Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
B
called Bragg wavelength and the wavelength is then ex-
pressed as follows:
B
= 2n, (3)
where n is the effective refractive index of the FBG and
is the modulation pitch of the refractive index of the FBG.
The Bragg wavelength
B
changes by longitudinal strain
z
applied to the FBG, and the Bragg wavelength-shift
B
is
expressed as follows [3]:
B
= (1 p
12
)
z
, (4)
68 S. Takashima et al. / Sensors and Actuators A 116 (2004) 6674
Fig. 2. Schematic diagram of the ow measurement system.
where p
12
is the elasto-optic constant of the optical ber and
is approximately 0.22. This yields the strain sensitivity of
1.2 pm/.
To obtain the shift
B
, the outputs of the interferometer
are used, and the outputs V
m
(m = 1, 2, 3) are expressed as
follows [9]:
V
m
=
m
V
in
+Re[()] =
m
V
in
[1 + cos(
MZI
+
m
)],
(5)
where () is the auto-correlation function of light wave
reected by the FBG sensor, V
in
is the voltage corresponding
to optical power reected by the FBG sensor, and
m
is
the coefcient compensating differences of photodetector
sensitivities and obtained from preliminary experiments. If
the split ratios of the 2 2 and 3 3 couplers are 1:1 and
1:1:1, respectively, one can obtain
1
= 0,
2
= 2/3 and
3
= 2/3, and the outputs V
1
, V
2
and V
3
are derived as
follows:
_
_
V
1
=
1
V
in
(1 + cos
MZI
)
V
2
=
2
V
in
_
1 + cos
_
MZI
+
2
3
__
V
3
=
3
V
in
_
1 + cos
_
MZI
2
3
__
(6)
The signal
MZI
can be then calculated using the following
equation:
MZI
=
2L
B
= tan
1
3(V
2
V
3
)
V
2
+V
3
2V
1
, (7)
where L is the optical path difference of the interferometer.
The relationship between the signal variation
MZI
and the
shift
B
is expressed as follows:
MZI
=
2L
B
+
B
2L
B
=
2L
B
(
B
+
B
)
B
2L
2
B
B
, (8)
where the term (
B
+
B
) was assumed to be nearly equal
to
B
because the shift
B
is signicantly smaller than
B
. An accidental loss of optical power while measure-
ment, which causes problems in optical intensity modula-
tion type sensors, is admissible in some measure because
the wavelength-to-phase sensitivity (=2L/
2
B
) depends
on only the path difference L.
3. Noise estimation of the FBG sensor with
interferometric detection
There are some reports about the noise estimation of FBG
sensors with interferometric detection. However the noise
estimation reports about the interferometric detection using
a 2 2 and a 3 3 couplers have not been presented.
3.1. Noise of the photodetector
Fig. 3 shows the circuit diagram of the photodetector.
The noise of the photodetector is dened by the noise of
a photodiode and a transimpedance amplier. The noise of
the photodetector is determined by thermal noises due to the
feedback resistance R
f
of the transimpedance amplier R
f
and shunt resistance R
sh
of the photodiode and by shot noise
due to the output current i
m
(=V
m
/R
f
) of the photodiode. The
Fig. 3. Circuit diagram of the photodetector.
S. Takashima et al. / Sensors and Actuators A 116 (2004) 6674 69
equivalent input noise voltage and current of the Op-Amp
are neglected because they are a few orders smaller than
the other noises. The RMS of the noise voltage V
N,m
(m =
1, 2, 3) can be then written as follows:
V
N,m
= R
f
_
4k
B
TB
w
R
sh
+2qi
m
B
w
+
4k
B
TB
w
R
f
, (9)
where k
B
is the Boltzmann constant (1.39 10
23
J/K), T
is the absolute temperature (300 K), B
w
is the equivalent
noise bandwidth of the photodetector (2.4 kHz) and q is the
electronic charge (1.6 10
19
C).
