Sensors For Everyday Life
Sensors For Everyday Life
Subhas Chandra Mukhopadhyay
Octavian Adrian Postolache
Krishanthi P. Jayasundera
Akshya K. Swain Editors
Sensors for
Everyday
Life
Environmental and Food Engineering
Smart Sensors, Measurement
and Instrumentation
Volume 23
Series editor
Subhas Chandra Mukhopadhyay
Department of Engineering, Faculty of Science and Engineering
Macquarie University
Sydney, NSW
Australia
e-mail: [email protected]
More information about this series at http://www.springer.com/series/10617
Subhas Chandra Mukhopadhyay
Octavian Adrian Postolache
Krishanthi P. Jayasundera
Akshya K. Swain
Editors
123
Editors
Subhas Chandra Mukhopadhyay Krishanthi P. Jayasundera
Department of Engineering, Faculty of Institute of Fundamental Sciences
Science and Engineering Massey University
Macquarie University Palmerston North
Sydney, NSW New Zealand
Australia
Akshya K. Swain
Octavian Adrian Postolache Department of Electrical and Computer
Instituto de Telecomunicações Engineering
Lisbon University of Auckland
Portugal Auckland
New Zealand
and
ISCTE-IUL
Lisbon
Portugal
Sensors play a pivotal role in our everyday life. They gather data on environment,
and information on weather, traffic congestion, air pollution, water pollution, etc. is
obtained; they gather data on human body, and information on health, treatment or
therapy outcomes is obtained; they gather data on objects, and information for
monitoring and control of these objects is obtained; they gather data on subjects or
objects functions, and information for better decisions, control and action is
obtained. For instance, the weather information is used to choose adequate clothes,
the battery level sensor permits smartphone power management optimization, and
the level of blood glucose allows better healthcare management. Data collected
through sensors enhance our lives and our connections to each other and with our
environment, allow real-time monitoring of many phenomena around us, provide
information about quality of products and services, improve the equipment control
based on sensorized interfaces and contribute to increase knowledge on physical
and chemical world.
The advances in electronics, embedded controller, technology for communica-
tion as well as the progress towards a better informed, knowledge based society
increase the demand for small size, affordable sensors that allow accurate and
reliable data recording, processing, storing and communication. The work contains
invited chapters from renowned experts, working in sensors’ field, and it is split into
two books that present several technologies and applications of sensors in Envi-
ronmental and Food Engineering (ISBN 978-3-319-47322-2) and for Healthcare
Settings (ISBN 978-3-319-47319-2).
The book Sensors for Everyday Life—Environmental and Food Engineering
describes novel sensors and sensing systems developed for environment monitoring
and food production and quality assessment.
Environmental quality refers to characteristics from natural environment as well
as from the built environment (i.e. city air and/or water pollution, concentration of
nitrate from the soil in cultivated fields). Environmental quality plays an important
role in health and well-being of the populations. Degraded environmental quality as
produced by air and water pollution may affect our lives, directly or through the
food we eat. In food engineering various sensors are used for assessment of
v
vi Preface
In Chap. 1, a novel method for the simultaneous determination of NOx and soot in
the exhaust of diesel engines during the periodical technical inspection roadwor-
thiness test is presented. A multi-wavelength light extinction measurement, in a
setup similar to an opacimeter with high sensitivity, and a mathematical inversion
algorithm are used to obtain the concentrations from the extinction readings.
Analytical technique of the fine particles using atomic emission spectroscopy
system for an environmental pollution monitoring is presented in Chap. 2.
Laser-induced breakdown spectroscopy (LIBS) system and the helium-microwave-
induced plasma-atomic emission spectroscopy (He-MIP-AES) system are used for
characterizations and real-time measurement of the air chemical compositions and
particle size.
Chapter 3 presents sensors and method for automatic fault detection in heating
ventilating and air conditioning (HVAC) systems. This is important mainly in smart
buildings context as the indoor condition in these buildings is mainly related with
the capabilities and reliability of HVAC systems.
New optical fiber humidity sensor is described in Chap. 4. Different humidity
sensors that have been developed by now are presented focusing in the different
optical structures and materials that have been used for improving sensitivity and
resolutions of these sensors.
The measurement of air gas concentration represents an important field of
application of sensing technologies. In Chap. 5 of the book, a review on the oxygen
gas sensing technologies is presented with focus on potentiometric, amperometric,
paramagnetic and tunable diode laser spectroscopy (TDLS) sensors. Theoretical
aspects and operational basic of these technologies, system requirements as well as
limitations of the methods are discussed in this chapter.
A low-cost sensor node based on interdigital capacitive sensor for nitrate and
nitrite in surface and ground water concentration detection is presented in Chap. 6.
This sensor is important for agronomy as well as for water pollution assessment.
Nitrates may be present in high concentration in ground and surface water as a
result of intensive agriculture, disposal of human and animal sewage and industrial
wastes.
In Chap. 7 an intelligent wireless sensor network system designed to monitor
various parameter in palm oil plantation for improvement in the controlled polli-
nation process is presented. The system helps in making decision related to best
time for pollination process. The inaccuracy in determining pollination readiness
Preface vii
of the oil palm flower could potentially cause a detrimental effect on the palm oil
industry in the long run.
The following two chapters present sensors for determination of quality and
quantity of water for drinking purpose. In Chap. 8, a reflectometer and a Doppler
radar systems for detection of water level in septic tank is described. A novel S3
(Small Sensor System) nanowire device for the detection of complex mixtures of
bacteria in potable water is presented in Chap. 9.
Next three chapters describe sensors used in food productions and quality
assessment. A novel approach to monitor the quality of milk products, based on
electromagnetic wave spectroscopy is presented in Chap. 10. The system use vector
network analyser to capture spectral signatures in the form of scattering parameters
from electromagnetic wave sensors. Data on reliability testing is presented. A new,
rapid, portable, easy-to-use, economic and non-destructive fouling based on
nanowire technology device to control the presence of false grated Parmigiano
Reggiano cheese is described in Chap. 11. A review on the conventional techniques
and dielectric spectroscopy for analyzing food products is presented in Chap. 12,
focusing on the application of dielectric spectroscopy in fats and oils adulteration
detection.
Different wireless sensor network architecture is implemented nowadays to
perform distributed measurement tasks for environment monitoring with increase in
space resolution. Big challenge in these implementations continues to be wireless
interference and radio-frequency (RF) spectrum crowding. Chapter 13 focuses on a
technique for optical-based RF interference cancellation. In this chapter several
system architectures are presented and a sample of their experimental performance
and the key characteristics of this technique and the future prospects for this
technology, focusing specifically on photonic integrated circuits are discussed.
A scheme is proposed in Chap. 14 that can reduce the performance difference
between cluster heads (CHs) involved in inter-cluster communication on IEEE 802.
15.4 cluster-based wireless sensor networks (WSNs) under spatial non-uniform
traffic condition where the CHs have various amount of traffic. This reduced the
energy consumption and simplified processing mechanism to achieve long WSNs
lifetime under limited network resource condition.
We do sincerely hope that the readers will find this special issue interesting and
useful in their research on sensors and wireless sensor networks for environment
monitoring, food production and quality assessment.
We want to acknowledge all the authors for their contribution and for sharing
of their knowledge. We hope that the works presented in this book will stimulate
further research related to sensors for everyday life.
ix
x About the Editors
xiii
xiv Contents
Nanowire (S3) Device for the Quality Control of Drinking Water . . . . . 179
Estefanía Núñez Carmona, Matteo Soprani and Veronica Sberveglieri
Milk Quality Monitoring Using Electromagnetic Wave Sensors . . . . . . . 205
Keyur H. Joshi, Alex Mason, Olga Korostynska
and Ahmed Al-Shamma’a
Grated Parmigiano Reggiano Cheese: Authenticity Determination
and Characterization by a Novel Nanowire Device (S3)
and GC-MS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Veronica Sberveglieri, Manohar P. Bhandari, Andrea Pulvirenti
and Estefania Núñez Carmona
Lard Detection in Edible Oil Using Dielectric Spectroscopy . . . . . . . . . . 245
Masyitah Amat Sairin, Samsuzana Abd Aziz,
Nina Naquiah Ahmad Nizar, Nurul Adilah Abdul Latiff,
Alyani Ismail, Dzulkifly Mat Hashim and Fakhrul Zaman Rokhani
Optical-Based Interference Cancellation in Wireless
Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Matthew P. Chang, Jingyi (Jenny) Sun, Monica Lu, Eric Blow
and Paul R. Prucnal
Traffic Adaptive Channel Access Scheme for IEEE802.15.4
Cluster-Based WSNs Under Spatial Non-uniform
Traffic Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
Akiyuki Yamauchi, Kazuo Mori and Hideo Kobayashi
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
Determination of NOX and Soot
Concentrations Using a Multi-wavelength
Opacimeter
Abstract A novel approach to measure both the particle and the NO2 concentra-
tion in the exhaust of diesel engines during the roadworthiness test in the periodical
technical inspection is presented. It is based on a multi-wavelength extinction
measurement and a mathematical inversion algorithm to obtain the concentrations
from the extinction readings. Such individual concentration values can deliver
valuable insight into the cause of engine or exhaust aftertreatment defects. Fur-
thermore the extended opacimeter provides future-proofness, if nitrous gas emis-
sions are incorporated in the roadworthiness regulations. In addition to a detailed
description of the multi-wavelength approach this chapter provides an overview of
particle and nitrous gas emissions by diesel engines, the related legislation, the
extinction measurement using standard opacimeters, and the physical background
for this optical measurement method. The applicability of the multi-wavelength
method is derived mathematically and validated with first experimental results as
well as with simulations.
1 Introduction
The emissions of vehicles equipped with internal combustion engines pose a sig-
nificant environmental problem and cause severe health effects. A big portion of the
engines run on diesel fuel. In some European countries they even surpass the amount
of gasoline vehicles with a relation of 70 to 30 % [1]. Across Europe the average
fraction of newly registered diesel cars reached 58 % in 2011 [2]. While the main
pollutant in the exhaust of gasoline engines is carbon dioxide (CO2), the most
H. Axmann (✉)
AVL DiTEST GmbH, Alte Poststraße 156, 8020 Graz, Austria
e-mail: [email protected]
A. Bergmann ⋅ B. Eichberger
Institute of Electronic Sensor Systems, University of Technology,
Inffeldgasse 10/II, 8010 Graz, Austria
relevant pollutants of diesel engines are particulate matter (PM) and nitrous gases
(NOX). CO2 is a green house gas, whereas particles and NO2 are toxic components.
Diesel particulate matter consists mainly of carbonaceous particles, ashes and
unburned fuel and oil droplets [3]. It has been lately proven to be carcinogenic [4].
Furthermore the black carbon (BC) fraction of the particles contributes to global
warming, and when deposited on snowpacks the particles lead to accelerated
melting through absorption of the sunlight [5].
Nitrous gases consist of nitrogen monoxide (NO) and nitrogen dioxide (NO2).
During the combustion process of diesel fuel mostly NO is produced. By the
catalyzed particle filter systems in modern vehicles a significant amount of NO is
oxidized to NO2 [6–8]. NO2 is a brown, toxic gas, which may lead to mucosal
irritation. It is a main contributor to smog in big cities [9] and is—in aqueous
solution—a strong acid leading to acid rain [10].
Governments all over the world reacted to this threat by introducing limits for
the allowed concentration of said pollutants in the exhaust of internal combustion
engines. Since the first emission regulations in the early 1970s the limit values have
been significantly reduced [10–12]. This is shown in Fig. 1 using the European
Emission Standards (EURO 1-6) as an example. To conform to these emission
limits the engine manufacturers have to optimize the combustion process and to add
new exhaust aftertreatment systems. Still the efficiency of these measures can
deteriorate over time. Accordingly periodic emission checks are necessary to ensure
the compliance of the vehicles to the emission limits. Such emission checks are
typically performed during the periodical technical inspection (PTI) in garages or
vehicle inspection institutions.
For the case of diesel vehicles the regulation-compliant functionality of the
engine and the exhaust system is determined from the opacity of the exhaust gases.
The corresponding measurement device is the opacity meter, commonly just called
opacimeter, which is based on an optical measurement. Just as the first emission
0.12
PM limit [g/km]
0.1
0.08
EURO 2
0.06
EURO 3
0.04
0.02 EURO 4
EURO 5/6
0
0 0.2 0.4 0.6 0.8 1
NOx limit [g/km]
Determination of NOX and Soot Concentrations Using … 3
limits its measurement principle dates back to the 1970s [13]: A measurement
chamber of defined length is filled with exhaust gas, and the attenuation of light by
the sample is measured. While this was working perfectly for the older diesel
vehicles emitting tight black clouds of soot, it is reaching its limits for modern
models [14]. Equipped with diesel particle filters (DPFs) they emit much less, very
small particles. Accordingly the exhaust plumes are effectively transparent, only
with a very small amount of opacity. Therefore updated or new measurement
technologies are needed in order to perform meaningful and reliable measurements.
An example for a new technology is light scattering. It is also an optical mea-
surement method, measuring the amount of light deflected by the particles in the
exhaust gases. Due to its measurement principle it is mainly sensitive to particles
[15]. Although it proved to be well applicable for PTI exhaust measurements, it has
not been approved by the governments yet. Some other new measurement methods
have been proposed, but without success [16, 17]. As an alternative researchers
focused on updated versions of the opacimeter featuring an increased sensitivity.
Generally it is possible to lower the detection limit by at least one order of mag-
nitude with moderate means.
However opacimeters measure the combinational absorbing effect of all exhaust
components within the green wavelength range, where the human eye has its
maximum sensitivity. Originally this was by intention. It was the goal to measure an
overall visual obstruction produced by the vehicle. Besides, for older vehicles, most
of the effect could be attributed to the PM part. With the reduction of the particle
emissions the attenuating effect of the brown NO2 can surpass that of the particles.
In such a case it might often be desirable to obtain additional information about the
source of the measured signal in order to quickly identify the related defects in the
exhaust system. Furthermore getting individual figures for PM and NOX or NO2
can be advantageous regarding the individual limitation of these two criteria pol-
lutants. A one year study performed by the International Motor Vehicle Inspection
Committee (CITA) in 2011 already investigated PTI test equipment for individual
measurement of said components [13].
As shown in the following sections such extended measurement results can be
achieved by an opacimeter enhanced with a multi-wavelength light source.
2 The Opacimeter
I at the detector. By relating the two intensities I and I0 to each other the opacity
N can be calculated as
N = 1 − ðI ̸I0 Þ. ð1Þ
Accordingly one can determine the concentration of the absorbing material from
the opacity measurement, if the extinction coefficient σ ext is known and constant.
The following section describes the physical background of said values and shows
typical values for PM and NO2.
When light travels through a medium, energy is partly removed from the light beam
by the obstacles (particles, gas molecules and atoms) in the medium. This results in
an attenuation of the light beam. Such an effect is called extinction [19]. It is the
combination of absorption, where energy is transferred from the incident beam to
the obstacles, and scattering, where some parts of the light beam are deflected from
the original path and thereby also removed from the incident beam. Mathematically
this can be described in terms of the cross sections
Determination of NOX and Soot Concentrations Using … 5
where σ abs and σ sca are the absorption and the scattering cross sections respectively
[20]. The concept of a cross section is used to describe the ability of an obstacle to
remove light from an incident beam in a simple way. It can be understood as the
area shadowed from the incident beam. For the case of exhaust the main contributor
to extinction is absorption. This is attributable to the strong absorbing effects of the
primary components of the exhaust. Soot particles have a scattering share of only
10–30 % [21, 22]. Gases like NO2 even have a negligible scattering effect [23]. This
is the reason why, although physically incorrect, it has become customary to call
k absorption coefficient rather than extinction coefficient.
The total extinction coefficient of a mixture of substances with differing
extinction coefficients ki can be determined by summation [24]. As in (3) ki can be
expressed by the products of the individual cross sections σ ext, i and the respective
concentrations ci . Accordingly the total extinction coefficient k can be calculated as
k = ∑i ki = ∑i ci σ ext, i . ð5Þ
As said before the scattering effect of gases like NO2 is negligible. Its extinction
cross section is therefore almost identical to the absorption cross section, which is
shown in Fig. 3 for the wavelength range from 250 to 800 nm, given in
6 H. Axmann et al.
section of NO2
8 · 10 -19
6 · 10 -19
4 · 10
-19
2 · 10
-19
0
300 400 500 600 700 800
Wavelength λ [nm]
cm2/molecule [26]. It has been measured at a pressure of 1000 mbar and a tem-
perature of 293 K.
Generally the uncertainty is in the order of a few percent. Near 250 nm the
uncertainty increases, as oscillations occur that are physically implausible. The peak
around 316 nm seems to be an artefact, too. Both phenomena are not of importance
for the actual work. The high oscillations around the peak at approximately 400 nm
however are factious. Thus it is crucial to utilize data with high wavelength reso-
lution in this area in order to achieve proper results. Temperature and pressure also
have a notable influence on the absorption behavior. Those influences can be
compensated using empirical models [25, 27].
15
10
0
200 400 600 800 1,000
Wavelength λ [nm]
size distributions of diesel exhaust [3, 31] predict deviations in the order of a few
percent.
The absorption spectrum of NO2 and the extinction spectrum of soot are very
different, as obvious from Figs. 3 and 4. Accordingly their corresponding extinction
effects differ substantially for e.g. red, green and blue light. This is used in the
multi-wavelength approach presented in the following section.
holds. Herein λi is the wavelength of each light source, kNO2 , ksoot , kother and ki are
the individual extinction coefficients of NO2, soot particles and other components
and the total extinction coefficient at the given wavelength λi respectively. The
equation can be rewritten as in (5) using the product of the concentrations cx and the
extinction cross sections σ x , leading to
ki = cNO2 σ NO2 ðλi Þ + csoot σ soot ðλi Þ + cother σ other ðλi Þ. ð7Þ
When performing three measurements of the same sample (i.e. the concentra-
tions cx remain constant during the whole measurement time) an equation system
with three equations and also three unknowns is obtained. This of course requires
knowledge of all extinction cross sections. In such a case, the matrix equation
0 1 0 10 1
k1 σ NO2 ðλ1 Þ σ soot ðλ1 Þ σ other ðλ1 Þ cNO2
@ k2 A = @ σ NO2 ðλ2 Þ σ soot ðλ2 Þ σ other ðλ2 Þ A@ csoot A ð8Þ
k3 σ NO2 ðλ3 Þ σ soot ðλ3 Þ σ other ðλ3 Þ cother
5 Technical Realization
sensitivity in the order of Δk = 0.001 m-1 the optical path length and hence the
measurement chamber should not be too small. We used a cylindrical cell with a
diameter of 2.4 cm and a length of 400 mm, where the axis of the cylinder coin-
cides with the optical axis of the system. The inlet for the smoke sample is placed in
the middle of the chamber. The exhaust is filled into the chamber by the dynamic
pressure in the exhaust pipe. It leaves the chamber through the in-/outlet for the
light beam.
A flow of sheath air transports the exhaust sample to the smoke outlet. The
sheath air flow is produced from ambient air by filtration using a high-efficiency
particulate air (HEPA) filter. On the one hand the sheath air protects the optical
setup from contamination, on the other hand it defines the length of the optical path.
As the optical path length directly influences the measurement value, it is crucial to
keep the optical path length constant. Therefore the sheath air flow needs proper
fluid dynamical considerations.
In Fig. 6 a multiphysical complex fluid dynamic simulation result is shown. The
ratio of exhaust gas entering through the middle section of the measurement cell to
the air flow rate through the sheath air channels defines the optical path. In order to
quantify a representative exhaust gas sample within the measurement cell, the
distribution of the analytes should be homogeneous, independently of the dynamic
pressure in the exhaust pipe. A proper fluid dynamic design and hence an iterative
optimization approach by means of complex fluid dynamics is a necessary pre-
requisite to obtain meaningful measurement results.
For the light source it is advantageous to use a realization that combines the three
colors into a single housing. That ensures comparable optical behavior of the
68.847
Particle velocity [m/s]
63.162
57.478
51.793
46.108
40.423
34.739
29.054
23.369
17.684
12.000
6.315
0.630
Fig. 6 Multiphysical complex fluid dynamic simulation of the measurement cell. The points
represent particle ensembles with particle velocities scaled by the color. A homogeneous velocity
distribution within the optical path guarantees representative samples of the exhaust gas portion to
be quantified
10 H. Axmann et al.
Relative intensity
0.6
0.4
0.2
0
400 450 500 550 600 650 700
Wavelength λ [nm]
Blue Green Red
different colors due to their proximity. To increase signal-to-noise ratio and reduce
the challenges for amplification on the receiver side, the light source should have a
high intensity. This is achievable with a commercially available high power LED,
like for instance the OSRAM LRTB. Its three dominant wavelengths are at 632,
523 and 455 nm for the red, the green and the blue color respectively (see Fig. 7).
The incident light is collimated by means of an aspherical lens (optical
parameters: d = 18 mm, f = 15 mm, bfl = 11.10 mm, Md = 6 mm), resulting in a
beam divergence of typically less than 8°. If needed the beam divergence can be
adapted by changing the distance between the lens and the LED.
On the receiver side an optical detector converts the incident light into the
electrical domain. For this purpose a PIN photodiode with a large optical area of
7 mm2 is used. A collimating lens, identical to the one on the transmitter side, is
placed in front of the photodiode, to maximize the incident intensity and overcome
losses due to misalignments produced by vibrations and shocks. As the measure-
ment signal is directly proportional to the source light intensity, a stable source is of
high importance. Nevertheless if fluctuations occur they can be compensated via a
reference photodiode, measuring the current source intensity.
exhaust sample
TX RX
LED photodiode
measurement path
ITX IRX
Transmitter Receiver
optical reference path
side side
reference
I
measurement
0 t
power at the transmitter side. As a consequence, at least two precision resistors are
needed: one is part of the voltage controlled current source for the transmitter LED,
another one is part of the transimpedance amplifier, converting the current output of
the photodiode into a voltage output for the ADC.
Figure 9 represents the actual setup. The electronic instrumentation relies on the
ratiometric principle to a great extent. A set of two identical optical receivers
replaces the optical chopper of Fig. 8, each one including its dedicated tran-
simpedance converter. A total of three precision resistors (RREF, TX , RREF, RX, 1 and
RREF, RX, 2 ) are required, which is acceptable because these parts are available with
excellent specifications for tolerance, temperature coefficient and long-term
stability.
The optical output power of the LED correlates with its forward current. It is
adjusted by the setting of the DAC, its reference voltage and the transconductance
of the voltage controlled current source. A digital control loop, implemented in the
firmware of the microcontroller, uses the output voltage VRX, 2 of the optical ref-
erence path to set the LED current ITX and keep it at the desired set level.
At the receiver side there are two photodiodes, each with its dedicated tran-
simpedance amplifier for the current-to-voltage conversion. A critical part for the
overall performance of the measurement setup is the matching of the optical
receivers and their interface electronics. The photodiodes operate in current (short
circuit) mode. Their light-to-current transfer ratio is therefore highly linear over
several decades of intensity. Thermal influences are low and further reduced by
keeping both photodiodes at the same temperature.
The influence of dark current and bias currents at the receiver side are com-
pensated by a reference measurement with the transmitter LED turned off. Minor
deviations are canceled out by digital post processing; abnormally high deviations
indicate a fault situation and are handled by self-diagnostic functions.
REF
RREF,RX,1 voltage
exhaust sample
VRX,1
V+
REF IRX,1
voltage RREF,RX,2 ADC
measurement path
ITX
DAC
optical reference path VRX,2
Voltage IRX,2
micro- RREF,TX controlled
micro-
processor current source
processor
control VCTL (transconductance Photo detectors & control
amplifier) transimpedance amplifiers
The measurement setup provides three extinction coefficients, each for one of the
three used wavelengths λi . Equation (8) provides the equation system for deter-
mining the concentrations of three absorbing species from such a measurement. As
mentioned above there is generally no exact solution for this equation system.
Furthermore the third typical absorber in addition to soot and NO2 can often not
easily be identified. We therefore use the reduced equation system
0 1 0 1
k1 σ NO2 ðλ1 Þ σ soot ðλ1 Þ
@ k2 A = @ σ NO2 ðλ2 Þ c
σ soot ðλ2 Þ A NO2 , ð9Þ
csoot
k3 σ NO2 ðλ3 Þ σ soot ðλ3 Þ
which focusses on the two criteria pollutants particles and NO2. In the discussion
section we will show that this generally delivers good results for diesel engine
exhaust.
The concentrations can be determined from (9) by inversion. This can be most
easily achieved by multiplication with the pseudo inverse
A + = ðAT AÞ − 1 AT , ð10Þ
where A denotes the cross section matrix and AT its transpose. It will yield the best
fit for the solution in a least squares sense.
The cross section values σ x ðλi Þ for NO2 and soot can be taken from Figs. 3 and 4
respectively, as long as monochromatic light sources are used. The graphs however
use different units for the cross section values. So a conversion of one of the cross
sections to the unit of the other one is needed. We chose to use the unit m2/g and
converted the cross section for NO2 as follows:
m2 1 cm2 NA molecules ̸mol
σ NO2 = 4 σ NO2 ð11Þ
g 10 molecule MNO2 g ̸mol
Herein MNO2 = 46.005 g/mol is the molar mass of NO2 and NA = 6.022
1023 molecules/mol is the Avogadro constant.
Since we use a LED as a light source, the single colors are clearly not
monochromatic. In this case the corresponding nominal intensities I0 at the
wavelengths λi have to be replaced by continuous nominal intensity spectra I0 ðλÞ
from Fig. 7, where at each wavelength λ (2) holds:
On the detector the single intensities IðλÞ are integrated with respect to the
spectral efficiency γðλÞ of the detector:
14 H. Axmann et al.
Z Z
I= γðλÞIðλÞdλ = γðλÞI0 ðλÞexpð − kLeff Þdλ ð13Þ
In order to set up the equation system, the extinction coefficient is again replaced
by the product of cross section and concentration, now denoted as σ x ðλi Þ and ĉ
respectively. However in the case of a polychromatic light source the strict pro-
portionality does not hold for the mean values. This is obvious from the following
equation:
R
γðλÞI0 ðλÞexpð − cσ x ðλÞLeff Þdλ
expð − ĉσ x ðλi ÞLeff Þ = R ð15Þ
γðλÞI0 ðλÞdλ
Despite the limited proportionality the model in (9) provides a good approxi-
mation to the real behavior, as shown by the simulations in the following discussion
section.
The multi-wavelength approach has been evaluated using simulations and basic
measurements. The results are presented and discussed in the following sections.
Our first measurements have been performed with a new, highly sensitive
opacimeter, to ensure that reliable measurements of the extinction produced by NO2
are possible even at low concentrations. The opacimeter was equipped with a single
color light source (green light with Gaussian intensity spectrum, μ = 560 nm,
σ = 8 nm). The resolution was 0.001 m-1. The results are shown in Fig. 10.
Conforming to theory the measured extinction value is linearly related to the
NO2 concentration. The smallest resolvable NO2 concentration is in the order of
10 ppm. Compared to simulations the measured extinction coefficient is a bit lower.
The simulated value at 1000 ppm was approximately 0.2 m-1, whereas the
Determination of NOX and Soot Concentrations Using … 15
0.08
0.06
0.04
0.02
Measurement
Linear fit
0
0 200 400 600 800 1,000
NO2 concentration [ppm]
measured value is 0.12 m-1. The deviations can probably be attributed to different
ambient conditions (pressure and temperature).
The simulations have been performed in MATLAB using the model from (9) with
the cross section spectra presented in Figs. 3 and 4. Combinations of NO2 at
concentrations from 100 to 1000 ppm and soot particles at concentrations from 1 to
1000 mg/m3 have been tested. The results for the estimation of the NO2 concen-
tration from the simulated values are depicted in Fig. 11 for the noise-free case in
the form of the relative deviation from the actual concentration.
1.5
NO2 error [%]
0.5
1
0 10
20
0 5
10 00 10 0
2 00
3 00 20 0
4 00 40 0
5 00
6 00 60 0
80 0
NO2 concentration 7 800 00 10
00
0 soot concentration
9 00
[ppm] 10 [mg/m3]
16 H. Axmann et al.
10 2
NO2
soot
10 3
10 −2 10 −1 10 0 10 1 10 2 10 3
Cross section ratio σsoot(λg)/σNO2(λg)
For most combinations the error is below 1 %. Only if the contribution of one
component is very low, e.g. 100 ppm of NO2, the error can increase to almost 2 %.
The reason for this error in the noise-free system is the limited proportionality due
to polychromatic light. Consequently the systematic error can be reduced, when
light sources with a narrow wavelength spectrum or quasi-monochromatic light
sources like lasers are used.
If noise is applied the errors generally rise to a few percent. The relative error for
1 % of noise is depicted in Fig. 12. Larger errors between 10 % to 40 % for the NO2
concentrations are only found when the contribution of soot is at least twenty times
higher than that of NO2. Solely if the share of NO2 is smaller than 1/100 and hence
negligible the mean error rises above 40 %. For the particle concentration small
errors below 2.5 % can be achieved for all combinations where csoot ≥ 10 mg/m3.
6.3 Discussion
primary contributors to the extinction. Other strong absorbers present in the exhaust
sample might lead to deteriorated results. For diesel exhaust this is a rare case.
Known examples of additional absorbers are white and blue smoke. White smoke,
produced by condensed water, can be eliminated by proper conditioning of the
engine prior to the PTI and by heated sample lines. By keeping the exhaust tem-
perature above the due point of exhaust gases, which is in the area of 45–55 °C
[14], water is hindered from condensing. When using a temperature above 100 °C
already condensed droplets can be evaporated again. Blue smoke is typically related
to an engine defect. It can occur for instance, if oil leaks into the cylinder [35–38].
If required, such hydrocarbon emissions can be eliminated by integration of an
oxidation catalyst (“catalytic stripper”) into the heated sample line [39]. Hence,
when implementing such measures, an overall good applicability to real engine
emissions can be assumed.
For the real-world exhaust samples comparison measurements with dedicated
NO2 sensors are planned to gain insights into the quality of the measurement
results. Although such dedicated devices will of course yield more accurate results
[40], the proposed method can provide at least valuable qualitative results without
the need for extra measurement equipment. Furthermore little or no additional
conditioning of the exhaust sample is needed in comparison to a standard
opacimeter. Finally the main functionality of the opacimeter, i.e. measuring the
exhaust opacity, is not degraded in any way.
7 Conclusion
Emission regulation covering the whole lifetime of vehicles is a key for a clean and
healthy environment. Two of the main pollutants of combustion engines, namely
soot and nitrous oxides, depend strongly on the state of health of the combustion
engine as well as of the exhaust aftertreatment system. Within this chapter a novel
method for the simultaneous determination of NO2 and soot in the exhaust of diesel
engines with emphasis on periodical technical inspection (PTI) was elaborated in
detail. It is based on a spectroscopic multi-wavelength light extinction measurement
in a setup similar to an opacimeter with high sensitivity. Preliminary experimental
results have proven its ability to reliably detect small concentrations of NO2 with a
limit of detection (LOD) as low as 10 ppm. The experimental research with the
focus on the simultaneous determination of the second pollutant soot is ongoing.
The overall results are promising and indicate the potential to gain further infor-
mation on the pollutants contained in exhaust without the need for an additional
dedicated sensor in PTI applications.
18 H. Axmann et al.
References
1. J. Linn, Explaining the adoption of diesel fuel passenger cars in europe (RFF, Washington,
2014)
2. M. Cames, E. Helmers, Critical evaluation of the european diesel car boom—global
comparison, environmental effects and various national strategies. Environ. Sci. Eur. 25(15),
1–22 (2013)
3. D.B. Kittelson, Engines and nanoparticles: a review. J. Aerosol Sci. 29(5/6), 575–588 (1998)
4. International Agency for Research on Cancer, World Health Organization, Diesel engine
exhaust carcinogenic. Press Release No. 213 (2012)
5. Y. Qian, T.J. Yasunari, S.J. Doherty, M.G. Flanner, W.K.M. Lau, J. Ming et al.,
Light-absorbing particles in snow and ice: measurement and modeling of climatic and
hydrological impact. Adv. Atmos. Sci. 32(1), 64–91 (2015)
6. G.P. Merker, C. Schwarz, R. Teichmann, in Grundlagen Verbrennungsmotoren: Funktion-
sweise, Simulation, Messtechnik, 7th edn. (Springer Vieweg, Wiesbaden, 2014)
7. G.A. Bishop, D.H. Stedman, Emissions of nitrogen dioxide from modern diesel vehicles, WIT
Trans. Ecol. Environ. 116 (2008)
8. T.R. Dallmann, R.A. Harley, T.W. Kirchstetter, Effects of diesel particle filter retrofits and
accelerated fleet turnover on drayage truck emissions at the port of Oakland. Environ. Sci.
Technol. 45(24), 10773–10779 (2011)
9. N.K. Verma, S.K. Khanna, B. Kapila, Comprehensive Chemistry, vol. I. Class XI. (Laxmi
Publications, New Delhi, 2010)
10. K. Mollenhauer, H. Tschöke, Handbook of Diesel Engines, 3rd edn. (Springer, Berlin
Heidelberg, 2007)
11. C.D. Rakopoulos, E.G. Giakoumis, Diesel Engine Transient Operation: Principles of
Operation and Simulation Analysis (Springer, London, 2009)
12. European Commission: Commission regulation (EC) No 692/2008 of 18 July 2008
implementing and amending regulation (EC) No 715/2007 on type-approval of motor
vehicles with respect to emissions from light passenger and commercial vehicles (Euro 5 and
Euro 6) and on access to vehicle repair and maintenance information. OJ L 199, (2008),
pp. 1–136
13. P. Boulter, P. Buekenhoudt, P. Stricker, H.-J. Mäurer, C. Nolte, T. Ost et al., TEDDIE—A
new roadworthiness emission test for diesel vehicles involving NO, NO2, and PM
measurements, final report, (CITA, 2011)
14. P. Eastwood, Particulate Emissions from Vehicles (Wiley, Chichester, 2008)
15. H. Axmann, A. Bergmann, B. Eichberger, Measurement of ultrafine exhaust particles using
light scattering, in 2013 Seventh International Conference on Sensing Technology
(ICST2013) (IEEE, 2008), pp. 937–941
16. G. Hauser, G. Hauser, Entwicklung eines Rußmessgerätes für Kfz-Werkstätten und
AU-Prüfung, final report, phase 1 (E.A.S.T. Solutions GbR, Twieflingen, Germany, 2009)
17. I. Bennett, J. Tikkanen, L. Ntziachristos, A new sensor for long term monitoring of ultrafine
particles in a number of applications, in Annual Aerosol Science Conference (The Aerosol
Society, 2010)
18. L. Bergmann, C. Schaefer, H. Niedrig, Lehrbuch der Experimentalphysik, vol. 3: Optik.
Wellen- und Teilchenoptik. Berlin: de Gruyter, 10 ed., 2004
19. H.C. van de Hulst, Light scattering by small particles (Dover Publications, New York, 1981)
20. C.F. Bohren, D.R. Huffman, Absorption and scattering of light by small particles (Wiley,
New York, 1998)
21. C.M. Sorensen, Light scattering by fractal aggregates: a review. Aerosol Sci. Technol. 35(2),
648–687 (2001)
22. A.C. Szkarlat, S.M. Japar, Optical and chemical properties of particle emissions from on-road
vehicles. J. Air Pollut. Control Assoc. 33(6), 592–597 (2012)
Determination of NOX and Soot Concentrations Using … 19
Abstract This chapter describes the elemental analytical technique of the fine
particles using atomic emission spectroscopy system for an environmental pollution
monitoring. Based on our previous measurement reports, differences of the usage
between the laser-induced breakdown spectroscopy (LIBS) system and the
helium-microwave-induced plasma-atomic emission spectroscopy (He-MIP-AES)
system are explained. Both systems were developed to allow to fulfil the criteria
prescribed by the Ministry of Environment, Japan, for measuring the chemical
components of particulate matter (PM) by introducing additional original sampling
systems. In current environmental monitoring systems, PMs are typically collected
on trapping filters placed across Japan and classified as either suspended particulate
matter (SPM) or PM2.5 depending on the size. The collected PMs are subsequently
analysed with automated measurement instruments such as a piezo balance and
with methods such as beta ray attenuation and light scattering. While these mea-
surement methods allow the mass concentration of PMs in the air to be obtained at
hourly intervals, the chemical composition of individual particles is analysed with
time-intensive laboratory procedures. In contrast, the LIBS and He-MIP-AES
measurement system allow the chemical compositions and particle sizes to be
measured simultaneously in real time.
1 Introduction
measurement sample is wet than helium plasma. However, helium is the ideal
support gas because the organic compounds eluted from the gas chromatograph are
then almost completely atomized in the plasma, which produces a line emission
spectrum. In contrast to the Ar-ICP system, the He-MIP-AES system employs
helium as the ionized gas for excitation. The metastable level of helium (19.8 eV) is
greater than argon (11.5 eV), and its high excitation energy allows the analysis of a
wide range of elements, including high-energy ionizing elements such as the
halogen family.
2 Motivation
With the rapid growth and industrialization of Asia’s economy, fossil fuel emis-
sions from factories, power plants and automobile are emitting into the atmosphere.
For preventing growing aerosol pollution and global climate change problems,
numerous organizations are actively involved in research programs for the
development of processes that utilize biodiesel sources. It is known that a diesel
engine is more fuel efficient than a gasoline engine, and further, electronically
controlled common-rail injection contribute to clean burning. Hence,
cleaner-burning engine often produce smaller soot; nanosized particles as a
byproduct. It is well known that inhalation of fine particle matter causes several
health problems. Particles that are smaller than 2.5 μm in size (PM2.5) can pene-
trate the gas-exchange region of the lung. However, it is difficult to obtain the size
and composition information of the fine particles in real time. We have developed
and conducted experiments using following measuring methods to overcome the
difficulty. Laser-induce breakdown spectroscopy (LIBS) is one of the useful
methods for the determination of the elemental composition of the fine particles.
Our research group has been developing LIBS techniques that have potential
applications in various fields. However LIBS itself is not enough to obtain the
particle size information. Figure 1 shows photograph of the MEMS fabrication
room and SEM images of the particles deposited on Si membrane porous filter
which gathered in the clean room.
Figure 2 illustrates for one example of the reason why LIBS particle size
measurement with the range smaller than laser spot size is hard to do. In case LIBS
measurement are conducted for fine particles, we only know the plasma intensity
information. Because of the intensity means total amount of the particles, it may be
only one particle, otherwise the plasma emission irradiated from number of
particles.
Based on the requirement for particle size information, we applied laser-induced
incandescence (LII) technique with LIBS. In previous our researches, the LII decay
simulation model was based on one of the most detailed studies provided by
Michelsen [22] (Fig. 3).
24 S. Ikezawa et al.
When using the LII technique, laser pulse was irradiated on particles then
thermal radiation was recorded. LII technique is based on the analysis of cooling
behaviour of particles after irradiation with the laser pulse. By using
Stefan-Boltzmann law for a black body, the LII signal; profile of the intensity decay
time depending on the particle size. Figure 4 shows solution algorism to obtain the
LII signal as a function of time-temperature curve.
Development of the Atomic Emission Spectroscopy System … 25
Fig. 3 Difference of the radiation cooling time associated with the radiation rate as a function of
surface to volume ratio
Fig. 4 Solution algorism to obtain the LII signal as a function of time-temperature curve
According to the Michelsen’s model, the energy balance for the interaction of a
particle with laser is given by
where QInt represents the sensible energy storage in particle, Qabs is the rate of the
laser energy absorption, Qrad is the radiation rate by blackbody emission, Qcond
represents the energy dissipation rate by conduction, Qsub represents the rate of
energy loss by sublimation of carbon clusters and accounts for the energy con-
sumption during photodesorption of the annealed particle to form small carbon
clusters, Qann represents the energy production rate by particle annealing, and Qox
represents the energy generation rate by oxidation. Each terms of this energy flow
26 S. Ikezawa et al.
rate equation were accounted for by laser beam temporal intensity profile, time
dependence of the particle temperature, and initial states of the particle.
Time-derivative term of the particle temperature yields
dT 6
= ðQ − Qrad − Qcond − Qsub + Qann + Qox Þ ð2Þ
dt πD3 ρS cS abs
3 Device Designs
Delay/pulse
Streak camera generator
Spectrograph
Condenser lens PC
Nd:YAG laser
Particle sample
Fig. 6 Schematic of the MIP system using Beenakker cavity and two-way spectroscopic analysis
system
28 S. Ikezawa et al.
B. He-MIP-AES Method
Figure 6 shows a schematic of the MIP system. The He-MIP-AES system is a
reconstruction of an existing particle analyzer system (PT1000, Yokogawa Electric
Co., Tokyo, Japan) [23]. The He-MIP system utilizes microwaves with a frequency
of 2.45 GHz in TM010 mode. The microwaves are introduced into a Beenakker-type
cylindrical cavity resonator. Electromagnetic energy at 70–200 W is condensed at
the quartz discharge tube, which is set at the center of the cavity. The plasma
stability depends on the temperature stability of the cavity. In order to keep the
cavity temperature at 20 °C, a laminar airflow at 20 L/min is formed on the surface
of the discharge tube for cooling, and a Peltier device is attached to the discharge
tube for fine control. Preventing the introduction of other gases into the cavity is
also important for plasma stability. Helium gas is introduced to the sampling
introduction system at 45 ± 10 mL/min and to the aspirator at 300 mL/min for
optimal operation. Contaminating gases lead to shrinking of the plasma and elec-
tromagnetic polarization. This may cause insufficient plasma processing of the
samples, which may lead to failure of the quantitative analysis from the matrix
effect, molecular ion effect, and quenching phenomenon. In order to maintain
plasma purity, special techniques need to be carefully applied to collecting samples
and introducing particulates. Therefore, the particle is exposed under a helium
atmosphere at the sample introduction area of the system.
The particle is then aspirated with helium gas, and the aggregate is unpacked at
the point of the inflow area into the fast-flowing helium gas. Considering the location
and velocity of the sample particles, the fastest helium flow of the centre axis for the
cylinder (inner diameter is 3.5 mm) reaches 1.108 m/s. The flow velocity at the
centre is twice the average flow speed at a 300 mL/min flow rate. This means that the
minimum transit time of particles inside the helium plasma over a distance of 10 mm
is about 9 ms. The minimum transit time requires the optical system to measure each
particle within 9 ms. The optical system employs an originally designed switching
module to detect particles in order to reveal their chemical composition at high
sensitivities (0.33 nm resolution) or to detecting particles with unknown chemical
compositions over broad wavelength ranges (250–680 nm, 3.3 nm resolution). The
optical system for the plasma emission spectrum combines a monochromator and
photomultiplier; these are used in the high-sensitivity detection mode. In the
high-sensitivity mode, the wavelength of the spectrum peak should be a narrow band
around a known spectral line when presetting the photomultiplier. The intensity
signal of the plasma emission produced from each particle is converted to the third
root value by the electrical circuit to obtain a diameter-equivalent value. When
detecting the chemical composition of the particles, a multi-channel spectrometer is
used, and the spectra are recorded with the charged-coupled device (CCD) array.
The trigger signal output has a delay at the oscilloscope of 20 ns. The CCD requires
8–9 µs to reset its operation and 1.9 ms to process data. Thus, if the gate time of the
spectrometer is 1 ms, one measuring cycle takes 2.909 ms (344 Hz). The system
operation time is clearly within 9 ms.
Development of the Atomic Emission Spectroscopy System … 29
From shorter
wavelength side
10
From longer
wavelength side
5
Calibration
wavelength
0
253.2 253.3 253.4 253.5 253.6 253.7 253.8 253.9 254.0 254.1 254.2
Wavelength [nm]
0.4
From longer
0.3 wavelength side
Calibration
0.2 wavelength
0.1
0
312.6 312.7 312.8 312.9 313.0 313.1 313.2 313.3 313.4 313.5 313.6
Wavelength [nm]
30 S. Ikezawa et al.
From longer
0.6 wavelength side
0.4 Calibration
wavelength
0.2
0
435.4 435.5 435.6 435.7 435.8 435.9 436.0 436.1 436.2 436.3 436.4
Wavelength [nm]
0.1
0
545.6 545.7 545.8 545.9 546.0 546.1 546.2 546.3 546.4 546.5 546.6
Wavelength [nm]
[nm]
0
-0.1
-0.2
-0.3
200.0 300.0 400.0 500.0 600.0
Wavelength [nm]
Four particle sizes (diameter: 2.5, 5.0, 10, and 20 µm) were used in the exper-
iment (Table 1). For conversion to the size of a pure silicon particle, the conversion
factor from the silicon dioxide particle was set to 0.7615 based on the volume
fractions of the constituent elements. The detection wavelength for silicon was set
to 288.18 nm.
C. Introduction of Intensity Correction Factor for Other Elements
The equivalent particle diameter D has a proportional relationship to the output
voltage of the third root amplifier V. However, the intensity correction factor K for
other elements with different wavelengths should be as given in (4) b
ased on previous research works [23].
D = KV ð4Þ
Table 1 Experimental results for output voltage at greatest frequency of third root amplifier and
particle diameter
Particle diameter variation
SiO2 (µm) 2.5 5.0 10.0 20.0
Si (µm) 1.9 3.8 7.6 15.2
Output voltage of third root amplifier (V) 1.7 3.8 7.5 14.7
32 S. Ikezawa et al.
D. Limit of Detection
The limit of detection (LOD) for a particle size in the system is defined according
to the relative intensity of the atomic spectrum (intensity correction factor) and the
ground noise level. Table 2 presents a reference chart for several elements.
The measurements was taken on the rooftop deck of a building at our university
campus in Kitakyushu, Japan. The height of the roof top deck is about 10 m from
the ground. Suspended particles were collected over a 7-h period from 10:00 to
17:00 with a flow rate of 5 L/min and deposited on a filter with a pore diameter of
0.4 µm.
Figure 14 shows the premeasurement plasma spectrum of the He-MIP system
with a multi-channel spectrometer being used for wide-range spectrum finding.
Besides the helium carrier gas (He), hydrogen (H), oxygen (O) and hydroxyl radical
(•OH) were also observed. This is because of the measurement was conducted
under atmospheric pressure conditions, and moisture in the air or adhering to the
inner tube wall was introduced.
The measurement results for particles suspended in air using the He-MIP-AES
instrument are reported in Figs. 15, 16, 17, and 18. These figures show the event
number of plasma emissions with respect to individual particle sizes for elemental
Fig. 14 Plasma spectrum of He-MIP system using multi-channel spectrometer without measured
particles
34 S. Ikezawa et al.
Fig. 15 Counts of particles greater than 5 µm in size for elemental signals obtained from plasma
spectral emissions
Fig. 16 Counts of particles greater than 2 µm and less than 5 µm in size for elemental signals
obtained from plasma spectral emissions
signals classified by the equivalent grain size diameter D. The diameter was derived
from the output voltage of the third root amplifier with an intensity correction
factor.
Development of the Atomic Emission Spectroscopy System … 35
Fig. 17 Counts of particles greater than 1 µm and less than 2 µm in size for elemental signals
obtained from plasma spectral emissions
Fig. 18 Counts of particles less than 1 µm in size for elemental signals obtained from plasma
spectral emissions
36 S. Ikezawa et al.
6 Conclusions
Suspended particle measurement was demonstrated using the He-MIP system with
two-way spectroscopic analysis. The measurement results can be explained as
follows. For the particles with a diameter of over 5 µm, silicon (Si) was the most
often observed element followed by carbon (C) and potassium (K). For particles in
the size range of 2–5 µm, Si was the most often observed element followed by C,
K, zinc (Zn), iron (Fe), magnesium (Mg), and boron (B). In the size range of 1–
2 µm, C was the most often observed element followed by Fe, B, Si, sodium (Na),
phosphorus (P), Mg, zirconium (Zr), Zn, molybdenum (Mo), aluminum (Al),
nitrogen (N), silver (Ag), and manganese (Mn). Nanoparticles with diameters in the
sub-micrometer range were also detected. A large number of elements were
detected, e.g., Fe, Mg, Zn, B, C, Au, chromium (Cr), Na, copper (Cu), calcium
(Ca), Mo, chlorine (Cl), Ag, arsenic (As), vanadium (V), Si, Mn, lead (Pb), and
gallium (Ga). The results obtained with the He-MIP system confirmed that the
nanoparticles contained rare and precious metals that are not produced around the
location where measurements were conducted. This indicates that many of the
nanoparticles were either related to industrial activities or had been transported over
several thousands of kilometers. While primary particles remain suspended for a
few days and then diffuse over several kilometers, secondary particles remain in
suspension for a few weeks to a few months. Thus, it was inferred that the measured
nanoparticles had undergone reactions, in contrast to other micro-sized particles,
over repeated cycles of aggregation and disaggregation. The above method allows
the sizes and chemical compositions of the PM to be detected simultaneously; this
may aid in gaining an understanding regarding the interactions between yellow sand
and nanoparticles and the adhesive properties of carbonaceous materials in terms of
nanoparticles deposited on micro-sized particles.
