ULTRA-LOW POWER ENERGY HARVESTING WIRELESS SENSOR NETWORK DESIGN by

ULTRA-LOW POWER ENERGY HARVESTING WIRELESS SENSOR NETWORK DESIGN by
ULTRA-LOW POWER ENERGY HARVESTING WIRELESS SENSOR NETWORK
DESIGN
by
ZHENG CHENYU
B.S., Kansas State University, 2012
A THESIS
Submitted in partial fulfillment of the requirements for the degree
MASTER OF SCIENCE
Department of Electrical and Computer Engineering
College of Engineering
KANSAS STATE UNIVERSITY
Manhattan, Kansas
2015
Approved by:
Co-Major Professor
Balasubramaniam Natarajan
Approved by:
Co-Major Professor
William B. Kuhn
Copyright
ZHENG CHENYU
2015
Abstract
This thesis presents an energy harvesting wireless sensor network (EHWSN) architecture
customized for use within a space suit. The contribution of this research spans both physical
(PHY) layer energy harvesting transceiver design and appropriate medium access control (MAC)
layer solutions. The EHWSN architecture consists of a star topology with two types of
transceiver nodes: a powered Gateway Radio (GR) node and multiple energy harvesting (EH)
Bio-Sensor Radio (BSR) nodes. A GR node works as a central controller to receive data from
BSR nodes and manages the EHWSN via command packets; low power BSR nodes work to
obtain biological signals, packetize the data and transmit it to the GR node.
To demonstrate the feasibility of an EHWSN at the PHY layer, a representative BSR node is
designed and implemented. The BSR node is powered by a thermal energy harvesting system
(TEHS) which exploits the difference between the temperatures of a space suit's cooling garment
and the astronaut's body. It is shown that through appropriate control of the duty-cycle in
transmission and receiving modes, it is possible for the transceiver to operate with less than
1mW power generated by the TEHS. A super capacitor, energy storage of TEHS, acts as an
energy buffer between TEHS and power consuming units (processing units and transceiver
radio). The super capacitor charges when a BSR node is in sleep mode and discharges when the
node is active. The node switches from sleep mode to active mode whenever the super capacitor
is fully charged. A voltage level monitor detects the system's energy level by measuring voltage
across the super capacitor.
Since the power generated by the TEHS is extremely low(less than 1mW) and a BSR node
consumes relatively high power (approximately 250mW) during active mode, a BSR node must
work under an extremely low duty cycle (approximately 0.4%). This ultra-low duty cycle
complicates MAC layer design because a BSR node must sleep for more than 99.6% of overall
operation time. Another challenge for MAC layer design is the inability to predict when the BSR
node awakens from sleep mode due to unpredictability of the harvested energy. Therefore, two
feasible MAC layer designs, CSA (carrier sense ALOHA based)-MAC and GRI (gateway radio
initialized)-MAC, are proposed in this thesis.
Table of Contents
Copyright ........................................................................................................................................ ii
Abstract .......................................................................................................................................... iii
List of Figures ................................................................................................................................ vi
List of Tables ................................................................................................................................. ix
Acknowledgements ......................................................................................................................... x
Chapter 1 - Introduction .................................................................................................................. 1
1.1 Project Background............................................................................................................... 1
1.2 Thesis Goals .......................................................................................................................... 3
1.3 Understanding EHWSN ........................................................................................................ 3
1.4 EHWSN Design Overview ................................................................................................... 4
1.5 Thesis Contributions and Organization ................................................................................ 4
Chapter 2 - Physical (PHY) Layer .................................................................................................. 6
2.1 Overall Description of Hardware System ............................................................................. 6
2.2 Daughter Board ..................................................................................................................... 9
2.2.1 Daughter Board Description .......................................................................................... 9
2.2.2 Sensors ......................................................................................................................... 10
2.2.3 Thermal Energy Harvesting System ............................................................................ 10
2.2.4 Testing Setup and Results ............................................................................................ 11
2.3 K-State-NASA Body Area Network Development Board ................................................. 17
2.3.1 Hardware Description of Mother Board ...................................................................... 17
2.3.2 Software Description of Mother Board ........................................................................ 19
2.3.3 Testing Results and Analysis ....................................................................................... 26
2.3.4 Duty Cycle Analysis .................................................................................................... 28
2.3.5 Summary of PHY Layer Design .................................................................................. 32
Chapter 3 - Recent MAC Layer Protocol Survey and New MAC Protocol Layer Design .......... 34
3.1 Survey of Energy Efficient MAC Protocols ....................................................................... 35
3.1.1 Energy Considerations ................................................................................................. 35
3.1.2 MAC Layer for Duty-cycled WSNs ............................................................................ 36
3.2 Type I MAC layer protocol design ..................................................................................... 41
iv
3.2.1 Attributes...................................................................................................................... 41
3.2.2 CS-ALOHA Based MAC Layer Design ...................................................................... 42
3.2.3 Simulation and Analysis .............................................................................................. 46
3.2.4 Design Drawbacks ....................................................................................................... 48
3.3 Type II MAC layer protocol design .................................................................................... 49
3.3.1 BSR Initiated Based MAC Layer Design .................................................................... 49
3.3.2 Simulation and Analysis .............................................................................................. 54
3.3.3 Design Drawbacks ....................................................................................................... 57
3.4 Conclusions ......................................................................................................................... 58
Chapter 4 - Conclusion and Future Work ..................................................................................... 59
4.1 Summary ............................................................................................................................. 59
4.2 Future Work ........................................................................................................................ 60
References ..................................................................................................................................... 61
Appendix A - Table of Acronyms ................................................................................................ 64
v
List of Figures
Figure 1-1: Field Tasks [2] ............................................................................................................. 2
Figure 1-2: Using commercial sensors to obtain EMG signals and ACC signals. [4].................... 2
Figure 1-3: Overview of a Star Topology Network ........................................................................ 4
Figure 2-1: Block Diagram of a BSR Node .................................................................................... 7
Figure 2-2: Block Diagram of a GR Node ...................................................................................... 8
Figure 2-3: Daughter Board Top View ........................................................................................... 9
Figure 2-4: Daughter Board Block Diagram .................................................................................. 9
Figure 2-5: Thermal Energy Harvesting System .......................................................................... 11
Figure 2-6: Measurement setup with power supply...................................................................... 12
Figure 2-7: Measurement set up with TEGs ................................................................................. 12
Figure 2-8: Combination of MB and DB ...................................................................................... 13
Figure 2-9: Testing set up model .................................................................................................. 13
Figure 2-10: Real testing set up .................................................................................................... 14
Figure 2-11: Charging time vs. Voltage across super capacitor (using 100mV power supply or
TEG at input to step up converter) ........................................................................................ 14
Figure 2-12: Charging current into super capacitor (using 100mV power supply vs. using TEGs)
............................................................................................................................................... 15
Figure 2-13: Output power measurement set up ........................................................................... 16
Figure 2-14: Top view of mother board ........................................................................................ 17
Figure 2-15: Block diagram of mother board ............................................................................... 17
Figure 2-16: Status register fields ................................................................................................. 19
Figure 2-17: Programming register fields ..................................................................................... 19
Figure 2-18: Software design overview for transmission mode ................................................... 19
Figure 2-19: Data Packet .............................................................................................................. 21
Figure 2-20: Command Packet ..................................................................................................... 21
Figure 2-21 Block Diagram of Receive Design ............................................................................ 22
Figure 2-22: Demodulation test results with 10kbit/s incoming data rate (the top line shows the
original incoming data and second line shows the data after demodulation) [13]................ 23
Figure 2-23: Demodulation test results with 30kbit/s incoming data rate (the top line shows the
original incoming data and second line shows the data out from demodulation) [13] ......... 23
vi
Figure 2-24: Block Diagram of Edge Detection Method.............................................................. 24
Figure 2-25: Block diagram of packet processing ........................................................................ 25
Figure 2-26: Time Sequence VS Current (S: Sleep mode; D: Data Sampling mode; AM: Active
mode) .................................................................................................................................... 26
Figure 2-27: Current measurement set up ..................................................................................... 26
Figure 2-28: Combinational test output (top yellow signal shows a transmitted packet and the
blue signal shows the voltage change across the super capacitor) ........................................ 27
Figure 2-29: Current flow during NSM ........................................................................................ 29
Figure 2-30: MDSR vs. Number of sample times during TTNSM (ACC sensor and EMG sensor)
............................................................................................................................................... 31
Figure 2-31: MDSR vs. Time spent in Rx modes (ACC sensor and EMG sensor)...................... 32
