Available online at [Link].

com

ScienceDirect
Energy Procedia 56 (2014) 414 – 422

11th Eco-Energy and Materials Science and Engineering (11th EMSES)

A survey on RF energy harvesting: circuits and protocols

Prusayon Nintanavongsaa
a Department of Computer Engineering, Rajamangala University of Technology Thanyaburi, Thailand

Abstract
Recent advancement in semiconductor technology and fabrication process enable realization of the concept of Radio Frequency
(RF) energy harvesting. RF energy harvesting, a process in which energy contained in electromagnetic waves is converted into
useful electrical energy, will help realize perennially operating sensors. With energy replenishment capability and protocol design,
RF energy harvesting sensors can attain the desirable characteristics of sensor design, lifetime and network performance. This
paper investigates detailed aspects of recent research on RF energy harvesting circuits and protocols. We also discuss the impact
of energy replenishment capability and protocol design on RF energy harvesting sensor networks.

©c 2014
2014Elsevier
The Authors. Published
Ltd. This is an openby Elsevier
access Ltd.
article under the CC BY-NC-ND license
Peer-review under responsibility of COE of Sustainalble Energy System, Rajamangala University of Technology Thanyaburi
([Link]
(RMUTT).
Peer-review under responsibility of COE of Sustainalble Energy System, Rajamangala University of Technology Thanyaburi (RMUTT)
Keywords: RF harvesting; Survey; Energy harvesting; Protocol; Wireless power transfer.

1. Introduction

One of major design considerations that hinders wireless sensor network from its fullest potential is the energy
constraint. Wireless sensor network is usually designed to have the longest possible lifetime while sacrificing sensing
capability due to duty cycling of sensors. However, this is no longer the case when energy harvesting technology
is introduced. Energy harvesting is the process of scavenging ambient energy from sources in the surrounding envi-
ronment. It is an attractive method for overcoming the energy limitations of conventional battery powered wireless
devices. This can instrument potentially leading to significant reduction in the costs associated with replacing bat-
teries periodically. Moreover, in some deployments, owing to the sensor location, battery replacement may be both
practically and economically infeasible, or may involve significant risks to human life.
RF energy harvesting is poised to alleviate some of these concerns by allowing sensors to re-charge energy storage
capacitors from the incident RF radiation. Since energy is replenishable, the design of wireless sensor network is no
longer bounded by the lifetime requirement and the network performance can be tuned with less energy constraint in
mind. However, The main challenge faced in harvesting RF energy is the free-space path loss of the transmitted signal
with distance. The Friis transmission equation relates the received (Pr ) and transmitted (Pt ) power with the distance

∗ Corresponding author. Tel.: +66-254-93460; fax: +66-254-93462.

E-mail address: prusayon.n@[Link]

1876-6102 © 2014 Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
([Link]
Peer-review under responsibility of COE of Sustainalble Energy System, Rajamangala University of Technology Thanyaburi (RMUTT)
doi:10.1016/[Link].2014.07.174
 Prusayon Nintanavongsa / Energy Procedia 56 (2014) 414 – 422 415

KƵƚ

DĂƚĐŚŝŶŐEĞƚǁŽƌŬ sŽůƚĂŐĞDƵůƚŝƉůŝĞƌ ŶĞƌŐLJƐƚŽƌĂŐĞ

(a) (b)

Fig. 1. Ambient RF energy harvesting (a) and RF energy harvesting module (DTV band) (b)

R as:

λ 2
Pr = Pt G t G r ( ) (1)
4πR

where Gt and Gr are antenna gains, and λ is the wavelength of the transmitted signal. The received signal strength,
diminishes with the square of the distance, requires special sensitivity considerations in the circuit design. Moreover,
FCC regulations limit the maximum transmission power in specific frequency bands. For example, in the 900 MHz
band, this maximum threshold is 4 W. Even at this highest setting, the received power at a moderate distance of 20 m
is attenuated down to only 10 μW.
The core contributions of this work can be summarized as follows:

• We present the fundamental design of RF energy harvesting circuit as well as advanced research in RF energy
harvesting circuit design.
• We provide specifically designed protocols for RF energy harvesting networks that complement RF energy
circuit design developed earlier.

