1768 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO.

5, MAY 2015

A Multi-Band Stacked RF Energy Harvester

With RF-to-DC Efficiency Up to 84%
Véronique Kuhn, Cyril Lahuec, Fabrice Seguin, and Christian Person, Member, IEEE

Abstract—The aim of this paper is to show the possibility to har- TABLE I

vest RF energy to supply wireless sensor networks in an outdoor TYPICAL ELECTRONIC DEVICES CONSUMPTION [2]
environment. In those conditions, the number of existing RF bands
is unpredictable. The RF circuit has to harvest all the potential
RF energy present and cannot be designed for a single RF tone. In
this paper, the designed RF harvester adds powers coming from
an unlimited number of sub-frequency bands. The harvester’s
output voltage ratios increase with the number of RF bands.
As an application example, a 4-RF band rectenna is designed.
The system harvests energy from GSM900 (Global System for
Mobile Communications), GSM1800, UMTS (Universal Mobile
Telecommunications System) and WiFi bands simultaneously.
RF-to-dc conversion efficiency is measured at 62% for a cumula- Unlike most energy sources, the EM or RF energy sources
tive 10-dBm input power homogeneously widespread over the are continuously available. Although being ubiquitous in our
four RF bands and reaches 84% at 5.8 dBm. The relative error daily live, the drawback of RF sources is that the power density
between the measured dc output power with all four RF bands ON
and the ideal sum of each of the four RF bands power contribution
they can provide is low [3]. In order to assess the practicable
is less than 3%. It is shown that the RF-to-dc conversion efficiency locations of ambient RF energy harvesters, the RF power
is more than doubled compared to that measured with a single RF density level in the sensor environment must first be quantified.
source, thanks to the proposed rectifier architecture. RF power density levels reports such as [3]–[6] have summa-
Index Terms—Multiband, rectenna, rectifier, RF energy har- rized measurements from several European countries. In an
vesting, sensor. urban situation, for emission and reception points distanced
from 50 to 200 m, the RF power density ranges from 1 to
0.15 mW/m for GSM900 and GSM1800 base stations and
I. INTRODUCTION from 0.5 to 0.1 mW/m for a UMTS base station. For distances
over 300 m, the RF power density is 10 W/m at maximum.
I N RECENT years, the decrease of electronic components
consumption has led to the development of wireless de-
vices. An interesting application concerns wireless sensor net-
These values have been calculated from data of the French
National Frequencies Agency (ANFR) [7].
Previous works on RF harvesting circuits focused on a
works (WSNs). The wireless sensors communicate with each
single operating frequency [4], [8]–[10]. When multiple RF
other and gather different data such as temperature, pressure, hu-
energy sources are available, the amount of energy harvested
midity, etc. They are used in various scenarios such as intelligent
can be increased if the system is designed to work over a
monitoring systems for office, domotic, medical, or military ap-
wide frequency band as proposed in this paper. From theory
plications [1]. Most applications require small low-power sen-
analysis supported by simulation results and measurements,
sors for autonomy, environmental, and electromagnetic (EM)
an RF harvester has been designed to cover the GSM900,
pollutions. The power consumption of typical wireless elec-
GSM1800, UMTS2100, and WiFi bands. The GSM900 en-
tronic devices is illustrated in Table I [2]. Today these sensors
tire band is from 876 to 959 MHz including UpLink (UL)
are powered by batteries. Despite significant progress, batteries
and DownLink (DL) modes. The GSM1800 is from 1710 to
still have a limited lifetime and their replacement is often com-
1880 MHz (including UL and DL modes). The UMTS band
plicated. This explains the motivation to find alternative ways to
is from 1920 to 2170 MHz (including UL and DL modes),
power these objects. A promising method consists of harvesting
and finally the WiFi band is from 2.41 to 2.48 GHz. The
energy from the ambient environment of the sensor nodes. For
RF harvester measurements show that the proposed stacked
instance, thermal gradients, mechanical vibrations, light, or EM
multi-band RF harvester is able to supply typical electronic
waves could serve such a purpose.
devices such as those listed in Table I.
This paper is organized as follows. Section II presents the
Manuscript received July 21, 2014; revised December 15, 2014; accepted
state-of-the-art and the rectifier impact on the multiband RF
March 14, 2015. Date of publication April 06, 2015; date of current version
May 04, 2015. This work was supported by Pracom. harvester architecture. Section III explains the theoretical anal-
The authors are with the Lab-STICC, Telecom-Bretagne, 29238 Brest Cedex ysis of the designed rectifier in the multiband RF harvester.
3, Brittany, France (e-mail: [email protected]).
Section IV presents the multiband RF harvester simulations and
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. Section V shows the multiband RF harvester measurements.
Digital Object Identifier 10.1109/TMTT.2015.2416233 Section VI concludes this paper.

