Over the past decade, fiber optic sensors have been gaining importance in different sectors of industry, mainly due to their unique properties that are unavailable with conventional electrical sensors. Whilst a brooder introduction of fiber sensing solutions into practical systems is desired by end-users, such an introduction is often limited by their complexities [1
Z. Jin and M. Song, “Fiber Grating Sensor Array Interrogation With Time-Delayed Sampling of a Wavelength-Scanned Fiber Laser,” IEEE Photon. Technol. Lett.
16(8), 1924–1926 (2004). [CrossRef]
P. Tsai, F. Sun, G. Xiao, Z. Zhang, S. Rahimi, and D. Ban, “A New Fiber-Bragg-Grating Sensor Interrogation System Deploying Free-Spectral-Range-Matching Scheme With High Precision and Fast Detection Rate,” IEEE Photon. Technol. Lett.
20(4), 300–302 (2008). [CrossRef]
] and, consequently, the prohibitive cost of available fiber-optic sensor interrogation systems. There is a considerable effort in the industry to reduce the cost of fiber-optic signal interrogators. For example, several efforts have been made by various companies to miniaturize conventional grating interrogators and thus reduce the cost. While the cost of miniaturized grating spectrometers is considerably lower when compared to their conventional counterparts, such systems still rely on relatively expensive and low-production-volume components such as gratings, liner InGaAa arrays, and SLEDs. The research and development of cost-effective, compact, and reliable signal interrogation concepts is, therefore, essential for any further successful introduction of fiber sensing technology into practical applications.
This paper presents a system for the interrogation of fiber optic sensors that change their distinctive spectral characteristics within a narrow band of optical spectrum. The proposed system can, for example, interrogate fiber Bragg grating (FBG), all-fiber Fabry-Perot (AFFP), and similar sensors. It utilizes the wavelength sweep of a standard distributed feedback laser diode (DFB). DFB diodes have been successfully applied in the past for fiber-optic sensor integration [4
J. Chang, Q. Wang, X. Zhang, D. Huo, L. Ma, X. Liu, T. Liu, and C. Wang, “A Fiber Bragg Grating Acceleration Sensor Interrogated by a DFB Laser Diode,” Laser Phys.
19(1), 134–137 (2009). [CrossRef]
D. Tosi, M. Olivero, and G. Perrone, “Low-cost fiber Bragg grating vibroacoustic sensor for voice and heartbeat detection,” Appl. Opt.
47(28), 5123–5129 (2008). [CrossRef]
], however the described systems usually employed a significant number of external components or were limited to dynamic measurements [7
T. Liu, C. Wang, Y. Wei, Y Zhao, D Huo, Y Shang, Z Wang, and Y Ning, “Fibre optic sensors for mine hazard detection,” J. Phys.: Conf. Ser.
178, 012004 (2009). [CrossRef]
]. We demonstrate that modern integrated telecommunication laser diode modules, intended for use in dense wavelength division multiplexing (DWDM) systems, can be directly applied to absolute and high-resolution spectrally resolved interrogation of fiber optic sensors, while requiring only a small number of additional low-cost opto-electronic components.
2. The physical configuration of the proposed interrogation system
The physical (hardware) configuration of the proposed sensor interrogation system is presented in Fig. 1
. This system was built around a Oclaro LC25-EW series DWDM laser module, which is an integrated photonic device that includes a DFB laser diode, wavelength locker, optical isolator, and temperature sensor (negative temperature coefficient thermistor -NTC), all mounted on a thermoelectric cooler (TEC), and packaged within standard 14 pin butterfly case. Such modules are standard and cost-effectve devices frequently used in today’s telecommunication systems. In addition to the laser module, only two PIN photo-detectors (detector 1 and detector 2) and a 2x2 coupler were required to complete the photonic part of the proposed sensor integration system. Detector 1 receives optical power reflected back from the sensor, whilst detector 2 measures the total power emitted by the laser diode.