The rest of this paper is organized as follows: In Section 2, we present various designs of RF energy harvesting
circuits. RF energy harvesting protocols are described in Section 3. Finally, Section 4 concludes this work.

2. RF Energy Harvesting Circuits

RF energy harvesting circuit is primarily based on the voltage multiplier circuit, which was invented by Heinrich
Greinacher in 1919. There are two major configurations of RF energy harvesting circuit, Villard voltage doubler,
sometimes also called Cockcroft-Walton voltage multiplier, and Dickson voltage multiplier. According to [1], Both
Villard and Dickson topology reveal no significant difference in performance. However, Dickson topology employs
parallel configuration of capacitors in each stage, reducing the circuit impedance, makes the matching task simpler.
Figure 1(a) illustrates the process of ambient RF energy harvesting. The incident RF power is converted into DC
power by the voltage multiplier while matching network, composed of inductive and capacitive elements, ensures the
maximum power delivery from antenna to voltage multiplier. The energy storage provides smooth power delivery
to the load, and as a reserve for durations when external energy is insufficient or unavailable. Such a design needs
to be carefully crafted: Increasing the number of multiplier stages gives higher voltage at the load, and yet reduces
the current through the final load branch. This may result in unacceptable charging delays for the energy storage
capacitor. Conversely, fewer stages of the multiplier will ensure quick charging of the capacitor, but the voltage
generated across it may be insufficient to drive the sensor mote (at least 1.8 V that becomes the +Vcc for Mica2
sensors). Along similar lines, a slight change in the matching circuit parameters alters significantly the frequency
range in which the efficiency of the energy conversion is maximum, often by several MHz. Hence, RF harvesting
circuits involve a complex interplay of design choices, which must be considered together. The authors address this
problem by considering a multi-stage design of the voltage multiplier, whose operating points are decided by solving
an optimization framework.
416 Prusayon Nintanavongsa / Energy Procedia 56 (2014) 414 – 422

Ĩϭ;džͿ ĨϮ;džͿ

ĨĨŝĐŝĞŶĐLJ;йͿ
ɲ ੘ ɴ
ZĞĐĞŝǀĞĚZ&WŽǁĞƌ;ĚŵͿ

Fig. 2. Efficiency curves of two energy harvesting sister-circuits, for LPD and HPD

2.1. Dual-stage RF Energy Harvesting Circuit Design

In [2], the authors proposed a dual-stage RF energy harvesting circuit design, one stage is most efficient at ex-
tremely low input RF power (say, low power design or LPD) while another performs better at comparatively higher
range (say, high power design or HPD). The prototype is shown in Figure 1(b). The study on impact of an individual
parameter, i.e., choice of diodes, number of stages, RF input power, and load impedance, on the output of the energy
harvesting circuit is investigated. This study is then used in the optimization framework of the energy harvesting
circuit. The authors conclude the impact of individual parameter on RF energy harvesting circuit as follows:

• Since the RF energy harvesting circuit has to be able to operate with weak input RF power, diodes with lowest
possible turn on voltage are preferable. Moreover, since the energy harvesting circuit is operating in high
frequencies, diodes with a very fast switching time need to be used. Schottky diodes, Avogo Technologies
HSMS-2852 and HSMS-2822, are chosen for LPD and HPD, respectively.
• The number of rectifier stages has a major influence on the output voltage of the energy harvesting circuit.
The output voltage is directly proportional to the number of stages used in the energy harvesting circuit. The
circuit yields higher efficiency as the number of stages increases. However, as more stages are introduced, the
peak of the efficiency curve also shifts towards the higher power region. Moreover, the voltage gain decreases
as number of stages increases due to parasitic effect of the constituent capacitors of each stage, and finally it
becomes negligible.
• It is important that the load impedance be carefully selected for a specific energy harvesting circuit. The circuit
yields the optimal efficiency at a particular load value, that is, the circuits efficiency decreases dramatically if
the load value is too low or too high.
• RF energy harvesting circuit exhibits non-linearity due to non-linear devices incorporated, i.e., Schottky diodes.
This implies that the impedance of the energy harvesting circuit varies with the amount of power received from
the antenna. Since the maximum power transfer occurs when the circuit is matched with the antenna, the
impedance matching is usually performed at the a particular input power.