0018-9480 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
KUHN et al.: MULTI-BAND STACKED RF ENERGY HARVESTER WITH RF-TO-DC EFFICIENCY UP TO 84% 1769

II. STATE-OF-THE-ART: RECTIFIER IMPACT ON THE

MULTIBAND RF HARVESTER ARCHITECTURE
The goal of an ambient RF energy harvester is to convert the
RF energy received from RF ambient sources into dc power.
Such harvesters are called rectenna, i.e., rectifying antenna. A
typical rectenna circuit consists of a receiving antenna followed
by an RF bandpass filter, a rectifier, a low-pass filter, and a
load. Previous works focusing on RF harvesting circuits have
been presented for a single RF tone [4], [8]–[10]. The func-
tionality of such harvesters is drastically degraded if the op-
erating frequency is changed from the optimal resonance fre-
quency. If the sensor location is the outdoor environment, a
single-tone rectenna is not suitable to supply the sensor. Indeed,
in an outdoor environment, the predominant frequencies may
significantly differ from one location to another. Other works
report on multiple single frequency structures [11]–[18]. This
type of rectenna benefits from the accumulation of RF radiation
for several frequencies and a higher amount of energy can be
harvested [13].

A. Multiband RF Harvester Topology Choice

In order to harvest energy from multi RF bands, several RF
harvester topologies can be investigated. The main difference
lies in the RF bandpass filter design. The filter functionality is
to match the antenna impedance and the conjugate impedance
of the rectifier input. Furthermore, the rectifier input impedance
varies as a function of the frequency and the incident power. The
antenna impedance may also vary as a function of the frequency.
Thus, it is easier to adapt impedances at a single frequency than
over an RF band. Based on this, the adaptation on multiple RF
bands induces two kinds of losses: the ones due to impedance
mismatch and those due to the filter complexity.
The losses due to the difficulties to match the impedance over
a large bandwidth can be illustrated using [15] and [12]. They Fig. 1. Multi-band RF harvesters: (a) with only one designed RF bandpass filter
both use the architecture shown in Fig. 1(a). Due to the imped- for an RF broadband, (b) for multiple single frequencies, and (c) proposed ar-
ances variation, an RF band of a few hundred of MHz induces chitecture.
impedance mismatch, and thus a decrease of the RF-to-dc con-
version efficiency. The RF-to-dc conversion efficiency is only
8% on a 1550-MHz RF band for an incident power of 20 dBm The losses due to the filter components are too important to
in [15]. However, this efficiency is almost doubled to 15% in obtain the correct adaptations between the antenna and the
[12] for the same incident power, and same topology, but on rectifier. As the aim is to harvest several entire RF bands, this
a 300-MHz RF band. Thus, in order to improve the RF-to-dc topology has not been chosen.
conversion efficiency, it is better to harvest energy from several Finally, harvesting multiple RF frequencies can be done by
small RF bands rather than a single large one. stacking several rectennas. In this case, the RF bandpass filter
The losses due to the filter complexity can then be illus- is designed to select a specific frequency and the rectennas dc
trated with data from dual-frequency RF harvesters described outputs voltages are added to the same load [16], [17]. In most
in [10], [14], [15], [18], or [30]. They used the architecture publications about this architecture, the quality assessment of
of Fig. 1(b). The perfect adaptation on multiple frequencies the output dc combination is not taken into account. In [17], with
requires high-order RF bandpass filters with drawbacks con- an incident power of 15 dBm per frequency, a 45% RF-to-dc
cerning induced complexity. Even for frequencies far apart (at conversion efficiency is obtained at 0.9 and 1.8 GHz and 25% at
least one octave), the RF-to-dc conversion efficiency remains 2.45 GHz. Despite these interesting results, the cumulative effi-
below 30%, except for [18], which reaches 40 % at 1.8 and ciency is not given if the two or three RF tones are present. Fur-
2.1 GHz. This is achieved by means of a complex filtering com- thermore, the issue of this topology is that no energy is harvested
posed of a low-pass filter and three stopband filters. Moreover, outside the selected frequencies. Moreover, this architecture is
the functionality of such harvesters is drastically degraded if the not suitable for compact applications due to the number of an-
operating frequencies are changed from the optimal resonance tennas used. Fig. 1(c) illustrates the architecture proposed in this
frequency. Furthermore, increasing the number of adaptive paper. The association of the RF bandpass filter, the rectifier, and
frequencies drastically decreases the RF-to-dc efficiency [15]. the low-pass filter is termed an RF branch. The number of RF
1770 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 5, MAY 2015

Fig. 3. Rectifier schematic proposal.