Figure 3-1: Packet Collision (part of packet overlaps) ................................................................. 35
Figure 3-2: S-MAC Sender-Receiver Communication................................................................. 37
Figure 3-3: Comparison of time line between B-MAC and X-MAC ........................................... 39
Figure 3-4: Time diagram of RI-MAC ......................................................................................... 40
Figure 3-5: Overview of a Star Topology Network ...................................................................... 42
Figure 3-6: BSR Node Process ..................................................................................................... 44
Figure 3-7: Receiver Node (GR node) Process............................................................................. 45
Figure 3-8: Timing Flow Diagram ................................................................................................ 45
Figure 3-9: Packet Collision Probability vs. Number of BSR Nodes ........................................... 47
Figure 3-10: Packet Collision Probability vs. Number of BSR Nodes with CS ........................... 47
Figure 3-11: Radio locations for intra-suit wireless propagation [27] .......................................... 48
Figure 3-12: Beacon Frame .......................................................................................................... 49
Figure 3-13: Concepts of GWRI-MAC Design ............................................................................ 50
Figure 3-14: Working Process of GR node................................................................................... 51
Figure 3-15: Working Process of BSR node................................................................................. 52
Figure 3-16: Timing Diagram of Scenario one ............................................................................. 52
Figure 3-17: Timing Diagram of Scenario two............................................................................. 53
Figure 3-18: Timing Diagram of Scenario three........................................................................... 54
Figure 3-19: Condition 1: Detect intended beacon within MLT .................................................. 55
vii
Figure 3-20: Condition 2: Detect intended beacon in extended listening time (Estimated sleep
time < 2xT_oscillation) ......................................................................................................... 56
Figure 3-21: Condition 2: Detect intended beacon in next waking-up time ................................. 56
Figure 3-22: ALT vs. Number of BSR nodes ............................................................................... 57
viii
List of Tables
Table 1-1: Energy harvesting techniques and harvested power [28] .............................................. 4
Table 2-1: Part Name and Part Number .......................................................................................... 6
Table 2-2: Sampling rate and number of sensors needed for an EHWSN within a spacesuit ...... 10
Table 2-3: Vout and its working range ......................................................................................... 11
Table 2-4: TEHS overall output power measurement .................................................................. 16
Table 2-5: Current measurement results ....................................................................................... 27
Table 2-6: PHY layer features summary ...................................................................................... 33
Table 3-1: Summary of Metric of CSA-MAC and GRI-MAC ..................................................... 58
ix
Acknowledgements
I would like to express my appreciation for the support and assistance of my major professors,
Dr. Bala Natarajan, Dr. William B. Kuhn and the other members of my graduate committee: Dr.
Dwight Day, and Dr. Don Gruenbacher. I would also like to express my thanks to my team
members: Charles Carlson, Levi Riley. Without their help, this project could not have been done.
Finally thanks to my parents and my friends for supporting me wholeheartedly throughout my
master’s program.
This work was supported by NASA/EPSCoR (NNX11AM05A). Any opinions, findings and
conclusions or recommendations in this material are those of the author(s) and do not necessarily
reflect the views of NASA.
x
Chapter 1 - Introduction
1.1 Project Background
The physical and mental health of astronauts on a space mission is often challenged by the harsh
operating environment. Various devices have been developed to track and predict astronaut's
health status in order to both maximize their working efficiency as well as to ensure their safety
[1]. Research for this thesis correlates to the NASA EPSCoR project (NNX11AM05A) whose
primary goal is to study the feasibility and implement a customized low power and low data rate
wireless sensor network (WSN) for use inside a space suit for biomedical application. The
project is divided into five tasks.
Task 1 determines physiological signals that can predict astronaut's fatigue during space
exploration mission and develop associated sensors to detect those physiological signals. Task 1
is a joint effort by Kinesiology department and ECE department to develop and utilize a
collection of field tests in the form of an obstacle course in which individual tests mimic typical
EVAs [2]. Field tasks include lifting weighted boxes, climbing ladders and stairs, and
transporting weighted objects in a wheelbarrow (see Figure 1-1). For this task effort,
commercial biomedical sensors (i.e., Delsys Trigno Electromyograph [3] and 3-aixs
accelerometer) were placed on meaningful spots on the body (see Figure 1-2) to acquire data for
further analysis.
Task 2 mainly focuses on testing and simulating the radio environment within a space suit.
Associated tasks include measuring path loss inside a space suit with different transmission
frequency range (315M, 433M, and 916M) and finding capable energy harvesting technologies
that can be used inside the space suit to provide electrical energy to the transceiver node
designed in Task 3.
Task3 is to develop an ultra-low power energy harvesting wireless sensor network (EHWSN) to
obtain the biological signals identified by Task1 and transmit the data to space suit backpack
1
computer for further analysis. Besides, energy harvesting (EH) transceiver node used in the
EHWSN is also designed in this task.
Task 4 is to create a compact package for the transceiver node designed in Task 3 and biosensors designed in Task 1.
Task 5 mainly focuses on public outreach.
Figure 1-1: Field Tasks [2]
Figure 1-2: Using commercial sensors to obtain EMG signals and ACC signals. [4]
2
1.2 Thesis Goals
The work in this thesis is primarily related to Task 3 of the project. The goals of this research
include designing an energy harvesting system (EHS) which can be used within a spacesuit,
developing an energy harvesting (EH) transceiver node which can work with an extremely low
duty cycle (≈ 0.4%), and presenting a feasible medium access control (MAC) layer design.
1.3 Understanding EHWSN
A wireless sensor network (WSN) is comprised of nodes that have processing units, sensors,
antenna, power source, and radio frequency integrated circuit (RFIC) as radio transceivers [14].
WSN typically use battery power as a power source while EHWSN use an energy harvesting
system (EHS) as a power source, converting energy from the environment to electrical energy to
power the sensor nodes. Energy from environment can be converted into electrical energy using
various energy harvesting (EH) techniques (see Table1-1). For example, photovoltaic cell that
converts solar energy to electrical energy is most widely used in our daily life. Solar keyboard
and solar calculator are two familiar applications of photovoltaic cell. Although the output power
from the EHS is extremely low (uW~mW) and various over time, with the development of ultra
low power electronics and energy storage techniques (e.g. low leakage super capacitor), EHWSN
becomes reality and attracts more and more researchers' attentions [5] [6] [7]. However,
according to literature, there are very few MAC layer designs which we can find suitable for the
EHWSN especially for the piezoelectric or thermoelectric techniques based EHWSN.
EHWSN is used for this thesis because 1) batteries are forbidden inside space suits because of
safety concerns, thereby preventing use of battery based WSN, 2) EHS does not require regular
replacement, thereby reducing nodes’ maintenance tasks, and 3) EHS is more environmentally
friendly compared to batteries [6][8]. Preliminary study in [9] finds that thermoelectric
generation is the most suitable EH method for using inside a spacesuit compared with other
viable methods (piezoelectric and automatic watches), although the method has low output
voltage (≈ 100𝑚𝑉) (solution for this drawback is shown in Section 2.2.3).
3
Table 1-1: Energy harvesting techniques and harvested power [28]
Harvesting techniques
Source
Efficiency
Harvested power
Photovoltaic
Ambient light
5 ~ 30%
10 uW/cm2 ~10 mW/cm2
Piezoelectric
Vibration/motion
1 ~ 10%
4 uW/cm2 ~100 uW/cm2
Thermoelectric
Thermal energy (human)
0.1~3%
30 uW/cm2
Antennas
RF (cell phone)
50 %
0.1 uW/cm2
1.4 EHWSN Design Overview
The EHWSN designed in this thesis follows a star topology, in which a gateway radio (GR)
works as central node and multiple bio-sensors radio (BSR) work as sub nodes (see Figure 1-3).
A subtle difference from traditional EHWSN in which all the nodes are EH nodes, is that the GR
node in our EHWSN is powered by the backpack battery of space suit and thereby not energy
constrained. Since BSR nodes powered by EHS are extremely power constrained, one major goal
of our MAC layer design is to shift power consumption pressure from BSR nodes to GR node as
much as possible, which is different from the constraints on existing MAC layer designs.
Figure 1-3: Overview of a Star Topology Network
1.5 Thesis Contributions and Organization
This thesis has following contributions:
4
First, we design a feasible thermal energy harvesting system (TEHS) that can be used within a
spacesuit. Since a thermal electric generator (TEG) needs a temperature differential to generate
electrical power, one approach is to exploit the difference between astronaut's body temperature
and the cooling garment that is worn. However, the key challenge is to ensure that there is
adequate physical contact between the TEG terminals and the body/cooling garment. A creative
solution to this problem has been designed and implemented as discussed in Section 2.2.4.
Second, we design and implement a transceiver node that can operate in energy constrained
environments. Specifically, the transceiver node works with extremely low duty cycle (≈ 0.4%)
by staying in sleep mode for most of the operational time. Analysis in Section 2.3.4 presents that
power generates during sleep (≈1mW) and power consumes (≈250mW) during active modes.
Both hardware and software design aspects of a transceiver node are discussed in Chapter 2.
Additionally, at the end of Chapter 2, we illustrate a way to calculate maximum sampling rate
which can be supported by our EH transceiver node. Based on the simulation results our system
can achieve as high as 4.5Hz sampling rate by using current EH transceiver node, although the
rate up to 300Hz may be feasible in the future as noted in the conclusion in section 4.2.
Finally, thanks to our special EHWSN construction with one power-unconstrained node (GR
node) and multiple extremely power-constrained nodes (BSR nodes), very few existing media
access control (MAC) layer design can be directly used in our system. Two feasible MAC layer
designs CSA (carrier sense ALOHA based)-MAC and GRI (gateway radio initialized)-MAC, are
present for our EHWSN in Chapter 3. Since the wake up time of EH transceiver node (BSR
node) is unpredictable, both MAC layer designs are asynchronous designs in which BSR nodes
have not to transmit data packets in fixed time slots. In CSA-MAC, if the channel is detected to
be clear, a BSR node can transmit a packet whenever the node is ready and a GR node keep
listening to the channel all the time. In GRI-MAC, a GR node always keeps transmitting beacons
to inform an awake BSR node to transmit a data packet. Both MAC designs are modified from
existing MAC layer designs to shift the power consumption pressure from power-constrained
nodes (BSR node) to power-unconstrained node (GR node). The details of MAC layer design are
provided in Chapter 3.
5
Chapter 2 - Physical (PHY) Layer
2.1 Overall Description of Hardware System
The hardware system of the energy harvesting wireless sensor network (EHWSN) studied in this
thesis consists of two types of transceiver nodes: Bio-Sensor Radio (BSR) node and Gateway
Radio (GR) node.
BSR Node
Multiple BSR nodes, which are placed on various locations on the human body to obtain
different types of vital signals, are present in the EHWSN. After collecting biomedical signal,
digitizing them, and converting the signals to data packets, BSR nodes send the packets to the
GR node. A BSR node is comprised of two main parts (see Figure 2-1): a daughter board (DB)
and K-State-NASA Body Area Network Development Board (KANDB), also known as the