The optimization framework is developed to determine the switchover point between these two sub-circuits, LPD
and HPD, so that the efficiency of the energy harvesting module is maximized in the operational incident power
range of −20 dBm to 20 dBm. The efficiency is defined as a ratio of DC output power of energy harvesting circuit to
incidental RF power at the antenna.
Figure 2 shows the two efficiency curves of energy harvesting sub-circuits. The efficiency curves f1 (x) and f2 (x)
belong to LPD and HPD circuits, respectively. The crossover point, γ, is the point where one of these two circuits
become the lead contributor to the total harvested energy. Thus, the LPD is operational if the RF input power is lower
than γ, otherwise the HPD circuit is operational.
As shown in Figure 2, at each particular crossover point γ, the total area under efficiency curve is the cumulative
sum of the area under the two distinct efficiency curves corresponding to the LPD and HPD designs, one on either
side of the crossover point γ. The total area under efficiency curve is hence,
 γ  β
Areatotal = f1 (x) dx + f2 (x) dx. (2)
α γ
 Prusayon Nintanavongsa / Energy Procedia 56 (2014) 414 – 422 417

The number of rectifier stages influences the minimum required voltage at the input in order to obtain a certain output
sufficient to drive a sensor mote. The authors consider various number of rectifier stages (N), ranging from 1 to
12 stages in this optimization framework. Moreover, the efficiency curve is also a function of impedance matching
network, consisting of inductor (L) and capacitor (C). Hence, The crossover point, γ, can be determined as follows:
 γ  β
γ = arg max { f1 (N1 , L, C, x) dx + f2 (N2 , L, C, x) dx} (3)
γ α γ

The dual-stage design yields almost double the efficiency than that of a major commercially available energy
harvesting circuit, Powercast P1100 [3], in the low incident power range (−20 dBm to 7 dBm) and Mica2 mote [4]
can be perpetually operated when their duty-cycle is carefully selected based on the incident RF power (as low as
−6 dBm).

3. RF Energy Harvesting Protocols

Energy efficient protocols have been extensively explored in the recent past, with a comprehensive classification
and survey on this topic presented in [5]. These protocols work on an assumption that energy is limited and ex-
haustible. Consequently, the effort is primarily shifted towards prolonging the network lifetime. However, with
energy harvesting capability, there is a need of fresh perspective on protocol design. Specific to the scenario of RF
energy transfer, the protocol proposed in [6], and its subsequent analytical model in [7], adopts a duty-cycle based on
the proportion of harvested energy. However, this protocol requires a strict centralized base station control and relies
on out-of-band RF power transfer, which can be expensive. In this section, various RF energy harvesting protocol
designs are presented.

3.1. Cross-layer Protocol for RF Energy Harvesting Sensors

In [8], the authors present two flavors of cross-layer protocols, Device-agnostic (DA) and Device-specific (DS).
These protocols determine the routing paths and the harvesting transmission duty cycle at each hop under different
conditions: The DA scheme relies purely on the local measurements on the harvesting capability of a node after the
sensors are deployed, and is useful for single-flow networks. The DS scheme provides a joint hardware-software
optimization by allowing the selection of the energy storing capacitor, apart from the route and duty cycle determi-
nation. Both the schemes rely on a rich set of device-level experimental studies that help provide exact performance
characteristics in practical scenarios.

