Fiber optic sensors have been the topic of intense research for over two decades. Fiber optic sensor technologies offer unique advantages over their electrical equivalents such as, for example, complete electrical passivity, robustness, small size, fully-dielectrically designed capability of performing quasi-distributed, multiplexed or fully-distributed sensing. However, the introduction of fiber sensing technology in industrial and other real applications have proved to be relatively slow and limited. The reasons for these limitations can be found in demanding sensor designs and/or complex photonic signal processing systems that are required for the extraction of measured parameters from optical sensor signals. Typically, the fiber optic sensor signal processing relies on spectrally-resolved techniques, white light interferometers, non-linear scattering, or similar complex photonic systems. These systems need to compete with other traditionally available solutions in terms of cost and performance. The development of cost effective and efficient concepts is, therefore, essential for increased introduction of fiber sensing technology into practical applications.
Recent developments in optical telecommunications, particular development of fiber to the home systems (FTTH), local area networks (LAN), and storage area networks (SAN), have resulted in widespread use of fiber-coupled components, such as simple Fabry-Perot (FP) laser diodes, and PIN diodes. The massive use of these components has significantly reduced their cost and allowed for their extensive use in cost sensitive systems. FTTH/LAN/SAN components could also provide opportunities for building cost-effective fiber sensor systems. If multiplexed or quasi-distributed configurations could be achieved, the cost per sensor could be further reduced to a level comparable to current electrical sensor equivalents.
Fiber optic sensing concepts that can take direct advantage of FTTH/LAN/SAN components include methods involving direct-time-of-flight measurements or methods based on the phase tracing of the radiofrequency modulated optical carrier. The radiofrequency modulated optical carrier approach has been known for many years and has been studied in several references [1
W. J. Rowe, E. O. Rausch, and P. D. Dean, “Embedded optical fiber strain sensor for composite structure applications,” Proc. SPIE
718, 266–273 (1986).
J. Plucinski, P. Wierzba, and B. B. Kosmowski, “Time-of-flight optic sensors for strain and temperature measurement,” Proc. SPIE
5952, 379–382 (2005).
]. Besides limited extension of the radiofrequency modulated approach to multisensory systems by application of optical switches [7
G. Thursby, F. Dong, B. Culshaw, G. Massaro, B. Glisic, and D. Inaudi, “An improved fibre optic strain sensor for gas tank monitoring with rf subcarrier phase and I&Q demodulation technique,” Proc. SPIE
5758, 381–389 (2005).
], it has been mainly used only in a single sensor configuration. Approaches similar to radiofrequency modulation were also reported in [8
V. Lyöri, A.Kilpelä, G. Duan, A Mäntyniemi, and J. Kostamovaara, “Pulsed time-of-flight radar for fiber-optic strain sensing,” Rev. Sci. Instrum.
78, 024705-1–024705-8 (2007).
] that involved the direct time-of-flight measurements in the fiber. While the latter approach could be applied to quasi-distributed sensing, it relies on more complex optoelectronic components, custom design integrated circuits for signal processing, and provides limited measurement resolution.
This paper presents the design of a sensor system that takes full advantage of low cost FTTH/LAN/SAN components and can measure optical path-length variations (OPV) in multiple segments, along the sensing fiber. The proposed system is based on the application of a radiofrequency modulated optical carrier signal that is launched into a sensing singlemode fiber containing multiple mirrors. The OPV is then obtained by comparing the phases of the reflected signals using a local oscillator. Simple and cost-effective fiber mirrors were also developed in order to allow the application of relatively low power FP lasers within quasi-distributed system. The proposed system allows for the quasi-distributive sensing of OPV caused by either strain or temperature change of the sensing optical fiber. The proposed quasi-distributed approach further reduces the cost per installed sensor and, thus, makes the proposed system suitable for cost sensitive applications.
2. Basic concept of the quasi distributed sensor based on radio-frequency mixing of a modulated optical carrier
The basic concept of this sensor system is presented in Fig. 1
. The system is composed of sensing optical fiber that is divided into sub-segments, separated by low-loss semi reflective fiber mirrors and an opto-electronic signal processor that determines OPV between individual fiber mirrors. When using a standard 50:50 coupler, two arrays of sensor segments can be simultaneously addressed by a single signal processor.