mother board (MB). Daughter board tasks include providing power to an entire BSR node,
acquiring vital signals from human body, and being a part of the antenna as discussed in [9];
mother board tasks include converting and storing incoming data from daughter board, building
and transmitting a data packet to GR node, managing power utilization, and responding to the
incoming command packet from GR node. Major components on the BSR node are shown in
Table 2-1.
Table 2-1: Part Name and Part Number
Daughter Board
Part Name
Part Number
Accelerometer Sensor
ADXL330
Ultralow Voltage Step-up Converter
LTC3108
Super Capacitor
AVX BestCap (0050mF/5.5V)
Mother Board
Part Name
Part Number
Field Programmable Gate Array (FPGA)
Actel-AGL1000
Microcontroller (uC)
ATmega 1284P
RF Micro-transceiver (RFIC)
K-State research IC
6
Figure 2-1: Block Diagram of a BSR Node
7
GR Node
Only one GR node is present in each EHWSN. The primary function of the GR node (see Figure
2-2) is to receive and decode data packets from BSR nodes and then transmit the data part to a
computer in the backpack of a space suit. A GR node acts as the WSN organizer via command
packets. It consists of a mother board only because the GR node is directly powered by the
space suit’s backpack and the node does not need to obtain vital signals.
Figure 2-2: Block Diagram of a BSR Node
Since the BSR node and GR node have similar hardware and software design and EHWSN
efficiency is limited by BSR node's performance, the remainder of this chapter primarily focuses
on the BSR node. The daughter board is discussed first followed by a description of the mother
board.
8
2.2 Daughter Board
2.2.1 Daughter Board Description
As mentioned, the daughter board has two main tasks: obtain vital signals from the human body
using a built-in sensor and provide enough power to KANDB using the Energy Harvesting (EH)
System. In addition, the daughter board contains antenna circuit as discussed in [9]. A top view
of daughter board is shown in Figure 2-3 and a block diagram is shown in Figure2-4.
Figure 2-3: Daughter Board Top View
Figure 2-4: Daughter Board Block Diagram
9
2.2.2 Sensors
Various sensors are needed to monitor the health condition of an astronaut. Sensor types and
their reasonable sampling rates are shown in Table 2-2.
Table 2-2: Sampling rate and number of sensors needed for an EHWSN within a spacesuit
Sampling Rate
Types of Sensor
(samples/second)
units
Electromyography (EMG)
500
Multiple (TBD)
Electrocardiogram (ECG)
250
1
Respiration Rate Senor (RRS)
200
1
Pulse Oximeter (PO)
240
1
Accelerometer (ACC)
50 ~ 200
Multiple (TBD)
2.2.3 Thermal Energy Harvesting System
As shown in Figure 2-5, the Thermal Energy Harvesting System (TEHS), the core part of a
daughter board, consists of a thermal electric generator (TEG), voltage step-up converter, and
super cap (50mF).
When body temperature is applied to the “hot” side of a TEG and low temperature is applied to
the “cold”, the TEG can be used as an energy convertor to transfer thermal energy to electrical
energy. Larger temperature difference between two sides of the TEG will produce higher output
power [9]. Since the output voltage of TEGs is too low to be used (approximately 90 mV) to
power the mother board directly, an ultralow-voltage step-up converter (UVSC) must be
employed. UVSC can operate from inputs at least 20 mV and provide selectable output voltage
(Vout) of 4.1V, or 5V [10]. The UVSC output (Vout1) connects with a super capacitor which
acts as the energy storage element of this system. The energy condition is monitored by
measuring voltage across the super capacitor using an A to D converter on the mother board and
also within the UVSC on the daughter board. Different Vout settings have correspondingly
different working ranges (see Table2-3). When Vout (voltage across the super capacitor) falls
10
below the working range, a switch in the UVSC turns off in order to shut down the mother board.
For example, if Vout is set to 4.1V, the switch is turned off when voltage across the super
capacitor falls below 3.7V. The switch is turned back on when voltage across the super capacitor
rises above 3.8V. Therefore, voltage of the super capacitor must stay within working range to
ensure smooth operation of the entire system. The mother board monitors the voltage to regulate
the voltage and avoid this condition.
Figure 2-5: Thermal Energy Harvesting System
Table 2-3: Vout and its working range
Vout
Working Range
4.1V
3.8V ~ 4.1V
5.0V
4.6V ~ 5.0V
2.2.4 Testing Setup and Results
1. Charging Current with Power supply
Because duty cycle is limited by the ratio of charging current to the super capacitor with the
mother board in sleep mode and discharging current consumed by mother board in active mode
(see Section 2.3.4), charging current is the key parameter in order to evaluate the TEHS. To
access the maximum charging current possibly with the UVSC, a power-supply with 100mV was
used to mimic the TEG (see Figure 2-6). Because charging current is difficult to measure
directly, the current was found by using Equation 2.2.4-1.
i=
𝑑𝑣
𝑑𝑡
C,
11
(2.2.4-1)
where 𝑣 is the voltage across the super capacitor, t is the time of super capacitor charging, and C
is the value of the super capacitor (50mF). After UVSC is powered by the power supply, super
capacitor begins charging. t is recoded starting when 2V is observed at the testing point with
steps of 0.2V and stops being recorded when the super capacitor is fully charged (5V).
Measured results of 𝑣 and t are shown in Figure 2-11. The charging current found using Equation
2.2.4-1 at different charging voltages is shown Figure 2-12.
Figure 2-6: Measurement setup with power supply
2. Charging Current with TEGs
In this test, two paralleled TEGs are used as a power source to power the voltage step-up
converter (see Figure 2-7). Since TEG needs temperature difference to create output power, the
hot side of the TEG should touch the high temperature object (body skin) and the cold side
should touch the low temperature object (ice bag). However, TEGs are located between mother
board and daughter board (see Figure 2-8) made from thermal isolation material, thereby making
TEGs impossible touching with high or low temperature object directly.
Figure 2-7: Measurement set up with TEGs
12
Figure 2-8: Combination of mother board and daughter board
A model designed to provide maximum thermal flow is shown in Figure 2-9. On one side, a
0.019 Inch thickness copper plate bent with U shape to bypass the mother board connects cold
side of TEG with ice bag (25 ̊ F); on the other side, thermal starting from arm (91 ̊ F) going
through another 0.076 Inch thickness copperplate and vias filled with thermal paste to the hot
side of TEG. Since the copper is thermal conductor with very high thermal conductivity (401 W/
(m K)) and is reasonable thick, very little temperature is changed when thermal flows through
copperplate. A testing set up is shown in Figure 2-10. Measuring steps are similar as above and
the measurement results are shown in Figure 2-11, and Figure 2-12. From the testing results we
can see that, when testing with TEGs, it needs around 8 minutes to charge the super capacitor
from 2V to 5V, and the average charging current is around 250uA. The testing with TEGs has
better performance probably because the TEGs have lower output impedance than the power
supply at the low voltages used.
Figure 2-9: Testing set up model
13
Figure 2-10: Real testing set up
Figure 2-11: Charging time vs. Voltage across super capacitor (using 100mV power supply or
TEG at input to step up converter)
14
Figure 2-12: Charging current into super capacitor (using 100mV power supply vs. using TEGs)
3. TEHS Overall Output Power
To further assess the thermal energy harvesting output capabilities of the TEGs, a load resistor
was placed across the super capacitor to determine average power available with 100% duty
cycle operation (Figure 2-13). One of three conditions must happen when adjusting load resistor
value: 1) the voltage across super capacitor decreases, indicating that load power is greater than
TEHS overall output power; 2) the voltage increases, indicating that load power is less than
TEHS overall output power; 3) the voltage remains stable, indicating that load power equals to
TEHS overall output power. The resistor value which keeps the voltage remaining stable is
recorded at Table 2-4 and the TEHS overall output power can be calculated by 𝑃𝑇𝐸𝐻𝑆 = 𝑃𝑙𝑜𝑎𝑑 =
𝑉 2 /𝑅𝑙𝑜𝑎𝑑 .
15
Figure 2-13: Output power measurement set up
Table 2-4: TEHS overall output power measurement
Load resistor
Voltage at testing point
Load power
TEHS output power
(𝑅𝑙𝑜𝑎𝑑 )
(𝑉)
(𝑃𝑙𝑜𝑎𝑑 )
(𝑃𝑇𝐸𝐻𝑆 )
28kΩ
4.95V
0.87mW
0.87mW
Based on previous test results (with TEGs), the output power is about 0.94mW which is a little
bit higher than 𝑃𝑇𝐸𝐻𝑆 measured in this test. Considering the temperature varies over time, this
little error is acceptable. However, the temperature of ice bag is around -4 ̊C, which must be
lower than the temperature of cooling garment, so further test within a real space suit is needed
to get more realistic 𝑃𝑇𝐸𝐻𝑆 .
16
2.3 K-State-NASA Body Area Network Development Board
2.3.1 Hardware Description of Mother Board
The mother board, shown in Figure2-14 (top view) and Figure2-15 (block diagram), mainly
contains two regulators, a microcontroller (uC), a RFIC, and a field programmable gate array
(FPGA).
Figure 2-14: Top view of mother board
Figure 2-15: Block diagram of mother board
17
Regulators
Two regulators in the mother board work with input voltage range spans of 3.7V to 5.0V.
Regulator 1 converts incoming voltage to 3.3 V to power the uC, the FPGA, and the RFIC.
Regulator 2 converts incoming voltage to 1.5V to power the FPGA core logic. Both regulators
contain an enable pin that can be used to activate or deactivate the output voltage. The 3.3V
regulator is active constantly to power uC and draws a quiescent current of approximately 2uA
[11]. The enable pin of the 1.5V regulator is controlled by the uC to power down the FPGA
during sleep.
uC
Three main tasks of the uC include operating as a power manager, functioning as an analog to
digital converter (ADC), and working as a data storage buffer. The uC can individually turn the
power supply on or off for each component (Regulator2, RFIC, FPGA, and sensor) on a BSR
node. When acting as a power manager, the uC places board in different types of modes (details
in Section 2.3.2) according to the current energy level that it monitors by measuring the voltage
across the super capacitor on the BSR daughter board. When acting as an ADC and a storage
buffer, the uC converts input analog signals to 8-bit digital signals and then stores the digital
signals in the data buffer, as described in Section 2.3.2.
FPGA
FPGA has four main functions: data processing, packet formation, bi-direction communication
with uC, and RFIC programming to build or decode a data packet. See Section 2.3.2 for details
for each function.
RFIC and Energy Detection
The RF integrated circuit (RFIC) is used to modulate and demodulate incoming data onto 433
MHz carrier. The FPGA controls the RFIC by setting value of the 61-bit programmable register
(see Figure 2-17). When the RFIC is set to “receive” mode, three bits of the status register (see
Figure 2-16) can be used as a received signal strength indicator (RSSI) to indicate the link
quality.
18
Figure 2-16: Status register fields
Figure 2-17: Programming register fields
2.3.2 Software Description of Mother Board
2.3.2.1 Transmission mode
The primary functions of the mother board software include controlling the active mode/sleep
mode duty cycle, sampling sensor data, creating data packets, and controlling the RFIC to
transmit and receive these packets. In transmit mode, the software operation is described in
Figure 2-18.
Figure 2-18: Software design overview for transmission mode
A. ADC and On-chip Data Buffer
Since packet size reduction and sufficient precision (precision = V_ref/2bits ) are goals, an 8-bit
ADC (12mV precision with 3.3V reference voltage) is employed in this design. After one ADC
conversion, an 8-bit byte is stored in the on-chip data buffer. For specific type of sensor, it needs
19
to take more than one value per sample time. For example, an accelerometer (ACC) sensor needs
to obtain 3 values (x-axis, y-axis, and z-axis) per sample, thereby 3 bytes stored in the on-chip
data buffer for each sample time. Maximum buffer size is set to 256 bytes. Buffer size could not
be too large (e.g. over 300 bytes) because data must be transmitted within 0.4s or less due to
working range of UVSC (see Table 2-3) and high Tx mode current consumption (51mA). Buffer
size could not be too small (e.g. less than 50 bytes) in order to avoid transmit efficiency
reduction due to packet having fixed preamble length and CRC bits. When the super capacitor is
fully charged, uC sends data stored in the on-chip data buffer to FPGA via serial peripheral
interface (SPI) for creating a packet.
B. Packet Creation and RFIC Programming
Packet formats
In wireless communication, data is most often transmitted in the form of packets. Two types of
packets are used in this design: data packets (see Figure2-19) and command packets (see Figure
2-20). Each type is formatted to achieve good tradeoff between performance and data
transmission efficiency when operating at extremely low power.
(1) Data Packet:
A data packet consists of:

