A Robust Serial FBG Sensor Network with CDM Interrogation Allowing

Overlapping Spectra

Marek G¨otten

1,2

, Steffen Lochmann

, Andreas Ahrens

1 b

and C´esar Benavente-Peces

2 c

Bereich Elektrotechnik und Informatik, Hochschule Wismar, Phillip-M¨uller-Straße 14, Wismar, Germany

Escuela T´ecnica Superior de Ingenier´ıa y Sistemas de Telecomunicaci´on, Universidad Polit´ecnica de Madrid,

Crtra de Valenica, km 7, Madrid, Spain

Keywords:

Code-Division Multiplex, Fiber-Bragg-Gratings, Smart Structures, Serial Optical Sensor Networks, Optical

Autocorrelation.

Abstract:

Massive optical sensor networks gained a lot of attention in recent years. They offer new advances in the ﬁelds

of smart structures and health monitoring. All serial optical sensor networks rely on multiplexing techniques

that provide huge amounts of sensors in a single optical ﬁber. Wavelength-division multiplex (WDM) which

has been established in many applications, is restricted to the spectral width of the used light source that

needs to be shared by several non-overlapping ﬁber-Bragg-grating (FBG) spectra. Time-division multiplex

(TDM) uses short impulses and relies on different sensor round trip delays to distinguish each single FBG.

These short impulses and long round trip times lead to a low signal-to-noise ratio (SNR). Optical frequency-

domain reﬂectometry (OFDR) offers a high spatial resolution of FBGs but only within a short ﬁber length.

This contribution deals with a code-division multiplex (CDM) interrogation technique that provides numerous

sensors in a single optical ﬁber, a better SNR, and a long range of distributed sensing points. It requires

codes with good autocorrelation behavior which is characterized by certain criteria. The detectable criteria

are limited which narrows signiﬁcantly a search for best possible codes for the interrogation system. In this

contribution, practical implementation limits such as the trigger timing and the achievable SNR are studied.

Based on the introduced SNR deﬁnitions for CDM and WDM systems, a direct comparison is possible and

it shows the superiority of the proposed CDM scheme. A network with 25 sensors operating at the same

wavelength can provide a 2.67dB improvement compared to WDM

1 INTRODUCTION

In the ﬁeld of sensor technology, optical sensors,

such as FBGs, gain a lot a attention. Their prop-

erties of being light, small, immune to electromag-

netic interferences and capable of multiplexing, make

them the preferred choice for many sensing appli-

cations (Rajan, 2015). Especially the multiplexing

capability is in the focus of interest for applications

like health monitoring (Gliˇsi´c, 2016; Nawrot et al.,

2017) and smart-structures, where optical ﬁber sen-

sor networks can be integrated into materials (Braghin

et al., 2013; Kim, 2004). Popular multiplexing tech-

niques are WDM (Kersey et al., 1997), TDM (Wang

et al., 2012) or optical frequency-domain reﬂectom-

etry (OFDR) (Yamaguchi et al., 2015). This contri-

https://orcid.org/0000-0002-0938-2186

https://orcid.org/0000-0002-7664-9450

https://orcid.org/0000-0002-2734-890X

bution makes use of CDM that overcomes the restric-

tions of other multiplexing techniques. Unlike WDM,

the optical sensors can spectrally overlap within the

spectrum of the light source. Compared to TDM that

makes use of a single short light impulse, a CDM

interrogator collects several light impulses forming

a code which signiﬁcantly improves the SNR of the

system. While OFDR works within short link net-

works (Yamaguchi et al., 2015), CDM has the po-

tential of covering increased ﬁber lengths. The CDM

technique requires codes with a good autocorrelation

behavior (Abbenseth et al., 2016). The proposed in-

terrogator for serial ﬁber optical sensor networks con-

sisting of sensors operating at the same wavelength

relies on the principle of CDM. In this contribution,

the network can consist of several FBG sensors that

partly reﬂect incident light according to their grat-

ing parameters, applied strain or temperature changes.

