Sie sind auf Seite 1von 19

3338

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

A Dual Golay Complementary Pair of Sequences


for Improving the Performance of Phase-Shift
Pulse BOTDA Fiber Sensor
Mohd Saiful Dzulkefly Bin Zan and Tsuneo Horiguchi, Senior Member, IEEE

AbstractWe propose in this paper to configure a dual Golay


complementary pair of sequences (DGCP) by nesting one Golay
complementary pair of sequences (GCP) into the other GCP. We
show that the DGCP also has the unique property that the side
lobes of the correlation functions are cancelled exactly as the conventional GCP. The DGCP allows the simultaneous use of coded
RZ-pulses and coded NRZ-pulses in the pump light of the Brillouin optical time-domain analysis (BOTDA). We call them in this
paper coded discrete pulses and coded continuous pulses, respectively. The simultaneous use of both types of coded pulses for the
BOTDA makes it possible to employ virtually longer code than the
use of only one kind of coded pulses. We theoretically show the dependence of signal-to-noise ratio enhancement (SNRE) on the code
length of the DGCP as well as on the code lengths of the GCPs used
for the discrete and continuous pulses. Examination of the spatial
resolution is also given. Experimental results illustrate the theoretical considerations; about 7 dB of SNRE and 10-cm of spatial
resolution are achieved by using DGCP, being in accordance with
the theory.
Index TermsComplementary code, coded Brillouin optical
time-domain analysis (BOTDA), fiber optic sensor, stimulated
Brillouin scattering (SBS).

I. INTRODUCTION

RILLOUIN optical time-domain analysis (BOTDA) fiber


optic sensor has been developed to measure distributed
strain and temperature by utilizing stimulated Brillouin scattering (SBS) [1]. A BOTDA measurement system uses an intense pulse light (pump) and a counter propagating continuouswave light (probe) at a Stokes frequency injected into a fiber
optic cable. When the frequency difference between the pump
and the probe is tuned to the Brillouin frequency shift (BFS) of
the fiber, the probe is amplified by the pump through the process

Manuscript received February 27, 2012; revised May 27, 2012, July 24,
2012; accepted August 10, 2012. Date of publication August 27, 2012; date
of current version October 17, 2012. This work was supported in part by
KAKENHI(22560427) of Japan Society for the Promotion of Science (JSPS).
The work of M. S. D. Zan was supported in part by the Ministry of Higher
Education of Malaysia (MOHE), MARA Education Foundation (YPM), and
the Universiti Kebangsaan Malaysia (UKM).
M. S. D. Zan are with the Graduate School of Engineering and Science,
Shibaura Institute of Technology, 1358548 Tokyo, Japan, and also with the
Department of Electrical, Electronics and System Engineering, Universiti Kebangsaan Malaysiak, 43600, Malaysia (e-mail: nb11106@shibaura-it.ac.jp).
T. Horiguchi is with the Graduate School of Engineering and Science,
Shibaura Institute of Technology, 1358548 Tokyo, Japan.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JLT.2012.2215578

of the SBS. BOTDA measures the amplified probe waveform.


The BFS increases linearly with the changes in the strain and
temperature of the fiber. Therefore, the BOTDA can measure
local changes in the strain and temperature along the fiber. The
spatial resolution of the BOTDA was around 1 m at best at an
early stage of the development [2]. The 1 m resolution corresponds to the pump pulse width of 10 ns. Shortening the pulse
width to less than 10 ns, which is around the phonon lifetime
for silica glass fibers, causes the Brillouin gain to decrease significantly and consequently broadens the gain linewidth. Both
the gain decrease and the linewidth broadening make it increasingly difficult to measure BFS accurately.
However, this difficulty has been overcome by new approaches which include Brillouin optical correlation domain analysis (BOCDA) [3] and pulse pre-pump BOTDA
(PPP-BOTDA) [4]. The technique of BOCDA is based on
optical correlation, and has achieved a high spatial resolution of
1.6 mm [5]. The interrogation range of the BOCDA, however,
is limited to about
times of the spatial resolution unless
time-division pump-probe generation scheme is employed [6].
The PPP-BOTDA originates in experimental observations of
the sudden reduction in the Brillouin linewidth when the pulse
duration gets shorter than the phonon lifetime [7]. Later on,
this surprising behavior has been theoretically explained [8];
the light leakage from an optical pulse modulator has played
a role of pre-pump that has accumulated the phonons for the
SBS. Therefore, the PPP-BOTDA utilizes a weak pre-pump of
a few tens of nanoseconds in duration and the following high
peak-power pump of a short duration. The PPP-BOTDA makes
the short pulse pump fall so steeply to suppress a trail of the
Brillouin signal due to the presence of the transient phonons,
achieving high spatial resolution. The trail has been unwanted
response for the BOTDAs utilizing the light leakage from the
modulator [9] and that using a phase pulse imbedded in cw
pump [10].
BOTDAs that utilize the pre-pump more efficiently than
the previous PPP-BOTDA have been recently reported. They
include differential pulse-width pair BOTDA (DPP-BOTDA)
[11], optical differential parametric amplification BOTDA
(ODPA-BOTDA) [12] and phase shift pulse BOTDA
(PSP-BOTDA) [13], [26]. Three types of BOTDAs above
use a long pulse pair and measure the differential Brillouin gain
of the pulse pair. The DPP-BOTDA employs two separate long
pulses with a small pulse-width difference that determines the
spatial resolution. In a similar manner, the ODPA-BOTDA uses
two long pulses with a small pulse-width difference. However,

0733-8724/$31.00 2012 IEEE

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

the pulses are generated at the Stokes and the anti-Stokes


frequencies, being injected into a fiber simultaneously. In
contrast, the PSP-BOTDA uses two long pulses that have the
same pulse-width; one without phase shift and the other with
-phase shift at the last portion of the pulse. The width of
the phase shifted portion determines the PSP-BOTDA spatial
resolution. Since three BOTDAs above employ the pre-pump
much higher in peak power than the previous PPP-BOTDA, the
differential Brillouin gain with DPP-BOTDA, ODPA-BOTDA
and PSP-BOTDA is stronger than that obtained with the
PPP-BOTDA.
More recently, coded DPP-BOTDA [14] and coded
PSP-BOTDA have been demonstrated for improving the
signal-to-noise ratio (SNR). The coded DPP-BOTDA has
employed discrete pulses, i.e., return-to-zero (RZ) pulses,
for coded signals rather than continuous pulses, i.e., non-return-to-zero (NRZ) pulses in order to avoid interactions
between the preceding and the succeeding pulses through the
process of the SBS. For the same reason, we have employed
the discrete phase pulses separated by 100 ns for the coded
PSP-BOTDA [15]; we call it coded discrete-PSP-BOTDA.
The pulse pump separation, however, reduces the number
of coded pulses that can be injected into a fiber during the
round-trip-time (RTT), i.e., the time taken for the light to make
a round trip along the fiber.
We also have reported coded PSP-BOTDA that employs
coded continuous pulses, i.e., NRZ pulses [16]. We call the
BOTDA coded continuous-PSP-BOTDA. We have found that
a good correlation function is obtained even for the coded
continuous-PSP-BOTDA. However, the total duration of the
coded continuous pulses should be no more than a few tens
of nanoseconds, i.e., the phonon lifetime, hence the maximum
limit of the code length is about the quotient of the phonon
lifetime divided by the coded pulse width, being not so large.
Both schemes in [15] and [16] have employed complementary
sequences called Golay complementary pair (GCP) [17].
In this paper, a new set of sequences is proposed for combining of the coded discrete pulses and the coded continuous
pulses used in the PSP-BOTDA. We call the new set of sequences a dual Golay complementary pair of sequences (DGCP)
since the DGCP is configured by nesting one GCP into the other
GCP. Algorithms for creating and correlating of the DGCP are
presented. The correlation function of the combined DGCP is
shown to have the same characteristic with that of the conventional GCP; cancellation of the range side lobes. This special property being true even for coding of very short pulses
shows promise in sensing distributed strain and temperature
with a high spatial resolution. Furthermore, the employment of
the DGCP in the coded PSP-BOTDA system makes it possible
to increase the total code length that can be used for the measurement, which further increases the SNR. Numerical simulations
and experimental results are presented to show the advantage of
the employment of the DGCP in the PSP-BOTDA systems.
This paper is organized as follows. In Section II we
briefly outline the PSP-BOTDA and two kinds of coded
PSP-BOTDAs; the coded continuous- and coded discrete-PSP-BOTDAs. An example of the dual Golay coded
pump is also explained. In Section III we explain the generation

