Polynomial Method for PLL Controller Optimization

Ta-Chung Wang, Sanjay Lall, Tsung-Yu Chiou

TL;DR

This paper introduces a novel polynomial-based method for optimizing Phase-Locked Loop (PLL) controllers. By leveraging recent advancements in sum-of-squares (SOS) techniques and semidefinite programming, the authors can precisely estimate the domain of attraction (DoA) for nonlinear PLL systems, leading to controllers with improved stability and performance compared to traditional linear design approaches.

What Problem Does This Paper Address?

Phase-Locked Loops (PLLs) are fundamental components in modern electronic communication and control systems, crucial for tasks like carrier tracking. However, traditional PLL controller design often relies on linear models, which can lead to performance degradation or instability when the system operates far from its equilibrium point. This is particularly problematic because PLLs are inherently nonlinear systems with a limited domain of attraction (DoA). Accurately characterizing and expanding this DoA is essential for robust system performance, especially in the presence of noise and interference. Existing methods for analyzing the DoA, such as Lyapunov's method with quadratic functions, often provide only local stability guarantees or inner approximations that leave significant unexplored regions.

Methods

The core of this paper lies in the application of sum-of-squares (SOS) techniques and semidefinite programming (SDP) to analyze and design PLL controllers. The authors utilize these algebraic tools to:

  • Estimate the Domain of Attraction (DoA): Instead of relying on simple quadratic Lyapunov functions, they search for more general polynomial Lyapunov functions. This is achieved by converting the problem of finding a polynomial that is positive within a certain region into an SOS constraint, which can be efficiently solved using SDP. The concept of set advection is employed to propagate sets backward in time, allowing for the estimation of the DoA.
  • Find Polynomial Lyapunov Functions: The SOS framework enables the search for polynomial Lyapunov functions that serve as certificates for the stability of the system within its DoA.
  • Optimize Controller Parameters: Once a suitable Lyapunov function is found, it's used as a performance constraint within an optimization problem. This allows for the refinement of controller parameters to achieve a larger DoA and faster convergence rates, going beyond the limitations of linear design. The optimization problem is formulated as a semidefinite program.
  • Handle Nonlinearities: The method explicitly accounts for the nonlinear nature of the PLL, particularly the sinusoidal nonlinearity of the phase detector, which is approximated by polynomials for tractability within the SOS framework.

Datasets

Not discussed in the paper. The paper focuses on theoretical development and simulation/experimental validation of the proposed method on a second-order PLL model and real GPS data.

Key Insights and Ideas

  • Polynomials as Powerful Tools for Nonlinear Systems: The paper demonstrates that by using polynomial representations and SOS techniques, we can move beyond the limitations of linear analysis and achieve a more precise understanding and control of nonlinear systems like PLLs.
  • DoA Estimation is Key to Robust PLL Design: Accurately characterizing the DoA is not just an academic exercise; it directly translates to designing PLLs that are more resilient to initial conditions, noise, and disturbances, leading to better overall system performance.
  • Set Advection for Backward Propagation: The use of set advection provides a systematic way to "grow" the estimated DoA backward in time, offering a more comprehensive picture of the stable region.
  • Lyapunov Functions Beyond Quadratics: The ability to find more general polynomial Lyapunov functions significantly enhances the power of Lyapunov stability theory for complex nonlinear systems.

Key Results

  • Improved Domain of Attraction: The proposed nonlinear design approach results in approximately 20% larger guaranteed domain of attraction compared to traditional linear design methods.
  • Faster Convergence and Less Overshoot: Simulations and experimental results show that the nonlinear designed PLL exhibits less overshoot and a much faster convergence rate (up to five times faster) than the linear design.
  • Effective Polynomial Approximation: The method effectively approximates nonlinear functions (sine and cosine) using polynomials to enable SOS analysis within a bounded interval.
  • Validation with Real GPS Data: The effectiveness of the proposed method is validated using real GPS experimental results, demonstrating its practical applicability.

Limitations and Future Work

  • Polynomial Approximation Accuracy: The accuracy of the DoA estimation is dependent on the degree of the polynomials used to approximate nonlinear functions. While the paper uses degree-10 polynomials, higher degrees might be needed for greater accuracy, potentially increasing computational complexity.
  • Computational Complexity: Solving semidefinite programs can be computationally intensive, especially for higher-dimensional systems or higher-degree polynomials.
  • Future Work: The authors suggest further investigation into the relationship between the accuracy of advection and the degree of polynomials used in SOS constraints. They also mention the possibility of applying the method to third-order controller designs and exploring different performance requirements within various sub-level sets of the Lyapunov function. The paper also hints at future work on applying dynamic performance constraints and noise bandwidth constraints.

Summary updated

BibTeX
@inproceedings{wang2011polynomial,
  author = {Ta-Chung Wang and Sanjay Lall and Tsung-Yu Chiou},
  title = {{Polynomial Method for PLL Controller Optimization}},
  booktitle = {Sensors},
  year = {2011},
  url = {https://openreview.net/forum?id=dNpI00NZeG},
}
Figure