In the following, the explanation of the interrogation

Götten, M., Lochmann, S., Ahrens, A. and Benavente-Peces, C.

A Robust Serial FBG Sensor Network with CDM Interrogation Allowing Overlapping Spectra.

DOI: 10.5220/0008942900230028

In Proceedings of the 9th International Conference on Sensor Networks (SENSORNETS 2020), pages 23-28

ISBN: 978-989-758-403-9; ISSN: 2184-4380

principle is shown with the mentioned FBGs. Nev-

ertheless, the principle is not limited to FBGs but

rather works with every reﬂecting ﬁber optical sen-

sor. CDM or code-division multiple access (CDMA)

is already well known in the ﬁeld of telecommuni-

cation, as the works by (Lee and Miller, 1998) or

(Kitayama, 1998) indicate. Transferring this tech-

nique to sensor networks opens a new dimension of

size considering spectral overlapping of sensors. In

WDM systems, the spectral distribution of FBGs is

limited to the spectral width of the light source and

the operating range of each FBG that must not over-

lap with neighbouring FBGs during a measurement.

(Abbenseth et al., 2016) shows that CDM in opti-

cal sensor networks allows spectral overlapping with-

out losing track of the measured sensor. Against this

background the novelty of this contribution focuses

on practical implementation limits such as the trigger

timing and the achievable SNR. Based on the intro-

duced SNR deﬁnitions for CDM and WDM systems a

direct comparison is possible and it shows the supe-

riority of the proposed CDM scheme. The remaining

part of the paper is structured as follows: Section 2 ex-

plains the CDM interrogation scheme. Section 3 dis-

cusses the different parameters inﬂuencing the over-

all performance. At the end, Section 4 concludes the

work.

2 INTERROGATION PRINCIPLE

2.1 Interrogator Setup

Fig. 1 depicts the scheme of a CDM interrogator for

serial ﬁber optical sensor networks. A light source

provides the light that is launched into the sensor net-

work and reﬂected by the FBGs. The spectral band-

width should cover at least the operating region of

each sensor in the network. The light source emits

Light Source

Modulator

Modulator Modulator

Spectrometer Spectrometer

−

Differential Spectrum

Code

∆t

INV

Figure 1: Scheme of CDM interrogation system.

continouswave (CW) light that needs to be modulated

for a CDM interrogation. The modulator is driven

with code that is suitable for CDM; further details

can be found in section 2.2. The modulated light is

launched into the network with FBGs operating at the

same wavelength λ

. The reﬂected light travels back

and reaches the detector paths. It is split into two

equal parts that are each modulated with a code again.

This code is a time delayed version of the previous

used code for the left path, the direct path, and for the

right path, the inverted path, the code is time shifted

and inverted. After the second modulation the light

in each path is collected by a spectrometer, which

performs the integration part of the correlation pro-

cess. The inverted spectrum is subtracted from the

direct spectrum which equals a differential spectrum

that is used for peak detection. Further details about

the setup are described by (Abbenseth et al., 2016)

and (G¨otten et al., 2018).

2.2 CDM-interrogation Scheme

Having a deeper look into the maths of the interroga-

tor, the scheme can be reduced to Figure 2. Light

source (LS) and Modulator provide unipolar modu-

lated light. The two detection paths contain each a

modulator that works as a multiplication of the codes.

Only, if a logical ’1’ reaches the modulator and it let’s

light pass (also a logical ’1’), light reaches the sum,

that represents the spectrometer collecting incoming

light. Multiplication and summation equals mathe-

matically a correlation; a unipolar - unipolar auto-

correlation of the used code. Unipolar signal cod-

ing is the consequence of amplitude modulators ap-

plying simple on-off-keying of light. Introducing the

inverted path, where the autocorrelation is performed

with the inverted code, and calculating the difference,

the result is a unipolar-bipolar autocorrelation which

is explained in the work of (G¨otten et al., 2019) on the

example of this interrogation setup.

