In our fibre-optic sensor, a laser pulse with a peak power below that required for the onset of non-linear effects is launched down an optical fibre. Backscattered Rayleigh and spontaneous Brillouin radiation are extracted from the input end of the fibre, spectrally separated, photodetected simultaneously, displayed on an oscilloscope and/or stored on a computer for further processing. The intensity of the spontaneous Brillouin radiation shows a temperature sensitivity of approximately 0.32%/K at 293K while the Rayleigh signal shows negligible temperature sensitivity. When ratioed against a broadband Rayleigh signal, which is used to reduce coherent Rayleigh noise, a spatial-resolved measurement of the fibre’s temperature can be obtained [1
K. De Souza, G. P. Lees, P.C. Wait, and T. P. Newson, “A diode-pumped Landau-Placzek based distributed temperature sensor utilising an all-fibre Mach-Zehnder interferometer,” Electron. lett.
33, 2174–2196, (1996). [CrossRef]
A measure of sensor performance is its range. Extending the measurement range results in a degradation of the signal-to-noise since the value of the input optical power is limited by nonlinear effects. A technique that can effectively enhance the receiver sensitivity by improving the signal-to-noise can lead to extension of measurement range. The motivation for this work was to extend the range of our sensor by using a cost-effective optical preamplifier system. This paper reports on the theory and experimental use of an optical preamplifier system in a fibre-optic distributed temperature sensor, as a means of improving receiver sensitivity and consequently sensor range. The optical preamplifier system consists of an erbium-doped fibre amplifier with 27dB gain, a three-port circulator and a 47GHz bandwidth in-fibre grating. A 17dB optical signal-to-noise improvement of the spontaneous Brillouin signal in a 23km fibre-optic sensor is demonstrated.
Our fibre-optic distributed temperature sensor uses a direct detection system consisting of an erbium-doped preamplifier system (EDFA) having a gain of 27dB, a single-pass fibre-optic Mach Zehnder interferometer (SPMZ) to optically separate backscattered Rayleigh and Brillouin signals and a receiver based on a photodiode/transimpedance amplifier as shown schematically. The detected optical signal is displayed and averaged on a digital oscilloscope and downloaded and stored in a computer.
The major noise sources due to an optical EDFA preamplifier followed by a direct detection transimpedance receiver are:- (i) photodiode noise consisting of shot noise of the photocurrent and dark current. (ii) transimpedance amplifier noise comprising: thermal noise of the feedback resistor, thermal noise of the FET channel and shot noise of the FET gate leakage current. (iii) EDFA noise comprising: ASE-shot noise, Signal-ASE beat noise, and ASE-ASE beat noise. The corresponding mean square noise currents [2
M. Gold, “Design of long-range single mode OTDR,” J. Lightwave. Technol.
3, 38–43, (1985). [CrossRef]
E. Desurvire, Erbium-doped fiber amplifiers , (John Wiley & sons,1994) 207–305.
R. M. Howard and R. D. Jeffery, “On the noise of high-transimpedance amplifiers,” Opt. Quantum Electron.
19, 123–129, (1987). [CrossRef]
] and the electrical signal, referenced to the input terminal of a transimpedance amplifier have been recast and are shown in table 1
. An optical bandwidth much greater than the electrical bandwidth has been assumed.
The parameters and their values used in our experiment which is described in section 3 are given by:- Quantum efficiency of detector η=0.63, operating wavelength λ=1533nm, electrical bandwidth of detector Be
=20MHz, equivalent noise bandwidth
, optical bandwidth Bo
=47GHz (0.37nm), feedback resistance Rf
=50 kΩ, Boltzmann constant k=1.38e-23J/K, Planck constant h=6.63e-34Js, electronic charge e=1.6e-19C, EDFA gain G (dB)=27dB (so numerical gain G=501), ASE power loss via SPMZ Lase
=3dB, thermodynamic temperature T=296K, FET transconductance gm
=5mS, total input capacitance of detector Ct
=8pF, dark current Id
=0.15nA, gate leakage current, Ig
=10pA, input optical power to EDFA Ps
=15pW, spontaneous emission factor nsp
=1 for fully inverted medium [3
E. Desurvire, Erbium-doped fiber amplifiers , (John Wiley & sons,1994) 207–305.
] and the optical noise figure F, of a EDFA is given by [5
P. Morkel and R. I. Laming, “Theoretical modelling of erbium-doped fiber amplifiers with excited-state absorption,” Opt. Lett.