Fig. 1 Sensor interogation system (TEC-thermo-electric cooler, NTC- negative temperature coefficient thermistor, TIA- transimpedance amplifier, PCG-programmable current generator)
The electronics part further included a low-cost digital signal processor (DSP) (dsPIC33FJ256GP710), four transimpedance amplifiers (TIA), programmable current generators (PCG) for supplying DFB laser diode and TEC, and a temperature sensing circuit.
3. The basic operation and spectral range of the proposed interrogation system
The interrogated sensor’s spectral characteristic is obtained by a sweeping of the laser diode wavelength by temperature and drive-current modulation, whilst providing simultaneous measurement of the emitted wavelength. An appropriate algorithm, executed by DSP, then extracts any relevant spectral features, such as, for example, peak wavelength from the acquired spectral characteristics in order to determine the value of the measured parameter.
Typical telecommunication DFB laser diode wavelength dependence on the temperature is around 90 pm/°C. Similarly, the DFB diode drive current wavelength sensitivity corresponds to approximately 7 pm/mA. The temperature control of the laser diode, therefore, allows for the slow scanning of the laser wavelength within the range that exceeded 3.5 nm. However, the achievable interrogation system wavelength scanning range also depends on a useful range of wavelength locker spectral properties (shown later in Fig. 2
.) which is, in this particular case of an Oclaro LC25-EW module, around 2 nm. The diode drive current sweeping allows for fast wavelength scans within the 0.3 nm wavelength range. These estimated tuning ranges are chosen conservatively to allow operation of the laser module over the full industrial environmental temperature range.
Fig. 2 Typical measured response of LC25-EW module wavelength locker detectors to laser wavelength variation
4. Application of the wavelength locker for determining the emitted wavelength
Whilst the temperature and diode drive current could be directly correlated to the wavelength emitted by the DFB diode, these correlations cannot be used to obtain high-resolution and high accuracy measurements, predominately due to the laser diode’s aging effects, and a limited ability to obtain the exact temperature of the laser diode PN junction from the temperature measured by a thermistor built within the laser module (the thermistor mass and position relative to the laser chip PN junction generate unavoidable delays and static discrepancies in temperature measurements that result in errors, especially when dynamic variation of temperature and current are used to sweep the laser’s wavelength).
Repeatable and high-resolution measurement of the emitted wavelength was therefore obtained by application of the wavelength locker. The Oclaro LC25-EW module utilizes a wavelength locker composed of two photo-detectors which are illuminated by diverging the laser beam through a tiled Fabry-Perot etalon (Fig. 1
The tilting and proper geometrical configuration of the etalon provides for the detection of two different narrow-wavelength bands by each individual etalon photo-detector. Figure 2
shows practically recorded optical power at wavelength locker detectors 3 and 4 versus emitted wavelength for typical Oclaro LC25-EW module. During the normal wavelength locking operation in the DWDM systems (e.g. when a locker is used to stabilize the wavelength), the currents generated by both detectors are subtracted to form a wavelength error parameter, which is actively maintained at near its zero value by a closed loop control of the laser diode temperature (the wavelength at which both detectors produce identical currents, thus corresponding to the locking wavelength of the module).
However, the currents generated by both lockers’ detectors can be utilized for high-resolution wavelength measurements. In order to obtain stable wavelength measurements, independent of diode drive current, temperature and diode aging, the difference in optical powers incident on both locker’s detectors was first normalized by total power incident on both detectors:
represent the optical powers incident on the wavelength locker detectors, while w0(λ0)
represents the dimensionless parameter obtained at the wavelength λ0
The exact dependence of parameter w
on wavelength λ
over the entire wavelength range of interest must be known, in order to determine the wavelength λ0
from the known value of parameter w0
,. We refer to this dependence as wavelength locker function W(λ).
Wavelength locker function W(λ)
was experimentally acquired using a separate calibration process. The acquisition of W(λ)
was achieved by recording parameter w
during a slow, temperature-controlled wavelength sweep of the laser diode over the entire wavelength’s working range, whilst using an external high-resolution spectrum analyzer for obtaining referenced wavelength readings. Figure 3
presents a recorded wavelength locker function W(λ)
for typical Oclaro LC25-EW laser diode module.