3339

of the GCP, the auto-correlation process and its correlation


function. In Chapter IV we describe the algorithms to generate
the DGCP from two GCPs. The mutual correlation functions
of the DGCP are formulated, showing the special property
of producing zero range side lobes. Examples of the mutual
correlations of the DGCP are given in Section V. In Section VI
we describe a way to synthesize the pump composed of the
pre-pump pulse and dually coded pulses for the PSP-BOTDA.
Numerical simulations of the dual Golay codes PSP-BOTDA
are presented in Section VII. Experiments and discussions are
given in Section VIII. Section IX concludes the paper.
II. PSP-BOTDA SYSTEMS
A. PSP-BOTDA
In the PSP-BOTDA, the pump is composed of two kinds of
pulses; the first pulse of long duration, and the second pulse of
short duration. The second pulse width determines the spatial
resolution. The first pulse is around or longer than the phonon
lifetime so that the acoustic wave grows to its full extent through
the process of SBS. The second pulse follows immediately after
the first pulse in order to interact with the acoustic wave excited
by the first pulse. The second pulse is modulated with binary
phase shift keying (BPSK). Therefore, the following two pumps
are used: 1) the first pulse followed by the second pulse without
phase shift and 2) the first pulse followed by the second pulse
with shift in phase.
First, pump 1) is injected to the fiber and the response is measured. Second, the measurement with pump 2) follows. In both
measurements, a portion of the second pulse is backscattered
by the acoustic wave excited by the pre-pump; the first pulse.
For pump 1), the second pulse backscattered interferes with the
probe constructively, increasing the Brillouin gain. Contrary to
the case of pump 1), the second pulse backscattered in the case
of pump 2) interferes with the probe destructively, decreasing
the Brillouin gain. The increase and the decrease in the gain
have the same magnitude. Third, the response signal measured
with pump 2) is subtracted from that obtained with pump 1).
This process cancels the Brillouin amplified signals caused only
by the first pulse and only by the second pulse, leaving only
the Brillouin amplified signal caused by a conjunction of the
second pulse and the acoustic wave excited by the first pulse.
The differential Brillouin amplified signal of the PSP-BOTDA
is much greater than the Brillouin amplified signal for the previously reported PPP-BOTDA. This is because the pre-pump
pulse and the second pulse of the PSP-BOTDA have the same
power while for the PPP-BOTDA the power of the pre-pump is
about
of that of the second pulse. Furthermore, the subtraction process doubles the amplified signal of the PSP-BOTDA
since their signs are opposite as explained above.
B. Coded PSP-BOTDA
By modulating the second pulse of the pump with codes we
are able to improve the performance of the basic PSP-BOTDA
explained in the previous section. We have already reported
coded discrete-PSP-BOTDA [15] and coded continuous-PSPBOTDA [16]. As aforementioned, in this paper, we propose

3340

Fig. 1. Coded PSP-BOTDA pump configurations. (a) Coded continuous-PSPBOTDA. (b) Coded discrete-PSP-BOTDA. (c) Dual Golay codes PSP-BOTDA.

the new dual Golay complementary pair of sequences and introduce the dual Golay codes PSP-BOTDA employing the new
sequences [18].
Fig. 1 shows the examples of pump configurations in the case
of 4 bits in code length for the three types of the coded PSPBOTDAs. All pump pulses of the coded PSP-BOTDAs above
are phase coded. Phase coding offers 3 dB advantage in the
SNR over amplitude coding commonly employed in conventional coded BOTDAs. For the coded continuous-PSP-BOTDA
shown in Fig. 1(a), the first pulse of width is used to grow the
acoustic wave through the SBS process, while the second pulses
of width each, which are four pulses in the case of 4 bit code,
are used to measure the strain and temperature changes. Assigning 1 and
bit-elements to 0 and shifts in phase, the four
pulses are phase modulated with code series of
.
As for the coded discrete-PSP-BOTDA shown in Fig. 1(b), the
second pulse of width
following the first pulse of width
is assigned to one bit-element of the code. The four sets of the
first and the second pulses are generated and the second pulses
are phase-modulated with the code series of
. The
pulse sets are separated by in time so that we can ignore the
interactions between them through the process of the SBS. In
both pump configurations
determines the spatial resolution,
as has been reported in [16] and [15].
We have shown in Appendix A that in contrast to previous
intensity-modulation coded BOTDAs the coded continuous-PSP-BOTDA is immune to code pattern effects due to
nonlinear SBS amplification of NRZ pulses. We also have
shown in Appendices A and B the analytical and numerical
results of Brillouin gain spectra obtained with the coded continuous- and discrete-PSP-BOTDAs, demonstrating that their
Brillouin gain coefficients and spectral widths are approximately the same with those of the steady state SBS.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

The combination of the coded continuous pulses and the


coded discrete pulses creates the pulse pump of the new dual
Golay codes PSP-BOTDA, which is illustrated in Fig. 1(c). The
pump in Fig. 1(c) is configured by the following procedure;
First, four coded pulses in Fig. 1(a) are inserted into the second
pulse each in Fig. 1(b). Second, the sign of bit-elements of the
four pulses are inverted if the value of the bit-element shown
in Fig. 1(b) is
while the four pulses are left as they are if
the value of the bit-element in Fig. 1(b) is
. Then the sixteen
values of the bit-elements are obtained and depicted in Fig. 1(c).
It should be noted that in this example the code length used for
the dual Golay codes PSP-BOTDA is four times greater than
that used for the previous coded PSP-BOTDAs. The maximum
code length
for the coded continuous-PSP-BOTDA is
limited by the phonon lifetime while the maximum code length
for the coded discrete-PSP-BOTDA is limited by fiber
length. In contrast, the dual Golay codes PSP-BOTDA can
employ the DGCP whose code length is as great as
bits. Then, there is no practical increase in the measurement
time. Therefore, the dual Golay codes PSP-BOTDA potentially
has greater SNR than the other coded PSP-BOTDAs as well
as the conventional PSP-BOTDA. Furthermore, the mutual
correlations of the DGCP still hold the special properties
of the conventional GCP. This property will be clarified in
Section IV. At the best of our knowledge, this is the first report
that proposes DGCP and analyzes their use for the high spatial
resolution BOTDA theoretically and experimentally.
III. GOLAY COMPLEMENTARY PAIR (GCP)
The definition of a GCP is as shown below [17].
Definition: A pair of L-element sequences, A with elements
and B with elements
, is
said to be complementary if the sum of the auto-correlations for
the two sequences satisfies the following equations:
(1)
where
(2)
(3)
for k=0
otherwise.
In the equations above, is the code length of Golay codes A
and B, is the bit shift during the auto-correlation calculation.
The out-of-phase aperiodic auto-correlation coefficients sum to
zero, while the summation gives a peak value of
for
.
Let us consider a GCP, which consists of codes A and B. The
codes can be generated, for example, through a method called
recursion method, which can be expressed as follows:
(4)

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

where

3341

refers to the operator of the nesting process, and


for
for

Fig. 2. Correlation functions for

In (4), means concatenation, and overbar denotes bitwise


inversion of the code. The expressions of
and
denote
the new codes of A and B, respectively. An example of a GCP
of 8 bits in length is given by
(5)
which is generated from a pair of one bit code of
and
by using the recursion method three times. Fig. 2 shows
a correlation function of a GCP of 8 bits. The special property
of a GCP expressed by (1)(3) is clearly observed in Fig. 2.
The code gain of
and the zero side lobes are favorable for
distributed sensing.

(10)

and so on. From (8) and (9), the combination of the two pairs
results in four sequences of the DGCP of length MN.
In the dual Golay code PSP-BOTDA system, we distinguish
the two GCPs with different names for convenience; one pair is
the continuous pulse codes and the other pair is the discrete
pulse codes. One may immediately realize that these names
correspond to the configuration in Fig. 1(a) and (b), respectively.
Note that Fig. 1(a) and (b) include the long first pulse with a
pulse width
which is used to excite acoustic wave, as explained before. Since this section explains the methods to generate the DGCP, it is convenient to express the codes in terms
of bit value and also to exclude the first pulse. We assume the
and
codes and in (6) as the continuous pulse codes,
and codes C and D in (7) as the discrete pulse codes,
and
. Since each pulse coded with
or
is generated separately as shown in Fig. 1(b), the DGCP to the
dual Golay codes PSP-BOTDA system should include zero elements. Then, modifying the right sides of (8) and (9), we obtain
four sequences as follows:

(11a)

IV. DUAL GOLAY COMPLEMENTARY PAIR (DGCP)


A. Generating the DGCP From Two GCPs
The DGCP is generated by combining one GCP with the other
pair. Let us consider two GCPs; the first pair of codes A and B
and the second pair of codes C and D. These pairs and their
elements can be expressed by

(11b)

(12a)
(6)
(12b)
(7)
where
is the code length for codes A and B and N is the code
length for codes C and D. The process of combining the two
pairs is called nesting; the elements in and is multiplied by
each element of and to generate the DGCP which can be
explicitly expressed as
(8a)
(8b)
(9a)

Here, denotes a sequence of the zero elements of length


that is the interval between the discretely configured coded
pulses. Take note that the superscript refers to the type of the
code. We call a set of these four sequences DGCP. The four
sequences have the same length
that is given by
(13)
is the code length of the continuous pulse codes
where
and
is the code length of the discrete pulse codes.
Let us make an example to generate the (AC) code by using
(11a). Codes
and
are assumed to be the same 4 bit
codes as follows;
(14)

(9b)

(15)

3342

Fig. 3. Example of pulses modulated with (


pulses are omitted.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

) code of the DGCP. The first

The bit number


can be chosen as being any non-negative
integers as long as the bit interference can be neglected. Here we
assume
bits for simplification and the signal amplitude
is zero in the interval. Then the (AC) code becomes

Fig. 4. Mutual correlation between (


.

