Automated multichannel broadband spectrum ...

Int. J. Reasoning-based Intelligent Systems, Vol. 1, Nos. 1/2/3, 2004

43

Automated multichannel broadband spectrum analysis of fiber-optic grating sensors Plamen Balzhiev* Department of Radio Communications and Video Technologies, Faculty of Telecommunications, Technical University – Sofia, Bulgaria E-mail: [email protected] *Corresponding author

Wojtek J. Bock Canada Research Chair in Photonics, Université du Québec en Outaouais, Québec, Canada E-mail: [email protected]

Tinko Eftimov Department of Experimental Physics, Faculty of Physics, Plovdiv University, Plovdiv, Bulgaria E-mail: [email protected]

Rumen Arnaudov Department of Radio Communications and Video Technologies, Faculty of Telecommunications, Technical University – Sofia, Bulgaria E-mail: [email protected] Abstract: In this paper we report methods for spectrum analysis and data processing algorithms. An automated multi-channel spectrum measurement system is introduced with controlled fiber-optic signal switching and spectrum analysis with linear CCD photodiode array, diffraction grating and precise stepper motor. The designed system demands advanced measurement and data processing techniques. The paper reports the implemented methods for automated multi-channel measurements, accuracy improvement, noise cancellation techniques and fiber-optic grating sensor measurements Keywords: long-period grating sensors, fiber-optic sensor interrogation and multi-channel spectrum measurement, spectrum interrogation, linear CCD spectrometer. Reference to this paper should be made as follows: Balzhiev, P., Bock, W., Eftimov, T., Arnaudov, R., (2013) ‘Automated multichannel broadband spectrum analysis of fiber-optic grating sensors’, Int. J. Reasoning-based Intelligent Systems, Vol…, No. …., pp…... Biographical notes: P. Balzhiev received his Ph.D. in system automation and precise measurements from Technical University – Sofia, Bulgaria in 2012. His current research focuses on intelligent sensors, embedded measurement systems and fiber-optic sensors. He is also interested in wireless networs, data acquisition and control systems. P. Balzhiev received his Ph.D. in system automation and precise measurements from Technical University – Sofia, Bulgaria in 2012. His current research focuses on intelligent sensors, embedded measurement systems and fiber-optic sensors. He is also interested in wireless networs, data acquisition and control systems. Wojtek J. Bock (M’85–SM’90–F’03) is a full professor of Electrical Engineering at the Université du Québec en Outaouais (UQO), Canada. Since 2003 he is a Canada Research in Photonics and the Director of the Photonics Research Center at UQO. His research interests include fiber optic sensors and devices, multisensor systems, and precise measurement systems of non-electric quantities.

Copyright © 2004 Inderscience Enterprises Ltd.

44

T-H. WENG AND B. CHAPMAN

T. Eftimov is currently with the Department of Experimental Physics, Plovdiv University, Plovdiv, Bulgaria. His interests include optical fibers, polarization phenomena, intermodal interference, fiber gratings, and fiber optic sensors. R. Arnaudov is a full professor in Department of Radio Communications and Video Technologies at Faculty of Telecommunications, Technical University - Sofia. His interests are measurement techniques, intelligent sensors, control and data acquisition systems, GPS navigation.

1

INTRODUCTION

The basic scheme of the multi-channel spectral measurement system is shown in Figure. 1. The radiation of