Acknowledgment We are grateful to JGC Catalysts and Chemicals Ltd. for providing us with
silicon dioxide particles.
References
1. The National Institute for Environmental Studies (NIES) (2013) No. 345, 2013.2.21 <https://
www.nies.go.jp/whatsnew/2013/20130221/20130221.html>
2. K. Hara, J. Homma, K. Tamura, M. Inoue, K. Karita, E. Yano, Decreasing trends of
suspended particulate matter and PM2.5 concentrations in Tokyo, 1990-2010. J. Air Waste
Manag. Assoc. 63(6), 737–748 (2013)
3. J.H. Brown, K.M. Cook, F.G. Ney, T. Hatch, Influence of particle size upon the retention of
particulate matter in the human lung. Am. J. Public Health Nations Health, 450–458 (1950)
4. U.P. Kodavanti, M.C. Schladweiler, A.D. Ledbetter, W.P. Watkinson, M.J. Campen, D.W.
Winsett et al., The spontaneously hypertensive rat as a model of human cardiovascular
disease: evidence of exacerbated cardiopulmonary injury and oxidative stress from inhaled
emission particulate matter. Toxicol. Appl. Pharmacol. 164(3), 250–263 (2000)
Development of the Atomic Emission Spectroscopy System … 37
5. G.L. Squadrito, C. Rafael, B. Dellinger, W.A. Pryor, Quinoid redox cycling as a mechanism
for sustained free radical generation by inhaled airborne particulate matter. Free Radical Biol.
Med. 31(9), 1132–1138 (2001)
6. A.P. Carll, N. Haykal-Coates, D.W. Winsett, W.H. Rowan III, M.S. Hazari, A.D. Ledbetter
et al., Particulate matter inhalation exacerbates cardiopulmonary injury in a rat model of
isoproterenol-induced cardiomyopathy. Inhalation Toxicol. 22, 355–368 (2010)
7. S. Ikezawa, T. Ueda, Basic research on elemental and size analytical system for nano-sized
suspended particulate matter using contactless optical measurement techniques. Int. J. Meas.
Technol. Instrum. Eng. 3(1), 16–27 (2013)
8. S. Ikezawa, M. Wakamatsu, T. Ueda, Optical analytical technique for carbonaceous particles
using laser-induced electro-avalanche fluorescence and laser-induced incandescence. Sens.
Mater. 25(1), 57–77 (2013)
9. S. Ikezawa, M. Wakamatsu, T. Ueda, Development of semi-quantitative analytical system for
metal nanoparticle ink using laser-induced breakdown spectroscopy. Int. J. Smart Sens. Intell.
Syst. 5(1), 215–232 (2012)
10. S. Ikezawa, M. Wakamatsu, T. Ueda, Diesel particulate analysis using SEM-EDX and
laser-induced breakdown spectroscopy. Int. J. Smart Sens. Intell. Syst. 4(2), 174–188 (2011)
11. S. Ikezawa, M. Wakamatsu, Y. L’vovich Zimin, J. Pawlat, T. Ueda, Development of
carbonaceous particle size analyzer using laser-induced incandescence technology.
IEEJ. Trans. Sens. Micromach. 131(5), 171–177 (2011)
12. M. Wakamatsu, S. Ikezawa, T. Ueda, Particle element and size simultaneous measurement
using LIBS. IEEJ Trans. Sens. Micromach. 127(9), 397–402 (2007)
13. S. Ikezawa, J. Yamamoto, T. Ueda, Development of Helium-microwave-induced
plasma-atomic emission spectroscopy system with two-way spectroscopic analysis, in
Proceedings of the 9th International Conference on Sensing Technology, pp. 786–791 (2015)
14. J.D. Cobine, D.A. Wilbur, The electronic torch and related high frequency phenomena.
J. Appl. Phys. 22(6), 835–841 (1951)
15. M. Yamamoto, S. Murayama, UHF torch discharge as an excitation source. Spectrochim.
Acta 25A, 773–776 (1967)
16. H. Goto, K. Hirokawa, M. Suzuki, Spectrometric detection of various metals by
high-frequency plasma torch. Fresenius’ Zeitschrift fur analytische Chemie 225(2), 130–
137 (1967)
17. R. Nakashima, S. Sasaki, Determination of iron, manganese, and magnesium in high-purity
molybdenum by high-frequency plasma torch emission spectrometry. Anal. Chim. Acta 85
(1), 75–81 (1976)
18. G. Cristescu, R. Grigorovici, Optische temperaturbestimmungen an der hochfrequenten
fackelentladung. Naturwissenschaften 29(38), 571–572 (1941)
19. C.I.M. Beenakker, A cavity for microwave-induced plasmas operated in helium and argon at
atmospheric pressure. Spectrochimica Acta Part B 31(8–9), 483–486 (1976)
20. A.J. McCormack, S.C. Tong, W.D. Cooke, Sensitive selective gas chromatography detector
based on emission spectrometry of organic compounds. Anal. Chem. 37(12), 1470–1476
(1965)
21. S. Greenfield, H.D. McGeachin, P.B. Smith, Plasma emission sources in analytical
spectroscopy II. Talantn 22, 553–562 (1975). Pergamon Press
22. H.A. Michelsen, Understanding and predicting the temporal response of laser-induced
incandescence from carbonaceous particles. J. Chem. Phys. 118, 7012–7045 (2003)
23. H. Takahara, M. Iwasaki, Y. Tanibata, Particle analyzer system based on microwave-induced
plasma technology. IEEE Trans. Instrum. Meas. 44(3), 819–823 (1995)
Real-Time HVAC Sensor Monitoring
and Automatic Fault Detection System
1 Introduction
Heating, ventilation, and air conditioning (HVAC) systems are designed to provide
thermal comfort and acceptable indoor air quality in a range of commercial buildings
[1]. HVAC systems consume a large amount of energy throughout the world. For
Y. Guo (✉) ⋅ J. Li
Data61, CSIRO (Commonwealth Scientific and Industrial Research Organisation),
Marsfield, NSW 2122, Australia
e-mail: [email protected]
J. Wall ⋅ S. West
Energy, CSIRO (Commonwealth Scientific and Industrial Research Organisation),
Mayfield West, NSW 2304, Australia
Another class of solutions use model-based methods, which use analytical mathe-
matical models to compare and identify faults based on the sensor data sets [11–13].
Since every building is unique, it is not a simple task to set these rules or to
generate these analytical mathematical models. In addition, the task of setting these
thresholds used by such solutions to raise alarms is quite involved, and prone to
producing false alarms. Hence, we developed a statistical machine learning algo-
rithm based approach for the automated monitoring and fault detection in HVAC
systems [14–17]. Our approach uses probabilistic models that are constructed on
the probabilistic links between variables. Meanwhile, the probabilities are learnt
from the stored sensor monitoring data sets. It is an ideal representation for com-
bining prior knowledge and data, and can have much better flexibility and adapt-
ability when applied to HVAC system.
For a complex HVAC system, the number of sensors and actuators can number
in their thousands, and the selection of key sensors and actuators containing the
main feature of the system and reflect to important system faults, is crucial for the
success of our approach. This paper presents automated monitoring and fault
detection techniques, and a key sensor sets selection approach to optimise the fault
detection results. This methodology has been implemented and tested in
real-world commercial buildings and experimental results show that different types
of faults are detected successfully.
Fig. 2 The view of an Air Handling Unit (AHU) of a HVAC system from the Building
Management & Control System interface.
Fig. 3 Overview of the real-time sensor monitoring and fault detection approach
44 Y. Guo et al.
where j denotes the observation index, nj denotes the number of data (volume), and
a particle position matrix X
X = ½x1 . . . xi . . . xni ,
where i denotes the particle index, ni denotes the number of particles, high
dimensional voronoi tessellations are performed on the data such that each obser-
vation in Y is mapped to the nearest particle. In other words, each particle xi governs
a voronoi cell of the set Ci :
Using the concept of charged particles [22], the possibility of creating duplicate
partitions is minimised. Ideally each CXm would then return a unique partition of the
data such that
CX1 ≠ C X2 ≠ . . . ≠ CXm ,
where each partition CXm denotes an optimal partition returned by the mth swarm.
After the clustering process, the label matrix is fuzzified based on the distance
between particles and data D = DðX, YÞ. The fuzzy membership value for the jth
observation with respect to the ith cluster, uij , can be calculated as followings:
e − dij ̸ð2λi Þ
uij = ð1Þ
∑ni =i 1e − dij ̸ð2λi Þ
where dij is the distance between the ith particle to the jth observation, and λi
denotes the ith bandwidth of the cluster centre. Here λ1, ..., n can be optimised using a
compromise between the partition’s fuzzified dissimilarity
46 Y. Guo et al.
for each i and j. In other words, for each cluster, Ci , the optimum λi can be found by
solving a convex optimisation problem:
which optimised λi for all cells, i = 1, . . . , ni , that the Gaussian probability distri-
bution of the data governed in each corresponding voronoi cell is best described. In
this approach, we use non-linear least square to optimise Eq. (2).
The examples in Sect. 4 will show that the feature selection is a powerful tool
which can not only determine the feature cluster number, but also rank each feature
within each cluster. The feature with the highest ranking in each cluster is then
chosen as the key feature for fault detection.
Intelligent fault detection methods: The main processes of fault detection
algorithm work as two steps, training process step (or called learning), and fault
detection step (or called testing).
The historical sensor data which measures the normal operation status of the
system was collected for training process. The intelligent sensor and feature
selection process (as in Sect. 3A) is firstly implemented to prepare the training
datasets. A statistical machine learning approach then learns the relativity between
sensor measurements and system performance through these data. This approach
uses probabilistic models that consist of variables and probabilistic links, which can
denote the physical relationship between the sensor readings. Because of the
complexity of the HVAC system, multiple models are learnt during the training
process. The training process can be done during nights or weekends, and it nor-
mally takes about ten minutes over one week’s datasets for one building. Because
the system performance might change slowly because of season, weather, or other
conditions, the training process can be repeated at a regular basis. For instance, the
current training process can be done every week on tested buildings.
The fault detection process is to detect whether there is a fault in the system or
not (binary classification problem). After the training process, a learnt normality
model of the HVAC system can be used to detect system faults automatically. The
collected sensor data is periodically detected with the models built based on the
sensor features, and the similarity between current measurement and the historical
feature is calculated. If the algorithm finds that the difference is obvious, a fault
alarm is raised.
As the training process is a binary classification problem on time-series sensor
features, we implemented a methodology which is based on a combination of the
HMMs and the Support Vector Machines (SVM) algorithms [24]. HMMs are used
Real-Time HVAC Sensor Monitoring … 47
to denote the physical relationship between the sensor data in a dynamic system.
Meanwhile, SVM can handle the nonlinear behavior very well, and requires small
amount of training samples [16, 25, 26].
Furthermore, while the faults are detected as deviations in the normal operation,
they can be the input data for training one or more of the fault detection models to
learn patterns of faulty operation of the HVAC system wherein normal operation is
detected as deviations in the faulty operation. As a dynamic process, the normality
model and the fault models are adapted while the HVAC BMS dataset increases,
hence the performance of real-time monitoring and fault detection improves.
The selected features of the sensor data are tested on different pre-trained HVAC
normality models. Then the likelihood matrix is calculated as the indicator of the
similarity between current sensor readings and normal system sensor readings.
A couple of methodologies such as clustering and data fusion algorithms are also
implemented as the final stage in the analysis of the likelihood matrix, and make the
final decision as a sequence of binary value (e.g. Yes for normal, and No for
abnormal). More details of the training methodologies are in [6, 15, 27].
Main faults for HVAC systems: It is reported that a few top listed faults are the
major faults for HVAC systems [28]. Properly detect these faults can avoid up to
30% of energy waste in HVAC systems [29]. A short description of each fault is
listed as following, and our sensor monitoring and fault detection approach will
mainly focus on them.
• Hot water valve leaking or stuck: If the hot water valve stuck, hot air cannot be
provided properly. If an internal valve leaks, it can be very difficult to trou-
bleshoot and is often confused with a compressor that is not pumping to
capacity. Both a leaking valve and a failing compressor have the same symp-
toms—both the heating and cooling capacity of the system are diminished. This
is because the compressor continues to pump the air around and around inside
the leaking valve.
• Supply-air fan belt slipping: The supply-air fan belt slipping can lead to the
supply-air fan not running at setting speed.
• Outside air damper leaking or stuck: Damper not in the proper position. Outside
air cannot circulate properly.
• Return air damper leaking or stuck: Return air cannot circulate properly.
• Individual zone temperature sensor fault: The temperature sensor reading in a
zone is wrong.
4 Experimental Results
The intelligent real-time sensor monitoring and fault detection system has been
tested on several buildings. In this section, we will show the results in one of these
buildings, a large commercial building in Newcastle, Australia. The sensors are
sending real-time data to the monitoring and fault detection system with a 1-min
48 Y. Guo et al.
interval which is typical of the resolution of trended sensor data from a BMS
database [13]. Figure 4 shows an example of 15 sensors for an air handling unit
(AHU).
Fault detection experimental results: During August and September 2013, 35
faults in four different fault types were generated for one AHU. Most faults last over
six hours.
Table 1 listed the summary of the experimental results. Two types of faults, hot
water valve stuck and individual zone air temperature failure are 100 % detected.
But the success rate for the slipping supply-air fan belt fault is low. The lowest,
though, is the return air/outdoor air damper point stuck fault, where only one third
faults are detected. One of the main reasons is the measurement of the effects of
each fault. For some faults, the return air qualities are affected directly, and cor-
responding sensor measurements are collected, such as return air temperature. For
the return air damper, when it stuck at 70 %, the main facts that can change
accordingly should be outdoor air fraction or indoor air CO2. These facts were not
measured or saved in the datasets for modelling.
Sensor monitoring and feature selection results: We have also analyzed the
sensitivity of the sensor features for the fault detection in HVAC systems. The
parameters that we judge include the data sample rate, training window size,
Fig. 4 An example of the sensor data for an air handling unit in the Newcastle building
Real-Time HVAC Sensor Monitoring … 49
missing data tolerance, and minimal sensor subset performance. We chose the high
frequency datasets for comparison experiments. The original dataset’s sample rate
is at one reading per 5 s, which is higher than most real HVAC systems. From the
original trials, we observed that the training/testing window size is relevant to the
sample rate. Hence we analyze the effects of the combination of these two
parameters. We changed the sample rate to the following: 5 s, 20 s, 30 s, 40 s, 50 s,
60 s, 80 s, and 90 s. The training/testing window size changes between the fol-
lowing values: 10, 20, 50, 80, and 100. The combination of sensors is the same as
default (six sensors).
Table 2 summarize the performance of the combination of these two parameters.
Tick for successful detection, cross for unsuccessful detection. As we can see, when
the sample rate increases, the window size need to be shorter to get the successful
results. For the 90 s sample rate, the window size should be 20, which means that
the training and testing data covers a 30–min period.
Table 2 Sensitivity of the sample rate versus window size of the sensor monitoring rate. Ticks for
successful fault detection
Sample rate → 5s 20 s 30 s 40 s 50 s 60 s 80 s 90 s
Window size ↓
10 × × × × × × × ×
20 × × × √ √ √ √ √
50 × × √ √ √ √ √ ×
80 × √ √ × × × × ×
100 √ √ × × × × × ×
50 Y. Guo et al.
Case study for sensor selection and its effect on fault detection: In this
subsection, we will present one example to show the effect of sensor selection on
fault detection results. The detail of this fault is as following:
• Date: 20/11/2009
• Location: All AHUs for Newcastle building
• Description: This is an actual fault that occurred when the chilled water system
was under heavy load. Chiller 2 failed due to an overheating safety cut-out, and
Chiller 1 failed to ramp up due to some misconfigured set-points, causing the
chilled water temperature to rise to around 23 °, and in turn causing the chilled
water valves on many zones (AHU9 included) to open up to near 100 %.
This type of faults is relatively common for HVAC systems, and can lead to a
large waste of energy. The sensors that are selected to build the HMMs by the
intelligent algorithms described in Sect. 3. A are: hot water valve sensor, chilled
water valve sensor, supply air relative humidity sensor, supply air temperature
sensor, return air relative humidity sensor, return air temperature sensor.
HMMs are trained on normal fault-free historical data. Then the real-time sensor
monitoring data is fed in to calculate the likelihood. The lower the likelihood value,
the higher possibility a fault exists.
The fault detection results are shown in Figs. 5 and 6. Figure 5 is the likelihood
curves for the data on 20th Nov 2009, and the corresponding fault detection results
after a classification and clustering process is shown in Fig. 6. It is clear that the
fault, occurring after 12:00 pm, is convincingly detected, as shown by the likeli-
hood curves.
3000
2000
1000
-1000
-2000
-3000
-4000
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
2
class ID
0
09:00 12:00 15:00 18:00
Fig. 6 Detection results for the fault after clustering and classification on the likelihood curves
If we manually remove one sensor feature from the optimized sensor combi-
nation, the fault would then become difficult to be detected. One example is shown
in Figs. 7 and 8, for the same fault datasets—where the supply air temperature
sensor data is not used for the HMM learning process. The likelihood curves drop
to lower levels earlier that day, and the classified fault period is not correct. This
example shows the importance of proper sensor features. Our approach can auto-
matically select suitable sensor features rather than depend on domain expert, and
assure successful detection results.
The above example proves that ERCE, as the data-driven sensor and feature
selection algorithm, is crucial for the fault detection results. In fact, ERCE specifi-
cally performs well under varying seasonal conditions. The ERCE method selects
features that are unique and relevant to the faults in the HVAC system. For different
seasons, ERCE selects different combination of sensor data as the key features.
Pattern for the winter dataset is different from the pattern for the summer dataset. The
experimental results show that the sensor and feature selection process can ensure
-1500
-2000
-2500
loglikehood
-3000
-3500
-4000
-4500
-5000
-5500
-6000
09:00 12:00 15:00 18:00
2
class ID
0
09:00 12:00 15:00 18:00
the success of the fault detection results for the HVAC system. It is proven that the
ERCE method can improve the fault detection results, comparing to the results based
on other feature selection clustering method, such as EAC k-means [18].
References
1. M. Rolloos, HVAC systems and indoor air quality. Indoor Built Environ. 2(4), 204–212
(1993)
2. A.G. Department of industry, innovation and science, in Heating, Ventilation and
Air-conditioning (2016). http://industry.gov.au/Energy/EnergyEfficiency/Non-
residentialBuildings/HVAC/Pages/default.aspx
3. K.W. Roth, J. Dieckmann, S.D. Hamilton, W. Goetzler, in Energy Consumption Charac-
teristics of Commercial Building HVAC Systems Volume III : Energy Savings Potential
(2002)
4. V. Vakiloroaya, B. Samali, A. Fakhar, K. Pishghadam, A review of different strategies for
HVAC energy saving. Energy Convers. Manag. 77, 738–754 (2014)
5. N. Gas, Renewable Energy Annual 2002,” vol. 0603 (2002)
6. J. Wall, Y. Guo, J. Li, S. West, A dynamic machine learning-based technique for automated
fault detection in HVAC systems, in 2011 ASHRAE Annual Conference (2011)
7. T.M. Rossi, J.E. Braun, A statistical, rule-based fault detection and diagnostic method for
vapor compression air conditioners. HVAC&R Res. 3(1), 19–37 (1997)
8. J. Schein, S. Bushby, N. Castro, J. House, A rule-based fault detection method for air handling
units. Energy Build. 38, 1485–1492 (2006)
9. P.P. Angelov, R.A. Buswell, V.I. Hanby, J.A. Wright, A methodology for modeling HVAC
components using evolving fuzzy rules, in 2000 26th Annual Conference of the IEEE
Industrial Electronics Society IECON 2000, vol. 1, no. 1, (2000), pp. 247–252
10. Y. Song, Y. Akashi, J.-J. Yee, A development of easy-to-use tool for fault detection and
diagnosis in building air-conditioning systems. Energy Build. 40(2), 71–82 (2008)
11. B. Yu, D.H.C. Van Paassen, S. Riahy, General modeling for model-based FDD on building
HVAC system. Simul. Pract. Theory 9(6–8), 387–397 (2002)
12. J. Liang, R. Du, Model-based fault detection and diagnosis of HVAC systems using support
vector machine method. Int. J. Refrig. 30(6), 1104–1114 (2007)
13. H. Yoshida, S. Kumar, ARX and AFMM model-based on-line real-time data base diagnosis
of sudden fault in AHU of VAV system. Energy Convers. Manag. 40(11), 1191–1206 (1999)
14. D. Dehestani, S. Su, H. Nguyen, Y. Guo, J. Wall, F. Eftekhari, Comprehensive sensitivity
analysis of heat ventilating and air conditioning (HVAC) system based on neural network
model, in 10th International Conference on Healthy Buildings, Brisbane, Australia, 8–12 July
2012, pp. 6–11
15. Y. Guo, J. Li, S. West, J. Wall, G. Platt, System and method for detecting and/or diagnosing
faults in multi-variable systems, US20120185728 A1 (2012)
16. Y. Guo, J. Wall, J. Li, S. West, A machine learning approach for fault detection in
multi-variable systems, in The Tenth International Conference on Autonomous Agents and
Multiagent Systems, (2011), pp. 23–30
17. Y. Guo, D. Dehestani, J. Li, J. Wall, S. West, S. Su, Intelligent outlier detection for HVAC
system fault detection, in 10th International Conference on Healthy Buildings, Brisbane,
Australia, 8–12 July, 2012, no. Vapnic (1995), p. 2
18. M. Yuwono, Y. Guo, J. Wall, J. Li, S. West, G. Platt, S.W. Su, Unsupervised feature selection
using swarm intelligence and consensus clustering for automatic fault detection and diagnosis
in Heating Ventilation and Air Conditioning systems. Appl. Soft Comput. 34, 402–425 (2015)
54 Y. Guo et al.
19. Z. Yu, F. Haghighat, B.C.M. Fung, L. Zhou, A novel methodology for knowledge discovery
through mining associations between building operational data. Energy Build. 47, 430–440
(2012)
20. A. Kusiak, M. Li, F. Tang, Modeling and optimization of HVAC energy consumption. Appl.
Energy 87(10), 3092–3102 (2010)
21. N. Gaitani, C. Lehmann, M. Santamouris, G. Mihalakakou, P. Patargias, Using principal
component and cluster analysis in the heating evaluation of the school building sector. Appl.
Energy 87(6), 2079–2086 (2010)
22. M. Yuwono, S.W. Sir, B.D. Moulton, Y. Guo, H.T. Nguyen, An algorithm for scalable
clustering: ensemble rapid centroid estimation, in Proceedings of the 2014 IEEE Congress on
Evolutionary Computation, CEC 2014, (2014), pp. 1250–1257
23. M. Yuwono, S.W. Su, B.D. Moulton, H.T. Nguyen, Data clustering using variants of rapid
centroid estimation. IEEE Trans. Evol. Comput. 18(3), 366–377 (2014)
24. M.A. Hearst, S.T. Dumais, E. Osman, J. Platt, B. Scholkopf, Support vector machines. IEEE
Intell. Syst. 13(4), 18–28 (1998)
25. J. Li, Y. Guo, J. Wall, S. West, Fault Detection for HVAC Systems, in 10th International
Conference on Healthy Buildings, Brisbane, Australia, 8–12 July 2012, pp. 12–15
26. D. Dehestani, F. Eftekhari, Y. Guo, S. Ling, S. Su, H. Nguyen, Online support vector machine
application for model based fault detection and isolation of HVAC system. Int. J. Mach.
Learn. Comput. 1(1), 66–72 (2011)
27. S.R. West, Y. Guo, X.R. Wang, J. Wall, Automated fault detection and diagnosis of HVAC
subsystems using statistical machine learning, in 12th International Conference of the
International Building Performance Simulation Association (2011)
28. S. Katipamula, M.R. Brambley, Methods for fault detection, diagnostics, and prognostics for
building systems—A review, Part I. HVAC&R Res. 11(1), 3–26 (2005)
29. EIA, International Energy Outlook 2009 (2009)
High Sensitivity Optical Structures
for Relative Humidity Sensing
Abstract This chapter is focused in the different optical structures and materials
that have been used for the development of optical fiber humidity sensors. First, we
will start with a short introduction of what relative humidity is, and why it has been
extensively investigated. We will make also a brief summary of the different
options that have been developed by now, showing the evolution of this research
field. Then we will look more closely at the most used structures, the most common
materials and the devices having greater sensitivity and resolution.
1 Introduction
Fig. 1 Evolution of the optical fiber humidity sensors contributions (Source Scopus)
ature is increased [1]. In this chapter we will always refer to relative humidity
although sometimes it will be mentioned only as humidity.
In Fig. 1, it is shown the evolution of the contributions found in Scopus, related
to optical fiber humidity sensors, from 1982 to 2015. As it can be seen, the number
of contributions has increased continuously since 1999 keeping today a number
worthy of consideration.
This increasing interest in optical fiber humidity sensors is due to the intrinsic
advantages of using optical fibers as the sensing structure. These advantages
include: small size, lightness, the possibility of working on flammable environ-
ments, the ability of working on greater pressure and temperature ranges and, the
most important, their electromagnetic immunity.
2 Materials
Several materials have been used for the development of optical fiber humidity
sensors, from polymers to metal oxides, from hydrophilic to hydrophobic materials.
Hydrophobic materials, which are composed of non-polar molecules, present
greater contact angles than the hydrophilic ones. Both of them have been used for
the development of humidity sensors [2, 3].
However, for the development of optical fiber sensors, there are other behaviors
which provide better performance from an optical point of view. These behaviors
are called swelling and adsorption. Swelling is a process whereby a material
increases in size when subjected to increasing humidity. This increased thickness
provokes that the light traveling through the core of an optical fiber found a greater
effective refractive index, which will affect at the transmitted light. The other
physical phenomena, adsorption, is defined by the IUPAC as an increase in the
concentration of a dissolved substance at the interface of a condensed and a liquid
phase due to the operation of surface forces. Adsorption can also occur at the
interface of a condensed and a gaseous phase [4]. For humidity measurements, the
High Sensitivity Optical Structures for Relative Humidity Sensing 57
result of adsorption is the generation of a thin film of water over the surface of the
material. Swelling is common for some polymers while adsorption is the pre-
dominant effect in metal oxides.
It does not mean that the previously commented physical phenomenon are the
only ones that can be used for measuring relative humidity by an optical set-up.
There are other phenomenon used for humidity sensing as it can be fluorescence
[5], but in this chapter we will focus on evanescent wave sensors, fiber gratings,
sensors based on Lossy Mode Resonances and Fabry-Pèrot interferometers, with
coatings based on adsorption and swelling.
3 Conventional Methods
There are conventional methods for measuring the relative humidity and we will
make a brief overview of those methods. Different physical phenomena have been
used for this methods and they can be classified on these categories:
• Wet and dry bulb psychrometer
• Mechanical
• Electrical
• Infrared (IR) optical absorption hygrometer
The wet and dry bulb psychrometer consists of two thermometers, one of which
is covered with a damp wick (wet bulb). The difference between the measured
temperatures is due to water evaporation and latent heat transfer [1, 6] and the
relative humidity can be calculated with the psychrometric diagram.
Mechanical humidity sensors are based on the expansion/contraction of certain
materials. A well-known humidity sensor based on mechanical behavior is the
mechanical hygrometer. Common materials used include synthetic fibres and
human hair. The first mechanical hygrometers were formed by linking the material
with a weak tension by a spring, and a mechanical linkage operates a needle. Other
mechanical hygrometers were formed by linking the material to strain gauges or
other devices to measure the displacement caused by the change in humidity [6].
Resistive and capacitive-type humidity sensors are included into the electrical
category. Resistive-type humidity sensors are based on changes on the electrical
resistivity of some materials as conductive polymers, semiconductors or compos-
ites. This behavior has been explained in [7–9] by the water adsorbed on the oxide
surface in the molecular and hydroxyl forms. Furthermore, Boyle and Jones [7]
established that water demonstrates donor properties at adsorption in the molecular
form while it shows acceptor properties in the hydroxyl form [10]. In the same way
capacitive-type sensors are based on the dielectric changes of a thin film. Common
materials used for capacitive sensors are ceramics like alumina [11] and polymers
[12]. The capacitive- and resistive-based humidity sensors are the most commonly
used, with capacitive-based sensors dominating and making up nearly 75 % of the
commercial humidity sensor market [13].
58 J. Ascorbe et al.
Fig. 2 a Mechanical humidity sensor, b wet and dry bulb psychrometer, c capacitive humidity
sensor, d IR optical absorption method and e resistive humidity sensor
Finally, the infrared (IR) optical absorption hygrometer is based on the optical
absorption of the water vapor. It works by emitting light at two different wave-
lengths. One of them has negligible absorption and it is used as reference. The
transmission ratio between those two wavelengths provides a direct measurement of
water absorption, making this technique insensitive to contaminants [6] (Fig. 2).
There are many ways of measuring the relative humidity by means of optical
set-ups, one of them has been mentioned before. Here, we are going to explain in
detail those based on optical fibers and most specifically, those involving coatings
onto an optical fiber. Among those devices we will focus on Fabry-Pèrot inter-
ferometers and Lossy Mode Resonances based devices although some other
structures will be mentioned to have a sight of the differences between them and the
evolution that optical fiber humidity sensors have suffered. Optical fiber humidity
sensors (OFHS) can be classified by the physical phenomena they used as:
• Evanescent wave sensors
High Sensitivity Optical Structures for Relative Humidity Sensing 59
– Optical absorption
– Change on the refractive index
– Lossy mode resonances (LMRs)
• Fluorescence sensors
• Interferometric sensors
• Fiber gratings
First of all, we are going to introduce the earlier optical fiber humidity sensors,
classified as evanescent wave (EW) sensors. The evanescent field is created as a
result of the standing wave generated by the interference between the incident and
the reflected signals at the core/cladding interface, inside an optical fiber, due to the
Total Internal Reflection (TIR) phenomena [6]. The amplitude of the evanescent
field decays exponentially with the distance away from the core/cladding interface
and it can be calculated as:
where the penetration depth (dp) is defined as the depth at which the amplitude of
the evanescent field, E, has decayed to 1/e of the initial value E0 at the
core/cladding interface and this is given by
λ
dp = h i1 ̸2
ð2Þ
2πncore ðsin θ − ðnclad ̸ncore Þ2
2
where λ is the wavelength of the propagating signal in the optical fiber, θ is the
angle of incidence normal at the interface, ncore and nclad are the refractive indices
of the fiber core and cladding, respectively.
There are two different approaches: the first one based on changes on the optical
absorption and the second one based on changes of the refractive index. The main
difference between these two methods is that optical absorption causes changes at a
certain wavelength or in a wavelength range and when changing the refractive
index there are changes at all wavelengths. Earlier optical fiber humidity sensors
were developed by the first method since 1985 using, for example, cobalt chloride
[14] or other chemical reagents immobilized in a polymeric matrix, and cladding
removed optical fibers. The problem with these sensors is that the evanescent tail
has only a small fraction of power, and a long portion of fiber must be coated to
have a detectable change at the output end [15]. For that reason some other
structures have been developed to increase the amplitude of the evanescent field and
consequently the sensitivity of the device. For example U-bent fibers [15] or porous
60 J. Ascorbe et al.
Fig. 3 a Humidity sensor structure. Taper profile with ESA overlay and b Optical output
spectrum dependence with respect to the relative humidity with 23 bilayers deposited. Tapered
fiber waist diameter: 20 μm. Taper length: 2.2 mm. Nanofilm: [PDDA+/Poly-R−]. Reprinted from
Sens. Actuators B, vol. 122, no 2, Corres, J.M., Arregui, F.J., and Matías, I.R., Sensitivity
optimization of tapered optical fiber humidity sensors by means of tuning the thickness of
nanostructured sensitive coatings, pp. 442–449, (2007), with permission from Elsevier
fibers [16] have been used to develop more sensitive OFHS based on optical
absorption.
Several different approaches have been followed to develop evanescent wave
sensors by changing the refractive index of the coating. Removed plastic cladding
optical fibers [17], tapered single mode [18, 19] optical fibers or D-shape optical
fibers can be used for this kind of devices. Common materials used for these OFHS
are hydrogels as poly-hydroxyethyl methacrylate, poly-acrylamide, poly-N-vinyl
pyrrolidinone and agarose [17, 18]. As it can be seen in Fig. 3, extracted from [19]
changes are produced at all wavelengths with this method.
Optical fiber sensors based on Lossy Mode Resonances are a special case of
evanescent wave sensors and, more specifically, a special case of EW sensors based
on changes of the refractive index. Here, the material chosen for coating the optical
fiber has to meet two conditions: to be sensitive to the physical parameter for what
the sensor is developed and to meet the requirements to generate LMRs. This
method, which is included in the wavelength based techniques, overcome the
limitations of the previous EW methods, such as external noises or power fluctu-
ations [20].
Lossy Mode Resonances are a physical phenomenon which consists on the
absorption of the light at certain wavelength. It is important to remark that this
absorbed light is not related to the absorbed light mentioned before. For EW sensors
based on optical absorption, the absorption of the light is produced when the
coating is subjected to the physical parameter to be measured. However, once the
LMR has been generated, the optical absorption occurs without need of an external
High Sensitivity Optical Structures for Relative Humidity Sensing 61
As it has been mentioned before, metallic oxides are good candidates for the
generation of LMRs and they are also a good choice for humidity sensing because
of their absorption capability. Most of the research done for LMRs has been
developed coating a 200/225 µm core/cladding plastic cladding optical fiber. This
optical fiber allows to transmit light in the range 400–2200 nm and the cladding can
be removed by simple chemical methods. First LMR devices were developed by
coating an optical fiber with ITO and they were characterized as humidity sensors
[32] and also as refractometers [33]. Then a typical transmission set-up, Fig. 5, was
used while the optical fiber was subjected to changes on the relative humidity and
Fig. 4 a Scheme of lossy mode resonances generation, b dielectric and refractive index
conditions for LRSEP, SPR and LMR generation and c graphic representation of previously
mentioned conditions
the wavelength shift of the LMR was observed. A SEM image of the coating can be
seen in Fig. 6.
The wavelength shift of the LMR is caused by changes on the refractive index
(RI) of the coating. Due to the absorbed water on the ITO surface the effective
refractive index increases, and as the LMR wavelength depends on that, its
wavelength position shifts to greater wavelengths. It seems to be a general trend that
for an increase of the refractive index there is a redshift of the LMR, although there
might be exceptions. Main contribution for these OFHS is due to changes on the RI
of the coating, although it may be considered that the RI of the surrounding medium
is also altered, it is negligible by comparison. Figure 7 shows the dynamic beha-
viour of these kind of devices.
The previous situation uses the dependence of LMRs on the RI of the coating.
But OFHS can be also developed taking advantage of the dependence of LMRs on
High Sensitivity Optical Structures for Relative Humidity Sensing 63
Fig. 6 SEM image of a ITO coating and b In2O3 coating deposited onto the optical fiber core.
Reprinted from Lossy mode resonances toward the fabrication of optical fiber humidity sensors,
P. Sanchez et al., 2012, Meas. Sci. Technol. 23 014002 doi:10.1088/0957-0233/23/1/014002 with
permission from IOP Publishing
Fig. 7 Dynamical response of ITO sensors from 20 to 80 % of RH. Reprinted from Lossy mode
resonances toward the fabrication of optical fiber humidity sensors, P. Sanchez et al., 2012, Meas.
Sci. Technol. 23 014002 doi:10.1088/0957-0233/23/1/014002 with permission from IOP
Publishing
the RI of the surrounding medium. For that purpose, some devices have been
developed using ITO for the generation of LMRs and using other materials like
PAH/PAA as the sensitive layer [20, 28]. These devices combine the high sensi-
tivity of LMRs to changes on the external RI with the greater changes on the RI of
PAH/PAA when subjected to high humidity, improving the sensitivity of the device
to changes on the relative humidity. This combination of two different layers allows
to have the LMR located at greater wavelengths, which also enhances the sensi-
tivity of the device. Figure 8 shows the absorption spectrum for this device at two
different relative humidity values and its dynamic response.
64 J. Ascorbe et al.
Fig. 8 Spectral response of the a non-tuned and b tuned sensors for 20 and 90 % RH and
dynamical response of the sensors to changes in the RH of the external medium c non-tuned sensor
(20 PAH/PAA bilayers) and d tuned sensor (100 PAH/PAA bilayers). Reprinted from Sens.
Actuators B, vol. 146, no 1, Zamarreño, C.R., Hernaez, M., Del Villar, I., Matias, I.R., and
Arregui, F.J., Tunable humidity sensor based on ITO-coated optical fiber, pp. 414–417, (2010),
with permission from Elsevier and IOP Publishing
Standard SMFs has some advantages respect plastic cladding multimode optical
fiber, and one of them is its price. On the other hand it is more difficult to make a
SMF sensitive to external parameters because its cladding it is made of silica
instead of plastic. But there are some methods to obtain access to the evanescent
field of a SMF. One of these methods consists on tapering the SMF by applying
heat and stretching. The heat can be applied by means of a flame or by an electric
arc [34–36]. When a single-mode fiber (SMF) is tapered, the core/cladding interface
is redefined in such way that the SMF in the tapered region acts as a multimode
fiber, sustaining several LP0 m modes. In this case, the light propagation in the
waist of the taper extends to the whole fiber, which plays the role of the new core,
and the new cladding is the external medium; this fact eases the interaction of the
evanescent field with the outer medium so that small changes of the outer medium
influence the transmitted optical power. Some OFHSs have been developed using
this structure [19, 34–36], most of them based on optical power measurements [19,
35] and others based on resonances [36]. As it can be seen in Fig. 9 from [36], a
tapered SMF can be sensitive to external parameters if coated with the appropriate
material, indium tin oxide for this case.
There are other ways to gain access to the evanescent field of a SMF as using a
D-shape optical fiber [27], but here we are going to focus on another structure that
has been recently used [37], which consists of a cladding etched single mode optical
fiber (CE-SMF) coated with a semiconductor oxide thin-film. As contrast to tapered
SMF, that keeps the ratio core/cladding, in the CE-SMF the core is not narrowed,
High Sensitivity Optical Structures for Relative Humidity Sensing 65
Fig. 9 a Evolution of the transmitted optical spectrum of a tapered SMF coated with ITO for
b some cycles of 20–90 % RH and c optical spectra for some values of RH [36]
Fig. 10 Cladding-etched single mode optical fiber. Reprinted from Sens. Actuators B, vol. 233,
Ascorbe, J., Corres, J.M., Matias, I.R., and Arregui, F.J., High sensitivity humidity sensor based on
cladding-etched optical fiber and lossy mode resonances, pp. 7–16, (2016), with permission from
Elsevier
Fig. 11 a Some transmitted optical spectra for different thicknesses of In2O3 nano film coated
onto a cladding etched single mode optical fiber and b changes on the transmitted optical spectrum
for increasing thickness of In2O3 nano film
This coated CE-SMF was then connected to the light source and to an optical
spectrum analyzer (OSA). The light source was a super-luminescent emitting diode
(SLED) HP-83437A white light source which has four LEDs at 1200, 1310, 1430
and 1550 nm and the OSA was the HP-86142A, which has a wavelength range
between 600 and 1700 nm, although 1200–1600 nm range was used. The optical
fiber was introduced in a climatic chamber (Angelantoni ACS CH 250), as it can be
seen in Fig. 12.
Fig. 12 Experimental set-up for measuring changes on the transmitted optical spectrum while
relative humidity changes. Reprinted from Sens. Actuators B, vol. 233, Ascorbe, J., Corres, J.M.,
Matias, I.R., and Arregui, F.J., High sensitivity humidity sensor based on cladding-etched optical
fiber and lossy mode resonances, pp. 7–16, (2016), with permission from Elsevier
High Sensitivity Optical Structures for Relative Humidity Sensing 67
Fig. 13 a Optical spectra for some values of RH, b evolution of the transmitted optical spectrum
of a CE-SMF coated with ITO for c some cycles of 20–90 % RH
Fig. 14 Wavelength position of the LMR and relative humidity plotted simultaneously as a
function of time for previous device
Fig. 15 a Optical spectra for four cycles of 20–90 % RH, b relative humidity cycles inside the
climatic chamber and c optical spectra for some values of RH. Reprinted from Sens. Actuators B,
vol. 233, Ascorbe, J., Corres, J.M., Matias, I.R., and Arregui, F.J., High sensitivity humidity sensor
based on cladding-etched optical fiber and lossy mode resonances, pp. 7–16, (2016), with
permission from Elsevier
High Sensitivity Optical Structures for Relative Humidity Sensing 69
In this section of the chapter we are going to focus on optical fiber humidity sensors
based on Fabry-Pèrot interferometers (FPI). There are a lot of different ways to
generate interferometry. One of the most known interferometers is the
Mach-Zehnder interferometer, which measures the phase difference between light
propagating through two arms of different length [41]. These kind of interferom-
eters can be obtained by splitting the optical waveguide and recombining it later, by
splicing different kind of fibers [42] or by tapered optical fibers [43]. This inter-
ferometers [42, 43] are based on modal interferometry, which is generated between
core modes and cladding modes. But there is another kind of optical fiber inter-
ferometers which are called Fabry-Pèrot interferometers (FPIs). FPIs can be clas-
sified into extrinsic and intrinsic interferometers. In the intrinsic FPIs (IFPIs) the
cavity is contained within the fiber and in the extrinsic FPIs (EFPIs) sensors the
cavity is external to the fiber. Both of them are based on the interference caused by
multiple reflections of light between two reflecting surfaces. Constructive inter-
ference occurs if the transmitted beams are in phase, and this corresponds to a
high-transmission peak of the etalon. If the transmitted beams are out-of-phase,
destructive interference occurs and this corresponds to a transmission minimum.
Whether the multiply reflected beams are in phase or not depends on the wave-
length of the light, the angle the light travels through the etalon, the thickness of the
etalon and the refractive index of the material between the reflecting surfaces [44]
(Fig. 16).
Fabrication methods for EFPIs are inexpensive and simpler than those used for
the fabrication of IFPIs. Regarding the fabrication of EFPIs different approaches
have been followed to obtain the interferometer. Most used structure consists of an
air gap between two optical fibers [45, 46]. Hollow core fiber spliced to a
single-mode optical fiber (SMF) and filled with a sensitive material [47] has been
also used. But the most robust and reliable method for developing an EFPI is
generating a nano-cavity by coating the tip of the SMF [48–50]. Although we are
going to focus on a sputtered coating, other methods have been used for the
fabrication of Fabry-Pèrot interferometers, as dip-coating [47], e-beam evaporation
[51] or electrostatic self-assembly [49].
In an interferometric cavity the mirrors are formed by the refractive index dif-
ferences between the different optical media at either end. In this case by the
refractive index differences between the optical fiber and the SnO2 coating, and then
between the SnO2 coating and the external medium, which is air for this case. The
reflectance of each mirror will be approximately that determined by Fresnel’s law
for the case of normal incidence,
ðn1 − nÞ2
R1 = ð3Þ
ðn1 + nÞ2
70 J. Ascorbe et al.
ðn2 − nÞ2
R2 = ð4Þ
ðn2 + nÞ2
and
4πnL
θ= ð6Þ
λ
where L is the length of the etalon and λ is the wavelength of the excitation light
[48].
High Sensitivity Optical Structures for Relative Humidity Sensing 71
Here we are going to show the behaviour of a FPI obtained by coating the tip of an
optical fiber with tin dioxide [52]. Tin dioxide is a good choice for humidity sensing
because of its absorption capability and it is also a good choice for the FPI gen-
eration because of its high refractive index. First of all, in Fig. 17 it is shown a
simulation of how the interferometry is generated and how it changes as the
thickness of the etalon increases.
First of all, just one interference peak appears, which shifts to greater wave-
lengths, then another peak appears which also shifts to greater wavelengths, but it
shifts more slowly than the previous one and so and so forth. This difference in the
speed of the wavelength displacement makes that, for greater thicknesses there were
Fig. 17 a Evolution of the simulated spectrum as the thickness of the coating increases and
b simulated optical spectra for different thickness of the coating
72 J. Ascorbe et al.
Fig. 18 a Evolution of the reflected optical spectrum for b four cycles of 20–90 %RH, c some
optical spectra for different values of RH and d wavelength position of the interference peaks as a
function of the RH
High Sensitivity Optical Structures for Relative Humidity Sensing 73
Wavelength shifts of 100 nm have been achieved for changes on the relative
humidity from 20 to 90 %. The wavelength shifts have a dependence on the
wavelength, slightly greater displacements are obtained for interference peaks
located at greater wavelengths. This kind of devices have low hysteresis, high
linearity, good response time and high sensitivity. However, it must be carefully
selected the material which is going to be used for the development of this device. It
must meet the requirements to be sensitive to RH and to generate a good inter-
ferometer. Most semiconductors are sensitive to changes on relative humidity and
they will generate a better interferometer as its refractive index is greater.
Long period fiber gratings (LPGs) has some desirable characteristics for their use as
optical fiber sensors. LPGs structure consists on a periodic modification of the
refractive index of the optical fiber. Several approaches have been followed to
develop LPGs. Modification of the refractive index of the optical fiber can be
obtained by exposing to UV radiation hydrogen-loaded germanosilicate fibers [60,
61], or by methods based on the physical deformation of the fiber [62], diffusion of
the core dopants in special nitrogen-doped Ge free fibers [63, 64], or refractive
index variation produced by CO2 lasers [65], CO laser [63, 64], ion implantation
[66], or electrical discharges [62, 63, 67, 68] (Fig. 19).
Fig. 19 Scheme of a LPG, indicating their parameters, and its characteristic spectrum when the
light goes through the grating. Reprinted from Sens. Actuators B, vol. 227, Urrutia A. et al.,
Simultaneous measurement of humidity and temperature based on a partially coated optical fiber
long period grating, pp. 135–141, (2016), with permission from Elsevier
74 J. Ascorbe et al.
LPGs have been also studied as humidity sensors [69–74]. Different materials
have been checked as the sensitive layer for humidity sensing purposes. In [69],
poly(ethylene oxide)/cobalt chloride (PEO/CoCl2) as a hybrid hygrosensitive
cladding coating. The material parameters associated with the sensing mechanism
may include those of refractive index, absorption, and morphological alterations of
the overlaid material. Relative humidity variations in the range from 50 to 95 %
have been detected with a resolution better than 0.2 %. The response time constant
of the fiber sensor is of the order of a few hundred milliseconds. Maximum
wavelength shift of 6 nm have been obtained, but the response of this hybrid
coating is nonlinear. In [70] a hydrogel has been used as the sensitive material
obtaining a linear wavelength shift of 12 nm for humidity changing from 30 to
100 %. Other approach has been followed in [73] where the LPG has been coated
by means of electrostatic self-assembly method with PAH, PAA and silica nano-
spheres. Wavelength shifts of 14 nm have been obtained for RH varying from 10 to
80 %. Optical spectra for different values of RH are shown in Fig. 20. It has been
also checked the influence of temperature on the sensitivity, which has not
noticeable effect.
Finally, in [74] a simultaneous measurement of temperature and RH has been
accomplished by a half-coated LPG. When half of the LPG coating was chemically
removed the main attenuation band was split into two different contributions, one of
them due to the grating and the other due to the coating. When this semi-coated
LPG was also exposed to RH and temperature tests, the new two attenuation bands
presented different behaviors for humidity and temperature. This novel
dual-wavelength based sensing method enables the simultaneous measurement of
RH and temperature using only one LPG (Fig. 21).