Fig. 1. Basic concept of quasi-distributed radiofrequency optical sensor system.
The opto-electronic signal processor consist of an FP laser diode (LD), a PIN diode with a transimpedance amplifier (TIA), a limiting amplifier (LIA), a laser diode driver (LDD), an oscillator (OSC), a programmable timing unit (PTG), AND gates, a XOR gate, a low-pass filter (LPF), a low-frequency differential amplifier, an A/D converter, and a microcontroller. Data from the microcontroller is transmitted to the PC for real-time observation and/or data storage.
The timing diagram of the proposed system is shown in Fig. 2
. The oscillator OSC generates frequency fOSC
. The PTG periodically generates pulses over a duration T0
and with a repetition rate TP. These pulses and the clock oscillator are supplied to the AND gate to form bursts of pulses having frequency of fOSC
, and duration T0
. These bursts of pulses are used to drive the FP laser diode. Generated optical bursts are launched into the sensing fiber that is composed of multiple sensing segments of length L0
. Launched optical bursts are partially reflected from individual mirrors and returned back to the detector. After optical reception and amplification, the received bursts are brought to the second AND gate. The function of the second AND gate is to extract those bursts reflected by a specific (observed) fiber mirror. For this purpose, the timing unit generates another pulse with a duration of TSE
(that corresponds to T0
or less) that is delayed for time Td
relative to the first outgoing pulse (pulse used for laser diode triggering). Td
is set according to the distance d
between the signal processing unit and the position of the observed mirror:
where c is the speed of light in vacuum and n is the group index of the fundamental mode.
Fig. 2. Timing diagram of quasi-distributed fiber optic radiofrequency optical sensor system.
The phases of the extracted back-reflected burst and oscillator are compared by an XOR gate. After removing RF components from the signal at the XOR output (e.g. after low-pass filtering), the voltage becomes proportional to the distance between the signal processing unit and the observed mirror. This dependence is periodic and corresponds to one half of oscillator period Λ (each optical burst transverses the fiber in forward and backward directions). The period of static transfer characteristic in relation to OPV can then be expressed as:
Since small path-length changes are usually of interest, it is straightforward to adjust oscillator frequency fOSC
, or simply the sensor length, to obtain the desired unambiguous measurement range. The system thus allows for measuring/tracking of any small variations in optical path length between the signal processing unit and the arbitrary mirror in the system (by programming the timing unit, the delay of the extraction pulse relative to the outgoing burst can be adjusted and thus the position of arbitrary mirror can be observed/addressed). The measured distances (converted into voltages) between two neighboring mirrors are then subtracted to obtain OPV for individual sensor segments.
3. Practical implementation of the proposed system
3.1 Optical sensor array
The optical coupler in Fig. 1
has two ports. Both ports can be used to connect segmented sensing fibers (in this case one of the ports should contain a delay line that allows for unambiguous multiplexing of all sensor segments). In this way, we successfully addressed 20 sensor segments in total. However for reasons of better presentation, further analysis and presented results only apply to the system that contains only one sensor arm (composed of 10 mirrors forming nine sensors). The creation of in-fiber mirrors that define individual sensing segments proved to be an essential step for successfully designing of the proposed system. In order to apply standard low-cost telecommunication components to the system, the reflectivities of individual mirrors should be considerable, while insertion losses must be low. A typical sensitivity of telecom 1.25 Gbit/s PIN-TIA is -24 dBm at the bit error rate (BER) 10-12
. If standard 3 mW (+5 dBm) source is used, a total loss of about 29 dB can be tolerated by the transmitter receiver-pair. Since the coupler adds 6 dB to the total loss in the system, back reflection and other excess loss in the system should remain below 23 dB. If we intend to interrogate at least 10 mirrors (9 sensors) in one coupler arm with a small percent reflectivity, this implies use of mirrors with insertion loss below 1 dB. For example, if we use mirrors with 4% (-14 dB) reflectivity, the total remaining loss in the system must not exceed 9 dB or less than 0.45 dB per mirror (bursts pass each mirror twice). Furthermore, the FP laser can emit a relatively broad spectrum that can change with temperature. Low-loss, high reflectivity, broadband mirrors are, therefore, needed for the realization of this system. However the fiber mirror insertion loss requirements can be partially relaxed since the signal-to-noise condition required to achieve BER of 10-12
does not need to be fulfilled in this particular case (output voltage is average of many bursts).