4-byte sync word: "010101..." sequence and a "111" to indicate the end of the sync word

1-byte ID &Type: 6-bit ID that can define up to 64 BSR nodes and 2-bit type always ‘10’
for data packet

1-byte size: identifies the number of bytes in the "data" (up to 256)

4 to 256-byte data payload

4-byte CRC
20
Figure 2-19: Data Packet
(2) Command Packet:
A command packet consists of:

4-byte sync word: "010101..." sequence and "111" to indicate the end of the sync word

1-byte ID & Type & Parity Check: 6-bit ID that can define up to 64 sensors, 1-bit type
always ‘0’ for command packet, and 1-bit parity check

1-byte command & Parity Check: 7-bit Command (up to 128 commands) and 1-bit parity
check
Figure 2-20: Command Packet
The “Sync Word” is a moderate length sequence of “0101...” ending with “111” to allow a
receiver to synchronize to, confirm the incoming packet, and locate the beginning field. In [12]
[13], extensive analyses have been done to decide the necessary length of sync word. Since
multiple transmitters (BSR nodes) are used, the ID defines the “sending source” of the packet.
Two types of packet are used in the system: one for transmitting data and another for
transmitting command. “Type” determines whether a packet is a command packet or a data
packet. If a packet is a data packet, the “Size” indicates the number of bytes of data (command
packet is not well defined). “Data” stores data coming from on-chip data buffer. The last part of
21
data packet is Cyclic Redundancy Check (CRC). If a packet is a command packet, the command
and parity check field are combined to minimize the packet size.
2.3.2.2 Receiving Mode
The software design of the receiving mode focuses on the processing of incoming data after the
demodulation process is performed by the FPGA and is the subject of a parallel research project
covering demodulation bit synchronization, and packet detection [13]. The primary functions of
receiving mode software include synchronization of incoming data, detection of the sync word,
and packet processing. An overview block diagram of the design is shown in Figure 2-21.
Figure 2-21 Block Diagram of Receive Design
Synchronization of incoming data
Due to limitations of the ultra-low power demodulation method used, data out of the
demodulation step may be slightly distorted and with the incoming data rate increased, the
distortion becomes more and more serious (see Figure 2-22 and Figure 2-23) [13].If incoming
data is continuously sampled by the software design without taking corrective action, distortion
accumulates and bit error occurs. Distortion accumulation is avoided by calibrating the start time
of sampling whenever an edge (either positive of negative edge) of incoming data is detected
(See 2-24). However, errors still occur if incoming data rate is high. Based on the test in [13],
probability of error reception increases from 1% to 10% with incoming data rate changing from
10kbits/s to 30kbits/s. For this reason, the data rate used in this project is currently limited to
10kbits/s.
22
Figure 2-22: Demodulation test results with 10kbit/s incoming data rate (the top line shows the
original incoming data and second line shows the data after demodulation) [13]
Figure 2-23: Demodulation test results with 30kbit/s incoming data rate (the top line shows the
original incoming data and second line shows the data out from demodulation) [13]
23
Figure 2-24: Block Diagram of Edge Detection Method
Detection of sync word
Two main steps detect the sync word:

Verification that incoming data is a sync word

Determination of the end of sync work
As discussed in Section 2.3.2 Packet format, a sync word has a long sequence of “0101…,”
allowing distinction between sync word and noise. When a pattern of “01” repeated eight times
is detected, a sync word is assumed; then a pattern of “111” must be detected to indicate the end
of sync word and begin data processing.
Packet processing
The incoming packet is decoded by following the order of the packet fields. The block diagram
of packet processing is shown in Figure 2-25. First, the ID field is checked to ensure that the
packet originates from the desired node. Then, by checking the “Type” of the packet, a
determination must be made as to whether the packet is a command or a data packet. When a
command packet is received and the packet passes parity check, the command will be further
processed; when a data packet is received and passes CRC check, data reception is confirmed.
When an incoming packet does not pass the check (parity or CRC check), the receiver node
24
transmits a Re-send command packet to the originating node inform it to re-transmit the previous
packet.
Figure 2-25: Block diagram of packet processing
Operation modes and duty cycle considerations
A BSR node operates in three modes: sleep mode (SM), data sampling mode (DSM), and active
mode (AM) (Figure 2-26). During sleep mode, uC works in power-save mode with an external
low frequency crystal controlled 32 KHz clock to minimize power use while maintaining
accurate timing. All other components, including RFIC, Regulator2, and FPGA, are turned off.
The current in sleep mode (IS ) is approximately 6 ~ 10uA (data sheet) and the period of SM (TS )
is around 1 over the sensor data sampling rate ( 𝑓𝑑𝑠 ). When SM is finished, the system
immediately proceeds to DSM. The task of DSM is to convert analog signals from the daughter
board to 8-bit data words and store the data in an on-chip buffer. In DSM, the microcontroller
works with an 8MHz internal clock, and the current ( IDS ) of the system increases to
approximately 4 mA (data sheet). The period of DSM (TDS ) is very short (25~75 us) compared to
the SM (approximately 1 second). When DSM is finished, the system returns to SM. The system
proceeds to AM when the super capacitor is fully charged. During AM, the uC turns on the
FPGA and RFIC in order to transmit (Tx) or receive (Rx) a packet as previously described. The
period of AM (Tam ) can be calculated by
1
Tam = 𝐿𝑝𝑎𝑐𝑘𝑒𝑡 (𝑓 ) + TRx
𝑡𝑥
(2.3.4-1)
where 𝑓𝑡𝑥 is transmit bit rate, 𝐿𝑝𝑎𝑐𝑘𝑒𝑡 is the packet length, and TRx is node listening time. If a
BSR is a transmit-only node, the TRx would be zero. However, for the protocols discussed in
Chapter 3, a non-zero TRx would be used.
25
Figure 2-26: Time Sequence VS Current (SM: Sleep mode; DSM: Data Sampling mode; AM:
Active mode; NSM: Nominal sleep mode)
Since the data sampling and active mode currents are much higher than the charging current to
the energy harvesting storage capacitor, the most difficult part of design is achieving acceptable
bio sensor data rate. To quantity this constraint, we next look at the current consumption level of
the major components.
2.3.3 Testing Results and Analysis
1. Current consumption of different modes
Figure 2-27 depicts the measurement set up. Measured current consumptions for different modes
are recorded in Table 2-5
Figure 2-27: Current measurement set up
26
Table 2-5: Current measurement results
Current
Mode Types
sleep mode (SM)
Components Turned on
uC(sleep),Regulator 1&2,
(measurement)
8uA
Data sampling mode(DSM) uC(active), ACC sensor
4.7mA
Tx mode
uC(active), FPGA, RFIC, Regulator 1&2
51mA
Rx mode
uC(active), FPGA, RFIC, Regulator 1&2
30mA
2. Testing with combined mother board and daughter board
In this test, instead of a power supply, the daughter board was employed to power the mother
board. A test point was put on the super capacitor to monitor the voltage changing when the
mother board operated in different mode. Figure 2-28 shows that the tested voltage decreased
when a BSR node ran in AM (Tx mode at this test) and increased with much slower rate when it
returns to sleep mode. The packet transmission last 89.6 ms with 10kbits/s transmission rate and
the voltage across the super capacitor drops approximately 98mV, from which the current
consumption of packet transmission can be calculated to be 53.4mA.
Figure 2-28: Combinational test output (top yellow signal shows a transmitted packet and the
blue signal shows the voltage change across the super capacitor)
27
2.3.4 Duty Cycle Analysis
A. Duty Cycle Calculation
To simplify duty cycle calculation, SM and DSM can be combined as a Nominal Sleep Mode
(NSM). The total time of the NSM (TTNSM) is N(Ts + Tds ) ≈ N Ts , where N is the number of
DSMs between two AM. The duty cycle is then calculated as
Duty Cycle =
T𝑎𝑚
N(Ts +Tds )+T𝑎𝑚
≈
T𝑎𝑚
NTs
, (Ts ≫ Tds , NTs ≫ T𝑎𝑚 )
(2.3.4-1)
To find the period of active mode (T𝑎𝑚 ), Equation (2.2.4-1) can be written as
dt =
𝑑𝑣
𝑖
C
(2.3.4-2)
where 𝑑𝑣 is the voltage difference (∆𝑉𝑑𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑒 ) before and after AM mode, 𝑖 is the AM current
(𝐼𝑎𝑚 ), C is the value of the super capacitor (50mF), and dt is the discharging time of the super
capacitor equaled to the period of AM (T𝑎𝑚 ):
T𝑎𝑚 =
∆𝑉𝑑𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑒
𝐼𝑎𝑚
C
(2.3.4-3)
Similarly, the total time of NSM (NTs) can be written as
∆𝑉𝑐ℎ𝑎𝑟𝑔𝑒
N Ts = 𝐼
C
𝑐ℎ𝑎𝑟𝑔𝑒
(2.3.4-4)
where ∆𝑉𝑐ℎ𝑎𝑟𝑔𝑒 is the voltage difference before moving into NSM and after leaving NSM to
AM, and 𝐼𝑐ℎ𝑎𝑟𝑔𝑒 is the charging current of the super capacitor during NSM. The super capacitor
must be fully charged before moving into AM in order to compromise the power consumed
during AM. Therefore,
∆𝑉𝑐ℎ𝑎𝑟𝑔𝑒 = ∆𝑉𝑑𝑖𝑠𝑐ℎ𝑎𝑟𝑔𝑒
(2.3.4-5)
When Equations (2.3.4-3), (2.3.4-4), and (2.3.4-5) are combined, the new expression of duty
cycle is
Duty Cycle =
T𝑎𝑚
N Ts
=
𝐼𝑐ℎ𝑎𝑟𝑔𝑒
𝐼𝑎𝑚
(2.3.4-6)
When supply current comes out from the daughter board (𝐼𝐷𝐵 ), it splits into two branches: one
branch goes to the mother board to maintain NSM (𝐼𝑁𝑆𝑀 ) and another one goes to charge the
super capacitor (𝐼𝑐ℎ𝑎𝑟𝑔𝑒 ) (Figure 2-29).
28
Figure 2-29: Current flow during NSM
According to Figure 2-29,
𝐼𝑐ℎ𝑎𝑟𝑔𝑒 = 𝐼𝐷𝐵 − 𝐼𝑁𝑆𝑀
(2.3.4-7)
where 𝐼𝑁𝑆𝑀 can be calculated by
𝐼𝑁𝑆𝑀 =
Is Ts + Ids Tds
Ts +Tds
≈
Is Ts + Ids Tds
Ts
1
= Is + Ids Tds (T )
s
(2.3.4-8)
Because the SM period equals to the time between two DSMs (Ts = 1/𝑓𝑑𝑠 ), Equation (2.3.4-6)
can be written as
𝐼𝑁𝑆𝑀 = Is + Ids Tds 𝑓𝑑𝑠
(2.3.4-9)
B. Maximum Data Sampling Rate Analysis without Rx Mode
The maximum data sampling rate (MDSR), which defines the maximum data that can be taken
during TTNSM and successfully be transmitted during AM, is an indicator of system capability.
According to Equations (2.3.4-9) and (2.3.4-7), when the data sampling rate (𝑓𝑑𝑠 ) is increased,
𝐼𝑐ℎ𝑎𝑟𝑔𝑒 is decreased. Therefore, the harvested energy (𝐸ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑 ) decreased with fixed
harvesting time. In addition, the energy consumed (𝐸𝑐𝑜𝑛𝑠𝑢𝑚𝑒𝑑 ) during AM mode increased
because more data obtained during TTNSM period required transmission. Because 𝑓𝑑𝑠 cannot be
too high to make 𝐸𝑐𝑜𝑛𝑠𝑢𝑚𝑒𝑑 exceed 𝐸ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑 , 𝑓𝑑𝑠 was limited by the following inequality
𝐸ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑 ≥ 𝐸𝑐𝑜𝑛𝑠𝑢𝑚𝑒𝑑
1
(2.3.4-10)
Given an energy harvesting time (N Ts = N(𝑓 )), an energy harvesting current (𝐼𝑐ℎ𝑎𝑟𝑔𝑒 ), and an
𝑑𝑠
EHS output voltage (𝑉𝑜𝑢𝑡 ), the 𝐸ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑 can be derived as
29
𝐸ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑 = 𝑉𝑜𝑢𝑡 (N Ts )𝐼𝑐ℎ𝑎𝑟𝑔𝑒
(2.3.4-11)
When Equation (2.3.4-9) and Equation (2.3.4-7) were substituted into Equation (2.3.4-11), a new
expression of 𝐸ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑 was evolved. 𝑓𝑑𝑠 is shown as
𝐸ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑 = 𝑉𝑜𝑢𝑡 N(
𝐼𝐷𝐵 −𝐼𝑆
𝑓𝑑𝑠
− IDS TDS )
(2.3.4-12)
Because BSR nodes are power-constrained, ideal energy utilization uses all the energy to
transmit data without wasting energy to listen to the channel. The expression of Econsumed under
this ideal condition (without considering Rx mode) is shown as
Econsumed = Vin Tam Iam = V𝑖𝑛 Ttx Itx
(2.3.4-13)
where Vin is input voltage of mother board that equals to 𝑉𝑜𝑢𝑡 , Itx is the Tx mode current
consumption, and Ttx is the packet transmission time which can be calculated as
1
Ttx = (𝐿𝑝𝑟𝑒𝑎𝑚𝑏𝑙𝑒 + 𝐿𝑑𝑎𝑡𝑎 )(𝑓 )
(2.3.4-14)
𝑡𝑥
where Lpreamble is the length of preamble part (bits) of a packet and Ldata is the length of data
part (bits) of a packet. Ldata is decided by the number of data sampled during TTNSM and
calculated by
𝐿𝑑𝑎𝑡𝑎 = 8𝑛N
(2.3.4-15)
where n, which is variously based on the type of sensor, is the total number of data sampled
during each DSM. For example, the accelerometer sensor at each DSM must obtain three data
for x, y, and z axis, and ECG sensor obtains one data at each DSM. There is an 8 in Equation
(2.3.4-15), because of the use of an 8-bit ADC. When Equations (2.3.4-15) and (2.3.4-14) are
substituted into Equation (2.3.4-13), a new expression of 𝐸𝑐𝑜𝑛𝑠𝑢𝑚𝑒𝑑 can be written as
1
Econsumed = Vin Itx (𝐿𝑝𝑟𝑒𝑎𝑚𝑏𝑙𝑒 + 8𝑛N)(𝑓 )
𝑡𝑥
(2.3.4-16)
𝑓𝑑𝑠 was then computed by substituting Equations (2.3.4-12) and (2.3.4-16) into inequality (2.3.410):
𝑓𝑑𝑠 ≤
(𝐼𝐵𝐷 −𝐼𝑆 )𝑁𝑓𝑡𝑥
𝐼𝑡𝑥 (𝐿𝑝𝑟𝑒𝑎𝑚𝑏𝑒𝑙 +8𝑛𝑁)+𝑁𝑓𝑡𝑥 IDS TDS
(2.3.4-17)
Therefore, the MDSR can be derived as
𝑓𝑑𝑠_𝑚𝑎𝑥 =
(𝐼𝐵𝐷 −𝐼𝑆 )𝑁𝑓𝑡𝑥
𝐼𝑡𝑥 (𝐿𝑝𝑟𝑒𝑎𝑚𝑏𝑒𝑙 +8𝑛𝑁)+𝑁𝑓𝑡𝑥 IDS TDS
30
(2.3.4-18)
In Equation (2.3.4-18), two variables, 𝑛 and 𝑁, affect the maximum data sampling rate (MDSR).
Figure 2-30 shows that, with the number of DSM during the total time of the NSM (TTNSM)
increased (𝑁), MDSR is increased and approaching to a fixed value. A BSR node with EMG
sensor (𝑛 = 1) has 3 times higher MDSR than a BSR node with ACC sensor (𝑛 = 3) due to the
single byte converted in the EMG sensor case.
Figure 2-30: MDSR vs. Number of sample times during TTNSM
(ACC sensor and EMG sensor)
When the Rx mode was considered, a new expression of Econsumed was derived as
Econsumed = V𝑖𝑛 (Ttx Itx + Trx Irx )
(2.3.4-19)
𝑓𝑑𝑠 can be computed as
𝑓𝑑𝑠 ≤
(𝐼𝐵𝐷 −𝐼𝑆 )Nmax 𝑓𝑡𝑥
𝐼𝑡𝑥 (𝐿𝑝𝑟𝑒𝑎𝑚𝑏𝑒𝑙 +8𝑛Nmax )+𝑓𝑡𝑥 (Nmax IDS TDS+ TrxIrx )
(2.3.4-20)
Therefore, the MDSR can be derived as
𝑓𝑑𝑠_𝑚𝑎𝑥 =
(𝐼𝐵𝐷 −𝐼𝑆 )Nmax 𝑓𝑡𝑥
𝐼𝑡𝑥 (𝐿𝑝𝑟𝑒𝑎𝑚𝑏𝑒𝑙 +8𝑛Nmax )+𝑓𝑡𝑥 (Nmax IDS TDS+ TrxIrx )
(2.3.4-21)
Where Nmax is the maximum sampling time during TTNSM and with fixed listening time Nmax
can be calculated as
31
(𝐸 ′ ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑− V𝑖𝑛 Trx Irx )𝑓𝑡𝑥
Nmax = (
V𝑖𝑛 𝐼𝑡𝑥
1
− 𝐿𝑝𝑟𝑒𝑎𝑚𝑏𝑒𝑙 ) ( )
8𝑛
(2.3.4-22)
where 𝐸 ′ ℎ𝑎𝑟𝑣𝑒𝑠𝑡𝑒𝑑 is the maximum energy can be harvested within the TEHS working range (see
Table 2-3).
Figure 2-31 and Figure 2-32 show that with more time spending on Rx mode, MDSR becomes
lower and lower. MDSR of sensors goes to zero when the listening time is 0.66 s, because all the
harvesting energy are used to listen to the channel and no energy are used to transmit data. For a
practical system, Figure 2-31 suggests that listening time should not exceed about 100ms. This is
an important result relative to the design of MAC layer protocols in the following chapter.
Figure 2-31: MDSR vs. Time spent in Rx modes (ACC sensor and EMG sensor)
2.3.5 Summary of PHY Layer Design
At the beginning of this chapter, we introduced two types of transceiver nodes of our star
topology energy harvesting wireless sensor network (EHWSN): the bio-sensor radio (BSR) node
and the gateway radio (GR) node. Then, we discussed the packet design and the modulation
method employed. Finally, we tested the output power from energy harvesting system and
32
current consumption for different modes, and analyzed the maximum data sampling rate (MDSR)
under different conditions. A summary of PHY layer design used in the calculations is shown in
Table 2-6.
Table 2-6: PHY layer features summary
Network type
Star topology network
Node type
BSR node, GR node
Modulation method
FSK @ 10kbit/s
Carrier frequency
467MHz
Packet type
Data packet, Command packet
Data packet: 13~269 bytes
Packet size
Command Packet: 6 bytes
Energy harvesting system
0.87mW
output power
Sleep mode: 8uA
Current consumption
Tx mode: 51mA
(with 4.95V)
Rx mode: 30mA
Maximum data sampling rate
ACC Sensor: 1.5Hz
(without Rx mode)
EMG Sensor: 4.5Hz
Maximum data sampling rate
ACC Sensor: 1.3Hz
(with 100 ms Rx mode)
EMG Sensor: 3.9Hz
33
Chapter 3 - Recent MAC Layer Protocol Survey and New MAC
Protocol Layer Design
As demonstrated in Chapter 2, controlling the duty-cycle of wireless sensor nodes in the EH
network facilitates low energy consumption. The MAC layer design for this WSN can be
regarded as a duty-cycling MAC Protocol [21] [24]. The first portion of this chapter presents a
survey of duty-cycling MAC layer design roughly categorized as synchronous (e.g. WiseMAC
[15], S-MAC [16], T-MAC [17]) and asynchronous (e.g. B-MAC [18], X-MAC [19], RI-MAC
[20], and RW-MAC [21]). Given the extreme energy constraints on energy and the cost of
achieving synchrony, asynchronous MAC layer is the primary focus of this thesis.
In our EHWSN, as Bio-Sensor Radio (BSR) nodes are extremely power-constrained, their
energy usage has to be reduced or even eliminated. It would be prudent to use almost all the
energy stored by EHS for transmitting data. Gateway Radio (GR) node is a power-unconstrained
node. Therefore, the primary goal of our MAC layer design is to shift energy consumption
pressure from BSR nodes to the GR node as best as possible. Two asynchronous MAC protocol
designs that fit the PHY layer design: Type I MAC Layer (CSA-MAC layer) and Type II MAC
Layer (GRI-MAC Layer) are present in this chapter. Their properties (collision probability,
overhearing time, and idle listening time) are analyzed via simulations.
34
3.1 Survey of Energy Efficient MAC Protocols
Currently, plenty of MAC layer designs for WSNs attempt to obtain high energy efficiency by
duty-cycling each node in the WSNs [22]. Duty-cycling nodes between sleep mode and active
mode can reduce unnecessary power consumption. The following section introduces background
knowledge and describes several duty-cycling MAC layers.
3.1.1 Energy Considerations
Before individual MACs can be introduced, reasons for energy waste [22] [23] [24] in WSNs are
demonstrated below:

Packet Collision: Packet collision occurs when more than one packet is transmitted
simultaneously or when only part of a packet overlaps with other packets during
transmission (sees Figure 3-1). Packet collisions can be solved by re-transmitting the
packet. This solution, however, will at least double the amount of energy utilized to
transmit the packet.
Figure 3-1: Packet Collision (part of packet overlaps)

Idle Listening: Idle listening occurs when a node consumes energy in order to listen to
the idle channel while waiting for possible traffic.

Overhead: Energy is consumed by nodes sending the control packet before sending the
data packet.

Overhearing: A node may waste energy listening to a signal not destined for it.

Overemitting: A node may transmit a packet to a receiver that is not ready to receive it.
35
3.1.2 MAC Layer for Duty-cycled WSNs
In order to prolong the life and increase energy efficiency of the sensor node, sensor nodes cycle
between a sleep mode and an active mode. MAC protocols developed for this kind of dutycycled WSNs can be categorized as synchronous or asynchronous [19] [22]. In synchronous
WSNs, duty cycles of sender nodes are pre-scheduled by receiver nodes via synchronization
information; in asynchronous WSNs, duty cycles of sender nodes are independent of receiver
nodes' status.
A. Synchronized MAC Layer for Duty- cycled WSN
Synchronous MAC protocols specify the period of wake-up and sleep durations to reduce
unnecessary time and energy wasted in idle listening. S-MAC is an example of this type of
MAC protocol.
Sensor-MAC (S-MAC)
S-MAC [16] aims to establish low-duty-cycle operation to achieve high energy efficiency. All
nodes in the network periodically sleep, wake up, receive/transmit data, and go back to sleep. To
reduce control over-head, they use virtual cluster techniques [19] to make the neighboring nodes
synchronize via exchange of schedule information. Therefore, the neighboring nodes can wake
up at the same time. After exchanging synchronization information, the nodes will send Ready to
Send-Clear to Send (RTS-CTS) packet at the end of a wake period to avoid data collision. The
send-data will follow the successful exchange of RTS-CTS, which is shown in Figure 3-3. Since
periodic sleep causes high latency, especially for multi-hop routing, adaptive listening is
introduced to reduce latency. When a node hears a transmission of its neighbor, it wakes up at
the end of the transmission to determine whether it is the next hop. If the node is the next-hop
node, its neighbor could pass the data immediately instead of waiting until the next scheduled
listening time.
36
Figure 3-2: S-MAC Sender-Receiver Communication [16]
Advantages: Sleep schedules reduce energy waste caused by idle listening. Carrier sense is used
to achieve collision avoidance.
Disadvantages: Adaptive listening causes idle listening and overhearing problems.
Synchronization information exchange also increases power consumption due to control packet
overhead.
Conclusion: For S-MAC, periods of sleep and awake are predefined and constant.
Unfortunately, these periods do not fit our system because sleep period of BSR nodes is
controlled by the energy monitor. Energy waste on the control packet overhead (sync
information exchange, CTS/RTS) cannot be accommodated by the energy harvesting system that
shoulders the burden of energy consumption.
B. Asynchronous MAC Layer for Duty- cycled WSN
The key advantage of Asynchronous MAC is that the sender and receiver can be completely
decoupled in their duty cycles [25], thereby removing the energy used for exchanging
synchronization information. Asynchronous MAC can be divided into two categories: senderinitiated MAC and receiver-initiated MAC.
a) Sender-Initiated MAC
Sender-initiated means the communication request starts from the sender side.
37
B-MAC
Berkeley Media Access Control (B-MAC) [18] is a Carrier Sense Multiple Access (CSMA)
protocol for low-power WSNs. In order to avoid packet collision, the channel must be
determined to be clear. This determination is referred to as Clear Channel Assessment (CCA). To
realize CCA, B-MAC uses an outlier algorithm comprised of nodes that take five samples during
the channel sampling period. The channel is declared to be busy if none of the five samples is
outlier whose energy level is significantly below the noise floor. The channel is declared to be
clear if outlier exists among the five samples.
Receiver node duty cycles its radio through periodic channel sampling, known as Low Power
Listening (LPL). Once the receiver wakes up, it checks channel activity. If a preamble is
detected, the receiver node stays awake to receive the incoming packet. Transmit nodes send the
preamble long enough to match the interval. For example, if the channel is checked every 100ms,
the preamble must be at least 100ms long for a node to wake up. The channel is also checked for
activities to wake up the receiver node. The process of timing line is indicated in Figure 3-3.
Advantages: LPL reduces the power consumption caused by idle listening. B-MAC has better
packet deliver rate, latency and energy consumption [18].
Disadvantages: A long preamble must be sent before the data is transmitted, consequently
decreasing transmission efficiency and increasing power consumption.
Conclusion: In our system of interest, the transmit node (BSR) is an extremely powerconstrained node. The long preamble consumes almost all the energy harvested in one sleep
cycle before data transmission occurs. Therefore, the B-MAC is unsuitable for our WSN.
X-MAC
To avoid the overhearing, X-MAC [19] sends a series of short preambles instead of an entire
long preamble used in B-MAC. For each short preamble, the target address is embedded, thereby
allowing the receiver node to know whether or not it is the target node without having to wait
until the end of the extended preamble. If the receiver node finds it is not the target node, it can
38
go back to sleep immediately to reduce energy cost due to overhearing; if the receiver node
finds it is the target node, it sends an Acknowledge (ACK) packet to the transmit node. In XMAC a short pause is inserted between short preambles to create a small window through which
the sender node can listen to the media. If a sender node received ACK packet from the receiver,
it stops sending the preamble and begins to send the data packet, as shown in Figure 3-3,
Advantages: Short preambles with embedded target ID solves the overhearing problem. X-MAC
has much better performance than B-MAC from the standpoint of the energy usage.
Disadvantages: The series of short preamble transmission dominates power consumption.
Conclusion: X-MAC cannot be used for the system in this study because of reasons similar to BMAC. Energy waste as a result of sending a series of short preambles is not feasible for BSR
nodes.
Figure 3-3: Comparison of time line between B-MAC and X-MAC [19]
39
b) Receiver-Initiated MAC
Receiver-initiated means the communication request starts from receiver side.
RI-MAC
In receiver-initiated MAC (RI-MAC) [20], a DATA frame transmission is initiated by the
intended receiver node of the DATA. Each receiver node in RI-MAC periodically wakes up at its
own time to turn on its radio and check channel status. If the channel is idle, the receiver node
broadcasts a beacon to inform that it is ready to receive a DATA frame. For the sender, a node
with pending data silently waits for a beacon from the receiver. When a sender node receives a
beacon from the intended receiver, it immediately sends the DATA frame. When the
transmission is complete, the sender will get an ACK from the receiver. If no DATA frame
comes after a beacon is broadcast, the node goes to sleep, as shown in Figure3-4.
Advantages: Sender does not occupy the channel until the receiver is ready to receive, which
handles a wide range of traffic loads more efficiently than B-MAC and X-MAC.
Disadvantages: Sender typically must idle listen for a long time to receive a beacon sent from
the receiver.
Conclusion: A long idle listening time for the sender nodes (BSR nodes) is not feasible for our
system.
Figure 3-4: Time diagram of RI-MAC [20]
40
3.2 Type I MAC layer protocol design
3.2.1 Attributes
Extremely Low Duty Cycle:
According to the duty-cycle calculation in Section 2.4, our EH sensor node has only
approximately 0.4% duty cycle that is limited by charging current (≈ 200uA) from the EH
system and discharging current (≈50mA) for data transmission. In this MAC layer design, the
sleep mode period of BSR node has to be maintained at least 250 times longer than the active
mode period. For example, if a transmit node spends 300ms to transmit a packet, it must sleep
approximately 1.25 minutes to recover its energy.
Single and Unpredictable Energy Source:
In our energy harvesting system, TEG is the only energy source. The TEG converts thermal
energy (due to temperature difference) to electrical energy to power the mother-board. Therefore,
when designing the MAC layer, the working range (see Table 2-3) of the system must be
considered by measuring voltage across the super cap in the daughter board.
Star Topology Network:
Our network can be considered as a Single Cluster Star Topology Network (SASTN) that
consists of one GR node and multiple BSR nodes (Figure3-5). In contrast to an ad hoc network,
in which every node plays an equivalent role, the GR node is a power-unconstrained node
(powered by the space suit) that works as central node to receive all data packets. BSR nodes are
power-constrained nodes (powered by EH system) that work as sub-nodes to transmit data
packets. During MAC protocol design, energy consumption must be shifted to the GR node as
best as possible.
41
Figure 3-5: Overview of a Star Topology Network
3.2.2 CS-ALOHA Based MAC Layer Design
Along with the discussions in previous subsections, any proposed MAC protocol for our
EHWSN must meet the following goals:

Low latency for data transmission

Simple control packet overhead

Achieve highest sampling rate with current duty-cycle

Shift power consumption pressure from BSR nodes to the GR node
We propose a CS-ALOHA based MAC protocol to meet these requirements.
3.2.2.1 Background
ALOHA Protocol
ALOHA protocol [26], a design for early satellite systems, allows each subscriber to transmit
data whenever necessary. This design eliminates idle listening for the subscriber or the BSR
node in our system. However, increasing numbers of nodes and length of packets may lead to
more packet collisions.
Carrier Sense
42
Carrier sense (CS) means that each node in the network monitors channel status before
information transmission, consequently greatly decreasing collision probability. In type I MAC
protocol, CS was realized by the Channel Energy Detection method utilized by other protocols,
including 802.15.4 [19]. In this method, each BSR node takes a sample of the channel and
compares it to the noise floor. If the sample is above the pre-set threshold, the channel is
considered to be busy; if the sample is below the pre-set threshold, the channel is considered to
be idle.
Energy Level Monitor
Due to limitations of the TEHS, our system’s working range is from 4.6V to 5.0V (voltage
across the Super Capacitor, and see Table 2-3 for details). If the voltage drops below 4.6V, the
power system unexpectedly shuts off the entire system. The energy status of this system can be
monitored by checking voltage across the Super Capacitor. If the voltage is above 4.95V, the
energy status is good for sending a data packet; if the voltage is below 4.65V, the energy status is
poor (too close to the 4.6V) and the system is forced into sleep mode.
3.2.2.2 Design Description
CS-ALOHA based MAC is sender-initiated asynchronous MAC protocol that combines the CS
method and the ALOHA protocol.
For the BSR node side, each node has distinct sleep-wake up periods. When a BSR node is ready
to transmit a data packet, the node turns on its radio to sense the channel. If the channel is idle,
the node transmits a data packet to the GR node; if the channel is busy, the node goes to sleep for
a period equal to the longest data packet transmission time (data pay load part contains 256 bytes
data) which is approximately 250ms with 10kbit/s transmission rate. The node then wakes up to
sense the channel again. After the BSR node receives the ACK command from the GR node
indicating correct reception of the data packet, the BSR node goes back to sleep for another
sleep-sampling period. Sometimes, because of low energy level, the BSR node goes back to
sleep mode without listening to the ACK command. The BSR node process is shown in Figure36.
43
Figure 3-6: BSR Node Process
Because the GR node is a power unconstrained, it listens almost continuously in order to receive
the data packet from BSR nodes, thereby making it possible for BSR nodes to transmit data
packet whenever they are ready to do so. Therefore, BSR node does not need to transmit a long
preamble or a series of short preamble in order to inform the receiver node that it is ready to
transmit (see B-MAC and X-MAC). Although this design causes the GR node to experience idle
listening, this design successfully shifts power consumption pressure from the BSR side to the
GR side, thereby meeting the goals of this study. After correctly receiving the data packet (pass
CRC checking), a GR node sends an ACK command packet to the BSR node to inform that the
transmission is complete. After that, the GR node continues to listen to the channel. The BSR
node process is shown in Figure3-7. A timing flow diagram is shown in Figure3-8.
44
Figure 3-7: Receiver Node (GR node) Process
Figure 3-8: Timing Flow Diagram
45
3.2.3 Simulation and Analysis
WSN in this study is a single-hop network. A BSR node transmits a data packet directly to the
GR node without going through intermediary nodes. Therefore there is no latency caused by
multiple-hops which are common in ad hoc networks. A BSR node immediately transmits a data
packet when the node is ready (high energy level). Therefore, no idle listening or overhearing
issues exists in this MAC layer design.
Ideally, packet collision will not occur in this MAC layer design, because a BSR node ensures
that the channel is clear before transmitting a data packet. Unfortunately, due to the hidden
terminal problem, carrier sense (CS) step does not always work. Therefore, packet collision
probability of this MAC layer without the CS step must be evaluated.
Without the CS step, CS-ALOHA MAC layer can be considered as a Pure ALOHA [26] scheme.
The probability of that n packets generated by nodes during a given packet duration interval is
assumed to be Poisson distributed, given by
Pr(𝑛) = (𝜆𝜏)𝑛 𝑒 −(𝜆𝜏) /𝑛!
(3.2.3-1)
where, 𝜆 is the mean arrival rate (packet/second) calculated by Equation (3.2.3-2), and 𝜏 is the
packet duration.
𝜆 = ∑𝑁
𝑖=1(
1
𝑇𝑐𝑦𝑐𝑙𝑒
)
(3.2.3-2)
where, N is the number of BSR nodes and 𝑇𝑐𝑦𝑐𝑙𝑒 is the average time of a sleep-wake up period.
Here, we assume that 𝑇𝑐𝑦𝑐𝑙𝑒 is the same across all BSRs. A BSR node transmits one packet
within each 𝑇𝑐𝑦𝑐𝑙𝑒 , thereby using 1/𝑇𝑐𝑦𝑐𝑙𝑒 to calculate mean arrival rate for one BSR node. 𝜆
increases with the number of BSR nodes.
A packet is assumed to be transmitted successfully if no other packets are transmitted during the
given packet duration interval (𝜏) [26]. The probability that zero packet is generated (i.e., no
collision) during this interval is given by
Pr(0) = 𝑒 −(𝜆𝜏)
(3.2.3-3)
Therefore, the probability of packet collision is given by
Pr[collsion] = 1 − Pr(0)
46
(3.2.3-4)
Figure 3-9 shows simulation results of packet collision probability with increasing number of
BSR nodes.
Figure 3-9: Packet Collision Probability vs. Number of BSR Nodes
After taking CS into consideration, the probability of packet collision can be derived by
Pr[collision] = (1 − P𝑠𝑢𝑐𝑐𝑒𝑠𝑠𝑓𝑢𝑙 )(1 − Pr(0))
(3.2.3-4)
where P𝑠𝑢𝑐𝑐𝑒𝑠𝑠𝑓𝑢𝑙 is the probability that CS gives a correct decision whether the channel is clear
or not. Figure 3-10 shows that, with P𝑠𝑢𝑐𝑐𝑒𝑠𝑠𝑓𝑢𝑙 increasing, the Pr [collision] is decreased.
Figure 3-10: Packet Collision Probability vs. Number of BSR Nodes with CS
47
3.2.4 Design Drawbacks
Due to the special usage environment, concerns exist regarding whether or not all nodes will
work within a space suit. A space suit consists of various layers, including multiple metallic
layers which are radio opaque at Ultra High Frequency (UHF) [27]. All wireless signals, which
are at UHF, can only be transmitted inside a space suit, and the signal is attenuated due to path
loss and antenna mismatch loss. The signal becomes weaker as distance increases. For example,
if a node located at the left wrist (14) (Figure 3-11) is transmitting a data packet, another node at
the left ankle (10) may consider the data packet to be a noise and consequently detect an idle
channel. In this case, packet collision is not avoided by sampling channel energy (CS), meaning
that the Type I MAC layer has potential for undetected packet collisions.
Figure 3-11: Radio locations for intra-suit wireless propagation [27]
48
3.3 Type II MAC layer protocol design
Type II MAC layer protocol is designed to eliminate packet collisions. A GR initiated based
MAC (GRI-MAC) layer is proposed in this study as a Type II MAC layer design in order to use
the GR node to control BSR nodes. BSR nodes transmit data packet only when they receive an
indication from the GR node. Every data transmission initiated by the GR node can guarantee
only one data packet transmission at a time, thereby eliminating the packet collision issue.
3.3.1 BSR Initiated Based MAC Layer Design
3.3.1.1 Definitions
Beacon: As shown in Figure 3-12, beacon contains sync word, ID, and Parity Check bit. Beacon
transmits from GR node to BSR nodes to indicate the BSR node with identical ID in order to
give a data packet transmission. Length of the beacon is calculated by Equation (3.3.1-1).
1
Beacon Length = 𝐵𝑙𝑒𝑛𝑔𝑡ℎ = (8𝑛𝑠𝑦𝑛𝑐 + 8)(𝑓 )
𝑡𝑥
(3.3.1-1)
where, 𝑛𝑠𝑦𝑛𝑐 is the number of bytes of sync word, and 𝑓𝑡𝑥 is transmission rate. 𝐵𝑙𝑒𝑛𝑔𝑡ℎ is 3.2ms
with 3 bytes sync word and 10kbit/s transmission rate.
Figure 3-12: Beacon Frame
Beacon Gap: A time period between beacons is called beacon gap (BG). During BG, the GR
node switches from transmit (TX) mode to receiving (RX) mode in order to detect channel
status. BG must be long enough to determine if a data packet is coming or just an idle channel.
BG length can be calculated as:
𝟏
Beacon Gap Length = 𝐵𝐺𝑙𝑒𝑛𝑔𝑡ℎ = (8𝑛𝑠𝑦𝑛𝑐 ) (𝒇 )
𝒕𝒙
(3.3.1-2)
where 𝑛𝑠𝑦𝑛𝑐 is the number of byte of sync word, and 𝑓𝑡𝑥 is the transmission rate. The current
𝐵𝐺𝑙𝑒𝑛𝑔𝑡ℎ is 2.4ms with 3 bytes sync word and 10kbit/s transmission rate.
49
Minimum Listening Time: Minimum listening time (MLT) is the minimum time required for
the BSR node to wake up to listen to the channel. MLT length must cover at least two lengths of
beacon and one BG length to ensure that at least one entire beacon is detected. MLT can be
calculated as:
𝑀𝐿𝑇 = 2𝐵𝑙𝑒𝑛𝑔𝑡ℎ + 𝐵𝐺𝑙𝑒𝑛𝑔𝑡ℎ
(3.3.1-3)
The current MLT is(2 × 3.2 + 3.6)𝑚𝑠 = 10𝑚𝑠
Figure 3-13: Concepts of GWRI-MAC Design
3.3.1.2 Design Description
GR based MAC (GRI-MAC) is a receiver-initiated asynchronous MAC protocol.
GR node process:
A GR node keeps the radio on in order to periodically broadcast beacon and detects channel. GR
node’s working process is shown in Figure 3-14. After broadcasting a beacon, a GR node
switches from Tx mode to Rx mode with 𝐵𝐺𝑙𝑒𝑛𝑔𝑡ℎ to detect any coming packet. If no packet is
received during BG, the GR node switches to TX mode to broadcast a beacon with another ID.
The GR node broadcasts beacons with pre-set order (ID number from 1 to n), and begins again
when the beacon with the last ID number is sent. During a BG, if a GR node listens to an
incoming data packet, the GR node continues in Rx mode until the entire packet is received.
When data packet receiving is complete, the GR node sends an ACK command to intend BSR
node to indicate a correct reception. Then, the GR node continues to broadcast beacon with next
ID number.
50
Figure 3-14: Working Process of GR node
BSR node process:
A BSR turns on its radio to listen to beacons whenever the node is ready to transmit a data
packet. If no beacon is detected within MLT indicating, meaning that another BSR node data
packet is being transmitted, the BSR turns to sleep mode for a short period equal to the longest
data packet transmission time. After finding a beacon, a BSR node compares the beacon ID with
its own ID in order to estimate time of detecting the intended beacon. If the time is greater than
twice the Temperature Compensated Crystal Oscillator (TCXO) stabilized time (T_oscillation),
the BSR node sleeps for the estimated time. Otherwise the BSR node extends its listening time to
detect the intended beacon. When the intended beacon is found, the BSR node begins to transmit
a data packet. If the BSR node receives an ACK command or detects a low energy level, it goes
to a sleep-sampling period to wait for the next transmission cycle. The entire working process is
shown in Figure 3-15.
Timing diagrams for various scenarios are shown in Figure 3-16, Figure 3-17, and Figure 3-18.
Figure 3-16 demonstrates that BSR Node1 wakes up to listen to beacons while BSR Node3
transmits the data packet, thereby preventing Node1 from finding a beacon within MLT. After
Node1 listens to the channel for MLT, it goes to sleep for a period (equal to the longest data
packet transmission time) and then wakes up to detect the beacon again. BSR Node3 wakes up
earlier than Node1 and finds a beacon with only one ID number ahead of the expected beacon.
Therefore, Node3 extends its listening time until detecting the intended beacon.
51
Figure 3-15: Working Process of BSR node
Figure 3-16: Timing Diagram of Scenario one
52
Figure 3-17 shows that, although BSR Node6 wakes up approximately one beacon time earlier
than BSR Node3, Node3 transmits a data packet earlier because Node3 detects its intended
beacon (with ID of Node3) before Node6 does. During data packet transmission of Node3,
Node6 no longer finds beacons, so it goes to sleep for a short period (equal to the longest data
packet transmission time) and then wakes up to detect the beacon again.
Figure 3-17: Timing Diagram of Scenario two
Figure 3-18 illustrates that, BSR Node15 wakes up to detect beacon with the ID of Node2. After
estimating that the time of receiving beacon with ID of Node15 is longer than twice of TCXO
stabilized time, Node15 goes to sleep for an estimated time and wakes up approximately one
beacon ahead of its intended beacon. If another BSR node does a data packet transmission during
Node15 sleep time, the intended beacon for Node15 will be delayed. In this case that intended
beacon is delayed, node15 wakes up to detect a beacon and re-estimates another sleep time.
53
Figure 3-18: Timing Diagram of Scenario three
3.3.2 Simulation and Analysis
Although GRI-MAC layer design eliminates packet collision and increases system reliability, the
design incurs extra energy costs due to idle listening and overhearing. Average listening time
(ALT) for individual BSR nodes must be quantified, consequently identifying the corresponding
max-sampling rate by using Equation 2.3.6-18.
Average listening time:
ALT is the average time that a BSR node listens to find its intended beacon. The BSR node
process (Section 3.3.1) shows that the listening time for a BSR node has three possible
conditions: detect intended beacon within MLT (Figure 3-19), detect intended beacon in
extended listening time (estimated sleep time < 2xT_oscillation) (Figure 3-20), and detect
intended beacon in the next wake-up time (Figure 3-21). ALT can be calculated by the
following Equation (3.3.2-1):
𝑨𝑳𝑻 = ∑𝟑𝐢=𝟏 𝑻(𝒊) 𝑷𝒓(𝒊) = 𝑻(𝟏) 𝐏𝐫(𝟏) + 𝑻(𝟐) 𝐏𝐫(𝟐) + 𝑻(𝟑) 𝐏𝐫(𝟑)
where 𝑻(𝒊) is the listening time for each condition and 𝑷𝒓(𝒊) is the probability for each
condition.
54
(3.3.2-1)
In the first condition, the intended beacon can be found at the beginning of MLT or at the end
of 𝑀𝐿𝑇. Therefore, 𝑻(𝟏) must be a random number located within [𝐵𝑙𝑒𝑛𝑔𝑡ℎ , 2𝐵𝑙𝑒𝑛𝑔𝑡ℎ +
𝐵𝐺𝑙𝑒𝑛𝑔𝑡ℎ ]; 𝑻(𝟏)'s expectation value is 1.5𝐵𝑙𝑒𝑛𝑔𝑡ℎ + 0.5𝐵𝐺𝑙𝑒𝑛𝑔𝑡ℎ . The 𝐏𝐫(𝟏) is 1/N, where N is
the number of BSR nodes.
Figure 3-19: Condition 1: Detect intended beacon within MLT
In the second condition, 𝑻(𝟐) 𝐏𝐫(𝟐) can be calculated by the following equation:
𝟏
𝒆
𝑻(𝟐) 𝐏𝐫(𝟐) = (𝒆 ) ∑𝒆𝒊=𝟏 (𝑀𝐿𝑇 + 𝒊(𝐵𝑙𝑒𝑛𝑔𝑡ℎ + 𝐵𝐺𝑙𝑒𝑛𝑔𝑡ℎ )) (𝑵)
(3.3.2-2)
where e indicates the maximum number difference between detected beacon and intended
beacon under condition 2. After MLT, if the beacon number extends by 1, listening time extends
𝒆
by (𝐵𝑙𝑒𝑛𝑔𝑡ℎ + 𝐵𝐺𝑙𝑒𝑛𝑔𝑡ℎ ). (𝑵) reflecting the probability of the second condition.
55
Figure 3-20: Condition 2: Detect intended beacon in extended listening time (Estimated sleep
time < 2xT_oscillation)
In the third condition, the listening time, 𝑻(𝟑), is approximately 2 𝑀𝐿𝑇. The probability of the
𝑵−𝒆−𝟏
third condition,𝐏𝐫(𝟑) is(
𝑵
).
Figure 3-21: Condition 2: Detect intended beacon in next waking-up time
(Estimated sleep time < 2xT_oscillation)
Assuming T_oscillation is 10ms, a simulation result of ALT vs. Number of BSR nodes is shown
in Figure 3-22. ALT converges to 18ms because Condition 3 will dominate ALT calculation with
an increasing number of BSR nodes. From Figure 2-31 we can easily find the maximum data
56
sampling rate of our system working under current MAC layer design
(time spent in Rx mode is 18ms).
.
Figure 3-22: ALT vs. Number of BSR nodes
3.3.3 Design Drawbacks
This design contains three primary drawbacks:

Idle listening: BSR nodes must wake up for a period in order to listen to their beacons
before transmitting a data packet. Although the period is short and the wasted energy is
affordable, the BSR nodes must prolong sleep time in order to recover the extra energy
cost.

Overhearing: A BSR node likely listens to the beacons with IDs of other BSR nodes.
57

