Fiber sensor analyzer
TECHNICAL FIELD
The invention relates to the field of interrogating fiber optic sensors, such as Bragg grating sensors or fiber lasers. More specifically, it relates to measuring the optical spectrum output from several fiber optic sensors in a way that compensates for the delays in sampling time introduced by a scanned wavelength selection device such as a narrow bandpass optical filter.
BACKGROUND OF THE INVENTION
In many fiber sensor applications, such as those based on fiber Bragg gratings, excitation results in changes in the spectrum emitted, reflected or transmitted by the sensor. The interrogation of such sensors is usually carried out by measuring the optical intensity as a function of wavelength, and detecting features in this spectrum corresponding to the emission from the sensor by use of one of several analog or digital detection methods available to those trained in the field. To fit as many sensors as possible in a system, wavelength multiplexing is commonly used, distributing the sensor wavelengths over an interval of the optical spectrum.
It is common knowledge that one method of interrogating an optical spectrum is by scanning a tunable narrow-bandpass filter across the spectrum, and record the transmitted intensity as a function of the wavelength, as is described in US patent 5,818,585. A well known equivalent alternative is to use a scanned narrow-band source to interrogate the spectrum. One such system is described in US patent 5,401,956.
As the scanning wavelength selection device only transmit a narrow wavelength range at each point in time, there will be a delay in the time at which the response from sensors centered at different wavelengths are measured, due to the time used by the scanning device to travel between these wavelengths [Figure 2].
This effect may introduce severe sampling errors in applications where data from several sensors, assumed to be sampled at exactly the same time, are used in subsequent calculations. An example of one such application is advanced data analysis for structural health monitoring, where the combined signal from several Bragg grating
strain sensors placed on different locations on the structure are used for the purpose of calculating the total loading of the structure [A.E. Jensen et al. "Measurement of Global Loads on a Full-Scale SES Vessel Using Networks of Fiber Optic Sensors", J. Ship Research, Vol. 45, No. 3, Sept. 2001, pp.204-214]. The work reported on in this article used strain data without sampling phase corrections, which may introduce errors as demonstrated below.
Consider the following example. Assume that an identical sine signals at frequency / , x = sin(<2#), ω = 2πf , is observed on two fiber optic sensors. Also assume that the sensors emit wavelengths close to the minimum and maximum wavelength scanned. In addition to the signal x, the measurement will also reflect a phase shift introduced by the time delay between the sampling events. This time delay is normally less or equal to half the scan period, Δf ≤ 0.57; = 0.5//, assuming a that the wavelength selection device is scanned using a symmetric periodic function, such as a triangular dependence on wavelength as a function of time, and that both sensors are measured in the same ramp direction. Thus, the maximum measured difference between the signals may be approximated by
Δx » — Δt = ωcosiωt) At . dt '
The error is maximal at times where cos(<sf) = ±1 , corresponding to x = sin(β#) = 0 , thus f
Ax ≤ π— ,
J s where the phase error Ax is at its maximum for sampling time delays Δt close to 1/(2/. ) at times where the signal x is close to 0.
If the sampling phase error is desired be limited to 1% at all times, the sampling frequency needs to be fs = # 0.01 / = 314 / . Even a modest 10% error margin requires a sampling frequency 31 times the signal frequency.
The maximal scanning speed of the wavelength selection device, as well as limitations in throughput of some digital signal processing techniques for interrogating the resulting spectrum, limit the sampling frequency fs of interrogation systems. In
systems available today, typical sampling frequencies lie in the range from a few Hertz to a few hundred Hertz. Taking into account the sampling rates necessary to limit phase errors, the maximum signal frequency,/ that is available for these signal analysis techniques, may be well below 1 Hz in several of these systems, which is too little for several applications.