Fig. 2. Three topologies of rectifier: (a) series, (b) voltage doubler, and to switch on the numerous diodes of a Greinacher rectifier
(c) Greinacher. compared to the other rectifiers.
Based on this discussion, the rectifier must have a differential
output to prevent dc interferences between the RF branches. Its
branches is not restricted. A single access wideband antenna is number of diodes must also be limited to work in an outdoor
connected to the circuit input. It allows a more compact struc- environment since the RF density power is low.
ture. In order to improve the RF-to-dc conversion efficiency,
the circuit must take into account the impedance matching con- III. THEORETICAL ANALYSIS OF THE DESIGNED RECTIFIER
IN THE MULTIBAND RF HARVESTER
straints of the different components involved, i.e., parallel rec-
tifiers and antenna. In order to match each parallel rectifier at The topology of the multiband RF harvester has been chosen
the dedicated frequencies, RF bandpass filters are used. The and the constrains with the rectifier have been explained. The
impedance matching is done for a chosen bandwidth. It is con- goal now consists of finding the best compromise for the recti-
sequently less sensitive to RF spectrum availability. For each fier used in the multiband RF harvester chosen.
RF branch, the bandwidth is chosen to cover all the potential
RF power density in a standard RF band, such as GSM900 or A. Rectifier Topology Proposed
GSM1800. Finally, the RF branch dc outputs are connected to- The chosen rectifier needs to have a limited number of diodes.
gether in order to sum the harvested power. Any RF signals must Looking at the existing rectifiers, only the series or doubler recti-
be blocked by a low-pass filter to pass only the dc component fier are suitable. By suppressing the voltage doubler connection
to the output load. to the ground, this rectifier benefits from a differential output
as wanted. The modified voltage doubler is illustrated in Fig. 3.
B. Problematic Linked With Rectifier Topology The two diodes mounted in parallel allow the rectifier starting
from only one diode threshold voltage. During the negative half
Several rectifier topologies exist depending on the incident period of , turns on and turns off. Thus, capacitor
power and frequency of the incident signal. Some examples is charged through . Likewise, during the positive half
are given in Fig. 2. The series topology, Fig. 2(a), is pre- period of , turns off and turns on. Thus, capacitor
ferred to the other ones when considering low incident power is charged through with input current and charges stored
[14]–[16]. Other structures, such as Greinacher [11], illustrated in . Finally, the differential output voltage, , equals twice
in Fig. 2(c), have higher dc voltage levels than series structures . Considering the diode threshold voltages, and ,
for a given input power level. Furthermore, the dc combination is
of several RF branches directly affects the overall efficiency
of the architecture [15]. Indeed, the total power harvested with (1)
parallel connection is less than that with series connection [19].
Furthermore, the current equalization in parallel connection
is worse than voltage equalization in series connection [20].
B. DC Combination of all RF Input Contributions
These characteristics are due to the fact that a parallel connec-
tion creates more dc interferences between the RF branches The circuit can be extended to stages, as illustrated in Fig. 4.
compared to a series connection [19], [20]. Thus, due to their The first RF branch is referenced to the ground and the th RF
single-ended output, the series or the voltage doubler rectifier, branch to the th RF branch. The load is connected be-
illustrated in Fig. 2(a) and (b), are not suitable. Thanks to its tween the th circuit and the ground. It is now shown that voltage
differential output, the Greinacher rectifier is a good candidate. across the load is the sum of the input voltages .
The dc output voltages of several Greinacher rectifiers stacked Starting from the top rectifier in Fig. 4, the dc output current,
can be summed without creating dc interferences between each , is first expressed as a function of and . As-
RF branches, as demonstrated in [11]. Furthermore, the unused suming and since diodes and ,
RF branch is not seen as a load for the other one. Despite its form a voltage divider, voltage can be expressed as
interesting characteristics, the Greinacher rectifier has twice
as many diodes than a voltage doubler and four times more
than a series rectifier. Thus, higher incident power is necessary (2)
KUHN et al.: MULTI-BAND STACKED RF ENERGY HARVESTER WITH RF-TO-DC EFFICIENCY UP TO 84% 1771

where is the diode internal series resistance, is the dc

current at the output. The current in the Schottky diode ,
which is forward biased, is then expressed as