3.1.1. Device-agnostic (DA)

An experiment to characterize the relationship between the received power P at the sensor and the distance away
from the energy transmitter (ET) is conducted. It is observed that the charging rate for a given receiver is highly de-
pendent on its specific location and relative height difference with respect to the energy transmitter ET. Thus, classical
metrics such as shortest path, in which all nodes are considered to exhibit a homogeneous charging characteristic, do
not work well in a realistic setting. The route formation is initiated by the source node, and the proposed metric can be
combined with most existing routing protocols for WSNs. In the current implementation, the authors modify AODV
by including the tuple < T ch max
(k), ηmax
c h(k) > in the route request (RREQ) packet that travels over path k. Here, T ch (k)
max

represents the maximum charging time considering all the nodes currently traversed in the path k, and ηc h(k) is the
max

observed standard deviation for this maximum value. As the RREQ is forwarded by the sensors, they may update
max
the field T ch (k) if their own charging time is greater than the value contained in this field. Thus, for a sensor i, the
max
change T ch (k) = tch
i i
if tch > T ch
max
(k), is undertaken before broadcasting the RREQ to its neighbors. The destination
receives multiples RREQs representing the different paths traversed from the source. It now chooses the path, say ψ,
with the lowest value of the maximum charging times of the various paths. Thus,

ψ = min{T ch
max
(k)} ∀k
= min{max[tch
i
]} ∀i ∈ path k, ∀k, (4)
418 Prusayon Nintanavongsa / Energy Procedia 56 (2014) 414 – 422

20

Voltage (V)
3

Time (sec)
15

2 10

1 5

0 0
3.5
12
10 400 3 0.2
8 300 2.5
6 200 0.15
100 2 0.1
4 0 Voltage (V) 1.5 Capacitor size (F)
RF Power (dBm) Time (sec)

(a) (b)

Fig. 3. The functional representation of the charging phase (a) Effect of capacitor sizes during capacitor discharging (b)

Fig. 4. The flow diagram of the RREQ-forward procedure.

After the base station chooses the optimal path, it sends back the route reply (RREP) to the nodes of this path, defining
the charging (T ch ) and transmission times (T x ) that is common to all of them. Finally, the duration for charging T ch
and the frame length T f rame , where T f rame = T ch + T x , are derived through the optimization framework.

3.1.2. Device-specific (DS)

A set of experiments to characterize the relationship between the received power P, the capacitor C that will
be charged by the energy harvesting device, and the output voltage V up to which the capacitor can be charged, is
conducted through real measurements of Mica2 energy harvesting equipped sensor. Figure 3(a) shows the functional
representation of the charging phase (Mica2 with 100 mF capacitor). It is obvious that the charging characteristic
curve is not constant but varies with the level of received signal power. In other words, the characteristic of charging
curve is a function of received signal power and a storage capacitor size. Moreover, figure 3(b) shows the discharging
characteristic of the device with various capacitor sizes). Note that the operational time of the devices is a function of
the residual voltage and a storage capacitor size.
After deriving functional representation in both charging and discharging phase, the optimal values for the storage
capacitor C and the voltage V upto which the capacitor must be recharged, is derived through an optimization frame-
work. This implies that the sensor will operate with the optimal duty cycle if these parameters are in effect. The route
formation is then performed based on the assumption that a node with higher active time is preferred to forward the
packet among potential relay nodes. The RREQ-forward procedure is shown in Figure 4.
Both DA and DS protocols are evaluated against voltage-aware AODV-based scheme (VA) [9]. A thorough per-
formance evaluation was conducted with various important metrics, including duty cycles, number of energy trans-
mitters, network load and multiple flows. The duty cycle of DS scheme is much higher than that of DA scheme. In
most applications where throughput is of most concern, the DS scheme is preferable over DA and VA scheme due to
higher throughput yielded. The DS scheme also yield marginally higher latency than the VA scheme. However, the
significantly higher throughput of the DA scheme can easily offset the marginal latency incurred.
 Prusayon Nintanavongsa / Energy Procedia 56 (2014) 414 – 422 419