) and

(20)

(16)
),
equals 28 bits according
and the total code length of (
to (13). The form of the signal coded in the time domain is
shown in Fig. 3 as a function of a bit. Since the fourth bit of
the
is
, the elements of the fourth
are inverted,
as shown in Fig. 3.
Reference [19] has proposed the generation of combined
Barker codes, which is quite similar to the DGCP. However,
the combined Barker codes differ from the codes proposed in
this paper in two respects; First, the combined Barker codes
include no space code S that plays important role in suppressing
crosstalks between coded bit signals through the SBS as will
be explained later. Second, the auto-correlation function of the
combined Barker codes has side lobes more than 1. This means
that the combined Barker codes have lost the unique property
of the original Barker codes that the side lobes are not larger
than 1. In contrast, the proposed DGCP maintains the property
that the side lobes are canceled exactly, as will be proved in the
next sections.

codes for

for
, where refers to the bit shift during the
mutual correlation process and the subscript on the left side of
refers to the interval of sampling the dual Golay codes data
for the correlation. It is assumed that the elements
and
of co des (AC), (BC), (AD)
and (BD) are zeros for
. As an example, Fig. 4 illustrates the way to calculate the correlation by using (17) when
. The variable in Fig. 4 denotes the time
slot of the discrete pulse code element and is given by
(21)
The correlations of (ii) are expressed as
(22)
(23)

B. Dual Golay Codes Correlation (Mutual Correlation)


The process of correlating the dual Golay codes in the DGCP
during the pulse compression stage is called mutual correlation.
For the dual Golay codes PSP-BOTDA system, the process decodes the measured data. The procedure of the mutual correlations is divided into two groups.
1) Correlations with the pair of continuous pulse codes (
and
).
2) Correlations with the pair of discrete pulse codes (
and
).
The correlations of 1) are expressed by
(17)
(18)
(19)

(24)
(25)
for
. The variable on the left-hand side
of
indicates that the dual Golay codes data sampled with
the interval are correlated with the discrete pulse code. The
elements
, and
of
codes (AC), (AD), (BC) and (BD) are assumed to be zeros for
. An example is shown in Fig. 5, where the
way to calculate the correlation given by (22) is depicted, when
.
There are two methods to perform the mutual correlations by
using both the calculations of: 1) based on (17)(20) and those
of 2) based on (22)(25). The first method processes 1) and 2) in
this order, while the second method calculates them in reverse
order. They will be discussed in Sections II-C and II-D.

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

3343

(AC) in (22) with the calculated data


, we mutually correlate the
with
. In a similar manner, we mutually
correlate
with the
code by using (23). Summing
these correlations results in the final correlation of the DGCP as

Fig. 5. Mutual correlation between (


.

) and

(29)

codes for

Equation (29) shows that the peak value at zero-shift of the correlated code corresponds to
and that the side lobes
are canceled exactly.

C. First Method to Perform DGCP Mutual Correlation


(Method I)
In this method, codes (
) and (
) form one pair and
codes (
) and (
) the other pair. First, (
) and (
)
codes are mutually correlated with
and
by using
(17) and (18), respectively. Summing both correlated codes and
using (1)(3) give the following equation:

D. Second Method to Perform DGCP Mutual Correlation


(Method II)
The second method is analogous to the first method except
that for the second method, the mutual correlation starts with
the correlation with the discrete pulse codes. First, using (22),
the (AC) code is sampled at the interval T. The sampled data
is mutually correlated with
by using (22). Similarly, the
(AD) code is sampled at the interval and is mutually correlated with the
code by using (23). The summation of these
correlations is expressed as

(26)
notates the floor function and
means the largest
where
integer not greater than x. If we define the data generated from
(26) as
, it can be expressed as

(30)

(27)
) code is mutually correlated with
by
Second, (
using (19) and (
) code is mutually correlated with
code using (20). Just as with the derivation of (26), summing
both correlated codes above yields the following:

(28)
whose exThe data generated from (28) is defined as
pression can be similar to (27).
The third process is to mutually correlate the data
with
code by using (22) with modification: Replacing

with
has not been defined
Strictly speaking,
yet. However, hereafter we will use the expression of (30) for
simplicity since in (30) equals to zero when
. The
data generated from (30) are denoted by
and is given by

(31)
elements of
Equations (30) and (31) show that the first
agree with the
elements multiplied with
,
and the rest of
elements are zeros.
Second, in a similar manner, (BC) and (BD) codes are mutually correlated with the
and
codes by using (24),
and (25), respectively. The summation of the correlations is as
follows:

3344

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

(32)
whose exWe define the data generated from (32) as
pression can be similar to (31).
and
are mutually correlated
Third, the data
with
and
, respectively. The summation of these
correlations yields the total correlation of the DGCP as

Fig. 6. Mutual correlation functions for (


sum of (26).

) of (17), (

) of (18) and their

Fig. 7. Mutual correlation functions for (


sum of (28).

) of (19), (

) of (20) and their

(33)
It is not surprising that the final result of (33) agrees with that of
(29), since the correlations are linear systems and the calculation
order is exchangeable. From (29) and (33), we can clearly see
that the magnitude of correlation coefficient code for
is
four times the product of the code length of the continuous and
discrete pulse codes and that the side lobes are canceled exactly.
V. EXAMPLES OF MUTUAL CORRELATIONS
In this section, examples for both correlation methods
(methods I and II) are explained for a clearer view on the
mutual correlation equations provided. Consider a pair of the
continuous pulse codes
and
of the code length
, and a pair of the discrete pulse codes
and
of the code length
, with the bit interval
. The generated GCPs based on (4) and the generated
DGCP based on (11) and (12) are as follows:
(34)
(35)
(36)
(37)
(38)
(39)
(40)
(41)
based
Then the total code length of the dual Golay codes
on (13) is 14 bits and the time-slot of the discrete pulse code T
based on (21) is 7 bits.
A. Example of Method I
of (17), and
The mutual correlations between (AC) and
between (BC) and
of (18), and their summation of (26)
are illustrated in Fig. 6. In Fig. 6, the summed data
of
(26) have two peaks of
in amplitude for
and
, and have zeros for other values. The pattern of the

peaks is the same with


, except that the peaks are separated
by the value of
. The mutual correlations between (AD)
and
of (19), and between (BD) and
of (20), and their
summation are shown in Fig. 7, showing that the summation
based on (28) produces
with amplitude of
and the pattern is the same with
, as is the case explained
above.
In the next process, the data
shown in Fig. 6 and the
data
shown in Fig. 7 are mutually correlated with
and
, respectively. The mutual correlation functions and
their summation of (29) are shown in Fig. 8. From Fig. 8, we can
see that the amplitude of the summation of (29) at
equals
to 32, which corresponds to
. Furthermore, the
summation process produces no side lobes for other values.
B. Example of Method II
Using the same codes listed in (34)(41), an example based
on Method II is explained. The mutual correlations between
(
) and
of (22), and between (
) and
of (23),
and their summation of (30) are illustrated in Fig. 9. In Fig. 9, the
summed data
produce the peak amplitudes of
and
. The data pattern is the same with the
. Next,
in a similar manner, the mutual correlations between (
) and
of (24), and (
) and
of (25), and their summation

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

3345

Fig. 8. Mutual correlation functions for


their sum of (29) based on method I.

, and

, and

Fig. 11. Mutual correlation functions


their sum based on method II.

further to be employed in the dual Golay codes PSP-BOTDA


sensing system.
VI. DUAL GOLAY CODES PSP-BOTDA OPTICAL PULSE PUMP

Fig. 9. Mutual correlation functions for (


sum of (30).