In order to calculate the shot noise by i
m
, the visibility
should be known. Although the spectral prole of a FBG is
expressed as a function of hyperbolic sine and cosine, for
ease of calculation of the visibility , we assumed that the
FBG had a Gaussian spectral prole S() given as follows
[10]:
S() =
2
ln 2
m
V
in
B
exp
_
4 ln 2
2
B
(
B
)
2
_
, (10)
where is the frequency of the light wave,
B
is the Bragg
frequency,
B
is FWHM of the FBG in the frequency
domain. From Eq. (10) and WienerKhintchine theorem, the
visibility can be derived as follows:
= exp
_
2
4 ln c
2
2
B
L
2
_
, (11)
where c is the speed of the light wave in vacuum. From
Eqs. (6) and (11), we can obtain the dc value of i
m
and
calculate the shot noise by i
m
.
3.2. Quantization noise during A/D conversion
Quantization noise should be taken into consideration be-
cause the system shown in Fig. 2 uses digitized signals for
demodulation of the signal
MZI
. We assumed that quantiza-
tion noise was a white noise, and the RMS of quantization
noise V
Q
may be then expressed as [16]:
V
Q
=
V
LSB
12
, (12)
where V
LSB
is the resolution of voltage. The A/D converter
with 16 bit and 10 Vmeasurement range used in the system
yields V
LSB
= 3.05 10
4
V.
3.3. Intensity noise of the optical source
The uctuation in the intensity of the ASE output should
be considered. The intensity noise arise as a uctuation in the
voltage V
in
. In the ideal situation, it is found from Eqs. (6)
and (7) that the intensity noise is deleted. But in the practical
situation, the intensity noise arises due to the existence of
the above-mentioned noises. We assumed that the intensity
noise was white noise whose RMS was 1% of the output
power of the ASE.
Fig. 4. Noise estimation of the FBG sensor with the interferometric
detection.
3.4. Noise estimation of the FBG sensor
Three uncorrelated white noises were given for noise esti-
mation of the FBG sensor. The output voltages V
est,m
(m =
1, 2, 3) and the sensor signal
MZI
are then written as:
V
est,m
= V
m
+V
noise,m
(V
m
), (13)
MZI
+
N
= tan
1
3(V
est,2
V
est,3
)
V
est,2
+V
est,3
2V
est,1
, (14)
where V
noise,m
(V
m
) is white noise whose RMS is (V
2
N,m
+
V
2
Q
)
1/2
and
N
is the uctuation of the signal. The noise
spectral density is calculated from the RMS of
N
divided
by B
1/2
w
and then converted to the noise in the wavelength
domain using Eq. (8).
Fig. 4 shows the relationship between FWHM
B
of
the FBG in the wavelength domain and the noise spectral
density for sufcient optical power. It is found that the noise
spectral density is mainly determined by the noise of the
photodetector. The increasing noise spectral density with the
broad
B
is due to decrease in the visibility . At
B
above 0.4 nm, the noise spectral density for L = 3.169 mm
is much higher than that for L = 1.635 mm. To the contrary,
at
B
below 0.4 nm, the noise spectral density for L =
3.169 mm indicates lower level than that for L = 1.635 mm.
This is because the optical path difference L has inuences
on both the wavelength-to-phase sensitivity (=2L/
2
B
)
and the visibility . The MachZehnder interferometer for
L = 3.169 mm is expected to yield the lower noise density
than that for L = 1.635 mm because FBGs in the system
normally have
B
of 0.2 nm.
The noise spectral densities of the FBG sensor with the
interferometric detection were estimated to be 210
4
and
10
4
pm/(Hz)
1/2
for L = 1.635 and 3.169 mm, respectively.
These values can be converted to 1.66 10
1
and 8.33
10
2
n/(Hz)
1/2
, respectively in consideration of the strain
sensitivity of the FBG (1.2 pm/).
4. Laboratory experiments of the cross-correlation
owmeter and discussion
Fig. 5 illustrates an experimental conguration of the FBG
sensors and the bluff body in the ow measurement section.
70 S. Takashima et al. / Sensors and Actuators A 116 (2004) 6674
Fig. 5. Experimental congurations of the FBG sensors and the bluff
body in the ow measurement section.
Fig. 6. A comparison of waveforms detected with the reference strain
sensor and the FBG sensor.
The size d
w
of the bluff body is 3 mm. From the principle of
the cross-correlation function R
SCOT
(), the coherent signal
with broad bandwidth is desirable, and we tested several
congurations of the sensors and the bluff body for broad
bandwidth of signal. The parameters of the congurations
are d
b
, d
s
and d
c
, which are the distance between the bluff
body and the upstream sensor, the distance between two
sensors and the distance between the bluff body and the
center of the PVC pipe, respectively.