'Ĩ WŽǁĞƌƐƉĞĐƚƌĂůĚĞŶƐŝƚLJ͕
dϴ
,ĐŝƌĐƵŝƚ͛ƐĨƌĞƋƵĞŶĐLJ ^ϭ;ĨͿ
ƌĞƐƉŽŶƐĞ͕,;ĨͿ

ZĞĐĞŝǀĞĚƉŽǁĞƌ;ĚŵͿ
dϯ dϳ
dϭ
dϰ ƌŽƐƐŽǀĞƌƉŽŝŶƚ͕J
dϲ dϮ
dϱ
WŽǁĞƌƐƉĞĐƚƌĂůĚĞŶƐŝƚLJ͕
^Ϯ;ĨͿ
dϵ

;ŵнϭͬϮͿO͗ŵсϬ͕ϭ͕Ϯ͕͙
ŵO͗ŵсϬ͕ϭ͕Ϯ͕͙
H Ĩϭ ĨϮ H &ƌĞƋƵĞŶĐLJ;,njͿ
ĞĐŝƐŝŽŶďŽƵŶĚĂƌLJ

(a) (b)

Fig. 5. ETs within a node’s radio range (a) Two-tone wireless energy transfer (b)

3.2. Medium Access Control Protocol for Sensors Powered by Wireless Energy Transfer (RF-MAC)

Consider an RF energy harvesting network with multiple energy transmitters (ETs), shown in Figure 5(a). The
sensor (located at the center) can be charged by multiple ETs, ET1 to ET9. However, Nintanavongsa et al. [10]
showed that the joint action can only be beneficial if the arriving waves at the sensor are aligned in phase. Hence,
ET3, ET4, and ET9 may together transmit, both being at a multiple of the signal wavelength λ away (denoted Group
I which translates in a phase difference that results in ‘constructive’ interference). While the remaining ETs, being at
a multiple of the signal wavelength (m + 1/2)λ away (denoted Group II) , should not be transmitting since this will
cause ‘destructive’ interference at the sensor being charged.
The author proposed Medium Access Control (MAC) protocol, called as RF-MAC, that ensures optimal energy
delivery to the requesting node under this circumstance. In RF-MAC, a node broadcasts its request for energy (RFE)
packet containing its ID, and then waits to hear for the ETs in the neighborhood. These responses from ETs are called
cleared for energy (CFE) pulses, which are simple, time-separated energy beacons. These pulses may be transmitted
by more than one ET concurrently, as overlapping CFEs need not be distinguished. Rather, the concurrent emission of
the CFEs increases the received energy level at the sensor, and this indicates a higher number of potential transmitters
from the energy requesting sensor. The responding ETs are then classified into two sets, based on rough estimates of
their separation distance from the energy requesting node to minimize the impact of destructive interference as much
as possible. Each set of ETs is assigned a slightly different peak transmission frequency so that each set contributes
constructively to the level of RF energy received at the node.

3.2.1. Grouping of the responding ETs

The ETs that hear the RFE reply back with a single, constant energy pulse. Each concentric band has the choice
of one of two time slots in which this pulse may be emitted, beginning from the instant of completion of the RFE.
Referring to Figure 6(a), the first slot is allocated for CFE pulses sent by energy transmitter of Group I (note: all
Group I bands are shown shaded). Similarly, CFE pulses from energy transmitters of Group II are sent during
the second slot, i.e. ETs 1, 2 and 3 collectively lie in the second concentric (Group II). band and simultaneously
transmit their pulses in the second slot. The node that sent the initial RFE estimates the total energy that it will receive
based on the signal strength of the CFE pulses in the slot number in which they were received. This arrangement
of using the pulses allows the ETs to be simple in design, and removes the concern of collisions in the reply packet.
Unlike classical data communication, it is not important for the node to know which ET will transmit energy. Rather,
its energy calculations are based on how much energy is contributed by the two groups of ETs separately. The authors
Group I Group II
define this cumulative energy as ERX and ERX , respectively, which are calculated by the RFE issuing node
from the received pulses.
The purpose of differentiating the energy contribution from the two groups is useful in the next stage, where an
optimization function returns the center frequencies of the ETs. Let Group I ETs be centered at frequency f1 ,
and Group II ETs be centered at frequency f2 so that they can concurrently transfer energy without destructively
affecting each other. Also, a desirable goal is to have minimum separation f2 − f2 , as the spectrum is most efficiently
utilized.
420 Prusayon Nintanavongsa / Energy Procedia 56 (2014) 414 – 422