(3)

where is the thermal voltage, is the diode saturated cur-

rent, and is the ideality factor.
From the Bessel function,

(4)

the current in can be expressed as

Fig. 4. Multiband RF harvester schematic.

(5)
cross the depletion region. For the Metelics case, in the deple-
where and are tion region, is equal to 1, and if biased close to its threshold
voltage, is around 1.3 [21]. In those conditions, assuming that
(6) k , is equal to 1.8e , making neg-
ligible compared to . Furthermore,
(7) is always negligible compared to . Thus, (10) can then
be simplified as
The dc part of the current is then extracted,
(11)

By iteration, is expressed as a function of the incident RF

signals, ,

(12)
(8)
is a sum of all the input contributions, , and ,

From (8), it can be deduced that (13)

For high values of such as 1 V, is equal to 1 and

is equal to 0.07 V negligible compared to . In those con-
ditions, for values of below 0.2 V, is expressed as
(12) and for values above 0.2 V as (14),
(9)
(14)

Thanks to this sum, is not disturbed by dc interferences

between the different RF branches on, , or off,
(10) . Starting from (14), can be also expressed as
a function of . Indeed,
Equation (10) can be simplified considering the characteristics
of the chosen Schottky diode. In this paper, the Schottky (15)
Metelics MSS20-141 diode is chosen considering its attractive
(16)
characteristics, namely, a threshold voltage mV
and a junction capacitance pF. The saturated current Thus, is the sum of the previous output voltage
of this diode, , is 3.5 A, its internal series resistance, , and its input voltage amplitude. In a particular case, if is
is 20 . The ideality factor typically varies from 1 to 2, the same for each RF branch, it can be deduced that
depending on the fabrication process. This factor is mainly
accounting for carrier recombination as the charge carriers (17)
1772 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 5, MAY 2015

Fig. 6. Rectifier input impedance seen on the Smith chart.

Fig. 5. Rectifier input impedance : real part (upper plane) and imaginary
part (lower plane) depending on the incident power and on the frequency.

IV. MULTIBAND RF HARVESTER SIMULATIONS Fig. 7. RF bandpass filter of the harvester topology.

A. Matching Network Designed

TABLE II
Thanks to the maximum power transfer theorem, the highest TYPICAL SENSOR CHARACTERISTICS
power is transferred to the load if the source and load complex
impedances are complex conjugates. This is achieved by means
of an impedance matching network placed between the antenna
and the rectifier. Whenever a source or a load has a reactive
component, the adaptation depends on the frequency for which
it is designed. Thus, the matching network is also considered as
an RF bandpass filter. The most frequently used matching net-
works are the , the and the networks [22]. To design the
impedance matching network, the input impedance of the rec-
tifier, , is first determined. The receiving antenna is mod-
eled as a simple 50- resistor at the selected frequencies. The
RF harvester is simulated using the Advanced Design System feed. Table II summarizes the dc supply, power consumption,
(ADS) software from Agilent Technologies. To take into ac- and equivalent impedance of several recent sensors. Note that
count the nonlinear behavior of the rectifier, the harmonic-bal- the trend in minimum supply voltage is close to the threshold
ance method is employed. As shown in Fig. 5, varies with voltage around 0.3 V. Their equivalent impedances vary from
frequency and incident power, its Smith chart is illustrated in few k to few dozens of k . In order to choose a realistic load
Fig. 6. In order to match the antenna impedance and for the value for simulations and measures, an average value of 11 k
selected frequencies, the chosen RF bandpass filters topology is chosen. Next, the impedance matching network is designed
is a variant of a T-network filter (Fig. 7). In order to increase for this load.
the bandwidth and suppress undesired high frequencies close
to the resonance frequency, a capacitor in parallel with an in- C. Comparison Between Simulation and Theoretical Results
ductor replaces the usual shunt capacitor of the T-network [23]. The output voltages , for to , have been simu-
The chosen bandwidth is 200 MHz, in order to cover GSM900, lated as a function of the incident power set for each four fre-
GSM1800, UMTS, and WiFi bands. The component values of quencies. In Fig. 8, the simulation results highlight that the dc
the four RF bandpass filters have then been determined with an output voltage with the four RF branches activated, , is
ADS optimization option. four times higher than that with a single partial RF harvester
switched on. There is no loading effect between the different dc
B. Simulation Setup in ADS outputs. The minimum input power to activate the harvester is
The RF harvester has been designed and simulated using 20 dBm, which is in the power range available in urban envi-
ADS. In the simulations, the receiving antenna is modeled as a ronments.
power generator with four tones frequencies set to 0.9, 1.8, 2.1, Table III summarizes a few selected values of . Those
and 2.45 GHz and a 50- output impedance. are obtained through simulations accordingly to the incident
The low-pass filter is a simple 68-pF capacitor. The RF har- power value, . Fig. 8 confirms the correlation between (12)
vester load models the equivalent impedance of the sensor to and the simulation results. For equal to 20 dBm, is
KUHN et al.: MULTI-BAND STACKED RF ENERGY HARVESTER WITH RF-TO-DC EFFICIENCY UP TO 84% 1773

