Analysis of distributed temperature sensing based on Raman scattering using OTDR coding
and discrete Raman amplification
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2007 Meas. Sci. Technol. 18 3211
(http://iopscience.iop.org/0957-0233/18/10/S24)
Download details:
IP Address: 128.178.23.47
The article was downloaded on 01/11/2012 at 14:03
Please note that terms and conditions apply.
View the table of contents for this issue, or go to the journal homepage for more
Home Search Collections Journals About Contact us My IOPscience
IOP PUBLISHING MEASUREMENT SCIENCE AND TECHNOLOGY
Meas. Sci. Technol. 18 (2007) 3211–3218 doi:10.1088/0957-0233/18/10/S24
Analysis of distributed temperature
sensing based on Raman scattering using
OTDR coding and discrete Raman
amplification
Gabriele Bolognini
1
, Jonghan Park
2
, Marcelo A Soto
1
,
Namkyoo Park
2
and Fabrizio Di Pasquale
1
1
Scuola Superiore Sant’Anna, via G Moruzzi 1, 56124 Pisa, Italy
2
School of EECS, Seoul National University, Seoul, Korea
E-mail: gabriele.bolognini@cnit.it and nkpark@plaza.snu.ac.kr
Received 19 December 2006, in final form 13 February 2007
Published 12 September 2007
Online at stacks.iop.org/MST/18/3211
Abstract
The behaviour of distributed temperature sensors based on spontaneous
Raman scattering and coded OTDR (optical time domain reflectometry) is
studied both theoretically and experimentally; in particular a high
performance scheme has been implemented using amplitude modulation
according to Simplex coding, direct detection and additional use of lumped
Raman amplification to further extend the sensing range. An efficient and
cost-effective distributed temperature sensing system operating along 30 km
of dispersion-shifted fibre with 17 m spatial resolution and 5 K temperature
resolution is theoretically demonstrated and experimentally achieved using
255 bit Simplex coding and low-power commercially available laser diodes
(80 mW CW power). Use of lumped Raman amplification to produce
high-power coded pulses allows further 10 km distance enhancement,
resulting in a total measurement range of 40 km.
Keywords: fibre optics, sensors, Raman scattering, fibre testing, coding
1. Introduction
Raman-based distributed temperature sensors (DTS) have
been the subject of intense research for several years [1] and,
due to their wide range of practical applications, they are today
available as commercial products.
The sensing principle is based on optical time domain
reflectometry (OTDR) in which pump light pulses are sent
down along the sensing fibre and the backward propagating
light is detected. Most Raman-DTS systems based on
spontaneous Raman scattering (SRS) light monitor the ratio
of the temperature-dependent anti-Stokes to the Stokes light
intensities, thus allowing for a correction of most fibre link
losses. However, temperature assessment can also be based
on the ratio of the anti-Stokes to the Rayleigh backscattered
light intensities.
Being based on well-understood and mature technology
principles, the range of applications of Raman-based DTS
is to-date rapidly growing, mostly including monitoring of
oil/gas pipelines and power cables, fire detection systems
in tunnels and nuclear plants [2, 3], as well as aeronautic
applications [4].
Most Raman-based DTS commercial systems are based
on multimode fibre technology and are targeting sensing
distances within the range of 2–10 km.
For longer sensing distances, the development and
commercial exploitation of low-cost DTS systems is however
hindered by the low backscattered intensity of the anti-Stokes
light, which requires high peak power levels in the OTDR,
as well as high sensitivity detection schemes, in order to
make the sensor performance attractive. These features have
made other DTS systems more attractive and cost effective
for long distance applications, such as spontaneous Brillouin
scattering-based sensors [5].
We have recently shown that coded OTDR techniques
with fibre link optimization allow for enhanced performance
0957-0233/07/103211+08$30.00 © 2007 IOP Publishing Ltd Printed in the UK 3211
G Bolognini et al
in Raman-based DTS, avoiding the use of high peak power
pump pulses [6].
In particular, using 255 bit Simplex codes and low-power
commercially available laser diodes at 1.55 µm, we have
demonstrated a DTS system based on a dispersion-shifted
(DS) fibre with 17 m/3 K spatial/temperature resolution and
with a measurement range enhancement greater than 15 km
with respect to conventional OTDR-based DTS systems
operating on conventional single mode fibre. Note that
this has been achieved using off-shell conventional OTDR
hardware and standard third window optical communication
components, providing then a cost-effective solution with a
maximum sensing distance up to more than 30 km, using
dispersion-shifted fibre.
A possible way to further enhance the measurement range
without sacrificing spatial and temperature resolution, consists
in introducing optical amplification techniques.
Distributed Raman amplification, although very effective
in improving the performance of Brillouin-based DTS, does
not seem to be appropriate in the case of Raman DTS; this
is due to the wide spectral band of the used pump laser at
1.55 µm, which avoids the use of coherent detection, but
unfortunately does not allow one to effectively filter out the
amplified spontaneous emission added along the sensing fibre
by the distributed amplification process. However, discrete
(or lumped) optical amplifiers can be used to enhance the peak
power of the coded OTDR pulses.
In this paper, using lumped optical amplification in
combination with 255 bit Simplex coded pulses along zero-
dispersion-shifted sensing fibres, we demonstrate a sensing
range enhancement up to 10 km by amplifying the coded
OTDRsignalat1.55µm before launching it into the sensing
fibre; a measurement distance greater than 40 km is reported,
with spatial (temperature) resolution of 17 m (5 K).
Since standard erbium-doped fibre amplifiers and counter-
pumped discrete Raman amplifiers cannot be effectively
used for coded OTDR amplification (due to undesirable
waveform distortion related to slow transient effects), we have
developed a co-pumped low-noise discrete—or lumped—
Raman amplifier based on the DS fibre, which provides up
to 6 dB net gain to the coded OTDR signal. Standard fibre
Raman laser (FRL) technology at 1460 nm is used to co-pump
the discrete Raman amplifier; no penalties associated with
pump to OTDR signal noise transfer have been observed, due
to the limited bandwidth of the receiver.
The paper is organized as follows: we first introduce the
basic theory of Raman-based distributed temperature sensors
and then introduce the experimental set-up including optical-
coded OTDR and discrete Raman amplification. Analysis
of lumped amplification is then carried out and compared
with the most common optical amplification techniques.
Experimental results are hence presented, showing the
obtained enhancement in the measurement distance. Also
theoretical calculations have been carried out, well predicting
the S-coding gain as well as the performance enhancement due
to discrete optical amplification of the coded OTDR pulses.
2. Theory
In OTDR-based DTS systems, short laser pulses are sent
along the sensing fibre, and the backscattered light, containing
information about loss and temperature along the fibre, is
detected with high temporal resolution. In particular, as
the anti-Stokes line depends on both fibre temperature (due
to changes in phonon distribution) and fibre loss, the ratios
of anti-Stokes to Stokes light intensities or anti-Stokes light
(AS) to backscattered pump light (BS) intensities are typically
used for effectively distinguishing distributed temperature
variations from local losses.
In our measurements we use the ratio between the
backward propagating AS intensity I
AS
and the BS pump
intensity I
BS
, which can be approximately expressed (at the
scattering point) as
I
AS
I
BS
∝
exp
hν
R
kT
− 1
−1
, (1)
where h is the Planck constant, k is the Boltzmann constant, T
is the absolute temperature, and ν
R
is the separation between
Raman anti-Stokes and pump light frequencies.
Equation (1) must be properly integrated along the
distance and spectrum to take into account the wavelength-
dependent loss along the light path and the cross-section
spectrum of Raman anti-Stokes scattering.
In more detail, the amount of optical power incident at the
receiver deriving from pump Rayleigh backscattering and anti-
Stokes light scattering at distance z can be written respectively
as
P
BS
(z) = T
BS
W
2
R
BS
(z)P
0
exp
−2
z
0
α
BS
(ς) dς
(2)
and
P
AS
(z) = T
AS
W
2
R
AS
(z)P
0
×exp
−
z
0
[α
BS
(ς) + α
AS
(ς)]dς
, (3)
where W is the pulse width, P
0
is the input pump peak power,
α
BS
and α
AS
are the loss coefficients at BS and AS light
wavelengths, respectively. T
BS
and T
AS
are the transmittivity
values at the receiver end for AS and BS light, respectively.
R
BS
(z) is the Rayleigh backscattering coefficient at position z,
and can be written as
R
BS
(z) = S(z) · α
s
(z), (4)
where S(z) is the fibre capture factor at z (depending essentially
on the fibre type) and α
s
is the fraction of absorption coefficient
due to Rayleigh scattering (∝1/λ
4
) for BS light.
