Analysis of enhanced-performance fibre Brillouin ring laser for Brillouin sensing applications

In this work, we present an enhanced design for a Brillouin ring laser (BRL), which employs a double resonant cavity (DRC) with short fiber length, paired with a heterodyne-based wavelength-locking system, to be employed as a pump-probe source for Brillouin sensing. The enhanced source is compared to traditional long-cavity pump-probe source, showing a significantly lower relative intensity noise (~-145 dB/Hz in the whole 0– 800 MHz range), a narrow linewidth (10 kHz), and large tunability features, resulting in an effective pump-probe source in BOTDA systems, with an excellent pump-probe frequency stability (~200 Hz), which is uncommon for fiber lasers. The enhanced source showed an improved signal-to-noise ratio (SNR) of about 22 dB with respect to standard BRL schemes, resulting in an improved temperature/strain resolution in BOTDA applications up to 5.5 dB, with respect to previous high-noise BRL designs. © 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

. In partic value to span simplification with respect t layouts, BRL extracted from temperature/s the cavity, in further improv Fig. 1 schem In order t increase the p that can be ac of long cavity for both pump explored in [1 implemented inherent mul characterized sensing appli temperature/s the implemen cavity (DRC) signal stabilit where further [18], where fo interferometer system presen frequency shi tunable probe a narrow line uncommon fo sensor system

Double re
In BOTDA sy sensing fiber tune the pum remain stable cular, such sch the Brillouin n of the electro o OSB based B L resonators ad m the cavity [1 train resolution n particular wh vements. 1 along the cessary to shift will he optical probe directly influences the frequency resolution for the BFS, which, in its turn, determines both temperature and strain resolution [19,20].
The scheme of a BOTDA system employing our DRC-BRL and the stabilization scheme is shown in Fig. 1(a). The pump and probe are generated by the same DFB laser source (λ = 1553.26 nm, 350 kHz wide linewidth). In the pump branch, the DFB seed is amplified through an erbium-doped fiber amplifier (EDFA) and pulsed by an electro-optic modulator (EOM). In the probe branch, the BRL generating the tunable probe light is coupled into the λstabilization scheme described below and modulated by a low-bandwidth EOM. The stabilization scheme (explained below in detail) is used to lock the probe to the pump signal and tune their frequency shift over a range of frequencies. The Standard BOTDA traces are then obtained by acquiring the detected amplified probe light intensity through a PIN photodiode and a fast analog-to-digital-converter (ADC) with the subsequent data processing to reconstruct the Brillouin Gain Spectrum (BGS). The Stokes signal is selected through reflection by a fiber Bragg grating (FBG) filter (also reducing unwanted noise and and improving SNR of BOTDA traces as explained below in section 3.4) and a second optical circulator than couples it into the PIN photodiode. The FBG used in the set-up has a 6 GHz large 3dB reflection bandwidth.
The structure of the BRL is shown in Fig. 1(b): two optical couplers are used to define a fiber ring. Through one of the couplers (C1), the light from the laser pump is injected into the ring. Once the light circulating inside the cavity reaches a given intensity threshold, a counterpropagating and frequency downshifted optical signal (Stokes light) is generated and amplified by Stimulated Brillouin Scattering (SBS), and then amplified through multiple loop propagation inside the ring, further increasing intensity and narrowing its linewidth. The Stokes output is then coupled out through C1 and inserted into the BOTDA through an optical circulator (OC).
More in detail, in the DRC-BRL shown in Fig. 1(b), the pump is provided by a frequency stabilized DFB laser whose operating wavelength matches the ring resonant frequency and is coupled to an optical circulator (OC) and then, through an optical coupler (C1), enters the fiber ring cavity, which is given by a single-mode fiber spool (SMF) (< 10 m length). The BFS value of the fiber employed in the ring lies below 10 GHz so that, as will be described in the following sections, only the lower frequency side-band signal generated by the wavelength-locking scheme modulator lies within the BGS of the sensing fibers that commonly show BFS values ~10.6 GHz. The ring fiber is thermally stabilized within a laboratory case and located on an active vibration isolating table in order to reduce acoustic noise and thermal fluctuations that are typically responsible for low-frequency (<100 kHz) intensity and phase noise in fiber ring laser signals. To ensure optimal polarization coupling between pump and Stokes lightwaves, which is necessary to maximize SBS gain, two polarization controllers (PC1 and PC2) are used before the ring and inside it respectively. The counter-propagating (counterclockwise) Stokes light from the ring cavity is coupled out from the 3 rd port of OC as the BRL output. A second optical coupler (C2) allows to monitor both re-circulating pump and Stokes. When BRLs are used as pump-probe sources, the Stokes signal is subject to both intensity and phase noise which can degrade the overall performance [21]; in particular, the frequency noise in the Stokes signal is influenced by the so-called mode hopping. A schematic representation of the mode hopping effect is given in Fig. 2. In mode hopping, thermal noise and vibrations alter the cavity free spectral range (FSR), causing a shift in the dominant lasing mode [22].    wer as a function nt wavelength peak double resonan n be found betw separation bet we used a tuna response, as a n to 4 m). In F elength for a ca followed by l separation ∼5 t mode peak s g. 3(b