14, 1062–1064, (1989). [CrossRef] [PubMed]
Table 1. Description of various noise sources and their corresponding mean square currents
|Description||Mean square currents|
|Photodiode noise consisting of shot noise of the photocurrent and dark current.||
|Thermal noise of the feedback resistor||
|Thermal noise of the FET channel||
|Shot noise of the FET gate leakage current||=2e Ig Beq
|ASE-shot noise||=2 e2η(G Lase) F Bo Beq
|Signal-ASE beat noise||
|ASE-ASE beat noise||=(ηe)2 F2(G Lase)2Bo Beq
The optical signal-to-noise ratio SNRo can be expressed in terms of measured voltages ‘V’ as:
where N is the no. of averages or signal integrations performed on the data and n represents the various noises: shot, thermal, FET, etc. It is instructive to examine the variation of the noise sources for different receiver bandwidths, which determines spatial resolution of the sensor. Figure 1
(a) shows unamplified and optically amplified signal levels and the variation of photodiode noise, transimpedance amplifier noise, and EDFA noise with receiver bandwidth for G=27dB, N=40960 and Bo
=47GHz. The input signal levels reflect values of spontaneous Brillouin signals obtained under experimental conditions at the far end of our sensor. The decrease in the amplified and unamplified signal levels with receiver bandwidth is due to the fact that in a fibre-optic distributed temperature sensor, a receiver bandwidth is selected to match the input pulse width in order to maximize the SNRo
and the backscattered energy is inversely proportional to the receiver bandwidth. The amplified signal is 27dB larger than the unamplified signal. Examination of the noise plots shows that photodiode noise is negligible in comparison to the other two sources for the signal levels under consideration. Also, the total noise is due mainly to EDFA noise (which is predominantly ASE-ASE beat noise) and coincides closely with it. Figure 1
(b) shows the corresponding SNRo
without and with optical amplification. SNRo
decreases with receiver bandwidth due to increasing noise. Without amplification, the signal is just discernable i.e. SNRo
=0dB, up to about 12MHz. On the other hand, with optical preamplification, the SNRo
>0dB over all receiver bandwidths, implying signal discernability throughout. For a 20MHz receiver bandwidth, SNRo
of approximately 12dB and -5dB are obtained with and without preamplification respectively, giving an improvement in SNRo
Fig. 1. Plots showing: (a) the unamplified and optically amplified signals and the variation of RMS noise voltages of photodiode noise, transimpedance noise and EDFA noise with receiver bandwidth. (b) Corresponding variation of optical signal-to-noise ratio with and without optical preamplification. An improvement of 17dB is shown for a 20MHz receiver. G=27dB, N=40960 and Bo=47GHz.
Fig. 2. Schematic experimental set up and preamplifier configuration.
A Q-switched fibre laser generated a pulse (1.5GHz bandwidth, 1533nm, 18ns, 250mW peak power) that propagated along 23km of standard telecommunications fiber divided into five sections (17km, 4.3km, 0.5km, 0.45km, 0.45km). The fourth section consisting of 0.45km was placed into an oven with a measured temperature of 333K±0.1. The other four sections were kept adjacent to one another at a room temperature of 296K±0.1. Backscattered Rayleigh and spontaneous Brillouin signals were routed by a circulator to - and divided by a 50/50 coupler with one output spliced on to the optical preamplifier system. This allowed an easy comparison to be made between optically unamplified (port A) and amplified backscattered signals although introducing a 3dB loss of signal. Unamplified and amplified signals were filtered in turn by a thermally tuned SPMZ (22.3 GHz FSR<1.5dB insertion loss, 22dB rejection) to separate the Rayleigh and Brillouin signals. The Brillouin signals were detected with a receiver system having a detection bandwidth of 20MHz. The detected signal was averaged 8192 times, digitized and stored on computer. Data from the far end of the sensing fiber, where signal strength is reduced, were averaged 40960 times and similarly stored. The preamplifier system with a measured noise figure of approximately 3.3dB is detailed in Fig. 2
. Backscattered radiation from the test fiber was amplified by a 980nm-pumped erbium-doped fibre and along with forward amplified spontaneous emission (ASE) were transmitted by a circulator to - and reflected from a narrowband in-fibre grating (1531.6nm centre wavelength, 47GHz optical bandwidth, >99% reflectivity). Tuning of the grating to the signal wavelength was made possible with a micrometer-controlled strain stage. Reflected radiation emerging from port 3 was fed to the insulated SPMZ.