Fig. 10. Prototype photograph.

Fig. 8. Simulated dc output voltage as a function of the incident power set for
four RF tones and comparison with (12) for k .

TABLE III
CORRESPONDENCE BETWEEN INCIDENT POWER INPUT
AND VOLTAGE AMPLITUDE ARF,I

Fig. 11. Simulated and measured of the RF harvester at dBm.

Fig. 9. Error in percentage between values obtained with (12) and the simula-
tions results as a function of per RF branch.

close to the threshold voltage diodes so is equal to 1.3. For Fig. 12. Measured dc output voltage as a function of per generator for one,
other values, is equal to 1. Fig. 9 illustrates the error in two, three or four RF tones.
percentage between simulation results and values obtained with
(12). The percentage value is below 3% whatever values. V. MULTIBAND RF HARVESTER MEASUREMENTS
These simulation results also validate the assumptions done to To validate the approach described in the previous section,
simplify (12). a prototype harvesting energy from four energy bands is fab-
1774 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 5, MAY 2015

Fig. 14. Dipole antenna photograph.

Fig. 13. Measured and simulated RF-to-dc conversion efficiency as a function

of per generator for one, two, three, or four RF tones.

ricated and tested. The chosen bands are GSM900, GSM1800, Fig. 15. Measured gain of the dipole antenna as a function of the frequency.
UMTS, and WiFi with a 200-MHz RF band around each The
prototype is fabricated on a Rogers R04003 substrate ( ,
thickness 508 m), with 0402 SMD (surface mounted) compo-
nents, and is shown in Fig. 10.

A. RF Harvester Measurements

Fig. 11 illustrates the simulated and measured input port

voltage reflection coefficient, i.e., , of the RF harvester for a
15-dBm input level. These results confirm a good correlation
between simulations and measures. The impedance matching
is achieved at the specified frequencies, i.e., 0.9, 1.8, 2.1, and
2.45 GHz. In order to validate the RF harvester, four R&S
SMJ100A power generators are directly connected to its input
by means of a power combiner. The load, , is a 11-k
resistor. The four power generators are set at: 0.9, 1.8, 2.1, and
2.45 GHz. They deliver the same power from 20 to 0 dBm. Fig. 16. Measured dc output voltage as a function of distance for k .
As Fig. 12 shows the measured dc output voltages are in line
with the simulated ones illustrated in Fig. 8. Fig. 13 represents B. Rectenna Measurements in Controlled Environment
the simulated and measured RF-to-dc conversion efficiencies Next, the RF harvester is tested considering real antennas as
of the RF harvester as a function of the input power delivered RF feeders illuminated by a controlled multi-frequency source.
by each power generator. The proposed stacked harvester does The harvester is connected to a 10 10 cm wideband dipole
enhance the RF-to-dc conversion efficiency as the number of antenna with a reflector plane (Fig. 14). This characteristic ex-
sources increases. From one to four sources, the efficiency plains the high gain value of 6 dBi on the entire RF band from
improvement is 20% over almost all of the range per gen- 0.9 to 3 GHz, as shown in Fig. 15. Four power generators, (R&S
erator. This is due to a good impedance adaptation between the SM300, Agilent E4422B, HP 8620C, and Giga GP2218), each
antenna and all the RF branches and also an almost perfect dc with a 15-dBm incident power, , are connected to in-
sum. If some of the incident powers are null or do not have the dividual optimized dipole antennas operating at 0.9, 1.8, 2.1,
same value, the impedance adaptation, and thus, the RF-to-dc and 2.45 GHz, respectively. A voltmeter is used to measure the
conversion efficiency, are not optimum. dc output voltage across the 11-k load. The RF transmitters
KUHN et al.: MULTI-BAND STACKED RF ENERGY HARVESTER WITH RF-TO-DC EFFICIENCY UP TO 84% 1775