Wavelength locked DRC-BRL
One of the issues of practical DRC-BRL is that while fluctuations of the cavity length induced by environmental vibrations and thermal instabilities no longer cause mode hopping, they can still detune the pump resonance reducing overall lasing stability [21] alongside the combination of Kerr effect and mode pulling [22]. DRC-BRL lasing can be stabilized by a variety of techniques such as those that use feedback systems locking the cavity length [25] or the pump wavelength [26]. To counteract the detuning effect, improve the pump-probe frequency shift stability and reduce the intensity noise, we employed an active DRC-BRL lasing wavelength locking system which also succeeds at compensating frequency drift, further improving the frequency-shift noise.

Wavelength locking scheme
The employed scheme is depicted in Fig. 4. From the DRC-BRL described in Section 2, small fractions of both output pump and BRL Stokes lightwaves are redirected by two optical couplers (S1 and S2) into the 3-dB optical coupler S3, and their combined beating signal BRL f Δ is converted into an RF tone by a fast photo-detector (PD). A harmonic mixer is used to mix the RF tone at the pump-Stokes frequency difference (Δf BRL = f pump -f BRL < 10 GHz) with the signal from a tunable local oscillator whose frequency (f LO ) is always kept higher than Δf BRL (f LO >Δf BRL , in our experiment f LO = 10.7-11.5 GHz); two sidebands (f LO ± Δf BRL ) are resulted from harmonic mixing. A low-pass filter removes the (unwanted) upper sideband, while the frequency-difference RF component (f LO −Δf BRL ), which in our experiment is lying in the 100-900 MHz range, is used to drive a slow (<1 GHz bandwidth) Mach-Zehnder electro-optical modulator (EOM). Corresponding optical sidebands are thus generated in modulation of Stokes light by EOM, while a suitable biasing of the EOM ensures efficient suppression of the (Stokes) carrier (>20 dB suppression). Of the two sidebands in the modulated Stokes light, the frequency of the lower sideband (f LSB = f BRL -f LO + Δf BRL = f pumpf LO ) is shifted from the pump by the frequency of the tunable oscillator f LO . This way, it is possible to tune the pump-probe frequency shift from the initial Δf BRL automatically provided by the BRL to the values required to span the whole BGS of the sensing fiber. Since the BRL signal frequency < 10 GHz does not lie within the gain spectrum of the sensing fibers that are commonly employed, the upper sideband signal that is generated by the EOM does not interact with the pulsed pump in the BOTDA set-up and does not have to be suppressed. In addition, any frequency perturbation in the ring, which would change the BRL output to f BRL = f BRL + δf, would also lower the Δf BRL by -δf, causing the frequency that is fed into the EOM to change from (f LO −Δf BRL ) to f LO − (Δf BRL -δf), resulting in a modulation change of exactly δf. This would result in a perturbed lower sideband frequency equal to f LSB = f BRL + δf-f LO + Δf BRL -δf = f pump -f LO . This means that every perturbation in the BRL cavity is effectively compensated by the EOM, leaving the final pump-probe frequency shift unalterated. It's worth noting that OSB technique employs modulators having bandwidth >10-11 GHz to directly modulate the probe signal from the pump frequency values to those required to span the BGS of the sensing fiber.