Fig. 3 Measured and interpolated wavelength locker function W(λ) obtained during the calibration process for typical Oclaro LC25-EW DWDM module
Several modules from the same LC25-EW series intended for use at the same ITU channel were analyzed and it was noticed that wavelength locker functions, whilst similar, were insufficiently identifiable to allow for interchangeability of modules, without individual calibration of their wavelength locker functions. In high-resolution wavelength measurements the calibration process should, therefore, be performed for each individual laser module. After the completion of the calibration process, the wavelength locker function W(λ)
was interpolated by cubic splines as shown in Fig. 3
. Coefficients of cubic spline functions were stored into DSP memory for further use.
5. Measurement algorithm and the operation of an experimental fiber optic sensor interrogation system
In order to instigate the measurement system, the temperature of the diode was first swept for approximately 22 °C, whilst simultaneously performing recordings of the sensor’s reflectivity (e.g. recording of the ratio det 1/det 2). This temperature sweep induced change in the diode wavelength over a full 2 nm wide spectral range, and generated a lock-up table that linked the sensors reflectivity with DFB diode temperature. The DFB diode’s temperature, at which peak reflectivity occurred, was sought, and the DFB diode temperature was further set at this temperature, in order to approximately adjust the laser wavelength to the sensor’s peak wavelength. Once this initialization process was completed, a high repetition rate 48 mA current modulation (corresponding to about a 0.3 nm wavelength sweep range) was applied to the laser diode in order to rapidly scan the vicinity of the spectral peak. During these current sweeps, the optical powers at all four detectors (both wavelength locker detectors, photo-detector 1 and 2) were simultaneously recorded, and stored in the memory of the DSP for further processing. A slower, closed-loop control of the laser diode temperature was also applied by constantly adjusting the diode temperature, in such a way as to maintain the sensor’s wavelength characteristic’s peak position in the middle of the wavelength scanning range induced by a 48 mA current sweep. This provided for a fast tracking of the sensor’s peak wavelength position when only limited changes in peak position occurred under influence of the measured parameter.
In our particular case, the temperature scan over an entire 2 nm wavelength range took around 4 s, whilst the current sweeps (including corresponding on-line data processing) were executed at 30 Hz. The 30 Hz rate was limited, in our case, by DSP signal processor performance and could be further increased by higher performance DSP.
During each diode current sweep cycle, 25 detector’s readings were recorded over the 300 pm wavelength range (one reading included data set from all four detectors). For each individual reading, detector 1 and 2 values were first divided in order to obtain normalized sensor reflectivity. This normalization is required due to significant changes in laser diode output power caused by temperature and, in particular by current modulation. Similarly, for each reading, the parameter w0
was calculated according to Eq. (1), by measurement of the wavelength locker detector’s currents. The parameter w0
was further used to determine the exact wavelength λ0
at which particular reading was recorded, by numerically solving the equation:
was the wavelength locker function obtained in the calibration process and stored in the form of cubic splines in DSP memory, as described earlier. Since Eq. (2) yields, in general, more than one solution, a valid wavelength solution was selected by additional calculation and comparison of the parameter’s w
slope sign with the slope sign of the stored wavelength locker function W
at the obtained multiple wavelength solutions. Since current induced wavelength sweep is performed during each measurement cycle and since the emitted laser diode wavelength is proportional to the laser diode drive current I
, we determined the sign of the slope dw/dλ
by calculating the sign of the slope dw/dI
. In other words, the correct wavelength solution λ0
of Eq. (2) for particular reading fulfils the condition Sig[dW(λ)/dλ
λ = λ0
I = I0
, where I0
represents the laser diode sweep current corresponding to the reading at which value w0
was recorded. Utilization of this additional condition allowed for usage of the wavelength locker over the full 2 nm wide wavelength range.
The algorithm described above, essentially generated a lock-up table in the DSP memory that contained sensor’s reflectivity versus wavelength characteristics. This lock-up table was further approximated by a cubic least-square approximation around the peak value to obtain continuous sensor wavelength characteristics. This reduced the noise by taking advantage of a larger number of available measured points near the peak wavelength. An example of such practically measured and interpolated sensor’s wavelength characteristics, obtained by the scanning of a typical FBG sensor, is shown in Fig. 4
. Finally, the wavelength of the peak value in the interpolated sensor’s wavelength characteristics was searched out and converted into the measured parameter such as, for example, strain or temperature.
