference, it significantly outperforms both direct estimation and conventional iterative methods. In GNSS closed-loop tests, the proposed approach extends the anti-jamming margin to 82 dB J/S. Overall, the proposed method effectively balances estimation accuracy and processing latency, providing a practical solution for GNSS anti-jamming in high-dynamic environments. 1. Introduction Global Navigation Satellite System (GNSS) provides all-weather, high-precision navigation and timing services, which are widely deployed across critical sectors such as national defense, transportation, power grid management, and financial communications. However, due to the long-distance space propagation from medium-to-high Earth orbits (approximately 20,000 km), GNSS signals arrive at ground receivers with extremely low power levels, typically around −130 dBmW. Although GNSS incorporates strategies like spread-spectrum processing to improve processing gain and enhance signal resilience, these techniques cannot completely eliminate the inherent vulnerability of the signals. When the strength of an interference signal exceeds the receiver’s anti-jamming tolerance margin, the front-end Automatic Gain Control (AGC) circuit may saturate, or the sharp degradation in the carrier-to-noise ratio may cause the tracking loops to lose lock. This ultimately leads to the failure of the receiver’s positioning and timing functions. Relevant studies have shown that even low-power, compact jamming devices can easily disrupt unprotected GNSS receivers within a 6 to 9 km radius. With the continuous advancement of electronic countermeasure (ECM) technologies, the electromagnetic environment where GNSS operates has become increasingly complex. Beyond unintentional radio frequency (RF) interference, intentional interference constitutes the primary threat. Intentional interference is typically categorized into two main types, jamming and spoofing, as illustrated in Figure 1. Among the aforementioned interference types, jamming has emerged as the dominant form of interference due to its straightforward implementation and pronounced disruptive effects. Based on the interference bandwidth, jamming can be further categorized into narrowband and wideband interference. Narrowband jamming typically includes single-tone and multi-tone jamming, while wideband jamming primarily includes matched-spectrum and sweep-frequency jamming. Among these, Narrowband Interference and sweep-frequency interference are the most prevalent types. Narrowband Interference (NBI) primarily manifests as single-tone or narrowband signals, with its energy concentrated at specific frequencies. Although NBI is relatively easy to detect, its high energy concentration can severely distort the signal’s spectral structure, resulting in significant degradation of the receiver’s carrier-to-noise ratio (C/N0). Conversely, Sweep-Frequency Interference is a highly dynamic, non-stationary signal whose frequency varies rapidly over time in either a linear or nonlinear manner. Compared to static NBI, sweep-frequency interference is capable of traversing the entire GNSS signal bandwidth, leading to sever disruption on both the acquisition and tracking stages of the receiver. In the domain of GNSS anti-jamming techniques, the deployment of digital notch filters—particularly Infinite Impulse Response (IIR) notch filters—is widely recognized as the most prevalent and effective countermeasure against Narrowband Interference. The performance of the notch filter highly relies on the mismatch between its null frequency and the jamming center frequency; thus, the effectiveness of this suppression technique is strictly dependent upon the accuracy of the jamming frequency estimation. Even a marginal deviation in the estimated frequency can result in significant residual jamming energy leakage into the system, thereby severely degrading the receiver’s performance. Therefore, ultra-precise and real-time frequency estimation is a critical prerequisite for the successful implementation of notch-based interference mitigation techniques. Although traditional Maximum Likelihood Estimation (MLE) provides high frequency estimation accuracy, it entails substantial computational overhead that hinders real-time implementation. In contrast, simple Fast Fourier Transform (FFT) operation is computationally efficient, but is fundamentally limited by the picket-fence effect, which restricts high-precision frequency estimation. To address this challenge, various frequency offset estimation algorithms based on interpolation, curve fitting, and iterative optimization have been proposed. One important optimization direction is based on RIFE. Wang, X. et al. proposed a shifted RIFE algorithm that forcibly moves the spectral peak intensity to the intermediate region between two quantized frequency points via a precise frequency-domain shift [1]. This mechanism effectively avoids the worst-case spectral leakage scenarios inherent in conventional RIFE algorithms when the true frequency lies exactly halfway between two bins. To further enhance estimation performance under severe low SNR conditions, J. Wang et al. proposed a multi-point RIFE algorithm utilizing three or more spectral lines for weighted interpolation [2], successfully decreasing the variance of the estimate by leveraging adjacent side-lobe information. Niu et al. proposed a three-line interpolation algorithm that compensates for the main-lobe broadening of commonly used windows via advanced polynomial fitting [3]. For pulse signals, Jia Luo et al. designed a cascaded architecture that combines M-Rife for coarse estimation with Maximum Likelihood Estimation (MLE) for fine calibration [4]. This approach significantly reduces the search space of the MLE, achieving an absolute accuracy better than 0.3 Hz without exhaustive searching. Furthermore, Cheng et al. introduced