4.1. Performances of the FBG sensor
Fig. 6 shows a comparison of waveforms detected with a
reference strain sensor (KYOWA, KFG-20-120-C1-11) and
the FBG sensor. The sensors were attached on a metal can-
tilever in this experiment. The waveforms of both the FBG
sensor and the reference sensor are similar as shown in this
gure. The difference between signal amplitudes of two sen-
sors is due to the difference of coupling of the sensors.
Fig. 7 shows the noise spectral density of the FBG sensor
as a function of
B
. This gure indicates that the experi-
Fig. 7. Experimental result of noise evaluation of the FBG sensor with
the interferometric detection.
mental noise spectral density is higher than the estimate at
narrow
B
. The reason for this difference is considered to
be due to the difference of characteristics of the photode-
tectors and quality (the split ratio) of the couplers used in
the MachZehnder interferometers. On one hand, at
B
of 1.2 nm, the experimental noise spectral density is lower
than the estimate. This difference is due to assumption of
the spectral prole of the FBG. The minimum noise density
of 4 10
4
pm/(Hz)
1/2
was achieved for both L = 1.635
and 3.169 mm at
B
of 0.2 nm. This value corresponds to
0.33 n/(Hz)
1/2
and is sufcient low to detect vortices in the
uid.
4.2. Flow measurement
4.2.1. Effect of the bluff body and Karman vortex shedding
frequency
To validate the effect of the bluff body, we compared
the signals with/without the bluff body. Fig. 8 shows the
measured waveforms, calculated R
SCOT
() and magnitude
squared coherence (MSC) for d
s
= 20 mm and the bulk
velocity v
b
= 1.0 m/s. The bulk velocity is dened as the
volumetric ow divided by the cross-sectional area of the
pipe. MSC indicates the similarity of two signals in the
frequency domain and dened as follows:
MSC(f) =
|G
xy
(f)|
2
G
xx
(f)G
yy
(f)
. (15)
MSC was used for evaluation of the bandwidth of the de-
tected signal. From Fig. 8a and b, it is clear that the ampli-
tude of the measured waveform with the bluff body is much
higher than that without the bluff body. The bluff body inu-
ences on both R
SCOT
() and MSC for the better. MSC with
the bluff body has the maximum of 0.9 around frequency
of 60 Hz although that without the bluff body does not have
such a peak. This coherent signal is due to Karman vortex.
As a result, R
SCOT
() with the bluff body has the sharper
and larger peak than that without the bluff body.
4.2.2. Inuence of the congurations: d
s
, d
b
and d
c
Fig. 9 shows MSC and R
SCOT
() obtained by the experi-
ments for various congurations and the bulk velocity v
b
=
1.0 m/s.
At rst, inuence of the distance d
s
is discussed. With
the shorter distance d
s
, the bandwidth of the coherent signal
becomes broader and the peak of the function R
SCOT
() be-
comes larger and sharper. This result means that vortices are
attenuated and degraded with its convection. Consequently
the distance d
s
is desirable to be as short as possible in a
view point of signal quality. On the other hand, the longer
distance d
s
is desirable for the estimation of velocity if the
resolution of the time delay is xed. The optimum d
s
de-
pends on the specications of the ow measurement system,
including the minimum and maximum measurement veloc-
ities and time resolution (sampling frequency of the A/D
converter).
S. Takashima et al. / Sensors and Actuators A 116 (2004) 6674 71
Fig. 8. Comparison of results for d
s
= 20 mm and v
b
= 1.0 m/s: (a) measured waveforms and R
SCOT
() without the bluff body; (b) magnitude squared
coherence without the bluff body; (c) measured waveforms and R
SCOT
() with the bluff body for d
b
= 20 mm and d
c
= 0 mm; (d) magnitude squared
coherence with the bluff body for d
b
= 20 mm and d
c
= 0 mm.
Next, we discuss the inuence of the distance d
b
. Shorter
distance d
b
makes the bandwidth of the coherent signal nar-
row as shown in Fig. 9a. In the region behind the bluff body,
the bandwidth of the vortex signal is narrow because the
signal is not disturbed. As a result, the upstream sensor de-
tects vortices with narrow bandwidth and the bandwidth of
the coherent signal becomes narrow.