Fig. 4 Example of FBG sensor’s reflectivity versus wavelength obtained by the current sweep and processing of raw data recorded by all four detectors
6. Experimental results
As a demonstration of a practical performance, we applied the proposed signal interrogation system to temperature sensing using a commercially-available (Welltech Inc.) FBG with peak wavelength near the locking point of DWDM laser module. The experimental FBG sensor had a full width at half maximum (FWHM) of around 0.65 nm, and peak reflectivity corresponding to approximately 4%. The FBG sensor was placed into the temperature calibrator together with reference Pt100 temperature probe. The FBG and reference Pt100 sensor’s temperature was then simultaneously increased by 0.1, 0.2, 0.5 and 1.1 °C for several minutes, as presented in Fig. 5
Fig. 5 Demonstration of FBG temperature sensor interrogation; the temperature was increased and reduced back to the initial temperature for 0.1, 0.2, 0.5, and 1.1°C
demonstrates a system resolution of about 0.1 °C, which corresponds to a wavelength resolution of around 1 pm. This measurement was obtained at 1 Hz bandwidth (about 30 raw measurements were averaged). Further reduction of bandwidth would additionally increase the resolution.
Another example of the interrogation system’s performance is shown in Figs. 6
. An FBG was bonded to a metal plate and exposed to the cyclical strain variation. Figure 6
shows the interrogation system response when the metal plate was cyclically strained for 1 με whilst utilizing an output filter with 1 Hz corner frequency.
Fig. 6 Change of peak wavelength when FBG was cyclically strained for 1 με (output filter corner frequency was set at 1 Hz).
Fig. 7 Dynamic performance of proposed integration system: A steel plate containing interrogated FBG and reference electrical strain gauge was exposed to nearly harmonic excitation bursts (the excitation signal within burst had frequency of 1 Hz and 5 Hz).
The strain induced changes in the measured wavelength’s peak position were above the noise level and corresponded to 1 pm, which is consistent with a typical FBG strain sensitivity of 1pm/με at 1550 nm.
shows the interrogation system’s response at full 30 Hz sample rate (no output filter was utilized). A steel plate containing interrogated FBG was dynamically excited by short duration nearly harmonic burst. In the upper part of the figure, the applied strain peak-to-peak amplitude was 75 με and the frequency of (nearly) harmonic excitation corresponded to 1 Hz. The lower figure shows similar burst excitation, but with 5 Hz excitation frequency and peak-to-peak amplitude of about 120 με.
The resolution in this case can be estimated to be 5 pm. Full sampling rate can be, however, only achieved for dynamic changes with limited amplitude. If the dynamic changes of the wavelength’s peak position exceed the wavelength range that can be covered by diode drive current sweep (approximately 300 pm in our case), temperature adjustment of the laser diode wavelength must be used for sensor interrogation, which reduces the system’s bandwidth to approximately 0.25 Hz (full 2 nm wavelength range temperature scan cycle can be completed with built in TEC in approximately 4 s).
In the last experimental example we interrogated 1 mm long AFFP interferometer, used as a temperature sensor (the sensor was build at the fiber tip using etch-and-splice technique described in [8
E. Cibula and D. Donlagic, “Low-loss semi-reflective in-fiber mirrors,” Opt. Express
18(11), 12017–12026 (2010). [CrossRef]
]). The peak wavelength shift for such AFFP was similar to FBG and corresponded to 11 pm/°C. Figure 8
shows result of sensor interrogation when we exposed the sensor to larger temperature changes. First temperature was cyclically changed for 10, than for 20 and at the end for 40 °C. Pt100 temperature probe was again used as temperature reference. None of the experiments showed any measurable hysteresis during the performed measurements.
