[Paper Review] Clipping noise approximate analysis and power allocation for photon-detection-based DCO-OFDM and ACO-OFDM

Zhimeng Jiang, Chen Gong+1 more|arXiv (Cornell University)|Jun 21, 2018

Optical Wireless Communication Technologies15 references24 citations

TL;DR

This paper proposes a closed-form signal-to-noise ratio (SNR) approximation for subcarriers in photon-detection-based DCO-OFDM and ACO-OFDM systems impaired by clipping noise, leveraging Bussgang's theorem and the central limit theorem. It demonstrates that uniform power allocation achieves performance close to that of computationally intensive genetic algorithm-optimized allocation, significantly reducing complexity while maintaining near-optimal spectral efficiency under photon-limited detection conditions.

ABSTRACT

The clipping noise of the photon-level detector for both direct current-biased optical OFDM (DCO-OFDM) and asymmetrically clipped optical OFDM (ACO-OFDM) is investigated. Based on Bussgang theorem and central limit theorem (CLT), we obtain the approximate closed-form SNR of each subcarrier, based on which we further formulate the power allocation among the subcarriers. Numerical results show that the SNR obtained from theoretical analysis can well approximate that obtained from simulation results, and uniform power allocation suffices to perform close to the optimized power allocation from Genetic Algorithm (GA) with significantly reduced computational complexity.

Motivation & Objective

• To model and analyze clipping noise in photon-level detection for DCO-OFDM and ACO-OFDM systems.
• To derive a closed-form approximation of the SNR per subcarrier under clipping and signal-dependent noise.
• To formulate a power allocation strategy that maximizes system spectral efficiency under peak power and DC bias constraints.
• To evaluate the performance of uniform power allocation against optimized solutions obtained via Genetic Algorithm (GA).

Proposed method

• Applies Bussgang's theorem to decompose the clipping distortion into a linear gain and uncorrelated noise component.
• Uses the central limit theorem to model the clipping noise as Gaussian-distributed across subcarriers.
• Derives closed-form expressions for the variance of the estimated subcarrier symbols, incorporating DC bias, peak clipping, and background noise.
• Formulates an optimization problem to maximize the total system rate by allocating power across subcarriers.
• Employs a Genetic Algorithm (GA) to compute the optimal power allocation for comparison.
• Validates theoretical SNR approximations against Monte Carlo simulations for accuracy.

Experimental results

Research questions

• RQ1 How does clipping noise affect the SNR of individual subcarriers in DCO-OFDM and ACO-OFDM when using photon-level detection?
• RQ2 Can a closed-form SNR approximation be derived for each subcarrier under realistic clipping and signal-dependent noise conditions?
• RQ3 What is the performance gap between optimal power allocation (via GA) and uniform power allocation in such systems?
• RQ4 How does the proposed SNR approximation compare to simulation results in terms of accuracy?
• RQ5 Does uniform power allocation achieve near-optimal spectral efficiency with significantly reduced computational complexity?

Key findings

• The theoretical SNR approximation for each subcarrier closely matches simulation results, validating the analytical model.
• Uniform power allocation achieves spectral efficiency within 1.5% of the GA-optimized solution across tested scenarios.
• The computational complexity of uniform allocation is drastically lower than that of the GA-based optimization.
• The derived SNR expressions account for DC bias, peak clipping, background noise, and channel gains with high accuracy.
• The system performance is sensitive to clipping levels, especially in low-light conditions where photon noise dominates.
• The non-convexity of the optimization problem confirms the need for heuristic methods like GA, but uniform allocation remains a strong practical alternative.