ASE

LPG4

System description

LPG3

2.1

Refl.4

LPG2

2 MULTI-CHANNEL FIBER-OPTIC SPECTRUM MEASUREMENT SYSTEM

Refl.1 LPG1

Over the past decade optical fiber-based sensors are gaining significant progress and popularity. Optical fiber gratings are of the type of Fiber Bragg gratings (FBGs) and Long Period gratings (LPGs) (Srimannarayana, 2008). Because of the large periodicity, LPGs are usually easier to fabricate in mass production in comparison with the FBGs. The LPGs have found applications in various devices like equalizers for erbium doped fiber amplifiers, band-rejection filters and sensors for strain, temperature and refractive index (Choi, 2006). Because of their great width, spectral multiplexing is limited; the absence of a reflected signal demands detection of center wavelength shifts in a noisy minimum; and since resonance coupling in LPGs is to a cladding mode, the fiber typically has to be stripped, which creates challenges for long-term reliability and packaging. On the other hand, the LPGs are sensitive to a number of physical quantities such as surrounding refractive index, hydrostatic pressure, bending and twisting. They therefore offer significant application opportunities. However, in order to reduce the effective price per sensor, simple and efficient multiplexing systems must be developed. Wavelength and time-division multiplexing are well advanced with FBG sensor networks (Murphy, 2001; Alves, 2003), but comparatively little has been reported on the multiplexing of LPGs (Guan, 2007). Also a precise stepper motor is introduced to extend measured spectrum range and resolution by rotating the diffraction grating. Precise stepping drives are presented in (Mikhov, 2005; Mikhov, 2008), where improvement in position accuracy and microstepping control is applied. In this paper we report on the further development of a previously proposed spectrally and spatially multiplexed sensor network using an InGaAs CCD photodiode array and opto-mechanic switches (Balzhiev, 2012). We also present results of the implemented methods on multi-channel sensor monitoring, accuracy improvement and noise cancellation techniques. Additially, an automated analysis of the monitored LPG spectra is introduced. It includes fiber-optic sensor identification by its spectral minimum and detection of any changes and shifts in the monitored signals.

a C+L band amplified spontaneous emission (ASE) broadband source (Joinwit) is coupled to port 1 of a 3-port optical circulator. Light reflected from the end of each channel is redirected from port 2 to port 3 and then to the diffraction-grating-based spectrometer and to the CCD photodiode array detector (Askins, 1995). At port 2 there is an arrangement of three electrically controlled 1x2 fiber optic switches that allow an arbitrary access to four sensing channels. These channels can accommodate up to four LPGs depending on their bandwidth and sensitivity to a particular physical quantity. At the end of each channel there is a tunable reflector which returns light back to the sensing channel, and through port 3 the light is collimated onto a 600 lines/mm diffraction grating so the spectrum is observed by the CCD photodiode array.

SW2 Fiber-optic Circulator 1

2

SW3

SW1

3 Attenuator

Diffraction Precise Stepper Grating Motor

CCD Detection Unit

Control Unit

Figure 1 Multi-channel spectrum measurement system

The four measured fiber-optic channels are set in the following configuration – in Ch.1 two LPG sensors are placed, Ch2 and Ch.3 investigate single LPG and Ch.4 is utilized to perform reference signal measurement and system calibration with the ASE light source. 2.2

Detection and control devices

45

TOWARDS OPTIMISATION OF OPENMP CODES FOR SYNCHRONISATION AND DATA REUSE

R ( )   f 1 (n) f 2 (n   )

(1)



R ( max )  max With a cross-correlation function (1) the exact signal shift is calculated and any difference between signals is analyzed (Vaseghi, 2009). The multiple spatially shifted measurement of an identical signal may result in the increased spectral resolution. On Figure 3a two measurements with spatial shift are presented and the cross-correlation function R(τ) is calculated. The optimal signal match is achieved at maximum of R(τ) and the exact shift (τ) is calculated. 50 x10^3

1.0

Signal 1

Rmax=0.9923 tau=18

40

Intensity, I

The detection unit is based on a 512-pixel InGaAs CCD (G9204-512D – Hamamatsu Photonics) linear array with an integrated low-noise charge-amplifier featuring high sensitivity, a low dark current and high stability in the 8001750 nm spectral range. Two high-speed capacitive-based analog-digital converters (ADCs) transform the analog data from the CCD sensor into 16-bit corresponding digital values. The obtained data from the CCD array is filtered and further transmitted via an USB interface to a personal computer. The interrogation system is controlled and configured by an application using a LabView programming environment (Bitter, 2007) which allows an individual setup for various parameters - integration time (τ), sensor sensitivity (s), conversion speed, data communication speed (r). Also start, stop and pause functions are available for manual configuration and more precise measurement. Dark current can be subtracted after averaging, reference and the current signal can be read, and the signal-to-reference ratio can be presented in dB on the screen (Mikhov, 2009).

0.9

Signal 2 Cross-correlation function

0.8 0.7 0.6

30

0.5 0.4

20

0.3 10

0.2 0.1

τ 0

0.0 0

А

50

100

150

200 250 300 Pixel Number, n

350

50 x10^3

400

450

500

Filtered Signal

Intensity, I

40 30 20

Figure 2 Photograph of CCD Detection Unit 10

The communication protocol between personal computer and devices is command-based – via the LabView application a command is transmitted to the CCD detection unit and it responds with a corresponding packet of data or device status. If command is intended to adjust or to acquire status for fiber-optic switches’ position or stepper motor position then the CCD detection unit retransmits the command to the external control unit. The developed LabView application automatically configures diffraction grating angle position via the precise stepper motor and the measured channel. After the current channel is measured and properly visualized, the program automatically configures the next channel for a subsequent measurement. 2.3 Correlation analysis for accuracy improvements and noise cancellation

Correlation analysis in two separate spatially shifted signals is introduced to increase an accuracy of spectral measurement and to reduce the signal noises. The measured broadband signals from the fiber-optic sensors are spectrally resolved by a diffraction grating and are projected on the linear CCD photodiode array. To perform a spatial shift of the signal a precise stepper motor is implemented. It rotates the diffraction grating with 0.1deg accuracy.

0

B

0

50

100

150

200

250

300

350

400

450

500

Pixel Number, n

Figure 3 Correlation analysis of two shifted spectra signals - A, Noise suppression with spatially shifted measurements - B

To filter any noise resulting from the signal conversion or from the photodiode array and channel inequalities, an averaging low-pass filter with respect to the spatial signal shift is designed (2). It averages N-spectrally shifted measurements from the initial signal (Farhang-Boroujeny, 1998). Since the correlation function is preliminarily calculated and the exact shift is acknowledged, the average signal is calculated and noises are filtered.

1 N  S k (n   k ) N  1 k 0 0

S ( n) 

 k 0

(2)

By introducing this filtering scheme the acquired spectrum preserves the narrow minima in the measured grating sensors but also suppresses noises due to conversion or channel inequalities in the CCD array. The resulting filtered signal is presented on Figure 3b.

46


3 EXPERIMENTAL SET-UP AND MEASUREMENT RESULTS

0

Signal w/o Force Signal with Force

-5 LPG1

-10

LPG2

Loss, a (dB)

The designed multi-channel broadband spectrum measurement system is tested by analysing four different long period grating sensors arranged in three channels. On Ch.4 only a tuneable reflector was connected and this channel was utilized as a reference signal.

∆=2.6nm

-15 -20 a=7.8dB

-25

L= 100mm

Multi-channel spectra measurement 2

Channel 1 Channel 2

-3

-30 1480

1500

1520

Losses, a (dB)

Channel 3

1540 1560 1580 1600 Wavelength,  (nm)

1620

1640

-8

Figure 6 Channel 1 spectral response changes of the LPG1 and LPG2 sensors under stress

-13 -18 -23

-2 1505

1525

1545 1565 1585 Wavelength,  (nm)

1605

1625

Figure 4 Joint multi-channel spectral response measurement

A joint multi-channel graphic of spectral responses from all of the measured LPG sensors is presented on Figure 4. It is derived from direct measurement of every channel signal spectrum, and then the transmission losses are calculated with the reference to the fibre-optic source signal spectrum (3).