) of (22), (

) of (23), and their

This section explains a way to configure the pulse pump of


the dual Golay codes PSP-BOTDA system. There are two types
of GCP based coded PSP-BOTDA sensing systems that have
been proposed in our laboratory and reported recently. They are
coded continuous- and coded discrete-PSP-BOTDAs [16], [15]
whose pump configuration have already been shown in Fig. 1(a)
and (b), respectively.
In both measurement systems, the pulse pump is comprised
of two kinds of pulses with different widths. The width of the
first pulse is long enough to ensure an efficient acoustic wave
excitation. The second pulse is phase coded with the typical
GCP and appended next to the first pulse for the measurement
purpose. In both systems, the width of one element of the second
pulses
corresponds to the spatial resolution of the system,
, as follows:
(42)

Fig. 10. Mutual correlation functions for (


sum of (32).

) of (24), (

) of (25) and their

of (32) are calculated. The results are depicted in Fig. 10. The
summed data have the peaks of 4 and
in amplitudes and the
data pattern is the same with
.
and
,
Finally, the mutual correlations between
and between
and
, and their summation are calculated by using (33), being depicted in Fig. 11. The summed data
agree with that obtained by using Method I, as explained before.
Therefore, we can confirm that both methods I and II can be applied to perform the correlation calculations of the DGCP and

where is light velocity in the fiber. The configuration of the


pulse pump of the dual Golay PSP-BOTDA is based on the combination of those two pumps as mentioned in the previous sections. By combining the pump configurations shown in Fig. 1(a)
and (b), we can produce the pump for the dual Golay codes
PSP-BOTDA system, as is shown in Fig. 12 that is identical
with Fig. 1(c). In Fig. 12, as has already been indicated,
is
the width of the first pulse,
is the width of one element of
the dual Golay code pulses, and is the interval time between
adjacent groups of pulses. The use of the pump based on DGCP
in Fig. 12 makes a greater contribution to the increase in the
SNR of the Brillouin signal than the previous pumps shown in
Fig. 1(a) and (b). This is because a dual Golay code makes it
possible to generate more coded pulses within the RTT, and to
increase the SNR with the number of the coded pulses.
The value of is crucial for accurate and efficient measurements. It is set long enough so that no interference between separate groups of pulses occurs through SBS process. However,

3346

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

Fig. 12. Example of a pump configuration of dual Golay codes PSP-BOTDA,


) code with
is used. The first pulses each
where (
precede the dually coded second pulses.

Fig. 14. Model for dual Golay codes PSP-BOTDA and fibers.

TABLE I
PARAMETERS USED FOR THE SIMULATIONS

Fig. 13. Schematic of BOTDA.

long
decreases the code length which is available within a
given period of transmitting the pump. It has been shown in [15]
and [20] that the choice of
to 100 ns makes a practical
compromise between these two factors which affect the performance of the coded discrete-PSP-BOTDA system. Similar results have been reported in the case of RZ coded pulses BOTDA
with 1-m spatial resolution [21].
VII. DUAL GOLAY CODES PSP-BOTDA SIMULATIONS
A. Simulation Setup
The SBS process in BOTDA involves the interaction among
pulse pump, cw Stokes and acoustic wave. The configuration
of the basic BOTDA system is depicted in Fig. 13. The pump
with frequency
is injected at the distance
, moving to
the
direction across a test fiber, while a continuous Stokes
wave that we call the probe with frequency
is injected at the
distance of
, moving to the direction of
across the test
fiber. The backward SBS wave is generated from the periodical
fluctuation of the dielectric constant of the fiber material due
to the pressure wave, i.e., acoustic wave excited by the beating
between the pump and the probe. Complying with the energy
conservation law and the phase matching condition, the probe
is amplified when the frequency difference between the pump
and the probe coincides with the Brillouin frequency shift which
is given by [22]
(43)
is the acoustic
Here, is the refractive index of the fiber,
velocity and
is the pump wavelength. The amplified probe
propagates to the optical coupler and finally detected at a
detector.
The dual Golay codes PSP-BOTDA system resembles the
basic BOTDA system except that the pulse pump used in the
measurement process is phase coded with the dual Golay codes
and the detected signals are processed with the codes. The configuration of the dual Golay codes PSP-BOTDA simulations
is shown in Fig. 14 where the test fibers are also shown. The

lengths of the test fiber


used were 0.3 m, 1 m, 3 m and 10
m. The 10 m fibers connected between the test fibers were used
as the reference. The difference in the BFS
between the test
fiber and the reference fiber was set to 300 MHz, which simulates a change in 0.6% strain. The optical power of the first pulse
was set at the same level with that of the second pulses
.
The values of the parameters used in the simulations are listed
in Table I.
Take note that two conditions, Condition A and Condition
B were considered in the simulations. Under Condition A, the
code length of the continuous pulse codes was fixed at 4 and
the code length of the discrete pulse codes was varied from 1 to
32 bits. In contrast, under Condition B, the code length of the
discrete was fixed at 4 and the code length of the continuous
pulse codes was varied from 1 to 32 bits.
The physical interactions between the three waves in the fiber
optic cable can be modeled as follows: [8]
(44a)
(44b)
(44c)
, and
are the normalized pump, Stokes, and
where
acoustic fields, respectively. is the normalized frequency de-

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

Fig. 15. Simulated correlation output of Brillouin amplified signals under


Condition A.

Fig. 17. Simulated relative SNRs versus code length

3347

under Condition A.

TABLE II
RISING EDGE LENGTH OF 3 M STRAINED FIBER UNDER CONDITION A

Fig. 16. Simulated Brillouin amplified signal power versus code length
under Condition A.

tuning expressed by
where
and
are the frequencies of pump and Stokes wave, respectively, and
is the Brillouin frequency shift.
B. Simulation Results (Condition A)
The Brillouin amplifications for each
of Condition A
are shown in Fig. 15, where the horizontal axis represents the
distance and the vertical axis the Brillouin optical signal. It appears that the Brillouin optical power detected at the 0.3 m, 1 m,
3 m and 10 m strained fibers increases with the increment of the
discrete pulse code length
. At a glance, the rising and the
falling edges for each strained length are clearly identified, and
no degradation in spatial resolution can be found for
to 32. Details on these observations will be given later.
Further analysis was performed by investigating the signal
power increment and the relative signal-to-noise ratio (SNR).
These results are depicted in Figs. 16 and 17, respectively. The
signal power increment was evaluated by the ratio of the signal
power
to the power for
. Here we note that
the value
in Fig. 16 corresponded to
the sum of two sets of measurements with a 4-bit GCP having
two sequences, while
was calculated by one set
of measurements with the DGCP of
and
having four sequences; then the number of code sequences used
to modulate the transmitting pump waves was the same among

the measurements of
and
; thus the measurement time was the same among all measurements with
ranging from 1 to 32. Fig. 16 shows a linear increment of the relative power with respect to the discrete code length. One can
see that the slope of the relative power increment is near to
1, being almost the same with that in linear system although
the PSP-BOTDA system is based on optical nonlinear phenomenon. This result encourages us to study further.
However, it should be noted that the noise power also increases during the mutual correlations for the decoding. Assuming that the noise power distribution is Gaussian, the noise
in terms of optical power increases in proportion to
, for
constant
. Therefore, at constant
, the relative SNR
can be obtained by dividing the signal power
by noise power
. Thus, the relative SNRs calculated are
shown in Fig. 17. The data in Fig. 17 can be well approximated
as follows:
(45)
This improvement in the relative SNR is about twice that of
intensity-modulation simplex-code BOTDA [21] and is twice
that of intensity-modulation complementary-code BOTDA
[14]. The factor 2 comes from the difference in modulation
methods: phase shift modulation for the PSP-BOTDA and
intensity modulation for the conventional BOTDAs.
The rising edge was also investigated to evaluate the spatial
resolution of the system. The results are listed in Table II, which
shows the rising edge length of the 3 m strained fiber as the
representative. The rising edge length was obtained by transforming the time required for the Brillouin signal to rise from
10% to 90% of its final value.
The calculated length of around 8.2 cm is a little bit shorter
than the theoretical spatial resolution of 10 cm which is obtained
from (42). This is because we have used the 10% to 90% rise
time for the evaluation. From Table II, we can conclude that
the increment of the discrete code length does not deteriorate

3348

Fig. 18. Simulated correlation output of Brillouin amplified signals under


Condition B.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

Fig. 20. Simulated Brillouin amplified signal power versus code length
under Condition B.

Fig. 19. Magnified rising edge region of 3 m strained region of Fig. 18.

the spatial resolution of the Dual Golay PSP-BOTDA system.