How to select these frequencies f1 and f2 is explained next, which takes into account two important physical
layer characteristics of the energy transfer. The first is the spectrum response of the energy harvesting circuit that
is connected to the sensor nodes, shown by the envelope H( f ) in the frequency domain in Figure 5(b). The power
spectral density (PSD) of the two groups of ETs is the other concern, represented by S 1 ( f ) and S 2 ( f ), respectively,
for Group I and Group II. These shapes are observed by the sensor node from the incoming pulses from the
ETs. Thus, the bandwidth 2ε of the transmission spectrum (centered at f1 and f2 ) must be selected in such a way
their is minimum overlap between their individual spectra, and yet contained within the envelope of H( f ) to affect the
maximum level of power transfer. The authors use the following optimization assuming the transmission spectrum of
the ETs occupies a bandwidth of 2ε.

3.2.2. Optimization function for frequency assignment

Group I Group II
The optimization for frequency assignment aims to maximize the energy transfer ERX Max
= ERX + ERX to
the requesting sensor node. The energy transferred by the RF signal at a given frequency point is the product of the
power spectral density and the circuit frequency response, i.e., S 1/2 ( f )H( f ). Thus, the useful components that need
to be maximized are the first two terms of (6), which give the constructive energy contribution of the ETs of the two
groups.

Given : S 1 ( f ), S 2 ( f ), and H( f )
To find : f1 , f2 (5)
To Maximize :
 f1 +ε  f2 +ε
Max
ERX = S 1 ( f )H( f )d f + S 2 ( f )H( f )d f
f1 −ε f2 −ε
 γ  f1 +ε
−( S 2 ( f )H( f ) d f + S 1 ( f )H( f ) d f ) (6)
f2 −ε γ

destructive interference

Subject to :
d(S 1 ( f )H( f )) 
 f =γ < 0 (7)
df
d(S 2 ( f )H( f )) 
 f =γ > 0 (8)
df

The two constraints of the above optimization ensure that the spectrum shapes of the Group I and Group II
ETs does not overlap completely. At the point of the intersection of the PSD curves S 1 ( f ) and S 2 ( f ), which is called
the cross-over point γ, the slope of the curves must be positive and negative, respectively. This is calculated by
differentiating the respective PSD plots at γ, to ensure that one of them increases (positive slope) while the other falls
(negative slope). With the resulting dual-frequency wireless energy transfer, both groups of ETs can be simultaneously
active. The final part of this stage involves letting the ETs know that they are cleared for energy transmission through
an Acknowledgement (ACK) packet. Simulation results reveal that RF-MAC largely outperforms the unslotted CSMA
in both average harvested energy and average network throughput, specifically, 112% average network throughput
improvement.