A secondary effect, also introducing measurement errors, was not taken into account in the example above. Since a signal is measured by changes in the sensor wavelength, this also influences the time, relative to the onset of the scan function, at which the sensor is sampled. In a case where the sensor is sampled on an increasing flank of the scan function, and that the signal moves the sensor towards increasing wavelengths, the time between the samples is longer than the period of the scan function. Correspondingly, a signal resulting in the sensor wavelength moving in the opposite direction of the scan function flank results in a shorter time between the samples. Thus, the signal is sampled at irregular intervals (irregular sampling grid). If not compensated for, this effect may introduce detectable false higher harmonic frequencies in the measurement.
Thus, there is a need for an interrogation system that corrects for these errors introduced by the scanned wavelength selection device.
SUMMARY OF THE INVENTION
Accordingly, the invention is directed to a system and method for compensating for the phase errors when sampling multiple fiber optic sensors using a scanned wavelength selection device. The compensation is carried out using a timer that accurately determines the time at which the measurement on each sensor was made, or time of arrival at each sensor, and an interpolation technique for transporting the values for each sensor reading within a scan to a common time.
There are several interpolation techniques available to those versed in the field, once the accurate detection time has been determined. In principle, these methods involve one or several steps in a process of interpolating data sampled on an irregular sample grid to a regular sample grid common for all sensors. In simple and direct methods, this process may be carried out in a single interpolation process. More advanced techniques may require first interpolating the data sampled on an irregular grid to a regular grid centered at the times corresponding to the average sensor
wavelength, before applying digital signal processing techniques such as a Fourier- based band limited sample rate expander, a sample shifter, and a decimator, to estimate the sensor value at the regular sample grid common to all sensors. Fourier-based interpolation techniques that operate directly on samples on an irregular grid are also available.
To illustrate the benefits of the invention, a simple linear interpolation is applied to the above example, disregarding any effects of irregular sampling. In this case, the interpolation value to time t of the sine signal x sampled at time t + Δt, Δt = 0.57; = 0.5// is
x = — (sin(β?(t + Δt))+ sin(ω(t - Δt)))
Thus, the error is
Ax = shι(ωt)—x
= sin(fi#)[l - cos(<»Δt)]
The table below compares the approximate sampling rate / , as estimated in the example above, necessary to obtain a given maximum phase error
Thus, an apparent advantage of the technique is that to accurately sample a 10 Hz signal with 1 % maximum phase error, a sampling rate of 3.14 kHz is needed for an uncompensated system, which is much higher than the sampling rates available in most
such systems today, while a linearly phase compensated system will require a relatively modest 220 Hz sampling rate. Even better performance may be attained through more advanced methods for estimating the measurement value at time t, as outlined above. Furthermore, the accuracy of the instrument is best when the scan characteristics of the wavelength selection device is continuously measured and used in the calculation of the compensated sensor wavelength.
A further advantage of the invention is that by use of a signal from a master clock the timing devices on several such units running in parallel may be synchronized. This allows the coordinated measurement of phase compensated wavelengths for more sensors than what is possible by a single unit.
Furthermore, the methods introduced by the invention may also be applied to interrogation systems where an optical multiplexer is used to couple a plurality of fiber sensor channels to the same light source and/or detection system. In this case, the scan phase correction algorithm is used to estimate the measurement value at a common time also for measurements made during different sampling periods.
To achieve these and other advantages in accordance with the invention, as embodied and broadly described, a system according to this invention includes a scanned wavelength selection device, a scan function generator, a fiber optic distribution network to distribute the light to one or several independent optical channels containing fiber sensors or an optical multiplexer, a receiver for measuring the optical spectra returned from the fiber sensors on each channel, a peak detector on each independent channel for interpreting the spectra, a timing device for accurately storing the time at which each sensor peak was detected, a processor for determining the wavelengths and measured signal where the processor uses timing and other relevant information to calculate the sampling phase compensated sensor wavelength. In stead of determining the wavelength it may also be possible to determine equivalent information related to this, such as temperature and strain, directly without first determining the wavelength. More specifically the invention is characterized as described in the accompanying independent claim. The invention is described in more detail below with reference to the accompanying drawings, illustrating the invention by way of example.