Thus, we can expect that the code length of the discrete pulse
codes can be set longer than 32 bits to further improve the SNR.
C. Simulation Results (Condition B)
The Brillouin amplification results for each
under Condition B are shown in Fig. 18. As observed in the figure, the Brillouin signal increases with the increment of the code length of
. However, the amount of the increment starts to decrease
considerably when
is set to 16 and 32 bits, especially at
is set to 16
the 0.3 m strained region. Furthermore, when
and 32 bits, there are noises in the region around the rising and
the falling edges. The magnification of the 3 m strained region
of Fig. 18 is shown in Fig. 19, where the Brillouin signals for
and 32 exhibit overshoots and undershoots that
degrade the spatial resolution of the system. The results indicate the appearance of the side lobes of the processed correlations. This is because the total duration of the pulses modulated
with the continuous pulse code (16 ns and 32 ns) is larger than
the phonon lifetime of around 10 ns. Thus, only the first 8 bit
pulses strongly interact with the acoustic wave while the rest
does not during the measurement. Note that in contrast to the
Condition B, the Brillouin signals under Condition A illustrated
in Fig. 15 showed little degradation in the spatial resolution bewas fixed at the value of 4 ns that is shorter than
cause
the phonon lifetime throughout the simulations.
The results of the signal power increment and the relative
SNRs under Condition B are depicted in Figs. 20 and 21, respec-

Fig. 21. Simulated relative SNRs versus code length

under Condition B.

tively. As in the calculation for Condition A, the value


in Fig. 20 corresponded to the sum of two
sets of measurements with a 4-bit GCP while
was obtained by one set of measurements with the DGCP of
and
; the GCP has two sequences while
the DGCP four sequences; thus the measurement time was the
ranging from 1 to 32.
same among all measurements with
In Fig. 20, the relative signal power shows a linear increment
up to 8 bits. When
equals to 16 and 32
trend with
bits, the occurrence of positive and negative overshoots causes
the increment trend to deviate from its linear form.
The relative SNRs were calculated by dividing the relative
since
power increment shown in Fig. 20 by the value
the noise power accumulated by the correlation process is prounder Condition B. We found from Fig. 21
portional to
up to
that the relative SNRs increase with the increase in
8 bits for the 0.3 m strained fiber and up to 16 bits for other
strained fibers. However, the relative SNRs begin to decrease at
. This is because the random noise power, prothe longer
cessed in the correlations, accumulates with the increase in the
, while the increase rate of the signal graducode length
ally decreases as shown in Fig. 20. Here we note that the values
for
and 16 are 8 ns and 16 ns, which
of
ns). This implies that when
are around the phonon lifetime (
is around the phonon lifetime the relative SNR reaches
the maximum, which we found from Fig. 21 is less than the

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

TABLE III
RISING EDGE LENGTH OF 3 M STRAINED FIBER UNDER CONDITION B

3349

D. Main Points of Simulations and Brillouin Gain Spectrum


It could be concluded from the results shown in (45) and (46)
that the SNR enhancement (SNRE) for the dual Golay codes
PSP-BOTDA is given by
(47)

value
by about 1 dB. We have also confirmed by other
simulations (Fig. C.1 of Appendix C) that this is also true for
ns and
ns.
narrower pulse widths:
Fig. 21 also shows that the relative SNRs for the 0.3 m
strained fiber drops more steeply than those for other strained
longer than 8 bits. This discrepancy
fibers in the range of
is explained by difference in the crosstalk from the neighboring
parts of the fiber. The side lobes of the correlations explained
above cause the crosstalk that makes apparent increase in the
signal. The crosstalk for the 0.3 m strained fiber is less than that
for the 1 m, 3 m, and 10 m strained fibers since the range of the
neighboring parts of the former fiber is narrower than those of
the latter fibers. Thus, it apparently causes the higher relative
SNRs for the 1 m, 3 m and 10 m strained fibers.
From the discussions above and Appendix C, the relative
SNR under Condition B can be approximately expressed by
(46)
represents the ratio of the SNR
where coefficient
. The coeffiimprovement to the ideal improvement of
cient also accounts for the acoustic wave attenuation. Conis much
sequently, is about unity when code time
shorter than the phonon lifetime (9 ns); gradually decreases
, as shown in Fig. C.2 of
to zero with the increase in
which is
Appendix C. Thus, if we increase
, initially rises in proporgiven by the product of and
, then takes its maximum and after that gradually
tion to
decreases. Since is dependent on the accumulation of the amplitudes of the decreasing acoustic wave, is a function of code
. Numerical simulations show that the values of
time
are almost the same for
0.25, 0.5, and 1 ns, as
shown in Fig. 33. In addition, Fig. 32 shows that for all cases,
takes its maximum when
is around the
the
phonon lifetime of 9 ns.
The results of further analysis on the rising edge length in the
case of the 3 m strained fiber are listed in Table III.
Table III shows that the length of the rising edge which is estimated from the 10% and 90% response of the Brillouin signal.
to 8 bits are
In Table III, the rising edge lengths for
around 8.2 cm, which are the same with that on Condition A.
and 32 bits, as exHowever, it starts to increase at
plained above.
From the discussion above, we can conclude that the total duration of the pulses modulated with the continuous pulse code
is limited at the value smaller than the phonon lifetime: the numerical results of the rising edge length in Table III
ns indicate that
and of the relative SNRs in Fig. 21 for
the effective total duration of the modulated pulses is 8 ns or
below.

Since accounts for the attenuation of the acoustic wave excited


by the first pulsed pump for the coded continuous-PSP-BOTDA
and we assume that interval time is long enough to avoid interactions among the acoustic waves excited by the first pulses,
depends on
but does not on
. Therefore, if
is
varied, SNRE takes its maximum
when
is
around the phonon lifetime as it would for
of (46).
Then, considering that
as can be seen from Fig. C.2,
is approximately given by
(48)
is given by dividing the phonon lifetime by durawhere
tion .
Maximum values of
may be limited by the time taken
for the pulse light to make a round-trip across the length of
the fiber (RTT: round-trip-time). BOTDAs usually inject pulsed
pump light into fibers repeatedly to improve the SNR with signal
averaging techniques. The period of the repetition should be
more than the RTT so that the Stokes signal amplified by one
pulsed pump should not superimpose on that amplified by the
next pulsed pump. When the repetition period is reduced to the
RTT, the maximum improvement in SNR is obtained via the
signal averaging within a given measurement time. Then, the
total coded pulse time that is proportional to
should not
exceed the RTT so that the maximum improvement in SNR may
not be reduced. This is required in both conventional and proposed coded BOTDAs. However, the proposed coding method
can virtually increase the code length by nesting two codes as in
a manner described before with just a little increase in the repetition time.
From the discussion above and (48), we found that the
SNR of the dual Golay PSP-BOTDA is approximately
times as large as SNR of the PSP-BOTDA
with
, i.e., single-pulse PSP-BOTDA.
Furthermore, PSP-BOTDA has twice SNR compared to the
intensity modulation BOTDA as explained before. Consequently, the SNR enhancement of the dual Golay PSP-BOTDA
is approximately
times that of the conventional high
spatial resolution RZ coded BOTDAs. Here, it is noted that we
should choose
so that
is less than the phonon
lifetime. Then ranges from 0.7 to 1 as shown in Fig. 33.
Tables II and III demonstrated that the spatial resolution of
the dual Golay codes PSP-BOTDA is given by (42) as in the
conventional BOTDAs provided that the continuous pulse code
time
is around and less than the phonon lifetime.
The Brillouin gain spectra (BGS) of the dual Golay codes
PSP-BOTDA can be that of steady state SBS since the coded
continuous- and discrete-PSP-BOTDAs, which are the bases
of the dual Golay codes PSP-BOTDA, have approximately the
same BGS with the steady state SBS spectrum, as analytically
verified in Appendices A and B. The BGS for the dual Golay

3350

Fig. 22. Brillouin gain spectra at the BFS transient range calculated with
m,
, and
. The values of other parameters are as shown
in Table I. Inset shows the time domain signal around the transient range.

Fig. 23. Experimental setup of the dual Golay codes PSP-BOTDA.