We investigated several options to create low-loss semi-reflective fiber mirrors. The use of standard polished connectors provided a repeatable and potentially-low-cost solution that can satisfy all the above requirements. The ferrule of an ordinarily prepared and polished connector was immersed in temperature-stabilized hydrofluoric acid (HF) for a predetermined time. During etching, the connector was perpendicular to the HF acid’s surface level, while the vessel with HF was being vibrated. The vibration of HF produced more repeatable results, probably due to the better depletion of etching by-products that reduces the local concentration of HF. The etching rate of the doped silica in the fiber core is higher than pure silica cladding [9
E. Cibula and D. Donlagić, “Miniature fiber-optic pressure sensor with a polymer diaphragm,” Appl. Opt.
44(14), 2736–2744 (2005).
]. The difference in the etching rates results in a cavity formation at the tip of the ferrule/fiber. When such a connector is mated with the non-etched connector a very short Fabry Perot cavity is formed. By measuring the etching time, good control over the cavity’s length and thereby reflectivity, can be obtained. Figure 3
shows typical reflectivity and transmission versus etching time in 40% HF at 23 °C for an FC connector mated with a non-etched connector at 1310 nm. The graph was obtained by progressive etching of the connector (the connector was etched, mated, and measured many times to obtain Fig. 3
Fig. 3. Reflectivity and transmission versus etching time for mirrors made out of polished connector (λ=1310 nm)
Low loss and broad-band reflective spectrum of the mirror can be realized since the FP cavity can be very short (below 1–2 µm). Maximal reflectivity within a range of over 8% was achieved in a repeatable way. The practical system used in the experimental setup contained, in total, 10 fiber connector mirrors in a single coupler arm. The last (10th) mirror in the arm was chemically coated with silver to maximize reflectivity. SMF-28 fiber was used when building the sensor network and the distances between individual mirrors were approximately 10 m. 7 m long sections of fibers between mirrors were fixed (glued) onto 60 cm long aluminum bars (Fig. 4
) to allow for testing the system for strain sensitivity. The cross-section of the aluminum bars was 15 mm×3 mm.
Fig. 4. Setup for determinating sensor resolution
More distant connectors in the arm were prepared in such a way as to have higher reflectivity compensating for the higher insertion losses in the network, while the mirrors closer to the FP diode were optimized for lower reflectivity, and lower insertion losses. The increase in mirror reflectivity towards the end of the network equalized the pulse amplitudes. Figure 3 was used as a guideline to determine those etching times that resulted in target mirror reflectivity/insertion losses. Figure 5
shows typical examples of transmission, reflectivity, and collateral losses for 9 etched connector mirrors used in one of the coupler arms. The collateral loss PCL
is defined as:
is incident optical power, PT
is transmitted optical power and PR
is reflected optical power.
The average reflectivity of the mirrors was 4.3%, while the average total insertion losses correspond to -0.36 dB, which is within the required power budget range for a 10 mirror array, as estimated at the beginning of this section. The collateral loss is low (not exceeding 5.5% or -0.24 dB in the worst case example), which makes the proposed fiber connector mirrors suitable for various fiber applications unrelated to the concept described in this paper.
Fig. 5. Results of 9 etched connector mirrors used in one coupler arm (a) transmission (b) reflectivity (c) relative collateral losses
The multiple double back reflections caused by linear sequence of mirrors did not have any detectable effect on the system operation. This was experimentally verified by severe bending of the individual sensing fibers, while observing voltages obtained at sensors prior the bend segment. This is actually expected, since the double back reflection that could cause signal degradation at the receiver, experiences at least three reflections from relatively low reflectivity fiber mirrors. If we assume average reflectivity of 4% per mirror, the total double back reflected fraction of forward propagating wave would be 0.04^3=0.000064 (or-42 dB), which is almost two orders of magnitude below receiver sensitivity.