Fig. 20 LPG spectra: a theoretical simulation for different refractive index values of the overlay;
b experimental results for different relative humidity levels. Reprinted from A fibre optic humidity
sensor based on a long-period fibre grating coated with a thin film of SiO2 nanospheres, Viegas, D.
et al., 2012, Meas. Sci. Technol. 20 034002, doi:10.1088/0957-0233/20/3/034002 with permission
from IOP Publishing
High Sensitivity Optical Structures for Relative Humidity Sensing 75
5 Conclusions
There are several ways to measure relative humidity by optical set-ups and by
optical fiber humidity sensors. Here, we have delved into lossy mode resonances
and Fabry-Pèrot interferometers, and LPGs have been briefly commented. These
structures have been chosen for this review because each of them presents an
optimal feature to function as an optical fiber sensor. Although there are more
structures that have not been analysed, some interesting conclusions can be
extracted from those that have been explained here.
First of all, polymeric coatings and inorganic salts usually present nonlinear
behaviour, especially at high relative humidity values where they have their greatest
sensitivity. Metal oxides and semiconductors oxides are a good choice for obtaining
linear responses and good sensitivities in the 20–90 % RH range. Their response
times are usually smaller than those of hygroscopic materials.
With regard to optical structures, there should be an agreement between the
wanted spectral width of the attenuation band and its displacement. LPFGs need a
fabrication process to obtain the grating which requires a specific machine. They are
76 J. Ascorbe et al.
References
1. P.R. Wiederhold, Water vapor measurement: methods and instrumentation, vol 1 (CRC Press,
1997)
2. A. Urrutia, P.J. Rivero, J. Goicoechea, F.J. Arregui, I.R. Matías, Humidity sensor based on a
long-period fiber grating coated with a hydrophobic thin film, in (EWOFS’10) Fourth
European Workshop on Optical Fibre Sensors (International Society for Optics and
Photonics, 2010), pp. 765320–765320
3. C. Bariain, I.R. Matı́as, F.J. Arregui, M. Lopez-Amo, Optical fiber humidity sensor based on a
tapered fiber coated with agarose gel. Sens. Actuators B: Chem. 69(1), 127–131 (2000)
4. Glossary of atmospheric chemistry terms (Recommendations 1990), Pure Appl. Chem. 62,
2167 (1990). doi:10.1351/goldbook.A00155. ISBN 0-9678550-9-8
5. M.M. Choi, O.L. Tse, Humidity-sensitive optode membrane based on a fluorescent dye
immobilized in gelatin film. Anal. Chim. Acta 378(1), 127–134 (1999)
6. T.L. Yeo, T. Sun, K.T.V. Grattan, Fibre-optic sensor technologies for humidity and moisture
measurement. Sens. Actuators, A 144(2), 280–295 (2008)
7. J.F. Boyle, K.A. Jones, The effects of CO, water vapor and surface temperature on the
conductivity of a SnO2 gas sensor. J. Electron. Mater. 6(6), 717–733 (1977)
8. C. Pijolat, R. Lalauze, Influence of adsorbed hydroxyl species on the electrical conductance of
SnO 2. Sens. Actuators 14(1), 27–33 (1988)
9. E.W. Thornton, P.G. Harrison, Tin oxide surfaces. Part 1.—surface hydroxyl groups and the
chemisorption of carbon dioxide and carbon monoxide on tin (IV) oxide. J. Chem. Soc.
Faraday Trans. 1: Phys. Chem. Condens. Phases 71, 461–472 (1975)
10. G. Korotchenkov, V. Brynzari, S. Dmitriev, Electrical behavior of SnO2 thin films in humid
atmosphere. Sens. Actuators B: Chem. 54(3), 197–201 (1999)
11. Y. Kim, B. Jung, H. Lee, H. Kim, K. Lee, H. Park, Capacitive humidity sensor design based
on anodic aluminum oxide. Sens. Actuators B: Chem. 141(2), 441–446 (2009)
12. Y. Sakai, Y. Sadaoka, M. Matsuguchi, Humidity sensors based on polymer thin films. Sens.
Actuators B: Chem. 35(1), 85–90 (1996)
13. K.T.V. Grattan, T. Sun, Fiber optic sensor technology: an overview. Sens. Actuators, A 82(1),
40–61 (2000)
14. A.P. Russell, K.S. Fletcher, Optical sensor for determination of moisture. Anal. Chim. Acta
170, 209–216 (1985)
15. S.K. Khijwania, K.L. Srinivasan, J.P. Singh, Performance optimized optical fiber sensor for
humidity measurement. Opt. Eng. 44(3), 034401–034401 (2005)
16. B.D. MacCraith, Enhanced evanescent sensors based on sol-gel derived porous glass coating.
Sens. Actuators, B 11, 29–33 (1993)
High Sensitivity Optical Structures for Relative Humidity Sensing 77
17. F.J. Arregui, Z. Ciaurriz, M. Oneca, I.R. Matias, An experimental study about hydrogels for
the fabrication of optical fiber humidity sensors. Sens. Actuators B (2003)
18. C. Barian, I.R. Matis, F.J. Arregui, M. Lopez-Amo, Optical fiber humidity sensor based on a
tapered coated with agarose gel. Sens. Actuators, B 69, 127–131 (2000)
19. J.M. Corres, F.J. Arregui, I.R. Matías, Sensitivity optimization of tapered optical fiber
humidity sensors by means of tuning the thickness of nanostructured sensitive coatings. Sens.
Actuators B: Chem. 122(2), 442–449 (2007)
20. M. Hernaez, C.R. Zamarreño, C. Fernandez-Valdivielso, I. Del Villar, F.J. Arregui, I.R.
Matias, Agarose optical fibre humidity sensor based on electromagnetic resonance in the
infra-red region. Phys. Status Solidi (C) 7(12), 2767–2769 (2010)
21. M. Marciniak, J. Grzegorzewski, M. Szustakowski, Analysis of lossy mode cut-off conditions
in planar waveguides with semiconductor guiding layer. IEE Proc. J (Optoelectronics) 140(4),
247–252 (1993)
22. F. Yang, J.R. Sambles, Determination of the optical permittivity and thickness of absorbing
films using long range modes. J. Mod. Opt. 44(6), 1155–1163 (1997)
23. I. Del Villar, C.R. Zamarreño, P. Sanchez, M. Hernaez, C.F. Valdivielso, F.J. Arregui, I.R.
Matias, Generation of lossy mode resonances by deposition of high-refractive-index coatings
on uncladded multimode optical fibers. J. Opt. 12(9), 095503 (2010)
24. C.R. Zamarreño, I. Del Villar, P. Sanchez, M. Hernaez, C. Fernandez, I.R. Matias, F.
J. Arregui, Lossy-mode resonance-based refractometers by means of indium oxide coatings
fabricated onto optical fibers, in (EWOFS’10) Fourth European Workshop on Optical Fibre
Sensors ( International Society for Optics and Photonics, 2010), pp. 76531 W–76531 W
25. P. Sanchez, C.R. Zamarreño, M. Hernaez, I. del Villar, I.R. Matias, F.J. Arregui, Humidity
sensor fabricated by deposition of SnO2 layers onto optical fibers, in Fifth European
Workshop on Optical Fibre Sensors (International Society for Optics and Photonics, 2013),
pp. 87940C–87940C
26. A.B. Socorro, I. Del Villar, J.M. Corres, F.J. Arregui, Lossy mode resonance-based pH sensor
using a tapered single mode optical fiber coated with a polymeric nanostructure, in Sensors,
IEEE 2011 (IEEE, 2011), pp. 238–241
27. C. Ruiz Zamarreño, P. Zubiate, M. Sagües, I.R. Matias, F.J. Arregui, Experimental
demonstration of lossy mode resonance generation for transverse-magnetic and
transverse-electric polarizations. Opt. Lett. 38(14), 2481–2483 (2013)
28. M. Hernáez, I. Del Villar, C.R. Zamarreño, F.J. Arregui, I.R. Matias, Optical fiber
refractometers based on lossy mode resonances supported by TiO 2 coatings. Appl. Opt. 49
(20), 3980–3985 (2010)
29. H.P. Ho, W.W. Lam, Application of differential phase measurement technique to surface
plasmon resonance sensors. Sens. Actuators B: Chem. 96(3), 554–559 (2003)
30. S. Zeng, D. Baillargeat, H.P. Ho, K.T. Yong, Nanomaterials enhanced surface plasmon
resonance for biological and chemical sensing applications. Chem. Soc. Rev. 43(10), 3426–
3452 (2014)
31. J.B. González-Díaz, A. García-Martín, J.M. García-Martín, A. Cebollada, G. Armelles, B.
Sepúlveda, Y. Alaverdyan, M. Käll, Plasmonic Au/Co/Au nanosandwiches with enhanced
magneto-optical activity. Small 4(2), 202–205 (2008)
32. P. Sanchez, C.R. Zamarreño, M. Hernaez, I. Del Villar, C. Fernandez-Valdivielso, I.R.
Matias, F.J. Arregui, Lossy mode resonances toward the fabrication of optical fiber humidity
sensors. Meas. Sci. Technol. 23(1), 014002 (2012)
33. P. Zubiate, C.R. Zamarreño, I. Del Villar, I.R. Matias, F.J. Arregui, High sensitive
refractometers based on lossy mode resonances (LMRs) supported by ITO coated D-shaped
optical fibers. Opt. Express 23(6), 8045–8050 (2015)
34. J.M. Corres, F.J. Arregui, I.R. Matias, Design of humidity sensors based on tapered optical
fibers. J. Lightwave Technol. 24(11), 4329–4336 (2006)
35. C. Bariain, I.R. Matı́as, F.J. Arregui, M. Lopez-Amo, Optical fiber humidity sensor based on a
tapered fiber coated with agarose gel. Sens. Actuators B: Chem. 69(1), 127–131 (2000)
78 J. Ascorbe et al.
36. J. Ascorbe, J.M. Corres, F.J. Arregui, I.R. Matias, Optical fiber humidity sensor based on a
tapered fiber asymmetrically coated with indium tin oxide, in Sensors, 2014 IEEE (IEEE,
2014), pp. 1916–1919
37. J. Ascorbe, J.M. Corres, I.R. Mat, F.J. Arregui, Humidity sensor based on lossy mode
resonances on an etched single mode fiber, in 2015 9th International Conference on Sensing
Technology (ICST) (IEEE, 2015), pp. 365–368
38. H.S. Haddock, P.M. Shankar, R. Mutharasan, Fabrication of biconical tapered optical fibers
using hydrofluoric acid. Mater. Sci. Eng., B 97(1), 87–93 (2003)
39. H.J. Kbashi, Fabrication of submicron-diameter and taper fibers using chemical etching.
J. Mater. Sci. Technol. 28(4), 308–312 (2012)
40. S$AB
41. R.A. Becker, Traveling-wave electro-optic modulator with maximum bandwidth-length
product. Appl. Phys. Lett. 45(11), 1168–1170 (1984)
42. N. Liu, M. Hu, H. Sun, T. Gang, Z. Yang, Q. Rong, X. Qiao, A fiber-optic refractometer for
humidity measurements using an in-fiber Mach-Zehnder interferometer. Opt. Commun. 367,
1–5 (2016)
43. Q. Wang, W. Wei, M. Guo, Y. Zhao, Optimization of cascaded fiber tapered Mach-Zehnder
interferometer and refractive index sensing technology. Sens. Actuators B: Chem. 222, 159–
165 (2016)
44. G. Hernández, Fabry-Pèrot Interferometers, vol 3 (Cambridge University Press, 1988)
45. X. Chen, F. Shen, Z. Wang, Z. Huang, A. Wang, Micro-air-gap based intrinsic Fabry-Pèrot
interferometric fiber-optic sensor. Appl. Opt. 45(30), 7760–7766 (2006)
46. J. Wang, B. Dong, E. Lally, J. Gong, M. Han, A. Wang, Multiplexed high temperature
sensing with sapphire fiber air gap-based extrinsic Fabry-Pèrot interferometers. Opt. Lett. 35
(5), 619–621 (2010)
47. L.H. Chen, T. Li, C.C. Chan, R. Menon, P. Balamurali, M. Shaillender, B. Neu, X.M. Ang,
P. Zu, W.C. Wong, K.C. Leong, Chitosan based fiber-optic Fabry-Pèrot humidity sensor.
Sens. Actuators B: Chem. 169, 167–172 (2012)
48. J. Goicoechea, C.R. Zamarreño, I.R. Matias, F.J. Arregui, Utilization of white light
interferometry in pH sensing applications by mean of the fabrication of nanostructured
cavities. Sens. Actuators B: Chem. 138(2), 613–618 (2009)
49. J. Goicoechea, C.R. Zamarreño, I.R. Matías, F.J. Arregui, Study on white light optical fiber
interferometry for pH sensor applications, in Sensors, 2007 IEEE (IEEE, 2007), pp. 399–402
50. F.J. Arregui, Y. Liu, I.R. Matias, R.O. Claus, Optical fiber humidity sensor using a nano
Fabry-Pèrot cavity formed by the ionic self-assembly method. Sens. Actuators B: Chem. 59
(1), 54–59 (1999)
51. W. Xie, M. Yang, Y. Cheng, D. Li, Y. Zhang, Z. Zhuang, Optical fiber relative-humidity
sensor with evaporated dielectric coatings on fiber end-face. Opt. Fiber Technol. 20(4), 314–
319 (2014)
52. J. Ascorbe, C. Sanz, J.M. Corres, F.J. Arregui, I.R. Mat, S.C. Mukhopadhyay, High sensitivity
extrinsic Fabry-Pèrot interferometer for humidity sensing, in 2015 9th International
Conference on Sensing Technology (ICST) (IEEE, 2015), pp. 143–146
53. B.C. Yadav, N.K. Pandey, A.K. Srivastava, P. Sharma, Optical humidity sensors based on
titania films fabricated by sol–gel and thermal evaporation methods. Meas. Sci. Technol. 18
(1), 260 (2006)
54. G.W. Bundrett, Criteria of Moisture Control (London, Butterworths, 1990), p 19
55. T. Morimoto, M. Nagao, F. Tokuda, The relation between the amounts of chemisorbed and
physisorbed water on metal oxides. J. Phys. Chem. 73, 243–248 (1969)
56. T. King, Water, Miracle of Nature (Macmillan, New York, 1953)
57. W.M. Sears, The effect of oxygen stoichiometry on the humidity sensing characteristics of
bismuth iron molybdate. Sens. Actuators B: Chem. 67(1), 161–172 (2000)
58. P.M. Faia, C.S. Furtado, Effect of composition on electrical response to humidity of TiO 2:
ZnO sensors investigated by impedance spectroscopy. Sens. Actuators B: Chem. 181, 720–
729 (2013)
High Sensitivity Optical Structures for Relative Humidity Sensing 79
P. Shuk
1 Introduction
Oxygen is the third most abundant chemical element in the universe, after hydrogen
and helium, and is one of the most important elements required to sustain the life
[1]. Oxygen is vitally important not only to the existence of the human and animal
life with ∼3 million tons daily of oxygen consumption just for the respiration [2]
but also very critical for any combustion process at power generation, chemicals
production, heating, process control, safety, and quality. A major oxygen gas
sensing application is for the combustion process optimization. The air required for
the oxidation of hydrocarbons (fossil or bio fuels) is optimized to ensure the fuel
complete combustion with the highest possible efficiency [3, 4]. Oxygen excess in
the combustion in normally measured in the flue gas not far away from the fire ball
or box and the fuel/air ratio supplied to the burner is controlled to the ideal mini-
mum of oxygen or air excess (Fig. 1). The oxygen is also measured to minimize
runaway combustion sometimes leading to the explosion and to control the product
formation and quality. Some oxygen sensors work at ambient conditions where
oxygen is either dissolved in liquid or solid or oxygen measurements have to be
P. Shuk (✉)
Rosemount Liquid and Combustion Analysis, Emerson Process Management, Emerson
Electric Corp. 2400 Barranca Pkwy, Irvine, CA 92606, USA
e-mail: [email protected]
made under ambient conditions. Mostly these sensors are applied in medical, food
processing and waste management industries. The oxygen sensors used under
ambient conditions include Clark type cells [5], paramagnetic [6] and optical
sensors [7, 8].
The high temperature oxygen sensors can be based on the potentiometric or
amperometric electrochemical cells and resistive semiconductors [9, 10]. Poten-
tiometric sensors measure the equilibrium voltage of the cells according to the
Nernst equation and would require the reference. The amperometric oxygen sensor
with fixed applied voltage and oxygen diffusion barrier is based on electrical current
dependence on oxygen concentration and can be only successfully applied to very
clean environment and this technology was discussed in many details elsewhere
[11, 12].
The resistive oxygen sensor is based on semiconductors with the defect
concentrations varying exponentially with the oxygen concentration [13]. The
sensitivity of the surface to the particular gas can be as low as parts per billion
(ppb) [14–17] and it is highly desirable that metal oxide semiconductor sensors
have a large surface area, so as to adsorb as much of the target analyte gas on the
surface as possible, giving a stronger and more reliable response especially at low
concentrations. The applications of the semiconductors oxygen sensors in severe
combustion process would be highly limited because of sensitivity, selectivity and
reproducibility issues [10].
Four major oxygen gas sensing technologies are applied in an in situ or
extractive mode to the combustion process environment and will be discussed in
more details.
30 80
N 70
25 2
Combustion
Gas concentration (%)
HO
Gas concentration (%)
2 control
20 60
15 50
CO
2
10 40
CO O 30
5 2
0 20
-5 0 5 10
% Deficiency % Excess
Oxygen concentration (%)
(a) (b)
2- ..
O ,VO
2- ..
O , VO
Solid electrolyte Solid electrolyte
The process side of the cell is gas tight separated from the reference side with the
special high temperature ceramic/metal/glass seals and zirconia ceramic is used in
the form of disc (Fig. 3), tube or thimble.
Differential oxygen chemical potentials on the oxygen cell electrodes would
develop electromotive force (EMF), E, according to the Nernstian equation:
RT pðO2 Þ′′Process
E= ln , ð3Þ
4F pðO2 Þ′Ref
with R universal gas constant, T the process temperature in Kelvin (K) and F the
Faraday number.
By using fixed oxygen partial pressure on the reference electrode, e.g., air with
20.9 % O2 and p(O2) = 2.1 * 104 Pa, the sensor signal of the thermally balanced
oxygen sensor will be only depending on the sensor temperature (Fig. 4). Oxygen
potentiometric gas analysis was established by Mӧbius [21] and is also permitting
measurements of not free but chemically bonded oxygen concentration <1 ppm
(part per million) or <1 ppb (parts per billion) O2 released in the reducing condi-
tions in the process and CO/CO2-ratio (Fig. 5) or H2/H2O-ratio (Fig. 6) can be also
calculated based on the oxygen sensor signal.
Pt electrode film microstructure is highly depending on the sintering temperature
and formation at lower temperatures would create much finer open porosity (Fig. 7)
contributing to the better oxygen molecules diffusion to TPB interface but the
electrode/electrolyte interface should be well established and lower temperature
sintered electrodes might not be reliable in the severe process environment with the
2
80
60 4
E (mV)
6
40
8
10
20
15
0 20
600 800 1000 1200 1400
o
T ( C)
10-20 c(O2)
1500
(%)
c(CO)/c(CO 2)
10-15
103
C
102
1000
10
10-10
E (mV)
10-1
500 10-5
10-2
10 -5
10 -4 10-4
10-3 10-3
CO2 10-2
0.1
0.4
1
8
0
700 800 900 1000 1100 1200 1300
T (oC)
the much less differential temperature profile across the cell on the reference and
process sides compared to the tube or thimble ceramic shape and less thermal stress
considering a good thermal expansion coefficients (TEC) optimization of all
involved in the oxygen sensor packaging metallic and ceramic materials. The small
oxygen sensor is gas tight sealed at O2 analyzer probe end for in situ operation
(Fig. 9) or process flue gas can be alternatively delivered to the sensor using special
extractive gas guiding tube [22]. The special protective shield, internal calibration
line and reference gas line made from the special alloys like Inconel, Hastelloy or
stainless steel 316L were implemented in the probe design to resist severe high
temperature process environment. With these entire improvements zirconia probe
can very precise measure in situ oxygen concentration down to 10 ppm (Fig. 10).
A special oxygen cell with improved thermal balance and ceramic or metallic dif-
fuser for the process gas particulate filtration was designed to improve O2 analyzer
[20] delivering an excellent stability (<0.1 % O2) and reproducibility (±0.01 % O2)
in 2…5 % oxygen combustion concentration range (see Figs. 11 and 12).
Typical installation points of Zirconia O2-probe in a coal fired boiler as it can be
seeing from Fig. 13, would be located before or after heater exchanger
Oxygen Gas Sensing Technologies Application: A Comprehensive … 87
10-20 c(O2)
1400
%
c(H2)/c(H 2O)
103 10-15
1200 10 2
10
1000 1
10-1 10-10
800 10-2
E (mV)
10-3
600 10-4
10-5 10-5
o
1400 C 1200 oC o
1000 C
(economizer). For the combustion control and higher efficiency single or multiple
probe systems might be used depending on the duct size and economizer outlet.
Additional probes may be mounted downstream monitoring the air entrance to
ensure efficient operation of fans and dampers. In the combustion process due to
88 P. Shuk
O2-cell
centrifugal forces, fly ash is concentrated towards the duct outer wall following
bends in the duct and the installation of Zirconia O2 analyzers in these locations
should be avoid to reduce the probe body erosion in the high velocity and high
concentration abrasive fly ash environment. There is a good Zirconia O2-probe
installation practice (Fig. 14) reducing very fine fly ash accumulation inside the
oxygen sensor on the process electrode and on the protection filter on the probe end.
In the oil and gas fired applications these problems are not present.
Zirconia O2 analyzer was applied with very good success in the combustion
control of power generation boilers, hot stoves for steelmaking, heating and com-
bustion exhaust gas control of coke ovens for steelmaking, lime and cement kilns
Oxygen Gas Sensing Technologies Application: A Comprehensive … 89
10 ppm O2
2
±0.75% error or <±0.05% O 2 accuracy
150
E (mV)
100
50
0
0.0001 0.001 0.01 0.1 1 10
c (O ) (%)
2
10
6888 O 2 probe drift at 0.41%, 2.0 and 8.0% O 2 (3/5...9/8/14)
8.0±0.1 % O2
8
6
c (O ) (%)
2
2.00±0.06 % O2
2
0.41±0.01% O2
0
Mar/19/14 Apr/12/14 May/6/14 May/30/14 Jun/24/14 Jul/18/14 Aug/11/14 Sep/4/14
Date
70
Zirconia O 2 Analyzer reproducibility at 2...5% O 2 :
<±0.01% O 2
60
2% O2 2% O 2
E (mV)
50
3% O2 3% O2
# 1
4% O 2 4% O 2 # 2
40 # 3
# 4
5% O 2 # 5
30
0 5 10 15 20
t (min)
Secondary
600-700 C
o
combustion zone
Zirconia
O2-probe Primary
Economizer combustion zone
installation
options
o
~400 C
Flue gas
dominating the combustion market because of the high reliability, accuracy in the
most of applications and relative low price. Ametek, Rosemount Liquid and
Combustion Analysis, Yokogawa and ABB are the major players on the in situ
Oxygen Gas Sensing Technologies Application: A Comprehensive … 91
Zirconia
O2-probe
Flue gas
s
ga
ue
Fl
s
ga
ue
Fl
Fig. 16 Diode laser types and combustion gases versus wavelength [30]
combustion environment due to the robust, compact, low-cost TDL sources, and the
convenient overlap of TDL wavelengths with absorption bands of O2. CO, NOx,
H2O, CO2 combustion gases. TDLS with the ability to interface directly with the
process gas and eliminating the need for the very high cost and high maintenance
sample handling systems would be ideal for the real time dynamic measurement of
the process conditions. With the federal, state and private companies multi billions
investments TDL analyzers have now comprehensively established themselves as a
core measurement technology within the portfolio of the gas analysis techniques
available on the market [29, 31–33].
The infrared absorption spectrum of the flue gases is like a fingerprint, providing
oxygen identification at ∼760 nm (Fig. 17) and measurements in a wide concen-
tration range down to ∼80…100 ppm as a detection limit.
With the longer optical path length a higher absorption and better sensitivity will
be achieved. Laser wavelength can be changed over a wide range by changing the
laser operation temperature and over a narrow range by changing the applied
current to the laser. Tunable lasers enabling the miniaturization of transmission and
receiving units, provide highly sensitive, quantitative measurements with the fast
response time without the need for the frequent calibration. TDL analyzers have
been shown to be capable of the effective, accurate measurement even loosing up to
CO
2330 nm
CO 2
1960 nm
NO
1800 nm
CO
1570nm
H 2O
1390nm
O2
760nm
Fig. 19 A typical installation of TDL O2-analyzer across the combustion duct [40]
96 P. Shuk
Fig. 20 Alignment
improvement with flexible
bellows without affecting the
process seal [41]
across the duct, would be unlikely much more representative than zirconia O2-
analyzer single point measurement. With the increasing temperature oxygen
absorption intensity would be reduced [43] and this might bring some challenge in
the interpretation of weaker intensity at higher temperature and temperature vari-
ation compensation affecting oxygen concentration calculation and contributing to
signal to noise ratio reduction. Temperature and pressure variation in the process
even claimed to be compensated might bring an additional error in the oxygen
measurements. Calibration and validation of TDL O2 analyzer on line is still
waiting for a conclusive answer. Implemented in some TDL instrumentation a
reference cell has a different temperature compared to the process and is very short
to provide very reliable in situ validation. Alignment across the stack and IR light
reflection at high temperature combined with the process windows fouling and wide
background radiation from the fire ball or box would bring many challenges in the
application and might reduce service life of the TDL O2-analyzer. Fiber coupled
Oxygen Gas Sensing Technologies Application: A Comprehensive … 97
laser solution increases the optical noise by a factor of 5–7 compared to the free air
lasers [43] and would reduce the sensitivity with detection limit ∼500 ppm. The
major players on TDL analyzers market are Neo Monitors (Norway), Yokogawa
(Japan), Focused Photonics Inc. (China) and Siemens (Germany) (Fig. 22).
One significant difference between traditional Zirconia O2-sensor single point
measurements and TDL O2 analyzer path average measurements is that the path
measurement does not provide spatial information [33]. Zirconia O2 probes are
providing a measurement at a particular point in the process with very useful
information identifying variations in O2 concentration levels caused by burner to
burner (unbalanced burners) and this information can be successfully used for the
burner diagnostics.
More details on TDLS theoretical background and application can be found in
the latest reviews [29, 30, 44–47]. Performance and application option of some of
TDL O2 analyzers on the market is summarized in Table 2.
With wide applicability, TDL O2 analyzers are enjoying great popularity but
being ∼4−5 times more expensive compared to the zirconia O2 analyzers and
additional, very expensive and sophisticated up to 4 months installation [32, 33] are
generally covering right now special chemicals production and very high temper-
ature (>1400 °C) applications where zirconia O2-analyzers would be rejected by
severe temperature environment or the probe might contaminate the chemicals in
the production. An extractive zirconia O2 analyzer might be considered as an
alternative less expensive option for these applications. Majority of the combustion
systems is still using zirconia oxygen O2 analyzers.
98 P. Shuk
Oxygen with two unpaired electrons is well known paramagnetic gas and would be
attracted into a strong magnetic field. These paramagnetic properties enable rela-
tively selective oxygen measurements method used by industries over fifty years
and applied in three different methodologies: magnetodynamic, thermomagnetic or
magnetic wind and non-contact Quinke method [48]. The magnetodynamic oxygen
Oxygen Gas Sensing Technologies Application: A Comprehensive … 99
sensors are dumbbell or of the magnetic pressure type. In the dumbbell type O2-
sensor incorporates two nitrogen-filled glass spheres mounted on the rotating sus-
pension (Fig. 23). This assembly is in a strong magnetic field and with the gas
sample containing oxygen introduced into the magnetic field, the oxygen molecules
will tend to approach the strongest part of the magnetic field pushing the dumbbell
aside and causing the suspending wire to be twisted. This twist is detected as the
movement of the light from the reflective mirror located at the suspending wire
centre (Figs. 23 and 24). The output from the photo-sensor would be fed back to a
coil around the suspension assembly restoring torque to the original “zero” position.
The current would be directly proportional to the magnetic susceptibility of the
sample oxygen gas and oxygen partial pressure. The applied current will be pro-
portional to the oxygen concentration in the gas mixture and would be used for the
precise oxygen calculation. This method with the high linearity and good response
is however highly sensitive to the different mechanical shocks, like vibration. In the
magnetic pressure type O2-sensor a small amount of reference gas like pure
nitrogen or air is externally supplied through a small hole made on one of the
magnetic poles. When the sample gas is introduced into uneven magnetic field,
paramagnetic oxygen gas would be attracted toward the strongest part of the
magnetic field, in which the magnetic poles have come closer to each other. The
increase in pressure resulting from the attraction of oxygen can be detected with a
capacitor microphone type detector or a mass flow sensor. In order to ensure the
stable signals detection, the two magnetic poles are alternately excited to amplify
the alternating current. The selectivity of this method is pretty high but a reference
gas would be required for these measurements. Thermomagnetic O2-sensor (mag-
netic wind method) has two chambers with heating wire element located at the
centre. A magnetic field would be provided only on the measurement chamber.
After the analysing oxygen gas is entering the measurement chamber, oxygen
would be attracted by the magnetic field and then heated using the heating wire
element. The magnetic oxygen flow would be cooling the heating wire element as
its intensity varies proportional to the oxygen concentration. The resistance changes
during this process would be picked up as unbalanced voltage at the Wheatstone
bridge to measure the oxygen concentration. The magnetic wind type analyzer has
no movable parts and is more stable in the vibration environment. Carbon dioxide
with the some positive interference for oxygen by the small variation in the con-
centration can be compensated.
Quincke O2-detector utilizes a pneumatic Wheatstone bridge for the measure-
ments using a differential pressure or flow to determine the oxygen concentration.
Reference gas is used so the sample does not come in contact with the sensor. The
reference gas is divided into two flow paths recombining at the reference gas outlet,
where the sample gas is also introduced. A magnetic field is in one arm of the
reference gas outlet creating a back pressure related to the presence of the oxygen in
the gas sample.
Measurement range for the paramagnetic O2-sensor is typically between 0.05 %
and 100 % O2 and sensor can be very precise under certain conditions [48] but
would require quite expensive gas sample conditioning system to stabilize the
process gas flow, pressure and temperature and to remove hazardous gases and
dust. The major disadvantages of the paramagnetic oxygen sensors are the price,
low stability in the vibration environment and relative low accuracy (± 0.10…
0.25 % O2). Paramagnetic analyzers have been historically limited to the extractive
application with general use (Fig. 25) or intrinsically safe instrumentation version
(Fig. 26) and maximum external operation temperature up to 120−145 °C.
Oxygen Gas Sensing Technologies Application: A Comprehensive … 101
nature of the oxygen limiting the detection to >100 ppm O2 and moving parts in
the design highly limiting applications especially in high temperature and humid
vibration environment. Paramagnetic analyzers are lately often expelled by the new
advanced Tunable Diode Laser (TDL) analyzers (see Sect. 3 of this review) pro-
viding more expensive but likely more reliable and more precise measurements
according to TDLS publications.
Paramagnetic oxygen analyzers are applied to the measurement of trace oxygen
concentration in various manufacturing processes, city gas quality control, safety
control at various plants, oxygen concentration control of flammable gas containing
mixtures and coke dry quenching (CDQ) plants for the steelmaking.
The electrochemical O2-sensors with liquid electrolyte date back to the 50th and
were used for the environmental oxygen monitoring providing warning of abnormal
conditions in industrial safety environments, i.e., mines, oil production facilities and
chemical plants. Oxygen measuring instruments containing electrochemical sensors
based on consumable lead anodes have been available for over 30 years [56]. These
type electrochemical oxygen sensors are metal-air battery type with the diffusion
limited barrier [56, 57] and oxygen sensor generates a current proportional to the
oxygen diffusion to the cathode limited by the diffusion film or capillary barrier
(Fig. 28). Such barriers are typically made of thin, low-porosity Teflon membranes.
Besides offering a mechanical protection to the sensor, the membrane performs the
additional function as a filter. Selecting the correct pore size of the membrane and
capillary is necessary to transfer the proper amount of gas molecules. The pore size
should be optimized to allow enough gas molecules to reach the sensing electrode
but be small enough to prevent liquid electrolyte from leaking out or drying too
quickly.
Oxygen Gas Sensing Technologies Application: A Comprehensive … 103
O2 O2
Anode
Oxygen in contact with the cathode is reduced to hydroxyl ions, with a balancing
reaction of lead oxidation at the anode:
than one or two years. Some components of the flue gas like carbon dioxide (CO2)
and chlorine (Cl2) might increase the rate of the oxidation or would react with the
lead shortening the life of the sensor. A dust protector is normally installed on the
top of the sensor (Fig. 30) to reduce diffuser hole or film pluggage and a scrubber
filter is sometimes installed in the front of the sensor to improve the sensor
selectivity.
There is a limited selection of the filters with the different efficiency degrees. The
most commonly used filter medium is activated charcoal filtering most chemicals
except of carbon monoxide (CO) and hydrogen (H2).
This oxygen electrochemical technology with the liquid electrolyte has been one
of the key technologies enabling the availability of the small, compact instru-
mentation for different applications with the estimated over 50 million sensors sold
world-wide since 1980 [56] and this technology is still continuing to dominate
industrial safety application. Although lead was recognized as offering an attractive
combination of suitable properties some metals like cadmium, bismuth, antimony
and copper were identified as potential lead free candidates with only copper being
available in a high surface area wire form. With a number of alternative approaches
offering longer lifetime for lead-free oxygen sensors a little market penetration was
achieved so far. More details on the electrochemical oxygen gas sensing technology
with liquid electrolyte latest development can be found in the recent review [56].
These electrochemical oxygen sensors are mostly supplied by City Technology Ltd
(UK) and Drägerwerk AG (Germany) and are used primary for the industrial safety
or oxygen monitoring at room temperature in portable analyzers and Continuous
Emissions Monitoring Systems (CEMS) with the gas conditioning employed.
Acknowledgment The author would like to thank Rosemount Liquid and Combustion Analysis
for the research program support.
Oxygen Gas Sensing Technologies Application: A Comprehensive … 105
References
1. J.C.G. Walker, The oxygen cycle, in The natural environment and the biogeochemical cycles,
ed. by O. Hutzinger (Springer, Berlin, Heidelberg, New York, 1980), pp. 87–104
2. N. Lane, Oxygen: the molecule that made the world (Oxford University Press, New York,
2002)
3. N. Docquier, S. Candel, Process control and sensors: a review. Prog. Energy Combust. Sci.
28, 107–150 (2002)
4. P. Shuk, Process Zirconia oxygen analyzer—state of art. Tech. Mess. 77, 19–23 (2010)
5. C. Wu, T. Yasukawa, H. Shiku, T. Matsue, Fabrication of miniature Clark oxygen sensor
integrated with microstructure. Sens. Actuators B110, 342–349 (2005)
6. C. Manning, R.F. Keeling, J.P. Severinghaus, Precise atmospheric oxygen measurements with
a paramagnetic oxygen analyzer. Global Biogeochem. Cycles 13, 1107–1115 (1999)
7. A. Mills, Optical oxygen sensors utilizing the luminescence of platinum metals complexes.
Platinum Met. Rev. 41, 115–127 (1997)
8. S. Chu, Y.L. Lo, T.W. Sung, Review on recent developments of fluorescent oxygen and
carbon dioxide optical fiber sensors. Photonic Sens. 1, 234–250 (2011)
9. W.C. Maskell, B.C.H. Steele, Solid state potentiometric oxygen gas sensors. J. Appl.
Electrochem. 16, 475–489 (1986)
10. R. Ramamoorthy, P.K. Dutta, S.A. Akbar, Oxygen sensors: materials, methods, designs and
applications, J. Mater. Sci. 38: 4271–4282
11. W.C. Maskell, Inorganic solid state chemically sensitive devices: electrochemical oxygen gas
sensors. J. Phys. E: Sci. Instrum. 20, 1156–1168 (1987)
12. W. Göpel, G. Reinhardt, M. Rösch, Trends in the development of solid state amperometric
and potentiometric high temperature sensors. Solid State Ionics 136–137, 519–531 (2000)
13. J.W. Fergus, Doping and defect association in oxides for use in oxygen sensors. J. Mater. Sci.
38, 4259–4270 (2003)
14. N. Barsan, M. Schweizer-Berberich, W. Göpel, Fundamental and practical aspects in the
design of nanoscaled SnO2 gas sensors: a status report. Fresenius J. Anal. Chem. 365, 287–
304 (1999)
15. N. Barsan, D. Koziej, U. Weimar, Metal oxide-based gas sensor research: How to? Sens.
Actuators B121, 18–35 (2007)
16. A. Oprea, N. Bârsan, U. Weimar, Work function changes in gas sensitive materials:
Fundamentals and applications. Sens. Actuators B142, 470–493 (2009)
17. H. Wang, L. Chen, J. Wang, Q. Sun, Y. Zhao, A micro oxygen sensor based on a nano sol-gel
TiO2 thin film. Sensors 14, 16423–16433 (2014)
18. H. Peters, H.H. Möbius, Procedure for the gas analysis at elevated temperatures using
galvanic solid electrolyte elements (Germ), DD-Patent 21673 (1961)
19. J. Weissbart, R. Ruka, Oxygen gauge. Rev. Sci. Instrum. 32, 593–595 (1961)
20. P. Shuk, E. Bailey, U. Guth, Zirconia oxygen sensor for the process application: state of art.
Sens. Transducer 90, 174–184 (2008)
21. H.H. Möbius, Basics of oxygen gas potentiometric analysis (Germ). Z. Phys. Chem. 230,
396–416 (1965)
22. http://www.fujielectric.com/products/instruments/library/catalog/box/doc/ECNO341b.pdf
23. D.M. Haaland, Internal-reference solid-electrolyte oxygen sensor. Anal. Chem. 49, 1813–
1817 (1977)
24. H. Kaneko, T. Okamura, H. Taimatsu, Y. Matsuki, H. Nishida, Performance of a miniature
zirconia oxygen sensor with a Pd-PdO internal reference. Sens. Actuators B108, 331–334
(2005)
25. J.V. Spirig, R. Ramamoorthy, S.A. Akbar, J.L. Routbort, D. Singh, P.K. Dutta, High
temperature zirconia oxygen sensor with sealed metal/metal oxide internal reference. Sens.
Actuators B124, 192–201 (2007)
106 P. Shuk
26. K. Dunst, G. Jasinski, P. Jasinski, Potentiometric oxygen sensor with solid state reference
electrode. Metrol. Meas. Syst. 21, 205–216 (2014)
27. E.D. Hinkley, P.L. Kelley, Detection of air pollutants with Tunable Diode Lasers. Science
171, 635–639 (1971)
28. P. Werle, A review of recent advances in semiconductor laser based gas monitors.
Spectrochim. Acta A54, 197–236 (1998)
29. P.A. Martin, Near-infrared diode laser spectroscopy in chemical process and environmental
air monitoring. Chem. Soc. Rev. 31, 201–210 (2002)
30. R.K. Hanson, Applications of quantitative sensor laser to kinetics, propulsion and practical
energy systems. Proc. Combust. Inst. 33, 1–40 (2011)
31. S.J. Rye, TDL: One Size No Longer Fits All, 59th Analysis Division Symposium, ISA, vol.
503 (Curran Associates Proc, Red Hook, 2014), pp. 1–16
32. J.D. Tate, Development of in situ analysis for the chemical industry. DOE report
(DE-FC36-02ID14428) (2006)
33. J.D. Tate, Advanced Combustion Diagnostics and Control for Furnaces, Fired Heaters and
Boilers. DOE report (DE-FG36-06GO16093) (2010)
34. Y. Krishna, S. O’Byrne, J.J. Kurtz, Baseline correction for stray light in log-ratio diode laser
absorption measurements. Appl. Opt. 53, 4128–4135 (2014)
35. S. Langridge, On line measurement of oxygen: review and new developments, in 55th
Analysis Division Symposium, ISA, vol. 481. Curran Associates Proc, Red Hook (2010),
pp. 113–127
36. H. Li, G.B. Rieker, X. Liu, J.B. Jeffries, R.K. Hanson, Extension of wavelength-modulation
spectroscopy to large modulation depth for diode laser absorption measurements in
high-pressure gases. Appl. Opt. 45, 1052–1060 (2006)
37. G.B. Rieker, J.B. Jeffries, R.K. Hanson, Calibration-free wavelength-modulation spectroscopy
for measurements of gas temperature and concentration in harsh environments. Appl. Opt. 48,
5546–5560 (2009)
38. G.B. Rieker, Wavelength-modulation spectroscopy for measurements of gas temperature and
concentration in harsh environments, Ph.D. thesis, Stanford University (2009)
39. D. Farnesi, A. Barucci, G.C. Righini, S. Berneschi, S. Soria, G. Nunzi Conti, Optical
frequency conversion in silica - whispering-gallery-mode microspherical resonators. Phys.
Rev. Lett. 112, 093901 (2014)
40. http://www.yokogawa.com/an/faq/tdls/spec_1.htm
41. http://www.appliedmc.com/content/images/TDLS200_BU_WEB.pdf
42. http://us.mt.com/dam/MTPRO/Brochures/Leporello_GPro500_en.pdf
43. http://specanalytics.com/index.php?route=information/information&information_id=13
44. R.F. Curl, F.K. Tittel, Tunable infrared laser spectroscopy. Ann. Rep. Prog. Chem C98,
219–272 (2002)
45. M. Lackner, Tunable diode laser absorption spectroscopy (TDLAS) in the process industries—
a review. Rev. Chem. Eng. 23, 65–147 (2011)
46. F.K. Tittel, R. Lewicki, R. Lascola, S. McWhorter, Emerging infrared laser absorption
spectroscopic techniques for gas analysis, in Trace Analysis of Specialty and Electronic
Gases, ed. by W.M. Geiger, M.W. Raynor (John Wiley & Sons Inc, New York, 2013),
pp. 71–109
47. K. Kohse-Hӧinghaus, R.S. Barlow, M. Alden, J. Wolfrum, Combustion at the Focus: Laser
Diagnostics and Control. Proc. Combust. Inst. 30, 89–123 (2005)
48. R.P. Kovacich, N.A. Martin, M.G. Clift, C. Stocks, I. Gaskin, J. Hobby, Highly accurate
measurement of oxygen using a paramagnetic gas sensor. Meas. Sci. Technol. 17, 1579–1585
(2006)
49. http://www.systechillinois.com/en/paramagnetic-cells_54.html
50. http://www2.emersonprocess.com/siteadmincenter/PM%20Rosemount%20Analytical%
20Documents/PGA_Manual_755A_199707.pdf
51. http://www.yokogawa.com/an/download/general/GS11P03A05_01E.pdf
52. http://www.keison.co.uk/products/servomex/Oxy1900.pdf
Oxygen Gas Sensing Technologies Application: A Comprehensive … 107
53. http://www2.emersonprocess.com/siteadmincenter/PM%20Rosemount%20Analytical%
20Documents/PGA_PDS_X-STREAM_XEGP.pdf
54. C. Manning, R.F. Keeling, J.P. Severinghaus, Precise atmospheric oxygen measurements with
paramagnetic oxygen analyzer. Global Biogeochem. Cycles 13, 1107–1115 (1999)
55. S. Vonderschmidt, J. Muller (2010) A novel micro paramagnetic oxygen sensor. In: Micro
Electro Mechanical Systems (MEMS). 23rd IEEE International Conference, pp. 903–906
56. M. Willett, Oxygen sensing for industrial safety—evolution and new approaches. Sensors 14,
6084–6103 (2014)
57. K.S. Goto, Solid state electrochemistry and its applications to sensors and electronic devices
(Elsevier Science Publishers, New York, 2013)
58. http://www.citytech.com/index.html
59. http://www.hsmsearch.com/page_527804.asp
Application of Practical Nitrate Sensor
Based on Electrochemical Impedance
Spectroscopy
1 Introduction
Fresh water is an important part of New Zealand’s way of life and economy. The
water quality of lakes, rivers, streams, and aquifers in New Zealand is variable and
depends mainly on the dominant land use in the catchment. Water quality is good in
the areas of indigenous vegetation and less intensive use of land, however, it is
poorer where there are pressures from urban and agricultural land use [1]. Rivers in
New Zealand has reduced water clarity and aquatic life, and higher concentrations
of nutrients and Escherichia coli (E.coli) bacteria [1].
The growth of population and increasing agricultural land use have placed
increasing pressure on waterways. Between 1990 and 2012, the estimated amount
of nitrogen (N) that leached into the soil from agriculture increased 29 % [1]. The
reason for this was due to the numbers of livestock (6.4 million dairy cattle and 29.5
million sheep [2]) (and, therefore, urine contains high concentrations of nitrogen
(N)) and the use of N fertilizer. Between 1989 and 2013, total N concentrations in
rivers increased between 12 to 60 % in the 77 monitored sites around the country
which was monitored by NIWA, the National Institute of Water and Atmospheric
Research. The physical impact of excessive N concentrations in New Zealand rivers
can be nuisance plantopic algae (includes blue green algae), benthic algae (includes
periphyton) and macrophytes(rooted plants) growth. This growth reduces oxygen
levels in the water, hampers river flows, and smothers the riverbed and plant life,
which fish and other aquatic animals depend on for food and habitat. About 49 % of
monitored river sites currently have enough N to trigger periphyton growth. High
levels of N can also be harmful to fish [1]. A common form of nitrogen is nitrate,
nitrite, and ammonium. In the human body, nitrite can react with secondary or
tertiary amines forming nitrosamines, which are recognized as carcinogenic agents.
When existing in high concentrations in the blood, nitrite can also react with iron(II)
in hemoglobin to form methemoglobin, which has no ability to carry oxygen [3–5].
Nitrate can also have the same toxic effects because it can be reduced to nitrite by
bacterial colonies that reside within the mouth [6]. Anthropogenic actions, such as
agriculture, discharge of industrial and domestic wastewaters and vehicle emis-
sions, have caused the increases in the concentrations of inorganic nitrogenous
compounds in aquatic ecosystems [4, 6, 7]. Ingested nitrate from polluted drinking
waters can cause methemoglobinemia, particularly in young infants, and also have a
potential role in developing cancers of the digestive tract due to the formation of
nitrosamines [4]. The regular monitoring of the levels of nitrate in water is,
therefore, relevant because of their potentially harmful impact on both the envi-
ronment and human health.
2 Motivation
New Zealand’s lakes, rivers, and groundwater are regularly monitored by regional
councils, NIWA, and community interest groups. New Zealand’s river network has
a total stream length of 199,641 km. Regional councils monitor river water quality
to manage environmental impacts. NIWA also monitors water quality at 77 sites on
35 major river systems. Figure 1 indicates the different monitoring sites by NIWA.
These sites tend to be in catchments dominated by agricultural land use. Regional
councils measure a range of parameters at each of the sites they monitor. The
parameters include turbidity, rainfall, different nutrients (nitrate, nitrite, ammonium,
and phosphate), pH, Dissolved Oxygen (DO).
Application of Practical Nitrate Sensor Based … 111
Weirs (Fig. 2a) have been installed at various locations around Tuapaka
(Fig. 2b) to monitor stream flow and water quality. These sites will be used to test
and validate the sensors under field conditions. Stream and river water contain
sediment and organic material. Under the current manual sampling regime, samples
are filtered through a 0.45 µm filter prior to nitrate analysis. It will, therefore, be
important to test the sensors performance in the presence of sediments and organic
material. The sensors functionality and robustness will also need to be tested under
the harsh weather conditions experienced in the remote stream and river locations.
Technicon Auto-Analyzer uses continuous flow system (CFS), which samples are
taken automatically from the sample tray and pump adds the appropriate reagents.