Fig. 17. Measured dc output power and RF-to-dc conversion efficiency as a

function of the total RF power density for four RF sources emitting.

Fig. 19. Outdoor measurements. (a) Test-bench. (b) Photograph.

TABLE IV
FOUR RF-BAND HARVESTER MAIN CHARACTERISTICS

Fig. 18. Measured output power as a function of distance and error between
ideal and implemented power summations in %.

and the rectenna are separated from 50 to 250 cm by a step of

50 cm and the dc voltage is measured. Fig. 16 shows the dc
output voltage as a function of this distance with only one source
emitting at either 0.9, 1.8, 2.1, or 2.4 GHz, the others sources
being switched off. This measurement is then repeated with the
four sources switched on simultaneously, the results of which
are also shown in Fig. 16. From this figure, it is clear that the dc
output voltage with the four sources switched on is almost equal
to the dc sum of each of the four RF tones’ dc contribution. The
dc output voltage measured with the four RF tones ON is also where is the input power level at each frequency ,
above the threshold voltage of typical sensors. is the measured gain of the antenna, is the speed of light, and
In Fig. 17, the RF-to-dc conversion efficiency is plotted as a is equal to 0.9, 1.8, 2.1, and 2.4 GHz.
function of the received RF power density. The power density This efficiency is higher than 50% for RF power density over
is defined as the power per unit area normal to the direction of 1 W/cm and goes up to 84% at 8 W/cm RF power den-
propagation usually expressed in units of watts per square meter, sity. Fig. 17 also shows that the dc output power grows with
W/m . Here, the RF power density, , is calculated as RF power density. Fig. 18 shows the dc output power based on
the dc output measurements illustrated in Fig. 16. The RF har-
vester is able to directly supply all the electronic devices listed
(18)
in Tables I and II as power and voltage requirements are met.
The relative error between the dc output power calculated based
1776 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 5, MAY 2015

TABLE V
PERFORMANCE COMPARISON

Fig. 21. Measured RF-to-dc conversion efficiency as a function of the incident

Fig. 20. Measured RF-to-dc conversion efficiency as a function of the incident power for two RF sources ON simultaneously for the prototype and [14].
power for four frequencies.
15 W with k , enough to power most of the sensors
on measurements with all the RF bands ON and the ideal sum of in Table II.
each of the four RF branches power contribution is also plotted
in Fig. 18. The error is less than 2%. This result confirms that the D. Performance Comparison
dc summation is done without any dc loading effects between Table IV summarizes the four-RF band harvester main char-
the four dc outputs. acteristics. Table V and Fig. 20 summarize this work along with
previously published works reporting measurements. Only pub-
C. Rectenna Measurements in Outdoor Environment
lications with measurements considering the incident power,
The rectenna has been tested in outdoor conditions, as illus- , below 0 dBm have been taken into account. Fig. 20 illus-
trated in Fig. 19(a) and (b). The RF harvester is placed 50 m trates the RF-to-dc conversion efficiency as a function of the
away from a GSM1800 and UMTS base station. A dosimeter incident power for an RF source ON set either at 0.9, 1.8, 2.1,
is placed near the RF harvester. The RF power density is mea- or 2.45 GHz. The efficiencies obtained are almost the same for
sured at 1.2 mW/m and 375 W/m for GSM1800 and UMTS each of the four frequencies. These results are compared to other
base stations, respectively. The dc output power is measured at works done in the same conditions. The multiband RF harvester
KUHN et al.: MULTI-BAND STACKED RF ENERGY HARVESTER WITH RF-TO-DC EFFICIENCY UP TO 84% 1777