R
AS
(z) is the temperature-dependent Raman backscatter-
ing coefficient [7], and its expression can be written as
R
AS
(z) = C · g
R
(z)
ν
p
∈S
p
S
p
(ν
p
) dν
p
=
ν
2
−ν
p
ν
1
−ν
p
σ
R
(ν
R
)B(ν
R
+ ν
p
)
exp
hν
R
k
B
T(z)
− 1
dν
R
,
(5)
where g
R
(z) is the fibre Raman gain efficiency, S
p
(ν
p
)isthe
pump optical power spectrum, σ
R
(ν
R
) is the normalized cross-
section of Raman scattering, B(ν) is the spectral band-pass
filter shape, C is a constant proportionality factor, h and k
B
are the Planck and Boltzmann constants, T(z)isthefibre
temperature at position z. Double integration is performed
to take into account spectral widths of pump light and Raman
3212
Analysis of distributed temperature sensing based on Raman scattering
DSP PC
LD
driver
TIA
APD
BPF
DSF -4 DSF -5
10 km
1 km 29 km
TCC
FP-LD
sig
IN
sig OUT
ADC
LRA
DSF -1 DSF-2 DSF-3
21.8 km 1 km
TCC
Figure 1. Experimental set-up implementing Raman-based DTS with coded OTDR.
scattering effect. In addition, equation (5) accounts for the
non-ideal spectral response of the band-pass filter used to
separate the AS light from BS light.
In order to compensate for possible changes in fibre loss
due to connectors, bends and so on, temperature distribution
is derived by using the ratio of P
BS
(z)toP
AS
(z), giving
P
BS
(z)
P
AS
(z)
=
T
BS
T
AS
R
BS
(z)
R
AS
(z)
exp
−
z
0
[α
BS
(ς) −α
AS
(ς)]dς
.
(6)
Using equation (6) it is difficult to estimate the absolute
fibre temperature, but it is possible to accurately measure
the temperature variations along the whole optical fibre with
respect to the known absolute temperature (300 K in our case)
at the fibre input.
Coding techniques in OTDR schemes provide substantial
SNR improvements, which can be quantified by the coding
gain, defined as the ratio of SNR obtained with coded OTDR
to SNR obtained with conventional ODTR at a given distance,
and equal total number of measured traces.
Regarding the Simplex coding scheme, used in this work,
the coding process is implemented by amplitude modulating a
laser diode with sequences of pulses, according to code words
which are row vectors of an unipolar S-matrix [8]. Decoding
is achieved by a simple linear algebra operation on the set
of measured coded OTDR traces obtained from the different
code words. This provides an effective noise reduction in the
decoded trace, with respect to the single-pulse case, and a
consequent increase in the signal-to-noise ratio with respect
to conventional, single-pulse OTDR, for the same number
of acquired traces. The signal-to-noise ratio enhancement
provided by coding is quantified by the coding gain G
cod
,
which, for an arbitrary code length L and under the assumption
of large receiver bandwidth and zero-mean uncorrelated noise,
is given by
G
COD
=
L +1
2
√
L
. (7)
In a more realistic case, considering a narrow receiver
bandwidth B, and assuming ideal low-pass response [8], the
coding gain can be written as
G
COD
=
L +1
2
√
L
1 −
1
L
2
L
n=1
(L −n)
sin(2nπBτ )
2nπBτ
−1
,
(8)
where τ is the pulse period. As can be readily verified, for
our used bandwidth (3 MHz) and pulse duration conditions
(100 ns), the two equations give negligible differences in
expected coding gain. However, for narrower receiver
bandwidths, significant further improvement in coding gain
can be achieved (at the expense of worse spatial resolution).
It is also worth noting that Simplex coding can provide
SNR enhancement without sacrificing spatial resolution unlike
correlation-based coding techniques which degrade spatial
resolution performance to some extent [8].
Finally, note that the longer coded pulse duration for
255 bit code words (25 µs) does not decrease the power
threshold for the onset of stimulated Raman scattering, since
for pulses longer than about 100 ns the threshold power is
nearly constant (∼1 W) and does not depend on the pulse
width [9, 10].
3. Experiment
Figure 1 shows the experimental set-up used for implementing
the Raman-based DTS system based on coded OTDR and
lumped Raman amplification of OTDR coded pulses. The
laser diode at 1550 nm was intensity modulated (direct current
modulation through a laser driver module) by an in-house-built
PC-controlled OTDR board with a digital signal processor
(DSP), able to generate pulse patterns of Simplex coding [6],
with 100 ns single bit pulse width, as well as to implement
single pulses as required in conventional OTDR.
The light source was a commercially available Fabry–
Perot laser diode (FP-LD centred at 1550 nm, 80 mW
maximum output power, 10 nm FWHM, thus preventing
coherent speckles). The input pulses were injected into
the sensing fibre through an optical circulator, and the
backscattered light wave signals were then coupled to the
receiver after a large-bandwidth band-pass filter (bandwidths
are about 1420–1510 nm for anti-Stokes light and 1520–
1600 nm for Rayleigh-scattered pump light). Measured
transmittance spectra of BPF for the anti-Stokes and Stokes
light are reported in figure 2, showing a directivity higher than
∼30 dB for Rayleigh-backscattering port, and higher than
∼43 dB for anti-Stokes port, which permits the effective
separation of anti-Stokes and Rayleigh-scattering light.