Samples are passed through manifold for mixing, filtering and heating. The col-
orimeter then measure the absorbance of color in the reaction of sample and is
displayed in the recorder [8].
Figure 3 shows the Technicon Auto Analyzer used to determine both of nitrate
and ammonium concentration in the laboratory using the current method. The Auto
analyzer is set up with a heating bath at 28–29 °C and uses 520 nm filters in the
colorimeter. A red azo dye and an indo-phenol prussian blue dye are used to
measure nitrate and ammonium respectively. Reagents used for nitrate-N detection
include catalyst solution, sodium hydroxide solution, hydrazinum sulphate solution
and so on (as shown in Fig. 3b). Reagents are pumped into the system for about
30 min to ensure the system is flushed completely. In the sample tray (Fig. 3a), the
prepared nitrate standard and water samples are put in the sample tray. The probe
will take each sample in the preset time interval into the system. The samples and
reagents are mixed in the system (Fig. 3c) and the measurement is done by a
colorimeter (Fig. 3d). The colorimeter analyzes the intensity of the color and the
results are displayed on a chart recorder in the computer. This recorder baseline
should be set along with the colorimeter baseline control before the measurement
starts by standard calibration. This method requires expensive equipment and
trained staff to conduct measurements in addition to the limitation on water samples
collection, which can only be collected periodically. This approach risks missing
key changes in nitrate-N concentration when river-flow rapidly rises or falls.
Although high-frequency nitrate sample equipment is available, these cost in the
order of $30–50,000.
Typically nutrient and sediment concentrations change with increasing and
decreasing stream or river flow rates (Fig. 4). Therefore, a monthly or fortnightly
sampling regime may not adequately represent the concentration nutrient profile.
This could influence our understanding of the seasonal effects on nutrient loss as
well as the total loads of nutrients (i.e. kg of nitrate/ha) estimated to be leaving a
catchment. This information is critical for reginal councils to implement policy and
management around water quality in their catchments.
Application of Practical Nitrate Sensor Based … 113
Samples
Probe
internationally and in New Zealand. However, the whole set of the system is very
expensive (around $60–70 k).
Hanna Instruments is focused on developing electro-analytical instrumentation.
They provide the products like nitrate portable/benchtop photometer and nitrate ion
selective electrode. For example, Hanna HI96728 is a portable checker to detect
nitrate in freshwater based on the colorimetric method. A special Tungsten lamp is
used as the light source and silicon photocell with narrow band interference filter at
the wavelength of 525 nm used to detect the light. It uses the cadmium reduction
method to determine the nitrate-nitrogen concentration range from 0.0 to
30.0 mg/L. The price is at around $350 NZD.
Xylem is a world leader in providing compact instruments for water technology.
It has a broad range of products under different brands. YSI is one of Xylem brand,
which providing the environmental monitoring products. IQ SensorNet 182 is a
modular water quality terminal which allowed connecting additional sensors such
as 6884 Nitrate ISE sensor and it provides a continuously measurement of water
quality parameters like pH, Dissolved oxygen, temperature, conductivity, ammo-
nium, nitrate, potassium, TOC (total organic carbon), COD (chemical oxygen
demand), DOC (dissolved organic carbon share of TOC), BOD (biochemical
oxygen demand) and SAC (spectral absorption coefficient). It has an easy-to-read
digital display and wireless connection via radio transmission with a range of 100
meters. However, when it uses the nitrate sensor such as NitraLyt, the function of
real-time monitoring is disabled. Another Xylem brand—WTW, it provides online
measurement system—TresCon Analyzer which can monitor Ammonium, Nitrate,
and Nitrite continuously. A continuous water sample supply with low solids con-
tents is required for operating this system. For nitrate detection, the UV light is
absorbed by the nitrate ions and determines the nitrate concentration at a wave-
length of 254 nm.
ABB is a multinational corporation which mainly designs and manufactures
power and automation products, including flow measurement, gas and liquid ana-
lyzer and environmental monitoring systems. For example, UV Nitrate Monitor
AV455 provides a continuous analysis without chemical reagents. It can detect the
nitrate concentration at the range of 0–100 mg/L at a wavelength of 220 nm. It
requires low-maintenance and simple calibration with auto cleaning for the optical
component. It uses de-nitrification as a process to reduce the nitrate concentrations
for nitrate monitoring.
ASA Analytics Inc. specializes in the manufacture of the automatic chemical
analysis system. ChemScan 6101 Process Analyzer can monitor Ammonia,
Phosphate, Nitrate-Nitrite and Phosphate in surface water based on the
UV-spectrometric method. The light is absorbed by nitrate ions and into 256
wavelengths of 200–450 nm. The detection range of nitrate is from 0.5–20 mg/L.
Table 1 summarizes the commercially available nitrate detection product/system
in the market. Hanna HI 96728 is a portable device, but it still needs to collect the
sample and require the reagent for measurement of nitrate. The YSI IQ SensorNet
182 terminal provides the continuous monitoring, but it calls for an additional
sensor for nitrate measurement.
Application of Practical Nitrate Sensor Based … 117
The electrode pattern of the interdigital sensor can be repeated several times, to
get a stronger signal and keep the signal-to-noise ratio in an acceptable range [31].
The configuration of the conventional interdigital sensor is shown in Fig. 6.
When an AC signal is applied as an excitation voltage to the terminals, an
electric field is setup from positive to negative terminal. This electric field bulges
through the test sample via excitation electrode and is received by the sensing
electrode which carries useful information about the properties of MUT such as
impedance, density, chemical composition and so on.
Figure 7 shows the electric field formed between positive and negative elec-
trodes for different pitch lengths- the distance between two consecutive electrodes
of the same polarity. As it is illustrated in the figure, different pitch length (l1, l2 and
l3) shows different penetration depth. The penetration depth improves by increasing
the pitch length, at the cost of electric field strength that will get weaker with greater
distance between the electrodes at same potential (pitch length).
fringing electric field. Different geometries have been studied in research literature
[32–34]. Figure 8 shows the excitation pattern for a multi-sensing electrode in
interdigital sensor geometry.
The novel interdigital sensors have been fabricated based on different geometric
parameters. Table 2 shows geometric parameters of four different interdigital
sensors and Fig. 9 shows the schematic of 1-5-25 and 1-11-25 configuration of
newly designed planer interdigital sensors [35].
V
R= ð1Þ
I
Vt = V0 sin ωt ð2Þ
where Vt is the voltage difference at time t, V0 is the amplitude of the signal, and the
angular frequency is ω.It is given by (ω = 2πf) expressed in radians/second and
frequency, f, in hertz [57].
For a linear system, the response signal, It, has a phase shift, θ, with amplitude of
I0 which can be explained by
It = I0 sinðωt − θÞ ð3Þ
Equation 1 (Ohm’s Law) can be used to calculate the impedance of the system
given by;
Vt V0 sin ωt
Z= =
It I0 sinðωt − θÞ
ð4Þ
sinðωtÞ
= Z0
sinðωt − θÞ
From Eq. 4, it can also be expressed in term of Euler’s relationship given by;
Vt = V0 ejωt ð6Þ
It = I0 ejðωt − θÞ ð7Þ
Vt V0 ejωt
ZðωÞ = = jðωt − θÞ = Z0 ejθ ð8Þ
It I0 e
The impedance now is in the form of real part (Z0 cosθ) and imaginary part
(Z0 sinθ).
Calculated impedance characteristics can be analysed using Nyquist plot and
Bode plot.
The Nyquist plot is a popular format for evaluating electrochemical impedance data
such as electrolytic solution resistance (Rs), electrode polarization resistance (Rp)
and double layer capacitance (Cdl). These parameters will be discussed in the
following sections. Nyquist plot represents imaginary impedance component
against real impedance component at each excitation frequency. The Nyquist plot
offers several advantages. The effects of solution resistance can be observed easily
using this format. One major drawback of the Nyquist plot is that there is no
information about the frequency in the Nyquist plot which makes it difficult to
calculate the double layer capacitance [55]. The Bode plot represents the absolute
impedance and the phase shift versus frequency.
Randle’s equivalent circuit is the most frequently discussed equivalent circuit [58]
used to interpret EIS experimental results in electrical form as shown in Fig. 11. It
Application of Practical Nitrate Sensor Based … 123
Rct
Z ′ ðωÞ = Rs + ð11Þ
1 + ω2 R2ct Cdl
2
124 M.E.E. Alahi et al.
ωR2ct Cdl
Z ′′ ðωÞ = − ð12Þ
1 + ω2 R2ct Cdl
2
The interdigital sensor is very sensitive to ambient temperature and humidity. The
first experiment was conducted to measure the impedance characterization in a
temperature range between 10 to 40 °C and obtain the linear relationship between
temperature and impedance. The second experiment was performed to get the
relationship between temperature and impedance by varying the humidity in dif-
ferent thermal condition. Then, the sensor was tested to observe the impedance
response in prepared diluted water samples from nitrates forms: ammonium nitrates
(NH4NO3) by commercial equipment-LCR meter.
The sensor was tested in the Milli-Q water in the temperature range from 10 to
40 °C. The characteristics of the sensor can be extended to accommodate the
behavior of it outside the range. The objective of this experiment was to observe the
impedance response of the sensor to different temperatures.
Application of Practical Nitrate Sensor Based … 125
Figure 13 shows the experimental setup and it involved high precision Hioki
3522-50 LCR meter, Hioki 4-terminal probe 9140, mercury thermometer,
SCILOGEX MS 7-H550 Digital Hotplate and computer for data acquisition. The
hotplate had the maximum limit from ambient temperature to 550 °C. In order to
measure the water sample at a lower temperature (below room temperature), the
water sample was stored in the freezer initially to let its temperature fall close to
0 °C and then taken out for testing at room temperature. The mercury thermometer
was immersed in the water to obtain a continuous temperature reading. Meanwhile,
the sensing surface was also submerged in the water and starts to collect data when
the temperature reached to a certain degree. The heating plate was used to heat the
water sample to reach the desired temperature. The Parylene coated sensor
(1-5-50A) was used for this experiment.
The Hioki 3522-50 was used to obtain the measured parameters for investigating
the impedance values in the frequency range between 1 Hz to 100 kHz. The probe
9140 had two sets of terminals connected to the sensor and LCR meter. The HCUR
and LCUR were used to measure the current flowing through the sensor and HPOT
and LPOT were used to measure the voltage across the sensor. Table 4.4.1.1 illus-
trates the technical specifications of the Hioki 3522-50 LCR meter (Hioki 2001).
The parylene coated 1-5-50A sensor was immersed into the MilliQ water in varying
thermal conditions. The resistance (Rs) and reactance (X) part of the impedance
Thermometer
Computer
LCR Meter
Hotplate
were evaluated and plotted in Fig. 14. The X-axis shows resistance part of impe-
dance in Ω and the Y-axis shows the reactance part of the impedance in Ω. With the
increase in temperature, the total impedance is decreased. Each data point on the
plot indicates the impedance at a specific frequency.
Figure 15 shows the relationship between the resistance of impedance and the
frequency. The resistance is decreased with the increasing in temperature and the
sensitive range is significant till 400 Hz.
Figure 16 shows the reactance (X) of MilliQ to changing temperature plotted
against frequency. This parameter is used to measure the dielectric properties of the
MilliQ in different temperature.
The experiment was repeated at different times to observe impedance behavior
and to take average results for the measurement of temperature for Parylene coated
sensor. Table 3 shows the average measurement results for resistance (Rs) and
ΔR R40 − R10
Slope = = ðEqu. 4.3.2.1Þ ð14Þ
ΔT T40 − T10
where:
• R40: Resistance value measured at 40 °C
• R10: Resistance value measured at 10 °C
• T40: Temperature at 40 °C
• T10: Temperature at 10 °C
Therefore, the slope for resistance part is (42392.83–118037.76)/(40–10) =
–2521.498. Similarly, replacing the ΔX with ΔR, the slope for reactance part is
(9658.2–47125.95)/(40–10) = –1248.92. Figure 17 shows the relationship between
the temperature and Rs. The resistance of impedance drops nearly linearly with the
increase in temperature.
128 M.E.E. Alahi et al.
Fig. 17 Relationship between the temperature and resistance part of impedance Therefore, the
temperature can be measured by Eq. 4.3.2.2
RT − R20
Tðo CÞ = + T20 ðEqu. 4.3.2.2Þ ð15Þ
slope
where:
• RT: Measured resistance value
• R20: Resistance value at 20 °C (which is the reference value)
• T20: Reference temperature is at 20 °C
Incubator
Sensor
LCR Meter
Portable Temperature
and Humidity Sensor
the impedance value in the frequency range between 1 Hz to 100 kHz, and the
computer is used for data acquisition.
Figure 19 shows the Nyquist plot for the Parylene coated sensor test at 25 °C
with varying humidity. From the Fig. 19, the total impedance is reduced with the
increase in humidity. The relationship between the resistive part of impedance and
frequency at different humidity is shown in Fig. 20. The most sensitive area of the
sensor is from 1 Hz to 400 Hz.
200000
54%
Rs (Ω)
150000
61%
100000
76%
50000
85%
0
1
2.257
5.094
25.95
58.57
132.2
298.4
673.4
1520
3430
7743
17480
39440
89020
11.5
Frequency (Hz)
1.60E+05
1.40E+05
1.20E+05
0.1 mg/L
1.00E+05
0.2 mg/L
Zimag (Ω)
8.00E+04
0.3 mg/L
6.00E+04
0.4 mg/L
4.00E+04
0.5 mg/L
2.00E+04
0.00E+00
0.00E+00 5.00E+04 1.00E+05 1.50E+05 2.00E+05 2.50E+05 3.00E+05 3.50E+05 4.00E+05
Zreal (Ω)
The NH4NO3 solution was used as a standard solution to calculate the nitrate
concentration in the collected water samples.
132 M.E.E. Alahi et al.
4.00E+05
3.50E+05
3.00E+05
0.1 mg/L
2.50E+05
Zreal (Ω)
0.2 mg/L
2.00E+05
0.3 mg/L
1.50E+05
0.4 mg/L
1.00E+05
sensitive area 0.5 mg/L
5.00E+04
0.00E+00
1.592
2.535
4.037
6.428
10.24
16.3
25.95
41.32
65.79
104.8
166.8
265.6
422.9
673.4
1
1072
1707
2719
4329
6893
10970
17480
27830
44310
70550
Frequency (Hz)
Frequency (Hz)
10970
17480
27830
44310
70550
1.592
2.535
4.037
6.428
10.24
25.95
41.32
65.79
104.8
166.8
265.6
422.9
673.4
1072
1707
2719
4329
6893
16.3
1
0.00E+00
2.00E+04
4.00E+04
Z imagniary (Ω)
6.00E+04
0.1 mg/L
8.00E+04
0.2 mg/L
1.00E+05
0.3 mg/L
1.20E+05
0.4 mg/L
1.40E+05
0.5 mg/L
1.60E+05
R − R0.3
C= + C0.3 ð17Þ
− 115543
where R is the real impedance of water sample, R0.3 is the real part of the impe-
dance at 0.3 mg/L of NH4NO3, C0.3 represent the concentration of 0.3 mg/L, which
Application of Practical Nitrate Sensor Based … 133
is a reference. Substituting the value of R0.3 and C0.3 into Eq. (8), the concentration
was calculated by:
R − 48056.78
C= + 0.3 ð18Þ
− 115543
5 Conclusion
References
10. D.C. Siu, A. Henshall, Ion chromatographic determination of nitrate and nitrite in meat
products. J. Chromatogr. A 804, 157–160 (1998)
11. P. Niedzielski, I. Kurzyca, J. Siepak, A new tool for inorganic nitrogen speciation study:
Simultaneous determination of ammonium ion, nitrite and nitrate by ion chromatography with
post-column ammonium derivatization by Nessler reagent and diode-array detection in rain
water samples. Anal. Chim. Acta 577, 220–224 (2006)
12. E. Pagliano, M. Onor, E. Pitzalis, Z. Mester, R. Sturgeon, A. D’Ulivo, Quantification of nitrite
and nitrate in seawater by triethyloxonium tetrafluoroborate derivatization—Headspace
SPME GC–MS. Talanta 85, 2511–2516 (2011)
13. N. Öztekin, M.S. Nutku, F.B. Erim, Simultaneous determination of nitrite and nitrate in meat
products and vegetables by capillary electrophoresis. Food Chem. 76, 103–106 (2002)
14. F. Manea, A. Remes, C. Radovan, R. Pode, S. Picken, J. Schoonman, Simultaneous
electrochemical determination of nitrate and nitrite in aqueous solution using Ag-doped
zeolite-expanded graphite-epoxy electrode. Talanta 83, 66–71 (2010)
15. A. Serra, S. Jorge, C. Silveira, J. Moura, E. Jubete, E. Ochoteco et al., Cooperative use of
cytochrome cd 1 nitrite reductase and its redox partner cytochrome c 552 to improve the
selectivity of nitrite biosensing. Anal. Chim. Acta 693, 41–46 (2011)
16. C. Xia, C. Xiaolan, W. Ning, G. Lin, Hierarchical CuO nanochains: synthesis and their
electrocatalytic determination of nitrite. Anal. Chim. Acta 691, 43–47 (2011)
17. F. Hu, S. Chen, C. Wang, R. Yuan, D. Yuan, C. Wang, Study on the application of reduced
graphene oxide and multiwall carbon nanotubes hybrid materials for simultaneous determi-
nation of catechol, hydroquinone, p-cresol and nitrite. Anal. Chim. Acta 724, 40–46 (2012)
18. M. Sohail, R. De Marco, K. Lamb, E. Bakker, Thin layer coulometric determination of nitrate
in fresh waters. Anal. Chim. Acta 744, 39–44 (2012)
19. J. Chen, H. Chen, C. Zhou, J. Xu, F. Yuan, L. Wang, An efficient upconversion luminescence
energy transfer system for determination of trace amounts of nitrite based on NaYF 4: Yb 3+,
Er 3+ as donor. Anal. Chim. Acta 713, 111–114 (2012)
20. Z.-T. Jiang, Y.-X. Guo, R. Li, Spectrophotometric determination of trace nitrite with brilliant
cresyl blue using β-cyclodextrin as a sensitizer. Food Anal. Methods 3, 47–53 (2010)
21. M. Pandurangappa, Y. Venkataramanappa, Quantification of nitrite/nitrate in food stuff
samples using 2-aminobenzoic acid as a new amine in diazocoupling reaction. Food Anal.
Methods 4, 90–99 (2011)
22. N. Cabaleiro, I. De La Calle, S. Gil, F. Pena, M. Costas, C. Bendicho et al., Simultaneous
ultrasound-assisted emulsification–derivatization as a simple and miniaturized sample
preparation method for determination of nitrite in cosmetic samples by microvolume UV–
vis spectrophotomety. Talanta 83, 386–390 (2010)
23. H. Filik, D. Giray, B. Ceylan, R. Apak, A novel fiber optic spectrophotometric determination
of nitrite using Safranin O and cloud point extraction. Talanta 85, 1818–1824 (2011)
24. M.D. Patey, E.P. Achterberg, M.J. Rijkenberg, P.J. Statham, M. Mowlem, Interferences in the
analysis of nanomolar concentrations of nitrate and phosphate in oceanic waters. Anal. Chim.
Acta 673, 109–116 (2010)
25. S. Senra-Ferreiro, F. Pena-Pereira, I. Lavilla, C. Bendicho, Griess micro-assay for the
determination of nitrite by combining fibre optics-based cuvetteless UV–Vis
micro-spectrophotometry with liquid-phase microextraction. Anal. Chim. Acta 668, 195–
200 (2010)
26. B.S. Gentle, P.S. Ellis, M.R. Grace, I.D. McKelvie, Flow analysis methods for the direct
ultra-violet spectrophotometric measurement of nitrate and total nitrogen in freshwaters. Anal.
Chim. Acta 704, 116–122 (2011)
27. M. Noroozifar, M. Khorasani-Motlagh, A. Taheri, M. Homayoonfard, Application of
manganese (IV) dioxide microcolumn for determination and speciation of nitrite and nitrate
using a flow injection analysis–flame atomic absorption spectrometry system. Talanta 71,
359–364 (2007)
28. A.S. Abu-Abed, R.G. Lindquist, Capacitive interdigital sensor with inhomogeneous nematic
liquid crystal film. Progress In Electromagnetics Research B 7, 75–87 (2008)
Application of Practical Nitrate Sensor Based … 135
29. A.V. Mamishev, K. Sundara-Rajan, F. Yang, Y. Du, M. Zahn, Interdigital sensors and
transducers. Proc. IEEE 92, 808–845 (2004)
30. F. Lisdat, D. Schäfer, The use of electrochemical impedance spectroscopy for biosensing.
Anal. Bioanal. Chem. 391, 1555–1567 (2008)
31. Y. Chen, C. Zhu, M. Cao, T. Wang, Photoresponse of SnO2 nanobelts grown in situ on
interdigital electrodes. Nanotechnology 18, 285502 (2007)
32. A.M. Syaifudin, K. Jayasundera, S. Mukhopadhyay, A low cost novel sensing system for
detection of dangerous marine biotoxins in seafood. Sens. Actuators B: Chem. 137, 67–75
(2009)
33. S. Mukhopadhyay, P. Yu, Electromagnetic field computation using COMSOL Multiphysics
to evaluate the performance of novel interdigital sensors, in Applied Electromagnetics
Conference (AEMC), 2009, pp. 1–4
34. M. Yunus, S. Mukhopadhyay, K. Jayasundera, A novel planar interdigital sensor for
environmental monitoring, in Sensors, 2009 IEEE, 2009, pp. 105–110
35. H. Arwin, Application of ellipsometry techniques to biological materials. Thin Solid Films
519, 2589–2592 (2011)
36. Y. Yang, G. Chiesura, G. Luyckx, T. Vervust, F. Bossuyt, J. Vanfleteren et al., In situ on-line
cure monitoring of composites by embedded interdigital sensor, in 16th European Conference
on Composite Materials (ECCM-16), 2014
37. A.I. Zia, A.M. Syaifudin, S. Mukhopadhyay, P. Yu, I. Al-Bahadly, C.P. Gooneratne et al.,
Electrochemical impedance spectroscopy based MEMS sensors for phthalates detection in
water and juices, in Journal of Physics: Conference Series, 2013, p. 012026
38. A.I. Zia, S.C. Mukhopadhyay, P.-L. Yu, I. Al-Bahadly, C.P. Gooneratne, J. Kosel, Rapid and
molecular selective electrochemical sensing of phthalates in aqueous solution. Biosens.
Bioelectron. 67, 342–349 (2015)
39. A. Zia, S. Mukhopadhyay, I. Al-Bahadly, P. Yu, C.P. Gooneratne, J. Kosel, Introducing
molecular selectivity in rapid impedimetric sensing of phthalates, in 2014 IEEE International
Instrumentation and Measurement Technology Conference (I2MTC) Proceedings, 2014,
pp. 838–843
40. A. Syaifudin, K. Jayasundera, S. Mukhopadhyay, A novel planar interdigital sensor based
sensing and instrumentation for detection of dangerous contaminated chemical in seafood, in
Instrumentation and Measurement Technology Conference, 2009. I2MTC’09. IEEE, 2009,
pp. 701–706
41. J. Fischer, H. Dejmkova, J. Barek, Electrochemistry of pesticides and its analytical
applications. Curr. Org. Chem. 15, 2923–2935 (2011)
42. M. Khafaji, S. Shahrokhian, M. Ghalkhani, Electrochemistry of Levo-Thyroxin on
Edge-Plane pyrolytic graphite electrode: application to sensitive analytical determinations.
Electroanalysis 23, 1875–1880 (2011)
43. L. Li, F. Yang, J. Yu, X. Wang, L. Zhang, Y. Chen et al., In situ growth of ZnO nanowires on
Zn comb-shaped interdigitating electrodes and their photosensitive and gas-sensing charac-
teristics. Mater. Res. Bull. 47, 3971–3975 (2012)
44. M. Dhull, A. Arora, Design of mems based microheater for enhanced efficiency of gas
sensors. J. Therm. Eng. Appl. 2, 16–21 (2015)
45. S.C. Mukhopadhyay, C.P. Gooneratne, A novel planar-type biosensor for noninvasive meat
inspection. Sens. J. IEEE 7, 1340–1346 (2007)
46. M.S.A. Rahman, S.C. Mukhopadhyay, P.-L. Yu, J. Goicoechea, I.R. Matias, C.P. Gooneratne
et al., Detection of bacterial endotoxin in food: new planar interdigital sensors based
approach. J. Food Eng. 114, 346–360 (2013)
47. C. Xhoffer, K. Van den Bergh, H. Dillen, Electrochemistry: a powerful analytical tool in steel
research. Electrochim. Acta 49, 2825–2831 (2004)
48. C.V. Vidal, A.I. Muñoz, Effect of physico-chemical properties of simulated body fluids on the
electrochemical behaviour of CoCrMo alloy. Electrochim. Acta 56, 8239–8248 (2011)
49. X. Li, K. Toyoda, I. Ihara, Coagulation process of soymilk characterized by electrical
impedance spectroscopy. J. Food Eng. 105, 563–568 (2011)
136 M.E.E. Alahi et al.
Abstract The Wireless Sensors Network (WSN) is nowadays widely used to build
decision support systems to solve many real-world problems. One of the most
interesting fields having an increasing need of decision support systems is
agricultural environment monitoring. Agricultural environment monitoring has
become an important field of control and protection, providing real-time system and
control communication with the physical world. An intelligent and smart WSN
system can collect and process large amount of data from the beginning of the
monitoring and manage air quality, soil conditions, to weather situations. Wireless
sensors are used in a palm oil plantation to monitor the humidity, temperature and
other parameters of a mother palm to find the best time to start a controlled
pollination process. Pollination is the process of transferring pollen from the male
reproductive organ to the female reproductive organ of a flower. The inaccuracy in
determining pollination readiness of the oil palm flower could potentially cause a
detrimental effect on the palm oil industry in the long run.
1 Introduction
Palm oil has become increasingly popular as it provides rich sources of α and
β-carotene and vitamin E, which play important roles in blood coagulation, sup-
pression of cholesterol production and cancer inhibition. Besides its nutritional
value, it is a well-known raw material for oleo chemical industry and bio-fuel for
automobiles. Around 85 % of the palm oil production is concentrated in Indonesia
and Malaysia, but its use is rapidly spreading in whole world as China and India are
the main importing countries, next to the EU.
At present, Malaysia accounts for an overwhelming contribution to world’s palm
oil production and export which is 39 % and 44 %, respectively. A massive 17.73
million tons of palm oil and 2.13 tons of palm kernel oil have been produced from
about 4.49 million hectares of land. Malaysia has a vital role to play in achieving
the rising global need for oils and fats, as Malaysia is one of the major producer and
exporter countries of palm oil and palm oil products.
Pollination is an important process in oil palm breeding as correct method
increases yield and quality of palm oil production. During pollination process the
conditions need to be monitored thoroughly within a time frame to ensure genetic
purity that will increase yield. Currently, this monitoring process of pollination is
very manual and labour intensive job. Workers need to climb an average height of 3
stories building only to check the conditions of the pollination. Climbing is dan-
gerous as risk of falling down can cause disability and deaths.
Advances in sensor and wireless radio frequency (RF) technologies and their
convergence with the Internet of Things (IoT) offer vast opportunities for appli-
cation of sensor systems for agriculture industry. Emerging wireless technologies
with low power needs and low data rate capabilities have been developed which
suit precision agriculture [1–5]. The WSNs have become the most suitable tech-
nology to monitor and control the agricultural environment.
Wireless Sensor Network (WSN) is the ideal candidate to provide effective and
economically viable solutions for a large variety of applications ranging from health
monitoring, agriculture, environmental monitoring to military operations. WSN is a
very important component in IoT and integrates the knowledge of sensors,
automation control, digital network transmission, information storage and infor-
mation processing. WSN market is forecast to rise from $0.45 billion in 2012 to $2
billion in 2022 [6, 7].
Current pollination processes for oil palm pollination relies heavily on human
dependence. It involves manual checking of trees for pollination readiness and is
more of a guessing game rather than an exact science. For the process of oil palm
pollination to be efficient and cost-effective, a logical solution would be to automate
the process and reduce dependence on human tacit knowledge by transferring it into
a system. Therefore, a remote monitoring system for pollination readiness using
WSN is introduced to help the field workers reduce the dangerous task load besides
improving the quality and productivity of yield. Figure 1 shows an oil palm
plantation in Malaysia.
Using Wireless Sensor Networks to Determine Pollination … 139
alerting farmers at the onset of frost damage and providing better microclimate
awareness.
Pierce and Elliot developed a regional and on-farm sensor network that provides
remote, real-time monitoring and control of farming operations in two agricultural
applications, a weather monitoring network and an on-farm frost monitoring net-
work [9].
Precision Agriculture: The development of WSN applications in precision
agriculture makes it possible to increase efficiencies, productivity and profitability
in many agricultural production systems, while minimizing unintended impacts on
wildlife and the environment. The real time information obtained from the fields
can provide a solid base for farmers to adjust strategies at any time. Instead of
making decisions based in some hypothetical average condition, a precision agri-
culture approach recognizes differences and adjusts management actions
accordingly.
Greenhouse: Greenhouse plants and crops can benefit a lot from the use of
WSNs, because inside the greenhouse the crop conditions such as climate and soil
do not depend on external factors and the implementations are thus easier than in
outdoor applications. The first application of WSN in a greenhouse was reported in
the year 2003. It was a monitoring and control system developed using Bluetooth
protocol. Since then, several applications have been developed, most of them
making use of IEEE 802.15.4/ZigBee protocol. Gonda and Cugnasca presented a
proposal of a distributed greenhouse control and monitoring system using ZigBee
[10]. Lea-Cox et al. developed a WSN in a greenhouse, that integrates a number of
sensors which can measure substrate water, temperature, electrical conductivity,
daily photosynthetic radiation and leaf wetness in real-time [11].
Temperature
Humidity
Soil moisture is the water that is held in the spaces between soil particles. Soil
moisture data will be used to determine when to irrigate and amount of water for
each irrigation. Based on observation, most of the time, the fields are over irrigated.
Over irrigation prevents the nitrogen to be used by roots, and lack of oxygen in the
root area. Different crops need different level of moisture to maximize the
productivity.
Temperature is another important parameter in agriculture field. Low tempera-
tures will slow down the photosynthesis process, thus growth is slowed and this
results lower yields. Different plants and crops grow best at different temperature.
Humidity is one of the very important parameter in agriculture field. Relative
humidity (RH) is the ratio of actual water vapour content to the saturated water
vapour content at a given temperature and pressure in percentage. RH indirectly
affects leaf growth, photosynthesis, pollination and generate pests and various types
of plant diseases.
Temperature and humidity are closely linked in a greenhouse environment. The
decrease of the relative humidity is a sign of increased air temperature because cold
air has a lower moisture-holding capability than warmer air. Humidity control is
very important to prevent plant diseases. The range of healthy RH for plants is from
50 to 70 % [13]. If humidity is below 50 % for longer duration, the plant growth
will be affected due to loss of water from leaves might be faster.
Electrical conductivity will determine the suitability of an environment for the
plants and crops. EC estimates the amount of total salts or ions in the soil. In
general, EC shows the level of ability of the soil water to carry an electrical current.
In other words, EC level is an indication of the amount of nutrients available for
plants and crops to absorb.
Figure 3 shows one of the wireless sensors developed by MIMOS that can be
used to measure the above environment parameters.
Malaysia’s position as the world’s leading palm oil country has allowed the
industry to flourish in the way it has never been before. The results in the churning
out of a wide variety of products as a result of continuous R&D efforts. With the
escalation in the volume of palm oil products and the fruit itself, it becomes nec-
essary to produce high quality seedlings from high yield trees by automating and
transforming the current process of pollination to meet growing demands.
The current methods of pollination involve tacit knowledge of checking whether
a palm is suitable for pollination and manual labour being deployed in fields of vast
sizes spanning over thousands of hectares in the nation itself. The drawbacks of
such a system is late bagging during the pollination process that ultimately impacts
overall yield and the process, being labour dependent, relies on a sizeable number
of foreign manpower to work the fields. The current gaps are:
• Human dependence to monitor trees for within the plantation and determine if
they are ready for pollination.
• Inaccurate human-based decision making based on the determining of the
pollination window.
• Vast plantation areas to be monitored to catch the pollination window to
maximise yield.
Controlled pollination consists of transferring pollen from one plant to the receptive
female reproductive organs of another plant while excluding all other pollen.
Controlled pollination plays an important role in the commercial crops production,
especially in oil palm production, as it facilitates producing progeny with favorable
genetic trait inherited from the parent plants.
Controlled pollination also contributes to offspring generation with more genetic
uniformity compared to the naturally pollinated plants. The success of the polli-
nation process depends greatly on the ability to determine the period at which the
flower or floral organs is receptive towards the pollens, namely the pollination
period. The introduction of pollens out of the pollination period, either too early or
too late, will result in the abortion of the unfertilized flowers or the development of
pathernocarpic fruits and thus resulted in a poorly pollinated bunch. Thus, the
pollination stage must be accurately determined for the pollination process to take
place. The ability to correctly determine the pollination readiness of an oil palm
flower provides a significant advancement in the oil palm industry.
Research has shown that the average receptive period of a female flower lasts
from 3 to 10 days, depending on the species of the flower. When a female flower is
found to be in its receptive period, pollination is usually repeated on three
144 M.R.M. Kassim and A.N. Harun
Many wireless technologies were developed over the last few decades to solve
many real-world problems. Wireless technologies and sensor networks have
become the most suitable technologies to monitor and control the agricultural
environment. These technologies are capable of providing greater level of flexi-
bility, mobility at an affordable cost and low power consumption. There are more
than 10 wireless technologies available for industrial applications however the most
exploited wireless technologies are Wi-Fi, ZigBee and Bluetooth. The selection of
these technologies dependents on the nature of the applications.
Bluetooth: Bluetooth is a short range wireless communication intended to
replace portable or fixed device. Bluetooth operates at 2.4 GHz ISM band. Blue-
tooth has the capability to simultaneously handle data and voice transmissions and
used in applications such as hands-free handsets for voice calls.
Wi-Fi: Wi-Fi requires high frequency radio waves, usually 2.4 or 5 GHz for data
transmissions and supports several hundred meters between two places. Wi-Fi was
initially used in mobile computing devices in LANs but now diversified to more
applications such as Internet and consumer electronics. There are few generations
of Wi-Fi standards available and carter for different applications.
Using Wireless Sensor Networks to Determine Pollination … 145
8 Platform Architecture
network with mesh network capability [15]. Router has to be powered up to ensure
network is always available for the sensor node to send sensor data to the Gateway
[16]. Router use external battery that has enough capacity to provide power longer
life. Figure 5c shows MSCAN router product that has cable to connect to the
external battery [17] Fig. 6.
MSCAN Gateway: MSCAN Gateway has two main module; Collector and
Main MCU board. Collector is a ZigBee coordinator that control router and Sensor
Node in the wireless sensor network. Collector interfaces to the main CPU using
serial interface. Main CPU board has GPS module, GSM module and other control
circuit for the gateway.
Collector forwards sensor data in MBPS packet format to the Gateway main
CPU which sends the packet to LSMS server using GPRS or SMS over GSM
network. Gateway is located in the field, powered by battery and solar panel.
Figure 5d shows MSCAN gateway product.
9 System Architecture
ApplicaƟon
Layer PDA App Web App
Physical
Layer
wind speed and also internal weather conditions in a greenhouse such as temper-
ature, pH, relative humidity, moisture, CO2, EC, and the intensity of light. The
information collected in real-time is then converted and stored into a database
through a web server, which provides critical information to users.
Physical Layer: The physical layer consists of wireless environmental, soil
sensors (temperature, humidity, moisture, EC, pH and CO2), actuators and CCTV
cameras. These sensors will collect environmental and soil information from indoor
(e.g. greenhouses) and outdoor environment periodically and sent the data to sensor
manager.
The measurement data include temperature, humidity, soil and water pH, soil
moisture, rainfall quantity, EC, wind speed and direction, water level, water con-
sumption, etc. The data is further analyzed and converted into appropriate formats
and finally sent to Middle Layer for storing in databases.
Middle Layer: The middle layer (ML) supports communication between
physical layer and application layer. Currently this architecture is using ZigBee as
the wireless communication technology. The ML converts the data received from
the physical layer and store it in a database. The ML will also support the data
request from the application layer. The Web server and App server allows users to
monitor data processed by the components through a Web browser, PDA or smart
phone at any time.
Using Wireless Sensor Networks to Determine Pollination … 149
The sensor manager converts and analyses this data and store it in a database.
The web server will fetch the sensing data stored in the database at a fixed interval
and send it to the GUI. The users can view the sensor data through the GUI in the
Application Layer. The GUI will also display and send emails and SMS if there is
any alerts and warnings. The CCTV cameras will collect real-time image infor-
mation (real-time streaming) on the indoors and outdoors and send it to image
manager. The image manager will store it in the database. The user can access the
image data at anytime and anywhere through the GUI using PDAs or Smart Phones.
The web server used to host sites and deliver web pages content to clients who
can access them through the Internet. Another use of web server is to connect any
component of the physical layer to browse any monitoring or actuating device in
real time.
Application Layer: The application layer consists of various GUIs to provide
suitable information and support user requests. The users can access environmental
information and status through various platforms such as smart phone, web and
PDA. This layer also consists of monitoring support, actuating support, alerting
support and decision making support.
(a) The female inflorescences which are going to be pollinated are identified. The
female inflorescences which are anticipated to undergo pollination within
certain number of days are covered with a bag attached with a main humidity
sensor on the interior surface of the bag.
(b) A reference humidity sensor is installed inside another bag similar to the bag
covering the main humidity sensor. The bag containing the reference humidity
sensor is placed in a position similar to the bag containing the main humidity
sensor in terms of the height of the surrounding trees.
(c) The humidity value is taken by the main humidity sensor and the reference
humidity sensor at certain times of the day. Each reading of the humidity
detected will be compiled into a database in a matrix form, wherein the
humidity readings will be arranged in one row in the matrix. For each day, a
new row is added to the humidity matrixes of the main humidity [Hm] and the
reference humidity [Hr]. The humidity profiles are compared regularly to
determine whether pollination has occurred.
(d) A general formula for determining pollination is shown below, where polli-
nation is deemed to have occurred when:
In which [α] and [β] are the coefficient matrixes and [HR] is the triggering
matrix. The meaning of the above equation is that if a predefined combination
of the humidity of the main bag is greater than a predefined combination of the
humidity of the reference bag by the triggering (threshold) value, then polli-
nation is deemed to have occurred. The threshold humidity values are based
on pre-existing information of humidity values emitted from flowers ready for
pollination that were collected over the years using sensors in different
plantations.
(e) Once pollination is determined to have occurred, the system will notify the
user regarding the event on the GUI (Web) and also by sending a message to
the user’s mobile phone. Upon receiving the notification, the user will intro-
duce the male pollen to the female inflorescence.
(f) The bag covering the inflorescence is opened and the pollens of the desired
progeny are transferred to the inflorescence. This can be carried out, for
example, by applying the pollens onto the inflorescence by using a hand
puffer. In a preferred embodiment, the pollination has to be performed on the
same day or the subsequent day of the receipt of the notification. This will
complete the controlled pollination process Fig. 8.
Using Wireless Sensor Networks to Determine Pollination … 151
(a) Gateway installation - fixed pole and (b) Router installation use portable pole.
powered by car battery & solar.
The above mentioned architecture in Fig. 7 has been deployed to monitor the
humidity of a mother palm to find the best time to start a pollination process. Sensor
Node will send humidity value inside pollination bag at a predefine interval and
forward by Gateway to PMS using SMS/GPRS over GSM.
The GUI of PMS present important information such as list of data from sensor
node, sensor status, pollination alert list and sensor battery level. Figure 9 shows
the GUI for PMS. Each of the node in the GUI represent a sensor node installed at
(b)
(a)
palm oil frond. The colour of each node represent the pollination status of each
palm tree. The trees which are ready for pollination is represented in blue colour.
Figure 10 shows the router and gateway installed at the palm oil plantation.
Figure 11 shows the sensor node and pollination bag with main humidity sensor.
In Phase 1 of this project, 200 palm oil trees are deployed with wireless sensors.
Each palm oil tree will be tagged for identification purposes. The initial threshold
value for pollination readiness are based on values obtained from manual obser-
vation. In Phase 2, another 300 palm oil trees are deployed. The improved threshold
value from Phase 1 was used in this phase. The final results show that the threshold
value from Phase 1 is very consistent throughout this project.
The data collection after the certain days of bagging will be carefully monitored
to see whether any pollination readiness has taken place. Figure 12 shows a set of
data (humidity from a pollination bag) collected on the nth day after bagging.
Figure 13 shows the graph of humidity difference between the humidity value
from main sensor and reference sensor. This graph will help us to determine
whether pollination has taken place or not.
Once pollination have occurred, the system will notify the user regarding the
event on the GUI (Web) and also by sending a message to the user’s mobile phone.
Upon receiving the notification, the user will introduce the male pollen to the
female inflorescence on the same day of receipt of the notification to complete the
controlled pollination process.
In an oil palm plantation, 69 palm oil trees were selected randomly to install with
sensors. The purpose is to test pollination readiness (CP) success rate. In this test,
PMS manage to detect a total number of 64 trees means these oil palm trees are
ready for pollination. We can conclude that the IPMS success rate in detecting
pollination readiness is 93 %.
2:00:35am
4:00:01am
4:00:02am
6:59:08am
6:59:09am
7:59:07am
8:59:06am
8:59:07am
9:59:11am
2:00:34am
10:59:15am
154 M.R.M. Kassim and A.N. Harun
15.00
10.00
5.00
0.00
13 Conclusions
Using the key elements of information, technology, and management, PMS can be
used to increase production efficiency, improve product quality, conserve energy
and protect the environment. Moreover, PMS with minor customization can also be
applied to other agricultural environmental applications.
Some of the benefits of using smart agriculture system for oil palm pollination
are as follows:
• To be able to trigger pollination period within certain days.
• To increase yield.
• To increase productivity.
• Possible reduction of labour dependency.
• To be able to directly communicate with the sensor nodes to get real time data
using WSN.
WSN ecosystem can be deployed in various sectors provided that the end-to-end
devices and solutions are properly configured. WSN is the key for deployment of
IoT as it provides large pool of sensors to be configured to the end devices. Among
the critical issues in palm oil pollination readiness monitoring is the RF range due to
the coverage of palm leaves and also terrain of the plantation area [20, 21]. The
second issue is on the power management for the sensor node which uses battery to
operate. However, in this project we only target sensor node battery life up to
2 years due to battery life time limitation [22]. We can conclude that the PMS
success rate in detecting pollination readiness is more than 90 %. The network data
security and power management with efficient algorithms will also be given higher
priority in our future work.
Using Wireless Sensor Networks to Determine Pollination … 155
References
1. M.S. Chaudhari, R. Jaiswal, C. Birhade, V. Bhapkar, Wireless sensor network as a tool for
supporting agriculture in the precision irrigation. Int. J. Adv. Res. Comput. Commun. Eng. 4
(2), 311–314 (2015)
2. N. Wang, N. Zhang, M. Wang, Wireless sensors in agriculture and food industry—recent
development and future perspective, review. Comput. Electron. Agric. 50, 1–14 (2006)
3. M. Sherine, A. El-Kader, B.M. Mohammad El-Basioni, Precision farming solution in Egypt
using the wireless sensor network technology. Egyptian Inform. J. 14(3), 183–250
4. Sagar Lachure, Amol Bhagat, Jaykumar Lachure, Review on precision agriculture using
wireless sensor network. Int. J. Appl. Eng. Res. 10(20), 16560–16565 (2015)
5. S.R. Nandurkar, V.R. Thool, R.C. Thool, Design and development of precision agriculture
system using wireless sensor network, in 2014 First International Conference on Automation,
Control, Energy and Systems (ACES), 2014, pp. 1–6
6. P. Harrop, R. Das, Wireless sensor networks 2012–2022. IDTechEx (2012)
7. P. Harrop, Wireless sensor networks and the new Internet of things. Energy Harvest J. (2012)
8. T. Arampatzis et al., A survey of applications of wireless sensors and wireless sensor
networks, in Proceedings of 13th Mediterranean Conference on Control Automation, Turkey,
2005, pp. 719–724
9. F.J. Pierce, T.V. Elliott, Regional and on-farm wireless sensor networks for agricultural
systems in Eastern Washington. Comput. Electron. Agric. 61, 32–43 (2008)
10. L. Gonda, C.E. Cugnasca, A proposal of greenhouse control using wireless sensor networks,
in Proceedings of Computers in Agriculture and Natural Resources, 4th World Congress
Conference, Orlando, FL, USA, 24–26 July 2006
11. J.D. Lea-Cox, G. Kantor, J. Anhalt, A. Ristvey, D.S. Ross, A wireless sensor network for the
nursery and greenhouse industry, in Proceedings of Southern Nursery Association Research
Conference, Atlanta, GA, USA, 8–9 August 2007
12. S. Roy, D. Anurag, S. Bandyopadhyay, Testbed implementation of a pollution monitoring
system using wireless sensor network for the protection of public spaces. Recent Advances in
Broadband Integrated Network Operations and Services Management, pp. 263–276 (2011)
13. T. Ahonen, R. Virrankoski, M. Elmusrati, Greenhouse monitoring with wireless sensor
network. Department of Computer Science, University of Vaasa (2003)
14. M. Beldjehem, Toward a multi-hop, multi-path fault-tolerant and load balancing hierarchical
routing protocol for wireless sensor network. School of Information Technology and
Engineering, University of Ottawa, Ottawa, Canada (2013)
15. I. Stojmenovic, Ad hoc and sensor wireless networks 1 (2005)
16. S.K. Singh, M.P. Singh, D.K. Singh, Routing protocols in wireless sensor networks: a survey.
Int. J. Comput. Sci. Eng. Surv. (IJCSES) 1(2) (2010). doi:10.5121/ijcses.2010.120663
17. A. Kellner, O. Alfandi, D. Hogrefe, A survey on measures for secure routing in wireless
sensor networks. Ashdin Publ. Int. J. Sens. Netw. Data Commun. 1 (2012)
18. A. Henderson, A review of pollination studies in the Palmae. Bot. Rev. 52, 221–259 (1986)
19. G.A. Santos, P.A. Batugal, A. Othman, L. Baudouin, J.P. Labouisse (eds.), Manual on
standardized research techniques in coconut breeding (IPGRI, Rome, Italy, 1996)
20. M.R.H. Khan, R. Passerone, D. Macii, RF-level power consumption measurement (RF-PM)
for zigbee wireless sensor network-towards cross layer optimization. University of Trento,
Via Sommarive, I-38100 Povo, Italy
21. J. Tyzzer, Easing the challenge of RF. Nordic Semiconductor White Paper (2010)
22. I.M. Yusoff, A.N. Harun, S.S. Ramli, RF and battery performance study for WSN in palm oil
plantation applications. IEEE Conf. Wirel. Sens. (ICWiSE) 2014, 85–89 (2014)
23. C. Alippi, R. Camplani, C. Galperti, M. Roveri, A robust, adaptive, solar-powered WSN
framework for aquatic environmental monitoring. IEEE Sens. J. 11(1), 45–55 (2011)
Time Domain Reflectometer
for Measuring Liquid Waste Levels
in a Septic System
Abstract The objective of this study was to determine the level of water in a septic
tank by inserting a waveguide at the top of the tank. A TDR (i.e. waveguide) probe is
immersed in the water tank, through which a pulse signal is launched. The impe-
dance mismatch occurs at the air-water interface and the bottom of the tank, which
results in the reflection of the signal. This reflected signal is algebraically added to
the incident signal. From the time difference between the incident and reflected
signal, the distance between the TDR and the surface of water is calculated. The
distance is generally twice the time traveled by the wave. Thus, the calculated
distance helps in finding out the length of water in the tank. Additionally, the TDR
was compared with a Doppler radar system to determine if the Doppler system
would be a feasible non-contact measurement system as compared to the TDR.
1 Introduction
presence of toilet tissue. Some of this material breaks down readily and some of the
material floats. Since this material is saturated with water is should not significantly
affect the reflection from the air water interface.