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

second optical pulses are phase-coded. A polarization scrambler (PS) was used to hinder the polarization dependence of the
Brillouin amplification. The PS used in the experiment modulates the state of polarization (SOP) of light via modulation of
birefringence of the fiber inside it. We have ensured that the polarization-induced noise was suppressed by observing that the
noise was reduced to less than the white noise at the receiver
after averaging
signal waveforms. The modulated optical
signals were amplified using an erbium-doped fiber amplifier
(EDFA1) to set the peak power of the first and the second optical
pulses to be at 33 mW. The widths of the first and the second
optical pulses are 30 ns and 1 ns, respectively. The rise time of
the pulsed light is around 0.5 ns. The test fiber cables used in
the experiment were the dispersion shift fibers (DSF) with the
lengths of 0.3, 1, 3, and 10 m. Single mode fiber (SMF) cables
with the length of 10 m were connected between the DSF cables
as the references. The difference in BFS between the SMF and
the DSF cables is about 300 MHz, which corresponds to about
0.6% strain.
As for producing the probe lightwave, the electrical RF signal
with a frequency of 10.555 GHz from an RF oscillator (RF OSC)
was input to an optical single side-band modulator (SSBM). The
SSBM output the Stokes light (probe), whose frequency was
downshifted with the amount of 10.555 GHz. EDFA2 was used
to amplify the probe light to set the probe power to 2 mW. An
optical spectrum analyzer (OSA) was used to monitor the modulated light frequency. The Brillouin amplified signal was converted into an electrical signal and sent to the oscilloscope. The
signals were digitized and averaged by the oscilloscope. Then
the recorded signals were further processed by a computer for
the decoding (mutual correlation calculations).
B. Experimental Results (Condition A)

codes PSP-BOTDA with


bits and
bits
has been also evaluated by numerical simulations based on (44).
The simulated fiber had a 3 m section of 300 MHz change in the
BFS. Fig. 22 shows the BGS calculated at three points I, II, and
III in the BFS transition range. We found here that all BGS for
the 3 m section and for the reference are approximately the same
with the steady state BGS having gain widths as narrow as 35
MHz, demonstrating that the proposed coding method does not
induce distortions in the measured BGS.
VIII. DUAL GOLAY CODES PSP-BOTDA EXPERIMENTS
A. Experimental Setup
The experiments on the dual Golay PSP-BOTDA were
performed to be compared with the simulations to validate the
sensor systems efficiency. The experimental setup of the dual
Golay PSP-BOTDA system is shown in Fig. 23.
The light from an LD operating at a wavelength of 1550 nm
and at power of 6 mW was divided by a 1 2 optical coupler
(50:50) into one for the pump and the other for the probe. An
arbitrary waveform generator (AWG) was used to synthesize
the electrical signal which was composed of an electrical first
pulse and the dual Golay code pulses (second pulses). The lightwave for the pump was modulated by a Mach-Zehnder modulator (MZM) with the electrical signal from the AWG. Then, the
MZM was biased at the transmission minimum point so that the

The experiments under Condition A were performed for various values of


. The experimental results are depicted in
Fig. 24(a) and (b). The results in Fig. 24 are in good agreement
with the simulation shown in Fig. 15, although random noise
is superimposed on the experimental signals. We can observe
that the Brillouin signal peak power is almost the same for all
strained regions. Furthermore, as expected, little deterioration
can be found in the rising and the falling edges of the signal
waveforms even in the case of 32 bits of the discrete pulse code.
By using the same approach of calculating the theoretical rising
edge length explained in Chapter VII, we further analyzed the
experimental spatial resolution. The rising edge length evaluated at the 3 m strained region was 8.5 cm for all values of
, and 32 bits, being in good agreement with
the theoretical value of
cm.
The SNR enhancement (SNRE) measured for

bits and
bits is shown in Fig. 25. Notice that the
SNREs were obtained by using the SNR for
bit as a reference. This reference SNR corresponded to the SNR
for the single pulse PSP-BOTDA measured in the same measurement time. The reference SNR obtained from the experiment was 4.7 dB. The straight line (
with
) was also included in Fig. 25 for comparisons.
We observe that the SNRE increases linearly with respect to
. When
, we obtained SNRE as large as about

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

Fig. 24. Measured correlation output of Brillouin amplified signals under Con1, 4, and 16. (b)
2, 8, and 32.
dition A. (a)

3351

Fig. 26. Measured correlation output of Brillouin amplified signals under Con1, 4, and 16. (b)
2, 8, and 32.
dition B. (a)

DC-coupled MZM would improve the modulation characteristics and increase the SNRE in accordance with (45).
C. Experimental Results (Condition B)

Fig. 25. Measured SNR enhancement of dual Golay PSP-BOTDA referenced


to the single pulse-PSP-BOTDA SNR( 4.7 dB).

7 dB. However, when


and 32 bits, the SNRE
starts to deviate from its linearity. This slightly differs from the
simulations shown in Fig. 17. This discrepancy was due to the
modulation characteristics of the MZM used in this experiment.
The capacitor incorporated in the driving circuits of the MZM
cut off the low-frequency components of the modulated signals. This resulted in the inefficient modulation of the discretely
configured coded-pulses of more than 8-bit code. The use of a

Figs. 26(a) and (b) show the experimental results under Condition B. It can be found from the figures that the increase in
also contributes to the Brillouin power increment of the
system. However, as predicted, the increase rate of the Brillouin
amplified signal has dropped significantly when
and
32 at the 0.3 m strained region. This is because total time duration of those code length pulses has exceeded the phonon lifetime, as explained in Section VII.C. In addition, as predicted
from the simulation in Fig. 18, the overshoot and undershoot
noises can be seen clearly when
equals to 16 and 32 bits.
The SNREs for
bits and
bits are depicted in Fig. 27. The SNREs were obtained by using the SNR
for
bit as a reference, as in the case before.
As in Fig. 25, the reference SNR corresponded to the SNR for
the single pulse PSP-BOTDA measured in the same measurement time. The reference SNR obtained from the experiment
was 4.7 dB. The straight line (
with
) was also included in Fig. 27 for comparisons.
The SNRE for the strained lengths of 1 m, 3 m, and 10 m
increases with
up to 16 bits, while it begins to decrease
at 16 bits in the case of 0.3 m strained length. It is noticeable

3352

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

Fig. 27. Measured SNR enhancement of dual Golay PSP-BOTDA referenced


to the single pulse-PSP-BOTDA SNR( 4.7 dB).

that the results in Fig. 27 agree well with the simulated relative
SNR shown in Fig. 21 if we take account of the coding gain
with
bits. For 0.3 m test section, when
,
maximum SNRE of about 7 dB was obtained, while for other
test sections, when
, maximum SNRE was about
8 dB. As mentioned in Section VII.C, acoustic wave damping
explains the reason of the deviation from the straight line of the
SNRE, i.e., the drop in the SNRE for
longer than 8 bits.
Since the SNRE for all test sections increases with
up
to 8 bits under Condition B, we measured the BGS by using the
dual Golay code of
bits and
bits. The BGS
measured at the 3 m test section and at the reference fiber are
shown in Fig. 28(a) and (b). Both spectra measured were fitted
well with Lorentzian curves of the bandwidth as narrow as about
30 MHz, indicating the BGS obtained by the proposed coding
method were approximately the same with the steady state BGS.
Therefore, we confirmed that the use of NRZ coded pulses in the
proposed method does not induce distortions in the BGS.
The spatial resolution was also evaluated at the rising edge
of 3 m strained region. The result showed that the rising edge
length was 8.5 cm for
, and 8 bits. When
was set to 16 and 32 bits, the rising edges range over 62 cm
and 160 cm, respectively. We have confirmed from these experiments that we can obtain good SNRE and achieve a high spatial
resolution on condition that
is set less than the phonon
lifetime. Otherwise, however, not only the SNRE but also the
spatial resolution deteriorates.
To fully analyze the change in the BFS and verify the spatial resolution, we also measured the BFS distribution. Fig. 29
shows the BFS distribution measured when
and
, which agreed well with the pseudo-BFS-distribution made by splicing SMFs and DSFs alternatively. The transient BFS is shown in an inset (a) in Fig. 29, showing the transient length is about 10 cm, which is in good agreement with
the spatial resolution calculated from (42). We can also see in
the inset (b) in Fig. 29 that the variation in the measured BFS in
the 3-m section was estimated at 0.67 MHz (standard deviation)
that corresponded to the measurement accuracy of
strain or 0.56 C temperature. Therefore, it is confirmed from
the BFS distribution measurement that the high spatial resolution as short as 10 cm has been fully demonstrated.

Fig. 28. Measured Brillouin gain spectrum under Condition B with


bits. (a) 3 m test section (b) 10 m reference fiber.

Fig. 29. BFS distribution under Condition B with


bits. Insets (a) and
(b) show expanded trace of the rising edge and distribution at the 3 m section.

