warwick.ac.uk/lib-publications
Manuscript version: Author’s Accepted Manuscript
The version presented in WRAP is the author’s accepted manuscript and may differ from the
published version or Version of Record.
Persistent WRAP URL:
http://wrap.warwick.ac.uk/139584
How to cite:
Please refer to published version for the most recent bibliographic citation information.
If a published version is known of, the repository item page linked to above, will contain
details on accessing it.
Copyright and reuse:
The Warwick Research Archive Portal (WRAP) makes this work by researchers of the
University of Warwick available open access under the following conditions.
Copyright © and all moral rights to the version of the paper presented here belong to the
individual author(s) and/or other copyright owners. To the extent reasonable and
practicable the material made available in WRAP has been checked for eligibility before
being made available.
Copies of full items can be used for personal research or study, educational, or not-for-profit
purposes without prior permission or charge. Provided that the authors, title and full
bibliographic details are credited, a hyperlink and/or URL is given for the original metadata
page and the content is not changed in any way.
Publisher’s statement:
Please refer to the repository item page, publisher’s statement section, for further
information.
For more information, please contact the WRAP Team at: wrap@warwick.ac.uk.
2332-7804 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMBMC.2020.3009899, IEEE
Transactions on Molecular, Biological and Multi-Scale Communications
SUBMITTED TO IEEE TRANSACTIONS ON MOLECULAR, BIOLOGICAL, AND MULTI-SCALE COMMUNICATIONS 1
Molecular Signal Tracking and Detection Methods
in Fluid Dynamic Channels
Mahmoud Abbaszadeh, Iresha Atthanayake, Peter J. Thomas, and Weisi Guo, Senior Member, IEEE
Abstract—This method and data paper sets out the macro-scale
experimental techniques to acquire fluid dynamic knowledge
to inform molecular communication performance and design.
Fluid dynamic experiments capture latent features that allow
the receiver to detect coherent signal structures and infer trans-
mitted parameters for optimal decoding. This paper reviews two
powerful fluid dynamical measurement methodologies that can
be applied beneficially in the context of molecular signal tracking
and detection techniques. The two methods reviewed are Particle
Image Velocimetry (PIV) and Planar Laser-Induced Fluorescence
(PLIF). Step-by-step procedures for these techniques are outlined
as well as comparative evaluation in terms of performance
accuracy and practical complexity is offered. The relevant data
is available on IEEE DataPort to help in better understanding
of these methods.
Index Terms—molecular communication; experimentation;
macro-scale; fluid dynamics; PIV; PLIF.
I. INTRODUCTION
Experimental molecular communications (MC) is critical
for a number of research sectors, including: channel character-
ization [1]–[5], noise characterization for mutual information
analysis [6]–[9], and system design (e.g. receiver size, mobility
[8]). Experimental work is lacking at the macro-scale (≥
1mm), where molecular signals are subject to a variety of flow
associated processes, most of which are dynamic and inter-
related. Unlike the mass diffusion dominated regime (typically
in micro-scale, 1µm-1mm, and nano scale, ≤ 1µm), where the
channel and noise model are well understood even for different
modulation schemes [10]–[12], macro-scale continuum forces
make analysis challenging. Macro-scale research is useful
for a variety of underwater, gas/oil-pipe networks, chemical
engineering, and electromagnetically denied applications.
Research at macro-scale requires significant undertaking
and there is a growing body of work. Theoretical and simula-
tion work on molecular communications with turbulence has
shown that the fluid dynamic complexities cannot be ignored
[2], [13]. Experimentation is essential to capture realistic varia-
tional behaviour in fluid dynamics. It can enable us to 1) find
stable coherent structures in fluids that point towards better
M. Abbaszadeh, P. J. Thomas, and W. Guo are funded by the US AFOSR
grant FA9550-17-1-0056. (Corresponding author: Weisi Guo.)
M. Abbaszadeh and P. J. Thomas are with the School of En-
gineering, University of Warwick, Coventry CV4 7AL, U.K, (e-mail:
m.abbaszadeh@warwick.ac.uk; p.j.thomas@warwick.ac.uk)
I. U. Atthanayake is with the Mechanical Engineering Department, Open
University, Sri Lanka (e-mail: ireshairesha@yahoo.com ).
Weisi Guo is with the University of Warwick, Coventry CV4 7AL, U.K.,
also with The Alan Turing Institute, London NW1 2DB, U.K., and also with
the School of Aerospace, Transport and Manufacturing, Cranfield University,
Bedford MK43 0AL, U.K. (e-mail: weisi.guo@cranfeild.ac.uk).
modulation design (e.g., generated self-propagating structures
to increase symbol rate and transmission range [14]), and 2)
infer channel parameters to aid receiver signal decoding (e.g.
maximum likelihood estimation [15]).
An overview of experimental molecular communications is
given in [16]. Early prototyping experimental work started
with tabletop prototypes characterizing experimental through-
put [17]–[22] and noise processes [23] with crude chemical
sensors, which has now advanced to encoding in chemical
mixtures [24], [25] with mass spectrometer demodulation.
This coincides with parallel work in replicating pheromone
signals [26]. In our attempts to understand and improve the
achievable mutual information in macro-scale fluid dynamic
channels with complex processes, recent work (2017-19) has
characterized the evolving information structure in turbulence
[3] and tracked info-molecules using fluorescence [4], [27],
[28].
A. Key Metrics for Signal Measurement
In all the aforementioned work in macro-scale experimental
MC, measuring key attributes of molecular signals are essen-
tial. The key attributes are closely related to the manner in
which information is modulated to the molecular signal [1]. In
concentration-shift-keying (CSK), measuring the concentration
of the flow field is important. In pulse-position-modulation
(PPM) measuring the time of arrival difference between se-
quential pulses is important. We can generally either measure
the concentration directly, or measure the flow attributes to ex-
tract or infer the concentration, as well as other flow attributes.
It is worth mentioning that other chemical modulation schemes
that encode information in chemical structure (e.g. molecular
shift keying) and ratio of chemical mixtures in compounds
(e.g. isomer-based shift keying and pheromones), require a
mass spectrometer [24] or proprietary electronic nose [26].
In laminar flow, measured concentration or flow attributes
can be directly related to laminar flow parameters (e.g. flow
speed increases signal-to-noise ratio and throughput [4]). How-
ever, in turbulent flow, additional processes must be taken into
account in terms of the turbulence structure and the size of
the eddies. The terms eddy and vortex are used to describe the
swirling motion of a fluid. The word eddy is usually associated
with small-scale swirling entities whereas a vortex, such as a
large-scale tornado, contains eddies on smaller spatial scales.
Eddies are created, for instance, when a fluid is injected from a
syringe into another larger fluid volume due to the momentum
difference and shear stress at the fluid interface. If the injected
momentum is high enough mixing leads to a turbulent patch
Authorized licensed use limited to: University of Warwick. Downloaded on July 20,2020 at 07:28:26 UTC from IEEE Xplore. Restrictions apply.
2332-7804 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMBMC.2020.3009899, IEEE
Transactions on Molecular, Biological and Multi-Scale Communications
2 SUBMITTED TO IEEE TRANSACTIONS ON MOLECULAR, BIOLOGICAL, AND MULTI-SCALE COMMUNICATIONS
comprising eddies of different diameters. The diameter of
an eddy is its characteristic length scale. By analyzing the
size distribution of eddies within a turbulent flow one can
extract information about the features of turbulence. As the
turbulent patch evolves in time there exists a cascade involving
larger eddies breaking up in to smaller ones. This evolving
process leads to transport of energy from larger to smaller
scales, referred to as the energy cascade, until the energy is
dissipated at a critical length scale by the action of viscosity.
This smallest scale is known as the Taylor microscale. At
this scale fluid viscosity significantly affects the dynamics of
the turbulent eddies in the flow (see e.g. [29]). Small eddies,
near the Taylor microscale where viscous effects dominate,
do not contribute significantly to the transport of information.
This is considered as the lower bound for the molecular
communication capacity [3].
B. Novelty and Organization of Paper
Particle Image Velocimetry (PIV) and Planar Laser-Induced
Fluorescence (PLIF) are standard measurement techniques
in fluid dynamics [30], [31]. However, to the best of our
knowledge, this is the first MC paper to detail PIV and PLIF
methods for measuring flow field and concentration properties.
The protocol to apply these experimental methods and how to
post-process the acquired data are described in detail in this
paper.
This method and data paper builds on the authors’ own
work over the past 3 years funded by United States Air Force
Office for Scientific Research (US AFOSR), Defence Science
Technology Laboratory (DSTL), and EC H2020. We have
developed macro-scale experimental capability that can faith-
fully track molecular information to translate fluid-dynamic
knowledge in the context of MC.
We used our setup in Ref. [4], [9] to measure the infor-
mation rate in turbulent and laminar channel flows and to
characterize the noise model in these channels. In particular,
the setup in [4] uses the PIV method, whereas the PLIF
method is applied in [9]. Here, we discuss more general
aspects of the equipment and of the experimental procedures
such that corresponding studies can be reproduced more easily
and reliably elsewhere.
The paper is organized as follows: the experimental setup
is described in Section II. In Section III the PIV method for
the measurement of flow velocities is outlined together with
a brief discussion of alternative techniques. Correspondingly
Section IV introduces the PLIF method, and alternatives, for
the measurement of the concentration of fluorescent tracers
transported along with the flow. Section V and Section VI
contain, respectively, the discussion and the conclusion.
II. EXPERIMENTAL SET UP
A. Armfield Flume (Channel)
The Armfield flume (channel) used for our studies is
shown in Fig.1(a). The open channel is 15 m long, 0.3 m
wide, and 0.6 m deep. The channel has glass side walls
providing convenient optical access. The flow rate of the
channel is controlled by means of a pump. This enables
setting the mean flow velocity to a desired constant value. To
maintain a constant water level within the flume an inclined
plate is mounted at the channel outlet. The recirculating
liquid has to spill across this plate. By adjusting the angle of
the plate one can therefore set the water level inside the flume.
The injection system comprises a syringe-pump arrange-
ment illustrated in Fig. 1(b). The syringe pump is driven
by a computer-controlled stepper motor and enables ejecting
precisely defined volumes of fluorescent tracer liquid into the
flow field. The electronic components required to drive the
stepper motor, which is controlled by an Arduino program,
are contained in a small box.
The transmitter for the release of fluorescent tracer liquid is
displayed in Fig. 1(c). It constitutes a bent pipe with a diameter
of 5 mm. The inlet of the pipe is connected to the solenoid
valve and the outlet is located submersed under water inside
the channel. The transmitter is mounted such that its position
within the channel can be varied.
Fig. 1(d) displays the components forming the receiver of
our experimental set-up. The receiver comprises a laser to
generate a light sheet for the illumination of a cross section
of the flow and a video camera as a means for recording
the experiments for the subsequent data analysis using the
relevant methods. The purpose of these two individual receiver
components within the arrangement are discussed in detail in
Sections III and IV.
B. Reynolds Number
A Reynolds number is defined to characterize the overall
flow conditions within the channel. The physical meaning of
the Reynolds number is that it represents the approximate ratio
of inertial forces and viscous forces in the flow. For open-
channel flow the Reynolds number is defined as [32]:
Re =
u × L
ν
, (1)
where u is the mean flow velocity and ν is the kinematic
viscosity of the liquid. The quantity L represents the charac-
teristic length scale for open channel flow. It is equal to the
hydraulic radius of the channel given by A/p, where A is the
cross-sectional area of the flow and p is the wetted perimeter of
the channel [32]. For open channels the flow remains laminar
for, approximately, Re < 500 and it transitions to turbulent
flow above this critical value [32]. In our experiment, ν =
10
−6
m
2
s
−1
, u = 0.2 ms
−
1, and the Re=16000.
III. PARTICLE IMAGE VELOCIMETRY (PIV)
The most frequently used modern technique for the analysis
of flow fields is PIV. PIV is an optical technique used for
the measurement of flow velocities and it thereby provides a
tool for flow visualization. The method is referred to as non-
intrusive since it is not required to insert flow sensors in the
flow field that can potentially alter aspects of the dynamics
to be monitored. Technical details of the methodology are
Authorized licensed use limited to: University of Warwick. Downloaded on July 20,2020 at 07:28:26 UTC from IEEE Xplore. Restrictions apply.
2332-7804 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMBMC.2020.3009899, IEEE
Transactions on Molecular, Biological and Multi-Scale Communications
SUBMITTED TO IEEE TRANSACTIONS ON MOLECULAR, BIOLOGICAL, AND MULTI-SCALE COMMUNICATIONS 3
(a) Armfield Flume (Channel)
Glass
walls
0.6 m
(b) Injection System
(c) Transmitter
(d) Receiver
Fluorescence Tracer (Information Particles))
Power
Stepper
Motor
Box
Solenoid
Valve
Camera
Laser Sheet
Laser
Fluorescence
Tracer
(Rhodamine 6G)
Syringe
pump
Fig. 1: a) Armfield-flume (channel) configuration. It is 15 m long, 0.3 m wide, and 0.6 m deep, b) Injection system including
power, electronic boards and the syringe pump, c) Transmitter pipe, d) Receiver station.
summarized in [30]. The following steps are required for a
typical PIV measurement.
• Tracer: The flow is seeded with micron-sized tracer
particles. Ideally seeding particles are chosen that are
neutrally buoyant. This is required such that the particles
always faithfully follow the flow and therefore accurately
represent the flow velocity at their locations within the
flow field [33]. In liquids one can use, for instance,
hollow-glass spheres as seeding particles, which can be
silver-coated to increase their reflectivity as is the case in
the experiments of [4], [34]. If the flow is in a gas, then
it is not possible to satisfy the requirement for neutral
buoyancy and one typically uses, for instance, micron-
sized oil droplets as tracer particles.
• Imaging: A laser and an optical arrangement, comprising
a cylindrical lens, is used to generate a thin light sheet that
intersects the flow as illustrated in Fig. 2 (see Table I).
Tracer particles moving within the light sheet become
brightly illuminated. Thus, their motion can be recorded
by means of a video camera.
• Analysis: On any two successive video images taken
at a short time interval ∆t = t
0
− t apart the tracer
particles will appear at slightly shifted locations due to
their motion while following the flow field (see Fig. 2).
By analyzing from where to where particles have moved
within the image plane in the known time interval ∆t,
it is possible to infer the magnitude and the direction of
the particle velocity and the flow velocity. Once the basic
velocity field is known as a function of time, other quan-
tities such as time averaged velocities, the vorticity or
the turbulence characteristics can be obtained from post-
processing of the collected data. An in-depth example
showing such results from a PIV study is discussed in
[34].
A. Processing PIV images
1) Image processing: In order to process the PIV data,
the raw images (Fig. 3(a1)) taken in successive frames are
uploaded to the PIVlab software [35]. The Matlab based
PIVlab software is an open-source tool for the analysis and
post processing of PIV data. The full tutorial and MATLAB
application of PIVlab are at (https://pivlab.blogspot.com). The
Authorized licensed use limited to: University of Warwick. Downloaded on July 20,2020 at 07:28:26 UTC from IEEE Xplore. Restrictions apply.
2332-7804 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMBMC.2020.3009899, IEEE
Transactions on Molecular, Biological and Multi-Scale Communications
4 SUBMITTED TO IEEE TRANSACTIONS ON MOLECULAR, BIOLOGICAL, AND MULTI-SCALE COMMUNICATIONS
TABLE I: Summary of features of PIV and PLIF methods. Equipment needed for running experiment and websites for
purchasing those equipment.
Particle Image Velocimetry (PIV) Planar Laser-Induced Fluorescence (PLIF)
Features
• Non-intrusive
• Currently normally gives velocity in 2D plane but and 3D
methods are now becoming available
• Low resolution
• Hazardous (laser)
• Requires transparent liquid
• Non-intrusive
• Yields concentration distribution in a 2D plane
• Hazardous (laser)
• Requires transparent liquids
Equipment
• Laser (Pulsed frequency doubled Nd: YAG at 532 nm wave-
length)
• Cylindrical lens
• Camera (High-speed CMOS camera)
• Seeding particles (Hollow glass spheres, Polystyrene, Oxygen
bubbles)
• External hard drives
• Laser (Pulsed frequency doubled Nd: YAG at 532 nm wave-
length)
• Camera (High-speed CMOS camera)
• Notch filter for the camera (OD 6.0)
• Seeding particles (Rhodamine 6G)
• External hard drive
Website Dantecdynamics.com; TSI.com; Lavision.de; Oxfordlasers.com Dantecdynamics.com; TSI.com; Lavision.de; Oxfordlasers.com
Laser
Laser lens
Laser sheet
Illuminated
particles
Flow direction
with particles
Camera
Image plane
t
1st Frame
2nd frame
t+dt
Cross-correlation of
successive frames
Displacement vector for
each interrogation window
Fig. 2: Schematic of PIV set-up. A Laser sheet illuminates a
plane section of the flow. The flowing liquid contains small
tracer particles following the flow and therefore representing
the flow velocity. The particles are illuminated brightly by the
laser light when travelling within the light sheet. The camera
captures the motion of the tracer particles in successive frames.
Cross-correlation of successive images yields the velocity
field.
principles of PIV which are briefly summarized by means
of Fig. 3(a2) shows an image taken from a flow field. The
camera is focused on the area in the centre of the image, the
area of interest. Here the particles are in focus whereas they
are somewhat blurred towards the edge of the image frame.
Only the area of interest is considered for further analysis of
the flow field. To this end, the video images are divided into
small interrogation regions and corresponding interrogation
regions in successive video frames are compared. Figure 3(a3)
displays a PIV image that is divided into interrogation areas
of 15 × 20 pixels. A pair of two consecutive instantaneous
video images which are separated by time ∆t =0.01 s is
shown in Fig. 3(b). Each interrogation area of the second
image, taken at time t + ∆t, is shifted relative to the first
image taken at t. A cross correlation between the two images
is performed (see the MATLAB code and velocity vector
in Fig. 3(b)). That particular shift for which the correlation
function adopts its maximum is then taken as the direction
in which the particles within the small interrogation region
have collectively travelled. This direction together with the
time interval ∆t between the two video frames then defines
the magnitude and the direction of the flow velocity at the
location of the interrogation region [36].
2) Calibration: The PIV methodology requires calibration
to provide a conversion factor relating distances in terms of
pixels to their corresponding real-world distance in units of
length. For the calibration process a calibration image is used.
Figure 3(c1) displays a typical calibration image which, in this
particular case, consists of rows of black circles of constant
radius on a white background. The radius of the circles and
the distance between their centres represent known reference
lengths. Once the real distance of particle displacement is
known, a velocity vector for the considered interrogation area
is determined. A velocity-vector map over the whole image
area is obtained by repeating the procedure outlined in Section
III-A-1 for each interrogation area over the entire image.
Figure 3(c2) shows a typical velocity-vector field calculated
from a PIV image pair.
B. PIV in Comparison to Other Methods
There are alternative methods for velocity monitoring. One
other popular optical technique is Laser Doppler Anemometry
(LDA), frequently referred to as Laser Doppler Velocimetry
(LDV) [37]. Similar to PIV the LDV technique is a non-
intrusive methodology that requires optical access to the flow
field but no sensors within the flow field. Other common non-
optical methods are Hot-Wire Anemometry (HWA) [38], [39],
and Ultrasonic Velocity Profilers (UVP) [40].
However, for most applications the PIV technique has
advantages over all the other methods since it gives the 2D
flow field and it is also non-intrusive. Note that with more
sophisticated technical arrangements 3D flow field monitoring
by means of PIV is nowadays increasingly becoming a viable
option.
Authorized licensed use limited to: University of Warwick. Downloaded on July 20,2020 at 07:28:26 UTC from IEEE Xplore. Restrictions apply.