The TDR technique was introduced in the early 1980s. Since its introduction, it
has gained interest by many fields of science for various applications [3]. TDR has
wide variety of applications and is very efficient compared to other measurement
techniques. TDR is also used for measuring impedance of the transmission lines,
dielectric constant of the medium of transmission, electrical conductivity and types
and position of faults in the transmission lines. A brief mention of some of the
previous work performed in the area. Tomi Engdahl designed a simple TDR using
74AC14 IC, which was primarily used to detect faults along the cable. He also
analyzed the outputs and classified them into different types of distortions [4]. Jim
Bartling worked on a “Low cost, high resolution Time Domain Reflectometer”
which was designed using a 16-bit microcontroller and a Charge Time Measurement
Unit (CTMU) peripheral on-chip that is used to measure time with a high degree
precision and resolution. The main goal of his work was to find out transmission-line
length, location of faults due to opens and location of faults due to shorts [5]. Jim
Bartling also developed a “Time Domain Reflectometer for liquid level measure-
ment”. He has designed a TDR using PIC microcontroller and used it with a probe to
measure the level of liquid. Time measurement is again accomplished using CTMU.
It provides information of the fluid composition as well [6]. Brian Kenner and John
Wettroth designed a TDR using Intel 87C51 microcontroller to measure length and
termination impedance of co-axial cables commonly used in computer networks,
like Ethernet, Arcnet etc. A high-speed comparator is enabled to measure the length.
The software consists of approximately 3 K of 8051 assembly language [7].
James R. Andrews analyzed “Time Domain Reflectometry and Timer Domain
Transmission Measurement fundamentals” in 2004. The main aim of his study was
to identify the nature of the Device Under Test (DUT) [8]. Scott B. Jones, Jon M.
Wraith and Dani Or discussed “Time domain reflectometry measurement principles
and applications” in 2002. Their main motive was to determine the porous media
water content based on dielectric properties and electrical conductivity of the
medium [3]. Kevin M O’Conner and Charles H. Dowding worked on “Real-Time
monitoring of infrastructure using TDR technology”. His study reveals the summary
of principles involved in various applications of TDR. This work was primarily
intended to provide background on the principles involved in geotechnical and
infrastructure applications of TDR [9]. Xinbao Yu and Xiong Yu studied on
“Measurement of simulated scour by Time domain reflectometry” in 2006. “TDR
instrument and analyses framework can be potentially refined into a useful tool for
bridge scour surveillance”. Their study shows the determination of dielectric con-
stant, electrical conductivity, the depth of the scour and properties of the soil sedi-
ments (porosity, density) [10]. A. Thomsen, B. Hansen and K. Schelde proposed a
TDR technique for measuring water level in tanks collecting surface water runoff
and soil moisture in their paper “Application of TDR to water level measurement” in
2000. They designed a two-wire closed loop TDR probe with balun transformer for
transforming impedance from 50 to 200 Ω. The water level calibration depended on
160 S.R. Mamidi et al.
2 Scope of Work
The objective of this study is to determine the level of water in a septic tank by
placing a waveguide at the top of the tank. A TDR (i.e. waveguide) probe is
immersed in the water tank, through which a pulse signal is launched. The impe-
dance mismatch occurs at the air-water interface and the bottom of the tank, which
results in the reflection of the signal. This reflected signal is algebraically added to
the incident signal. From the time difference between the incident and reflected
signal, the distance between the TDR and the surface of water is calculated. The
distance is generally twice the time traveled by the wave. Thus, the calculated
distance helps in finding out the length of water in the tank. Additionally, the TDR
was compared with a Doppler radar system to determine if the Doppler system
would be a feasible non-contact measurement system as compared to the TDR.
TDR was basically developed to check for faults in cables. That where it got its
name as ‘cable tester’ [3]. TDR is also called guided-wave RADAR, i.e. the pos-
itive pulse generated by TDR, travels along the cable or a metal rod under test. The
signal waveform is seen on the Oscilloscope, which is connected at the source and
DUT connection. This explains the basic TDR setup. It generates a step signal or an
impulse with fast rising and falling edges. The rise and fall time of the pulse
matters, the steeper the slope the greater the resolution [10].
Once the incident pulse is fed into the DUT, the waveform displayed by the
Oscilloscope is monitored. If the load impedance matches the characteristic
impedance of the line, no reflection is seen. Since the whole signal transmitted is
absorbed at the load. The only thing seen on the Oscilloscope screen is the incident
pulse. But, if the load impedance does not match the characteristic impedance of the
Time Domain Reflectometer for Measuring Liquid … 161
Fig. 1 (Top) Oscilloscope display when load impedance matches characteristic impedance.
(Bottom) Oscilloscope display when load impedance mismatches characteristic impedance [14]
line, part of the incident pulse is reflected back to the source, which is recorded by
the Oscilloscope as in the Fig. 1. This reflected pulse can be easily noticed as there
is a time difference between incident and reflected pulses [14].
For cable testing, a cable of known length with velocity of propagation, Vp, is
taken, and reflections from discontinuities are observed on the Oscilloscope screen.
The time difference and the voltage difference between the incident and the
reflected peaks are noted, from which the distance of the discontinuity from the
source can be calibrated. The time difference is the time taken by the wave to travel
the round trip (i.e. two times the length, L) between source and discontinuity.
Therefore the time difference is:
2L
t= ð1Þ
v
where, L is the length of the discontinuity from the source, t is the time difference
(total time taken from source to discontinuity and again back to source) and v is the
velocity of light in the medium. TDR is also used to measure the properties of the
materials surrounded by the probe. Typical properties are impedance, dielectric
constant, its conductivity and velocity of propagation.
In the previous section, TDR basics and transmission line basics were discussed.
Their functions and various applications have been analyzed. In this chapter the
proposed circuit and the experimental setup for measuring the water level in the
tank will be studied.
The TDR circuit discussed starts Wolke’s [15] TDR circuit. This simple circuit
comprises of a Schmitt trigger for generation of a square wave which acts as the
source of incident voltage pulse.
The Fig. 2 shows the block diagram of the experimental setup. The setup is
divided into three main parts. The signal generation, TDR probe and Oscilloscope.
The signal generation is carried out by the simple TDR circuit built in the labo-
ratory. The Oscilloscope is used to check the waveforms and also to extract the data
in the digital form. The probe used here is a bi-axial probe built using two brass
rods. All these three parts are inter-connected using a BNC Tee-connector.
In the sections below, all the parts will be discussed. The Oscilloscope used here
is Agilent Technologies Digital Storage Oscilloscope DSO1102B. It is a 2-channel,
100 MHz Input bandwidth oscilloscopes with a sample rate of 500 MSa/s–1 GSa/s
[16]. Signal generation is done by a simple TDR circuit, built using 74AC14 IC as
shown in the Fig. 3. It generates a square wave with fast rising and falling edges,
Time Domain Reflectometer for Measuring Liquid … 163
Fig. 3 Schematic of TDR circuit [15]. Signal generation is done by a simple TDR circuit, built
using 74AC14 IC
164 S.R. Mamidi et al.
which is transmitted to the biaxial cable that is immersed in the water tank.
This TDR circuit has the Schmitt trigger in the circuit which yields the square wave.
The TDR probe used here is a bi-axial probe (i.e. 2-wire transmission line). It is
made out of two 3/32 brass rods which are 36 inches long and have 2.38 mm
diameter. These two brass rods which have to be immersed in water and are
connected to a coaxial cable at one end. First, the outer conductor of the coaxial
cable is unbraided and connected to one of the brass rods, and the inner conductor is
connected to the other rod using crimp connectors. The main stretch of the bi-axial
cable in targeted to be at 50 Ω in air to minimize the reflections when the tank has
no water in it. The connections to the bi-axial cable create a static reflection node
which does not change.
3 Experimental Procedure
The TDR circuit and probe built are now used to calibrate the level of water in the
tank. Firstly, the TDR launches the voltage pulse into the probe. This incident pulse
reflects at the bottom of the tank and the air-water interface, due to change in
impedance. As the probe is open at the end, reflection happens at the end of the
probe. At air-water interface, due to large difference in the dielectric constants of air
and water (i.e. 1 and 80.4 respectively) impedance mismatch occurs. The incident
and reflected signals are captured on the Oscilloscope screen. The time difference
between the two peaks (i.e. the incident and reflected peaks) is found out, which in
return helps to find out the height of water in the tank.
The Fig. 4 shows the setup in the water tank. The propagation velocity of the
probe inserted in the medium must be known to find out the height of water. The
wave propagates through coaxial cable, air and water mediums. The velocity of
propagation is different in each section. The total time will be the summation of
travel times in all the phases (coaxial cable, air and water). Then from the measured
travel time, it will be easy to find out the level of water in the tank [12].
The coaxial cable used here has a solid polyethylene dielectric with dielectric
constant 2.23, and characteristic impedance 50 Ω. The velocity factor of coaxial
cable is 67 %, i.e. the velocity of propagation of wave in coaxial cable is 0.67 times
the velocity of propagation in air (3 × 108 m/s). Therefore, the velocity of prop-
agation of wave in coaxial cable is 2 × 108 m/s.
The relative dielectric constant of water is 80.4. The speed of EM wave prop-
agating through a dielectric medium is given by:
V = c ̸ðε1r ̸2 Þ, ð2Þ
Fig. 4 Probe setup in the tank describing the various reflection points
2*Length
time = ð4Þ
velocity
x ðl − hÞ h
t=2 + + ð5Þ
Vc Va Vw
The experiment was carried out under laboratory conditions to simplify the
experimental environment. Below is the Fig. 5, which shows experimental setup
and TDR circuit. During initial testing of the TDR circuit, a square wave which did
not have clean edges was obtained. There was little bit ringing on the edges, which
will be used to identify the location of the reflected signal in order to calculate the
distance between the source and surface of water. The square wave generated by the
TDR circuit has amplitude of about 5.2 V and frequency 2.84 kHz.
The waveform, in Fig. 6, is the screen-shot of the Oscilloscope. Here, the x-axis
represents time and y-axis represents voltage. The first peak here is the incident
peak, whereas the second peak is the sum of incident and reflected peaks. The other
peaks are the ringing at the edge of the square wave. Later, water is added into the
tank, and the readings are noted at regular intervals (for every 100 ml addition of
water). The time difference between the incident and reflected peaks is observed.
The height of the water in the tank is also noted manually by measuring it with a
scale, which will later be used for verification. Table 1 shows the data from the
experiment. These results are taken directly from the oscilloscope output. This is
because, as the water height increases, the signal travels through more water than
before. Therefore, the time taken to travel the complete path also increases. The
bottom of the tank, i.e. the probe termination and air-water interface are both
considered as discontinuities. But due to the close proximity of the discontinuities
placed at close proximity to each other the reflection was smoothed into a single
discontinuity. As the water level increases, the distance between the discontinuities
Time Domain Reflectometer for Measuring Liquid … 167
Fig. 5 Experimental setup for the TDR experiment. The image on the right is the pulse generator.
The graduated cylinder allows for testing against multiple contaminated water types
also increases and therefore there are two reflections which are observed. For an
example, many points can be made out from Fig. 7. The red line represents the x-y
plot at 100 ml and blue line at 200 ml. The second peak reflects from the bottom of
168 S.R. Mamidi et al.
Fig. 7 Waveforms at 100 ml (red line) and 200 ml (blue line) water
the tank. We can see that as the height of the water increases, the reflected signal
shifts toward right. Figure 8 is the waveform at 800 ml water, i.e. 27.4 cm height
and Fig. 9 is the data for the pulse edge of the 34 cm height (1000 ml).
Table 2 shows the data obtained during the measurement of the water height and
Fig. 10 shows a plot of the actual water column height versus the calculated height.
For many application this amount of accuracy is sufficient. Many wastewater
storage tank systems for the transportation industry only need to know when the
tank is full or when the tank is empty. The wave-guides used in this experiment
were covered in a thin Teflon® tubing. This tubing allowed the wave-guides to be
placed in both acidic and basic solutions with no change in the signal.
Time Domain Reflectometer for Measuring Liquid … 169
Fig. 10 a Empirical height versus time difference, b actual height versus time difference
2 ⋅ Vr ⋅ ftx
FD = ð6Þ
c0
where FD is the Doppler frequency, vr is the radial velocity of the target, ftx is the
transmitter frequency and c0 is the propagation velocity of the signal. For most
instances, c0 is given as 2.99 × 108 m/s.
Rearranging the equation to determine the radial velocity of the target
F p ⋅ c0
vr = . ð7Þ
2 ⋅ ftx
Fig. 11 Schematic of a
Doppler radar measurement
technique for a liquid
Horn
measurement system
Antenna
Waste Water
172 S.R. Mamidi et al.
To improve the gain and directivity during signal transmission and reception, an
antenna needed to be designed to match the transceiver. To keep the cost down, the
materials used for the horn antenna had to be readily available. The material for the
horn antenna consists of circuit board sheet and copper sheet. The design of the
horn antenna was performed using Antenna Magus software. The results of the
simulation of the horn antenna is shown in Fig. 12. The antenna had a gain of
12 dB in the design and a center frequency of 23 GHz. This missed the target by a
GHz but was sufficient for this experiment. The S11 (or reflection coefficient) at
24 GHz was still −22 dB.
An antenna not only helps with gain but more important it improves directivity.
Directivity is the ability to radiate maximum possible amount of microwave energy
into a particular direction i.e., to the direction where we are measuring the solid or
liquid level. It is almost impossible to design an ideal antenna because no matter
how well the antenna may be designed a considerable amount of energy will still be
radiated into all directions. The goal is to design an efficient antenna. The output of
the modeling software is shown in Figs. 13 and 14 demonstrating the resulting gain
and directionality along with the reflection coefficient. Figure 15 shows the final
antenna construction. The antenna was machined in segments and assembled by
Time Domain Reflectometer for Measuring Liquid … 173
Fig. 13 Output of the Antenna Magus software for the horn antenna used in the test. This design
has good directionality
soldering the segments together using copper strips. A copper flange was also
machined and then the horn was soldered to the flange.
The observed signal in an oscilloscope was below the resolution of the oscil-
loscope. This required an operational amplifier circuit LM 741 to increase the
output voltage range of the Doppler unit to allow for better matching to the
oscilloscope.
To get the distance to the top of the septic wastewater it will be necessary to
manipulate the Doppler equations to get the range i.e. the distance to the top of the
liquid to the opening of the horn antenna. The equations can be further manipulated
to provide this distance or range to the target
174 S.R. Mamidi et al.
Fig. 15 Pyramidal horn antenna used to improve signal to noise ratio. This is a fairly large
antenna
c0 ⋅ t r
r= , ð8Þ
2
where tr is the phase delay of the returning pulse and is typically calculated and
displayed using microcontroller technology. This measurement assumes, however,
that the signals emanating from the Doppler transceiver are transmitted and
received perpendicular to the target. Otherwise, an angular correction must be
employed to account for increased distance and increased phase delay. The Doppler
system consisted of a Ma/Com MACS-007801 transceiver. A +18 dBi horn
antenna was fabricated using copper shim stock to focus the signal toward the liquid
column. The horn antenna and mounting over the water tank used in these initial
tests is shown along with the instrumentation. Subsequent Doppler signal amplitude
measurements were made, as shown in Fig. 13 and it was determined that for the
given range of liquid column heights the change in intensity was negligible and
would not be included in the instrument package.
This measurement assumes, however, that the signals emanating from the
Doppler transceiver are transmitted and received perpendicular to the target.
Otherwise, an angular correction must be employed to account for increased dis-
tance and increased phase delay. The Doppler system consisted of a Ma/Com
MACS-007801 transceiver. A +18 dBi horn antenna was fabricated using copper
shim stock to focus the signal toward the liquid column. The horn antenna used in
these initial tests is shown in Figs. 11 and 12 shows that antenna in the test fixture.
Subsequent Doppler signal amplitude measurements were made, as shown in
Fig. 13 and it was determined that for the given range of liquid column heights the
change in intensity was negligible and would not be included in the instrument
package (Figs. 16, 17).
Time Domain Reflectometer for Measuring Liquid … 175
Fig. 16 Measurement setup for radar system. Horn antenna is over the ¼ full water tank
12
11
f(x) = 0.09x + 8.49
Amplitude in Volts
10
9
Amplitude(V)
8
Linear (Amplitude(V))
6
0 5 10 15 20 25 30
Distance (cm)
Fig. 17 Results of the Doppler experiments showing the amplified signal versus the distance.
This signal was usable but have some issues with large outliers
6 Conclusions
The TDR technique has been developed for calibration of liquid level in the tank. It
important to remember that the material in a septic system is non-ideal. The TDR
probe used in this thesis is a bi-axial cable which is designed using two brass rods
encased in a Teflon® sheath. They were assembled in such a way that the char-
acteristic impedance was close enough to 50 Ω, which matches with the source
impedance. The transmission of the wave through the probe and its parameters were
studied. The TDR circuit was built with 74AC14 IC for signal generation. The
generated signal was transmitted through the bi-axial probe, where part of the
incident signal was reflected back to the source when the signal was incident to a
discontinuity. This TDR technique was tested in the laboratory, in a plastic grad-
uated cylinder from which a few conclusions were drawn. The TDR technique is
176 S.R. Mamidi et al.
simple and accurate for measurement of liquid level, and it is a technique that can
be used in situations that regular float-type or pressure based systems cannot. To
verify whether or not this was the most effective type of radar system, the results
were compared with a Doppler based radar. This comparison shows that the TDR
has significantly higher sensitivity than the Doppler. The installation and the use of
a TDR twin-axial probe also is much simpler than the antenna since it conceptually
could use a port on the side of the tank. Other types of probes could be constructed
an implemented allowing for a probe that had a lower chance of catching solids.
This probe was a first pass design and with some problems with impedance
matching though the entire system were observed. Future work will be on opti-
mizing the concept. Additional work will be on designing a measurement technique
that will reduce the need to use an oscilloscope to measure the time-based signal.
Modern high speed oscilloscope on a chip devices could also be used but most of
the devices that can be used to collect information like this are still quite expensive.
References
publication/222533231_TDR_application_for_automated_water_level_measurement_from_
Mariotte_reservoirs_in_tension_disc_infiltrometers
14. W.F. Kane, Embankment monitoring with time domain reflectometry, http://www.iti.
northwestern.edu/tdr/ publications/emwtdr.pdf
15. Time Domain Reflectometry. Hewlett hp Packard Application note 62, http://www.
hpmemoryproject.org/an/pdf/an_62.pdf
16. R.M. Heitsenrenther, Using microwave radar for long term remote sensing of ocean surface
dynamics (2009). Accessed 30 May 2015
17. A. Wolke, #88: Cheap and simple TDR using an oscilloscope and 74AC14 Schmitt Trigger
Inverter, https://www.youtube.com/watch?v=9cP6w2odGUc
18. P. Devine, Radar level measurement (Vega controls, West sussex, 2000). Accessed 30 May
2015
19. J. Fehrenbach, K. Griessbaum, Patent No. US 6,987,481 B2, United States of America (2006).
Accessed 29 May 2015
20. J.M. Fulford, L.W. Ester, J.W. Heaton, Accuracy of Radar Water Level Measurements
(2006), http://www.waterlog.com/media/pdfs/USGS-Study.pdf. Accessed 30 May 2015, from
Waterlog
Nanowire (S3) Device for the Quality
Control of Drinking Water
Abstract The control and sanitation of water it’s a worldwide concerning problem.
In particular is in developing countries where the need is more evident due to the
lack of sources and appropriate structures to cope with the outbreaks of waterborne
diseases. In most of cases, the cause of these significant epidemiological events
have microbial origin. Despite of it the situation of water supplies and sanitation has
improve all over the world in the last decades. Nanowire technology has already
shown their ability to perform very effective and fast monitoring microbiological
spoilage and quality control. The aim of this study was to test the ability of a novel
S3 (Small Sensor System) nanowire device for the detection of complex mixtures of
bacteria in potable water in order to approach into a real condition, in cooperation
with GC-MS- SPME technique. The achieved results notably advocate the use of
EN as a very easy to use, fast and accurate tool in water quality control.
1 Introduction
Water is one of the most basic chemical compounds, on Earth, it occupies the 71 %
of the total surface area. Even though it provides no calories or organic nutrients is
essential to humans and any kind of life. The amount of fresh water on earth is
limited and its quality is under constant threat. Preserving the quality of fresh water
is important for the drinking-water supply and food production (Table 1).
Nowadays in the developed continents such as Europe and North America
access to drinking water is easy, directly from aquifers, and at a low cost. This leads
to a great waste of this resource, to which there is no longer the proper social and
Conversely 700 million people still lack a ready access to improved sources of
drinking water and some 2.5 billion people do not use an improved sanitation
facility (Figs. 1 and 2).
About 2.6 billion people—half the developing world—lack even a simple ‘im-
proved’ latrine and 1.1 billion people has no access to any type of improved
drinking source of water. As a direct consequence [2]:
182 E. Núñez Carmona et al.
• 1.6 million people die every year from diarrhoeal diseases (including cholera)
attributable to lack of access to safe drinking water and basic sanitation and
90 % of these are children under 5, mostly in developing countries;
• 160 million people are infected with schistosomiasis causing tens of thousands
of deaths yearly; 500 million people are at risk of trachoma from which 146
million are threatened by blindness and 6 million are visually impaired;
• intestinal helminths (ascariasis, trichuriasis and hookworm infection) are
plaguing the developing world due to inadequate drinking water, sanitation and
hygiene with 133 million suffering from high intensity intestinal helminths
infections; there are around 1.5 million cases of clinical hepatitis every year.
Between 1990 and 2014 the drinking water coverage is growing up from 76 to
89 % (Figs. 3 and 4).
Regarding all continents’ situation for the drinking water coverage is possible to
remark in particular:
• Europe. In Europe the drinking water coverage is optimal; the range is from 91
to 100 %;
• North America, South America, Asia and Oceania. The North America, South
America, Asia and Oceania situation is similar to the Europe one; the range is
from 91 to 100 %;
• Sub-Saharan Africa. In a further 35 countries, 26 of which are in sub-Saharan
Africa, coverage of improved drinking water supply was between 50 and 75 %.
Access to drinking water from an improved source is significantly higher in
urban than in rural areas. In rural areas, in virtually the entire developing world,
drinking water coverage from an improved source remains unacceptably low.
Fig. 3 World population with and without access to an improved drinking water source in 1990,
2004 and 2015
Nanowire (S3) Device for the Quality Control of Drinking Water 183
Fig. 4 World population with and without access to improved sanitation in 1990, 2004 and 2015
Urban drinking water coverage remained the same from 1990 to 2004 at 95 %,
whereas in rural areas coverage increased to 73 % in 2004 from 64 % in 1990. In 27
developing countries, less than 50 % of the rural population have access to
improved drinking water.
3 Sensors Background
One of the types of sensors most frequently used is the metal oxide (MOX or
MOS). The working principle of this technology is that in the presence of a vari-
ation in gas concentration ΔCgas produce a change of electrical conductivity Δσ,
and, therefore, a change of resistance ΔR, for which the measured parameter in this
case is the sensor resistance, R. In other words the principle on which is based the
MOS consists in a variation of the conductivity of the metal oxide in the presence of
volatile compounds compared to the value assumed by the same conductivity in
reference conditions.
The characteristics that should possess an ideal sensor are the following:
• High sensitivity to chemicals;
• High selectivity (low noise) and stability;
• Low sensitivity to humidity and temperature;
• High reproducibility and reliability;
• Reaction time and return short;
• Strength, durability, ease of calibration;
• Small size.
184 E. Núñez Carmona et al.
4 Nanowire Sensors
Metal oxides in the form of nanowires are interesting for their peculiar morphology
and their exceptional crystalline features, the first assuring a high surface to volume
ratio necessary to maximize surface related properties such as the ones governing
chemical sensing transduction principles, while the latter guarantees stable crys-
tallinity and therefore electrical properties over long-term operation, i.e., a required
quality for an industrial application of any kind of sensor in real environments.
Different experimental techniques that may lead to the formation of these quasi
one-dimensional structures. First of all there are two main groups that can be
distinguished: top down and bottom up technologies. The first starts from bulk
structures, reducing them to nanometric dimensions usually with lithographic
techniques, while the latter involve a direct assembly in the desired morphology that
can be obtained from vapor or liquid phase.
At the beginning the research was focused on the vapour phase methods that
were producing, with cheap instrumentation, high quality nanostructures in terms of
crystallinity and stoichiometry. We have thoroughly studied the deposition using
evaporation and condensation from powder in controlled environments using a
different experimental set up.
Tin oxide was preferred over other oxides thanks to its well known chemical
sensing properties and the easy preparation conditions, but indium and zinc oxide
were also studied [6–11].
The experimental procedure consists in the evaporation of the powder (metal or
metal oxide) at high temperatures in a controlled atmosphere at pressures lower than
hundreds of millibar and the following mass transport of the vapour towards the
substrates kept at lower temperatures with respect to the source evaporation region.
Argon, an inert gas, is used for the mass transport in order to avoid unwanted
reactions with the oxide vapour. Critical variables are the distance between the
source and the substrates and the temperature gradient that are controlling
the supersaturation conditions in the vapour phase causing the condensation in the
desired morphology (Fig. 5).
186 E. Núñez Carmona et al.
Fig. 5 Scheme of a Nanowire sensor, from the top to the bottom: Pt electrodes (grey colour), TiW
Pads (brown colour), ZnO Nanowires, Alumina substrate (white colour), Pt heating element (grey
colour)
5 Device Design
The S3 (Gas Sensor System srl—Spin-Off University of Brescia, Italy) used during
the experimental analysis is constituted of: Auto-sampler headspace system
HT280T (HTA srl, Brescia, Italy). The auto-sampler supports a 40 loading sites
carousel and a shaking oven, to provide the equilibrium of the samples head space.
The sensor’s chamber is a thermally controlled chamber with 9 loading sensors
places (Fig. 6).
The sensor array installed in the used instrument is described in Table 2. The
study has been carried out using an hybrid array composed by 3 Nanowires and 3
RGTO.
The advantage of having an hybrids array enhance the performance of the
instrument, that possess the capacity to detect a widest range of volatile chemical
compounds.
Nanowire (S3) Device for the Quality Control of Drinking Water 187
6 Aim
The study and introduction of innovative technologies, that make possible the
access to drinking water and sanitation, are fundamental for the poor countries’
growth and for the welfare maintenance in the developed countries. The wastewater
purification process in the developing countries is the only possible way to make
contaminated water drinkable and, at the same time, it allows to save water in the
aquifers and to increase the availability of drinking water and also to improve the
sanitation conditions.
The improvement of the water quality control systems in both developing and
developed countries is one of the better approaches to reach this objective. This
improvement will not have just an impact in the enhancement of the water sources
for agriculture and industry (incomes and outcomes), but the most important will
have a big impact on the final consumer.
188 E. Núñez Carmona et al.
Currently used methods require high tech lab equipment, high level trained lab
staff, a big investment of money and which is more important requires time. The
aim of this work was to establish a new, rapid and economic fouling, based on the
cooperation between the use of the new S3 device, nanowire tech, and chemical
analytical techniques to detect the presence of complex microbial contaminant sets
in water.
This is possible by using the innovative e-sensing device: Small Sensors System
(S3. S3 device equipped with nanowire tech is based in the analysis of the head
space. The Nanowires are gas sensors based on semiconducting metal oxide are a
promising technology thanks to their high sensitivity to a broad range of chemicals,
reduces size, weight, and low power consumption. These entire characteristic
makes S3 device equipped with nanowire sensor an optimum portable device for
the attainment of on field test. The optimization of the beginning wastewater
purification process and the management of potable water is obtained through the
identification of volatile metabolites (VMs) and volatile organic compounds
(VOCs) released from the microorganisms’ metabolism, that typically contaminate
water [12–21].
Samples of tap water were sterilized and inoculated with varying rates of the
reference microorganisms, to verify the threshold of the instrument to detect the
presence of bacteria in an environment as close as possible to the real one. The
samples were subjected to conventional microbiological methods aimed at assess-
ing the magnitude of microbial growth, such as measurement of pH, turbidity and
seeding on selective media. The results obtained were then compared with those
resulting from the analysis conducted by means of electronic nose. This comparison
enabled to investigate the ability of the instrument to perceive precociously the
presence of volatile organic compounds resulting from the development of the
bacterial load of the substrate to different microbial loads of departure. These tests
were conducted at first individually considering the different bacteria under study;
later were analyzed bacterial mixtures simulating a real case of multiple contami-
nations. Finally, the samples associated with better outcomes have been subjected
to analysis by GC-MS-SPME technique in order to determine the chemical nature
of the organic compounds most frequently detected from the microbial mixtures
examined.
Nanowire (S3) Device for the Quality Control of Drinking Water 189
The vials were placed in a randomized mode into the HT280T carousel. Each vial
was incubated at 40 °C for 10 min into the HT280T oven, by shaking it during all
the incubation in order to reach the equilibrium of the headspace. The sample
headspace (2 ml) was then extracted from the vial in static headspace path and
injected into the carrier flow (speed 4 ml/min) through a properly modified gas
chromatography injector (kept at 40 °C to prevent any condensation). Using syn-
thetic chromatographic air with a continuous flow rate of 10 ml/min performed the
sensor baseline and the recovery time was 28 min [23]. The sensor baseline was
performed by using synthetic chromatographic air with a continuous flow rate of
10 ml/min and the recovery time was 28 min. In every carousel 4 of the 40 loading
positions were dedicated to vials containing 500 µl of butanol. This compound act
as the internal standard of the instrument.
7.1.3 pH
A control of the pH was done for all the three dilutions. The pH was measured each
hour during the first 24 h of analysis. Whereas During their grow the E. coli pro-
duce organic acids that will acidify the pH of the water.
190 E. Núñez Carmona et al.
7.1.4 Photometer
The used microorganisms in this work are Escherichia coli, Salmonella Typhy-
murium, Listeria monocytogenes. Cultures were developed in 7 ml of Brain Heart
Infusion (BHI) (OXOID) liquid media [22]. The cultures were incubated at 30 °C
during 48 h. Subsequently the cultures were centrifuged at 3000 rpm during 5 min,
the supernatant was discarded and the cells were re-suspended in 3 ml of physio-
logical solution. A second step of centrifugation at the same conditions was per-
formed in order to clean any trace of media from the suspension. At the end of the
process the supernatant was discarded and the pellet was suspended in 4 ml of
sterilized tap water. The sterilization process was carried out at 121 °C during
15 min at one over pressure atmosphere in order to eliminate all the microorgan-
isms present in the tape water.
Once culture had been cleaned, the turbidity has been adjusted adding sterilized
tape water until it reached the same turbidity as the third standard of the McFarland
series. These standards are used as a reference to adjust the turbidity of bacterial
suspensions in order to have a number of bacteria within a given range. Number 3
of McFarland standard corresponds with a bacterial concentration of 9 × 108
CFU/ml. These microbial solutions were subsequently used to prepare the bacterial
combinations as it’s indicated in Table 3 to inoculate the samples with a total
concentration of 9 × 102 CFU/ml.
A total volume of 400 ml of sterilized tap water for each mix of bacteria was
inoculated with the suspension of cells prepared before until it reaches 9 × 102
CFU/ml. Once prepared all three dilutions were incubated in the dark at 30 °C
during the 4 days of analysis (T0, T2, T3 and T7).
A control sample was also performed using sterilized tap water. In addition, in
this case the parameters of incubation were the same followed by the inoculated
samples.
The same procedure as the epigraph 7.1.2 was used for the analysis of samples and
data.
8.1.1 pH and OD
pH
Series1
7.10
7.00 Linear (Series1)
6.90
6.80
6.70
0.550 0.600 0.650 0.700
O. D. (600 nm)
As the microorganisms develop inside the water, the OD increases because the
turbidity increases as well. The metabolic activities carried out by the microor-
ganisms during their grow causes the production of a big amount of organic acids
that subsequently decrease the pH values. These two variables are inversely related
so a test of correlation between pH and OD was performed for the 3 dilutions. In the
three cases, the negative correlation between the 2 variables were confirmed
obtaining a R2 > 0.98 (Fig. 7).
In the two following score plot (Figs. 8 and 9) the control samples are compared,
separately, with samples contaminated with E. coli and P. fluorescens in the days
T0 and T3. All the concentrations studied were considered in the statistical analysis.
For both PCA is possible to observe three clusters of dots clearly separated from
each other, the first of which (at the top left—black) regarding the samples inoc-
ulated with the microbial species of reference, the second covering not contami-
nated samples (blue) and the third (right, upper extremity—green) relative to the
internal standard (pure butanol, 500 μL).
In particular in the second chart (Fig. 9) can be observed a progressive trend line
of the statistical objects included in the cluster of E. coli against the cluster of
analytical blank. This result is in agreement with the nature of the samples subject
to analysis, which are referable to all microbial concentrations studied. In fact the
tail observed is due to samples with a lower microbial load, which will be asso-
ciated with a lower odorous intensity, arranging itself in proximity of the
non-inoculated samples. The experimental investigations were prolonged in order
to determine whether the olfactory system was able to distinguish different
microbial agents of contamination, different microbial loads and/or the different
days of experimentation. In general the instrument has proved to be able to carry
out, at fixed concentration, a proper distinction of individual species concerned over
the several days of analysis, remarkable results were obtained for E. coli at very low
concentrations (Fig. 10).
Nanowire (S3) Device for the Quality Control of Drinking Water 193
Fig. 8 Score plot PCA samples of water contaminated with P. fluorescens (all concentrations, T0)
compared with samples not contaminated and those of BuOH (internal standard)
Fig. 9 Score plot PCA samples of water contaminated with E. coli (all concentrations, T3)
compared with samples not contaminated and those of BuOH (internal standard)
From the chart above (Fig. 10) its possible to observe three homogeneous
groups of statistical objects, among them adequately separated along the first
principal component (PC1), associated with E. coli at a concentration of 9 × 102
CFU/ml. The PCA demonstrates the ability of the instrument to properly dis-
criminate the microbial presence in water for the days of experimentation T0, T1
and T3. The change in the flavor profile detected in the course of time by the
electronic nose could be due to a gradual accumulation of volatile secondary
metabolites. It is possible to relate that response to a release in the surrounding
medium of odorous compounds, which in turn can be associated with phenomena
194 E. Núñez Carmona et al.
Fig. 10 PCA Score plot: comparison of water samples contaminated with E. coli set up at the
time T0, T1 and T3
of cell lysis, frequently observed in advanced stages rather than in the more pre-
cocious. In the following graph (Fig. 11) are compared the results obtained E. coli
and S. Typhymurium (T0 and T4).
The concentration considered is 9 × 108 CFU/ml. The graph shows some
distinction of the cluster relative to E. coli from those of S. Typhymurium most
central with respect to the first, which on the contrary undergo a clear separation
along the first principal component as a function of the different days of experi-
mentation, T0 and T4. The identification of four separate groups demonstrates the
potential of the instrument to discriminate between different microorganisms and
Fig. 11 PCA Score plot of comparison of water samples contaminated with E. coli and samples
contaminated with S. Typhymurium. Days considered: T0 and T4; concentration: 9 × 108 CFU/ml
Nanowire (S3) Device for the Quality Control of Drinking Water 195
Fig. 12 PCA Score plot of a comparison of water samples contaminated with E. coli and samples
contaminated with L. monocytogenes. Concentrations considered: 9 × 102 CFU/ml and 9 × 105
CFU/ml; day: T1
for the same microorganism, among the different days of analysis considered. The
greater separation of the clusters may indicate a greater metabolic reactivity of
E. coli than S. Typhymurium as well as a greater ability to adapt and proliferation
within the medium water, generally poor in nutrients. In the next figure is showed
(Fig. 12) the scores for the water samples inoculated with E. coli and L. monocy-
togenes at the time T1 compared with the different concentrations 102 CFU/ml and
105 CFU/ml.
Although in the figure is not possible to make a proper distinction between the
different microbial loads is still evident discrimination of the various species con-
sidered in two clusters clearly separated along the first principal component (PC1).
In general the lower concentrations have proved to be those associated to the
best instrumental performance in terms of the distinction of the microbial kind than
the different days of analysis. This could be due to the by saturating effect of the
sensor with the higher microbial loads, in absolute terms, greater emissions of
volatile compounds could lead to a reduction in the resolving power of the
instrument exceeding a certain concentration threshold value.
Fig. 13 PCA score plot of water samples contaminated with E. coli, L. monocytogenes e
S. Typhymurium—EC + LM + ST (T3/T7) compared with control samples and BuOH
Fig. 14 PCA score plot, comparison of water samples on 3 different microbial mixtures:
EC + LM + ST, EC + LM, EC + ST. Day: T2
Nanowire (S3) Device for the Quality Control of Drinking Water 197
Fig. 15 PCA score plot showing T0, T2, T3, T7 for the microbial mix EC + ST
samples obtained on different days (T0, T2, T3, T7). The best results were obtained
for the mixture of E. coli and S. Typhymurium (EC + ST) (Fig. 14).
In the chart below (Fig. 15) is shown the distribution of clusters for the microbial
mixture EC + ST within the information space identified by the principal com-
ponents PC1, PC2 and PC3 for everyday trials of analysis considered.
From the graph con be observed the correct distinction of the samples prepared
in the days of analysis as demonstrated by the identification of four separate
clusters.
Fig. 16 PCA score plot comparison between the samples of T0, T2, T3 for the microbial mix
EC + LM + ST
198 E. Núñez Carmona et al.
Fig. 17 Score plot of PCA: Comparison between the samples during the analysis time of T0, T2,
T3, T7 for the microbial mix EC + LM
Parallel results to that just described above have also been obtained for mixtures
microbial.
EC + LM + ST of which are reported the relative score plot (Figs. 16 and 17).
Table 4 Main VOCs detected by each microbial mixture by GC-MS; Days: T0, T2, T3, T4
Microbial VOCs
blends
EC + LM + ST Indole, 1-pentanolo, 1-nonantiol, Nonanal, 1-dodecen-3-ol, 1-dodecanol,
3-metilbutanol, n-pentadecanol
EC + ST Indole, Benzaldeide, 3-metilbutanol, Nonanal, Decanal, Benzoic acid,
Butirric acid, 2-etilesanol, n-amilisovalerate, Acetilmetilcarbinol,
1-pentanol, 1-dodecanol, 1-tetradecanol
EC + LM Indole, Benzaldeide, Butanoic acid, Decanal, 1-pentanol, 1-ottantiol,
Nonanale, 3-metilbutanol
LM + ST Benzaldeide, Eptanal, 1-tridecen, n-tridecanol, 3-metilbutanol,
3-metilbutilformiate
The compound ever more frequently isolated from mixtures containing E. coli is
indole. Indole is a heterocyclic compound of empirical formula C8H7N. At room
temperature is presented as a white solid colorless with characteristic aroma. It is in
fact one of the compounds that give the typical smell of the fecal matter along with
several other indole-derivatives such as skatole. It can be found in low percentages
in combination with other substances such as aromatic component of many foods
such as Camembert, Swiss cheese, coconut, black and green tea, rum and toasted
hazelnuts [24]. Various microorganisms (over 85 different species among
Gram-positive and Gram-negative) secrete the indole as a by-product of protein
degradation. It constitutes one of the characteristic metabolites of E. coli, as such
frequently used as a marker for the identification/detection of the bacterium in
different culture media (indole test).
In the following histogram (Fig. 18) is shown the trend of indole emission for
each one of the mixtures studied. The maximum abundance of indole is recorded,
for all mixtures, at T0 and then gradually fell down. This phenomenon is
60000
50000
40000
30000
20000
10000
0
T0 T1 T3 T7
EC+LM+ST 49892 9785 1200 0
EC+LM 60120 45201 21954 0
EC+ST 49852 32154 22182 0
LM+ST 20198 613 400 0
TIME
Fig. 18 Yield of indole emission for each of the microbial mixtures studied throughout the week
of testing
200 E. Núñez Carmona et al.
80000
Indole 3-Methylbutanole Decanale Benzaldeyde
70000
60000
ABUNDANCE
50000
40000
30000
20000
10000
0
T0 T1 T3 T7
Indole 45892 20785 12000 0
3-Methylbutanole 0 60120 50954 0
Decanale 0 76154 19182 0
Benzaldeyde 0 0 TIME 14000 40198
Fig. 19 Yield of the emission of indole, 3-methyl butanol, decanal and benzaldehyde throughout
the week of testing. Mixtures of microbial reference: EC + ST (indole and 3-metilbutanol),
EC + LM (decanal) and LM + ST (benzaldehyde)
attributable to the microbial transfer from a rich culture media (BHI) to the water a
poorer growth medium (water) and therefore less able to support the cell metabo-
lism. It is conceivable that the biosynthetic capacity of the bacteria decreases as a
function of permanence within the system in response to a gradual decreasing of
nutrients with the consequent increase of the cell mortality. The data collected from
the mix EC + ST and EC + LM are examples of the above explanation.
The following graph (Fig. 19) shows the abundance trend versus time of indole
and three other selected VOCs, 3-metilbutanol, decanal and benzaldehyde, con-
sidered of major importance for the various mixtures considered.
In Fig. 19 can be observed 3-metilbutanol and decanal with a similar trend to
indole with a progressive reduction in the synthesis up to complete absence in last
day of analysis.
Whatever distinguishes these compounds from indole is the absence of pro-
duction in the first day of experimentation that might suggest a late synthesis in the
course of bacterial growth. It is possibly facilitated by phenomena of synergy or
competition among different microbial species once these are arranged in the same
medium of growth. An opposite trend with respect to that described above is
recorded for the benzaldehyde whose synthesis is until the time T2 to increase and
become maximum at the time T7.
Nanowire (S3) Device for the Quality Control of Drinking Water 201
9 Conclusions
During this investigation was possible to confirm the potential application of the S3
device to perform earliest and on-situ analysis of the water microbiological quality.
The results achieved demonstrate a capacity to discriminate reliably between
contaminated samples with samples of sterile water, so much as regards the indi-
vidual microorganisms as for mixtures microbial.
The most significant performance was obtained distinguishing different days of
testing (associated at different stages of microbial growth). The artificial olfactory
system (S3) has shown a remarkable versatility in finely distinguish the samples of
lower bacterial concentrations (102 CFU/ml) from samples of sterile water.
In general, the most promising responses were observed in the analysis of the
microbial samples with a concentration of 102 CFU/ml. This is due to a possible
saturation effect exerted on the sensors from bacterial higher concentrations, such as
those associated with a higher emission of volatile compounds.
Predictably mixing the bacteria under study was accompanied by a general
reduction in the ability of the device to distinguish easily the samples that share the
same microorganism. However, it was possible to correctly discriminate and
reconfirm for a given mixture, the different stages of the culture development.
It is necessary to pursue a process of continuous technological improvement
through the design and implementation of sensors, more sensitive and possibly
specific to selected VOCs they appropriate for use as a marker of contamination.
Extending the investigations for the identification and quantification of the
organic substances of microbial origin in contexts gradually more and more com-
plex would allow increasing the knowledge regarding any synergy/antagonism
between microorganisms colonizing the same ecological niche.
The future perspective will be to realize a device, possibly portable, that will be
able to provide, with high degree of sensitivity, a reliable answer in real time about
the hygienic status of drinking water by the simple analysis of odor. The artificial
olfactory system may be used as a screening tool for the distinction of water
samples subject to risk contamination and then proceed accordingly to microbio-
logical investigation.
This kind of approach will offer a considerable savings in terms of time and
money. It will allow sustained control of water sources indicating in situations of
real risk when will be important to perform traditional microbiology and chemical
investigation (longer and more expensive) [24, 25].
References
2. Meeting the MDG drinking water and sanitation target: the urban and rural challenge of the
decade. I. WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation,
WHO Library Cataloguing-in-Publication Data. World Health Organization, UNICEF (2015).
ISBN: 92 4 156325 7
3. N. Barsan, U. Weimar, Conduction model of metal oxide gas sensors. J. Electroceram. 7,
143–167 (2001)
4. P.T. Moseley, J.O.W. Norris, D.E. Williams, Techniques and Mechanism in Gas Sensing
(Adam Hilger, Bristol, UK, 1991)
5. S.R. Morrison, The Chemical Physics of Surfaces, 2nd edn. (Plenum Press, New York, NY,
USA, 1977)
6. S. Bianchi, E. Comini, M. Ferroni, G. Faglia, A. Vomiero, G. Sberveglieri, Indium oxide
quasi-monodimensional low temperature gas sensor. Sensors Actuators B Chem. 118, 204–
207 (2006)
7. G. Sberveglieri, C. Baratto, E. Comini, G. Faglia, M. Ferroni, A. Ponzoni, A. Vomiero,
Synthesis and characterization of semiconducting nanowires for gas sensing. Sensors
Actuators B Chem. 121, 208–213 (2007)
8. E. Comini, G. Faglia, M. Ferroni, G. Sberveglieri, Gas sensing properties of zinc oxide
nanostructures prepared by thermal evaporation. Appl. Phys. Mater. Sci. Process 88, 45–48
(2007)
9. E. Comini, S. Bianchi, G. Faglia, M. Ferroni, A. Vomiero, G. Sberveglieri, Functional
nanowires of tin oxide. Appl. Phys. Mater. Sci. Process 89, 73–76 (2007)
10. E. Comini, C. Baratto, G. Faglia, M. Ferroni, G. Sberveglieri, Single crystal ZnO nanowires
as optical and conductometric chemical sensor. Appl. Phys. Mater. Sci. Process 40, 7255–
7259 (2007)
11. E. Comini, G. Faglia, M. Ferroni, A. Ponzoni, A. Vomiero, G. Sberveglieri, Metal oxide
nanowires: preparation and application in gas sensing. J. Mol. Catal. A Chem. 305, 170–177
(2009)
12. E. Núñez Carmona, V. Sberveglieri, A. Pulvirenti, Detection of microorganisms in water and
different food matrix by electronic nose, in Proceedings of the 2013 Seventh International
Conference on Sensing Technology, Wellington, NZ, pp. 703–707, Dec 2013
13. V. Sberveglieri, E. Comini, D. Zappa, E. Núñez Carmona, A. Pulvirenti, Electronic nose tor
the early detection of different types of indigenous mold contamination in green coffee, in
Proceedings on the Seventh International Conference on Sensing Technology, Wellington,
NZ, pp. 465–469, Dec 2013
14. V. Sberveglieri, E. Núñez Carmona, E. Comini, A. Ponzoni, D. Zappa, O. Pirrotta, A.
Pulvirenti, A novel electronic nose as adaptable device to judge microbiological quality and
safety in foodstuff. BioMed Res. Int. 2014, 1–6 (2014)
15. E. Núñez Carmona, V. Sberveglieri, A. Ponzoni, D. Zappa, A. Pulvirenti, Detection of
microbial contamination in potable water by Nanowire technology, in Proceedings of the 8th
International Conference on Sensing Technology, Liverpool, UK, pp. 136–139, Sept 2014
16. V. Sberveglieri, A. Pulvirenti, E. Comini, E. Núñez Carmona, What happens at the aroma of
coffee beans after roasting? MOX nanowire technology by Novel Electronic Nose to discover
the finger print, in Proceedings of the 8th International Conference on Sensing Technology,
Liverpool, UK, pp. 136–139, Sept 2014
17. E. Núñez Carmona, V. Sberveglieri, E. Comini, D. Zappa, A. Pulvirenti, Nanowire
technology for the detection of microorganism in potable water. Proc. Eng. 1453–1456
(2014). ISSN: 1877-7058
18. V. Sberveglieri, E. Núñez Carmona, D. Zappa, E. Comini, A. Pulvirenti, Classification of
different roasting processes by MOX nanowire. Proc. Eng. 572–575 (2014). ISSN: 1877-7058
19. V. Sberveglieri, M. Falasconi, E. Gobbi, E. Núñez Carmona, G. Zambotti, A. Pulvirenti,
Candida milleri detected by electronic nose in tomato sauce. Proc. Eng. 584–587 (2014).