an improved algorithm based on Phase Angle Interpolation (PAI-Rife) [5]; with the introduction of the frequency deviation factor, this method mitigates phase wrapping errors and improves estimation linearity. Jia et al. innovatively proposed the SP-Rife algorithm, which utilizes Spectral Refinement (Zoom-FFT) as a digital down-converter to isolate the target band before applying phase angle interpolation [6], demonstrating exceptional robustness and resolution even under heavy noise. Another important refinement direction is based on the Quinn algorithm. The Quinn algorithm utilizes complex spectral line phase relationships for frequency estimation, but it inherently lacks robustness at low SNRs. To resolve this, Liu et al. proposed an improved Quinn algorithm utilizing amplitude and phase interpolation alongside a segmented spectrum-shifting strategy [7]. This method avoids the nonlinear distortion near the Nyquist frequency and DC components, reducing computational complexity while maintaining high fidelity. X. Zhang et al. proposed a joint estimation scheme cascading the frequency-domain Quinn algorithm with time-domain pilots [8], effectively resolving the phase ambiguities that purely frequency-domain methods often encounter. To enhance accuracy in specific high-resolution scenarios such as FMCW radar ranging, Duan et al. [9] and Fernando M. Janeiro [10] explored cascaded refinement schemes integrating the Fast Chirp-Z Transform (CZT) with the Quinn algorithm, where CZT is leveraged to “zoom in” on a target frequency band, and arbitrary resolution can be achieved without the substantial memory overhead associated with an N-point FFT. Under non-ideal channel conditions, Xie et al. adopted a minimax nonlinear transformation design [11], thereby enhancing the algorithm’s statistical robustness against harmonic distortion and heavy-tailed non-Gaussian impulse noise. Additionally, significant progress has been made in high-precision iterative methods and theoretical analysis. Zhe Rao et al. proposed a secondary correction method (RP method) that combines amplitude ratio and segmented phase difference, where the former isolates the main lobe and the latter enables ultra-fine tuning [12]. Chen, K. et al. adopted a hierarchical processing strategy to propose a third-order frequency offset estimation method [13], which is particularly effective for tracking dynamic signals with significant Doppler rates, such as mechanical vibrations. Z. Chen et al. introduced an adaptive tracking algorithm based on an improved Kalman filter [14], facilitating the continuous smoothing of frequency estimates over time. Liao et al. investigated the underlying mathematical derivation of frequency offset estimation based on all-phase FFT (apFFT) dual-spectral-line interpolation [15]. Owing to the inherent zero phase difference at the main spectral line in apFFT, phase-based interpolation can be effectively applied. Furthermore, they proved that the vast majority of interpolation estimators possess analytical solutions, thereby introducing a completely new set of mathematically unbiased estimators. To tackle the computational complexity of iterative algorithms, Lu et al. proposed an Iterative Interpolated DFT (IpDFT) method that dynamically reconstructs and subtracts the dominant spectral leakage [16], successfully unmasking weak, closely spaced interference components. Wei M. introduced a binary search mechanism (Ds-IpDTFT) to replace complex polynomial root-finding, relieving the reliance on hardware-unfriendly operations [17]. Yang et al. utilized Semidefinite Relaxation (SDR) technology to relax the non-convex cost function of MLE [18], providing reliable global initial values that prevent the Gauss–Newton iteration from becoming trapped in local minima. Liao et al. proved that the interpolated estimator admits an analytical solution [19], thereby introducing an unbiased estimator. Although significant progress have been made to the development of frequency estimation algorithms, the design of the efficient hardware implementation platforms is also crucial for their practical deployment. Wang et al. pioneered its deployment on FPGAs by designing a highly pipelined architecture that optimizes the utilization of dedicated DSP slices [1]. Ye et al. detailed the hardware implementation of the M-Rife algorithm on FPGAs and successfully applied it to FMCW LiDAR ranging [20], their work verified that RIFE-class algorithms could operate at high sampling rates with minimal Look-Up Table (LUT) and Flip-Flop (FF) resource consumption, demonstrating their feasibility for real-time hardware implementation. For active anti-jamming filter design, Same et al. proposed using simplified Welch power spectrum estimation combined with a chained Infinite Impulse Response (IIR) notch filter structure [21], which facilitates the adaptive tracking and suppression of multiple narrowband interference tones with extremely steep roll-off characteristics. Furthermore, Gomes et al. proposed a software-defined antenna array platform based on the heterogeneous Zynq UltraScale+ RFSoC [22], utilizing the parallel processing capabilities of the Programmable Logic (PL) for high-speed computation (e.g., beamforming and filtering) while placing complex control logic and dynamic weight updates on the Processing System (PS) side. This heterogeneous hardware–software co-design paradigm enhances system flexibility, resource efficiency, and real-time response speed. Based on the above discussion, to address the inherent conflict between high-precision frequency estimation and microsecond-level real-time responsiveness, a novel interference mitigation scheme based on a Newton-iteration-enhanced three-spectral-line RIFE method and a heterogeneous hardware architecture is proposed. The core contributions of this paper are explicitly summarized as follows: (1) A Cascaded High-Precision Frequency Estimation Algorithm: We propose a novel Newton-iteration-enhanced three-spectral-line RIFE algorithm. By utilizing the three-spectral-line RIFE for coarse estimation and Newton iteration for fine refinement, this cascaded architecture effectively overcomes the limitations of traditional iterative algorithms, including slow convergence and high sensitivity to initial values. Furthermore, it achieves near-CRLB high-precision frequency estimates with a significantly reduced number of FFT points, making it exceptionally well-suited for tracking highly dynamic sweep-frequency interference. (2) A Heterogeneous “Fast-Slow Loop Separation” Hardware Architecture: To address the performance degradation and excessive computational loop delays inherent in pure software-based SDR platforms, a tunable, low-latency hardware architecture based on joint PS-PL (Processing System–Programmable Logic) co-design is developed. Specifically, to ensure a microsecond-level rapid response to incoming interference signals, the critical “fast loop”—encompassing interference detection, coarse frequency estimation, and real-time notch filtering—is deployed entirely within the PL (FPGA) fabric. Simultaneously, the PS (ARM Cortex) manages a supervisory “slow loop” responsible for Newton iterative refinement, high-level system control, and the precise calibration of filter coefficients. (3) An Effective Trade-off Among Accuracy, Latency, and Resource Efficiency: The proposed software-hardware co-design strategy fundamentally breaks the latency bottleneck of traditional high-precision iterative algorithms. By decoupling the ultimate estimation accuracy from the physical FFT size, the proposed platform drastically reduces hardware resource consumption (e.g., BRAM utilization) while significantly extending the overall anti-jamming margin of the GNSS receiver under severe dynamic conditions. Table 1 summarizes the advantages and disadvantages of different algorithms in terms of estimation accuracy, computational complexity, and response speed. The remainder of this paper is organized as follows: Section 2 establishes the theoretical framework, detailing the principles of digital notch filters and the rigorous mathematical derivation of the proposed Newton-refined RIFE algorithm. Section 3 describes the hardware implementation, focusing on the joint PS–PL scheduling mechanism and the heterogeneous design of the anti-jamming processor. Section 4 presents the experimental validation. The performance of the proposed algorithm is comprehensively evaluated against conventional methods under both fixed-frequency and sweep-frequency interference scenarios, followed by a closed-loop GNSS positioning test. Finally, Section 5 concludes the paper by summarizing the key contributions and outlining directions for future research. 2. Methods In the case of a sweep-frequency or other similar jamming signal, minor frequency misalignment between the notch null and the interference center frequency can result in undesired attenuation of the GNSS signal, necessitating the real-time precise estimation of the interference center frequency. The proposed high-precision, real-time anti-jamming scheme, based on a Newton-iteration-enhanced three-spectral-line RIFE algorithm, is designed to resolve the inherent conflict between ultra-high-frequency estimation accuracy and microsecond-level hardware responsiveness in GNSS receivers. This is particularly critical when dealing with highly dynamic interference, such as sweep-frequency jamming. The methodology adopts a hierarchical processing architecture. Initially, an improved three-spectral-line RIFE algorithm is employed to execute a fast coarse estimation by calculating the approximate frequency center within the main lobe. This step effectively mitigates the FFT picket-fence effect while introducing minimal computational overhead. Subsequently, a fine refinement stage is introduced based on the Newton iteration method. By exploiting the intrinsic properties of the Fourier transform, this stage achieves quadratic convergence, allowing the estimation accuracy to approach the Cramér–Rao lower bound (CRLB) within a limited number of iterations. In summary, the proposed methodology, as shown in Figure 2, can improve the performance the notch filter from two aspects, including the accuracy and real-time capability. 2.1. Impact of Estimation Error on Anti-Jamming Performance In conventional anti-jamming architectures, digital notch filtering is widely adopted as the primary technique for suppressing narrowband interference. A notch filter is specifically designed to attenuate signals at targeted frequencies, exhibiting deep rejection at a single narrowband frequency while maintaining an extremely narrow stopband, thereby minimizing distortion to adjacent useful signal components. Such filters can be implemented using either Finite Impulse Response (FIR) or Infinite Impulse Response (IIR) structures. In this work, a single-sided IIR notch filter is employed for interference suppression. The operating principle of an IIR notch filter is based on the placement of conjugate zeros and poles in the z-plane. To achieve strong attenuation at the interference frequency, the zeros are typically located on the unit circle to generate a deep spectral null. Correspondingly, the poles are placed slightly inside the unit circle at the same angular frequency, ensuring system stability while preserving a narrow stopband. The general transfer function of the filter can be expressed as follows:(1)H(z)=1−ejω0z−11−pejω0z−1 where z−1 denotes the unit-delay operator in the Z-domain, and ω0 represents the notch angular frequency (in radians per sample), which is related to the physical frequency f0 (in Hz) and the sampling frequency fs (in Hz) by ω0=2πf0/fs. The parameter p is the pole radius satisfying 0