BRIEF DESCRIPTION OF THE DRAWINGS
FIGURE 1 : An embodiment of the invention with N independent optical channels FIGURE 2 : An illustration of sampling phase correction.
FIGURE 3: An embodiment of the invention with an optical multiplexer switch to select between optical channels.
FIGURE 4: An embodiment where the sensors are Fiber Bragg Gratings illuminated by an external source or the sensors are Fiber Lasers. FIGURE 5: An embodiment where the sensors are Fiber Bragg Gratings illuminated by an external source or the sensors are Fiber Lasers, and the system selects between sensor channels by use of an optical multiplexer switch.
FIGURE 6: The coordination of several systems by having one system act as a Master unit synchronizing the clocks of one or several Slave systems. FIGURE 7: A variant of the embodiment in Figure 1 where the spectrum transmitted by Fiber Bragg Grating sensors is interrogated.
DETAILED DESCRIPTION OF THE INVENTION
The preferred embodiment of the invention is described below, with the help of the figures. An embodiment pertinent to the interrogation of Bragg grating fiber sensors is shown in Figure 1. The interrogation of other types of sensors may require changes in the embodiment, see examples in figures 3 to 5 and 7. The elements discussed are each given a number for clear identification. For brevity, elements with similar functions are given the same reference number in all figures. The description is of the preferred embodiments, although several other equivalent embodiments of the invention should be apparent to those versed in the field. A wavelength selection device is formed by the combination of a broad band source (1), which may be a superluminescent diode, an erbium-doped fiber amplifier or a wavelength tunable laser source of similar function, possibly in combination with a tunable narrow bandpass filter (4), preferably a Fabry-Perot filter. Preferably, an optical isolator (2) is inserted between the Fabry-Perot filter and the source, blocking the light returned from the filter.
The wavelength is scanned over a chosen range of wavelengths corresponding to the wavelengths of the sensors. Also, the change in wavelengths resulting from the
scanning represents a preferably periodic function of wavelengths along the time axis. The preferred function according to the invention is periodic and symmetric over one period of time, but other functions are also possible.
To reduce measurement noise introduced by source polarization, a depolarizing module (3) may be inserted in the light path. This depolarizer may be constructed by passing through one or a plurality of fiberoptic couplers, where light passing through a coupler is fed back into the same coupler by connecting the coupler arms in a loop. This depolarization device may be inserted at any location from the source to the fiberoptic distribution network (5). According to the preferred embodiment illustrated in figure 1 a timer (6) generates a signal that is used to generate a periodic ramp function (7). The ramp function controls the spectral position of the filter pass band, and thereby which narrow portion of the spectrum illuminates the sensors. The timer may be operated in one of two modes. The master mode, where the ramp function is controlled solely by the internal timer, or the slave mode, where the onset of new ramps are controlled by an external synchronization signal (8). A system operating in the master mode may operate alone, or, through the output of a synchronization signal (9), control the synchronous operation of one or several units operating in the slave mode [Figure 6]. A system in the slave mode will need an external synchronization signal to operate. Systems without this functionality operate in the master mode.
The output from the wavelength selection device is then distributed to one or several channels by means of an optical distribution network (5).This network is normally constructed using fiber optic couplers.
The optical input to a single channel is then passed from the "left-side" of an optical coupler (10) to one or more fiber optic sensors (11) connected to one or more legs of the "right-side" of the coupler. The light reflected from the sensors (11) are returned through the optical coupler (10) and the intensity is measured on a detector (12).
When the spectrum passing through the wavelength selection device matches the reflected spectrum of a Bragg grating, a peak in the intensity is observed at the detector (12). The position of the peak is detected by a peak detector (13), which may be made from analog electronic components or from a digital signal processor analyzing
the signal, e.g. where digitalization of the optical intensity spectrum and digital signal processing techniques are used to detect the sensor wavelengths
When a peak is detected, a signal to a memory buffer (14) initiates the storage of a data record. The record contains the time of detection, the channel number, and diagnostic data. The time of detection preferably identifies which scan number, in a series of repeated scans, the measurement was made. This feature is necessary when analyzing data from several units or in systems using an optical multiplexer, and simplifies subsequent signal processing in all cases. Preferably, the buffer is sufficiently large to hold records from several scans, reducing the chance of missing data points due to data transfer latency.