Frequen
In order to ve of the pumpintensity nois a local oscilla resulting RF t acquired spec timescales: 1 measurement, (FWHM) is ∼ same order o can be highly and on the lon Similar lin oscillator rang with the DRC probe source. In addition assessing app linewidth, thu hence employ sential for lute BRL BRL. We linewidth of the probe lightwave. A Mach-Zehnder interferometer with a 12 km long fiber delay line in one arm and a 150 MHz acusto-optic modulator in the second arm is used for this purpose with a fast (12 GHz bandwidth) photodetector and ESA for the analysis of the beat signal. The optical path difference between the two arms allows to measure signal having spectral linewidth values down to about 6 kHz [27]. The attained full width at half maximum (FWHM) of the spectrum for signal power of 0.75 mW, pump power of 12.6 dBm and ~200 μW extracted from the Mach-Zehnder intereferometer (shown in Fig. 6) results to be ~10 kHz, thus significantly narrower than the original DFB laser linewidth (350 kHz). The observed narrowing of the Stokes signal is due to the combined influence of the acoustic damping and the cavity feedback of the ring resonator [13]. The expected Stokes spectral linewidth value S ν Δ , considering situations with FSR values comparable to the BGS bandwidth, that is the case of short-cavity layout, is given by [13] is the pump linewidth, B ν Δ is the Brillouin linewidth. Γ is a parameter depending on the length L and coupling ratio of the cavity R and is defined as: / ln . c nL R Γ = Using the experimental values of the used BRL cavity, we obtained an expected Stokes linewidth approximately of 5 kHz which is in line with self-heterodyne measurements. Outside the frequency range reported in Fig. 5, the signal is within the noise due to the SNR of the measurement set-up, which limited the measure sensitivity to about −80 dBm; however, the attained power levels were 20dB below the peak values, and no additional peaks were observed. Furthermore, we have performed spectral linewidth measurements on BRL signal with different input pump power values above the laser threshold; in particular, we have carried out measurements of the BRL signal with power ranging from 0.5 to 1.2 mW and no significant variations of the spectral linewidth values have been observed.
One notable advantage of this active stabilization scheme is that, thanks to the removal of the noisy fluctuation and the optical filtering provided by the DRC-BRL cavity, the Stokes light exhibits a linewidth that is narrower than the original DFB pump laser. Moreover, as described above, in this scheme the anti-Stokes sideband, which is represented by the upper sideband created by the modulator, has frequency of f USB = f BRL + f LO -Δf BRL which does not lie within the Brillouin sensing region (unlike e.g. direct DFB modulation used in standard OSB methods) and thus it doesn't need to be filtered out, allowing for simpler detection schemes. In addition, since the output from the BRL cavity is especially close to the pumpprobe frequency shift, the EOM only needs to tune/modulate at a frequency f LO -Δf BRL , which is in the range of 100 to 900 MHz for very large tuning ranges. This means that it is possible to employ a slower-response EOM (hundreds of MHz) compared to the faster EOMs needed in direct modulation schemes (>10 GHz bandwidth). Finally, the re-circulating pump extracted from the Brillouin ring laser cavity experiences a similar filtering effect as the DRC-BRL Stokes light. The proposed scheme hence allows us for a wide, stable tuning of the frequency shift that is expected to allow for an accurate reconstruction of the Brillouin gain spectrum in BOTDA, as it will be shown in the next section. Please note that the wavelength locking scheme circuitry can be easy implemented using commercial low-cost electronic components.

Intensity
In addition to stabilized DR exhibit larger longer cavity pump-signal n stabilization f for the stabili well as the pu signal power where P is t large number light using a f spectrum ana subtracting th The result BRL, free-run 7, it can be no low frequency RIN across t frequency ma the RIN level levels, i.e. bet other hand, R and remain ap frequency spe were higher th becomes evid stabilizing the that are typica especially at l

