🔬 physics

Investigation of the Asymptotic Properties of Active Impedance in Large Finite Array Antennas

This paper introduces an improved, memory-efficient full-wave solver to demonstrate that even large finite arrays (up to 1000 elements) may exhibit active impedance behaviors distinct from infinite arrays, and proposes accurate predictors to determine when simplified unit cell approximations are valid.

Original authors: Harald Hultin, Lucas Åkerstedt, B. L. G. Jonsson

Published 2026-08-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Harald Hultin, Lucas Åkerstedt, B. L. G. Jonsson

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Antenna arrays are the silent workhorses of modern communication, from the cell towers that keep our phones connected to the radar systems that guide aircraft. Instead of relying on a single, massive dish, these systems use dozens, hundreds, or even thousands of small, identical antennas arranged in a grid. By working together, they can steer radio beams electronically with incredible speed and precision, a capability essential for everything from 5G networks to advanced radar. However, designing these systems presents a tricky physical puzzle. When an antenna sits in the middle of a vast grid, it does not behave exactly as it would if it were standing alone. It is constantly influenced by its neighbors, a phenomenon known as mutual coupling. The radio waves bouncing off nearby elements change how the central antenna receives and transmits signals. Engineers have long relied on a simple rule of thumb: if an array is large enough, specifically if it has a certain number of elements, the center antenna will eventually stop feeling the influence of the edges and begin to behave as if it were part of an endless, infinite grid. This assumption allows designers to simulate a tiny, manageable piece of the array and apply those results to the whole system, saving immense amounts of time and computing power.

But is this rule of thumb actually true? A recent study by researchers at KTH Royal Institute of Technology and Saab Surveillance suggests that the answer is more complicated than previously thought. The team set out to investigate exactly how large an array needs to be before the center element truly settles into a stable, predictable state. To do this, they first had to build a better tool for the job. Simulating thousands of antennas at once is a massive computational challenge that often crashes standard computer programs or takes days to run. The researchers developed a new, highly efficient computer solver that can handle these giant grids without breaking a sweat. By organizing the data in a smarter way and using a specific mathematical technique to speed up the calculations, they created a system that runs eight times faster than their previous best attempts while using far less computer memory. This new tool allowed them to run full, detailed simulations of arrays with up to 1,000 elements, a scale that was previously too difficult to model with such precision.

Using this powerful new solver, the team tested the long-held belief that a 10-by-10 grid is sufficient for the center antenna to act like it is in an infinite array. They examined different types of antenna elements, including bowtie shapes and Vivaldi designs, across various frequencies. The results were surprising. Even in arrays with 1,000 elements, the behavior of the center antenna did not always match the predictions made by infinite-array theories. The influence of the edges, where the array simply stops, continued to ripple inward, altering the electrical properties of the central element in ways that standard approximations missed. In some cases, the error in these approximations remained significant even for very large grids, particularly for wideband antennas that operate over a broad range of frequencies. The study found that the "infinite" behavior assumed by engineers is not a guaranteed destination that every array reaches, regardless of size.

To help engineers navigate this uncertainty, the researchers proposed two new ways to predict how accurate an approximation will be before they commit to a full simulation. The first method involves looking at the passive connections between the antennas—essentially measuring how much one element talks to another when the system is not actively transmitting. By analyzing these quiet interactions, they could estimate how much the active performance of the center element would deviate from the ideal. The second method relied on a theoretical prediction of how these interactions should fade with distance, based on established physics. Both methods proved to be remarkably effective. They offered a practical way to decide whether a simplified model would work for a specific design, moving beyond the blunt instrument of "is it big enough?" to a more nuanced understanding of "how big is big enough for this specific antenna?"

The study concludes that while the idea of an infinite array is a useful starting point, it is not a universal truth for all antenna designs. The researchers demonstrated that for many modern, wideband systems, the edge effects persist much longer than the old rules suggested. This means that for critical applications, relying solely on simplified models could lead to unexpected performance issues. Instead, the new predictors provide a reliable safety check, allowing designers to know exactly when a full, detailed simulation is necessary and when a simpler approximation will suffice. By replacing guesswork with precise measurement and prediction, this work offers a clearer path forward for designing the complex antenna arrays that will power the next generation of wireless technology.

Technical Summary: Investigation of the Asymptotic Properties of Active Impedance in Large Finite Array Antennas

Problem Statement
The design and simulation of large finite array antennas present a significant computational challenge. While full-wave simulations are accurate, they become computationally prohibitive for arrays with thousands of elements. Consequently, engineers often rely on approximations, such as unit cell methods derived from infinite array theory, assuming that the central region of a sufficiently large finite array behaves identically to an infinite array. A common heuristic suggests that a $10 \times 10$ array (or $5\lambda \times 5\lambda$ ) is sufficient for this approximation. However, the accuracy of this assumption varies depending on element geometry, scan direction, and bandwidth. There is a lack of nuanced tools to determine when these approximations are valid for specific active impedance parameters, particularly the Active Reflection Coefficient (ARC), and how edge effects influence convergence in finite arrays.