In the case of the offset of the bluff body d
c
= 2.5 mm,
turbulent ow is generated because ow with different ve-
locities are mixed. The function R
SCOT
() is expected to
have a sharp peak because turbulent ow signal is thought
to have broad bandwidth. From Fig. 9a, it is clear that the
bandwidth of the coherent signal for d
c
= 2.5 mm is broader
Fig. 9. Inuence of the congurations on the vortex signals: (a) magnitude squared coherence; (b) cross-correlation function R
SCOT
().
than that for d
c
= 0 mm. However, the maximum of MSC
for d
c
= 2.5 mm is smaller than that for d
c
= 0 mm. The
function R
SCOT
() for d
c
= 2.5 mm indicates worse perfor-
mance on sharpness than that for d
c
= 0 mm. This may be
due to coherence degradation of the signal.
4.2.3. Linearity of the owmeter
Fig. 10 shows the experimental relationship between the
bulk velocity v
b
and the measured velocity v
meas
. All results
with the bluff body indicate better linearity than that without
the bluff body. The result for d
s
= 20, d
b
= 20 and d
c
=
0 mm exhibits the best accuracy with the relative error of
5%.
72 S. Takashima et al. / Sensors and Actuators A 116 (2004) 6674
Fig. 10. Linearity of the owmeter: (ac) reference vs. measured velocities; (df) relative error from the linear t.
4.2.4. Karman vortex
Karman vortex shedding frequency can be also utilized
for velocity estimation. The relationship between the bulk
velocity v
b
and Karman vortex shedding frequency f
K
is
expressed as follows:
f
K
=
S
t
d
w
v
b
, (16)
where S
t
is the Strouhal number and is usually constant at
ambient temperature. Fig. 11 displays the relationship be-
Fig. 11. Bulk velocity vs. Karman vortex shedding frequency for d
s
= 20,
d
b
= 20 and d
c
= 0 mm.
tween the bulk velocity and Karman vortex shedding fre-
quency for d
s
= 20, d
b
= 20 and d
c
= 0 mm. It shows a
good linearity. The Strouhal number is obtained to be 0.168.
The problem with ow measurement using Karman vortex
shedding frequency is that this method is susceptive to ex-
ternal vibration noise.
The owmeter presented here can measure the ow ve-
locity from both the time delay and Karman vortex shed-
ding frequency. The owmeter is expected to be reliable in
various situations by combining these techniques.
Fig. 12. The square of the bulk velocity vs. RMS of Bragg wavelength-shift
of the FBG sensor.
S. Takashima et al. / Sensors and Actuators A 116 (2004) 6674 73
Fig. 13. Denition of the coherence indicator P
H
/P
L
.
4.2.5. Minimum detectable velocity of the system
The minimum detectable velocity is determined by the
noise level of the FBG sensor and the coherence of the vor-
tices. Fig. 12 shows RMS of the Bragg wavelength-shift of
the FBG sensor as a function of the square of the bulk veloc-
ity. The relationship is linear, and the minimum detectable
Fig. 14. Bulk velocity vs. the coherence indicator P
H
/P
L
: (a) the result
for d
s
= 20 and d
c
= 0 mm; (b) the result for d
b
= 20 and d
c
= 0 mm;
(c) the result for d
b
= 20 and d
c
= 2.5 mm.
velocity is about 0.03 m/s in a view point of the noise level
of the FBG sensor in the experiment. To evaluate the coher-
ence of the vortices, an indicator P
H
/P
L
is dened as shown
in Fig. 13, and the minimum detectable velocity of the sys-
tem is a velocity when the indicator P
H
/P
L
is 1. Fig. 14
shows the experimental results of the indicator P
H
/P
L
as a
function of the bulk velocity. From this gure, the minimum
detectable velocity is found to be about 0.05 m/s in the ex-
periment. Consequently, the minimum detectable velocity of
the system is about 0.05 m/s.
5. Conclusion
We presented the principle of the cross-correlation
owmeter using FBG sensors. The interferometric detec-
tion was used as a FBG wavelength-shift detection method
for dynamic measurement. The noise spectral density of
the FBG sensor with the interferometric detection was dis-
cussed and it was estimated to be 2 10
4
pm/(Hz)
1/2
.