IX. CONCLUSION
An algorithm has been presented for synthesizing a dual
Golay complementary pair of sequences (DGCP) for measuring
distributed strain and temperature based on Phase Shift Pulse
Brillouin Optical Time Domain Analysis (PSP-BOTDA) with
a high spatial resolution. The proposed DGCP is configured by
nesting one Golay complementary pair of sequences (GCP) into
the other one. We call the former pair of sequences continuous

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

pulse codes and the latter discrete pulse codes, which are used
to generate the pulsed pump with NRZ-pulses and RZ-pulses,
respectively.
Introducing a proper space time between the coded discrete
pulses avoids the interference noise through the SBS process.
However, the introduction of the space time increases the time
slot for one bit, decreasing the code length that is limited by the
RTT.
Constructing the coded continuous pulses with the total duration less than the phonon lifetime makes it possible to obtain
good correlation properties of the Brillouin signals. However,
the maximum code length is limited.
The proposed DGCP enables us to use the RZ-pulses and
NRZ-pulses simultaneously and advantageously for coding
the pump of the PSP-BOTDA, and to use longer codes than
the conventional codes within a given period; let code lengths
of the continuous and the discrete pulse codes be
and
, the code length of each dual Golay code in the DGCP
is
. Then, the SNR enhancement, i.e., the coding
gain of the proposed method is about
, while that
of the conventional coded BOTDA based on intensity modulations is
. Therefore, the SNR enhancement of the
proposed coding method is
times as large as that of
the conventional coding method. The coefficient
represents the efficiency of the SNR improvement with the
continuous pulse codes and also accounts for the attenuation of
the acoustic wave that is excited by the first pulsed pump of the
PSP-BOTDA. When the product of
and the second pulse
duration
is larger than the phonon lifetime, the quantity
decreases to less than 0.7 and the spatial resolution get worse
than the theoretical one; thus
should be less than the
phonon lifetime; then ranges from 0.7 to 1.
A new coded PSP-BOTDA system based on the DGC has
been demonstrated by both simulations and experiments. From
the results, we have clarified that the power of the correlated
Brillouin signal has increased with increase in the code length
of the new code provided that the total duration of the pulses
coded by the continuous pulse codes is less than the phonon
lifetime. The increase in the Brillouin signal power raises the
signal-to-noise ratio and is to BOTDAs benefit to attain high
spatial resolution. Initial experiments for
have demonstrated about 7-dB enhancement in the optical
signal-to-noise ratios when compared to the single pulse PSPBOTDA. The same amount of enhancement was also achieved
when
. We also have attained a spatial
resolution of 10 cm.
In this paper, we have proposed the combination of two GCPs
for high spatial resolution BOTDA. It should be noted, however,
that the combination of the codes is not limited to the complementary codes. Other types of codes which include Barker codes
and Hadamard codes can be combined in a similar manner presented in this paper.
Finally, we note here that in this study only short length of
fiber of about 70 m was evaluated with the use of a few tens of
mW in pump peak power. Further work is needed to improve
the performance of BOTDA by extending the fiber length and
by raising the pump power. However, it is known that the pump

3353

depletion effects or nonlocal effects would tend to limit the performance for longer fiber and for higher input optical power.
We believe that the solutions of these problems for conventional
coded BOTDAs, which include a method to use both Stokes
and anti-Stokes lights simultaneously as cw probe [23], [24],
and a method to employ Frequency-Division Multiplexing [25],
would also work well for the BOTDA employing the coding
method proposed.
APPENDIX A
In the following, we give the signal for the coded continuous-PSP-BOTDA which is obtained by subtracting the signal
measured with a code from the signal
measured
with a code , where
and
or
, with
.
We note here that all parameters of analytical expressions in this
Appendix are normalized according to [8].
Let
represent an increment in the Stokes light field
produced by SBS after passage of a time t from the time when
the pulsed pump enters into a local region along an optical fiber.
When
, the field
for
and that for
are the same. Then, the subtracted signal becomes zero. Therefore we have only to derive the signal
for
,
where
is the first pump
pulse width and is the code element pump pulse width. When
, the Stokes light field increment
is obtained from (44b) as follows:
(A1)
The factor 1/2 in (A.1) comes from the fact that the Stokes light
propagates in the direction of
while the pump light in the
opposite.
The acoustic wave field
in (A.1) is obtained from (44c)
as follows:
(A2)
where denotes the normalized frequency dewith
tuning. If we express the pulsed pump field as
and assume
is constant, (A2) is transformed into
(A3)
with
(A4)
(A5)
Here we note that is due to the acoustic wave excited by the
first pulsed pump, while
is produced by the sequence of
the coded pulsed pump, thus being dependent on the bit pattern
of the code.

3354

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

The Stokes light power increment


is given by
(A6)
where c.c denotes complex conjugate. For simplicity, the factor
of proportionality in the right side of (A6) was set to unity. Substituting (A1) and (A2) into (A6), we obtain the Stokes light
power increment
when the code is employed as

(A7)
On the other hand, when the code
is used, in (A7) and
in
are replaced with
and
, thus yielding
the Stokes light power increment as:

(A8)
Consequently, the differential Stokes light power is given by

(A9)
Here, we find that the subtraction above exactly cancels the
bit pattern dependent term
.
In contrast, when the intensity modulation format is employed for coding the pump as in the conventional intensity
modulation based BOTDA, codes
and
instead of codes and
are used for measuring
and
, respectively. Then, the bit pattern dependent term does
not vanish through the subtraction. This is the reason why the
coded PSP-BOTDA has better correlation characteristics than
the conventional intensity modulation based BOTDAs when
the NRZ coded pulses are used.
Substituting (A4) into (A9) yields the explicit expression of
the differential signal of the coded continuous-PSP-BOTDA:

Fig. 30. Numerical results of BGS at the test section of the fiber for coded
continuous-PSP-BOTDA.

35 MHz in the case of 1 bit to 16 bits, the spectrum width has


narrowed to about 30 MHz in the case of 32 bits. However, there
is little distortion in the spectrum shape.
APPENDIX B
In the following, we give the signal for the coded discretePSP-BOTDA. As can be seen from Figs. 1(a) and (b), when
the code length is one bit, the pump configuration of the coded
discrete-PSP-BOTDA is the same with that of the coded continuous-PSP-BOTDA. Therefore, one-bit signal of the coded discrete-PSP-BOTDA is given by (A.10) with
as follows:

(B1)
If we assume typical conditions for measurement with high
spatial resolution BOTDA: the one-bit pulse width is far less
), the first pump pulse
than the phonon lifetime (i.e.,
width is much greater than the phonon lifetime (i.e.,
), and the frequency detuning is within the Brillouin gain
bandwidth (i.e.,
), then (B1) becomes
(B2)

(A10)
We find from (A10) that the spectrum profile of the differential signal is almost the same with that of the steady state, that
is equivalent to what is given at the first row of the right side of
(A10). We also find that the differential signal decreases with the
order of the code element in accordance to
where
time is normalized by phonon lifetime. Thus the maximum code
length in duration is limited up to about the phonon lifetime.
The spectra of the differential signals also have been numerically evaluated based on (46) and shown in Fig. 30. Then the
calculation condition was similar to that provided in Table I. It
can be observed that the peak Brillouin power increases linearly
with code length up to 8 bits. However, it starts to deviate from
its linearity. The reason for this to occur is as what has been explained in Section VII.C. While the spectrum width stays around

The factor 2 in (B2) comes from two measurements for the


subtraction process. Therefore, it is found from (B2) with eliminating the factor 2 that the Brillouin gain spectrum of the coded
discrete-PSP-BOTDA for one bit is identical to that in the steady
state in terms of both magnitude and bandwidth.
If we configure the pulsed pump of the coded discrete-PSPBOTDA as shown in Fig. 1(b) with separation much longer
than the phonon lifetime, each pulsed pump for one bit does
not interfere with the others. Then, the bit patterning effects are
reduced and the Brillouin signal for the k-th bit is given by the
right side of (B.2) with replacing by .
The spectra of the differential signals are evaluated numerically as in a manner of Appendix A, and are illustrated in Fig. 31.
The Brillouin power for each code length is normalized to the
peak Brillouin power for 1-bit code. It is observed that the power

ZAN AND HORIGUCHI: DGCP FOR IMPROVING PERFORMANCE OF PHASE-SHIFT PULSE BOTDA FIBER SENSOR

3355

code lengths
were set up to 64 bits for
ns and
up to 128 bits for
ns. Fig. 32 shows
as a
function of code time
for the 0.3 m section. It can be
seen from Fig. 32 that for all cases,
takes maxima
when
is around the phonon lifetime of 9 ns.
Fig. 33 shows coefficient
that is obtained by dividing
shown in Fig. 32 by
according to its
definition given by (46). We found from the figure that code
time dependence of is almost the same for different pulse
durations of
and 1 ns. We also found that when
is maximum, that is when
ns or 9
ns (phonon lifetime) according to Fig. 32, takes the value of
about 0.8 or 0.7 even for different
values.
Fig. 31. Numerical results of BGS at the test section of the fiber for coded
discrete-PSP-BOTDA.