The reflected optical power versus time after launching 90 ns burst into the described sensor array is shown in Fig. 6
(last sensor was in part chemically silvered and thus had higher reflectivity).
Fig. 6. Reflected optical signal from 10 mirrors as measured by a digital communication analyzer HP83480A.
Special care was also taken to provide initial distances among fiber mirrors that are multiples of period Λ. This ensured that all bursts returned from mirrors in the network experienced the same initial phase difference relative to the oscillator and, thus, provided the same initial position in the sensor system transfer function. The sensor operating point must be set in the linear range of static characteristic and shall remain there over the entire practical measurement range. The precision of the initial sensor segment length adjustment is thus not critical, and typical corresponds to about 10–20% of the ΛPFE
, or about 1 cm.
Finally, the coupler could also be replaced by a circulator, which would reduce losses in the system and would thus allow for the addressing of larger, single arm sensor array (instead of two-arms sensor array based on a coupler). While options on coupler and circulator are comparable in terms of optical power budgets and, thereby, the number of practically addressable sensors, the current cost of the circulator is comparable to the cost of the present entire sensor system.
3.2 Opto-electronics signal processor
During the design of the experimental system we relied on common and cost-effective standard opto-electronic and electronic components. Components that could significantly increase the cost of the final system, like for example temperature stabilized oscillator or avalanche diode, were deliberately avoided even though improved performance of the system could be achieved by employment of such components.
We used a standard telecom 1310 nm 3 mW fiber pigtailed FP laser (FL-3340) produced by Optoway Inc. as optical source. Optical detection was performed by standard telecom 1.25 Gbit/s PIN-TIA (Optoway Inc. PT-7330). Standard telecom integrated circuits produced by Maxim Inc. were used to drive the laser diode (MAX 3869 was used as LDD) and amplify/convert received signals (MAX 3768 was used as LIA). The 1.25 GHz oscillator produced by FOX Electronics (FXO-PC536R-1250.00) was used as oscillator. The high-speed part of the circuit was based on differential low-voltage positive emitter coupled logic (LVPECL). Temperature compensated AND gates (On-Semiconductor Inc. MC100EP05) and XOR gates (On-Semiconductor Inc. MC100EP08) with differential inputs and outputs were employed in the circuit design. Differential LVPECL design was used to help reduce noise, improve the rise-times of the phase comparing gates and, thereby, provide better linearity of the sensory system, and to double useful output voltages. Both differential XOR outputs were filtered by passive 2nd order RC filters and then directly subtracted by an instrumentation amplifier. The used RF filter had cutoff frequency at 15 kHz. This frequency can be set arbitrarily according to the desired system bandwidth, and resolution of the system. Another low-frequency signal conditioning amplifier was added to remove any remaining DC offset and to amplify signals to a level suitable for AD conversion. An Analog Device Inc. programmable delay generator ADF4026 was used to generate timing signals for AND gates control. The programming of the ADF4026, data acquisition, communication, and other control functions were performed by a microcontroller.
3.3 Measurement algorithm and sensor sensitivity
The PTG (ADF4026) was programmed to generate at its laser-triggering output pulses with duration T0
=90 ns and with repetition rate of TP
=1.2 µs (TP
was increased to 2.4 µs when an array with 20 sensors was integrated). The sensor burst extraction duration TSE
was set at 75 ns to avoid possible transients at the detector, upon arrival of the sensor burst. The timed delay Td
between the laser diode triggering pulse and sensor burst extraction pulse was continuously re-programmed by the microcontroller to cyclically address all mirrors in the network. Once the specific time delay Td
was programmed into the PTG, the voltage at the filtered XOR output was left to settle for 300 µs before it was sampled, converted into digital value, and stored in the microcontroller’s memory or sent to the PC. The variations between individual sensor OPV’s were obtained by subtracting measured voltage values obtained from bursts reflected from neighboring mirrors. The entire PTG programming/reading/measurement cycle for a 10 mirrors network took between 9 ms and 9 s, depending on the degree of digital filtering used to further limit the measured bandwidth and improve sensor resolution. Alternatively, when only a specific sensor segment was of interest, the PTG was programmed to cyclically address only two neighboring mirrors representing the targeted sensor segment. This improves bandwidth and signal to noise ratio, but allows for simultaneous observation of only one specific sensor in the network. Since telecom LIAs use dynamic adjustment of the threshold (decision) levels, long pauses between pulses bursts can lead to distraction of LIA’s decision threshold adjustment circuit. When programming the PTG, care must be taken that the bursts coming to the receiver are not separated by long pauses. Essentially, when a burst from the most distant mirror is received, the new burst should be generated promptly to keep the detector in a constant state of active reception. Similarly, the use of telecommunication LIA requires reasonable equalization of optical power among individual bursts, which is achieved by proper mirror construction, as described earlier in the text. The OPV is determined directly from the voltage at filtered XOR output. When the launched pulses are shifted in time for one half of the oscillator period, the XOR output pulse width will shift from zero to the full width, and the averaged voltage at XOR output will change from a value corresponding to logical zero to a value corresponding to logical one. The relationship between the OPV and averaged (filtered) output voltage can then be described as:
where Λ is the oscillator period length in the fiber, fOSC
oscillator frequency, n group index of the fundamental mode, c speed of light in vacuum, D duty ratio (TSE
voltage of logical high level, UL0
voltage of logical low level, and ΔUL
is difference between both levels. The first multiplication by 2 is due to the periodic nature the transfer function (change in OPV for Λ/2 causes full swing in output voltage). The second multiplication factor 2 comes in our particular case from the use of differential XOR gates (both inverted and non-inverted outputs are filtered and subtracted by low-frequency differential amplifier, which doubles the output voltage).
To obtain correlation between physical fiber length elongation (PFE) and output voltage change, the OPV in Eq. (4)
should be doubled, since each optical burst transverses the fiber in forward and backward direction. Furthermore, Eq. (4)
should be multiplied by k
, a coefficient that correlates physical fiber length change with the OPV (around 0.71 for silica fiber [10
K. Abe, K. Yoshida, O. Daneshvar, and J. J. Carr, “Photo-Elastic Correction Factor for Fiber Strain Measurements in a Cable Under Tensile Load,” J. Lightwave Technol.
13(1), 1–5 (1995).
If we assume ΔUL
=750 mV (this voltage was measured in our system, typical differential output voltage range defined by LVPECL specification is between 525 mV and 900 mV), fOSC
=1.25 GHz, TSE
=0.056, n=1.45 and k=0.71 we obtain:
where S presents sensitivity of the sensor system. When the sensor is used as a strain measurement system, the relationship between the strain ε and voltage can be expressed as:
where L presents the active length of the fiber used as a strain sensor. Similarly, temperature sensitivity can be obtained from Eq. (4)
. If we neglect fiber elongation due to the temperature change and we only take into account fiber refractive index temperature dependence (ratio between both is about 1:50), the OPV can be approximated by OPV
, and Eq. (4)
In our practical experimental setup, using LVPECL logic with sensors having active length L
and with dn
for silica 10-5
fiber, we obtain from Eq. (7)
and Eq. (8)
Strain sensitivity: 10.1 µV/µε
Temperature sensitivity: 143 µV/K
It should, however, be noted that when the fiber is glued on the surface of measured material (for example aluminum) the fiber will follow the thermal expansion of such material. In this case the temperature effects due to fiber expansion can be significant and can be taken into account using Eq. (7)
4. Experimental results and discussion
4.1 Sensor static characteristics
shows measured voltages (obtained by cyclically addressing all mirrors in the network) as a function of the lead-in fiber’s elongation (fiber preceding the first mirror).
Fig. 7. Measured sensors static characteristics.
The static characteristic has a periodic, almost triangular shape, with rounded tops near minimum and maximum values. The unambiguous PFE
measurement range corresponds to one half period of the transfer characteristics and was measured to be ΛPFE
. The ΛPFE
is related to ΛOPV
by a photo-elastic factor of 0.71, as already explained in section 3.3. The measured ΛPFE
*0.71 ratio corresponds to 8.24 cm while the Eq. (2)
predicts this value to be 8.27 cm. The absolute value of the static characteristic’s slope corresponds to approximately 1.4 µV/µm. This is also in reasonable agreement with the calculated sensitivity obtained from Eq. (6)
, which predicts 1.45 µV/µm. The rounded shape of static transfer characteristics near its extremes is probably caused by the limited rise-times of logical gate used to compare reflected optical signals with the oscillator. Faster gates and LIA or higher power laser diode would probably further improve the rise times in the system and further improve linearity near the characteristics’ extremes. An almost linear response was, however, obtained within an approximately 75% range of the transfer characteristics.