SNR improvement in BOTDA applications
By evaluating the signal to noise ratio (SNR) of the probe signal in BOTDA applications, it is possible to provide a quantitative estimate on the effect of the potential SNR improvement thanks to reduced RIN that would arise by using the DRC-BRL instead of the LC-BRL as the probe source for a BOTDA sensor. In a typical BOTDA configuration (refer to Fig. 1(a)), the SNR of the detected probe signal at the photodetector can be expressed as [ δ are the noise variances given respectively by thermal noise and shot noise of the photodetector, spontaneous-spontaneous and signal-spontaneous beat noise of the EDFA and the RIN of the optical probe. Note that the FBG placed at receiver side contributes removing out-of-band light noise sources possibly impairing the SNR (e.g. ASE from EDFA and related ASE-signal beat noise, Rayleigh back-scattered pump light etc). The 3 dB bandwidth of FBG (6 GHz) was large enough to keep the Stokes signal frequency far from the slope of FBG reflection spectrum (thus also limiting the possible noise contributions generated by the phase-tointensity conversion of signal phase noise due to thermal drift of the FBG bandwidth). Unlike with probes from most low-RIN sources, where the main limit to the SNR is due to the spontaneous-signal beat noise 2 sp s δ − , in probe signals generated by a noisy fiber ring laser the most important contribution is given by the RIN [30]. The SNR, together with the Brillouin linewidth B δν , determines the minimum detectable frequency shift of the reconstructed Brillouin gain spectrum fit for the acquired BOTDA trace at a given fiber location. Actually, the frequency resolution in BFS measurements on standard BOTDA systems can be found using the following expression: and from this value it is hence easy to estimate corresponding temperature and strain resolutions, which are respectively given by [31]: where T C and S C are the linear temperature and strain coefficients, respectively, while represent Brillouin frequency shifts of sensing fiber at reference temperature and of unstrained sensing fiber, respectively [31].
It is hence possible to provide an estimation of attainable resolution improvements for BOTDA applications by using a wavelength-locked DRC-BRL instead of a higher-RIN LC BRL. In the following analysis, the major noise source is assumed to be given by the laser RIN, and other noise sources (see Eq. (10)) are considered of smaller extent with respect to the laser intensity fluctuations. Clearly, while this approach is valid only with noisy laser sources such as fiber lasers (as for our BRL), it gives a good estimation about whether the BRL intensity noise can be a limiting parameter for BOTDA applications.
The resulting probe SNR due to source intensity fluctuations (SNR RIN ) has actually been calculated by integrating the measured RIN values over the BOTDA receiver bandwidth (125 MHz). This resulted in a SNR RIN of ~38.7 dB for the standard long-cavity BRL, and of ~61 dB for the λ-locked DRC-BRL. The short-cavity double-resonance λ-locked BRL scheme hence allows a notable SNRRIN improvement of ~22.3 dB.
Referring to the BOTDA set-up shown in Fig. 1(a), and to the details and results of experiments reported in [12] for the standard BRL configuration, it is hence possible to infer the SNR and resolution improvements attained by using the new DRC-BRL scheme. Actually, it can be easily seen from Eqs. (10)-(12) (in the assumption of RIN as the prevalent component in probe fluctuations and in detected SNR), that the resolution improvement (in terms of frequency, temperature or stain resolution) achievable with the DRC-BRL can reach a value up to 5.5 dB.

Conclusions
In conclusion, we reported on an enhanced performance fiber Brillouin ring laser (BRL) exploiting a doubly resonant cavity (DRC) using a short single-mode fiber length, combined with an optical wavelength locking technique based on heterodyne detection. The novel DRC-BRL configuration has shown to provide a highly stable and tunable probe light, which can be highly suitable for BOTDA sensing applications. The probe source intensity noise characterization showed a measured RIN of ~-145 dB/Hz across the whole 0-800 MHz range; the measured laser linewidth resulted to be only 10 kHz, and an excellent pump-probe frequency stability was observed (pump-probe stability was 200 Hz with 10 ms integration time and 400 Hz with 120 s integration time). It's worth noting that, unlike in the PLL and OSB layouts, BRL resonators add a beneficial linewidth narrowing effect on the probe signal extracted from the cavity which can be used to further improve BFS and consequently temperature/strain resolution. The DRC-BRL light showed an improved intensity-noise signal-to-noise ratio (SNR) of about 22.3 dB with respect to standard BRL schemes, resulting in improved temperature/strain resolutions in BOTDA applications up to 5.5 dB with respect to previous high-noise BRL designs. The carried out analysis then indicates that stabilized DRC-BRL could be successfully employed as dual pump-probe source in accurate Brillouin optical time-domain sensor system applications, as an efficient and cost-effective alternative to the solutions based on PLL and OSB techniques, with a narrow linewidth (10 kHz) and an exceptional pump probe stability (~200Hz) which is uncommon for fiber lasers.