ACKNOWLEDGMENT
The authors would like to thank D. Uchiyama and T. Sasaki
for useful discussions and experimental help.
REFERENCES

Fig. 32. Relative SNR versus code time

Fig. 33. Coefficient

vs code time

for different

for different

increases linearly with the increase in the code length. Also, as


expected, the shape of the spectra was not distorted and their
widths were found to stay about 35 MHz; as narrow as the value
for the steady state SBS.
APPENDIX C
In the following, we give the relative
and the coefficient in (46) which are numerically evaluated based on (44)
for various
of 0.25, 0.5 and 1 ns. The numerical calculations
were performed based on Condition B of Table I except that the

[1] T. Horiguchi, K. Shimizu, T. Kurashima, M. Tateda, and Y. Koyamada,


Development of a distributed sensing technique using Brillouin scattering, J. Lightw. Technol., vol. 13, no. 7, pp. 12961302, Jul. 1995.
[2] T. Horiguchi, T. Kurashima, and Y. Koyamada, 1 m spatial resolution
measurement of distributed Brillouin frequency shift in single-mode
fibers, in Tech. Dig. OFM, 1994, pp. 7376.
[3] K. Hotate and T. Hasegawa, Measurement of Brillouin gain spectrum distribution along an optical fiber using correlation-based techniqueProposal, experiment and simulation, IEICE Trans. Electron., vol. E83-C, no. 3, pp. 405412, Mar. 2000.
[4] C. H. Li, K. Nishiguchi, M. Miyatake, A. Makita, M. Yokoyama, K.
Kishida, T. Mizutani, and N. Takeda, PPP-BOTDA method to achieve
2 cm spatial resolution in Brillouin distributed measuring technique
Tech. Rep. IEICE. OFT 108(32), May 2008, pp. 5560, (in Japanese).
[5] K. Y. Song, Z. He, and K. Hotate, Distributed strain measurement
with millimeter-order spatial resolution based on Brillouin optical correlation domain analysis, Opt. Lett., vol. 31, no. 17, pp. 25262528,
Sept. 2006.
[6] T. Yamauchi and K. Hotate, Distributed and dynamic strain measurement by BOCDA with time-division pump-probe generation scheme,
in Proc. Conf. Lasers and Electro-Optics, San Francisco, CA, May
2004.
[7] X. Bao and A. Brown, Characterization of the Brillouin-loss spectrum
-ns) pulses, Opt. Lett.,
of single-mode fibers by use of very short (
vol. 24, no. 8, pp. 510512, Apr. 1999.
[8] V. Lecoeuche, D. J. Webb, C. N. Pannell, and D. A. Jackson, Transient
response in high-resolution Brillouin-based distributed sensing using
probe pulses shorter than the acoustic relaxation time, Opt. Lett., vol.
25, no. 3, pp. 156158, Feb. 2000.
[9] S. V. Ashfar, G. A. Ferrier, X. Bao, and L. Chen, Effect of the finite
extinction ratio of an electro-optic modulator on the performance of
distributed probe-pump Brillouin sensor system, Opt. Lett., vol. 28,
no. 16, pp. 14181420, Aug. 2003.
[10] S. Foaleng-Mafang, J.-C. Beugnot, and L. Thevenaz, Optimized
configuration for high resolution distributed sensing using Brillouin
echoes, in Proc. 20th Int. Conf. Opt. Fibre Sensors, Edinburgh, UK,
Oct. 2009, pp. 75032C175032C4.
[11] W. Li, X. Bao, Y. Li, and L. Chen, Differential pulse-width pair
BOTDA for high spatial resolution sensing, Opt. Exp., vol. 16, no.
26, pp. 2161621625, Dec. 2008.
[12] Y. Li, X. Bao, Y. Dong, and L. Chen, A novel distributed Brillouin
sensor based on optical differential parametric amplification, J.
Lightw. Technol., vol. 28, no. 18, pp. 26212626, Sep. 2010.
[13] T. Horiguchi, R. Muroi, A. Iwasaka, K. Wakao, and Y. Miyamoto,
Negative Brillouin gain and its application to distributed fiber sensing
(presented poster style), in Proc. 33rd Eur. Conf. Exh. Opt. Commun.,
Sep. 2007, P018.
[14] H. Liang, W. Li, N. Linze, L. Chen, and X. Bao, High-resolution DPPBOTDA over 50 km LEAF using return-to-zero coded pulses, Opt.
Lett., vol. 35, no. 10, pp. 15031505, May 2010.

3356

[15] R. Muroi, T. Horiguchi, Y. Miyamoto, Y. Sato, A. Tachibana, and


A. Takakura, PSP-BOTDA based on Golay code, IEICE Trans.
Commun., vol. J91-B, no. 11, pp. 14931501, Nov. 2008, (in Japanese).
[16] D. Uchiyama, T. Horiguchi, H. Ando, Y. Okumoto, T. Sasaki, and Y.
Sawai, Signal to Noise Improvement Ratio for Coded PSP-BOTDA
Tech. Rep. IEICE. OFT 109 (377), Jan. 2010, pp. 3338, (in Japanese).
[17] M. J. E. Golay, Complementary series, IRE Trans. Inf. Theory, vol.
IT-7, pp. 8287, Apr. 1961.
[18] M. S. D. B. Zan, T. Sasaki, T. Horiguchi, and D. Uchiyama, Phase
shift pulse Brillouin optical time domain analysis (PSP-BOTDA) employing dual Golay codes, in Proc. Int. Conf. Comput. Commun. Eng.,
Kuala Lumpur, Malaysia, May 2010, pp. 16.
[19] S. E. El-Khamy, On generation and measurement of combined Barker
codes, Int. J. Electron., vol. 50, no. 6, pp. 429436, 1981.
[20] T. Sasaki, T. Horiguchi, M. S. D. Zan, and D. Uchiyama, Coded
pulse interval dependence of the characteristics in discretely coded
PSP-BOTDA, IEICE Trans. Commun., vol. J94-B, no. 11, pp.
14811489, Nov. 2011, (in Japanese).
[21] M. A. Soto, G. Bolognini, and F. Di. Pasquale, Analysis of pulse modulation format in coded BOTDA sensors, Opt. Exp., vol. 18, no. 14,
pp. 1487814892, Jul. 2010.
[22] R. W. Boyd, Nonlinear Optics, 3rd ed. New York: Academic, 2008.
[23] A. Minardo, R. Bernini, and L. Zeni, A simple technique for reducing
pump depletion in long-range distributed Brillouin fiber sensors, IEEE
Sens. J., vol. 9, no. 6, pp. 633634, Jun. 2009.
[24] M. A. Soto, G. Bolognini, F. D. Pasquale, and L. Thevenaz, Simplex
coded BOTDA fiber sensor with 1 m spatial resolution over a 50 km
range, Opt. Lett., vol. 35, no. 2, pp. 259261, Jan. 2010.
[25] Y. Dong, L. Chen, and X. Bao, Extending the sensing range of Brillouin optical time-domain analysis combining frequency-division multiplexing and in-line EDFAs, J. Lightw. Technol., vol. 30, no. 8, pp.
11611167, Apr. 2012.
[26] T. Horiguchi, R. Muroi, A. Iwasaka, K. Wakao, and Y. Miyamoto,
BOTDA utilizing phase-shift pulse, IEICE Trans. Commun., vol.
J91-B, no. 2, pp. 207216, Feb. 2008, (in Japanese).

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 21, NOVEMBER 1, 2012

Mohd Saiful Dzulkefly Bin Zan was born in Malaysia on December 5, 1983.
He received the B.E. degree from Waseda University, Tokyo, Japan, in 2006,
and the M.E. degree from Shibaura Institute of Technology, Tokyo, Japan, in
2011, where he is currently working toward the Ph.D. degree.
He was a Product Engineer with Toshiba Electronics Malaysia (TEM) in
2006. Since 2007, he has been a Tutor with the Department of Electrical, Electronics and System Engineering, Universiti Kebangsaan Malaysia (UKM). His
doctoral work is focused on the Brillouin-based sensor.
Mr. Bin Zan is a student member of the Institute of Electronics, Information
and Communication Engineers (IEICE) of Japan.

Tsuneo Horiguchi (M87SM96) was born in Tokyo, Japan, on June 5, 1953.


He received the B.E. and Dr.Eng. degrees from the University of Tokyo, Tokyo,
Japan, in 1976 and 1988, respectively.
In 1976, he joined Ibaraki Electrical Communication Laboratories, NTT,
Ibaraki, Japan, where he engaged in research and development of measurement
technique and measuring instrument for evaluating the transmission characteristics of optical fiber cables. Since 1988, he has been also involved in the
field of optical fiber distributed sensing. From 1999 to 2001, he was the Project
Manager of the Advanced Transmission Media Project of NTT Access Service
Systems Laboratories. Since April 2002, he has been a Professor with the
Department of Electrical Communication, Shibaura Institute of Technology,
Tokyo, Japan. His primary research interests include optical distributed sensing
based on nonlinear effects and intelligent signal processing for optical sensors
and optical communications.
Prof. Horiguchi is a Member of the Institute of Electronics, Information and
Communication Engineers (IEICE) of Japan and the Optical Society of Japan
(OSJ).