ISSN: 1877-7058
Nanowire (S3) Device for the Quality Control of Drinking Water 203
Abstract This chapter presents a novel approach to monitor the quality of milk
products, based on electromagnetic wave spectroscopy. A comparative analysis is
made to demonstrate the effectiveness of using microwave sensors over the other
types, existing in the wide field of sensing technology. Three broadly used com-
mercial varieties of milk, namely skimmed, semi-skimmed, and whole milk types
are considered for the test measurements. The overall quality parameters of these
products obtained from the market are comparatively measured in terms of their
composition and spoilage with reference to ageing. The experiments carried out
have illustrated that the sensor was able to distinguish one milk type from another.
Moreover, it was also able to differentiate between fresh and aged milk samples of a
given milk type as the number of days passes. The methodology used here employs
Vector Network Analyser to capture spectral signatures in the form of scattering
parameters from electromagnetic wave sensors. These data are then analysed to
evaluate quality monitoring process achieved by these sensors. This work offers a
potential platform for an economical, less complicated, and real-time milk quality
control mechanism that can be employed outside of the laboratories at medium or
large scale retailers in milk supply chain hierarchy.
1 Introduction
Milk, along with other dairy products, is consumed by over 6 billion people around
the world [1]. This shows the importance of milk and milk products, as an integral
part in daily life, especially for those who follow a vegetarian diet, inevitably due to
their nutritional values. Milk supplies nutrients like good quality proteins, fat, car-
bohydrates, vitamins and minerals in significant amount than any other single food
[2]. Dairy industries, that procure milk from dairy farms to process the raw milk in
order to produce pasteurised milk, cheese, butter, ice-cream etc., also play a very
significant part in overall economy of many developing as well as developed
countries. The interest in milk products is continuously growing around the globe
over the years because of several factors like changing dietary habits, improving
nutritional awareness, rising population and overall development in the economy [3].
This growth can be observed from the fact that the milk production, worldwide, has
increased by over 50 %, from 482 million tons as in 1982 to 754 million tons as in
2012, in 30 years, which is projected to increase by 175 million tons (23 %) by 2024
with comparison to that of the base years (2012–2014) [4, 5]. Milk processing at
industrial level, following milk production at dairy farms, is next important stage in
the supply chain of all dairy products [6]. Milk, because of being highly nutritious
food, offers a very good environment for the microorganisms called bacterial
pathogens to develop, breed and grow, eventually causing spoilage of milk. This
phenomenon leads to short shelf-life of milk itself as well as other dairy products
made by processing the milk. Besides shortening the overall edible life span, bacteria
adulteration of any milk products can also cause consumers to contract various food
borne diseases when such affected products are consumed. This can be eliminated to
an extent by milk processing which enables its preservation over longer time-period
and also helps to avoid the problems pertaining to health deterioration because of the
lack in its quality or due to spoilage. Processing of raw milk ensures that its original
nutritional value is maintained which also allows its preservation. Milk with
improved overall natural quality is more preferred over the milk products that have
artificially added nutritious values by processing. The increasing intake of dairy
products provides vital health benefits to a large population of the globe especially
that of the developing countries, in spite of the fact that millions of people in these
nations still cannot afford better quality diets which are expensive [7]. These high
costs of production are not only because of the high value raw milk but its subsequent
stage of milk processing also adds to overall retail value of the end product.
Various milk quality testing techniques are available in industries and are being
developed within research laboratories. The tests focusing on adulteration of milk
are predominant in the industries, besides certain general tests applied to check
overall quality of milk; like phosphatase test used on pasteurised milk and the
acidity development test done on Ultra High Temperature (U.H.T.) processed milk
[3]. There is also growing interest in techniques that are specifically aimed at
determining the contamination within milk, such as Detergent Residue Testing [8].
Microwave sensors are used for a wide variety of applications including mea-
surement of distance, movement, shape, and particle size, but the largest group of
applications is concerned with the measurement of material properties [9]. Novel
applications for these types of sensor have been introduced in food and water
quality monitoring—for meat products [10], vegetable oils [11], and analysis of
water contaminants [12] in the past. The other domain of application for these
sensors is health monitoring where various applications have been introduced
which includes the real-time indication of neurological impairments during surg-
eries [13, 14], and in detection of specific biomarkers in human cerebrospinal fluid
[15]. The fundamental principle of EM wave sensors that operate at microwave and
Milk Quality Monitoring Using Electromagnetic Wave Sensors 207
radio frequencies (RF) is based on the interaction of these signals with the material
under the test. The object under test changes the actual velocity of the electro-
magnetic wave signal, by either attenuating or reflecting it. The advantage of
microwave sensors is that they can be implemented cheaply for a wide range of
applications in a non-destructive yet effective manner and are capable of measuring
without physical contact over a short distance, with the help of penetrating waves
without creating any health hazards [16].
This chapter reports on a novel measuring technique that assesses overall quality
of the milk by employing EM wave spectroscopy approach. Bespoke EM sensors
that operate at microwave frequencies were developed and tested. The quality
assessment is done in terms of classification of the milk type and detection of
spoilage or bacterial adulteration developing over the days within the milk samples.
The three categories of milk majorly sold in the market; namely whole milk,
semi-skimmed milk and skimmed milk were tested. The sensor developed for this
purpose was a resonator cavity designed to radiate within microwave range of
frequencies between few hundred MHz to few GHz. The full laboratory setup
makes the use of Vector Network Analyser (VNA) connected with the EM wave
cavity to record the spectral signatures. The results of these test experiments, as
explained in the following sections, are quite promising for further design and
development of an all-in-one system. This innovative system for quality monitoring
of milk products using EM wave spectroscopy is real-time, with less complexity
and bearable costs, in comparison with many other existing methods currently in
practise. This does not only classify one type of milk from another but also dis-
tinguishes between fresh and aged milk samples under test conditions.
Milk is a perishable product, hence, spoils very quickly. Its nature of having lower
acidity and higher nutrients make it the perfect breeding platform for bacterial
micro-organisms, including those which are responsible for food poisoning, known
as pathogens. Bacteria contaminate the milk, and their destruction is the main
reason for milk processing which enables milk preservation. This can be achieved
by fermentation, heating, cooling, and removal of water or concentration or sepa-
ration of various components from milk, to produce foods like butter and cheese
[17]. Table 1 lists the nutritional content of milk types used for making test
samples.
Milk quality control is the use of approved test methods to ensure standards
concerning the milk and milk products are regulated to accepted levels. This takes
the account of composition and purity as well as the level of different
micro-organisms within milk. It includes testing milk and milk products for quality
and monitoring that milk products, processors and marketing agencies adhere to
208 K.H. Joshi et al.
accepted codes of practices [3]. The milk producer expects a fair price in accor-
dance with the quality of milk that they produce. The milk processor pays the
producer and assures that the milk received for processing is of normal composition
and is suitable for processing into various dairy products. The consumer expects to
pay a fair price for milk and various other milk products of acceptable to excellent
quality.
To design a methodology and for developing a system that can test the overall
quality of milk products, it is required to understand the basics of the milk cate-
gories under consideration. Here, the three commercially predominant categories of
milk in consumer market—whole milk, skimmed milk and semi-skimmed milk are
explained along with their spoilage patterns. The Dairy Council, UK, classifies the
three types based on the fat content that they carry. Skimmed milk should not have
fat in excess of 0.3 g per 100 ml whereas whole milk is expected to have a min-
imum of 3.5 g per 100 ml, while, the fat value in semi-skimmed milk ranges
between 1.5 and 1.8 g of the 100 ml of product [18]. Several parameters for the
spoilage of milk were studied and discussed including bacterial growth for both the
milk types, and the observations showed that the bacterial growth patterns in the
skimmed milk did not significantly vary from those in the whole milk. In addition
to that, it was also confirmed that the milk spoilage bacteria grow at similar rates, in
both, skimmed and whole milk types but have different metabolic behaviours
within the two of them [19]. This also supports the results achieved in this work and
the results of other researchers who observed similar rates of growth in skimmed
and whole types [20–22]. These findings lead us to a common conclusion that the
different shelf-lives of skimmed and whole milk types are not characterized by
different bacterial growth rates within them.
In the past two decades, the evolvement in high-speed instrumental testing
techniques to measure milk and dairy product composition provided the required
platform for improvement in overall effectiveness and accuracy of classical
chemical analysis methods [23]. For example, the development of Gas Chro-
matography Olfactrometry (GCO), in combination with mass spectrometry is the
development of analytical tools that can be correlated with aroma/odour perception
of foods. GCO basically detects aroma compounds developed due to degradation of
Milk Quality Monitoring Using Electromagnetic Wave Sensors 209
proteins and amino acids. In improving the milk preservation techniques, spoilage
detection plays a vital role.
Milk spoilage can be defined in many ways and its measure could be different each
time depending upon the context it is discussed in. For example, the simplest defi-
nition of quality product, for a common buyer it would be, a product that tastes good
and smells good [24]. Many times we may find that the milk which is expired as per
its ‘Use By’ date, labelled on its packaging, is not actually outdated and sometimes it
is the opposite way around. This vagueness troubles both, the consumers as well as
the manufacturers. Buyers are mostly reluctant to purchase products close to their
dates of expiry. This ultimately imposes a negative financial impact on dairy industry
[25]. Therefore, there’s a developing need for an accurate milk spoilage detection
method as much as there’s a growing demand in preventing the wastage of milk and
the illnesses occurring due to deteriorated milk consumption.
Shown in the Fig. 1, is the basic block diagram of milk supply chain. To understand
where the crucial challenges and problems lie in the system, it is necessary to study
this hierarchy including the inherent difficulties involved in it. This ranges from the
procurement level of milk taking place in dairy-farms, to the transport and handling
of milk, and stretches to the end consumer level. Primary quality concerns rise at
various stages of this chain, including the retailers that sell the packaged and
processed milk products received from the dairies. For example milk is tested at
both, procurement and packaging levels, by the industries for required amount of
quality and its assurance, but after each phase of transportation it is most likely that
the quality of milk product is altered, especially when the distance is longer and the
quantity is large.
Fragmented Milk Production, where milk production takes place in an unorga-
nized sector, is a big challenge and raises concerns when it comes to milk quality
and milk procurement over many economies like India—which is the largest pro-
ducer as well as consumer of milk, having the largest livestock worldwide [26]. The
research report by the team also adds that milk supply chains face difficulties in
transporting milk, especially in large quantity, under desired conditions over longer
distances (>200 km) while ensuring the quality of milk products.
This research work is motivated by the need to eliminate the above mentioned
adversities faced by the dairy industry, milk supply chain and the consumers with
regards to the contamination, spoilage and overall quality of milk in terms of
constituents present, including milk adulteration. The issues concerning suitable
levels of fat, added sugar and flavouring in milk also needs to be addressed [7],
where no economical, effective yet easy-to-use method currently exists to achieve
these goals [25]. In the past two decades, the evolvement in high-speed instrumental
testing techniques to measure milk and dairy product composition provided the
required platform for improvement in overall effectiveness and accuracy of classical
chemical analysis methods [23].
In October 2015, raw milk produced by a dairy in state of California, USA, was
ordered to be immediately recalled and quarantined, by a state veterinarian,
whereas consumers were asked to dispose of any products from the given batch of
code date Oct 24 [27]. The order was enforced after confirmed detection of
Campylobacter type bacteria were found in raw whole milk, by the inspectors at
routine sample test. Infection caused by these bacteria could show no symptoms at
all or it could harm the immune system and be life threatening too. Mainly it
includes symptoms like diarrhoea, cramps in stomach, or fever after 2–5 days of
being exposed to the bacteria and remains around a week. Less likely yet possible
symptoms could be joint pain or swelling. Also, a rare illness known as
Guillain-Barré syndrome which brings weakness and paralysis can also occur some
weeks after the initial sickness.
This is one of many stories, which surface regularly from around the world,
related to milk, its contamination, spoilage or adulteration. Food Safety and Stan-
dards Authority of India (FSSAI) defines adulteration in food as, ‘An act of
intentionally debasing the quality of food offered for sale either by admixture or
substitution of inferior substances or by the removal of some valuable ingredient’
[28].
Adulteration can be in either form:
• Deliberate, which is done to gain commercial benefits, like addition of water or
• Accidental, that occurs incidentally due to several reasons like inappropriate
handling.
Around 68 % of milk in India, which is the largest producer of milk worldwide
with 16 % share of global milk production [4], doesn’t meet the quality standards as
recommended [29].
Milk Quality Monitoring Using Electromagnetic Wave Sensors 211
This work seeks to address the issues faced at consumer end of the milk production
supply chain hierarchy as at the milk procurement level dairy industries already
have the robust quality assessment techniques suitable just within the industries as
explained in following Sect. 3 of this chapter.
With this novel sensor application at retailers’ end, the sellers can implement
random checks for the packaged products from a given batch of packaged milk
without having to check every single pack of the same batch. Therefore, it is
broadly applied at the link just before the very end consumer filling in the gap that
the dairy industry has after long transport phase where milk is most likely to
get altered in its quality over longer distances as explained earlier. This can also
eliminate the need to send the test samples to laboratories and awaiting their quality
test results for a day or two, offering a real-time solution to the problem.
• Electrical methods are traditional and use the changes in current through an
amperometric sensor as a measure of determining spoilage. This technique
employs a potentiostat and two electrodes which are immersed in milk samples
having different concentrations of bacterial inoculums and containing methylene
blue [35]. In these types of method spoilage is detected when growing meta-
bolism due to multiplying bacteria leaves a coloured solution into colourless
[25]. The microbial metabolism causes the reduction in methylene blue, which
eventually results into change of current through the electrodes. The Detection
Time (DT) required to sense the detectable change of current gives the measure
of present microorganisms at a given time. This is simpler method as compared
to the orthodox bacteria plating techniques with the only drawback that it
requires continuous supervision.
• Optical test methods, which are conventional, are costly approach to determine
milk spoilage because of the reason that fresh milk micelles have sizes that require
scattering setups to include sources and detectors that operate at wavelengths
λ < 300 nm, which are on the deep Ultraviolet (UV) range, covering the visible
(VIS) range [36]. An approach, with low-cost optical capillaries, to test the quality
of milk involves an optical sensor configuration along with the fibre optical setup
used a simple photonic system with optical capillaries to determine the milk
quality by observing under certain heating conditions [37]. It proved that the
optical capillary is a suitable medium to analyse liquids that show high scattering
of light, e.g. milk. The overall setup used there is still more complex than the
proposed technique based on EM wave sensor in this chapter. Also, such instru-
mentation is relatively costlier, compared to the proposed work here, as that is
made up of a light source, fibre optical link, sensing head, and two photo-detection
units which are controlled by an intelligent detection-and-control system. It, pri-
marily, focused on the variation in quality resulting from predefined storage
conditions. This process requires the heating of milk, as the test takes into account
the influence of temperature on the structure of milk, and hence local heating to the
capillary is introduced [38], which results in quite a large instrument setup.
• Protein count or fat count is also one approach to determine the spoilage in
milk as by measuring the protein or fat counts [39]. Pseudomonas aeruginosa
(P. aeruginosa) have been correlated to proteolytic activities in all food systems
by many other studies as well. It produces a large amount of proteinases and
lipases as well as plays an important role in producing by-products by taking
part in lipid and protein breakdown along with temperature effects and storage
time conditions [40]. So a good amount of technical ability and knowledge is
required, which makes this method less favourite amongst end users. The
industrial approach for determination of protein levels is known as Dumas
method, which is more preferred as it is enhanced version of the standard
214 K.H. Joshi et al.
Kjeldahl method [41]. Both the methods first need to convert measured nitrogen
concentration to a protein concentration, as they don’t directly measure the
protein, which leads to inaccuracies.
As one report from Food and Agriculture Organization of the United Nations
suggests that even after the efforts over past half-century, there still exists the need
for an internationally accepted methods with supporting data [42]. In fact, as it
further describes, the development of new methods for analysing specific compo-
sition elements that yields energy (like protein, fat or carbohydrates) has further
increased the complexity and made this need greater than ever before.
Milk sensing technologies, currently being used, still lacks the intelligent
approaches in practice [43]. As after the determination of good or poor quality of
milk, most of the systems don’t offer the automated diversion of poor quality of
milk so as to maintain the high standards of their products at manufacturers end.
From the above discussion, it is clear that there is scope and also need of
improvement in milk quality testing methods. When it is hard to overcome all the
limitations in all the characteristics as shown in Table 1, there is still a way to
minimize the overall shortcomings of existing approaches.
The use of EM wave sensor can offer good balance among the overall aspects
including performance, costs, reliability and less complexity, in comparison with
the currently used standards. In this chapter, as described in following section, the
attempt is to address these limitations by the design and development of a sensory
system that is able to distinguish among given types of milk based on:
• Fat composition (five different samples ranging from 3.6 to 0.1 g per 100 ml)
and
• Spoilage detection between fresh and old milk samples
Table 2 summarises these current advancements, as discussed above, for
detecting milk quality using various measures, along with the main parameter that
they detect to enable the test, its main limitation and primary advantages as further
explained in [25]. The techniques are categorised based on whether they are
spoilage detection methods or composition tests. This list can be more extensive
with increasing amount of interest in the field of food quality assessment, especially
with the help of various sensing technologies. This work is an attempt to address the
shortcomings that the current technologies have while giving optimum results for
determining the milk quality.
Milk Quality Monitoring Using Electromagnetic Wave Sensors 215
The antennas used for wireless and mobile communication operate on the EM wave
transmission and reception. The EM wave based sensors, as used in this work, are
the similar devices customized for different type of applications of sensing and
detection. Wide range of variety exists within this group of sensors based on the
way they are made, type of operation or range of frequency they work upon. EM
wave or microwave sensors are mainly categorized into two groups, based on their
applications. One group is of the radiometer, topographic and radar sensors for
applications related to distance, movement and shape measurement [44]. The sec-
ond group, which is related to this research work, is of the sensors like resonance
cavity, waveguide sensors and transmission or reflection sensors, which are used for
measuring material property, especially those of liquids and gases flowing or held
through within them [9].
EM wave sensing or microwave sensors operate based upon the fact that the
material under test condition, e.g. milk sample in this case, when placed into
proximity or in direct contact with a microwave sensor, interacts with the EM
waves that can be specifically correlated with the properties of the given material
(milk). Particularly, the sensing is based on the interaction of propagating or res-
onating modes with the liquid under test. Due to this interaction, the permittivity of
the material changes and it reflects itself as a frequency change, attenuation of the
signal or a phase shift [45].
The material to be tested with this methodology is exposed to the electromag-
netic field generated by the sensor fed through a source. For example, the 2-port
microwave resonator cavity is designed to hold the test-tube containing milk sample
to carry out the quality sensing, as depicted in Fig. 2. The changes captured in the
sensor’s spectral responses, through a network analyser, then become indication of
the differences among samples under consideration [46]. These differences could be
based on the composition, i.e. milk types, or could be changes between the
adulterated (spoiled) and fresh samples of the same milk product.
Fig. 2 An EM wave
resonator cavity sensor
Milk Quality Monitoring Using Electromagnetic Wave Sensors 217
These types of sensor have following primary advantages over the types that
have been discussed in earlier sections:
• Less complexity in installation of experimental/lab setups
• Universal applications (i.e. not specific to each bacteria/lipid/protein type)
• Wide range of variety according to needs of the application
• Better accuracy and less costs
For development of a new technology that can evaluate milk quality, it can be
important to possess the understanding of dielectric properties of milk. The
dielectric constants for both, the raw milk (with 100 % concentration of milk) and
the diluted milk (with 70 % milk concentration to deionized water) decreases with
increasing frequency, more so at the lower frequency range of 10–4500 MHz, but
dielectric constants for raw milk decreased more rapidly than in case of the diluted
one, at room temperature of 22 °C [47]. For raw milk it was 97.7, 68.1, and 65.9,
whereas 93.3, 70.9 and 69.1 for 70 % milk solution, at 10 MHz, 915 MHz, and
2450 MHz, respectively. The dielectric loss factor can be another potential indi-
cator for prediction of milk concentration as well as freshness. Since, the adulter-
ation of milk with water dilutes the ionic concentration within milk, the loss factor
decreases with the increase in water content, which is decrease in the milk
concentration.
Another work, concludes, that both, the dielectric constant as well as the loss
factor, for cow milk, are affected by the protein content and also the temperature of
milk. The frequency of electric field influences these two values, too. The dielectric
constant decreased with increase in the frequency within range 10–4500 MHz [48].
The dielectric loss factor increased with increasing temperature below around fre-
quency values of 600 MHz and decreased above frequencies of 1000 MHz. Poor
correlation was seen between the dielectric constant and the protein content in the
region of 100–600 MHz frequency values which means that this frequency range is
insufficient for being used to develop milk protein detector on the basis of dielectric
constant values.
This work takes into account the literature studied and uses resonator cavity as an
EM wave sensor operating at microwave frequencies to evaluate milk quality. The
procedures adopted to carry out the milk sensing and quality tests using this sensor,
are explained below.
218 K.H. Joshi et al.
Three milk types, namely skimmed, semi-skimmed and whole milk were bought
from a local chain of stores for the tests. Five samples were made from each of this
categories and labelled as shown in Fig. 3. To ensure that we have more accurate
results and more no. of points for Fat versus Frequency graphs, another two cat-
egories were made by mixing two types of milk respectively. That is 50 % whole
milk was mixed with 50 % semi-skimmed milk and 50 % of skimmed milk to the
another 50 % of semi-skimmed milk which ultimately gives us 2 more varieties (5
in total) in terms of fat content. Each sample was of 15 ml in size, held in a
polypropylene plastic test-tube.
The samples were made at room temperature and the storage conditions were
governed in a temperature controlled climate with minimum fluctuations. For the better
accuracy of captured spectral data, 5 repetitions were made for each of the 5 samples for
all 5 categories of milk giving eventually 125 iterations to be mathematically analysed.
The test-tubes as shown in Fig. 3, were labelled as W-1, W-2, …, W-5 for whole milk,
SS-1, SS-2, …, SS-5 for semi-skimmed milk and S-1, S-2, …, S-5 for skimmed milk as
abbreviated, respectively. Similarly semi-skimmed + skimmed milk solution and
semi-skimmed + whole milk solutions were labelled SSS-1, SSS-2, …, SSS-5 and
SSW-1, SSW-2, …, SSW-5, respectively. Table 3 lists the test measurement specifi-
cations with sample storage condition.
For the measurements with EM wave cavity resonator sensor a slot that can be
inserted with the milk samples in 15 ml test-tubes, was used as shown in Fig. 4.
The 2-port cavity was calibrated with the VNA, prior to measurement to enable
accurate measurements of S11 (reflected back signal to input port-1) and S21
(transmitted signal from input port-1 to output port-2). The captured raw data of
spectral signatures consisted of 60,000 point values each for different frequency
within the range given, i.e. 10 MHz–15 GHz. Laser beam thermometer along with
the digital probe thermometer were used to ensure stable temperature conditions
before, during and after the measurements as can be seen from the illustration.
For the measurements a 2-port EM wave cavity was used as shown in Fig. 4
along with a VNA. Wide frequency sweep (10 MHz–15 GHz) allows more spectral
Fig. 3 Sample preparation for milk quality testing using EM wave spectroscopy
Milk Quality Monitoring Using Electromagnetic Wave Sensors 219
Fig. 4 EM wave cavity testing milk sample using VNA in temperature controlled climate
The EM wave cavity sensor was modeled and simulated with the help of High
Frequency Structure Simulator (HFSS), as shown in Fig. 5, before the development
of its fabricated hardware. The resonance frequencies for TE mode and TM mode
equations for microwave cavity are given by Eq. (1) and Eq. (2), respectively [49]:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
′ 2
1 Xmn pπ 2
fmnp = pffiffiffiffiffi + ð1Þ
2π μϵ R L
220 K.H. Joshi et al.
Fig. 6 3-D Polar plot of EM wave cavity for the frequency sweep 10 MHz–15 GHz
Milk Quality Monitoring Using Electromagnetic Wave Sensors 221
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 Xmn 2 pπ 2
fmnp = pffiffiffiffiffi + ð2Þ
2π μϵ R L
where, R and L are radius and length for a cylindrical cavity, respectively. Xmn
represents nth zero of mth Bessel function and X′mn represents nth zero of derivative
of mth Bessel function. µ and ϵ are permeability and permittivity of the medium,
respectively. Here, p is the no. of half waves in the axial direction.
Figure 6 illustrates the field polar plot in 3-dimensions showing how the radiated
EM wave energy spreads through the cavity. The primary difference between the
microwave sensors and communication antennas is their far-field patterns. As we
can see from the plot, in the case of a sensor, radiation is mainly near-field whereas
it is opposite in case of an antenna. The reason behind that is, an antenna is
supposed to radiate over a distance to enable wireless communication between two
points while a sensor is ideally meant to sense the material in its nearest proximity
with or without physical connection with the material being sensed.
6 Results Discussion
As shown in Fig. 7 for whole milk, there is a clearly visible distinction between the
curves for S21 plot of Day-1 and Day-7. Five repetitions for each sample were
made. Similarly, Figs. 8 and 9 show the graphs for semi-skimmed and skimmed
milk, respectively. These plots are identical with minor shifts for all 3 given types
Fig. 7 S21 variations between Day-1 (Blue) and Day-7 (Grey) for whole milk
222 K.H. Joshi et al.
Fig. 8 S21 variations between Day-1 (Green) and Day-7 (Grey) for semi-skimmed milk
Fig. 9 S21 Variations between Day-1 (Red) & Day-7 (Grey) for skimmed Milk
of milk and shows effective results to distinguish between the aged and fresh milk
samples.
Figure 10, on the other hand, distinguishes among the 5 various samples made
from 3 categories, based on their fat content, of milk and labelled as S, SSS, SS,
SSW and W as explained earlier. The figure shows spectral graph of S21 against the
fat content of milk. The biggest advantage of this novel technique is, it does not
Milk Quality Monitoring Using Electromagnetic Wave Sensors 223
Fig. 10 S21 variations with respect to fat contents of 5 types of milk samples (Day-1)
Fig. 11 S21 variations with respect to protein contents of 5 types of milk samples (Day-1)
only detect ageing but also distinguishes quite well among the types of milk as can
be seen below. These 5 samples are achieved by 3 different milk types; skimmed,
semi-skimmed and whole, and another two by adding 50 % of each, Skimmed milk
with Semi-skimmed milk, and 50 % of each Semi-skimmed milk with the Whole
milk, in order of fat content values.
We see the linear regression with correlation value, R2 = 98 %, at the frequency
value of 5.4598 GHz. This technique is less complex, and therefore, quick-time
result oriented method than other methodologies previously explained in this arti-
cle. The results achieved in this preliminary work of the whole project give accurate
detection and efficient outcomes.
Figure 11 shows the S21 plot versus the protein content graph of the five types of
milk. Here, unlike in case of fat content variations, the linear regression gives
224 K.H. Joshi et al.
7 Future Work
8 Conclusion
be made standalone and applied to large or medium scale retailers’ level of milk
supply chain hierarchy to avoid ambiguities and wastage of large quantity of milk
as well as prevent any food-borne illness occurring due to spoilage or adulteration
of milk.
This, with further optimasation, can help the large-scale or medium-scale
retailers, who receive the milk from dairies for selling in packaged form after long
stages of transport, to help allow the quality assurance and also to ensure that the
customers are getting what they are paying for. This work not only will establish a
strong base for the laboratory based research but also must aid towards the problem
solving that the dairy industries and end consumers face due to either complexity or
costliness of existing standard practices for milk quality control.
References
15. M. Fok, M. Bashir, H. Fraser, N. Strouther, A. Mason, A novel microwave sensor to detect
specific biomarkers in human cerebrospinal fluid and their relationship to cellular ischemia
during thoracoabdominal aortic aneurysm repair. J. Med. Syst. 39(4) (2015)
16. S.G. Bjarnadottir, K. Lunde, O. Alvseike, A. Mason, A.I. Al-Shamma’a, Assessing quality
parameters in dry-cured ham using microwave spectroscopy. Meat Sci. 108, 109–114 (2015)
17. P. Fellows, A. Hampton, Small-Scale Food Processing—A Guide for Appropriate Equipment
(FAO, Intermediate Technology Publications, London, 1992)
18. Milk Factsheet, The Dairy Council (The Dairy Council, London, 2016)
19. H.C. Deeth, T. Khusniati, N. Datta, R.B. Wallace, Spoilage patterns of skim and whole milks.
J. Dairy Res. 69(2), 227–241 (2002)
20. J.J. Janzen, J.R. Bishop, aB Bodine, Relationship of protease activity to shelf-life of skim and
whole milk. J. Dairy Sci. 65(12), 2237–2240 (1982)
21. J.V. Brown, H.M.P. Ranjith, G.A. Prentice, Comparative shelf-lives of skimmed,
semi-skimmed and whole milks. Int. J. Dairy Technol. 37(4), 2–5 (1984)
22. R.E. Chandler, S.Y. Ng, R.R. Hull, Bacterial spoilage of specialty pasteurised milk products.
CSIRO Food Res. Q. 50, 11–14 (1990)
23. D.M. Barbano, J.M. Lynch, Major advances in testing of dairy products: milk component and
dairy product attribute testing. J. Dairy Sci. 89(4), 1189–1194 (2006)
24. S.C. Murphy, Shelf-Life of Milk Products—Microbial Spoilage—The Evaluation of Shelf-Life
(Ithaca, New York, 2009)
25. M. Lu, Y. Shiau, J. Wong, R. Lin, H. Kravis, T. Blackmon, T. Pakzad, T. Jen, A. Cheng,
J. Chang, E. Ong, N. Sarfaraz, N.S. Wang, Milk spoilage: methods and practices of detecting
milk quality. Food Nutr. Sci. 04(07), 113–123 (2013)
26. NPCS Team, Market Research Report on Milk Processing & Dairy Products in India (Butter,
Yogurt, UHT Milk, Cheese, Ice Cream, Ghee & Other Products). NIIR Project Consultancy
Services (2014)
27. News Desk, Raw Milk Recalled in California for Campylobacter Contamination. Food Safety
News (2015). http://www.foodsafetynews.com/2015/10/raw-milk-recalled-in-california-for-
campylobacter-contamination/#.VnJ4MFduTOs
28. Food Safety and Standards Authority of India, Adulteration in Milk and Milk Products,
FSSAI (2014). http://www.fssai.gov.in/Portals/0/Pdf/Adulteration(26.02.14).pdf. Accessed 03
Apr 2016
29. S. Saxena, Is the Milk You’re Drinking Safe? 4 Easy Ways to Spot Adulteration, NDTV
Food, 18 Mar 2016
30. T.R. Klaenhammer, Bacteriocins of lactic acid bacteria. Biochimie 70(3), 337–349 (1988)
31. R. Guntupalli, R.S. Lakshmanan, J. Hu, T.S. Huang, J.M. Barbaree, V. Vodyanoy, B.A. Chin,
Rapid and sensitive magnetoelastic biosensors for the detection of Salmonella typhimurium in
a mixed microbial population. J. Microbiol. Methods 70(1), 112–118 (2007)
32. C. Ruan, K. Zeng, Ca Grimes, A mass-sensitive pH sensor based on a stimuli-responsive
polymer. Anal. Chim. Acta 497(1-2), 123–131 (2003)
33. J.E. Haugen, K. Rudi, S. Langsrud, S. Bredholt, Application of gas-sensor array technology
for detection and monitoring of growth of spoilage bacteria in milk: a model study. Anal.
Chim. Acta 565(1), 10–16 (2006)
34. H.M. Al-Qadiri, M. Lin, Ma. Al-Holy, aG Cavinato, Ba Rasco, Monitoring quality loss of
pasteurized skim milk using visible and short wavelength near-infrared spectroscopy and
multivariate analysis. J. Dairy Sci. 91(3), 950–958 (2008)
35. Y.G. Lee, H.Y. Wu, C.L. Hsu, H.J. Liang, C.J. Yuan, H. Der Jang, A rapid and selective
method for monitoring the growth of coliforms in milk using the combination of
amperometric sensor and reducing of methylene blue. Sensors Actuators B Chem. 141(2),
575–580 (2009)
36. P. Dress, M. Belz, K.F. Klein, K.T.V. Grattan, H. Franke, Water-core waveguide for pollution
measurements in the deep ultraviolet. Appl. Opt. 37(21), 4991–4997 (1998)
37. M. Borecki, M. Szmidt, M.K. Paw, M. Beb, T. Niemiec, P. Wrzosek, A method for testing the
quality of milk using optical capillaries. Photonics Lett. Pol. 1(1), 37–39 (2009)
Milk Quality Monitoring Using Electromagnetic Wave Sensors 227
Abstract Parmigiano Reggiano (PR) cheese is produced from the bovine milk in a
limited geographic area in northern Italy. It is one of the oldest traditional cheeses
manufactured in Europe (XII century) and is still one of the most esteemed Pro-
tected Designation of Origin (PDO) cheeses of Italy. The Parmigiano Reggiano
Cheese Consortium (CFPR) is the institutional system that covers all the cheese
dairy factories and controls the cheese production in terms of cow feeding, cheese
manufacture and ripening processes. The aim of this study was to establish a new,
rapid, portable, easy-to-use, economic and non-destructive fouling based on
nanowire technology device (S3) to control the presence of false grated PR. All the
analyzed samples were collected from the Consortium that picked the samples
produced in different European countries (three samples from Germany, one from
Ireland, one from Latvia and one from Lithuania). The achieved samples were
analyzed with the cooperation of multidisciplinary technologies like Colorimeter,
pH Meter, GC-MS with SPME and the new nanowire gas sensor device (S3). The
applied techniques were able to create a specific database for the S3 device, capable
to give results in a few minutes. The use of a Small Sensor System (S3) device to
determine the specific volatile organic compounds (VOCs) of the grated PR is a
promising approach offering, at the same time, a simpler, faster and easier to handle
(no specialized technicians are required) solution with respect to the analytical
techniques (GC-MS, HPLC). However, from the commercial point-of-view, low
cost and reduced power consumption are the fundamental features that should entail
a sensor device. In this study, S3 device was applied and found to be efficient in the
discrimination of the different grated PR cheese samples. In particular, the dis-
crimination was checked quali-quantitatively thanks to the analysis performed with
the GC-MS-SPME technique. In fact, the achieved results demonstrated the
1 Introduction
Parmigiano Reggiano is a hard cheese, slowly matured, made from the raw
semi-skimmed milk from cows whose diet consists mainly of fodder from the
original area (Reggio Emilia, Parma, Modena, Bologna and Mantova). The milk
may not be subjected to a chemical treatment and is not allowed for the use of
additives [1–4].
Through this characteristics a form of cheese is produced called as Parmigiano
Reggiano, a Protected Designation of Origin (PDO) product, according to the
European EEC Reg. 2081/92 and the recognition of Reg. (EC) No 1107/96 by its
disciplinary and as evidenced by the Consortium.
This organization, therefore, has the task to protect the Parmigiano Reggiano
from its imitations and many frauds that are being reported around the world. In
fact, the designation of origin of the cheese ‘Parmigiano Reggiano’ extends to the
grated type cheese.
In addition, the cheese must have some special characteristics: fat content in dry
matter must not be less than 32 %; aging not less than 12 months and within the
limits fixed by the production standard.
It must not contain additives; organoleptic characteristics must be consistent
with the definitions laid down by the production standard; humidity not less than
25 % and not more than 35 %; aspect: not crumbly, homogeneous, particles of
diameter less than 0.5 mm not exceeding 25 %; rind quantity not greater than 18 %;
amino acid composition: specific of ‘Parmigiano Reggiano’ [4, 5].
All these characteristics will be treated deeply in the next epigraphs of the
current chapter.
The link between Parmigiano Reggiano and its area of origin is essential. Parmi-
giano Reggiano comes from the territory and the tradition of man. It is produced
today just as nine centuries ago: the same ingredients (milk, salt, rennet), the same
care and passion, the same area of origin.
The area of production of Parmigiano Reggiano (Fig. 1) is located in the pro-
vinces of Parma, Reggio Emilia, Modena, Bologna to the left of the Reno River and
Mantova to the right of the River Po.
Grated Parmigiano Reggiano Cheese: Authenticity … 231
This area must be involved either in the milk production or its transformation
into the cheese.
The milk for the production of Parmigiano Reggiano comes from the cattle fed with
natural fodder and the production process starts from the raw milk. The actual
process is carried out using the same ancient technique but in modern settlements
with new machinery.
The raw material is the evening milk, which is allowed to stand in large stainless
steel tanks, as shown in the Fig. 2, to bring out the cream, and the whole milk from
the morning milking. In the morning, it proceeds with mixing the semi-skimmed
milk and the whole milk in copper vats shaped like an inverted bells.
The whey and the calf rennet are added after the milk has been heated to a
temperature of 33 °C. After the mongrel coagulation, which takes place in 15 min,
and thanks to the spino (Fig. 2), the curd is broken into small fragments having the
size of a grain of rice and is cooked up to 55 °C [1].
After these operations, the mass is left to rest for 30 min, then extracted with a
linen canvas of 1 m2 and is placed in the molds, where it undergoes a pressing step.
After a few hours, the wet cloth is changed with a new one and then is removed
to insert a matrix for stamping the identification data of the form on the crust: month
and year of manufacture, the serial number of the dairy and the dots-label for
Parmigiano Reggiano. Salting is carried out by bathing in brine for 25–30 days.
Maturation takes place in appropriate constant temperature and humidity
warehouses, where the forms are constantly brushed and turned; to perform a
homogenous seasoning and to prevent the overgrowth of surface mold.
Only after a minimum period of 12 months, the forms that will be found correct
by the examination of the Consortium Body Control become full Parmigiano
Reggiano.
232 V. Sberveglieri et al.
Fig. 2 a The addition of rennet, b the dowelling, c collection in the linen bands, d the molding,
e salting and f the curing in a controlled environment
The use of a Small Sensor System (S3) device, to determine the specific VOCs
of grated PR is a promising approach offering, at the same time, a simpler, faster
and easy to handle (no specialized technicians are required) solution with respect to
the analytical techniques (GC-MS, HPLC).
However, from the commercial point-of-view, low cost and reduced power
consumption devices are the fundamental features.
Gas chemical sensor arrays or Electronic Noses (EN) have been identified in the
past few years as the valuable candidates for different applications including food
quality control.
1.3 Aim
The aim of this study was to establish a new, rapid, portable, easy-to-use, economic
and non-destructive fouling based on the nanowire technology device S3 to control
the presence of false grated PR. All the grated cheeses were prepared using the same
grinding machine using a predefined mesh dimension (ASTM series, mesh n°35).
The possibility of detecting the addition of cheese (with a big range of different
seasoned cheeses from different country of origin) to the grated PR cheese has been
always studied by means of the destructive chemical methods.
Thus the application of a novel nanowire device S3 can serve as an effective
means to determine the authenticity of a PR cheese and also to characterize its
quality parameters at the same time. Six different foreign cheese samples originated
from Latvia, Lithuania, Ireland and Germany (Hartkase, Hartkase 1, Hartkase 11)
were analyzed.
The preparation of the samples (Table 1) was carried out as follows: 2 g of grated
cheese was placed in a 20 ml chromatographic vial. Subsequently all the vials were
covered with the Silicon-PTFE septum and crimped with an aluminium crimp. At
the same time, the water activity was measured in order to control and assure that
this parameter will not influence the sensor response.
At SENSOR laboratory, the studies on chemical sensors started in 1988 with the
improvement of thin films and then of a new technique for the planning of thin films
with an extremely porous structure.
In 2002, quasi-one-dimensional metal oxide nanostructures were achieved with
several advantages compared to their thin and thick film counterparts such as large
surface-to-volume ratio, lateral dimensions comparable to the surface charge region,
and superior stability when in the single crystal structure.
The novel nanowire device S3 is a production of the SENSOR laboratory.
The Small Sensor System (S3) device (Fig. 3) is equipped with a thermally
controlled sensor chamber of 20 ml internal volume where there are placed 6 MOX
gas sensors.
Three of these sensors were prepared with the RGTO thin film technology [6],
and the other three constructed with the nanowire technology [7]. In fact, two of the
three nanowires in the array are zinc oxide sensors but with different operating
temperatures and the third one is a tin oxide sensor [8].
The vials containing different samples were placed in a randomized mode into
the HT280T carousel. Each vial was incubated at 40 °C for 10 min into the
HT280T oven, by shaking it throughout the incubation.
The sample headspace (4 ml) was then extracted from the vial in the static
headspace path and injected into the carried flow (speed 4 ml/min) through a
properly modified gas chromatography injector (kept at 40 °C to prevent any
condensation).
Using synthetic chromatographic air with a continuous flow rate of 10 ml/min,
the sensor baseline was performed and the recovery time was 28 min. The data
analysis was run by means of Principal Component Analysis (PCA), operated with
the Nose Pattern Editor software (SACMI Imola scarl, Imola, Italy).
The scheme of the sensor device based on an alumina transducers 2 × 2 mm2
with the heater on the backside is shown in Fig. 4 together with the SEM images of
nanowires. Tin and Zinc oxide nanowires exhibit a remarkable crystalline quality
and a very high length-to-width ratio, resulting in an enhanced sensing
performance.
The nanowire sensors thus manifesting a very high length-to-width ratio create a
3 dimensional network (Fig. 4). In this way, the adsorption surface is increased in a
huge amount enhancing the response of the instrument and decreasing the
threshold.
They exhibit a remarkable crystalline quality as well. These characteristics result
in an enhanced performance and a long-term stability for the sustained operations.
The data analysis was run by means of Principal Component Analysis (PCA).
Data were processed by EDA software, a home written software developed in
MATLAB® at SENSOR laboratory [9].
Fig. 4 Left Schematic sketch of the conductometric device. Alumina substrate is white colored in
the middle, platinum heater and contacts are in metallic gray. SEM images of Tin (center) and Zinc
(right) oxide nanowires
236 V. Sberveglieri et al.
Feature extraction is usually unsupported, the most common example being the
Principal Component Analysis. PCA can, by itself, transfer a good visual idea of the
interclass split-up; in fact, this is how it is usually applied.
The goal was to obtain a good visual separation of the clusters, not just to have
high estimate accuracy. To this end, PCA was systematically used to judge the
clusters separation [10, 11].
3.1 S3 Device
The data were processed by the software ‘Nose Pattern Editor’. In Fig. 5, PCA
score plot can be observed showing the comparison of all the samples for the
foreign cheeses. The instrument was able to well differentiate the groups of the
samples, here represented by different colors.
The two PC components alone covered over 60 % of the total information. This
difference is then explained by the analysis first conducted with the GC-MS in
Grated Parmigiano Reggiano Cheese: Authenticity … 237
which several compounds are observed between the different samples and also
different amounts of the same compounds (Fig. 5).
The next step was then to compare these foreign samples with the pulp and crust
composition of the real Parmigiano again using the PCA analysis. From the
obtained PCA (Fig. 6), the clusters of the samples were observed to be quite
compact.
From the figure, it is seen that the foreign samples are not confused with the real
Parmigiano Reggiano cheese samples, but some similarities in the foreign cheese
samples with the crust or the pulp are found. In fact, in the left side the black dots
Fig. 6 PCA score of foreign cheeses in comparison with 100 % pulp and 100 % crust of
Parmigiano Reggiano
238 V. Sberveglieri et al.
are the detached samples of the crust and near these are all Hartkase1 and Hart-
kase11 samples.
These samples are more similar to Parmigiano Reggiano crust compared to the
pulp or to other foreign samples, confirming the fact that probably these samples
have very high percentages of the crust that characterize them.
On the other hand, the PCA score plot includes the pulp cluster of the real
Parmigiano close to the clusters of Ireland, Hartkase and Lithuania. Probably these
samples show the most similarities to the pulp of Parmigiano Reggiano over the
crust.
The figures below report the frequency histograms of the compounds present in
different samples divided into two parts to allow an easier viewing of the graphs
(Figs. 7 and 8).
The samples analyzed are very different from the true Parmigiano: even with a
fast sensory analysis many atypical features are observed later confirmed by the gas
chromatography, which included the analysis of different true Parmigiano Reggiano
cheese samples.
In particular, there are very significant compounds present such as butanoic acid
derivatives that characterize especially the Ireland grated sample.
The most abundant compounds in this sample are propanoic acid, butanoic acid
and derivatives, tridecyl-3-butyric acid ester, acetic acid phenylmethyl ester.
These compounds are typical for Cheddar cheese, therefore the Irish sample
could possess a small percentage of real Parmigiano (in fact possesses some sim-
ilarities: 1-butanol-2methyl, 2-nonanone, octanoic acid ethyl ester and propanedioic
Fig. 8 Frequency histogram of foreign cheeses compounds from 24.1 to 59.5 min
acid) but for the major part it seems to have been mixed with the Cheddar cheese. It
is also very different from the other foreign samples.
The German samples have similarities between them and even consist of dif-
ferent compounds. They are very different from the true Parmigiano, small analo-
gies may be found with the crust as confirmed by the PCA provided by the
Electronic Nose and could probably contain Emmental or Camembert as confirmed
by the comparison with the bibliography.
The Latvian sample is the only one to get closer to the true profile of Parmigiano
Reggiano (see for example: 2-heptanone) which presents the typical compounds of
Parmigiano both from the pulp and the crust.
However, the analogy is greater for the crust (see the presence of 2-pentanol,
2-heptadecanone, 2-octanone, 2-nonanone, octanoic acid ethyl ester, 2-nonanol,
2-undecanone, heptanoic acid, octanoic acid).
Significant is the presence of dodecanoic acid which denotes a typical aroma of
the cheese crust and the presence of 5-methyl-2-hexanol, wood aroma denoted by
the real Parmigiano crust during ripening in which the forms remain for months on
the wooden planks.
The only foreign presence of phenylmethyl acetic acid ester and anethole in the
samples makes us suspect that the foreign samples and imitations are mixed with
the small percentages of Parmigiano indicating the commercial fraud.
Here the compounds found in the chromatograms have been classified selecting
only those compounds that characterize the different origins of the foreign cheese
samples defining the aroma.
• Latvia: The compounds that characterize the aroma of the cheese from Latvia
were:
240 V. Sberveglieri et al.
4 Conclusion
The fight against counterfeiting has been a priority for the protection of PDO
products. The globalization of markets not only provides several opportunities but
also many risks: the ability to package the product in the world makes the making
of documentary controls increasingly difficult.
In this sense, the Parmigiano Reggiano Cheese Consortium is working to change
the rules of the production of Parmigiano Reggiano, to make the packaging com-
pulsory in the area and to develop the sophisticated solutions for the traceability of
the product.
In parallel, however, the target is to find the technical and organizational solu-
tions that increase the chances of the cheese downstream control that enters into the
commercial circulation, in terms of compliance with the same standards imposed by
the regulations and the way it is included in the agreements signed between the
Consortium and the packers.
In addition to this, several foreign samples and imitation of Parmigiano Reg-
giano are widely available in the market. The grated suspects were examined and
found to be mixed with the PR. Using the SPME-GC-MS and Electronic Nose
242 V. Sberveglieri et al.
techniques, the most significant compounds of the different samples were identified,
observing many differences in the chromatogram and in the PCA score plots
obtained from the Electronic Nose. In particular, the butanoic acid compounds and
derivatives were found to characterize especially the Ireland grated sample.
This sample presented different levels of those compounds in comparison to the
other samples. It is more abundant in propanoic acid, butanoic acid and derivatives,
tridecyl-3-butyric acid ester and acetic acid phenylmethyl ester.
These compounds are typical for the Cheddar cheese, therefore the Irish sample
could have possessed a small percentage of real Parmigiano but for the most part, it
seems to have been mixed with the Cheddar. The German samples are very different
from the true Parmigiano; little analogies may be found with the crust as confirmed
by the PCA provided by the Electronic Nose [15].
They could probably contain Emmental or Camembert cheese as confirmed by
the comparison with the published literature. The Latvian sample is the only one to
get closer to the true profile. It, in fact, presents the typical compounds of Parmi-
giano both from the pulp and the crust.
S3 device is a promising tool for the quality discrimination of different cheeses
and other foodstuffs. This ability to distinguish cheese quality by an easy-to-use,
rapid, economic, portable, non-destructive and low power consumption S3 device is
crucial as it can be utilized to validate if a given sample is according to the law and
standards in order to assess the fraudulent or imitative production.
The nanowire based S3 Electronic Nose represents a simple, fast, economic and
portable tool for the quality control and authenticity evaluation of grated PR cheese.
