See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/5579201

High-Spatial-Resolution Distributed Strain Measurement in Optical Fiber with

Rayleigh Scatter

Article in Applied Optics · May 1998

DOI: 10.1364/AO.37.001735 · Source: PubMed

CITATIONS READS
619 2,346

2 authors:

Mark Froggatt Jason P Moore

Intuitive Surgical NASA
82 PUBLICATIONS 3,265 CITATIONS 26 PUBLICATIONS 1,221 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Jason P Moore on 24 May 2016.

The user has requested enhancement of the downloaded file.

High-spatial-resolution distributed strain
measurement in optical fiber with Rayleigh scatter

Mark Froggatt and Jason Moore

A method of measuring strain over 30-cm intervals to an accuracy of 10 microstrain in unaltered low-loss
communications-grade single-mode optical fiber is presented. The method uses a tunable external
cavity diode laser to measure the reflected intensity of a reflector–fiber system as a function of wave-
length. This measurement is performed with no strain applied to the fiber to produce a reference and
then again after a strain has been induced. Cross correlation of the Rayleigh scatter spectra from a
selected section of fiber in the strained and unstrained states determines the spectral shift resulting from
the applied strain.
OCIS codes: 060.2370, 060.2920, 060.2430, 060.2300, 290.5870.

1. Introduction tuations along the core are modeled as a Bragg

Most proposed methods of measuring strain in an
optical fiber rely on some locally induced change in
the fiber that forms the sensor. Extrinsic Fabry–
Perot interferometers require the construction of
small air gaps within the fiber.1 Intrinsic Fabry–
Perot interferometers require the construction of two
reflecting surfaces in the optical fiber generally by
use of a fusion splice.2 Bragg gratings are produced
by UV exposure causing periodic changes in the index
of refraction of the fiber core.3
Methods that do not use local changes in the fiber
rely on the inherent back scatter or birefringence in
the optical fiber.4 Raman scattering has been used
to measure the distributed temperature in an optical
fiber.5 Rayleigh scattering has also been used to
measure distributed temperature in optical fiber.6
Brilloun scattering that is due to thermally generated
acoustic waves propagating in the fiber core has also
been suggested as a means of measuring temperature
and strain in an optical fiber.7
The method described here uses the Rayleigh scat-
ter in a standard single-mode communications fiber.
Rayleigh scatter occurs as a result of random fluctu-
ations of the index of refraction in the fiber core.
The scatter from these random fluctuations has been
modeled as a one-dimensional change in the permit-
tivity as a function of distance.4 These random fluc-
The authors are with NASA Langley Research Center, Hamp-
ton, Virginia 23681-0001.
Received 11 June 1997; revised manuscript received 20 October
1997.
grating with a random variation of amplitude and
phase along the grating length. Although a random
Bragg grating appears to be an oxymoron, the model
works well to describe the reflections received from
an optical fiber. The system used to measure the
reflections in the optical fiber is a variation of a sys-
tem described by Glombitza and Brinkmeyer.8 The
random Bragg grating model works because a
narrow-linewidth source is used, and therefore only
the component of reflection at the exciting wave-
length is detected. The spectral nature of the mea-
surement makes relatively high strain resolutions
possible @10 parts per million ~ppm! over 30 cm# with
comparatively low spatial resolution ~0.6 cm!. To
achieve similar strain sensitivity with a standard op-
tical time domain reflectometry system, a spatial res-
olution of 3 mm is needed.
2. Theory
The problem considered has essentially three compo-
nents, and we discuss them in succession. The first
component is to derive a description of the light scat-
tered in the backward direction from a permittivity
that is a random function of distance along the fiber
length. The second component of this problem is the
analysis of a system in which a strong ~relative to the
Rayleigh scatter! reflector has been added to the sys-
tem so that we can recover both the phase and the
amplitude of the Rayleigh-backscattered light. In
this second analysis, we also address the effects of
measuring over a finite range of wavelengths. The
third component addresses some implementation ef-
fects such as discrete sampling of the measurement