Unipolar-bipolar correlations result in a scaled ver-

sion of bipolar-bipolar correlations, as shown by

(O'Farrell and Lochmann, 1994). The advantage is

the ability of the autocorrelation function (ACF) to

have low sidelobes. The products of the unipolar-

Mod S

Mod R

−

∑

Figure 2: Principle of SIK.

SENSORNETS 2020 - 9th International Conference on Sensor Networks

unipolar correlation that are summed up can only be

0 or 1. The summed up products of bipolar-bipolar

correlations can be -1 or 1. Therefore, the sum rep-

resented by each sidelobe can be close to zero if the

mixture of -1 and 1 products is more or less equally

distributed. In the unipolar-unipolar sum, the prod-

ucts can only add up, as no -1 product appears.

The ACF has the highest value in the autocorrelation

peak for a time shift τ = 0. Considering the value of

the ACF as amount of light in the differential spec-

trum, the code reﬂected by an FBG that reaches the

modulator exactlysynchronousis the majority of light

after calculating the difference. Assuming an ideal

synchronization, in the direct path the whole light can

pass the modulator and in the inverted path no light

passes the modulator. Therefore, the difference be-

tween the sum of all light impulses of the code and

no light results in a high value, mathematically the

amount of chips of the code. Other reﬂected codes

reach the detection path asynchronous to the modula-

tors and therefore, not all light can pass or is blocked.

Ideally, the amount of light passing each modulator is

equal, so the difference results in zero. Depending on

the code, the asynchronous reﬂected codes by other

FBGs do not appear in the differential spectrum. This

is achieved by a code that provides an ACF with very

low sidelobes, ideally equal to zero, hence orthogonal

or quasi-orthogonal codes.

2.3 Improvements of Interrogator Setup

The two detecting paths have to be identical to obtain

a correct unipolar-bipolar correlation. Setting up the

interrogator in a testbed requires identical equipment

for both paths. To overcome this difﬁculty, the de-

tection path is realized in a serial manner with only

ASE Source

SMF Modulator S FBGs

Pol. Controller

Modulator R

SMF

Spectrometer

∆t

Pattern Generator

Trigger

CH0

CH1

Figure 3: Scheme of a serial CDM interrogation system.

one modulator and one spectrometer. Direct and in-

verted path are now performed sequentially, where

each measurement is repeated two times with direct

and inverted code (Figure 3).

An ampliﬁed spontaneous emission (ASE) broadband

light source provides the light that covers the full op-

erating range of the FBGs in the network. The mod-

ulator S is driven with a code from a pattern gen-

erator. A circulator is responsible to guide direct

light into the network and out of it into the detec-

tion path. The code is reﬂected by each FBG and

has a different time delay introduced by different lo-

cations in the serial network. After the circulator, the

light passes a polarization controller, the modulator R

and reaches the spectrometer, which is a charge cou-

pled device (CCD).The state of polarization (SOP) of

the reﬂected light is unknown while the modulator is

polarization dependent. Therefore, the polarization

controller helps to improve the modulation result of

the modulator. Modulator R is driven with the time

shifted code. The time delay ∆t matches the time light

takes to travel from the modulator S via a distinct sen-

sor to modulator R. Then the code is synchronized

to that particular sensor. The pattern generator stores

the direct and inverted code and is triggered by the

spectrometer. The spectrometer emits a trigger sig-

nal when the integration time starts. Then the pat-

tern generator emits through channel 0 the code and

through channel 1 the time shifted code. The code

is repeated so that the whole integration time is ﬁlled

with the code. The time delay is introduced either

by a channel time delay which is limited to 20ns by

the pattern generator settings or by rotating the code

in the memory registers. Combining both methods, a

coarse time shift is introduced by rotation of the code.

Hence the shifting step equals the chip length. A ﬁne

time shift is added with the channel time delay of the

pattern generator. The result is gapless adjustment

of required time delays. The spectrum is stored on a

computer and a second measurement with the inverted

code is performed. Both spectra are subtracted from

each other and the differential spectrum is used for

peak detection. (G¨otten et al., 2019) provides more

details about the correlation process with this inter-

rogator.