In our future works, we propose to illustrate an amplified database of the grated
PR for the direct application of S3 in the cheese-chain productions and also to
perform a data fusion by integrating different results for a deeper analysis.
References
M.A. Sairin (✉) ⋅ N.N.A. Nizar ⋅ A. Ismail ⋅ D.M. Hashim ⋅ F.Z. Rokhani (✉)
Halal Product Research Institute, Universiti Putra Malaysia, Seri Kembangan, Malaysia
e-mail: [email protected]
F.Z. Rokhani
e-mail: [email protected]
M.A. Sairin ⋅ S.A. Aziz ⋅ N.A.A. Latiff ⋅ A. Ismail ⋅ F.Z. Rokhani
Faculty of Engineering, Universiti Putra Malaysia, Seri Kembangan, Malaysia
1 Introduction
Fig. 1 Breakdown of food fraud being reported based on food ingredient category [2]
Figure 2, shows the breakdown of the world consumption for major fats and oils
in 2015 as reported by IHS Inc [17]. It can be observed that animal-based fats and
oils such butter, lard, tallow and grease ranked fourth most consumed fats and oil
globally [17]. According to [16], animal fats were used immensely for domestic
frying due to the desirable flavors they impart on some foods. For example, tallow
and lard was added to french-fried potatoes and pie-crusts respectively, to enhance
the flavor. In addition, animal fats are used as ingredient in the formulation of food
products such as breads and cakes due to exhibiting good stability and economical
[16, 18]. While both genuine and industrially-modified randomized lard are com-
mercially available [19], they are also used to blend with vegetable oils to produce
cost-effective margarines, shortenings and other oil-based foods [20].
In general, adulteration of edible oil causes two great concerns to the consumer.
First, animal-based fats and oils adulteration in edible oil present great concern to
followers of vegetarian diet such as Hindus or Buddhist [21]. In particular, lard
adulteration prompt concern to followers of Islamic and Judaism religions as both
religions prohibit the consumption of lard and pork [22, 23]. Several reported cases
include shortening, butter, margarine and other specialty food oils made of veg-
etable oils blended with lard or tallow [24, 25], goat and pig body fat detected in
vegetable ghee and cow/buffalo ghee respectively [26, 27] and recently, sunflower
oil contaminated with chicken fat [28].
Second, health issues concern, which can be very vital to people who have
allergies to a certain types of food as well as the relationship of dietary cholesterol
and saturated fatty acid to coronary heart disease [29]. In 1997, European Com-
mission reported that around 103,400 tonnes of olive oil adulterated with estimated
20,680 tonnes of hazelnut oil were to be placed on the market [30]. The presence of
nut oils could cause complication for people with food allergies to nuts [2]. In [31],
non-optimal intakes of polyunsaturated fat, trans fat and saturated fat each con-
tribute to significant coronary heart disease.
It is clear that there is high dependency of consumers throughout the world to
oils and fats. With the vast consumption of oils and fats, greedy manufacturers
attempted to reap profits by illicitly adulterating original oils with alternative
substances leading to numbers of food fraud cases. While in certain cases oil
adulteration is intended to enhance food flavor or give stability as ingredient, these
may cause serious health related issues or unable to be consumed by certain group
of people. Thus, concern in adulteration in oils and fat should not be taken lightly or
ignored. It is crucial to develop a verification system that can distinguish adulterated
edible oils to combat food fraud. This thus can induce a healthier and safer eating
lifestyle among the citizen of all ages, races and religions.
Table 1 Summary of conventional techniques used in lard detection on several types of samples
with different detection limit, and spectrum measured
Category Techniques Detection Spectrum Sample
limit measured
Labeling-based PCR – – Raw meats and fats [32]
– – Pork derivatives in sausages
and the casings, bread and
biscuits [33]
Spectral-based Electronic Nose 1.0 % Volatile RBD palm olein adulterated
compounds with lard [42]
spectrum
DSC 0.5 % Thermal Beef fat and chicken fat
spectrum adulterated with lard [37]
GCMS & – Mass Lard, chicken fat, beef fat and
EA-IRMS spectrum mutton fat [43]
GCFID 0.5 % Ion intensity Beef fat and chicken fat
spectrum adulterated with lard [44]
GC-TOF-MS – Mass Chicken fat, lard, goat fat and
spectrum cattle fat [45]
Liquid 2.0 % Mass Palm oil, palm kernel oil and
chromatography spectrum canola oil adulterated with
lard, beef tallow and chicken
fat [46]
FTIR 4.0 %a Absorbance Lard adulteration in chocolate
spectrum and chocolate products [10]
1.0 %a Pork adulteration in beef
meatball [47]
1.0 %a Cod-liver oil adulterated with
lard [48]
a
17.0 % Lard adulteration in cake
formulation [49]
1.0 % Lard blended with mutton and
cow body fats [50]
10.0 % Lard mixed with body fats of
chicken, lamb and cow [51]
– Lard, beef, chicken and
mutton fats, cod liver oil and
other vegetable oil [20]
a
assumed based on the lowest lard adulteration concentration investigated
cod-liver oil [48], lard adulteration in cake formulation [49], lard in mixture of body
fats of mutton, cow and chicken [50, 51] and to differentiate lard from other edible
oil [20]. Generally, these studies utilize the frequency region of 4000 cm−1 until
650 cm−1 which lies in the infrared spectrum of frequency. Table 1 summarizes all
of the techniques that have been discussed with their respective samples.
Lard Detection in Edible Oil Using Dielectric Spectroscopy 251
Aside from the data collected from measurements techniques discussed in previous
topic, data analysis is important to enhance the detection ability. The main goal of
this data analysis is to represent sample’s data in a way that the differentiation
between samples is significantly distinct.
Several examples of data analysis techniques that have been used in previous
studies are partial least squares (PLS), discriminant analysis, principal component
analysis (PCA), cluster analysis (CA), stepwise multiple linear regression (SMLR),
least significant differences (LSD), analysis of variance (ANOVA), linear regres-
sion analysis, and Pearson’s correlation analysis. Other statistical analysis tech-
niques include mean difference (MD), standard deviation (SD), standard error
(SE) and t-test. Some studies used one analysis technique while others used mul-
tiple analysis techniques at different stages during the data analysis process. Table 2
summarizes the different types of data analysis techniques and their respective
functions.
As an example, a study to analyze the adulteration of lard in cod-liver oil utilized
the PLS technique to develop a calibration model for quantitative measurement of
Fig. 3 a FTIR spectra of cod-liver oil adulterated with lard at different concentrations and
b Coomans plot of pure cod-liver oil (squares) and cod-liver oil spiked with different proportion of
lard (triangles) [48]
the adulterant, together with discriminant analysis method to classify cod-liver oil
samples from other animal fats, beef, chicken, mutton and lard. Figure 3 shows the
Coomans plot obtained from the discriminant analysis for the classification of pure
cod-liver oil and the ones spiked with lard. From the plot, it is observed that there is
clear group separation between pure cod-liver oil (squares) and the one that is
spiked with lard (triangles). Other than that, the values of root mean standard error
of cross validation (RMSECV) and coefficient of determination (R2) were used as
the validity criteria for the calibration model. The validation was further investi-
gated using the MD and standard deviation of difference (SDD) for accuracy and
reproducibility.
In this case, PLS is successfully used to detect the level of adulterant and
discriminant analysis allows one to make a classification of cod-liver oil and
potential adulterants of animal oils.
Figure 3a shows the data of FTIR spectra for different samples before data
analysis was applied. Through naked eye, it is challenging to analyze, let alone to
be able to detect the lard adulterated samples. A Coomans plot after discriminant
analysis is applied in Fig. 3b showing the clear separation of lard adulterated with
the original samples, thus enhance the detection ability. Multiple data analysis will
not only enhance detection by developing a regression model but can also provide
different information about the data collected from measurements like the accuracy,
repeatability, significance difference among the samples by hypothesis testing and
many more [61].
2 Dielectric Spectroscopy
A dielectric material is materials that are neither conductors nor insulators that can
be polarized by an applied electric field [62]. A dielectric material does not have
loosely bound electrons like a conductor material. When a dielectric material is
Lard Detection in Edible Oil Using Dielectric Spectroscopy 253
placed in an electric field, electric charges in the material slightly shift from their
average equilibrium position. This cause a phenomenon called polarization, in
which the positive charges was displaced towards the electric field, while negative
charges shift on the opposite direction as illustrated in Fig. 4 [63]. The polarization
creates internal electric field that reduces the overall field within and increases the
capacitance. Dielectric spectroscopy or spectral permittivity is the measure of the
ability of the material to resist (resistance) or store (capacitance) the electric field
created [64]. Dielectric constant also known as relative permittivity is the permit-
tivity of the material in relative to that of free space [65, 66].
According to [68], the dielectric parameter (Ɛ) can be calculated from the fol-
lowing equation
Cp 1
εr = −j ð1Þ
C0 ωC0 Rp
where, Ɛ′r is the relative dielectric constant and Ɛ′′r is the dielectric loss.
254 M.A. Sairin et al.
Thus, dielectric constant can be defined as the ratio between the capacitance of
the dielectric material, Cp, with the capacitance of that in free space or vacuum, C0,
as shown in Eq. (3) below:
Cp
εr = ð3Þ
C0
Temperature dependence also relates with the polarization of the molecules with the
electromagnetic field in terms of viscosity and water phases of the food material. At
low temperature, high viscosity leads to longer relaxation time and little dipole
rotation and thus results in low dielectric constant [69, 92]. More on this will be
discussed in the next section.
While in food materials with high water content, the dielectric constant highly
dependent on the phase of the water in the food. Figure 5 illustrate the temperature
dependence of dielectric value for various food materials from freezing to melting
state [93]. The dielectric value of all the food materials experience abrupt change
from very low to very high at around the melting point temperature of –5 to 0 °C,
suggesting that no free water molecule is present during freezing and otherwise
during melting. As temperature increases, the dielectric constant reduces following
dielectric profile of water.
Fig. 5 Temperature
dependence of dielectric value
for various food substance at
2.8 GHz [93]
Dielectric spectroscopy has been applied to variety of agriculture and food prod-
ucts. The understanding of dielectric properties of food products becomes
increasingly important for consistent and predictable product, process and equip-
ment development [69]. Dielectric spectroscopy has been used to analyze and
characterize fruits and vegetables, meat and seafood, milk and dairy products, oil
and many more.
In [84], a study on the frequency dependence of the dielectric constant and loss
factor of fresh fruits using a network analyzer in the range from 200 MHz to
20 GHz was conducted. The study also explored the possibility of sensing the
maturity of fresh peaches. Investigation in [85] suggests that higher moisture
content in fruit results in slightly higher dielectric constant, and decreased in
dielectric constant with the increased of temperature or frequencies, for most fruits.
In [95], the permittivity and conductivity of various beef and chicken were
measured within the frequency range of 1 Hz–1 MHz. The work observes the
differences in fresh raw meat of commercially purchased with that after being
cooked in microwave. Raw meat showed higher dielectric constant measurements
than cooked meat, potentially due to high moisture content in raw meat.
A study on milk and dairy product in [96], investigated the dielectric properties
of UHT skim, low fat and homogenized whole milk. The study concludes that the
dielectric constant for all types of milk are close together in the high frequency
region of 1–10 GHz, and display larger differences at lower frequency range.
Figure 7 illustrate the dielectric constant of whole milk, low fat milk, skim milk and
water at frequency range from 1–10 GHz. It can be observed that water gives the
Fig. 7 Dielectric constant profile for whole milk, low fat milk, skim milk and comparison with
water [96]
258 M.A. Sairin et al.
Several works have been reported on the usage of dielectric spectroscopy to study
the dielectric properties of oils and fats. A research group from Japan investigated
on discriminating olive oil adulterated with vegetable oil in the frequency range of
101 Hz–1 MHz [97]. The olive oil is spiked with other vegetable oils, soy, corn,
canola, sesame and perilla oils with different concentration from 0 to 100 %.
Figure 8 shows a PCA score plot for soy oil and perilla oil adulterated in olive oil,
respectively. Leveraging on data analysis technique, the adulterated oil can be
easily detected from the original oil as can be observed in Fig. 8.
A study edible oils and fatty acids as a function of frequency, temperature,
moisture and composition has been reported in [92]. In this study, 10 edible oils and
6 fatty acids were used and the dielectric properties were measured over the fre-
quency range 100 Hz–1 MHz. The effects of temperature (20–45 °C), moisture
content (0.02–0.31 %), and fatty acid component on dielectric properties of oils
were investigated. From the investigation, it can be concluded that the dielectric
constant of oils affected by their fatty acid composition significantly, the dielectric
constant of the oils increased with increasing moisture content, this effect dimin-
ished as frequency increases, and both dielectric constant and loss factor of edible
oils demonstrated a gradual decrease with increasing temperature, as well as the
dielectric constant of fatty acids.
Fig. 8 PCA score plot for different concentration of soy oil in olive oil (left) and PCA score plot
for different concentration of perilla oil in olive oil [97]
Lard Detection in Edible Oil Using Dielectric Spectroscopy 259
In most oil and fats, different frequencies result in little changes in the dielectric
constant. Figure 10 shows the dielectric values dependency on frequencies for
different fatty acids (i.e. caprylic, linolenic, linoleic, oleic, palmitic and stearic) and
also of a few vegetable oils (i.e., sesame, soy, corn, canola, olive, safflower and
Table 3 Summary of previous studies on dielectric spectroscopy for oils and fats
Study Sample Frequency
Bansal (2001) Rapeseed-mustard oil 100 kHz–8.93 GHz
Inoue (2002) Soybean oil 1 kHz–100 kHz,
50 kHz–30 MHz
Lizhi (2008) Vegetable oil & fatty acids 100 Hz–1 MHz
Cataldo (2009) Vegetable oil 100 kHz–1 GHz
Lizhi (2010) Olive oil adulterated with other vegetable oil 100 Hz–1 MHz
260 M.A. Sairin et al.
Fig. 10 Spectra of dielectric constant of saturated and unsaturated fatty acids at 75 °C (left) and of
various oils at 25 °C (right) [92]
sunflower) [92]. It is explained that the dielectric spectra indicated general plateau
and maximum value in frequency range between 100 Hz and 500 kHz suggesting
that there exist equilibrium among the orientation of oil molecule and the electric
field at low frequencies. At higher frequency, the dielectric spectrum is observed to
decrease gradually. This is explained in [69] that at higher frequency, the dipoles
are not reacting to the electric field as they are unable to follow the rapid field
reversal.
Fig. 11 Effect of temperature on dielectric constant (left) and loss factor (right) of soybean oil
[92]
Lard Detection in Edible Oil Using Dielectric Spectroscopy 261
Work in [92] also studied the effect of moisture on dielectric constant. The oil
moisture content ranges from 0.02 to 0.31 % (w/w), was determined in accordance
with Air Oven Method 1.4.1-2013 [100]. Figure 12 shows the effect of moisture
content on dielectric constant of corn oil. As discussed earlier, water is a polar
molecule with higher dielectric constant than water, thus lead to increase of
dielectric constant with increasing moisture content [69].
The major animal fats, also termed meat fats, traded world-wide are produced from
pigs (Sus scrofa), termed as lard or rendered pork fat, from cattle (Bos Taurus) or
sheep (Ovis aries) termed tallow, and from poultry, primarily chickens (Gallus
gallus) termed as poultry fat. Tallow from domestic cattle is known as beef tallow,
whereas that from sheep is termed mutton tallow. “Acylglycerol” is the primary
component of animal lipids, which may exist as either fat or oils. At room tem-
perature, 24 °C, the former are primarily solid and the latter are liquid [16].
Lard is the fat rendered from fresh, uncontaminated, healthy fatty tissues from
the meat of sound swine (Sus scrofa) at the time of slaughter, and hence, fit for
human consumption. It is also imperative that the fat is pure and without the
presence of bones, isolated skin, blood or any other viscera. Edible tallow (drip-
ping), on the other hand, is the product obtained by rendering the clean, sound fatty
tissues, including trimming and cutting fat, attendant muscles and bones of bovine
and/or sheep in good health at the time of slaughter and fit for human consumption
[101].
In our recent study, sample of animal fats which are lard, beef tallow, mutton
tallow and chicken fat were extracted by rendering of adipose tissue of animals
collected from local slaughter houses and market according to the method proposed
by [37]. The rendering was done at 90–100 °C for two hours in the oven. The
extracted fats was filtered through double-folded muslin cloth and dried over
anhydrous sodium sulphate (Na2SO4) to remove residual moisture through filter
paper. The filtered samples were stored in a freezer at –20 °C before further
analysis. Before being used for analysis, all animal fats were thawed at 60 °C until
they melted. As some of the animal fats are in solid form at room temperature, a
solvent is needed for uniform sampling. Hexane is used as a standard oil-based
solvent. Polar solvent dissolve polar compound, while non-polar solvents dissolve
non-polar compounds [102]. As oil and fats are non-polar compounds, non-polar
solvents are needed [103]. Common extraction solvents include hexane, methylene
chloride, methy isobuthyl ketone (MIBK) and xylene [102]. Thus, the concentra-
tions mentioned in the discussion of dielectric spectroscopy would indicate the
concentration of each fat and oil in hexane.
Two works has been done for the lard detection in animal fats using dielectric
spectroscopy. The first work utilizes the higher frequency region, 200 MHz–
20 GHz [104], while the second study utilizes lower region frequency, 100 Hz–
100 kHz [105].
Fig. 13 Dielectric constant of individual fats; a beef tallow, b chicken fat, c lard, d palm oil and
e soy bean oil [104]
Figure 14 shows the comparison on the dielectric constant and loss factor between
beef tallow sample, lard sample and beef adulterated with lard sample. It can be
observed that frequency between 8.2–14.2 GHz shows most clear separation of the
dielectric constant between beef tallow and samples that contain lard, pure lard and
beef adulterated with lard. While the loss factor shows separation in different fre-
quency range, which is between 10.2–2.2 GHz.
It is concluded that the dielectric constant of all individual fats samples
decreased as the frequency increased for all concentration. Lard adulterated sample
can be distinguished from non-adulterated samples by its dielectric constant and
loss factor spectral pattern at certain frequency range.
264 M.A. Sairin et al.
Fig. 14 Comparison of a dielectric constant and b loss factor between beef tallow sample, lard
sample and beef adulterated with lard sample [104]
The study of the animal fats at low frequency was conducted at room temperature as
well. Animal fats used in this study were beef tallow, chicken fat, mutton tallow and
lard. An Agilent 16452A liquid dielectric test fixture was used as the probe and
Fig. 15 RSD values for dielectric measurements at frequency range of 10 Hz–100 kHz for
different animal fats: a chicken, b lard, c beef and d mutton [105]
Lard Detection in Edible Oil Using Dielectric Spectroscopy 265
connected to an Agilent 4263B LCR meter. The dielectric values were measured at
six discreet frequencies; 100 Hz, 120 Hz, 1 kHz, 10 kHz, 20 kHz and 100 kHz.
This study utilizes data analysis of Relative Standard Deviation (RSD),
ANOVA, and PCA. Figure 15 illustrate the RSD values obtained from all of the
animal fats.
It can observed that the RSD values of all samples are well below 1.8 %, which
is very good for precision, as the minimum RSD required for normal analysis is
around 10 % [106]. Table 4 shows the result from ANOVA analysis, which shows
that at every frequency, the p-value is less than 0.05 indicating that different animal
fat samples have significant difference at these frequencies. Figure 16 shows a PCA
score plot for four different types of animal fats; i.e., beef tallow, mutton tallow,
lard and chicken fat. From the observation on the PCA score plot, it is clearly
depicted a well separation between different types of animal.
From this study, it is concluded that the dielectrics value is not a function of
frequencies but a function of sample’s concentration levels. Using ANOVA, it is
statistically concluded that there is significant difference between different groups of
Fig. 16 PCA score plot for different types of animal fats in which B (beef tallow), M (mutton
tallow), L (lard) and C (chicken fat) can be observe was well separated between each other [105]
266 M.A. Sairin et al.
sample, i.e., animal fats. PCA technique positively show clear group separation
among different animal fats, thus highlight the ability of dielectric spectroscopy in
lard detection.
4 Conclusion
This chapter has reviewed the motivation of detecting lard in lard adulterated food
product such as edible oil, several conventional techniques that have been used in
lard detection and a number of data analysis that have been applied on lard
detection study. The factors influencing dielectric spectroscopy technique and use
of dielectric spectroscopy on food products have been discussed. In addition to the
applications presented on food products, the dielectric spectroscopy measurements
on oil and fats was also discussed in detail with the factors influencing the dielectric
constant and loss factor. Work has also been performed to distinguish lard from
other animal fats by using the dielectric spectroscopy method with data analysis to
help enhance lard detection ability.
References
1. J. Spink, D.C. Moyer, Background: defining the public health threat of food fraud. J. Food
Sci. 76(9), 1–7 (2011)
2. R. Johnson, Food Fraud and Economically Motivated Adulteration of Food and Food
Ingredients, Congr. Res. Serv. Rep., R43358, 1–40 (2014)
3. V. Lakshmi, Food adulteration. Int. J. Sci. Invent. Today 56(2), 72–77 (2012)
4. S. Lauterbach ,J.A. Albrecht, NF94-186 functions of baking ingredients, Hist. Mater. from
Univ. Nebraska-Lincoln Ext., (1994)
5. R.V. Rios, M. Durigan, F. Pessanha, P.F. De Almeida, C.L. Viana, S. Caetano, Application
of fats in some food products. Food Sci. Technol. 34(1), 3–15 (2014)
6. Vegetable Oils and Oilseeds: A Review (The Commonwealth Secretariat, London, 1968)
7. C.J. Forde, M. Meaney, J.B. Carrigan, C. Mills, S. Boland, A. Hernon, Biobased fats
(Lipids) and oils from biomass as a source of bioenergy, in Bio-energy Research: Advances
and Applications, ed. by V.K. Gupta, C.P. Kubichek, J. Saddler, F. Xu, M.G. Tuohy
(Elsevier, Amsterdam, 2014), pp. 185–201
8. F.D. Gunstone, Major edible uses of oils and fats, in Oils and Fats in the Food Industry
(Wiley-Blackwell, 2008), pp. 113–135
9. P. He, X. Wan, C. Wang, Y. Jiao, Determination of Animal Oil Added in Vegetable Oil By
Standard Chemical Method Coupled With Image Texture Analysis Technology. Int.
J. Innov. Comput. Innnov. Control 10(1), 67–80 (2014)
10. Y.B. Che Man, Z.A. Syahariza, M.E.S. Mirghani, S. Jinap, J. Bakar, Analysis of potential
lard adulteration in chocolate and chocolate products using Fourier transform infrared
spectroscopy, Food Chem. 90(4), 815–819 (2005)
11. N. Blechman, Extra virgin suicide—the adulteration of italian olive oil (2014). http://www.
nytimes.com/interactive/2014/01/24/opnion/food-chains-extra-virgin-suicide.html
12. N.H. Jenkins, Virgin Territory—Exploring the World of Olive Oil (Houghton Mifflin
Harcourt, New York, 2015)
Lard Detection in Edible Oil Using Dielectric Spectroscopy 267
Saturated, Polyunsaturated, and Trans Fat on Global Burdens of Coronary Heart Disease.
J. Am. Heart Assoc. 5(1), e002891 (2016)
32. A.A. Aida, Y.B.C. Man, C.M.V.L. Wong, A.R. Raha, R. Son, Analysis of raw meats and
fats of pigs using polymerase chain reaction for Halal authentication. Meat Sci. 69(1), 47–52
(2005)
33. Y.B. Che Man, A.A. Aida, A.R. Raha, R. Son, Identification of pork derivatives in food
products by species-specific polymerase chain reaction (PCR) for halal verification, Food
Control, 18(7), 885–889 (2007)
34. J. Sneddon, D.J. Butcher, Advances in atomic spectroscopy, 1st edn. ( Elsevier Science,
2002)
35. J.M. Hollas, Modern Spectroscopy. (John Wiley & Sons, Inc., 2004)
36. G. Hohne, W.F. Hemmingher, H.J. Flammersheim, Differential Scanning Calorimetry
(Springer-Verlag, Berlin Heidelberg, New York, 2003)
37. J.M.N. Marikkar, O.M. Lai, H.M. Ghazali, Y.B. Che Man, Detection of lard and randomized
lard as adulterants in refined-bleached-deodorized palm oil by differential scanning
calorimetry. J. Am. Oil Chem. Soc. 78(11), 1113–1119 (2001)
38. E. Coni, M. Di Pasquale, P. Coppolelli, A. Bocca, Detection of animal fats in butter by
differential scanning calorimetry: A pilot study. J. Am. Oil Chem. Soc. 71(8), 807–810
(1994)
39. O. Dahimi, A.A. Rahim, S.M. Abdulkarim, M.S. Hassan, S.B.T.Z. Hashari, A. Siti
Mashitoh, S. Saadi, Multivariate statistical analysis treatment of DSC thermal properties for
animal fat adulteration. Food Chem. 158, 132–138 (2014)
40. R.L. Wixom, C.W. Gehrke, V.G. Berezkin, J. Janak, Chromatography: A New Discipline of
Science, in Chromatography : A science of discovery, vol. 1, ed. by R.L. Wixom,
C.W. Gehrke, (John Wiley & Sons, Inc., Hoboken, NJ, USA, 2015)
41. H. Alam, S.H. Saeed, C. Engg, Electronic nose in food and healh applications: A review. Int.
J. Comput. Corp. Res. 2(6), 1–17 (2012)
42. M. Nurjuliana, Y.B. Che Man, D. Mat Hashim, Analysis of lard’s aroma by an electronic
nose for rapid Halal authentication, JAOCS. J. Am. Oil Chem. Soc. 88(1), 75–82 (2011)
43. N.N. Ahmad Nizar, J.M. Nazrim Marikkar, D.M. Hashim, Differentiation of lard, chicken
fat, beef fat and mutton fat by GCMS and EA-IRMS techniques. J Oleo Sci, 62(7), 459–464
(2013)
44. O. Dahimi, M.S. Hassan, A.A. Rahimm, S.M. Abdulkarim, Differentiation of lard from other
edible fats by Gas Chromatography-Flame Ionisation Detector (GC-FID) and Chemometrics,
J. Food Pharm. Sci. 2(2), 27–31 (2014)
45. D. Indrasti, Y.B. Che Man, S. Mustafa, D.M. Hashim, Lard detection based on fatty acids
profile using comprehensive gas chromatography hyphenated with time-of-flight mass
spectrometry. Food Chem. 122(4), 1273–1277 (2010)
46. J.M.N. Marikkar, H.M. Ghazali, Y.B. Che Man, T.S.G. Peiris, O.M. Lai, Distinguishing lard
from other animal fats in admixtures of some vegetable oils using liquid chromatographic
data coupled with multivariate data analysis. Food Chem. 91(1), 5–14 (2005)
47. A. Rohman, E. Sismindari, Y. Erwanto, Y.B. Che Man, Analysis of pork adulteration in beef
meatball using Fourier transform infrared (FTIR) spectroscopy, Meat Sci. 88(1), 91–95
(2011)
48. A. Rohman, Y.B. Che Man, Analysis of Cod-Liver Oil Adulteration Using Fourier
Transform Infrared (FTIR) Spectroscopy. J. Am. Oil Chem. Soc. 86(12), 1149–1153 (2009)
49. Z.A. Syahariza, Y.B. Che Man, J. Selamat, J. Bakar, Detection of lard adulteration in cake
formulation by Fourier transform infrared (FTIR) spectroscopy. Food Chem. 92(2), 365–371
(2005)
50. I. Jaswir, M.E.S. Mirghani, T.H. Hassan, M.Z.M. Said, Determination of Lard in Mixture of
Body Fats of Mutton and and Cow by Fourier Transform Infrared Spectroscopy. J. Oleo Sci.
52(12), 633–638 (2003)
Lard Detection in Edible Oil Using Dielectric Spectroscopy 269
51. Y.B. Che Man, M.E.S. Mirghani, Detection of lard mixed with body fats of chicken, lamb,
and cow by fourier transform infrared spectroscopy. J. Am. Oil Chem. Soc. 78(7), 753–761
(2001)
52. R. Peng, Analysis of Variance, in Statistical Procedures (2009), pp. 327–374
53. M.H. Katz, Multivariable analysis: a practical guide for clinicians (New York, 2011)
54. R. Shier, Paired t-tests. (Mathematics Learning Support Centre, 2004), pp. 1–3
55. J. Poulsen, A. French, Discriminant Function Analysis (DA). J. Forensic Sci. 56(2), 297–301
(1996)
56. L.I. Smith, A tutorial on principal components analysis introduction, Statistics (Ber), 51, 52
(2002)
57. P.-N. Tan, M. Steinbach, V. Kumar, Chap 8: cluster analysis: basic concepts and algorithms,
in Introduction to Data Mining (2005), p. Chapter 8
58. D.M. Lane, D. Scott, M. Hebl, R. Guerra, D. Osherson, H. Zimmer, Introduction to Statistics
(1974)
59. R.D. Tobias, An introduction to partial least squares regression, in Proceedings. Annual SAS
Users Group International Conference, 20th, (Orlando, FL, 1995), pp. 2–5
60. SPSS: Stepwise linear regression, School of Geography, University of Leeds. http://www.
geog.leeds.ac.uk/courses/other/statistics/spss/stepwise
61. S. Nielsen, Food analysis laboratory manual. J. Chem. Inf. Model. 53, 160 (1989)
62. R. Simpson, Engineering aspects of thermal food processing (Taylor & Francis Group, Boca
Raton, 2009)
63. C.G. Pearce, Polarization of Dielectrics, EM Fields 2, 196AD. http://www-users.aston.ac.uk/
∼pearcecg/Teaching/PDF/LEC2.PDF
64. J. Mijovic, B.D. Fitz, Dielectric spectroscopy of reactive polymers, Mater. Sci. 2(5),
pp. 1531–9 (1998)
65. S.E. Braslavsky, Glossary of terms used in photochemistry, 3rd edition. Pure Appl. Chem.
79(3), 293–465 (2007)
66. Wave Propagation Standards Committee: IEEE Standard Definitions of Terms for Radio
Wave Propagation, vol. 1997 (1997)
67. R. Nave, Dielectrics, Hyperphysics. http://www.hyperphysics.phy-astr.gsu.edu/hbase/hph.
html
68. Agilent 16452A Liquid Test Fixture Operation and Service Manual (Japan, 2007)
69. A.K. Datta, G. Sumnu, G.S.V. Raghavan, Dielectric properties of foods, in Engineering
Properties of Foods, 3rd edn., ed. by M.A. Rao, S.S.H. Rizvi, A.K. Datta (Taylor & Francis
Group, Boca Raton, 2005), pp. 501–566
70. A.R. Von Hippel, Dielectrics and Waves (Cambridge, Mass., MIT. Press, Cambridge, 1954)
71. A.R. Von Hippel, Dielectric Materials and Applications (The Technology Press of MIT and
John Wiley, New York, 1954)
72. S.O. Nelson, L.F. Charity, Frequency dependence of energy absorption by insects and grain
in electric fields. Trans. Am. Soc. Agric. Eng. 15, 1099–1102 (1972)
73. S. Nelson, L.E. Stetson, C.W. Schlaphoff, A general computer program for precise
calculation of dielectric properties from short-circuited waveguide measurements, IEEE
Trans. Instrum. Meas. IM–23(4), 455–460 (1974)
74. C. Gabriel, E.H. Grant, I.R. Young, Use of time domain spectroscopy for measuring
dielectric properties with a coaxial probe. J. Phys. E. 19(10), 843–846 (1986)
75. Y. Wang, T.D. Wig, J. Tang, L.M. Hallberg, Dielectric properties of foods relevant to RF
and microwave pasteurization and sterilization. J. Food Eng. 57(3), 257–268 (2003)
76. S. Mashimo, S. Kuwabara, S. Yagihara, K. Higasi, Dielectric relaxation time and structure of
bound water in biological materials. J. Phys. Chem. 91(25), 6337–6338 (1987)
77. A.W. Kraswzewski, S. Trabelsi, S.O. Nelson, Microwave dielectric properties of wheat, in
Proceedings of 30th Microwave Power Symposium (1995), pp. 90–93
78. R.B. Keam, W.S. Holmes, Uncertainty analysis measurement of complex permittivity using
microstrip transmission line, in SBMO/IEEE MTT-S IMOC’95 Proceedings (1995), pp. 137–
142
270 M.A. Sairin et al.
79. F.M. Ghannouchi, R.G. Bosisio, Measurement of microwave permittivity using a six-port
reflectometer with an open-ended coaxial line. IEEE Trans. Instrum. Meas. 38(2), 505–508
(1989)
80. H. Wakamatsu, A Dielectric Spectrometer for Liquid Using the Electromagnetic Induction,
Hewlett-Packard J. 1997, 1–10 (1997)
81. J.L. Jorgensen, A.R. Edison, S.O. Nelson, L.E. Stetson, A bridge method for dielectric
measurements of grain and seed in the 50–250 MHz range. Trans. ASAE 13(1), 18–20
(1970)
82. A.C. Metaxas, R.J. Meredith, Industrial Microwave Heating (Peter Peregrinus Ltd., London,
1983)
83. S.O. Nelson, Dielectric spectroscopy in agriculture, J. Non. Cryst. Solids, 351(33–36) SPEC.
ISS., 2940–2944, (2005)
84. S.O. Nelson, Dielectric properties of agricultural products-measurements and applications.
IEEE Trans. Instrum. Meas. 41(1), 385–387 (1991)
85. M.E. Sosa-Morales, L. Valerio-Junco, A.L. López-Malo , H.S. García, Dielectric properties
of foods: Reported data in the 21st century and their potential applications, LWT-Food Sci.
Technol. 43(8), 1169–1179 (2010)
86. C. Gabriel, C. Gabriel, S. Gabriel, S. Gabriel, E. Corthout, E. Corthout, The dielectric
properties of biological tissues: I. Literature survey. Phys. Med. Biol. 41(11), 2231–2249
(1996)
87. R. Hollertz, Dielectric properties of wood fibre components relevant for electrical insulation
applications (KTH Royal Institute of Technology, Stockholm, Sweden, 2014)
88. F. Liu, I. Turner, E. Siores, P. Groombridge, A numerical and experimental investigation of
the microwave heating of polymer materials inside a ridge waveguide.pdf. J. Microw. Power
Electromagn. Energy. 31(2), 71–82 (1996)
89. M. Arai, J.G.P. Binner, G.E. Carr, T.E. Cross, High temperature dielectric measurements on
ceramics, in Sixth International Conference on Dielectric Materials, Measurements and
Applications, 1992, (IET, Manchester, 1992), pp. 69–72
90. P. Hoekstra, A. Delaney, Dielectric properties of soils at UHF and microwave frequencies.
J. Geophys. Res. 79(11), 1699–1708 (1974)
91. S. Ryyniinen, The electromagnetic properties of food materials: a review of the basic
principles. J. Food Eng. 26(4), 409–429 (1995)
92. H. Lizhi, K. Toyoda, I. Ihara, Dielectric properties of edible oils and fatty acids as a function
of frequency, temperature, moisture and composition. J. Food Eng. 88(2), 151–158 (2008)
93. N. Bengtsson, P. Risman, Dielectric properties of foods at 3 ghz as determined by a cavity
perturbation technique–II measurements on food materials. J. Microw. Power 6(2), 107–123
(1971)
94. R.E. Mudgett, S.A. Goldblith, D.I.C. Wang, W.B. Westphal, Dielectric behavior of a
semisolid food at low, intermediate and high moisture content.pdf. J. Microw. Power 15(1),
27–36 (1980)
95. B. Bodakian, F.X. Hart, The dielectric properties of meat. IEEE Trans. Dielectr. Electr.
Insul. 1(2), 181–187 (1994)
96. A.C. Nunes, X. Bohigas, J. Tejada, Dielectric study of milk for frequencies between 1 and
20 GHz. J. Food Eng. 76(2), 250–255 (2006)
97. H. Lizhi, K. Toyoda, I. Ihara, Discrimination of olive oil adulterated with vegetable oils
using dielectric spectroscopy. J. Food Eng. 96(2), 167–171 (2010)
98. A.K. Bansal, P.J. Singh, K.S. Sharma, S. Kumar, P.R. Kumar, Dielectric properties of
different varieties of rapeseed-mustard oil at different temperatures. Indian J. Pure Appl.
Phys. 39, 532–540 (2001)
99. A. Cataldo, E. Piuzzi, G. Cannazza, E. De Benedetto, L. Tarricone, On the use of dielectric
spectroscopy for quality control of vegetable oils, in 19th IMEKO World Congr. 2009, vol.
1 (2009), pp. 1–5
100. The JOCS Standard Methods for the Analysis of Fats, Oils and Related Materials Second
English Edition. Japan Oil Chemists’ Society (2013)
Lard Detection in Edible Oil Using Dielectric Spectroscopy 271
101. Codex Standard for Named Animal Fats (CODEX-STAN 211–1999), in Codex Alimentarius
Commission: Fats, Oils & Related Products (Secretariat of the Joint FAO/WHO Food
Standards Programme, Rome, 2001)
102. E.M.S.F., G.M.F.I.I. James W Robinson, Concepts of Instrumental Analytical Chemistry, in
Undergraduate Instrumental Analysis, 6th ed., New York: Taylor & Francis Group, 2005,
pp. 1–1107
103. R.J. Stokes, D.F. Evans, Fundamentals of Interfacial Engineering (Wiley Inc, New York,
1997)
104. N.H.A. Latiff, Lard detection using dielectric spectroscopy (Universiti Putra Malaysia, 2010)
105. M.A. Sairin, N. Naquiah, A. Nizar, S.A. Aziz, D.M. Hashim, A. Samples, Potential of
Dielectric Spectroscopy Measurement for Lard Detection, pp. 1–6
106. V. Thomsen, D. Schatzlein, D. Mercuro, Limits of detection in spectroscopy. Pure Appl.
Chem. 18(12), 112–114 (2003)
Optical-Based Interference Cancellation
in Wireless Sensor Networks
1 Introduction
Wireless sensor networks can provide valuable insight into the patterns and processes
of our surroundings that we were formerly unaware of. Yet, connecting thousands or
millions of wireless devices in a seamless network brings a unique set of challenges
that must first be overcome before this promise can be fulfilled. Chief among them
are the problems of wireless interference and RF spectrum crowding.
Wireless interference is the greatest problem in wireless communications because
of the very nature of our shared transmission medium. Interference mitigation tech-
niques involve sharing other radio resources such as time (time-division duplexing,
TDD), codes (code-division duplexing, CDD), and frequency (frequency-division
duplexing, FDD). These techniques have successfully permitted multiple users to
communicate without interfering with each other, but they come at a cost: RF spec-
trum, the universal currency of wireless communications. As data rates and the num-
ber of connected devices continue to rise exponentially [1], the limited RF spectrum
has grown extremely crowded, and the price for bandwidth worldwide has soared
into the billions of U.S. dollars for 10s of MHz of spectrum. Similarly, the number
of communication bands and the market for RF front-end components, particularly
filters, the gateways to different bands, has followed suite (see Fig. 1).
Adding more bands and filters may seem like a solution, but this trend is not sus-
tainable. Most RF components, like filters, operate in a single, fixed frequency band,
meaning each band requires a dedicated chain of RF components. As a result, RF
front-ends, like those in Fig. 2, have grown complex, bulky, and expensive. Mean-
while, limited circuit board real estate (and profit margins) in size-sensitive plat-
forms like handsets mean that hardware designers have to sacrifice functionality for
the expanding RF front-end. Finally, no amount of additional filters actually creates
capacity, it only divides it up further.
New methods of operating in the presence of interference will be required in
widescale wireless sensor networks, as well as in the larger scope of 5G. These
methods will need to work in tandem with the traditional TDD, FDD, and CDD
techniques. Furthermore, any new hardware must be tunable and wideband, lest they
be vulnerable to the same scaling limitations of existing front-ends and risk facing
irrelevance in 5G.
Fig. 1 a The projected growth of revenue for RF front-end components, dominated by filters.
Taken from [2]. b The rise in the number of different band filters. Source Triquint
Optical-Based Interference Cancellation in Wireless Sensor Networks 275
Fig. 2 RF front-end architecture for the 700–900 MHz bands, showing the front-end complexity
Fig. 3 Different forms of interference mitigation techniques: (left) cancellation, (middle) spatial
isolation, (right) spectral isolation or filtering
Optical-Based Interference Cancellation in Wireless Sensor Networks 277
The ability of a system to cancel interference depends strongly upon the source of the
interference, the coupling channel between the interference source and the receiver,
herein called the interference channel, and any apriori knowledge about the inter-
ference and the desired signal. In a general interference cancellation scenario, the
received signal consists of a signal of interest, s(t), corrupted by an unknown interfer-
ence signal, n(t), which is assumed to be strong enough to block the receiver (Eq. 1).
The purpose of the interference cancellation system is to extract the interference from
the aggregate signal and subsequently subtract the extracted interference, np (t), from
the aggregate signal, recovering the signal of interest. The output is given by Eq. 2,
where the term in brackets represents the residual interference.
Apriori knowledge can consist of anything from the waveform of the interference to
its statistical nature and will in general dictate the signal processing that is required
to extract the interference from the aggregate signal.
Interference can be divided into two categories: known and unknown. When
the interference signal is unknown, statistical methods must be used to distinguish
the interference from the desired signal. This is known as blind source separation.
Assuming that the signal of interest and the interference are statistically independent,
techniques such as principal component analysis (PCA) and independent component
analysis (ICA) are often employed. These techniques aim to separate the aggregate
signal into two or more components, demarcated by their statistical properties. An
interference canceller would then eliminate the components related to the interfer-
ence. Much research has been devoted to the topic of blind source separation, and
we refer the readers to these sources [3–5]. Herein, we focus on the situation where
the interference signal is either known or has been successfully extracted.
278 M.P. Chang et al.
2.2 Self-interference
(d) (e)
nal, but more complex channel effects, such as multipath, reflections within the RF
circuitry, amplifier noise, and nonlinear distortion are also present [8, 9]. The self-
interference at the receiver can be given by
∞
∑
n(t) = (𝛼n n0 (t − 𝜏n )) + w(t) + nnl (t) (3)
n=1
where n0 (t) represents the original interferer at the transmitter, 𝛼n and 𝜏n are the
channel attenuation and delay for the nth multipath component, w(t) represents wide-
band amplifier noise, and nnl (t) represents all nonlinear byproducts. The ability of
the system to estimate and mimic these effects determines the maximum amount of
interference cancellation achievable.
In this section, we give a brief introduction to analog optical signal processing and
make a case for using optics to perform interference cancellation based on its ability
to satisfy the four key characteristics of an interference canceller:
∙ Wideband operation
∙ Precise amplitude and delay (phase) control
∙ Low latency tunability
∙ High dynamic range
Despite the omnipresence of digital electronics in our world, there has recently been a
strong resurgence of interest in RF analog signal processing. This is because convert-
ing an analog RF signal into a digital signal with high fidelity is extremely demand-
ing. It requires a large number of bits to achieve high dynamic range and a large
sampling rate to represent high bandwidths. These tall technical requirements make
it cost prohibitive, if not impossible, to implement a truly software-defined radio
(SDR) today.
Following the development of the laser and the optical fiber in the late 1960s,
researchers began investigating the potential of optics to perform high-fidelity trans-
mission, distribution, and processing of analog RF signals [10]. In this process, the
RF signal is first modulated onto an optical carrier (historically 1310 and 1550 nm
because of propagation characteristics in optical fiber), which was generated by a
laser or an LED. The modulated optical carrier is transmitted through an extremely
low-loss optical fiber over long distances to a destination, where it is converted back
into an RF signal by a photodetector. The goal was to capture the natural advantages
of optical fiber, namely very low propagation loss and RF frequency-independent
propagation (i.e. wide bandwidths). Today, optical fiber exhibits losses as low as
0.5 dB/km and carries aggregate data rates over 1 Tbps; in contrast, RF cables have
losses of 360 dB/km at 2 GHz [11] and gets even lossier at higher frequencies. Mil-
itary and other weight sensitive platforms also sought lightweight optical fibers to
replace heavy RF cables (1.7 kg/km vs. 80–600 kg/km) [12].
Soon after data transmission using fiber-optics had proven successful, researchers
turned towards using optics to perform actual RF signal processing functions (see.
Fig. 7). In 1991, the term microwave photonics was coined [13] to describe the
general field of analog RF signal processing, distribution, and generation using
optics and photonics, and has since continued to captivate the research community.
Microwave photonics seeks to leverage the physics of optics to achieve functional-
282 M.P. Chang et al.
Fig. 7 A general analog optical link, showing RF input and output, and optical signal processing
and distribution in the middle
ity and performance that is impossible to obtain using RF components. These key
performance advantages include high dynamic range, low loss fiber, low suscepti-
bility to electromagnetic interference, ease of reconfigurability, and perhaps most
importantly wide bandwidth [11, 14–16].
The performance of a microwave photonic link, generically shown in Fig. 7, is
determined by three main parts: the electrical-optical (EO) conversion, the interme-
diate optical signal processing, and the optical-electrical (OE) conversion. To speak
a common language with RF engineers, microwave photonic links are often char-
acterized using traditional RF figures of merit: link loss, noise figure, and dynamic
range. Because the EO and OE conversions determine the fidelity of the signal as it
transitions between the electrical and optical domains, the modulators and detector
devices that perform this conversion have a dominant effect on the link loss (or gain),
noise figure, and dynamic range. Over the years, microwave photonic links have
shown consistent improvement in all three figures of merit and commercial products
are now on the market [17–19]. The intermediate optical signal processing archi-
tecture determines the actual photonic circuit function. Microwave photonic imple-
mentations of FIR (or matched) filters [20–22], beamformers [22–25], wideband
phase shifters [21, 26, 27], true-time delays [16, 21, 28], and interference cancellers
[7, 29–33] have all been demonstrated.
The wide bandwidth of optics is one of the primary reasons why optics is the de facto
carrier of the Internet, and these benefits apply to wideband interference cancella-
tion as well. Optics is able to process extremely wide bandwidth signals because
Optical-Based Interference Cancellation in Wireless Sensor Networks 283
Fig. 8 A standard RF signal is essentially narrowband when modulated onto an optical carrier
To effect amplitude control, mirror and semiconductor based devices with tun-
able gain or insertion loss are commonly used. Commercial mirror-based variable
optical attenuators (VOAs) can achieve an RF attenuation resolution better than
0.01 dB/mV over a range of 30 dB, which is sufficient to obtain 50 dB of cancel-
lation given sufficient phase matching. Semiconductor optical amplifiers (SOA) and
electroabsorbers use a bias current or voltage to change a semiconductor’s optical
gain or absorption coefficient, respectively. SOAs can achieve similar gain resolu-
tion as a mirror-based VOA, but it is typically not a linear function of the SOA bias
current and is also dependent on the input optical power. Additionally, memory and
cross-gain saturation effects within the SOA can induce crosstalk between RF signals
modulated onto different wavelength-division multiplexed optical carriers.
Tunable optical delay lines have been researched intensively for their important
role in applications such as wideband true-time delays, FIR filters, beamforming,
and optical buffers [16, 38–41]. They have been implemented through a variety of
mechanisms such as mechanical positioning of mirrors, thermally-tuned refractive
index spiral waveguides, microring resonators, Bragg gratings, and switching delays.
Sub-picosecond delay resolution can be readily achieved by multiple delay technolo-
gies. At 10 GHz RF frequency, this corresponds to a phase shift of 0.01◦ , which is
sufficient to meet high cancellation depths.
Lastly, an interference canceller must be able to both resolve weak signals of inter-
est and deal with strong interferers simultaneously. Many microwave photonic links
have demonstrated experimental dynamic ranges surpassing 120 dB/Hz2∕3 [19] (here
we use third-order intermodulation spurious-free dynamic range or SFDR as the
representative metric), which meets the stringent requirements of many analog RF
applications.
