Abstract

With the aim of speeding up the measurement time of Brillouin distributed sensors, the use of compressive sensing (CS) has been recently proposed to reconstruct the Brillouin gain spectral shape based on a reduced number of spectral samples. However, the real benefits of using CS are still unclear in terms of the achieved sensing performance, especially when compared to the performance of a conventional Brillouin sensor operating with the same reduced number of spectral samples and with no use of CS. Here, a thorough analysis of the performance of compressive sensing applied to distributed Brillouin sensors is presented. Based on experimental and theoretical results, the usefulness of CS for Brillouin spectral reconstruction is analyzed by investigating its impact on the tradeoff between sensing performance parameters, such as measurement time, signal-to-noise ratio (SNR), and frequency estimation errors. For this, the sensing performance achieved by CS under uniform and random sampling scenarios is compared to the performance obtained by a classical interrogation scheme, under similar SNR and spectral sampling conditions (i.e., under the same measurement time, fiber length and spatial resolution). Results point out that both classical and CS-based Brillouin sensing approaches lead to the same figure of merit, verifying that the use of CS for Brillouin spectral reconstruction provides no real benefits to improve the overall performance of the sensor.