Abstract
Based on the Hilbert transform algorithm, a digital phase detector for sinusoidal signal is proposed. A half-band IIR filter implemented by the all-pass method approximates the Hilbert transform to realize the orthogonal decomposition of the signal. The Prosthaphaeresis formulas and Cordic algorithm are used to deal with the phase difference. The method can effectively avoid the influence of signal amplitude and frequency fluctuation on the phase detection results, and improve the precision of phase detector. The phase detector is designed and verified on Xilinx's Artix7 FPGA. Experiments show that the phase detector error is 0.3%.
Introduction
Phase-locked loops (PLLs) are foundational components in modern electronics, essential for wireless communication and high-speed sensor interfaces. At the core of any PLL is the Phase Detector (PD), which measures the phase difference between two signals. Traditional phase detection methods often suffer from sensitivity to noise, signal amplitude changes, and frequency instability. Fluctuations in input signal amplitude or frequency can lead to significant errors, degrading overall system performance. This paper presents a robust digital PD designed to maintain high precision despite such fluctuations while remaining computationally efficient for real-time hardware implementation.
Methodology
The proposed digital phase detector utilizes a multi-stage pipeline to extract phase information accurately. The overall architecture is illustrated in the block diagram below.
Hilbert Transform via IIR Filtering
The system first decomposes the input signal into two mutually orthogonal components (sine and cosine) using a Hilbert transform. To optimize for hardware efficiency, the authors implemented a half-band IIR filter using an all-pass method rather than standard, computationally expensive FIR filters. This approach achieves the necessary 90-degree phase shift with a lower filter order, as shown in the realization diagram.
Signal Processing and Phase Extraction
Following the orthogonal decomposition, Prosthaphaeresis formulas—trigonometric identities—are applied to process the four-channel signals. This step isolates the phase difference ($\theta$) into a format relating the signals' amplitudes ($A_1, A_2$) to $\sin(\theta)$ and $\cos(\theta)$.
Finally, the Coordinate Rotation Digital Computer (CORDIC) algorithm is employed to extract the angle $\theta$. CORDIC is highly efficient for FPGA implementation as it calculates trigonometric functions using only additions and bit-shifts.
Hardware Implementation
The design was implemented on a Xilinx Artix7 FPGA. To balance precision and resource usage, filter coefficients were quantized to 11 bits, while internal calculations utilized 22-bit precision to minimize truncation errors.
Key Results and Discussion
The proposed phase detector demonstrated significant performance advantages:
- High Precision: Experimental verification on the FPGA yielded a phase detector error of only 0.3%.
- Amplitude Independence: By utilizing both sine and cosine components of the phase difference, the system effectively cancels out the influence of signal amplitudes ($A_1$ and $A_2$), making it robust against volume fluctuations.
- Frequency Stability: Bode diagram analysis confirmed a constant 90-degree phase shift and 0dB gain within a frequency range of 148Hz to 2kHz.
- Fast Response: Simulation results showed the system reaches stability in approximately 2ms.
Conclusion
The Hilbert transform-based digital phase detector offers a high-precision, hardware-efficient solution for sinusoidal signal processing. While the current design is optimized for a specific frequency window (148Hz to 2kHz), its robustness against amplitude fluctuations makes it a strong candidate for integration into all-digital phase-locked loops (ADPLLs) for broadband communication. Future work may involve adapting the IIR filter design to support a wider frequency range.
Summary updated