3 INFLUENCES ON

INTERROGATION SCHEME

AND SETUP

The inﬂuences of a real testbed setup on the interro-

gation scheme have to be investigated. This section

A Robust Serial FBG Sensor Network with CDM Interrogation Allowing Overlapping Spectra

presents two inﬂuences that affect the CDM results of

the interrogator. As the interrogator needs to deal with

different time delays, chip lengths and other tempo-

ral adjustments, the time delay of triggers and pattern

generators has to be analyzed. Also the spectrometer,

as one of the main parts of the correlation process is

analyzed in more detail in comparison with a standard

WDM system. A closer look at the modulators can be

found in (G¨otten et al., 2019).

3.1 Trigger Scheme and Delays

Figure 4 shows the time behavior between integration

time of the spectrometer and the code output of the

second modulator. When the spectrometer starts to

measure the spectrum, it emits a trigger that is con-

nected to the data generator to start outputting the

code. This delay time between trigger and the elec-

trical code signal arriving at the ﬁrst modulator is in-

dicated by the blue part in the time diagram. This de-

lay occurs for every single measurement of the spec-

trometer. After the delay, the ﬁrst modulator starts

with the code. This light travels through the network

and is reﬂected at different locations by each FBG.

Therefore, this code reaches the second modulator at

different times, indicated by the light gray boxes in

the diagram. The second modulation is synchronized

to a speciﬁc reﬂected code and is time shifted to all

other reﬂected codes. The time shift is depicted by

the green part of the time diagram. The codes are re-

peated several times within one integration time. At

the end, a part of the code is not covered by the inte-

gration time t

, indicated by the hatched gray part of

the time diagram.

During the trigger delay and the time shift, the second

modulator is set to a logical ’0’. The ﬁrst modulator

is set to logical ’0’ during the trigger delay as well.

Then it starts modulating the light, according to the

code. In the autocorrelation region, where both codes

are multiplied by means of the second modulator, the

error,introduced by an aperiodic autocorrelation is in-

vestigated in (G¨otten et al., 2018). Since a logical ’0’

does not prevent all light to pass the modulators, the

spectrometer collects light at the beginning of the in-

tegration time. This light is attenuated by the offset

of the ﬁrst modulator, the reﬂectance of the FBGs and

Code

Code Code

Direct Spectrum Inverted Spectrum

Figure 4: Temporal trigger scheme analysis.

the offset of the second modulator. Nevertheless, the

collected light appears in the direct and in the inverted

spectrum. The subtraction of both spectra cancels out

this effect which shows the advantage of a sequence

inverse keying (SIK) approach. The error introduced

in the green region is depending on the code and the

time shift or rather the synchronization point. Re-

ﬂected codes can arrive before the second modulator

starts to switch as well as after the second modulator

starts to switch. For larger time shifts, the expected

error increases, as well as the standard deviation of

that error (G¨otten et al., 2018). For network lengths

in the kilometer range, time shifts of more than 5µs

can occur.

3.2 Spectral Inﬂuences of CDM

Compared to WDM

In comparison with a WDM system, in the CDM pro-

cessing all FBGs need to share the same wavelength.

In the testbed, a CCD spectrometer is implemented

that can collect only a certain amount of light that

is represented in counts per wavelength range in the

spectrum. In this testbed the spectrum is limited to

counts that need to be shared by FBGs, as well

as noise. The distribution inside a spectral peak for

CDM is depicted in Figure 5. The two columns rep-

resent the amount of counts per spectrum, if all FBG

are not strained, or heated so they all overlap. To

achieve a cancellation of all interferences, the amount

of light in both spectra (direct and inverted) needs to

be equal, since the result is the difference out of these

two. The light of the synchronized FBG (here FBG

I-1) appears only in the direct spectrum, so nothing is

subtracted from the desired signal. Therefore, the two

parts of each FBG signal that distribute equally over

FBG 0 FBG 0

FBG 1 FBG 1

FBG 2 FBG 2

FBG i FBG i

FBG I-2 FBG I-2

FBG I-1

65535 Counts

X Counts

−

N N

Dir. Spectrum Inv. Spectrum Diff. Spectrum

X = 2·

65535−

N −σ− ε

I +1

Figure 5: Spectral inﬂuences of a CDM approach with SIK.

