Optimizing Brillouin Optical Time-Domain Analyzers Based on Gain Spectrum Engineering

Distributed fiber sensing (DFS) is a unique mechanism that employs a whole fiber as a sensing medium. It has attracted a lot of interest due to the low fiber loss; small size; and the ability to be easily embedded in various structures, such as dams, bridges, and buildings, to perform environment surveillance as an artificial nerve system. In comparison to applying numerous traditional point sensors, such as fiber Bragg gratings (FBG), it provides a more efficient and cost-effective solution in a wide range of large-scale sensing tasks, such as infrastructure and structural health monitoring¹.

Current distributed sensors exploit different scattering mechanisms inside the fiber to measure temperature and strain distribution. Among them, DFS based on Brillouin scattering is the most attractive due to the striking advantages of the stimulated Brillouin scattering (SBS), such as high signal-to-noise ratio (SNR), low threshold, and the sensitivity to both temperature² and strain³. SBS can be classically described as an interaction between the incident optical continuous waves (CW), i.e., the pump, and the counter-propagating CW probe wave via an acoustic wave. According to the conservation of energy and momentum, the probe wave is frequency downshifted to the pump. This shift is called Brillouin frequency shift (BFS). Considering the finite lifetime of a 10 ns acoustic wave, there is a finite spectral distribution of the refracted wave, also called Brillouin gain spectrum (BGS), in which the BFS is the frequency difference between the pump wave and the peak center frequency. The interaction between the waves leads to a frequency down-shifted gain region and a frequency up-shifted loss region where the probe wave gets amplified and attenuated, respectively. For a standard single mode fiber (SSMF) in C-Band, the BFS is approximately 11 GHz and the BGS has a Lorentzian shape with an ultra-narrow full-width at half maximum (FWHM) of 10-30 MHz, which can be further reduced to 3.4 MHz with specific techniques^4,5,6,7. Based on these characteristics, SBS can also be applied in microwave photonics filters^8,9,10, optical filters¹¹, slow and fast light^12,13,14, and high resolution optical spectroscopy^7,15.

One of the most promising SBS applications is distributed Brillouin sensing. These sensors exploit the BFS dependence on temperature and strain. The first to be demonstrated was the Brillouin optical time-domain analyzer (BOTDA)¹⁶, which is the most consolidated time-domain distributed Brillouin sensing technique. It differs from the conventional CW-SBS interaction in that it exploits the SBS interaction between a pulsed pump wave and a probe CW so that the environmental information is locally interrogated on every fiber section. The pump or probe frequency is usually fixed while the probe or the pump frequency is scanned in the vicinity of the BFS. The probe power is recorded for BGS reconstruction and the BFS is ideally the peak frequency of the local BGS at each fiber section. However, due to the inevitable system noise, the BGS peak is usually ambiguous and a fitting algorithm has to be applied¹⁷, which leads to a certain estimation error in frequency¹⁸ and influences the measurand resolution.

Statistically, the BFS estimation error is inversely proportional to the system signal-to-noise ratio (SNR). The most straightforward way to enhance the SNR is to increase the pump and probe power. However, these are limited by modulation instability (MI)¹⁹ and non-local effects (NLE)^20,21 to ~20 dBm and -14 dBm, respectively. Numerous techniques, such as coding²² and Raman based inline-amplification²³ have been proposed to break these limits. Recently, it has been reported that this frequency error can be minimized by choosing a proper fitting algorithm²⁴. Relatedly, measurements exploiting the Brillouin phase and a linear fitting algorithm are also reported to have a reduced frequency error²⁵, which indicates the potential of a well-engineered BGS to sensing performance enhancement. Another option to enhance SNR is noise reduction. However, according to the traditional point of view, the sensing system noise comes mainly from the detector (i.e., common-mode noise, including dark noise, shot noise, etc.)^26,27 and improvement is less likely.

The basic idea of this paper is to engineer the BGS by the superposition of a conventional BGS with two symmetric Brillouin loss spectra (BLS) (see Figure 1). In comparison to a conventional BGS spectrum, which follows a Lorentzian shape, the engineered spectrum is sharper and more robust with the same level of system noise. Thus, the noise has less influence on the determination of the peak frequency. This can be verified by collecting and fitting the BGS measurement data a statistically significant number of times. Because of this better resistance to the noise, sensing performance enhancements are achieved, including the sensing range by 60% and doubled measurand resolution, i.e., a 50% reduced frequency error. Due to the involvement of the measurement with Brillouin loss interaction in part of the engineered BGS, a direct comparison of the trace noise with and without Brillouin interaction is made. Owing to the circumvention of the excess Brillouin noise, the trace with the engineered BGS is much clearer.

Figure 1: Schematic of an engineered BGS by the superposition of one Brillouin gain and two symmetric Brillouin loss spectra. Please click here to view a larger version of this figure.