3.3. A Dual-band Wireless Energy Transfer Protocol for Heterogeneous Sensor Networks Powered by RF Energy
Harvesting (RF-HSN)

In [11], the authors consider a network architecture that consists of two types of RF energy harvesting capable
sensors, shown in Figure 6(a). One class of sensors harvests RF energy on the DTV band (614 MHz), called Type
I, while another uses the 915 MHz ISM band, called Type II. While Type I has the advantage of scavenging
existing radiation without any need of new transmitter equipment, it is subject to the schedule followed by the TV
stations, and is highly location-dependent requiring a clear line of sight. As opposed to this, Type II can be carefully
controlled through dedicated ETs, though this introduces hardware requirements, impairs ongoing communication in
 Prusayon Nintanavongsa / Energy Procedia 56 (2014) 414 – 422 421

ϵϭϱD,nj͕͘WĐƵƌǀĞdLJƉĞ//

WŽǁĞƌ;ǁͿ
dzW/ĚĞǀŝĐĞƐ

d
dzW//ĚĞǀŝĐĞƐ ϲϭϰD,nj͕͘WĐƵƌǀĞdLJƉĞ/

ƚĐŚ dŝŵĞ;ƐͿ

(a) (b)

Fig. 6. Heterogeneous RF energy harvesting sensor network (a) and power curves of RF energy harvesting sensors in charging state (b)

the ISM band. The effective energy transfer under these circumstances is investigated and then used to design a
protocol called RF-HSN that optimizes the energy delivery to sensors with the optimal duty cycle. Heterogeneous
network raises several key concerns on how to manage effective charging among these sensors that harvest in multiple
bands. For instance, (i) how to sustain the communication among heterogeneous sensors when the primary source of
wireless energy suddenly becoming unavailable, (ii) how to optimally deliver wireless energy to the heterogeneous
sensor networks, and (iii) the challenges in aggregating the charging action of multiple sensors.
The energy transfer stage begins when the ET sends out the the Request to Charge (RTC) packet, at 915 MHz
(operating frequency of Mica2 mote), offering to transfer wireless energy to the sensors in the network. Consequently,
both Type I and Type II devices can hear the RTC packet sent by the ET. The sensors that received the RTC then
acknowledge the this packet by sending back a response, called as energy pulses. Once the ET receives the energy
pulses from the responding sensors, it estimates the average power that sensors will receive during the charging
process. After the charging duration is over, the ET monitors the average residual energy of the sensors and sends out
the RTC once the pre-set threshold is reached.

3.3.1. Grouping of the responding sensors

The sensors that hear the RTC packet reply back with a single, constant energy pulse called Clear to Charge (CTC),
sent out on either 614 MHz or 915 MHz, depending on the operating frequency of the harvesting module equipped.
The constant energy pulse is emitted in a time slot, beginning from the instant of completion of the RTC packet.
Consequently, Type I devices transmit constant energy pulses in the same time slot that Type II devices transmit
their constant energy pulses. Note that there will be no interference in differentiating the band of received energy
pulses at the ET since Type I and Type II devices respond to the RTC packet in different spectrum bands.
The ET that sent the initial RTC estimates the time it needs to transmit the energy signal (charging time) based
on the signal strength of the received CTC pulses on both frequencies. Unlike classical data communication, it is
not important for the ET to know the required energy of each sensor. Rather, the energy requirement calculations are
based on how much energy is needed by the two type of devices separately. The authors define this cumulative energy
T ype I T ype II
as ERX and ERX , respectively, which are calculated by the RTC issuing ET from the received pulses of both
frequency bands.

3.3.2. Optimization function for optimal energy delivery

T ype I T ype II
The aim of the optimization formulation is to maximize the energy transfer ERX Max
= ERX + ERX to the
coverage area of the ET with the highest duty cycle. The energy transferred by the RF signal at each frequency is
the time integral of the power (charging characteristic) curve at the corresponding frequency, shown in Figure 6(b).
Note that the power curve is a function of the received signal strength (RSS) and the charging time. Thus, the useful
components that need to be maximized are two terms of (10), which give the total energy transfer to both Type I
and Type II devices.
422 Prusayon Nintanavongsa / Energy Procedia 56 (2014) 414 – 422

Given : PcurveI (RS S I , tch ), PcurveII (RS S II , tch )

To find : tch (9)
To Maximize :