SENSORNETS 2020 - 9th International Conference on Sensor Networks

Table 1: Parameters of testbed components.

Parameter Value Description

C 2

= 65535 Maximum capacity of CCD spectrometer (BaySpec, FBGA)

I 25 Amount of serial sensors

N 3000 Mean of noise of spectrometer (measured)

σ 30 Standard deviation of noise (measured)

ε 0 Summary of other inﬂuences (modulators, shape of light spectrum, .. .)

both spectra for interference, is located only in the

direct spectrum for the synchronized signal. Addi-

tionally, the noise of each measurement, divided into

mean and standard deviation, is part of each spectrum

and consumes a certain amount of light. The differ-

ence spectrum contains now only the two parts of the

synchronized signal located in the direct spectrum and

two times the standard deviation, since subtracting a

random number from each other, the mean values can-

cel out and the standard deviation sum up. The theo-

retical available counts are calculated by the equation

in the Figure using the example of 2

counts.

Since a WDM does not require a second measurement

and a difference spectrum, the noise of the measure-

ment cannot cancel out. Having a look at the SNR of

the two different interrogation techniques, the advan-

tage of a CDM approach can be shown.

The SNR in the CDM system can be calculated by

SNR

CDM

C−

N − σ − ε

(I + 1) · 2 · σ

. (1)

C denotes the amount of counts provided in the spec-

trum, or rather the maximal possible height of a peak

in the spectrum minus the noise values (

N + σ) and

minus other inﬂuences ε. It is divided by the amount

of sensors I plus 1, as depicted in Figure 5. It is also

divided by two times the standard deviation σ that oc-

curs as the only part of the noise in the differential

spectrum. In comparison the SNR of a WDM system

is derived by

SNR

WDM

N + σ

, (2)

where the possible peak hight C is divided by the

complete noise consisting of mean value

N and stan-

dard deviation σ. The denominator of both SNRs in-

ﬂuences its value. The larger the denominator, the

worse the SNR. Using values from the testbed com-

ponents in this contribution (see. Table 1), it shows

the difference. For now, the other inﬂuences ε are set

to zero.

SNR

CDM

> SNR

WDM

(3)

− 3000− 30

(25+ 1) · 2 · 30

3000+ 30

(4)

40.07 b= 16.02dB > 21.63 b= 13.35dB (5)

Due to the cancelling of the mean noise value

N in the

differential spectrum, the SNR increases in compari-

son to a WDM system. To obtain the same SNR, as

the WDM approach, other inﬂuences can increase up

ε ≤ C− N − σ − SNR

WDM

· (I + 1) · 2σ (6)

≤ 2

− 3000− 30− 21.63· (25+ 1) · 2· 30

≈ 28762 Counts. (7)

Hence, these other inﬂuences on the spectrum of a

CDM approach can be up to 43.88% of the maxi-

mum amount of counts, so the maximal possible peak

height measurable by the implemented spectrometer.

The SNR remains larger than in a WDM interrogator

although the wavelengths are shared by all sensors in

the network.

4 CONCLUSIONS

In this work, a CDM interrogation scheme for serial

ﬁber optical sensor networks is introduced focusing

on CDM-related issues and advantages. The trigger

mechanism is temporally investigated. A delay intro-

duced by the trigger has no inﬂuence on the resulting

spectrum for peak detection. The spectral inﬂuences,

due to wavelength overlapping, are analyzed and the

advantage in terms of SNR are presented. The inter-

rogation scheme is very robust against other possible

imperfections. These imperfections such as the inﬂu-

ence of the modulators as well as the spectrum of the

source may accumulate up to 43.88% of the maxi-

mum measurable spectral power given by the spec-

trometer. Below this limit the CDM system performs

superior to WDM in terms of the achievable SNR. In-

cluding previous works focusing on other aspects of a

CDM approach, this CDM interrogation scheme is a

promising step in a new dimension of massive sensor

multiplexing in serial ﬁber optical sensor networks.