Fig. 8. Each sensor in the network was displaced for about 70 µm: a) raw voltages as measured by observation of individual mirrors; b) subtracted neighboring voltages representing displacements of individual sensor segments.
The PFE for individual sensors was obtained by subtracting those voltages obtained by observing reflections from neighboring mirrors. Figure 8
(a) demonstrates those measured voltages obtained by cyclically addressing all 10 mirrors while each sensor segment in the network was, one after another, strained and released for about 70 µm (bars in Fig. 4
were consecutively loaded with about 3 kg weights). Figure 8
(b) show the response of an individual sensor within the network after subtraction of those voltages obtained from the neighboring mirrors. No significant crosstalk among sensors can be observed.
4.2 The drift
The drift of the output voltages (measurement results) presents the main shortcoming of the proposed system, as it directly limits achievable measurement repeatability and, thereby, achievable absolute accuracy of the system. However, a more-detailed investigation of the experimental system, showed that the drift originates from several sources and that output signal fluctuations can be clearly attributed to temperature changes. We, therefore, placed the experimental electronics signal processor, and the sensor network into two separate temperature-controlled chambers. We stabilized the temperature of the sensor network to within ±0.5 °C, while we varied the temperature of the electronics signal processor. The results are shown in Fig. 9
and Fig. 10
. All voltages representing individual mirrors’ positions experienced similar temperature drift. When voltage signals obtained from neighboring mirrors were subtracted to calculate the sensor OPV, a significant part of the thermal-related drift was also canceled-out, as shown in Fig. 10
. The maximum residual drift after subtraction of neighboring mirror signals was equivalent to 20 µm/°C (expressed in PFE), which is equivalent to strain measurement drift of 3 µm/m/°C in our particular case (7 m of sensing fiber).
The cooling of individual components (using liquefied gas) within the circuit, showed that the majority of the drift comes from the temperature dependence of high speed (LVPECL) electronic components. This dependence probably originates from the temperature dependence of signal propagation delays within these components. Since all signals from different mirrors were routed through the same electronics paths, the propagation delay dependence was relatively effectively canceled-out by subtraction of neighboring mirrors’ voltages. The residual drift, which was not canceled-out by subtraction of neighboring mirrors’ voltages, can be further attributed to oscillator frequency drift, and other unidentified causes. The change in oscillator frequency generating phase difference among two neighboring mirrors that can be expressed as:
is the group index of the fundamental mode, c speed of light in vacuum, L0
sensor length (length of fiber between two mirrors). We measured oscillator frequency change separately, which proved to be about 2000 Hz for a temperature change of 10 °C. To convert the phase change Δφ
, Eq. (9)
should be multiplied by factor 2ΔUL
D/π (using the same reasoning as in the formulation of Eq. (4)
). 2000 Hz oscillator frequency change thus corresponds to a PFE change of about 27 µm/°C, which is in reasonable agreement with the measured temperature drift of 20 µm/°C.
Finally, a system drift over a 13 hour period was observed when the sensors and signal processor were both temperature stabilized within ± 0.5 °C at temperature 30 °C. The drifts of voltages (and equivalent PFE) obtained from individual mirrors, are shown in Fig. 11
, while Fig. 12
presents subtracted voltages representing individual sensors. When this drift is expressed in PFE, it corresponds to a maximum of 48 µm/13h or 7 µm/m/13h when it is expressed in equivalent strain drift for particular sensor configuration (assuming 7 m of sensing fiber).
Fig. 9. Output voltage change (a) for individual mirrors (10 mirrors in total) when the temperature of signal processor is increased for 10 °C (b).