In a microwave photonic link, the ceiling on the dynamic range is often set by the
optical modulator, because of its nonlinear transfer function. With direct modulation
(modulating the laser bias current using the RF signal), dynamic range is maximized
by using a laser with as linear a light-current relationship as possible, maximizing
gain, and minimizing relative intensity noise (RIN) [19, 44]. In direct modulation,
the dynamic range is also a strong function of frequency, declining sharply as the
operating frequency approaches the laser relaxation oscillation frequency [19, 45].
External modulators have the benefit of operating a continuous-wave laser and
therefore, do not have the dynamic range degradation with respect to frequency
that directly modulated links exhibit. However, both Mach-Zehnder modulators
(MZM) and electro-absorption modulators (EAM) exhibit nonlinear transfer func-
tions, which limit the link’s dynamic range at high powers. Because an MZM’s
transfer function is predictable, it can be linearized by pre-distorting the input RF
signal to eliminate the third-order intermodulation product [46]. Such linearized
links have demonstrated SFDRs larger than 130 dB/Hz4∕5 [19], albeit at limited fre-
quencies. Other techniques, such as low-biasing the MZM to lower noise and using
dual-parallel MZMs to cancel third-order intermodulation products, have also been
used to increase SFDR [13, 47, 48]. Overall, microwave photonic links can exhibit
extremely large dynamic ranges. However, careful consideration of the link appli-
cation requirements must be taken in account, as there is almost always a tradeoff
between dynamic range, gain, noise figure, and bandwidth.
(a) 10 kHz
65 dB
SOI
(b) 10 kHz
65 dB
SOI
(c)
40 dB
SOI
50 MHz
(d)
30 dB
SOI
40 MHz
Fig. 10 Experimental results of interference cancellation in the (top) 900 MHz band and (bottom)
2.4 GHz band for broadband (left) and narrowband (right) interfering signals. SOI = signal of
interest. Taken from [29]
288 M.P. Chang et al.
Simulation
Experiment
Simulation
Experiment
Fig. 11 Interference cancellation measured over 400 MHz in both the (top) 900 MHz band and
(bottom) 2.4 GHz band. Taken from [29]
and TDL to account for the interference channel’s frequency response, illustrating the
wide tunability of the optical canceller. In fact, any system operating frequency from
DC to 10 GHz can be selected by the VOA and TDL. The system achieves greater
than 65 dB of cancellation of a narrowband interferer, demonstrating the high ampli-
tude and delay precision of the optical components. For broadband cancellation, the
system achieves better than 40 and 30 dB of cancellation for a 50 MHz and 40 MHz
signal in the 900 MHz and 2.4 GHz bands, respectively. In all cases, the signal of
interest is recovered, despite the in-band interference, which is a key advantage of
interference cancellation over filtering. Because the cancellation system operates on
the RF-level of the communication system, the results are completely independent
of either signal of interest or interference signal modulation format.
The tradeoff between cancellation depth and bandwidth (see Sect. 2.3) can be
studied by observing cancellation for a fixed system state across a wide bandwidth,
as shown in Fig. 11. Cancellation degrades with bandwidth because of the inability
of a two-tap system to model the interference channel’s response at every frequency.
Nonetheless, the optical system has clearly demonstrated the ability to operate over
an extremely wide bandwidth.
A limitation of the previous system was that the optical delay line was tuned mechan-
ically, and therefore could not adapt to rapidly changing RF environments. The archi-
tecture featured in Fig. 12 utilizes similar principles to the interference canceller
in the previous section; however, unlike the previous architecture, a semiconduc-
tor optical amplifier (SOA) is inserted into the cancellation path to provide a fast
and fully electronic control interface. The SOA performs three key signal-processing
Optical-Based Interference Cancellation in Wireless Sensor Networks 289
Fig. 12 Architecture of an adaptive canceller using a semiconductor optical amplifier based tun-
able delay line and variable optical weight. BPF = Bandpass Filter, EDFA = Erbium-Doped Fiber
Amplifier, Mod = Modulator, PC = Polarization Controller
(b)
3
2
1
2
1
9 dB rise in noise floor is in agreement with estimated SOA noise figure, and is one
tradeoff of using the SOA.
The SOA’s fully electronic control variables are used to implement a computer-
controlled optimization algorithm to adaptively cancel interference. An example
experimental run of the algorithm is shown in Fig. 14a. A map of the output residual
interference power as a function of the SOA control variables is retroactively super-
imposed. Figure 14b demonstrates that the algorithm converges to the global min-
imum solution over a delay mismatch range of 83 ps (the maximum SOA-induced
delay range was 120 ps). We note that 120 ps is not a sufficient delay range to account
for delay variations in most RF environments, particularly if antennas are subject to
movement. Indeed, additional coarse delays (either optical or RF) will be required
to account for large delay variations. With regards to speed, the settling time of the
SOA corresponds to the inverse of the modulation bandwidth of the SOA, which is
about 1 µs for these devices. This can be improved by selecting SOAs designed for
high modulation bandwidths. The SOA-based tunable delay is thus well-suited to
react to rapid but small changes in delay, such as antenna vibrations.
As a bit of foreshadowing, the SOA-based interference canceller sheds light on a
possible path to achieving a fully-integrated interference canceller, by designing the
signal-processing functions of amplitude control, delay control, and signal inversion
into a semiconductor.
Optical-Based Interference Cancellation in Wireless Sensor Networks 291
(a) (b)
Fig. 14 a The SOA-based interference canceller adaptively canceling interference using the SOA
bias current and input optical power as control knobs. b The system converges to the global mini-
mum over a range of 83 ps. Insets show the output spectra as the algorithm progresses. Taken from
[30]
tude and phase to match. The goal of the optical canceller is to delay and attenuate
each interference reference copy n0copy(i) (t), with unique parameters 𝜙i and 𝛼i , respec-
tively, such that it resembles each ni (t) as closely as possible, thereby achieving the
maximum cancellation. To perform these adjustments, each cancellation path under-
goes delay matching using fixed optical fiber lengths and tunable optical delay lines,
and amplitude matching using variable optical attenuators, yielding 𝛼i n0copy(i) (t − 𝜙i ).
After tuning adjustments, both the cancellation and received signals are fed into
separate inputs of a custom-built 8 × 1 singlemode to multimode combiner (SM-
MM) combiner. The SM-MM combiner maps each singlemode input onto orthog-
onal spatial modes of a multimode fiber [52]. Since each optical signal resides in
an orthogonal spatial mode, all optical carriers can share wavelengths without suf-
fering coherent beat noise. This eliminates the need for a separate optical source
for each multipath component as well as the associated multiplexers/demultiplexers/
modulators that would be necessary otherwise. Specifically, without the SM-MM
combiner, this system, with N = 2 taps, would require three distinct optical sources
and modulators. Although this allows each laser and modulator to be tailored to opti-
mize different trade-offs, its scalability is limited.
Finally, the combined optical signal is detected by a multimode photodetector.
Because each optical signal resides in a different spatial mode of the multimode fiber,
Optical-Based Interference Cancellation in Wireless Sensor Networks 293
the photodetector incoherently sums each optical signal in the output RF photocur-
rent. This operation results in the subtraction of the self-interference copies from the
received signal, since the self-interference copies were initially inverted by a balun.
The system output can be represented as:
∑
N
y(t) = xRX (t) − 𝛼i n0copy(i) (t − 𝜙i )
i=1
{ }
∑
N
∑
N
= s(t) + ni (t) − 𝛼i n0copy(i) (t − 𝜙i )
i=1 i=1
where our goal is to reduce the term in brackets to 0 such that y(t) → s(t).
Figure 16 shows the results of the optical canceller used in an outdoor environ-
ment to uncover a weak signal of interest [51]. The canceller, wielding two can-
cellation paths, canceled the line-of-sight component of self-interference and the
strongest multipath component of the same self-interference. The system achieved
34 dB of narrowband cancellation and >29 dB of cancellation across 20 MHz of
instantaneous bandwidth. Multipath cancellation contributed more than 10 dB of
additional cancellation. This additional cancellation enabled the system to uncover
a weak sigal of interest received at 5 dB above the noise floor [51].
Although discrete optical cancellers achieve impressive performance, they are imprac-
tical for cost-sensitive commercial applications due to their large size, high cost, and
power hungry nature. Miniaturizing the optical canceller into a monolithic photonic
integrated circuit (PIC) can offer a panacea to these issues and pave avenues to new
applications. A PIC is compact and mobile, since it is a single semiconductor chip;
scalable and low cost, since it can leverage the economies of scale of the semicon-
ductor industry; and robust to thermal and vibrational shock, since it is purely solid-
state. With these advantages, the PIC canceller creates a win-win situation for both
performance and cost. Here, we introduce the design and demonstration, for the first
time, of a monolithically integrated PIC canceller.
The PIC canceller is designed to mimic the functionality of the discrete optical can-
celer while eliminating its weaknesses. First, the PIC canceller is optically self-
contained, meaning that it has monolithically integrated lasers and photodetectors
so that the electrical-optical and optical-electrical conversions are all performed on-
chip. This eliminates the expensive and tedious fiber-coupling process that drives
294 M.P. Chang et al.
Fig. 16 A weak signal of interest received at 5 dB above the noise floor is uncovered with the help
of the multi-tap optical canceller. The spectrum analyzer plots show the output of the canceller (top)
without cancellation, (middle) with cancellation using only the line-of-sight cancellation path, and
(bottom) with cancellation using both the line-of-sight and the multipath cancellation paths. Only
the last configuration is able to uncover the weak SOI
Optical-Based Interference Cancellation in Wireless Sensor Networks 295
packaging costs in optical components and other PICs. Instead, the PIC can be
packaged using standard IC packaging technology. Second, instead of optical fiber,
semiconductor optical waveguides carry the optical signal from one optical device
to the next (Fig. 17), greatly reducing the loss caused by cascaded fiber-couplings
between discrete optical components. This loss can be 1–2 dB per fiber-coupling,
which results in 2–4 dB of RF loss due to the square law relationship between opti-
cal and RF power. Link designers now have more margin in the loss budget and can
make tradeoffs to improve dynamic range, noise figure, or bandwidth.
As with the discrete optical canceller, the optical modulators are critical to the per-
formance of the PIC canceller. The PIC canceller erforms electrical-optical conver-
sion via direct modulation of etched facet Fabry-Perot lasers as the optical modulator
(see Fig. 18). Although, distributed feedback (DFB) lasers are usually the mono-
lithic laser source in PICs, the etched-facet laser design eliminates the requirement
for complicated electron-beam or stepper lithography, as well as one regrowth step,
which adds cost and reduces yield. The etched facet is created by dry-etching a trench
to produce a vertical sidewall that is sufficiently smooth to act as a reflective laser
Fig. 17 Mode-matched
ridge
(a) (b)
Dry-Etched
Trench
InP P-Contact
Fig. 19 a Schematic of the chip layout and composition. b Scanning Electron Microscope image
of a fabricated PIC
facet. The 5 µm wide trench is etched 3 µm deep into the existing ridge waveguide,
which connects the laser to the remainder of the photonic circuit. The lack of an
external modulator also reduces insertion loss and the need for an additional device;
however, further research on alternative laser and modulator devices is ongoing.
Architecturally speaking, each PIC canceller is composed of two paths: one for
the propagation of the reference interference signal (the cancellation path) and one
for the propagation of the corrupted received signal (see Fig. 19a). Both paths have
tunable amplitude and phase devices to maximize tunability range. All elements on
the PIC canceller are connected by integrated waveguides, which carry the opti-
cal/RF signal from one element to the next with minimum loss. Both paths consists
of a monolithic laser, tunable delay line, semiconductor optical amplifier (SOA), and
photodetector. As with the architecture presented in Sect. 4.2, the SOA plays a criti-
cal role in the functionality of the canceller. It overcomes the speed bottleneck that is
present in the discrete optical canceller while also implementing amplitude match-
ing of the RF signal via current injection. A waveguide photodetector converts the
processed optical signal back into an RF signal. Gold wirebonds are used to transfer
the RF signals to and from the PIC. Finally, the subtraction mechanism is performed
by using a differential transimpedance amplifier following the photodetectors.
The PIC is fabricated on a 1550 nm multiple quantum well laser-diode epitaxial
wafer. The active region of the wafer consists of 5 InGaAlAs quantum wells grown on
a semi-conducting InP substrate; the photoluminescence spectral peak of the wafer
is 1542 nm. The fabrication of the PIC requires five mask layers: ridge definition,
electrical isolation, trench etch, contact window opening, and p-contact deposition.
Following fabrication, the PICs are lapped, cleaved, and mounted to a silicon sub-
mount, which facilitates electrical contact and testing. Each cleaved die, which also
includes test structures and other elements, is 2.5 mm × 2 mm. A scanning electron
microscope image of a single PIC die is shown in Fig. 19b.
PICs were characterized for both DC optical performance and interference cancel-
lation. Self-interference cancellation tests were conducted by generating an interfer-
Optical-Based Interference Cancellation in Wireless Sensor Networks 297
(a) (b) 0
-5
Cancellation (dB)
125 mV
-10
No Cancellation
Vatt = 1.35 V -15
Transmitter Laser Off
-20
-25
-30
-35
Vatt = 1.37 V Vatt = 1.42 V -40
10 15 20 25 30 35 40
Frequency (MHz)
ence signal using a signal generator and modulating it onto both paths of the PIC. The
system was then tuned to optimize cancellation of the self-interference by adjusting
the laser and photodetector bias voltages to amplitude match the reference interfer-
ence with the interference in the corrupted received signal.
Preliminary results of the experiment are shown in Fig. 20a and demonstrate the
convergence of the PIC from zero cancellation to maximum cancellation. In these
experiments, the delay is matched ahead of time, and the amplitude is tuned from
completely unmatched (i.e. cancellation path is off) to optimum matching. As the PIC
improves the amplitude matching between the interference reference and the actual
interference signal, the residual interference at the PIC output is slowly reduced
until it hits the noise floor. The cancellation versus bandwidth tradeoff for the PIC
is shown in Fig. 20b, indicating that optimum cancellation is achieved over about
5 MHz instantaneous bandwidth and degrades sharply outside that bandwidth. The
degradation is a result of deviations between the frequency responses of the two dif-
ferent paths of the PIC. We believe that the deviation results from significant differ-
ences in parasitics, such as wirebond lengths and the traces on the silicon submount.
Fabrication adjustments to improve the chip bandwidth are currently underway.
For the past few decades, research on microwave photonics has primarily focused
on discrete fiber-optic systems. These systems were built to address niche appli-
cations, such as military and defense, where performance is the top priority over
cost. However, the recent years have witnessed a flurry of interest in high data rate
(Gbps) telecommunications, such as wireless sensor networks, broadband public
access points, and small cell and picocell backhaul. These applications are much
more cost-sensitive and are not appropriate for discrete fiber systems. The emerg-
ing field of integrated microwave photonics is poised to address these high data rate
298 M.P. Chang et al.
applications, while also meeting their price-point. The high volume demand of these
applications can also serve as the catalyst to generate the capital needed to propel
the integrated microwave photonic infrastructure to commercial readiness.
Integrated microwave photonics seeks to package the advantages of optics, such
as wide bandwidth and frequency-agnostic performance, into an integrated circuit,
where it can leverage the economies of scale of the semiconductor industry. The pri-
mary barriers to the commercial viability of such systems are the economics (yield)
and the fabrication of materials and devices, which must satisfy stringent applica-
tion requirements. As opposed to electronic ICs, where the transistor is the dominant
device, multiple optical devices are needed per photonic integrated circuit (PIC),
leading to a more difficult fabrication process. Unlike discrete microwave photonics,
where each device can be optimized on separate materials, integrated microwave
photonics compromises this degree of freedom for manufacturing simplicity. Yet,
great strides have been made in recent years to increase the number of devices inte-
grated on one PIC and reduce the killer defect density to increase yield (see Fig. 21).
Infinera, the first widely successful PIC company, has led the charge by fabricating a
PIC with over 400 devices, capable of aggregate data rates of 500 Gbps [53]. Indeed,
a Moore’s Law of Photonics has already been declared.
Photonic integrated circuits (PICs) are primarily being built in three core material
platforms: InP and GaAs (III-V), silicon-on-insulator, and silicon nitride. Each plat-
form has its unique advantages and disadvantages and should be selected based on
the application requirements; a review for each platform is provided in [55]. Multi-
Project Wafer (MPW) consortia for standardized PIC development, such as IMEC
and JeppiX in Europe, have made PIC research much more accessible to researchers
and developers worldwide. These initiatives consolidate resources and know-how to
make it affordable for industry and academic researchers to experiment with PICs.
A standard software ecosytem (layout, verification, extraction) and simulation plat-
form is also needed to unite the PIC community in a similar fashion to what was
achieved in silicon. Companies such as PhoeniX and PhotonDesign have begun this
effort.
It is an exciting time to be involved in integrated microwave photonics research
because manufacturing technology maturity, device performance, and market demand
Fig. 21 a Device count per PIC. Taken from [54]. b Killer defect density. Taken from [53]
Optical-Based Interference Cancellation in Wireless Sensor Networks 299
References
1. Cisco, Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2015–
2020 (2016)
2. Resonant Inc., Radically reducing the cost and size of cellphone RF filters to fuel the mobile
revolution, Tech. Rep. (2015)
3. S. Choi, A. Cichocki, A. Beloucharni, Second order nonstationary source separation. J. VLSI
signal Process. Syst. Signal, Image Video Technol. 32(1–2), 93–104
4. S. Lee, S. Choi, A. Cichocki, H.M. Park, Blind source separation and independent component
analysis: a review, Jan 2005
5. G. Fabrizio, A. Farina, Blind source separation with the generalised estimation of multipath
signals algorithm. IET Radar Sonar Navig. 8(9), 1255–1266 (2014)
6. S. Hong, J. Brand, J. Choi, M. Jain, J. Mehlman et al., Applications of self-interference can-
cellation in 5G and beyond. IEEE Commun. Mag. 52(2), 114–121 (2014)
7. M.P. Chang, P.R. Prucnal, Full-duplex spectrum sensing in cognitive radios using optical self-
interference cancellation, in 2015 9th International Conference on Sensors and Technology
(IEEE, 2015), pp. 341–344
8. M. Jain, J.I. Choi, T. Kim, D. Bharadia, S. Seth et al., Practical, real-time, full duplex wire-
less, in Procedings of 17th Annual International Conference on Mobile Computing Networks—
MobiCom ’11 (ACM Press, New York, New York, USA, 2011), p. 301
9. D. Bharadia, E. McMilin, S. Katti, Full duplex radios, in Proceedings of ACM SIGCOMM 2013
Conference—SIGCOMM ’13, vol. 43, no. 4 (ACM Press, 2013, New York, New York, USA),
p. 375
10. R.C. Williamson, R.D. Esman, RF photonics. J. Light. Technol. 26(9), 1145–1153 (2008)
11. J. Capmany, D. Novak, Microwave photonics combines two worlds, in National Photonics,
vol. 1, no. 6, June 2007, pp. 319–330
12. E. Granot, R. Weber, S. Tzadok, G. Gertel, N. Narkiss, Fibers vs. coax for RF delay line appli-
cations, in 2009 IEEE International Conference on Microwaves, Communications, Antennas
and Electronics System (IEEE, Nov 2009), pp. 1–2
13. D. Marpaung, High Dynamic Range Analog Photonic Links: Design and Implementation
(2009)
14. J. Yao, Microwave photonics. J. Light. Technol. 27(3), 314–335 (2009)
15. A.J. Seeds, K.J. Williams, Microwave photonics. Light. Technol. J. 24(12), 4628–4643 (2006)
16. R.A. Minasian, Photonic signal processing of microwave signals. IEEE Trans. Microw. Theory
Tech. 54(2), 832–846 (2006)
17. C.H. Cox, Gain and noise figure in analogue fibre-optic links, pp. 238–242 (1992)
18. C.H. Cox, E. Ackerman, R. Helkey, G.E. Betts, Techniques and performance of intensity-
modulated direct-detection of analog optical links. IEEE Trans. Microw. Theory Tech. 45(8),
1375–1383 (1997)
19. C. Cox, E. Ackerman, G. Betts, J. Prince, Limits on the performance of RF-over-fiber links
and their impact on device design. IEEE Trans. Microw. Theory Tech. 54(2), 906–920 (2006)
20. J. Capmany, B. Ortega, D. Pastor, A tutorial on microwave photonic filters. J. Light. Technol.
24(January), 201–230 (2006)
21. R.A. Minasian, Ultra-wideband and adaptive photonic signal processing of microwave signals.
IEEE J. Quantum Electron. 52(1), 1–13 (2016)
300 M.P. Chang et al.
22. J. Chang, M.P. Fok, R.M. Corey, J. Meister, P.R. Prucnal, Highly scalable adaptive photonic
beamformer using a single mode to multimode optical combiner. IEEE Microw. Wirel. Com-
ponents Lett. 23(10), 563–565 (2013)
23. J. Chang, J. Meister, P.R. Prucnal, Implementing a novel highly scalable adaptive photonic
beamformer using “blind” guided accelerated random search. J. Light. Technol. 32(20), 3623–
3629 (2014)
24. S. Fathpour, Silicon-photonics-based wideband radar beamforming: basic design. Opt. Eng.
49(1), 018201 (2010)
25. T.P. McKenna, J.A. Nanzer, T.R. Clark, Photonic beamsteering of a millimeter-wave array with
10-Gb/s data transmission. IEEE Photonics Technol. Lett. 26(14), 1407–1410 (2014)
26. J. Sancho, J. Lloret, I. Gasulla, S. Sales, J. Capmany, Fully tunable 360 microwave photonic
phase shifter based on a single semiconductor optical amplifier. Opt. Express, 19(18), 17–421
(2011)
27. W. Xue, S. Sales, J. Capmany, J. Mørk, Wideband 360 degrees microwave photonic phase
shifter based on slow light in semiconductor optical amplifiers. Opt. Express 18(6), 6156–6163
(2010)
28. K.L. Deng, K.I. Kang, I. Glask, P. Prucnal, A 1024-channel fast tunable delay line for ultrafast
all-optical TDM networks. IEEE Photon. Technol. Lett. 9(11), 1496–1498 (1997)
29. M.P. Chang, M. Fok, A. Hofmaier, P.R. Prucnal, Optical analog self-interference cancella-
tion using electro-absorption modulators. IEEE Microw. Wirel. Compon. Lett. 23(2), 99–101
(2013)
30. M.P. Chang, C.-L. Lee, B. Wu, P.R. Prucnal, Adaptive optical self-interference cancellation
using a semiconductor optical amplifier. IEEE Photonics Technol. Lett. 27(9), 1018–1021
(2015)
31. J. Chang, P.R. Prucnal, A novel analog photonic method for broadband multipath interference
cancellation. IEEE Microw. Wirel. Compon. Lett. 23(7), 377–379 (2013)
32. C. Cox, Transmit isolating photonic receive links: a new capability for antenna remoting, in
Optical Fiber Communication Conference on Fiber Optic Engineers 2011 (OSA, Washington,
D.C., Mar 2011), p. OThA1
33. C. Cox, E. Ackerman, Demonstration of a single-aperture, full-duplex communication system,
in 2013 IEEE Radio Wireless Symposium (IEEE, Jan 2013), pp. 148–150
34. R. Alferness, Waveguide electrooptic modulators. IEEE Trans. Microw. Theory Tech. 30(8),
1121–1137 (1982)
35. J.-W. Shi, C.-B. Huang, C.-L. Pan, Millimeter-wave photonic wireless links for very high data
rate communication. NPG Asia Mater. 3(4), 41–48 (2011)
36. M. Fice, E. Rouvalis, F. Van Dijk, A. Accard, F. Lelarge et al., 146-GHz millimeter-wave radio-
over-fiber photonic wireless transmission system. Opt. Express 20(2), 1769–1774 (2012)
37. A. Hirata, H. Ishii, T. Nagatsuma, Design and characterization of a 120-GHz millimeter-wave
antenna for integrated photonic transmitters. IEEE Trans. Microw. Theory Tech. 49(11), 2157–
2162 (2001)
38. M.P. Chang, N. Wang, B. Wu, P.R. Prucnal, A simultaneous variable optical weight and delay
in a semiconductor optical amplifier for microwave photonics. J. Light. Technol. 33(10), 2120–
2126 (2015)
39. P.A. Morton, J. Cardenas, J.B. Khurgin, M. Lipson, Fast thermal switching of wideband optical
delay line with no long-term transient. IEEE Photonics Technol. Lett. 24(6), 512–514 (2012)
40. F. Xia, L. Sekaric, Y. Vlasov, Resonantly enhanced all optical buffers on a silicon chip, in 2007
Photonics Switch. PS (2007), pp. 7–8
41. J.K.S. Poon, L. Zhu, G.A. DeRose, A. Yariv, Transmission and group delay of microring
coupled-resonator optical waveguides. Opt. Lett. 31(4), 456–458 (2006)
42. B. Jalali, S. Fathpour, Silicon Photonics. J. Light. Technol. 24(12), 4600–4615 (2006)
43. R. Soref, The past, present, and future of silicon photonics. IEEE J. Sel. Top. Quantum Elec-
tron. 12(6), 1678–1687 (2006)
44. G. Bodeep, T. Darcie, Semiconductor lasers versus external modulators: a comparison of non-
linear distortion for lightwave subcarrier CATV applications. IEEE Photonics Technol. Lett.
1(11), 401–403 (1989)
Optical-Based Interference Cancellation in Wireless Sensor Networks 301
45. K.Y. Lau, A. Yariv, Intermodulation distortion in a directly modulated semiconductor injection
laser. Appl. Phys. Lett. 45(10), 1034 (1984)
46. R. Childs, V. O’Byrne, Multichannel AM video transmission using a high-power Nd:YAG
laser and linearized external modulator. IEEE J. Sel. Areas Commun. 8(7), 1369–1376 (1990)
47. L.M. Johnson, H.V. Roussell, Reduction of intermodulation distortion in interferometric opti-
cal modulators. Opt. Lett. 13(10), 928 (1988)
48. S. Korotky, R. de Ridder, Dual parallel modulation schemes for low-distortion analog optical
transmission. IEEE J. Sel. Areas Commun. 8(7), 1377–1381 (1990)
49. J. Capmany, I. Gasulla, S. Sales, Microwave photonics: Harnessing slow light. Nat. Photonics
5(12), 731–733 (2011)
50. J.B. Khurgin, Slow light in various media: a tutorial. Adv. Opt. Photonics 2(3), 287 (2010)
51. J. Sun, M. Chang, R.P. Paul, Scalable optical self-interference cancellation system for over-
the-air multipath interference, IEEE Photonics J. (2016)
52. M.P. Fok, Y. Deng, K. Kravtsov, P.R. Prucnal, Signal beating elimination using single-mode
fiber to multimode fiber coupling. Opt. Lett. 36(23), 4578–4580 (2011)
53. F.A. Kish, D. Welch, R. Nagarajan, J.L. Pleumeekers, V. Lal et al., Current status of large-
scale InP photonic integrated circuits. IEEE J. Sel. Top. Quantum Electron. 17(6), 1470–1489
(2011)
54. M. Smit, J.V.D. Tol, M. Hill, Moore’s law in photonics, vol. 13, no. 1 (2012), pp. 1–13
55. D. Marpaung, C. Roeloffzen, R. Heideman, A. Leinse, S. Sales, J. Capmany, Integrated
microwave photonics. Laser Photonics Rev. 7(4), 506–538 (2013)
Traffic Adaptive Channel Access
Scheme for IEEE802.15.4 Cluster-Based
WSNs Under Spatial Non-uniform Traffic
Condition
1 Introduction
Due to limited network resources such as battery capacity and processing capa-
bility, wireless sensor networks (WSNs) absolutely require lower energy con-
sumption and simpler processing mechanism to achieve long network lifetime
under limited network resource condition [1, 2]. To satisfy these requirements, duty
cycle mechanism, under which sensor nodes (SNs) periodically put their commu-
nication functions in inactive state, and cluster-based network topology, under
which neighbouring SNs form a cluster and each SN transmits its measured data to
its cluster head (CH) and then the CH sends the aggregated data to a sink node (or a
parent CH), are deeply investigated in recent years [3, 4].
2 Related Work
IEEE 802.15.4 [5] specifies physical (PHY) and MAC protocols for low-rate
wireless personal area networks (LR-WPANs) and is often employed to lower layer
protocol in WSNs, such as ZigBee [14, 15], due to its lower energy consumption.
The beacon enabled mode specified by IEEE 802.15.4 MAC provides power saving
operation based on duty cycle mechanism by using superframe structure. The
superframe is divided into active and inactive periods, and its active period is also
divided into two periods: contention access period (CAP) and contention free
306 A. Yamauchi et al.
period (CFP). This chapter focuses on shared channel access with slotted
CSMA/CA in the CAP, and dedicated channel access in the CFP is out of scope.
The slotted CSMA/CA channel access normally adopts backoff operation to
prevent signal collisions on shared communication channel. Before transmitting
data, each sender randomly selects access delay interval from a backoff window
(BW) defined by a range of [0, 2BE-1]. The backoff exponent BE is initially set to
BEmin and is increased up to BEmax when clear channel assessment (CCA) detects
busy channel after a timeout of selected access delay interval. The transmitting data
is discarded after the number of the backoffs exceeds a predefined value
(‘macMaxCSMABackoffs’ in IEEE 802.15.4).
Equal transmission opportunity is generally given to all the nodes willing to
transmit their data in the slotted CSMA/CA access with normal backoff operation.
This is because the above control parameters such as BEmin, BEmax, and
macMaxCSMABackoffs commonly take the same values for all nodes. However,
this parameter setting and thus equal transmission opportunity for all nodes, is
undesirable to the cluster based WSNs with spatial non-uniform traffic.
IEEE 802.11e EDCA [9] is a typical well-known priority access scheme for shared
communication channels in WLANs, and can offer QoS provision to multi-medium
communication services in WLANs by utilizing backoff operation in CSMA/CA
mechanism.
The EDCA defines multiple access categories (ACs) depending on required
QoSs, such as transmission delay and so on. Transmission data is categorized into
one of these ACs based on its required QoS and is stored separately in the trans-
mission queue corresponding to each AC. The EDCA selects the transmission data
from one of the AC queues based on their priority order and then transmit wireless
frame containing selected data.
To offer priority channel access depending on the required QoS, the control
parameters, arbitration inter-frame space (AIFS) and contention window (CW) are
set to an adequate value for each AC according to its priority order. Specifically, the
AIFS and CW take smaller values for the AC with higher priority, and larger ones
for lower priority AC. Thus, the EDCA provides priority operation by adjusting
access delay interval before data transmission, and enables the QoS provision
required by each communication service.
Traffic Adaptive Channel Access Scheme for IEEE802.15.4 … 307
3 System Model
Let us consider a cluster-based WSN with one PCH and Ncch CCHs subordinating
some SNs, as illustrated in Fig. 1. All nodes in the WSN operate with duty cycle
mechanism in accordance with IEEE 802.15.4 beacon enabled mode, and the dif-
ferent time period in the IEEE 802.15.4 superframe is assigned as an active period
for intra-cluster communication to different cluster in order to avoid interference
between any intra-cluster communications. The active period for inter-cluster
communication is also assigned different time period from those for any
intra-cluster communications, as shown in Fig. 2. Therefore, this active period
Inter-cluster
communication
Ncch child
cluster heads (CCHs)
CCH1 CCH2 . . . CCH Ncch
cluster1 cluster Ncch
cluster2
...
Intra-cluster
SNs CCH Ncch
CCH Ncch (inactive) CAP CAP (inactive)
Time
4 Motivation
Contemplating the causes raising the above problem, the proposed scheme requires
two basic functions: traffic adaptive channel access control and autonomous dis-
tributed control.
For the traffic adaptive channel access, the CCHs with high/low traffic should
operate with high/low priority for shared channel access to supress the performance
difference between CCHs with different traffic. In addition, this priority control
should be achieved by using autonomous distributed control mechanism since
WSNs should avoid additional control signal exchange due to their limited network
resources, such as communication capacity, computational and battery power.
In the traditional priority access mechanism in WLAN systems, IEEE 802.11e, each
mobile terminal (MT) decides the channel access priority by selecting the size of
AIFS and CW in backoff operation for CSMA/CA channel access, as mentioned
before in Sect. 2.2. IEEE 802.15.4 MAC employs the same kind of backoff oper-
ation as IEEE 802.11 WLAN systems, and the BW for IEEE 802.15.4 MAC
corresponds to the CW for IEEE 802.11e. The size of both BW and CW is given by
the same control parameter of backoff exponent BE. Therefore, the proposed
scheme also controls the access priority by adjusting the BE.
The proposed scheme sets different access priority to the CCHs according to the
traffic conditions in their clusters. The priority control can be achieved by setting
different value to this BE at each CCH with different traffic load. Figure 3 illustrates
a conceptual diagram of the proposed autonomous traffic adaptive backoff control
scheme.
In the proposed priority control, as a basic framework, the high traffic loaded
CCHs obtain high priority state, being able to get more transmission opportunity,
by setting a smaller value to their BE. The smaller BE brings a shorter waiting
interval before channel accesses, leading to high priority in the shared channel
access. The low traffic loaded CCHs, on the other hand, take a larger value to their
310 A. Yamauchi et al.
BE in order to increase the BW size and this parameter setting forces them to wait
for a larger interval, leading to low priority state, being hard to get more trans-
mission opportunity. From these BE setting, the CCHs with higher traffic can obtain
more transmission opportunity than the CCHs with lower traffic, as illustrated in
Fig. 3. As a result, each CCH can transmit the data received from its SNs in
proportion to its own traffic load, and then, WSNs can suppress the non-uniform
transmission performance between CCHs, resulting in system performance
improvement.
IEEE 802.15.4 MAC defines two control parameters for the BE; BEmin and
BEmax. The BEmin adjustment could impact the access priority at the earlier stage of
backoff operation. BEmax adjustment, however, cannot influence at the earlier stage,
do only at the final stage. Therefore, in this work, we employ the BEmin to control
the priority levels for CCHs.
To achieve the above-mentioned priority access, each CCH needs to know its
current traffic status relative to other CCHs competing the same CAP. Therefore, as
an indicator of the relative traffic status, the proposed scheme introduces the length
LCAPend
queue (packets) of transmission queue at the end point of the CAP assigned to the
communication from the CCHs to their PCH.
At the end point of a given CAP, the data stored in the transmission queue shows
the data which the CCH cannot complete to transmit during that CAP. Monitoring
the amount of the non-completed data allows each CCH to recognize its trans-
mission status, good or poor status, reflecting the competition for the CAP with
Traffic Adaptive Channel Access Scheme for IEEE802.15.4 … 311
...
(monitoring) PCH
Tx. queue
Short
Long backoff CCH n
Backoff CS BEmin CS
Data increase Data Low traffic
Channel access at CCHn
other CCHs. Hence, each CCH can estimate its traffic status relative to other CCHs
by observing the transmission queue length LendCAP
queue .
Figure 4 partly illustrates an example for the proposed traffic status estimation,
where the CCH1 (upper one) with many SNs detects a long LCAPendqueue at the end of the
current CAP. Then, the CCH1 can estimate its traffic status at high traffic condition,
where it obtain insufficient transmission opportunity compared with its current load.
Conversely, the CCHn (lower one) with a small number of SNs has a short LCAPendqueue
which shows it can get sufficient transmission opportunity during that
CAP. Therefore the CCHn can cognize that it is in low traffic status. From these
considerations, the transmission queue length LCAPendqueue at the end of the CAP is
useful for estimating traffic status relatively to other CCHs.
estimate high traffic status relative to others when detecting the observed LCAPend
queue is
greater than the threshold THqueue. In contrast, they estimate their low traffic status
queue ≤ THqueue .
after detecting Nup consecutive LCAPend
For the specific procedure for controlling the BEmin, the CCHs under the high
traffic status decrease the value of BEmin by a decrement step ΔBEdown in order to
increase their access priority, whereas CCHs with the low traffic status increase the
BEmin by an increment step ΔBEup to lower their access priority. These adjustment
steps ΔBEdown and ΔBEup are predefined and constant control parameters. The
detailed controlling procedure for the BEmin and its adjustment range is given by:
8 before
< BEmin + ΔBEdown
> ; LCAPend
queue > THqueue
after before
BEmin = BEmin + ΔBEup queue ≤ THqueue in consecutive Nup
; LCAPend
>
: before ð1Þ
BEmin ; otherwise,
upperr
lower
subject to BEmin ≤ BEmin ≤ BEmin ,
lower upper
where BEmin and BEmin are a minimum and maximum value for BEmin. The
optimal settings for Nup, ΔBEdown and ΔBEup will be discussed in Sect. 7.2.
6 Performance Evaluations
The performance for the proposed autonomous traffic adaptive backoff control is
evaluated through computer simulation with following assumptions. The statistics
are collected from the inter-cluster communication involving one PCH and its Ncch
subordinate CCHs, as illustrated in Fig. 1.
The radio channels are assumed to be ideal without any consideration of propa-
gation loss, shadowing and fading fluctuation in this evaluation. Packet transmis-
sion error occurs only due to packet collision caused from simultaneous
transmission from multiple CCHs. The CCA at each CCH is perfect without impact
of hidden terminal problem.
The packets are continuously generated at each CCH following Poisson distribution
with an average cluster traffic load of Gx (packets/Lpkt), which corresponds to
Traffic Adaptive Channel Access Scheme for IEEE802.15.4 … 313
cluster traffic load. The Lpkt is a packet length and is same for all generated packets.
After the packet generation, the packet is stored in transmission buffer with a length
of Lbuffer (packets) at each CCH. The packet is discarded due to buffer overflow
when the buffer is full. At the buffer overflow the oldest packet is discarded.
Three different cluster traffic loads, Gx (x = ‘low’, ‘mid’, or ‘high’) are defined
to simulate spatial non-uniform traffic condition. Ncch CCHs are equally separated
into three CCH groups, and each CCH group has one of Gx without duplication
each other. Defining the ratio (Rlow : Rmid : Rhigh) for traffic loads (low : mid : high),
the traffic load Gx for the cluster in group x is given by using average network traffic
load G (packets/Lpkt):
Rx
Gx = ⋅ G ̸ NCCH ð2Þ
∑ Rx
x = low, mid, high
For transmission failure, the packets are retransmitted following IEEE 802.15.4
MAC specification.
We evaluate the packet dropping probability Pdrop and transmission delay Dtx as
transmission performance, where Pdrop and Dtx are given by a ratio of the number of
dropping packets due to transmission failure or buffer overflow at CCHs to that of
generated packets, and a time duration elapsed from packet generation at CCHs to
successful reception at the PCH, respectively. We also evaluate standard deviations
(SDs) σ drop and σ delay for the packet dropping probability Pdrops and the trans-
mission delay Dtxs observed at all the CCHs as an indicator showing uniformity in
the transmission performance among CCHs with different cluster traffic.
In the performance evaluation, we compare the above each performance for the
proposed scheme with that for the system employing fixed BEmin setting of 5, which
is called ‘standard system’ in this evaluation.
7 Evaluation Results
Using the parameter settings listed in Table 1, we evaluated the system perfor-
mance for the proposed scheme, comparing with the standard system. Note that
MAC parameter settings except shown in Table 1 employ their default values
specified in the IEEE standard.
314 A. Yamauchi et al.
Before evaluating the performance of the proposed scheme by comparing with the
standard system, optimal parameter settings for Nup, ΔBEup, and ΔBEdown should
be obtained to achieve better system performance for the proposed scheme.
Traffic Adaptive Channel Access Scheme for IEEE802.15.4 … 315
Fig. 5 Performance
non-uniformity between
CCHs with different cluster
traffic loads
Therefore, their optimal values are derived through computer simulation. Here, we
employ, as an evaluation metric, an average packet dropping probability Pave drop over
all CCHs in the network traffic region of 0.002 ≤ G ≤ 0.03, and its standard
deviation σ ave
drop for Ncch CCHs.
Firstly we determine an optimum value for the number of detection times Nup
required for estimating low traffic status. Figure 6 shows the Pave drop and σ drop per-
ave
formances for various Nup under the same conditions as Fig. 5. The packet dropping
probability Pave
drop in Fig. 6 is slightly better for smaller Nup. However, the standard
deviation σ ave
drop degrades at Nup of 1 and is better for Nup ≥ 2. In addition it is
almost the same for Nup ≥ 2, and thus, Nup of 2 is concluded to be optimum value
from these observations.
Next, Fig. 7 shows the Pave drop and σ drop performances for varying the increment
ave
step ΔBEup, to discuss its optimal value. The σ ave drop performance degrades with
increasing ΔBEup, while the Pave drop is almost equal for all ΔBEups. Therefore, these
results show the ΔBEup should be set to 1 as its optimum value.
Then, for optimization of the decrement step ΔBEdown, Fig. 8 shows the Pave drop
and σ ave
drop performances for various ΔBE down setting. Roughly speaking, both the
Pdrop and σ drop indicate almost the same performance regardless of ΔBEdown value,
ave ave
however, the larger ΔBEdown gives slight better σ avedrop performances. From these
results, therefore, we conclude ΔBEdown of 3 as its optimal setting.
Setting the optimum values derived above to Nup, ΔBEup, and ΔBEdown, Fig. 9
shows the BEmin behaviour for varying network traffic load G, under the same
conditions as Fig. 5. From Fig. 9, each CCH can adjust the BEmin to a smaller value
with higher traffic load and a larger value with lower traffic load in middle and high
network traffic load regions. This BEmin behaviour proves the proposed scheme
works well according to the traffic load estimated at each cluster.
The transmission performances Pdrop and Dtx are shown in Figs. 10 and 11 for the
proposed scheme (referred as ‘Traffic adaptive BE’ in Figs.) and the standard
system under the same conditions as Fig. 5. The performance for each traffic group,
Traffic Adaptive Channel Access Scheme for IEEE802.15.4 … 317
low, middle or high traffic, is separately shown when varying the network traffic
load G in these figures.
In Fig. 10, the packet dropping probability Pdrop is nearly equal between dif-
ferent traffic groups of CCHs for the proposed scheme, despite the standard system
with different Pdrop performance, and thus, the proposed scheme can suppress
318 A. Yamauchi et al.
non-uniformity in Pdrop across CH groups with different cluster traffic. For the
transmission delay Dtx, almost the same behaviour can be found in Fig. 11 as the
Pdrop. Hence, the non-uniformity in the transmission performances can be reduced
for all network traffic load regions, applying the proposed scheme.
To provides quantitative analysis of the uniformity in transmission performance
between CCH groups with different cluster traffic, Figs. 12 and 13 show the
standard deviations σ drop and σ delay of Pdrops and Dtxs observed at all the Ncch
CCHs. In Comparison with the standard system, both the standard deviations σ drop
and σ delay are largely reduced for high network traffic load regions by applying the
proposedscheme, and we can find in Figs. 12 and 13 significant reduction against
the transmission performance non-uniformity, provided by the proposed scheme.
This performance improvement is given by proper BEmin control based on current
cluster traffic offered to each CCH, which is a new mechanism provided by the
proposed scheme and is a main contribution of this work.
From the above results and their considerations, applying the proposed scheme
can provide quite effective mitigation against the non-uniformity in the transmission
performance between different traffic loaded CCHs on the inter-cluster communi-
cation in cluster-based WSNs under spatial non-uniform traffic.
In the previous section the effectiveness of the proposed scheme is basically pre-
sented for the typical network condition. Since the network has various operating
condition, this section examines the performance for the proposed scheme under for
various network operating conditions.
First, we examine the performance under various spatial non-uniform traffic
conditions, and then the packet dropping probability Pdrop is evaluated for the
conditions with spatial non-uniform traffic of (Rlow : Rmid : Rhigh) = (1.0: 1.0: 1.0),
(1.0: 1.25: 1.5), (1.0: 1.5: 2.0), (1.0: 2.0: 3.0), (1.0: 2.5: 4.0), and (1.0: 3.0: 4.0). The
Traffic Adaptive Channel Access Scheme for IEEE802.15.4 … 319
standard deviation σtraffic in offered cluster traffic loads under these traffic conditions
is 0.0, 0.2, 0.41, 0.82, 1.2, and 1.6, respectively, which indicates degree in
non-uniformity of the offered cluster traffic. Figure 14 shows the standard deviation
σ ave ave
drop of the Pdrop s at the CCHs, which is the same evaluation metric as shown in
Fig. 6, for varying the standard deviation σ traffic of offered traffic. We can see from
Fig. 14 that the proposed scheme provides superior σ ave drop performance to the
standard system for any σ traffic (spatial non-uniform traffic conditions). Moreover,
the performance improvement obtained by the proposed scheme increases with the
σ traffic. This result concludes that the proposed scheme can provide the effectiveness
under various spatial non-uniform traffic conditions, and the effectiveness gets
larger for the larger spatial non-uniform traffic conditions.
Next, we examine the transmission performance for various network scale,
which includes the number of CCHs Ncch competing the CAP in the inter-cluster
communication. The proposed scheme adjusts the BEmin according to the traffic
load offered to each CCH, and higher traffic CCHs tend to use smaller BEmin to
achieve higher priority in the channel access. However, the smaller BEmin likely
increases the possibility of channel access collision between competing CCHs, and
thus, the impact of smaller BEmin setting to the performance should be clarified for
the proposed scheme. Figures 15 and 16 show the packet dropping probability
performance for various number of competing CCHs. From the standard deviation
σ ave
drop shown in Fig. 15, the proposed scheme show better performance and can
achieve a reduction in the performance non-uniformity compared with the standard
system for any values of the Ncch. Therefore, an effectiveness of reducing the
performance non-uniformity can be obtained even under the conditions with larger
Ncch. However, the improvement is getting smaller with increasing Ncch, and thus,
the effectiveness gradually disappears with increasing Ncch. Additionally, the packet
dropping probability Pave
drop for the proposed scheme is slightly getting worse with
increasing Ncch, compared with the standard system, and therefore, the network
scale, such as the Ncch, gives no small impact to the transmission performance for
the proposed scheme. Hence, the proposed scheme has to pay some cost to achieve
uniformity in the transmission performance and requires countermeasure against
this drawback, which is left as our future work.
Traffic Adaptive Channel Access Scheme for IEEE802.15.4 … 321
8 Conclusion
Acknowledgments This work is partly supported by Japan Society for the Promotion of Science
(JSPS), Grant-in-Aid for Scientific Research (c) 25420366, 2013–2015.
References
A K
Akiyuki Yamauchi, 303 Kassim, Mohamed Rawidean Mohd, 137
Alahi, Md Eshrat E, 109 Kazuo Mori, 303
Al-Shamma’a, Ahmed, 205 Kocanda, Martin, 157
Arregui, Francisco J., 55 Korostynska, Olga, 205
Ascorbe, J., 55
Axmann, H., 1 L
Aziz, Samsuzana Abd, 245 Latiff, Nurul Adilah Abdul, 245
B M
Bergmann, A., 1 Mamidi, Shreya Reddy, 157
Bhandari, Manohar P., 229 Mason, Alex, 205
Blow, Eric, 273 Matthew P. Chang, 273
Bukka, Kaushik, 157 Matias, Ignacio R., 55
Burkitt, L., 109 Monica Lu, 273
Mukhopadhyay, Subhas, 109
C Mukhopadhyay, Subhas Chandra, 55
Carmona, Estefania Núñez, 229
Corres, J.M., 55 N
Nizar, Nina Naquiah Ahmad, 245
E Núñez Carmona, Estefanía, 179
Eichberger, B., 1
P
H Prucnal, Paul R., 273
Haji-Sheikh, Michael, 157 Pulvirenti, Andrea, 229
Harun, Ahmad Nizar, 137
Hashim, Dzulkifly Mat, 245 R
Hideo Kobayashi, 303 Rokhani, Fakhrul Zaman, 245
I S
Ikezawa, Satoshi, 21 Sairin, Masyitah Amat, 245
Ismail, Alyani, 245 Sberveglieri, Veronica, 179, 229
Shuk, P., 81
J Soprani, Matteo, 179
Jiaming Li, 39
Jingyi (Jenny) Sun, 273 T
Joshi, Keyur H., 205 Taherinezahdi, Mansour, 157
Jun Yamamoto, 21 Toshitsugu Ueda, 21
W Y
Wall, Josh, 39 Ying Guo, 39
West, Sam, 39
Z
X Zinger, Donald, 157
Xie Li, 109