The data stored in the memory buffer is then read into a system for calculating the wavelength (15). The wavelength is found using the time stamp and knowledge of wavelength output from the wavelength selection device as a function of time. Preferably, the system includes a wavelength reference for absolute determination of the sensor wavelength and elimination of drift in the wavelength selection device. Such a reference may be made from one or several temperature stabilized and strain isolated fiber Bragg gratings, by gas cells or etalons, as is well known in the art. Preferably, the data is checked for errors at this stage, identifying missing or erroneous data points. The wavelength data and timing information is passed to a system for phase correction of the wavelength (16). Several advanced methods for applying such corrections are available to those versed in the field (see e.g. T Strohmer (1993), Efficient Methods for Digital Signal and Image Reconstruction from Nonuniform Samples, PhD thesis, University of Vienna). The choice of the compensation method depend on the digital signal processing power available in the system, the amount of data to be processed, and the required accuracy.
An implementation illustrating the approach is shown in Figure 2. The topmost graph (17) shows a time- varying signal x experienced identically by two fiber sensors at different wavelengths λa and λb. The middle graph (18) shows the wavelength selected by the wavelength selection device as a function of time, while the bottom graph (19) shows the detector intensity on scan up (solid lines) and scan down (dotted lines). The measured values without phase correction is shown as circles. A phase corrected
wavelength^ of measurement number n on sensor α may be conveniently found from a linear extrapolation as follows
where λ
a t
a (n)] is the wavelength emitted by the wavelength selection device at the time of sensor detection t
a(ri) , and t(ή) is the common time to which all measured wavelengths are interpolated. The phase corrected wavelength value is then used to find the measurement value x from the sensor response function, shown as a cross (X) in the graph (17). Note that these steps may also be carried out in reverse order, that is that the measurement value x may first be found from an uncompensated sensor wavelength, before phase compensation of the measurement value is carried out using methods similar to those described above.
Also note that a similar method may be applied for extrapolating the measured wavelengths to a common time for multiplexed systems [Figures 3 and 5]. Typically, a multiplexer (20) addresses a new independent optical channel for each scan in a scan sequence. This process is then repeated after all channels that have been measured in course of N repetitions of the scan. In this case, phase corrected measurements may be made by linear extrapolation to a common time for all N scan repetitions, conveniently using the same method as above. Variants of the embodiment of the invention using fiber lasers as sensors is pertinent to a fiber laser interrogation system as given in Figures 4 and 5. In this case, the sensors (21) emit a wavelength signal that is interrogated using a scanning filter (4) and techniques similar to those described above.
The embodiments described above may, with minor modifications, be used to interrogate the transmitted spectrum of the fiber sensors. The principle is shown in figure 7, where the intensity of the spectrum transmitted through the fiber sensors are monitored on the detectors (12). Correspondingly, changes are made in the peak detection circuitry to detect the intensity "dips" in the transmitted spectrum (24). An embodiment pertinent to very large measurement systems is shown in Figure 6. Here, several units are synchronized by having a unit designated as master (22) emit a synchronization signal that initiates a new repitition of the scan function by
one or more slave units (23). By this method, the clocks on the different units will operate in approximate synchrony, allowing phase correction by extrapolation to a common sample time for all sensors in the system.
According to a practical embodiment of the system it may be provided in a modular form in which the sensors and interrogation parts may be independent of each other. Thus fiber laser or Bragg sensors may be chosen depending only of the object to be measured and may be sensors disclosed in the objects beforehand, while the interrogations instruments may be coupled to the sensors afterwards in a permanent or time limited manner. Other variations of the invention may also be contemplated within the scope of the invention, being limited by the accompanying claims.