Fig. 10. Heating of the optoelectronic signal processor for 10 °C - voltage differences obtained by addressing of neighboring mirrors (corresponding to individual sensors drift in the network)
Fig. 11. Voltages obtained from an individual mirror for long term testing
Fig. 12. Subtracted voltages of two neighboring mirrors for-long term drift
4.3 Sensitivity and bandwidth
Sensitivity was determined by the vertically loading of aluminum bars (containing 7 m of fiber) with calibrated weights to induce well-defined strain and, thereby, known PFE, as shown in Fig. 4
(a) demonstrates the typical measured response of an arbitrary sensor in the network when the measurement bandwidth was 15 kHz and the sensing fiber was elongated for 120 µm (corresponding to straining the bar for 17 µm/m). Figure 13
(b) shows a similar situation but when the measurement bandwidth was limited to 1 kHz and the sensor fiber was elongated for 35 µm (corresponding to straining the bar for 5 µm/m). Figure 13
(c) show the situation when measurement bandwidth was limited to 1 Hz and sensor fiber was elongated for 5.5 µm (corresponding to straining the bar for 0.8 µm/m). This additional bandwidth limitation was performed by digital filtering carried out by a microcontroller. While the absolute PFE measurement resolution is typically within the 10–100 µm range, a long-section of measurement fiber can easily be implemented to achieve strain resolution within the range or below 1 µm/m, as also shown by previous examples.
Fig. 13. Loading and unloading of a single sensor in the system by weights: a) Fiber is elongated for 120 µm, bandwidth of the system is 15 kHz b.) Fiber is elongated for 35 µm, bandwidth is 500 Hz c.) Fiber is elongated for 5.5 µm, bandwidth is 1 Hz.
A cost-effective quasi-distributed fiber optic system was presented, for the measurement of optical path length variation. The individual sensor segments were created by the introduction of semi reflective fiber mirrors down the sensing fiber. The system can be directly configured as a quasi-distributed long gauge strain or temperature sensor.
Simple and cost-effective fiber mirrors with adjustable reflectivity were created by controlled, short duration etching of standard FC/PC connectors. Mirrors with reflectivity within the range 0.1% to 8% and with insertion losses between 0.2 dB and 0.7 dB were produced in a repeatable way. The proposed optical part of the sensor array is, therefore, simple to implement and low-cost. The optical path-length variations between mirrors were measured by launching short-duration radio-frequency modulated optical bursts into sensing fiber. Phase differences among individual reflected pulses in the bursts were measured to determine the optical path-length variation. Eighteen sensing fiber segments (organized in two arrays containing 2×10 mirrors) were successfully addressed with single signal processor, while relying on a standard telecommunication PIN diode receiver, and Fabry-Perot laser diode.
The stability and achievable accuracy of the system is mainly determined by the stability of the optoelectronic signal processor. One of the main sources of this instability relates to the temperature drifts of the electronic components. These temperature drifts have several origins and the majority of them were successfully removed from the measurement signals by subtracting signals obtained from neighboring mirrors. This subtraction is inherent in the proposed sensor design, as it is necessary for obtaining the OPV of individual sensor segments. Even when using low-cost components, measurement instability caused by temperature change of the signal processor was lower than 20 µm/°C. The long-term instability of the system in the controlled temperature environment proved to be below 48 µm/13 hours. An optical path-length measurement resolution of about 4 µm (corresponding to mechanical elongation of the fiber for 5.5 µm) was demonstrated in practice at bandwidth of 1 Hz, 54 µm (mechanical elongation 75 µm) at a bandwidth of 1 kHz and 85 µm at a bandwidth of 15 kHz (mechanical elongation of the fiber for 120 µm). The system can also be easily operated in kilohertz or even within higher-measurement bandwidth range and is applicable to static and particularly quasi-static, quasi-distributed strain measurements. The ability to address at least 18 sensing segments using single signal processor allows for implementation of various strain sensing temperature compensation schemes that can effectively remove drifts caused by measurand thermal expansion and sensing fiber refractive index change. Such schemes might include for example sensor pairs, where one senor is exposed to the measured strain, while the other is used as a temperature reference or when sensor pair is exposed to oppositely signed strain (assuming that such configuration is available) while mounted in the way that provides the same temperature effect on both sensors. Since the long sections of sensing fiber can be employed to construct individual sensors within the sensor array, a sub-microstrain quasi-static resolution can also be achieved in practice.