Performance analysis of frequency shift estimation techniques in Brillouin distributed fiber sensors

The performance of post-processing techniques carried out on the Brillouin gain spectrum to estimate the Brillouin frequency shift (BFS) in standard Brillouin distributed sensors is evaluated. Curve fitting methods with standard functions such as polynomial and Lorentzian, as well as correlation techniques such as Lorentzian Cross-correlation and Cross Reference Plot Analysis (CRPA), are considered for the analysis. The fitting procedures and key parameters for each technique are optimized, and the performance in terms of BFS uncertainty, BFS offset error and processing time is compared by numerical simulations and through controlled experiments. Such a quantitative comparison is performed in varying conditions including signal-to-noise ratio (SNR), frequency measurement step, and BGS truncation. It is demonstrated that the Lorentzian cross-correlation technique results in the largest BFS offset error due to truncation, while exhibiting the smallest BFS uncertainty and the shortest processing time. A novel approach is proposed to compensate such a BFS offset error, which enables the Lorentzian cross-correlation technique to completely outperform other fitting methods.