Experimental noise spectral density of the FBG sensor was
4 10
4
pm/(Hz)
1/2
. The water ow experiments showed
that the owmeter had a good linearity at velocity range
from 0 to 1.0 m/s and the minimum detectable velocity of
0.05 m/s. Furthermore, Karman vortex shedding frequency
was also detected and indicated to have a linear relationship
with the bulk velocity. This can be also utilized to mea-
sure the ow velocity, and the ow measurement system
is expected to be reliable by combining two measurement
techniques.
The owmeter presented in this paper has many advan-
tages including passive nature, explosion-protection, EMI
immunity and capability of remote sensing. We try to apply
this owmeter to geophysical use because the advantages
are leveraged in the application. By combining the owme-
ter with FBG pressure and temperature sensors, FBG sensor
system for pressure, temperature and ow measurement in
a borehole may be constructed.
Acknowledgements
This research was supported by Ministry of Education,
Sports, Science and Culture, Japan (Grant-in-aid for Scien-
tic Research, 09305068). FBGs we used were fabricated at
the Venture Business Laboratory, Tohoku University.
References
[1] G. Meltz, Overview of ber grating-based sensors, in: Proceedings
of the SPIEs Symposium on Smart Structures and Materials, vol.
2838, 1996, pp. 222.
[2] Y.J. Rao, In-ber Bragg grating sensors, Meas. Sci. Technol. 8 (1997)
355375.
[3] A.D. Kersey, M.A. Davis, H.J. Patrick, M. LeBlanc, K.P. Koo, C.G.
Askins, M.A. Putnam, E.J. Friebele, Fiber grating sensors, J. Light-
wave Technol. 15 (8) (1997) 14421463.
74 S. Takashima et al. / Sensors and Actuators A 116 (2004) 6674
[4] M.A. Davis, A.D. Kersey, Matched-lter interrogation technique for
bre Bragg grating arrays, Electron. Lett. 31 (10) (1995) 822
823.
[5] K. Wood, T. Brown, R. Rogowski, B. Jensen, Fiber optic sensors
for health monitoring of morphing airframes. I. Bragg grating strain
and temperature sensor, Smart Mater. Struct. 9 (2000) 163169.
[6] V.M. Murukeshan, P.Y. Chan, L.S. Ong, L.K. Seah, Cure monitor-
ing of smart composites using ber Bragg grating based embedded
sensors, Sens. Actuators A 79 (2000) 153161.
[7] M.A. Davis, D.G. Bellemore, M.A. Putnam, A.D. Kersey, A 60
element ber Bragg grating sensors system, in: Proceedings of the
11th International Conference on Optical Fiber Sensors, Sapporo,
Japan, 2124 May 1996.
[8] A.D. Kersey, T.A. Berkoff, W.W. Morey, High-resolution bre-grating
based strain sensor with interferometric wavelength-shift detection,
Electron. Lett. 28 (3) (1992) 236238.
[9] M.D. Todd, G.A. Johnson, C.C. Chang, Passive, light
intensity-independent interferometric method for bre Bragg grating
interrogation, Electron. Lett. 35 (22) (1999) 19701971.
[10] H.S. Kim, R.P.H. Haaksman, T.P. Newson, D.J. Richardson, Noise
properties and phase resolution of interferometer systems interrogated
by narrowband ber ASE sources, J. Lightwave Technol. 17 (11)
(1999) 23272335.
[11] W. Jin, Y. Zhou, P.K.C. Chan, H.G. Xu, A bre-optic grating sensor
for the study of ow-induced vibrations, Sens. Actuators A 79 (2000)
3645.
[12] R.S. Weis, B.L. Bachim, Source-noise-induced resolution limits of
interferometric bre Bragg grating sensor demodulation systems,
Meas. Sci. Technol. 12 (2001) 782785.
[13] A. Worch, A clamp-on ultrasound cross-correlation ow meter for
one-phase ow, Meas. Sci. Technol. 9 (1998) 622630.
[14] T. Dyakowski, R.A. Williams, Measurement of particle velocity dis-
tribution in a vertical channel, Powder Technol. 77 (1993) 135142.
[15] G.C. Carter, Coherence and Time Delay Estimation, IEEE Press,
New Jersey, 1993.
[16] R.C. Kavanagh, J.M.D. Murphy, The effects of quantization noise and
sensor nonideality on digital differentiator-based rate measurement,
IEEE Trans. Instrum. Meas. 47 (6) (1998) 14571463.