A Robust Serial FBG Sensor Network with CDM Interrogation Allowing Overlapping Spectra

ACKNOWLEDGEMENTS

This work has been funded by the German Ministry

of Education and Research (No. 13FH030PX8).

REFERENCES

Abbenseth, S., Lochmann, S., Ahrens, A., and Rehm, B.

(2016). Serial FBG sensor network allowing over-

lapping spectra. In Lewis, E., editor, Sixth European

Workshop on Optical Fibre Sensors. SPIE.

Braghin, F., Cazzulani, G., Cinquemani, S., and Resta, F.

(2013). Potential of FBG Sensors for Vibration Con-

trol in Smart Structures. In 2013 IEEE International

Conference on Mechatronics (ICM). IEEE.

Gliˇsi´c, B. (2016). Performance and Health Monitoring of

civil structures and infrastructure using long-gauge

and distributed Fiber Optic Sensors. In 2016 18th In-

ternational Conference on Transparent Optical Net-

works (ICTON). IEEE.

G¨otten, M., Lochmann, S., and Ahrens, A. (2018). Analy-

sis of Non-Ideal Optical Correlation for Interrogating

Overlapping FBG Spectra. In 2018 Advances in Wire-

less and Optical Communications (RTUWO), pages

154–160. IEEE.

G¨otten, M., Lochmann, S., Ahrens, A., and Benavente-

Peces, C. (2019). Detection Limits of Optical Auto-

correlations with a CDM Interrogator for Overlapping

FBG Spectra. In 2019 International Interdisciplinary

PhD Workshop (IIPhDW), pages 106–109. IEEE.

Kersey, A. D., Davis, M. A., Patrick, H. J., LeBlanc,

M., Koo, K. P., Askins, C. G., Putnam, M. A., and

Friebele, E. J. (1997). Fiber grating sensors. Journal

of Lightwave Technology, 15(8):1442–1463.

Kim, K. S. (2004). Dynamic Strain Measurement with Fiber

Bragg Grating Sensor System for Smart Structure.

Key Engineering Materials, 270-273:2114–2119.

Kitayama, K. (1998). Code Division Multiplexing Light-

wave Networks Based upon Optical Code Conversion.

IEEE Journal on Selected Areas in Communications,

16(7):1309–1319.

Lee, J. S. and Miller, L. E. (1998). CDMA Systems Engi-

neering Handbook. Artech House.

Nawrot, U., Geernaert, T., Pauw, B. D., Anastasopoulos,

D., Reynders, E., Roeck, G. D., and Berghmans, F.

(2017). Mechanical strain-amplifying transducer for

ﬁber Bragg grating sensors with applications in struc-

tural health monitoring . In Chung, Y., Jin, W., Lee,

B., Canning, J., Nakamura, K., and Yuan, L., editors,

25th International Conference on Optical Fiber Sen-

sors. SPIE.

O'Farrell, T. and Lochmann, S. (1994). Performance anal-

ysis of an optical correlator receiver for SIK DS-

CDMA communication systems. Electronics Letters,

30(1):63–65.

Rajan, G., editor (2015). Optical Fiber Sensors: Advanced

Techniques and Applications. CRC PR INC.

Wang, Y., Gong, J., Dong, B., Wang, D. Y., Shillig, T. J., and

Wang, A. (2012). A Large Serial Time-Division Mul-

tiplexed Fiber Bragg Grating Sensor Network. Jour-

nal of Lightwave Technology, 30(17):2751–2756.

Yamaguchi, T., Arai, M., and Shinoda, Y. (2015). Develop-

ment of wavelength measurement system of multiplex

Fiber Bragg Gratings using Optical Frequency Do-

main Reﬂectometry. In 2015 9th International Con-

ference on Sensing Technology (ICST). IEEE.

SENSORNETS 2020 - 9th International Conference on Sensor Networks