In this paper we compare the performance of standard
OTDR, 255 bit S-coded OTDR and 255 bit S-coded OTDR
combined with a lumped Raman amplification scheme (LRA),
whose structure is schematically shown in figure 3, to provide
3213
G Bolognini et al
Figure 2. Transmittance spectra at BPF output.
low-noise optical amplification for the long coded pulse pattern
at 1550 nm, as explained below.
The receiver, described in figure 1, is based on a high
sensitivity InGaAs avalanche photodiode (APD), amplified by
a high gain transimpedance amplifier (TIA), with about 3 MHz
bandwidth. The estimated temperature sensitivity with this
set-up was about 0.65%/K(atT = 300 K).
After the receiver block, an analogue-to-digital converter
(ADC) was used to sample the incoming analogue data trace at
20 MHz. The OTDR traces resulting from coded pulses were
hence transmitted to the PC, where the de-coding process was
carried out. The spatial resolution (defined by the measured
10%–90% response time) for the current DTS system was
found to be 17 m, mainly limited by the TIA bandwidth.
Five spools of single-mode dispersion-shifted fibres
(DSF-1 to DSF-5) have been used (lengths are 21.8, 1,
29, 1 and 10 km, respectively). The fibres have been
spliced together to compose a total of 62.8 km sensing fibre
link. DSF-2 and DSF-4 have been put inside a temperature-
controlled chamber (TCC), while the other spools were kept at
room temperature (300 K). DSF are characterized by smaller
chromatic dispersion and higher Raman scattering coefficient
than standard SMF (showing similar loss coefficients), and
hence they have been found more suitable for long-range
Raman-based DTS [6].
4. Optical-coded pulse amplification
In order to increase the peak power at the DTS input, thus
permitting enhancement of the sensing distance in a Raman-
ISO
DSF
ISO
OTDR
coded
pulses
FRL
1460nm
Residual
Pump
1460 nm
WDM
4.7 km
4.7 km
Residual
Pump
1460 nm
WDMWDM
To optical
circulator
Figure 3. Scheme of co-pumped lumped Raman amplifier used to amplify long coded pulse sequences.
based DTS, the coded sequences at the output of the FP laser
have been optically amplified. The long used code words, and
the high required values of optical peak power impose serious
constraints for a suitable amplification technique, as explained
below. The proposed scheme is shown in figure 3,andis
based on a LRA. In this scheme, a WDM combiner couples
in the co-propagating direction the coded pulses from directly
modulated FP-LD (1550 nm, P
peak
= 80 mW) together with
a non-polarized FRL pump (1460 nm, P
max
= 1.9 W), into
a DSF spool (4.7 km, λ
0
= 1540 nm). At the LRA output,
two cascaded WDM combiners have been used to separate the
amplified OTDR pulses from the residual pump power with
high directivity, to avoid leakage of strong FRL residual light
into the sensing fibre.
The LRA stage provides a net gain of about 6 dB, allowing
one to attain about 320 mW peak power in coded pulse patterns,
with negligible power excursions within one codeword (note
that, as mentioned above, peak power remains below the
threshold value for the onset of detrimental nonlinear effects
[9, 10], thus allowing for spontaneous Raman temperature
sensing).
Note that the LRA shown in figure 3 isbasedonaco-
pumping scheme; actually, the use of other amplification
techniques, such as counter-pumped LRA or erbium-doped
fibre amplifiers (EDFAs) would not be effective due to
undesirable waveform distortions of the coded OTDR signal
at 1550 nm, induced by power transient excursions [11]
arising due to long codeword duration (about 25 µsfor
a 255 bit codeword). Transient power variation within
one codeword would cause significant impairment during
the decoding process, leading thus to the impossibility of
reconstructing the OTDR trace accurately. Semiconductor
optical amplifiers (SOAs) also do not appear suitable for coded
pulse amplification, due to typical lower values of saturation
output power.
In order to characterize the extent of transient effects for
the above-mentioned amplification techniques, a numerical
study has been carried out simulating the power waveform
at amplifier output in the case of 255 bit long coded
sequences, considering a high output power EDFA, a
counter-propagating LRA, and finally the co-propagating LRA
previously described. The numerical method used to assess
the transient dynamics of EDFAs is based on an average
inversion model [12], also called reservoir model, where the
upper level population of erbium ions and the exponential gain
coefficient are averaged along the fibre length, thus describing
the time-dependent gain of EDFA with a single ordinary
differential equation for the average inversion level, which
3214