Max
ERX = PcurveI (RS S I , tch )dt
tch

+ PcurveII (RS S II , tch )dt (10)
tch

Subject to :
d(PcurveI (RS S I , tch )) d(PcurveII (RS S II , tch )) d(Pdischarge(tch ))
+ − is maximum (11)
 dt dt dt
PcurveI (RS S I , tch )dt > EminI (12)
t
 ch
PcurveII (RS S II , tch )dt > EminII (13)
tch

With the resulting dual-frequency wireless energy transfer, both groups of devices can be simultaneously active.
The final part of this stage involves letting both Type I and Type II devices know that they are about to enter the
charging state through an Acknowledgement (ACK) packet. Simulation results show that RF-HSN offers over 59%
higher duty cycle and 66% average network throughput improvement over the classical CSMA MAC protocol.

4. Conclusions

RF energy harvesting sensor network is an emerging technology thanks to the recent advancement in semiconduc-
tor technology and fabrication process. In this paper, the concept of RF energy harvesting is presented as well as its
realization through RF energy harvesting circuit design. A recent dual-stage RF harvesting circuit design, aimed at
maximizing harvesting efficiency, is investigated. Various protocols specifically designed for RF energy harvesting
network are also discussed in details. These protocols offer benefits over existing protocol by taking energy replen-
ishment capability into an account as well as complementing the RF energy harvesting circuit design.

References

[1] H. Yan, J.G. Macias Montero, A. Akhnoukh, L.C.N. de Vreede and J.N. Burghart. An Integration Scheme for RF Power Harvesting. 8th
Annual Workshop on Semiconductor Advances for Future Electronics and Sensors, Veldhoven, the Netherlands, 2005.
[2] P. Nintanavongsa, U. Muncuk, D. R. Lewis, and K. R. Chowdhury, Design Optimization and Implementation for RF Energy Harvesting
Circuits. IEEE Journal on Emerging and Selected Topics in Circuits and Systems, vol. 2, no. 1, pp. 24–33, Mar. 2012.
[3] Powercast Corporation, Lifetime Power Evaluation and Development Kit. [Online] [Link]
[4] Crossbow Technology, Inc. [Online] [Link]
[5] Z. A. Eu, H.P. Tan, and W. K. G. Seah, Design and performance analysis of MAC schemes for wireless sensor networks powered by ambient
energy harvesting. Ad Hoc Networks, vol. 9, no. 3, pp. 300–323, 2011.
[6] J. Kim and J. W. Lee, Energy adaptive MAC protocol for wireless sensor networks with RF energy transfer. Proc. of IEEE Intl. Conference on
Ubiquitous and Future Networks (ICUFN), pp. 89–94, 2011.
[7] J. Kim and J. W. Lee, Performance analysis of the energy adaptive MAC protocol for wireless sensor networks with RF energy transfer. Proc.
of IEEE Intl. Coverage and Transmission Conference (ICTC), pp. 14–19, 2011.
[8] P. Nintanavongsa, R. Doost, M. Di Felice, and K. R. Chowdhury, Device characterization and cross-layer protocol design for RF energy
harvesting sensors. Elsevier Journal of Pervasive and Mobile Computing, vol. 9, no. 1, February 2013.
[9] K. Pappa, A. Athanasopoulos, E. Topalis, S. Koubias, Implementation of Power Aware Features in AODV for Ad Hoc Sensor Networks. A
Simulation Study. Proc. of IEEE Emerging Technology and Factory Automation, Apr 2007.
[10] P. Nintanavongsa, M.Y. Naderi, and K. R. Chowdhury, Medium Access Control Protocol Design for Sensors Powered by Wireless Energy
Transfer. Proc. of IEEE International Conference on Computer Communications (INFOCOM), pp. 150–154 , April 2013.
[11] P. Nintanavongsa, M.Y. Naderi, and K. R. Chowdhury, A Dual-band Wireless Energy Transfer Protocol for Heterogeneous Sensor Networks
Powered by RF Energy Harvesting. IEEE International Computer Science and Engineering Conference (ICSEC), pp. 400–405, September
2013.