 S ( n)   S dB (n)  10 log  S ( n)   ref 

Signal w/o Force

-7

∆l=2nm

-12 -17 L= 100mm -22 1480

1500

1520

(3)

20 15 10

-5

Signal with Force ∆=27nm

-7 -9 L= 100mm

-13 1480

0 150

1640

Signal w/o Force

-11

5

100

1620

-3 Loss, a (dB)

25

50

1600

1 ASE Reference Signal Channel 1 - LPG1 & LPG2

0

1540 1560 1580 Wavelength, l (nm)

Figure 7 Channel 2 spectral response changes of the LPG3 sensor under stress

-1

35 x10^3 30

Intensity, I

Signal with Force

1645

Loss, a (dB)

-28 1485

200 250 300 Pixel number, n

350

400

450

500

Figure 5 Channel 1 LPGs’ signal compared to the reference spectrum

The measurement results from the reference ASE broadband light source and the LPG spectra in Channel 1 are shown in Figure 5. The calculated spectral response of LPG1 and LPG2 connected in Channel 1 are presented on Fig.6. It also includes measured spectra with an applied external field on the fiber (bending), which results in a spectral shift. The same results of the monitored spectra for Ch.2 and Ch.3 are presented respectively in Figures 7-8.

1500

1520

1540 1560 1580 1600 Wavelength,  (nm)

1620

1640

Figure 8 Channel 3 spectral response changes of the LPG4 sensor under stress

In Ch.1 two LPG sensors are simultaneously monitored and analysed. They are particularly selected not to interfere with each other’s spectra even if an external field is applied. The LPG1 shows significant depth change in its transmission minimum at λ=1515nm, while LPG2 results in minimum shift of Δλ=2.6nm. Corresponding transmission minimum shifts for Ch.2 and Ch.3 are presented in Figure 7 and in Figure 8. The latter


47

sensor is particularly sensitive that result in a larger minimum shift of Δ=27nm.

4

SENSOR DETECTION AND SPECTRUM ANALYSIS

The presented multichannel fiber-optic measurement system simultaneously monitors multiple grating sensors distributed in up to three fiber-optic channels. In order to automate and to facilitate the data analysis of all these sensors an advanced signal processing techniques are introduced. For every fiber-optic grating sensor the transmission minimum is required to be precisely calculated and monitored (Trees, 2001). To determine the spectral band for every sensor and to localize the minima an advanced signal processing is introduced. Due to the presence of noise in the local minima which can disturb the data interpretation and accuracy it is combined with low-pass noise filtering (4).

1 N  k (i) S dB (n  i) 2 N i  N k (i )  {1, 2, 4, 2, 1} S f ( n) 

(4)

Where SdB(n) is the measured spectrum, relative to the reference signal; k(i) are the coefficients of the low-pass IIR filter; Sf(n) is the filtered signal and n is the discrete number of spectrum signal. The designed filter suppresses noise, however it reduces the spectral resolution of monitored spectrum. But it also contributes to the more accurate calculation of the local extrema (6) which can be easily determined from the first derivative of the filtered signal (5). In order to further filter any false detections of the local extrema due to noise in the signal, the step (δ) function is introduced (Choi, 2006).

S (n)  S f ( n   )  S f ( n   )

 2  sign{S (i  1)}  0 S (i ) is loc max   sign{S (i  1)}  0  (i )    sign{S (i  1)}  0 S (i ) is loc min  sign{S (i  1)}  0

Figure 9 Channel 1 Spectrum analysis of the LPG sensors with local extrema localization

(5)

Figure 10 Channel 2 Spectrum analysis of the LPG3 sensor with local extrema localization

The local maximum separates the spectral response characteristics between the LPG1 and LPG2 sensors. The same spectra analysis is applied for the Channel 2 and Channel 3 (Figure 10 and Figure 11), which employ single LPG sensor per channel. The spectral response of the fiberoptic sensor in Channel 3 extends the monitored spectrum range which additionally impedes the spectra analysis.

(6)

The presented signal analysis effectively identifies the local minima of the fiber-optic grating sensors. Also if multiple grating sensors are multiplexed into a single fiberoptic channel, the spectral windows for every sensor can be defined by the calculation of their local maxima, which determines the sensor operation ranges. The results from the applied spectra analysis for Channel 1 are presented in Figure 9. It includes the first derivative function ΔS(n) and the localized extrema. Figure 11 Channel 3 Spectrum analysis of LPG3 sensor with local extrema localization

48


Fiber-optic grating sensors are particularly sensitive to applied external fields and changes in surrounding refractive index. It results in slight shifts of the measured spectral minima. Therefore it is crucial to properly measure and to calculate these spectral shifts. Two different methods are proposed in this paper. The first employs a direct localization of the minimum and calculation of the shift utilizing the same calculation techniques of local extrema for the forthcoming measured spectra and specifying the exact shifts (7) (Porter, 2006).

 LPG   S K   S 0 min

min

(7)

Figure 13 Channel 3 Spectrum analysis of LPG3 sensor based on the local extrema localization

Figure 12 Channel 1 spectral shift detection with direct minima localization

The results of the applied method are presented in Figure 12 which includes two different spectra measurements – with and without the applied external bending on the fiberoptic sensors. The localized minima and their wavelength values are calculated. However this direct approach does not provide a high accuracy and reliability of the calculated results due to the fact that the monitored transmission minima contain high noise levels which can cause false minima detections. More reliable method of spectral shift detection is monitoring the steepest areas of the transmission minima in the spectrum of LPG sensors (Figure 13). It ensures a significant reduction in the noise levels and increased sensitivity to the minimal changes (Wang, 2008). The localization of the steepest points in the spectra is derived from the already calculated dS(n)/dn function (8).

dS (n) max dS (n) min

p.1 p.2

 LPG  mean1 ,  2 

(8)

(9)

Figure 14 Spectra shift calculation for Channel 1 using steepest area monitoring consept

The calculation of exact transmission minimum shift is derived from the formula (9), where Δλ is the spectrum shift at the selected points (p.1 and p2) as shown in Figure 14. The latter method involves more precise calculations of the spectrum shits and eliminates any noise interferences, which can cause false minimum localization.

5

CONCLUSIONS

The reported automated multi-channel broadband analysis system for the system of fiber-optic grating sensors is capable of simultaneously monitoring up to 12 long period gratings in groups of three spectrally multiplexed sensors per channel. An advanced signal processing and analysis is demonstrated using cross-correlation function and adaptive filtering techniques which effectively suppress noises while preserving the spectral resolution. It is also possible to significantly extend the measured spectrum range by rotating the diffraction grating with the precise


stepper motor and by exploiting the wide sensitivity range of the linear CCD photodiode array (800-1750 nm). The proposed methods for sensor signal detection and analysis of the fiber-optic grating sensors successfully define the spectral minima for every sensor and detect any shifts of them occurring under the influence of external parameters. The measurement results demonstrate high sensitivity of the implemented long period grating sensors to external strain fields on the fibers and also the ability of the developed system to detect and to analyze those spectral changes.

ACKNOWLEDGEMENT

The authors acknowledge the support if the Ministry of Education of Bulgaria under research project VU EES 303/ 07, the Natural Sciences and Engineering Research Council of Canada and of the Canada Research Chairs Program.

REFERENCES Murphy, D. F., Flavin, D. A., McBride, R., and Jones, J. D. C., (2001), ‘Interferometric Interrogation of In-Fiber Bragg Grating Sensors Without Mechanical Path Length Scanning’, Journal of Lightwave Technology, No.19, pp.1004-1009. Alves, J., Santos, J.L., Carvalho, A., and Lage, A., (2003), ‘Fiber Bragg Sensor Interrogation System Based on a CCD Spectrometer’, Proceedings of IEEE, Sensors, No.1, pp.909913. Choi, J., (2006), ‘Adaptive and Iterative Signal Processing in Communications’, Cambridge Press. Askins, C.G., Putnam, M.A., Friebele, E.J., (1995) ‘Instrumentation for Interrogating Many-element Fiber Bragg Grating Arrays’, Proceedings of SPIE, Vol.2444 pp.257-266. Srimannarayana, K., Shankar, M., Prasa, R.S., (2008) ‘Fiber Bragg Grating And Long Period Grating Sensor For Simultaneous Measurement And Discrimination Of Strain And Temperature Effects’, Optica Applicata, Vol. 38,No. 3.

49

Vaseghi, S., (2009), ‘Advanced Digital Signal Processing and Noise Reduction’, 4th Edition, Wiley Press. Farhang-Boroujeny, B., (1998), ‘Adaptive Filters – Theory and Applications’, Wiley Press. Bitter, R., Mohiuddin, T., Nawrocki, M., (2007), ‘LabVIEW – Advanced Programming Techniques’, 2nd Edition, CRC Press. Balzhiev, P., Bock, W.J., Eftimov, T.A., Arnaudov, R., (2012), ‘Automated multi-channel broadband spectrum analysis of fiber-optic grating sensors’, International Scientific Conference on Infomation, Communication and Energy Systems and Technologies, pp.309-313, Veliko Tarnovo, Bulgaria. Guan, Z.G., Zhang, A.P., Liao, R., (2007), ‘Wavelength Detection of Coherence-Multiplexed Fiber-Optic Sensors Based on Long-Period Grating Pairs’, IEEE Sensors Journal, No.7, pp.36-37. Tamhane, A. C., (2009), ‘Statistical Analysis of Designed Experiments: Theory and Applications’, Wiley Press. Mikhov, M., Nakov, P., (2005), ‘A Hybrid Step Motor Drive System with High Posi¬tioning Accuracy’, Proceedings of the International Conference on Electronic Devices and Systems, pp. 59-64, Brno, Czech Republic. Mikhov, M., Nakov, P., (2008), ‘Stepping Motor Drive for Precise Positioning Applications’, Proceedings of the International Scientific Conference on Infomation, Communication and Energy Systems and Technologies, Vol. 1, pp. 227-230, Nish, Serbia. Trees, H., (2001), ‘Detection, Estimation, and Modulation Theory, Part I: Detection, Estimation, and Linear Modulation Theory’,John Wiley & Sons, Inc. Guan, Z.-G., Zhang, A.P., Liao, R. and He, S., (2007), ‘Wavelength Detection of Coherence-Multiplexed Fiber-Optic Sensors Based on Long-Period Grating Pairs’, IEEE Sensors Journal, Vol.7, pp. 36-37. Murphy, R.P., James, S.W. and Tatam, R.P., (2007), ‘Multiplexing of Fiber-Optic Long-Period Grating-Based Interferometric Sensors’, Journal of Lightwave Technology, Vol.25, pp. 825829. Allsop, T., Earthrowl, T., Reeves, R., Webb, D. J. and Bennion, I., (2004), ‘The Interrogation and Multiplexing of Long-Period Grating Curvature Sensors Using a Bragg Grating Based, Derivative Spectroscopy Technique’, Measurement Science and Technology, Vol.15, pp. 44-48. Allsop, T., Webb, D.J., Bennion, I. and Sakamoto, T., (2003), ‘Bend- and Index-Insensitive Long-Period Grating In Progressive Three Layered Optical Fiber’, IEEE Electronic Letters. Vol.39, Issue. 6, pp. 508-509, 2003. Porter, J., Queener, H., Lin, J., (2003), ‘Adaptive Optics for Vision Science Principles, Practices, Design, and Applications’, Wiley Press. Wang, B., (2000), ‘Digital Signal Processing Techniques and Applications In Radar Image Processing’, Wiley Press.,