1 April 1998 y Vol. 37, No. 10 y APPLIED OPTICS 1735

and cross correlation as a means of comparing ran-
dom, complex spectra.
Throughout the analysis we assume that the per-
mittivity of the fiber is a function only of distance,
even though we know that there are random fluctu-
ations in both the azimuthal and the radial directions
as well. We neglect these complicating factors and
show experimentally that for low strain levels this is
a valid assumption. We also neglect dispersive ef-
fects within the fiber. For small wavelength ranges,
and short ~,100-m! lengths of fiber, this introduces
little error.
A. Reflection from a Random Variation of Permittivity
The system considered is a single-mode optical fiber in
which the permittivity of the core varies along the axis.
Beginning with the wave equation for steady-state
propagation as described in an earlier publication,9
]2E
]z 2 1b 11
2
F
Dε~z!
ε
E 5 0, G (1)
where E is the electric field, b is the wave number in
the waveguide, ε is the permittivity of the fiber core,
and Dε~z! is the variation of the permittivity of the
fiber core as a function of distance along the core.
We immediately make the assumption that the for-
ward part of the propagating field is unaffected by the
small modulation of the permittivity as a function of z.
In low-loss fiber, for propagation distances less than
100 m, this is an excellent approximation for the am-
plitude of the forward-propagating field. If the coher-
ence length of the laser is much longer than the length
over which the measurement is performed, then this is
also an acceptable approximation of the phase of the
forward-propagating field. The electric field can then
be written as the sum of a forward-traveling wave and
a backward-traveling wave:
E 5 E0 exp~ibz! 1 C~z, b!exp~2ibz!,
where E0 is the constant amplitude of the forward-
traveling wave and C~z, b! is the spatially varying
amplitude of the backward-traveling wave. Substi-
tuting Eq. ~2! into Eq. ~1! and applying the slowly
varying wave approximation ~i.e., d2Cydz2 > 0!, we
obtain
dC b Dε~z!
5 @E0 exp~i2bz! 1 C~z, b!#.
dz 2i ε
For weak scattering over short distances, it is a good
approximation that
uC~z, b!u ,, E0. (4)
Applying this approximation @inequality ~4!# and in-
tegrating both sides over all z yields,
E0b
* Dε~z!
`
C~2`, b! 5 exp~i2bz!dz. (5)
2i ε
2`
This is the same result found for Bragg gratings in
earlier research.10 This relation tells us that the
complex amplitude of the backscattered field from a
random fluctuation in the permittivity as a function
of distance is the Fourier transform of the permittiv-
ity evaluated at twice the spatial frequency ~wave
number! of the exciting field.
B. Measurement of the Complex Reflection over a Finite
Bandwidth
Measurement of the backscattered field as a function
of wave number should then give a means of deter-
mining the spatial variation of the permittivity
through an inverse Fourier transform. If the range
of wave numbers over which we can measure C~2`,
b! is limited, then we will see only the spatial fre-
quencies of Dε~z! that fall within the range of wave-
number values 2b0 , v , 2b0 1 2Db where b0 is the
minimum value of wave number, Db is range of wave-
number values over which the measurement is per-
formed, and v is the spatial frequency of the variation
in the permittivity.
To perform the inverse Fourier transform, the
phase as well as the amplitude of the backscattered
field is needed. We reduce the measurement to a
simple intensity measurement by introducing a ref-
erence reflection into the fiber at the location z 5 z0
and examining the intensity of the detected field in
(2)
the presence of the reflector. We then have a new
detected field:
b
* Dε~z!
`
Cd~b! 5 E0 exp~i2bz!dz 1 rE0 exp~i2bz0!,
2i ε
2`
(6)
where the reflection, r2 is also weak but of the order
of a few percent. The intensity at the detector is
(3)
then

Id 5 Î
ε
m
Cd~b!Cd*~b!

* *
5 G6
b2 ` Dε~z! Dε~z!
`

Î
r2 1 exp~i2bz!dz exp~2i2bz!dz
ε 2 4 ε ε

F
2` 2`
5 E0 . (7)

* *
m b Dε~z!
` Dε~z!
`
1 r i exp~i2bz0! exp~2i2bz!dz 2 exp~2i2bz0! exp~i2bz!dz
2 2`
ε 2`
ε

1736 APPLIED OPTICS y Vol. 37, No. 10 y 1 April 1998

We can see that a number of the terms in Eq. ~7!
have the form of Fourier transforms; however, the
exact meaning of each term is not immediately
clear. Another important aspect to consider is that
we measured the reflected intensity over only a
small range of b values, therefore we proceed by
taking a inverse Fourier transform of the intensity
as a function of wave-number b using finite limits
on our integration.
mittivity into its separate spatial frequency compo-
nents, and the integral of Eq. ~10! created a new
permittivity function constituted from the fre-
quency components of the original permittivity
function that were in the range of the measure-
ment.
Substituting Dε# ~ z! from Eq. ~11! into Eq. ~8! allows
us to then evaluate the Fourier-transform integrals
to obtain

*
b01Db

b02Db
Id~b!exp~2ibx!db 5 2r2 Îε 2
m
E0 Db exp~2ib0 x!sinc~Dbx!

1 Îε 2 b02
m
E0
4ε2 * *
b01Db `
Dε~z!exp~i2bz!dz
*
`
Dε~z!exp~2i2bz!dz exp~2ibx!db

* F
b02Db 2` 2`

1 Îε 2 b0
m
E0 r
2ε
i
b01Db
exp~i2bz0!
*
`
Dε~z!exp~2i2bz!dz exp~2ibx!

G
b02Db 2`

*
`
2 exp~2i2bz0! Dε~z!exp~i2bz!dz exp~2ibx! db, (8)
2`

*
where the value of b was regarded as nearly con- b01Db
stant when it does not appear in an exponential, Id~b!exp~2ibx!db 5
and we allowed b to vary over a range of 2Db cen- b02Db
tered around b0.

5 6
2r2Db exp~2ib0 x!sinc~Dbx!

S D
To produce tractable integrals, and to provide
some useful insight, we introduce the Fourier trans-
form of the change in permittivity as a function of Î ε 2
E0
1p 2
b02 `
2ε 2` * x
Dε# ~z!Dε# z 2 dz
2 . (12)

F S D S DG
distance: m
b0 x x
1 pr i Dε# z0 2 2 Dε# z0 1

*
` ε 2 2
Dε̃~v! 5 Dε~z!exp~ivz!dz. (9)
2`
Because Dε# ~z! is a random function, the autocorre-
lation value for x Þ 0 will be much smaller than the
The permittivity function that we are able to mea- average amplitude of Dε# ~z!, so that

U S DU
sure will contain only frequency components that
b0
*
` x
fall at twice our laser wavelength. As a result, the
function measured will be a filtered version of the Dε# ~z!Dε# z 2 dz
2ε 2

U F S D S DGU
2`
actual permittivity. Therefore we introduce an-
other new function that is the component of the x x
variation of the permittivity that has spatial fre- ,, r Dε# z0 2 2 Dε# z0 1 (13)
quency components we can measure. The function 2 2
is given by
for x Þ 0. Because we are interested in locations
only where x Þ 0 we can further simplify Eq. ~12! by

*
1 2~b01Db!
Dε# ~ z! 5 Dε̃~v!exp~2ivz!dv (10) dropping the sinc function also.

Î
p

*
2~b02Db! b01Db ε 2 pb0
Id~b!exp~2ibx!db 5 exp~ib0 x! E0 r
and has the property that b02Db
m ε

Dε# ~ z! 5
2
p * *
b01Db

b02Db
`

2`
Dε~z!exp~2i2bz!dz exp~i2bx!db. F S D S DG
3 i Dε# z0 2
x
2
2 Dε# * z0 1
x
2
. (14)

(11) By assuming that x . 2Llead, where Llead is the length

of fiber between the circulator and the reference re-
The integral transform in Eq. ~9! separated the per- flector, the location z0 2 xy2 then corresponds to a

1 April 1998 y Vol. 37, No. 10 y APPLIED OPTICS 1737

position not in the fiber and therefore does not con-
tribute a reflection. After performing a change of
variables in the integration, we obtain
*2Db
Db
Id~b 2 b0!exp~2ibx!db 5 E02r
cpb0
ni
S D x
Dε# * z0 1 ,
2
(15)
where c is the speed of light in a vacuum and n is the
index of refraction of the fiber.
We now have an expression that relates the inten-
sity as a function of wave number to the variation of
the permittivity as a function of distance down the
fiber. Note that the amplitude of the reference re-
flection rE0 acts as an effective interference gain and
accounts for the systems ability to detect the weak
Rayleigh backscatter.
C. Method of Implementing Strain Measurement
Recognizing that the spatial data contain phase as
well as amplitude information, we can compute the
spectrum of any section of the fiber by selecting the
corresponding region of Dε# *yε~xy2! and transforming
it back to the frequency domain. In this case, the
spectrum is the transform of a random complex func-
tion and results in another random complex function.
We accomplished strain measurements by measur-
ing the shift in the Rayleigh-scatter spectra. Be-
cause of the random nature of the spectra, the shift
could not be readily identified by simple visual in-
spection. Instead, a complex cross correlation be-
tween the spectrum of a strained section of optical
fiber and the spectrum of the same section measured
with no strain applied was used to identify the shift
in the strained spectrum.
In practice, the data acquisition and processing all
occur in a discrete manner. As a result one actually
computes a finite discrete Fourier transform and a
discrete cross correlation. To begin, we measure the
intensity at a discrete set of wave numbers to obtain
an array, I~ jdk! 5 Ij , where 0 # j , N is an integer,
N is the number of samples measured, and dk is the
separation in wave number between samples. If we
approximate that b > nk, then the spatial range over
which we can measure is given by
p l2
Lrange 5 < , (16)
2ndk 4ndl
where l is the free-space wavelength. The discrete
Fourier transform of the intensity is
S D
1 N21 2p
Ĩm 5 (
N j50
Ii exp 2imj
N
. (17)
Ĩm is a discrete function of space with the step size
given by
p l2
Lres 5 < . (18)
2nDk 4nDl
1738 APPLIED OPTICS y Vol. 37, No. 10 y 1 April 1998
Note that Dl is defined to be only half of the total
range of the change in wavelength. If we measure
two intensity arrays and transform them so that they
are functions of length, yielding Ĩ~1! ~2!
m and Ĩm , we can
use some Fourier-transform identities to compute the
cross correlation between the spectra of any subset of
points as
S D
2m
1 2p
I~1!9
j
~2!*9
^ IN92j 5 (
2pN9 m5m1
Ĩ~1! ~2!*
m Ĩm exp ijm
N9
, (19)
where N9 5 m2 2 m1 1 1 is the total number of points
selected from the spatial array and I~1!9
j , Ij
~2!9
are the
spectra of the selected section of fiber. The signal-
to-noise ratio of the cross correlation will increase as
the square root of the number of points selected, and
the spectral resolution will also increase with the
number of points selected. Of course as the number
of points per strain measurement is increased, the
spatial resolution is degraded. We end up with a
space-strain product that is determined by the wave-
length in the medium:
l
Lresεres 5 , (20)
4n
where εres is the strain resolution and is therefore
unitless. In some mechanical situations, a strain
resolution of 1 ppm is sufficient, yielding a spatial
resolution of 26 cm. For the vast majority of appli-
cations, however, strain resolutions of only 10 ppm
are required, and we obtain a spatial resolution of 2.6
cm.
When implementing the system, one finds that the
spatial and strain resolutions are not the limiting
factors of the system, but the spatial and strain
ranges are. The range of the strain measurement is
given by the continuous tuning range of the laser
source, so that
Dl
εrange 5 , (21)
l0
where Dl is the continuous tuning range of the laser
source and l0 is the center wavelength of the tuning.
3. Experiment
A schematic of the apparatus used for the measure-
ment of the Rayleigh scatter is shown in Fig. 1. The
source is a piezoelectrically tunable external cavity
laser diode with a 50-ms line width of 50 kHz and 800
mW of fiber-coupled power. A low-finesse Fabry–
Perot cavity is formed by two air-gap reflectors, and a
fiber of length Lref 5 12 m. This Fabry–Perot cavity
is used to trigger sampling of the backscattered
power at well-defined intervals of wavelength. The
length of the Fabry–Perot determines the wavelength
spacing
l2
dl 5 < 67 fm (22)
2nLref

Fig. 1. Schematic of the apparatus used to measure the spectra of
the Rayleigh backscatter. IyO, input– output.
between intensity measurements and thus the range
of measurement of the system, which after substitu-
tion into Eq. ~16! is one half the length of the Fabry–
Perot cavity. The lead length of fiber is the fiber
between the point where the forward- and backward-
traveling light is separated ~i.e., at the circulator! and
the reference reflector. The fiber-optic circulator is
not necessary for this measurement; however, if a
coupler is used instead, all optical paths, including
those of the reference leg and its associated detector,
must be considered for contributions to the measure-
ment.
Figure 2 shows the amplitude versus distance
plot of the sensing fiber. We made this plot by
Fig. 2. Plot of the amplitude of the Rayleigh scatter as a function
of distance of the optical fiber. The large spike at 1 m that is due
to a reflection in the circulator and a smaller reflection from the
end of the fiber are both visible. A spurious reflection at 5 m is
also present.
Fig. 3. System used to strain the optical fiber with minimal ap-
plication of radial stress by use of two translation stages and two
fixed blocks. Each 30-cm section as referred to in Figs. 4 and 5 are
labeled here.
sweeping the tunable diode laser wavelength over
0.16 nm and measuring the power at detector 2 as
a function of wavelength and then taking the am-
plitude of the Fourier transform of this data. Note
that the first meter has a higher Rayleigh scatter
because of the addition of the lead fiber scatter to
the sensing fiber’s first meter as predicted above.
The spike at the end of the meter is assumed to be
due to an internal reflection within the optical cir-
culator.
Figure 3 shows the system used to induce strain
in the sensing fiber. Two translation stages are
used to independently induce strain in separate sec-
tions of the same optical fiber. There is a 1.0-m
section of nonsensing fiber leading from the refer-
ence reflection to the sensing fiber. The Rayleigh
scatter in this section of fiber cannot be separated
from the Rayleigh scatter in the fiber-optic circula-
tor’s pigtail. The first 30 cm of sensing fiber is to
the left of the translation stage and is therefore not
strained. The fourth 30-cm section of sensing fiber
is between two fixed stages and is therefore not
strained. All the other sections are subject to
strain. We achieved effective compression by pre-
tensioning the fiber sections when the reference
measurement was taken.
Three overlaid complex cross-correlation ampli-
Fig. 4. Cross-correlation results for compression, no load, and
tension in the second section of the sensing fiber.
1 April 1998 y Vol. 37, No. 10 y APPLIED OPTICS 1739

Fig. 5. Comparison of the strain measured by the shift in the
Rayleigh-scatter spectra and the strain as measured by the micro-
meter displacement of the translation stage for each 30-cm section
of the sensing length.
tudes taken at three different strain levels of the
measure spectra with respect to a reference spectrum
for the second section of the sensing fiber are shown
in Fig. 4. The strain levels were 20, 0, and 220 ppm
as noted in the figure. Fifty points were selected for
the cross correlation, giving a best-case signal-to-
noise ratio of 7. The zero-strain cross correlation
has a signal-to-noise ratio of 5, matching well with
the expected results.
Figure 5 shows the parts per million wavelength
shift in the Rayleigh scatter and the strain level ex-
pected in that section based on the micrometer move-
ment for the seven sensing sections in the fiber. We
computed the shift in wavelength by using the loca-
tion of the maximum value of the cross correlation.
The data were taken with the left micrometer posi-
tioned for 20 ppm of effective compression, and the
right micrometer was positioned for 56 ppm of ten-
sion. Figure 5 clearly shows that the system can
sense different strains at different locations in the
same unaltered optical fiber.
1740 APPLIED OPTICS y Vol. 37, No. 10 y 1 April 1998
View publication stats
4. Conclusion
We have demonstrated distributed strain sensing
based on Rayleigh scatter in unaltered single-mode
optical fiber. The demonstration was for strains of
less than 60 microstrain and in a laboratory environ-
ment in which birefringence effects were minimized.
The very high strain resolutions of 5 microstrain over
30 cm make this method a promising beginning for
future development of strain sensors based on Ray-
leigh scatter.
References
1. K. A. Murphy, M. F. Gunther, A. M. Vengsarkar, and R. O.
Claus, “Quadrature phase-shifted, extrinsic Fabry–Perot opti-
cal fiber sensors,” Opt. Lett. 16, 273–275 ~1991!.
2. T. Yoshino, K. Kurosawa, K. Itoh, and T. Ose, “Fiber-optic
Fabry–Perot interferometer and its sensor applications,” IEEE
J. Quantum Electron. QE-18, 1624 –1633 ~1982!.
3. G. Meltz, W. W. Morey, and W. H. Glenn, “Formation of Bragg
gratings in optical fibers by a transverse holographic method,”
Opt. Lett. 14, 823– 825 ~1989!.
4. A. J. Rogers and V. A. Handerek, “Frequency-derived distrib-
uted optical-fiber sensing: Rayleigh backscatter analysis,”
Appl. Opt. 31, 4091– 4095 ~1992!.
5. J. P. Dakin, D. J. Pratt, G. W. Bibby, and J. N. Ross, “Distrib-
uted optical fibre Raman temperature sensor using a semicon-
ductor light source and detector,” Electron. Lett. 21, 569 –570
~1985!.
6. A. A. Boiarski, G. Pilate, T. Fink, and N. Nilsson, “Tempera-
ture measurements in power plant equipment using distrib-
uted fiber optic sensing,” IEEE Trans. Power Delivery 10,
1771–1778 ~1995!.
7. M. A. Davis and A. D. Kersey, “Simultaneous measurement of
temperature and strain using fiber Bragg gratings and Bril-
louin scattering,” in Distributed and Multiplexed Fiber Optic
Sensors VI, A. D. Kersey and J. P. Dakin, eds., Proc. SPIE
2838, 114 –123 ~1996!.
8. U. Glombitza and E. Brinkmeyer, “Coherent frequency-
domain reflectometry for characterization of single-mode
integrated-optical waveguides,” J. Lightwave Technol. 11,
1377–1384 ~1993!.
9. M. Froggatt, “Distributed measurement of the complex modu-
lation of a photoinduced Bragg grating in an optical fiber,”
Appl. Opt. 35, 5162–5164 ~1996!.
10. K. O. Hill, “Aperiodic distributed-parameter waveguides for
integrated optics,” Appl. Opt. 13, 1853–1856 ~1974!.