Fig. 8 Peak wavelength during high temperature changes on AFFP
Practically demonstrated slow-rate temperature induced wavelength scanning over a 2 nm range and fast-rate current induced scanning over a 0.3 nm span does not represent an extensive wavelength span, however it allows for practical interrogation of at least two important groups of fiber sensors: FBGs, and properly designed AFFP sensors.
FBGs have a typical strain sensitivity of 1 pm/με@1550nm, and a temperature sensitivity of around 11 pm/°C@1550nm. Most of the typical industrial applications require strain measurements up to between 1000 and 2000 με, which can be covered by the proposed system. It should also be stressed that we achieved a 2 nm wavelength range within a particular single channel (100 GHz) DWDM module. Selection of other, particularly multichannel modules, can easily and significantly expand this wavelength range. Readily available multichannel DWDM laser modules (based on a very similar design to that of the used module) can cover wavelength ranges beyond 3.5 nm (4 channels on 100 GHz ITU grid).
Similar results as with FBGs can also be achieved with AFFPs. AFFP sensors are usually produced by the creation of semi-refractive mirror(s) within the fiber as described, for example, in [8
E. Cibula and D. Donlagic, “Low-loss semi-reflective in-fiber mirrors,” Opt. Express
18(11), 12017–12026 (2010). [CrossRef]
C. E. Lee and H. F. Taylor, “Interferometric Optical Fibre Sensors Using Internal Mirrors,” Electron. Lett.
24(4), 193–194 (1988). [CrossRef]
].The free spectral range (FSR) of such AFFP sensors should be adjusted to generate (at least) one well-defined peak within the interrogator scanning range. At 1550 nm a 1 mm long AFFP yields, for example, FSR of 0.8 nm and can thus be straightforwardly interrogated by the proposed system as already demonstrated in Fig. 7
The presented concept allows for the design of a compact and cost-effective sensor interrogation system. The current total cost for all components utilized to build a prototype was below $600 USD (this estimate is based on low volume - sample prices). The majority of this cost estimate belongs to the cost of the laser module. Since such laser modules are produced today in high volumes and will likely be produced at even in higher volumes in the future due to the increasing demand for telecom industry, we can expect that a realistic component cost of the proposed interrogator would actually be significantly below $600 USD.
This paper presented a spectrally-resolved fiber optic sensor signal interrogation system that can interrogate FBGs, properly designed AFFP, and similar spectrally-resolved sensors that exhibit distinctive spectral signatures within a narrow wavelength band. The presented concept demonstrates the possibility of converting a standard telecom DWDM laser module into an effective optical sensor interrogator. Besides the DWDM laser module, the proposed sensor signal interrogation system requires only two additional standard telecom detectors, and a standard fiber coupler. The proposed signal interrogator concept can thus lead to compact, robust, and cost-efficient sensor interrogation devices. Furthermore, all the used photonics components comply with Telecordia standards, which allows for straightforward qualification of the proposed interrogator within the range of the demanding applications.
The experimentally built prototype, which was based on the Oclaro LC25-EW laser diode module, demonstrated an about 2 nm wide spectral scanning range with a spectral resolution of around 1 pm. The main limitation of the proposed system is in its relatively narrow wavelength scanning range that is mainly determined by the useful operating range of the wavelength locker and the tunability of DFB laser diode. Readily-available multichannel DWDM laser modules can, however, cover wavelength ranges beyond 3.5 nm. Further limitation of the proposed principle arises from its slow interrogation rate when changes in the sensor’s spectral characteristics exceed a wavelength range that can be covered by the current sweep of the DFB laser (approximately 0.3 nm), and thus requiring a low-rate thermal sweeping of the laser wavelength. The proposed combination of simultaneous current and temperature diode control can, however, significantly mitigate this limitation in a considerable number of practical measurement applications.
Whilst we demonstrated interrogation of a single sensor, the high optical power that emitted by a typical DWDM laser module, can easily be divided among many sensors. Extension of the proposed concept to multichannel signal integrator is, therefore, straightforward and would require only one additional 2xN coupler, one standard detector, and one 1x2 coupler per extra channel. If etalon with wider spectral characteristic is chosen, a few sensors in a single 1x2 coupler arm could also be interrogated.