Latency: Because of an increasing number of nodes, a BSR node requires longer time
waiting for its intended beacon (a node will turn to sleep mode first and turn back to Rx
mode if the waiting time is too long), thereby increasing the latency between packet
transmissions.
3.4 Conclusions
In this chapter, we provided a survey of energy efficient MAC protocols and proposed two MAC
layer designs which fit the EHWSN of interest in this thesis. CS-ALOHA MAC layer design has
high energy utilization and GRI-MAC layer design has high reliability. A summary of these two
MAC layer performance metrics designs is presented in Table 3-1.
Table 3-1: Summary of Metric of CSA-MAC and GRI-MAC
CS-ALOHA MAC
GRI-MAC
Packet collision
YES
NO
Idle listening
NO
YES
Overhearing
NO
YES
Overemitting
NO
NO
Overhead
YES
YES
Latency
NO
YES
Energy utilization
Higher
Lower
Reliable
Lower
Higher
Average listening time
0
18ms
MDSR (EMG sensor)
≈ 4.5𝐻𝑧
≈ 4.3𝐻𝑧
58
Chapter 4 - Conclusion and Future Work
4.1 Summary
Energy harvesting wireless sensor networks (EHWSNs) have gained popularity recently
due to their independence from utility power [5], longer transceiver node lifetime, and safety in
special environments (e.g., inside space suits). This thesis, which discusses both physical layer
and medium access control (MAC) layer design, demonstrates EHWSN design inside a space
suit. The EHWSN in this thesis considered a star topology network, including multiple biosensor radio (BSR) nodes and one gateway radio (GR) node. A GR node acts as a central node to
receive data packets sent by BSR nodes and organize the network. The gateway node is assumed
to be power-unconstrained.
The wireless sensor network (WSN) in this research is considered as an EHWSN because
all BSR nodes are powered by thermal energy harvesting system (TEHS) which provides very
low output power (≈ 1𝑚𝑊). An experiment is conducted to replicate real thermal-energy flow
in the proposed hardware to estimate output power during practical use. The measurement result
(output power = 0.87mW) is a key parameter to calculate the BSR node duty cycle.
Another key parameter for duty cycle calculation is BSR node-consumed power under
various modes of operation. A functional BSR node with transmit (Tx) mode and receive (Rx)
mode was implemented to measure power consumption under real situations. Based on the
measurements, Tx mode consumes approximately 250mW, Rx mode consumes approximately
150mW, and duty cycle is calculated to be approximately 0.4%. Results of duty cycle calculation
and maximum data sampling rate (MDSR) analysis based on measurements provided guidance
for the MAC layer design.
A survey was conducted to obtain a broad outline for existing MAC layer designs. After
considering previous analysis and modifying existing MAC layers, two MAC layer designs, CSA
(Carrier Sense ALOHA) -MAC layer and GRI (gateway radio initiated)-MAC layer, were
presented. CSA-MAC layer and GRI-MAC layer designs which shift power consumption
pressure from BSR nodes to the GR node are feasible for our extremely low duty cycle EHWSN.
CSA-MAC layer has a little bit higher maximum data sampling rate and GRI-MAC is more
reliable.
59
4.2 Future Work
Due to the extremely low duty cycle (≤ 0.4%) and low Tx rate (10kbit/s), MDSR in this
system is only 1.5Hz~ 4.5Hz, which is far less than biosensor signal sampling rate requirement
(Table 2-2). Two aspects including TEHS and Tx rate must be improved in order to realize
higher MDSR. Current TEHS having only 0.87mW overall output power is designed for using
only two parallel thermal electrical generators (TEGs) and more than two-thirds of harvested
energy is wasted in ultra-low voltage step-up converter (UVSC). In order to increase TEHS
output power, a new structure with more TEGs must be designed and a new USVC must be
found. Tx rate is limited primarily by the current demodulation method. As link budget studies
suggest significant link margin exist within the envisioned space suit application [27], new
methods can be employed to achieve higher Tx rate. If we can increase the Tx rate from 10kbit/s
to 1Mbit/s without changing anything else of this system, MDSR can reach as high as 300Hz
which will meet almost all the biosensor signals sampling rate requirements. Obtained MDSR
value currently is based primarily on calculation and simulation results. In order to obtain more
accurate MDSR of EHWSN under real situation, the two types of MAC layer designs proposed
in Chapter 3 should be implemented and tested.
60
References
[1] Day, D.; Xiongjie Dong; Kuhn, W.; Gruenbacher, D.; Natarajan, B.; Sobering, T.; Taj-Eldin,
M.; Warren, S.; Barstow, T.; Broxterman, R.; Stonestreet, A., "Biomedical sensing and wireless
technologies for long duration EVAs and precursor scout missions," Aerospace Conference,
2014 IEEE , vol., no., pp.1,14, 1-8 March 2014
doi: 10.1109/AERO.2014.6836290
[2] Song, Wen, Carl Ade, Ryan Broxterman, Thomas Nelson, Dana Gude, Thomas
Barstow, and Steve Warren. “Classification Algorithms Applied to Accelerometer Data as a
Means to Identify Subject Activities Related to Planetary Navigation Tasks,” HRP 2013,
NASA Human Research Program Investigators’ Workshop, February 11–14, 2013, Moody
Gardens Hotel, Galveston, TX.
[3] [Delsys] DelsysTrigno (Natick, MA).
[4] Xiongjie Dong. "A ZigBee-based Wireless Biomedical Sensor Network as a precursor to an
in-suit system for monitoring astronaut state of health". Thesis for Master of Science,
Department of Electrical and Computer Engineering, Kansas State University, Manhattan,
Kansas, 2014
[5] Akshay Uttama Nambi S., N.; V, Prabhakar T.; Venkatesha Prasad, R; S, Jamadagni H." Zero
Energy Network stack for Energy Harvested WSNs " eprint arXiv:1404.7330 04/2014
[6] Venkata, P.T.; Nambi, S.N.A.U.; Prasad, R.V.; Niemegeers, I., "Bond Graph Modeling for
Energy-Harvesting Wireless Sensor Networks," Computer , vol.45, no.9, pp.31,38, Sept. 2012
doi: 10.1109/MC.2012.250
[7] T.V. Prabhakar, S.N. Akshay Uttama Nambi, R. Venkatesha Prasad, S. Shilpa, K. Prakruthi,
and Ignas Niemegeers. " A Distributed Smart Application for Solar Powered WSNs" :
NETWORKING 2012, Part II, LNCS 7290, pp. 291–303, 2012.
[8] . ZigBee Green, http://www.zigbee.org/Standards/Overview.aspx
[9] Amelia Lynn Hodges. "Investigation of Antenna and Energy Harvesting Methods for Use
with A UHF Microtransceiver in A Biosensor Network" Thesis for Master of Science,
Department of Electrical and Computer Engineering, Kansas State University, 2013
[10] LTC3108 datasheet: http://cds.linear.com/docs/en/datasheet/3108fc.pdf
[11] TPS781 datasheet: http://www.ti.com/lit/ds/symlink/tps78101.pdf
[12] Zhang Xiaohu." VHF & UHF energy harvesting radio system physical and MAC layer
considerations" Thesis for Master of Science, Department of Electrical and Computer
Engineering, Kansas State University, Manhattan, Kansas 2009
61
[13] Charles Carlson, thesis Thesis for Master of Science (in preparation), Department of
Electrical and Computer Engineering, Kansas State University, Manhattan, Kansas 2014
[14] Bdiri, S.; Derbel, F.; Kanoun, O., "Wireless sensor nodes using energy harvesting and BMac protocol," Systems, Signals & Devices (SSD), 2013 10th International Multi-Conference
on , vol., no., pp.1,5, 18-21 March 2013
doi: 10.1109/SSD.2013.6564160
[15] C.C. Enz, A. El-Hoiydi, J.-D. Decotignie, V. Peiris: WiseMAC: An Ultralow-Power
Wireless Sensor Network Solution, IEEE Computer, Vol. 37, Issue 8 (August 2004).
[16] Wei Ye, J.Heidemann and D. Estrin: An Energy-Efficient MAC Protocol for Wireless
Sensor Networks, IEEE INFOCOM, New York, Vol. 2,pp. 1567-1576 (June 2002).
[17] Tijs van Dam, Koen Langendoen: An Adaptive Energy Efficient MAC Protocol for
Wireless Networks, in Proceedings of the First ACM Conference on Embedded Networked
Sensor Systems (November 2003).
[18] J. Polastre, J. Hill, D. Culler: Versatile low Power Media Access for Wireless Sensor
Networks, Proceedings of the 2nd ACM Conference on Embedded Networked Sensor Systems
(SenSys’04), Baltimore, MD, (November 2004)
[19] Dongyu Yang; Ying Qiu; Shining Li; Zhigang Li, "RW-MAC: An asynchronous receiverinitiated ultra low power MAC protocol for Wireless Sensor Networks," Wireless Sensor
Network, 2010. IET-WSN. IET International Conference on , vol., no., pp.393,398, 15-17 Nov.
2010doi: 10.1049/cp.2010.1085
[20] Yanjun Sun, Omer Gurewitz, David B. Johnson, "RI-MAC: A Receiver-Initiated
Asynchronous Duty Cycle MAC Protocol for Dynamic Traffic Loads in Wireless Sensor
Networks" SenSys’08, November 5–7, 2008, Raleigh, North Carolina, USA
[21] Dongyu Yang; Ying Qiu; Shining Li; Zhigang Li, "RW-MAC: An asynchronous receiverinitiated ultra low power MAC protocol for Wireless Sensor Networks," Wireless Sensor
Network, 2010. IET-WSN. IET International Conference on , vol., no., pp.393,398, 15-17 Nov.
2010 doi: 10.1049/cp.2010.1085
[22] Jaehyun Kim; Jeongseok On; Seoggyu Kim; Jaiyong Lee, "Performance Evaluation of
Synchronous and Asynchronous MAC Protocols for Wireless Sensor Networks," Sensor
Technologies and Applications, 2008. SENSORCOMM '08. Second International Conference
on , vol., no., pp.500,506, 25-31 Aug. 2008
doi: 10.1109/SENSORCOMM.2008.80
[23] Ilker Demirkol; Cem Ersoy; Fatih Alagöz, "MAC Protocols for Wireless Sensor Networks:
a Survey" IEEE Communications Magazine • April 2006
62
[24] Ali, M.; Bohm, A.; Jonsson, M., "Wireless Sensor Networks for Surveillance Applications –
A Comparative Survey of MAC Protocols," Wireless and Mobile Communications, 2008.
ICWMC '08. The Fourth International Conference on , vol., no., pp.399,403, July 27 2008-Aug.
1 2008
doi: 10.1109/ICWMC.2008.53
[25] Inwhee Joe; Hanjong Ryu, "A Patterned Preamble MAC Protocol for Wireless Sensor
Networks," Computer Communications and Networks, 2007. ICCCN 2007. Proceedings of 16th
International Conference on, vol., no., pp.1285,1290, 13-16 Aug. 2007
doi: 10.1109/ICCCN.2007.4317998
[26] Theodore S. Rappaport. "Multiple Access Techniques for Wireless Communications" in
Textbook of Wireless Communications Principles and Practice, 2ed ed. Prentice-Hall, PTR
Pub,1996, pp. 462-466.
[27] Taj-Eldin, M.; Kuhn, B.; Hodges, A.; Natarajan, B.; Peterson, G.; Alshetaiwi, M.; Ouyang,
S.; Sanchez, G.; Monfort-Nelson, E., "Wireless propagation measurements for astronaut body
area network," Wireless for Space and Extreme Environments (WiSEE), 2013 IEEE
International Conference on , vol., no., pp.1,7, 7-9 Nov. 2013
doi: 10.1109/WiSEE.2013.6737569
[28] Li Huang; Pop, V.; de Francisco, R.; Vullers, R.; Dolmans, G.; de Groot, H.; Imamura, K.,
"Ultra low power wireless and energy harvesting technologies — An ideal
combination," Communication Systems (ICCS), 2010 IEEE International Conference on , vol.,
no., pp.295,300, 17-19 Nov. 2010
doi: 10.1109/ICCS.2010.5686436
63
Appendix A - Table of Acronyms
ALT
Average Listening Time
ADC
ALT
AM
BG
B-MAC
BSR
CSA
CCA
CRC
CS
CS
CSMA
DB
DSM
ECG
EH
EHS
EHWSN
EMG
FPGA
GND
GR
GRI
GRI-MAC
KANDB
LPL
MAC
MB
MDSR
MLT
NSM
PHY
PO
RFIC
RI-MAC
Analog To Digital Converter
Average Listening Time
Active Mode
Beacon Gap
Berkeley Media Access Control
Bio-Sensor Radio
Carrier Sense ALOHA Based
Clear Channel Assessment
Cyclic Redundancy Check
Carrier Sense
Channel Energy
Carrier Sense Multiple Access
Daughter Board
Data Sampling Mode
Electrocardiogram
Energy Harvesting
Energy Harvesting System
Energy Harvesting Wireless Sensor Network
Electromyography
Field Programmable Gate Array
Ground
Gateway Radio
Gateway Radio Initialized
GR Initiated Based MAC
K-State-NASA Body Area Network Development Board
Low Power Listening
Medium Access Control
Mother Board
Maximum Data Sampling Rate
Minimum Listening Time
Nominal Sleep Mode
Physical
Pulse Oximeter
Radio Frequency Integrated Circuit
Receiver-Initiated MAC
64
RRS
RSSI
RTS-CTS
RX
SASTN
SM
SPI
TCXO
TEG
TEHS
TTNSM
TX
UHF
UVSC
Vout
WSN
Respiration Rate Senor
Received Signal Strength Indicator
Ready To Send-Clear To Send
Receiving
Single Cluster Star Topology Network
Sleep Mode
Serial Peripheral Interface
Temperature Compensated Crystal Oscillator
Thermal Electric Generator
Thermal Energy Harvesting System
total time of the NSM
Transmit
Ultra High Frequency
Ultralow-Voltage Step-Up Converter
Output Voltage
Wireless Sensor Network
65
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertising