September
15,
1993 /
Vol. 18, No.
18 / OPTICS
LETTERS
1561
32-km
distributed
temperature
sensor
based
on
Brillouin
loss
in
an
optical
fiber
X.
Bao, D.
J. Webb,
and D.
A. Jackson
Applied
Optics
Group,
Physics
Laboratory,
The
University,
Canterbury,
Kent CT2
7NR,
UK
Received
May
3, 1993
We present
a novel
distributed
temperature
sensor
that
uses the
temperature
dependence
of the
frequency
at
which
the loss
is maximized
in the
interaction
between
a cw
laser and
a pulsed
laser.
With a
32-km
sensing
length,
a temperature
resolution
of
1'C has
been achieved;
it is also
shown
that a
spatial
resolution
of 5
m may
be obtained.
Among
the hierarchy
of sensors,
the distributed
fiber-
optic
temperature
sensor
is unique,
as
it offers
con-
tinuous
sensing
over tens
of kilometers
with good
temperature
accuracy
and
high
spatial
resolution.
Distributed
temperature
sensors
(DTS's)
that
use
Raman
scattering
as
the sensing
mechanism
have
already
been
reported.`-
4
More
recently,
the
variation
with
temperature
of
the Stokes
frequency
shift in
Brillouin
scattering
has
been
proposed
as a suitable
mechanism
for a DTS.
5
A system
based
on
this technique
has been
demon-
strated
by Kurashima
et al.
6
Those
authors
reported
a
temperature
resolution
of 3
0
C
combined
with
a
spatial
resolution
of 100
m over
1.2
km of
fiber
for this
technique,
which
they have
termed
Brillouin
optical-
fiber
time-domain
analysis
(BOTDA).
Higher
spatial
resolution
and
temperature
accuracy
combined
with
a significant
increase
in
the sensing
range
have
re-
cently
been
reported
6
'
7
with
BOTDA.
Our
research
7
has
indicated
that
the maximum
range
when BOTDA
is used
at 1.3
4
um (without
optical
amplifiers)
is
restricted
to -20-25
km
because
of system
noise
(combined
electronic
and
optical)
and
the input
cw
beam
power
limit, which
is governed
by the pulsed
power
depletion
that
occurs through
the
Brillouin
interaction.
Although
at 22 km
the
sensing
length
of
the DTS
described
in Ref.
7 is
the largest
yet reported
to our
knowledge,
there
are several
applications
for
which a
longer sensing
length
combined
with
higher
spatial resolution
is desirable.
To
this end
we have
explored
the possibility
of
using Brillouin
loss
rather
than Brillouin
gain.
We have
found
that significant
improvements
in the performance
of DTS's
based
on
Brillouin
scattering
are
possible
with this
approach.
In this
Letter
we describe
a DTS
system
based
on
Brillouin
loss
that has
demonstrated
a
temperature
resolution
of
1
0
C with
a spatial
resolution
of 5 m
and
a sensing
length
of
more than
32 km.
The
system
operates
as
follows:
light
from a
tunable-frequency
cw laser
at a
frequency
VLi (the
L in
the subscript
represents
the Brillouin
loss
process)
is launched
into
one end
of the
sensing fiber.
The output
from
a pulsed
laser
at frequency
VL2
is injected
into
the other
end of
the sensing
fiber.
When
'L1 =
VL2
+ VB
('B
is Brillouin
frequency),
the
counterpropagating
beams
interact
through
the
Brillouin
gain mechanism.
8
The
pulsed
beam is
amplified
at the
expense
of the
cw beam;
hence the
intensity
of
the cw
beam will
be reduced
as
a result
of the
Brillouin
interaction.
If
the intensity
of
the cw
beam emerging
from
the fiber
is monitored
following
the launch
of
the
pulsed
beam,
a decrease
in
the intensity
will
be
observed
whenever
Brillouin
loss occurs.
The
time
delays
between
the
launch of
the pulsed
beam
and
those
regions
where
power is
transferred
from
the
cw
beam correspond
to round-trip
times
for
light
traveling
to and
from the
regions
of Brillouin
loss.
These times
provide
the
positional
information.
If
the laser
frequency
difference
is adjusted,
then the
cw light
will experience
loss in parts
of the
fiber
at
a different
temperature.
Hence
by slowly
scanning
one of
the laser
frequencies
it is
possible
to map
out
the
temperature
distribution
of the
whole
fiber.
In our previous
system
7
the
cw laser
frequency
was
lower
than
that of
the pulsed
laser
and therefore
experienced
gain.
When depletion
of
the pulsed
light
can be
neglected,
the two
approaches
are
equivalent
for sensing
purposes,
giving
signals
that differ
only
in
sign
and not
in magnitude.
However,
the Brillouin
loss
approach
is superior
when
much of
the fiber
is
at
the same
temperature,
as
may often
be the
case.
In
such a situation,
when Brillouin
gain
is used,
the
pulsed
beam
is depleted
by the
interaction,
resulting
in
a very
weak signal
from
the end
of the fiber
most
distant
from
the
pulsed source.
Conversely,
with
Brillouin
loss
the pulsed
beam
is increased
as
a result
of the interaction,
resulting
in a much
stronger
signal
from this
end of
the fiber
and thus
permitting
the
realization
of longer
sensing
lengths.
The
experimental
arrangement
is
illustrated
in
Fig. 1.
Both lasers
were solid-state
cw diode-pumped
Nd:YAG
ring lasers
emitting
close
to 1319
nm. The
maximum
launched
power
of the
cw beam
laser was
- 10
mW. The
frequency
of the
laser
could be
ad-
justed
by temperature
tuning
the cavity.
A
Bragg
cell
was used
to provide
short
optical
pulses
rang-
ing in
time from
50 to
400 ns.
The peak
launched
pulsed
power
from
the first-order
diffracted
beam
was
-5 mW.
This signal
is monitored
at photodetector
DI and
is also
used to
synchronize
the
start of
the
data
storage.
The
zero-order
beam
was mixed
with
0146-9592/93/181561-03$6.00/0
©
1993 Optical
Society
of America
1562
OPTICS LETTERS / Vol.
18, No. 18 / September
15, 1993
SENSING FIBER
PUMP
~S~
LASER
a ( i
,sj
F] ED 4, ~~A
U U 0 UE F
LASER
Om ~~~i
T2 (15m) T3 (500m
r F7
AOM
LJ3
j lDC1
~~~~~~~AB -F11 BC -F2
{
~~~~~~~~CD - F3 DE - F4
1 ~~~~~~~~EF
- Fs
D
2 ffI separation between
03 SPECTRUM
I dillerent
fiber spools
ANALYZER
Fig. 1.
Experimental arrangement:
AOM, acousto-optic
modulator.
light from the cw beam
laser by directional coupler
DC1 and detected with a fast
(20-GHz) detector.
The resultant beat frequency was
monitored with a
microwave spectrum analyzer.
Most
of the pulsed power was launched into
one
end of the 32.2-km
sensing fiber by a 90:10 coupler,
DC2, while the cw beam
was injected into the other
end of the fiber by a variable-ratio
coupler, DC3. The
cw light
emerging from the sensing fiber was
de-
tected
with a low-noise photoreceiver with a 125-MHz
bandwidth amplifier
D2 and monitored with a stor-
age oscilloscope.
Ideally the sensing fiber would
be
fabricated from a single
length of fiber to ensure that
the propagation constants,
such as the mode field
diameter, attenuation,
and dispersion, are the same
throughout the sensing length.
The actual 32.2-km
fiber sensing length consisted
of five different fiber
spools, F1-F5 (see Fig. 1), all spliced
together. The
specifications of the fiber
spools were different, which
meant that the Brillouin frequency
shifts were dif-
ferent even when all the fiber was at the
same tem-
perature. Most of the sensing fiber (32.2
km) was at
ambient
temperature. However, three separate
sec-
tions of the sensing fiber
were placed in temperature-
controlled ovens, as shown in Fig. 1.
There were
nine splices over
the whole fiber length. All optical
launching and delaunching were
achieved with 20X
microscope
objectives. Manually adjustable polar-
ization controllers were
used at both ends of the
sensing fiber to maximize the Brillouin
loss signal
over the whole fiber length; their use is
necessary, as
the Brillouin
interaction is polarization sensitive.
Figure 2(a) shows two sample
traces measured
with the storage oscilloscope monitoring
the cw beam
intensity,
synchronized with the launch of the pulsed
beam
into the fiber. The pulse width is 50
ns,
corresponding
to 5-m spatial resolution. The
data
shown by the
upper trace correspond to the condition
in which the Brillouin
loss is maximized for fibers
F1 and F2 (21.2 km)
at ambient temperature. The
sharp transition in the signal that occurs
at point C
(i.e.,
where fibers F2 and F3 are spliced) occurs
because the propagation
constants of the fibers are
different. The discrete
steps in the trace are caused
by the digital nature of the oscilloscope.
In the
bottom trace of Fig.
2(a) the beat frequency was
chosen to maximize the Brillouin loss at
46'C (first
oven); the third
oven temperature was 54'C.
To compare the performance obtained
by using
Brillouin loss
against that which used Brillouin gain,
we repeated
the experiment
with the
sensing fiber
subject
to the
same temperature
distribution.
Fig-
ure
2(b) shows
the best
trace
that we could
achieve
with gain;
in this
case the minimum
pulse width
that
could be
used was 100
ns, as the signal
was so weak.
The upper
and lower
traces of Fig.
2(b) measured
with Brillouin
gain can
be compared
directly
with
the traces
of Fig. 2(a) obtained
with Brillouin
loss.
Comparing
the upper
traces
of Figs. 2(a)
and 2(b),
we
see that
with the
gain process
only the
signal
at
position
B is observable,
the
signals in
the C-F
region
of the fiber being
totally undetectable.
Comparing
the bottom
traces
of Figs. 2(a)
and 2(b),
where the
beat frequency
is set
to the Brillouin
shift
correspond-
ing
to the temperature
of the first
oven (46'C) at
position
B,
we find that
the peak
for Brillouin
gain
is much
weaker and
the peak at E
is not observable.
(Note
that the peak seen
at E in the Brillouin
loss
spectrum
is at a temperature
of 54 'C and hence
is not
at its maximum.)
These
data were
taken with
the
sample averaging facility of
the oscilloscope; typically
20
averages were necessary.
The
high spatial
resolution of
the Brillouin
loss
technique
cannot be
appreciated
from Fig.
2(a), as
each sample
of the
storage
scope averages
over
a
distance
of 50 m.
To demonstrate
the
higher
spa-
tial resolution,
we performed
experiments
with fiber
lengths of 15, 10, and 5 m
placed in the oven at po-
sition
D at
24.1 km
(temperature
46'C).
Again the
frequency
of the cw beam was adjusted
to maximize
the
Brillouin loss; the measurements
were taken with
pulse
widths of
100 and 50
ns. The
results of
these
experiments
are summarized in Fig. 3. Figure
3(a)
shows
the trace that
results when the
15-mi length of
_
A _ _
_ I
_ _
_ C4E
- 5km
|
(a)
5km
(b)
Fig. 2. Oscilloscope traces
showing cw beam intensity
monitored at D2 for (a) Brillouin loss
and for (b) Brillouin
gain.
Time base, 50 us/division; cw wave frequency
opti-
mized for
fiber at ambient temperature
(upper traces) and
at 46 'C, the temperature
of the first oven (lower traces).
September
15, 1993
/
Vol. 18,
No.
18 /
OPTICS
LETTERS
, .
A- . .
''
. .
~~~SW
m
(a)
Ia
50m
|-
(a)
v
.
.
...
50m
(b)
50m
(C)
.
.
.
.
.11"
.(d)
10m
e
(d)
Fig.
3. Expanded
view
of
the
oscilloscope
traces
of
cw
beam
intensity
as a
function
of
time
for
a section
of
fiber
in
oven
2 (fiber
D) at
46'C.
One
sample
corresponds
to
0.5 m
[for
(a)-(c)]
and
to
0.1
m [for
(d)];
see
text
for
details.
80
70
a
0
C.)
0.
S
60
50
40
12.35
12.40
12.45
Beat Frequency
(GHz)
Fig.
4.
Laser
beat frequency
that
gives
maximum
Bril-
louin
loss
for the
fibers
in oven
1 (Fl,
filled
circles)
and
oven
3 (F4,
open
circles)
as a
function
of
oven
temperature.
fiber
is
in the
oven
and
the pulse
width
is
100 ns.
Here
the time
axis
has
been
expanded
to
provide
0.5
m
of fiber
per
sample.
Figure
3(b)
shows
the
trace
with
10
m of
fiber
in the
oven,
again
with
a
100-ns
pulse,
and
Fig.
3(c)
shows
the
trace
for the
5-m
length
of fiber
with
50-ns
pulses;
this
trace
is
expanded
further
in Fig.
3(d),
which
provides
0.1
m
of fiber
per
sample.
In
these
traces
the
number
of
averages
was
increased
to
512.
Most
workers
in
the
field
of
DTS's
define
the
spatial
resolution
as
the
rise (or
fall)
time
of the
received
signal.
Ap-
plying
this
definition
to
Fig.
3(d),
we
could
claim
a
spatial
resolution
of
better
than
3 m;
however
we
prefer
to define
the resolution
as
the
half-width
of
the minimum
detectable
signal,
which
in this
case
is
-5
m.
The
resolution
is actually
governed
by
the
rise time
of the
Bragg
cell
used
in these
experiments.
We
believe
that,
with
a faster
Bragg
cell,
a better
spa-
tial
resolution
could
be achieved.
The
fundamental
spatial
resolution
is
determined
by the
phonon
life-
time,
i.e., the
reciprocal
of the
Brillouin
half-width;
this
limits
the
spatial
resolution
to 1 to
2 m,
depend-
ing
on fiber
composition.
Figure
4 is a
plot of
the laser
beat
frequency
needed
to maximize
the Brillouin
loss
that occurs
in the
first
oven (at
10.1
km,
fiber Fl)
and
in the
third
oven
(at
27.4
km, fiber
F4)
as a
function
of oven
temperature.
For
these
data
the
rms
deviations
from
the
best
straight-line
fit
were
0.93 and
1.12'C,
respectively.
From
the
data,
the
temperature
coefficient
of the
Brillouin
frequency
shift was
determined
to be
1.25
±
0.10
MHz/'C.
According
to these
experimental
data,
the temperature
resolution
of
this
system
is ±
1 'C.
We have
realized
a DTS
based
on Brillouin
loss
that
has
a 32.2-km
sensing
length
with a
1 'C temperature
resolution
combined
with
a spatial
resolution
of
5 m.
The authors
gratefully
acknowledge
the
financial
support
of
the UK
Science
and
Engineering
Research
Council
and
the
Optoelectronics
Interdisciplinary
Re-
search
Centre,
Southampton,
UK.
References
1.
A. H. Hartog,
J. Lightwave
Technol.
LT-1,
498
(1983).
2.
J. P. Dakin,
D. J. Pratt,
G.
W. Bibby,
and
J. N.
Ross,
Electron.
Lett.
21, 569
(1985).
3. A.
H. Hartog,
A. P. Reach,
and M.
P. Gold,
Electron.
Lett.
21,
1061
(1985).
4.
T. Shiota
and
F. Wada,
Proc.
Soc.
Photo-Opt.
Instrum.
Eng.
1586,
13
(1991).
5.
D. Culverhouse,
F.
Farahi,
C. N.
Pannell,
and
D.
A.
Jackson,
Electron.
Lett. 25,
914
(1989).
6.
T. Kurashima,
T. Horiguchi,
and
M. Tateda,
Opt.
Lett.
15,
1038 (1990).
7. X.
Bao, D.
J. Webb,
and
D. A.
Jackson,
Opt.
Lett.
18,
552
(1993).
8.
T. Horiguchi
and M.
Tateda,
J. Lightwave
Technol.
7,
1170 (1989).
1563