Abstract
Optical beam center position on an array of detectors is an important (hidden) parameter that is essential not only from a tracking perspective, but is also important for optimal detection of Pulse Position Modulation symbols in free-space optical communications. In this paper, we have examined the beam position estimation problem for photon-counting detector arrays, and to this end, we have proposed and analyzed a number of non-Bayesian beam position estimators. These estimators are compared in terms of their mean-square error, bias and the probability of error performance. Additionally, the Cramer-Rao Lower Bounds (CRLB) of the tracking error is also derived, and the CRLB curves give us additional insights concerning the effect of number of detectors and the beam radius on mean-square error performance. Finally, the effect of beam position estimation on the probability of error performance is investigated, and our study concludes that the probability of error of the system is minimized when the beam position on the array is estimated as accurately as possible.
Introduction
Free-Space Optics (FSO) has emerged as a cornerstone for high-bandwidth 5G backhaul and satellite-to-ground links. However, the efficacy of FSO is often limited by the use of incredibly narrow laser beams. Physical disturbances—such as atmospheric turbulence, mechanical jitter, or the swaying of a drone—can cause the beam to drift off the receiver, leading to significant signal loss. Traditionally, systems have addressed this using dedicated hardware like Quadrant Photodetectors or by sacrificing bandwidth for "pilot symbols" to aid alignment. This research proposes a more efficient "two-for-one" hardware approach: using the same photon-counting detector array for both data reception and beam tracking, thereby eliminating the need for redundant hardware and improving data throughput.
System Architecture
The proposed system utilizes a detector array that captures incoming photons from a source, such as a drone-mounted laser. The counts from individual detectors in the array are processed to estimate the beam's center position while simultaneously decoding the communication signal.
As shown in the system model, the light hitting the array is modeled as a Gaussian beam, and the noise is treated as a Poisson process. The output of the detector array is fed into a threshold device for data detection, but the spatial distribution of these counts is the key to the tracking mechanism.
Beam Modeling and Estimation
To track the beam effectively, the system must estimate the center of the light intensity profile. The intensity of the laser beam follows a Gaussian distribution across the two-dimensional plane of the detector array.
The researchers evaluated several non-Bayesian estimators to determine the beam's center position $(x, y)$:
- Low-Complexity Trackers: These include the Maximum Detector Count (MDC), which simply identifies the brightest pixel, and the Centroid Estimator, which calculates a weighted average of positions. Adaptive Centroid Estimators (ACE) further refine this by non-linearly weighting counts to focus on the most intense part of the beam.
- High-Complexity Trackers: The Nonlinear Least Squares (NLS) and the Maximum Likelihood Estimator (MLE) represent the upper tier of performance. While mathematically optimal, they require solving complex optimization problems, often utilizing genetic algorithms to find the global maximum.
Key Insights and Performance
Through extensive numerical simulations and the derivation of the Cramér-Rao Lower Bound (CRLB), several critical insights were discovered:
- The "Goldilocks" Beam Radius: A thinner beam is not always superior. If the beam is significantly smaller than a single detector cell, the system loses the gradient information necessary to track sub-pixel movements. An optimal beam radius exists that balances signal concentration with tracking sensitivity.
- Diminishing Returns of Pixel Density: While increasing the number of detectors (e.g., from $4 \times 4$ to $8 \times 8$) improves accuracy, the gains diminish as the array becomes more granular. The computational cost of high-resolution arrays must be weighed against these marginal performance improvements.
- Tracking as a Prerequisite for Detection: The study confirms that accurate tracking directly impacts the Probability of Error in Pulse Position Modulation (PPM). By knowing the exact beam position, the receiver can weight the most relevant pixels more heavily, significantly lowering bit error rates.
Conclusion
The research demonstrates that the Maximum Likelihood Estimator (MLE) consistently provides the lowest Mean-Square Error (MSE), closely approaching the theoretical CRLB. While high-performing trackers like MLE and NLS offer superior reliability, their computational overhead remains a challenge for real-time jitter compensation. Future work will focus on mitigating the effects of atmospheric scintillation and optimizing these algorithms for low-latency applications in dynamic environments.
Summary updated