Methodology
To address these challenges, the authors developed an improved full-wave solver and utilized it to investigate the asymptotic properties of active impedance in large regular arrays.

Solver Improvements:
- Preconditioning Scheme: The authors adapted a preconditioning strategy for the Method of Moments (MoM) matrices of finite arrays. By exploiting the Hermitian conjugate properties of the Multi-Level Block-Toeplitz (MBT) impedance matrix ( $Z_A$ ), they demonstrated that the MLFFT operator commutes with the Hermitian conjugate. This allows the preconditioner to be applied efficiently without storing redundant data.
- Algorithm Selection: The improved preconditioning enables the use of the Preconditioned Conjugate Gradient (PCG) method instead of GMRES. PCG offers a significantly lower memory footprint (scaling linearly with problem size rather than iterations) and improved numerical stability.
- Data Structure: A new 4-dimensional data arrangement was implemented to optimize the Multi-Level Fast Fourier Transform (MLFFT) operations, reducing memory access jumps and improving cache efficiency.
- Performance: These improvements resulted in an 8-fold reduction in simulation time and a reduced memory footprint compared to previous Toeplitz-based solvers, enabling the simulation of arrays with up to 1000 elements.
Asymptotic Analysis and Predictors:
- The study analyzed the convergence of the active impedance of the center element as the array size increased for three different antenna types: bowtie, Vivaldi, and BoR (Body of Revolution).
- Two predictors were proposed to estimate the error ( $\xi$ $ξ$ ) between a finite array's center element impedance and its asymptotic (infinite) value:
  - $\xi_Z$ (Impedance-based): Derived from the passive mutual impedance between the center element and the nearest border elements.
  - $\xi_S$ (Scattering-based): Derived from the expected asymptotic decay of the scattering matrix ( $S$ ) coefficients, assuming a $\Delta_n^{-2}$ decay rate typical of infinite arrays.

Key Results

Solver Performance: The new PCG-based solver with the adapted preconditioner and 4D data structure achieved a factor of 8 speedup over previous methods while maintaining high numerical accuracy (RRMSE comparable to direct solvers for the center element).
Convergence Behavior: The study found that even for arrays with 1000 elements, the active impedance of the center element does not always reach the asymptotic values predicted by infinite array theory. Edge effects remain significant.
Failure of Universal Rules: The "10x10" or " $5\lambda \times 5\lambda$ " rule of thumb does not guarantee convergence for all element types. For instance, the Vivaldi array showed inconsistent decay and did not converge within the simulated sizes, whereas the bowtie array converged more rapidly.
Predictor Accuracy:
- The impedance-based predictor ( $\xi_Z$ ) performed very well, accurately tracking the deviation of the active impedance from its asymptotic value for both bowtie and Vivaldi arrays.
- The theoretical scattering decay predictor ( $\xi_S$ ), based on the $\Delta_n^{-2}$ assumption, provided reasonable estimates but showed deviations in specific cases (e.g., Vivaldi).
- Coupling Decay: Contrary to infinite array theory which predicts a $\Delta_n^{-2}$ decay for coupling coefficients, the actual coupling in the finite bowtie and BoR arrays decayed faster (approximately $\Delta_n^{-2.4}$ to $\Delta_n^{-2.6}$ ). However, this steeper decay in the finite $S$ -matrix did not translate directly to a better prediction of the active impedance error when used in the theoretical model.

Significance and Claims
The paper claims to provide a more nuanced understanding of when array approximations, such as unit cell methods, are appropriate. Rather than relying on blanket statements regarding array size (e.g., "10x10 is sufficient"), the authors propose using the passive inter-element impedance or the expected decay of scattering parameters as practical predictors.

The significance of this work lies in:

Computational Efficiency: Enabling full-wave simulations of very large finite arrays (up to 1000 elements) that were previously infeasible, using a stable and memory-efficient PCG solver.
Validation of Approximations: Demonstrating that infinite array asymptotic properties only weakly apply to finite arrays due to persistent edge effects, even at large scales.
Practical Tools: Offering the impedance-based predictor ( $\xi_Z$ ) as a reliable tool to estimate the error of macroscopic approximations without requiring a full array simulation, thereby aiding in the decision-making process for antenna design and simulation strategies.

More like this