(a) and (b) show the unamplified and amplified Brillouin signals respectively over 23km. Both were averaged 8192 times.
Fig. 3. (a) Unamplified and (b) 27dB amplified spontaneous Brillouin signal over 23km. Averages=8192.
Fig. 4. (a) unamplified and (b) 27dB amplified spontaneous Brillouin signal at far end of fibre sensor. Averages=40960.
In the unamplified case, the signal was very noisy and was just discernable up to about 10km. Beyond this, the receiver sensitivity was too poor and the signal was masked by the transimpedance amplifier noise. In contrast, the amplified Brillouin signal was discernable over the full range of the sensing fibre and attributed to dominant ASE-ASE beat noise. Figures 4
(a) and (b) show the unamplified and amplified Brillouin signals at the heated section of the sensing fibre. Both signals were averaged 40960 times. There was no indication of any temperature change in the unamplified case. An optical signal-to-noise ratio, SNRo
, of approximately -5dB was calculated from the plot. The improvement in the signal-to-noise of the detected preamplified Brillouin signal is visible in Fig. 4
(b). An increase in Brillouin power at the heated section was observed. The calculated mean value of SNRo
was 12dB. The improvement in SNRo
due to optical preamplification was approximately 17dB (12+5). This agrees with the theoretical value of 17dB calculated in Section 2.
An improvement of 17dB in the signal-to-noise via our preamplifier system is invaluable in reducing the measurement time and extending the range of the sensor. Without preamplification, for a given measurement range, performing many more signal integrations or averages can reduce the measuring error and produce the same results achievable with preamplification but at the expense of degrading measurement time. Indeed for a single-mode fiber with 0.2dB/km attenuation, a 17dB improvement in signal-to-noise translates to approximately 40km extension in measurement range. Though not presented in this paper, by removing the 50/50 coupler in Fig. 2
, ratioing the resulting amplified Brillouin signal against a broadband Rayleigh signal (~3nm) gave a temperature sensor with 6K resolution [6
K. De Souza and T. P. Newson, “Improvement of signal-to-noise capabilities of a distributed temperature sensor using optical preamplification,” Meas. Sci. Technol. , 12, 952–957, (2001).
]. The latter was due to imperfect filtering of the Rayleigh signal by the SPMZ. This can be improved by using a double-pass Mach-Zehnder interferometer [7
K. De Souza, P.C. Wait, and T. P. Newson, “A double-pass configured Mach-Zehnder interferometric optical filter for distributed sensing,” Electron. lett.
33, 2148–2150, (1997). [CrossRef]
], which in addition to providing improved rejection of the Rayleigh signal would also reduce the forward ASE transmitted to the receiver. The EDFA-based preamplifier system with its low noise figure of ~3.3dB is more aptly suited than electronic amplifiers such as avalanche photodiodes which have higher noise figures (typically 4-6 dB) for comparable detection bandwidth. The use of preamplification can further be applied to a recent technique in which a Raman pump pulse was used to amplify a probe pulse (which generated the backscattered spontaneous Brillouin signal), at some distance away from the input end of the sensing fibre [8
Y.T. Cho, M. N. Alahabi, M. J. Gunning, and T. P. Newson, “50km single-ended spontaneous Brillouin-based distributed temperature sensor exploiting pulsed Raman amplification,” Opt. Lett.
28, 1651–1653, (2003). [CrossRef] [PubMed]
]. An improvement in temperature resolution of 27°C (40°C to 13°C) was achieved at a distance of 50km. Further improvement in measurement range can be achieved by integrating an EDFA based optical preamplifier into the system.