designed presents an efficiency at least 15% better compared to [13] V. Kuhn, F. Seguin, C. Lahuec, and C. Person, “A multi-tone RF energy
state-of-the-art multi-band RF harvesters [11], [16], [29], [32]. harvester in body sensor area network context,” in Antennas Propag.
Conf., Loughborough, U.K., 2013, pp. 238–241.
Furthermore, the efficiency obtained with two RF sources ON is [14] H. Sun, Y.-X. Guo, M. He, and Z. Zhong, “A dual-band rectenna using
compared to [14] in Fig. 21. is the total incident power of the broadband yagi antenna array for ambient RF power harvesting,” IEEE
two sources. A highly selective filter with two frequencies is de- Antennas Wireless Propag. Lett., vol. 12, pp. 918–921, 2013.
[15] A. Collado and A. Georgiadis, “Conformal hybrid solar and electro-
signed in [14]. Both prototypes have similar results at very low magnetic (EM) energy harvesting rectenna,” IEEE Trans. Circuits Syst.
incident powers. However, from a total power of 10 dBm, the I, Reg. Papers, vol. 60, no. 8, p. 10, Aug. 2013.
efficiency of the proposed stacked rectenna is better than [14]. [16] M. Pinuela, P. D. Mitcheson, and S. Lucyszyn, “Ambient RF energy
harvesting in urban and semi-urban environments,” IEEE Trans. Mi-
At 5 dBm, the RF-to-dc efficiency is 49% for [14] and 60% crow. Theory Techn., vol. 61, no. 7, pp. 12–28, Jul. 2013.
for the prototype. [17] S. Keyrouz, H. J. Visser, and A. G. Tijhuis, “Multi-band simultaneous
radio frequency energy harvesting,” in Eur. Antennas Propag. Conf.,
Göteborg, Sweden, 2013, pp. 3058–3061.
VI. CONCLUSION [18] Y. H. Suh and K. Chang, “A high-efficiency dual-frequency rectenna
A four-RF band rectenna has been designed for RF energy for 2.45- and 5.8-GHz wireless power transmission,” IEEE Trans. Mi-
crow. Theory Techn., vol. 50, no. 7, pp. 1784–1789, Jul. 2002.
harvesting in the context of the WSN. A new architecture has [19] N. Shinohara and H. Matsumoto, “Experimental study of large
been designed covering the GSM900, GSM1800, UMTS, and rectenna array for microwave energy transmission,” IEEE Trans.
WiFi bands. The fabricated prototype shows a 84% of RF-to-dc Microw. Theory Techn., vol. 46, no. 3, pp. 261–268, Mar. 1998.
[20] T. Miura, N. Shinohara, and H. Matsumoto, “Experimental study of
conversion efficiency at 0-dBm input power set on each of the rectenna connection for microwave power transmission,” Electron.
four RF branches. The efficiency is more than doubled in the Commun., Jpn., vol. 84, no. 2, pp. 27–36, 2001.
presence of all of the RF sources compared to a single-tone [21] C. R. Crowell and S. M. Sze, “Current transport in metal–semi-
conductor barriers,” Solid State Electron., vol. 9, no. 11–12, pp.
rectenna. Final results confirm the RF harvester workability to 1035–1048, 1966.
supply small sensors. [22] M. Thompson and J. K. Fidler, “Determination of the impedance
matching domain of impedance matching networks,” IEEE Trans.
Circuits Syst. I, Reg. Papers, vol. 51, no. 10, pp. 2098–2106, Oct.
ACKNOWLEDGMENT 2004.
[23] H. Kanaya et al., “Energy harvesting circuit on a one-sided directional
The authors would like to thank the Satimo Company. The flexible antenna,” IEEE Microw. Wireless Compon. Lett., vol. 23, no.
supported from Pracom is greatly appreciated and acknowl- 3, pp. 164–166, Mar. 2013.
edged. [24] C. Chung and C. Yang, “An autocalibrated all-digital temperature
sensor for on-chip thermal monitoring,” IEEE Trans. Circuits Syst. II,
Exp. Briefs, vol. 58, no. 2, pp. 105–109, Feb. 2011.
REFERENCES [25] M. Hempstead, M. J. Lyons, D. Brooks, and G.-Y. Wei, “Survey of
[1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “Wire- hardware systems for wireless sensor networks,” J. Low Power Elec-
less sensor networks: A survey,” Comput. Netw., vol. 38, no. 4, pp. tron., vol. 4, no. 1, pp. 11–20, 2008.
393–422, 2002. [26] J. Van Rethy et al., “A low-power and low-voltage BBPLL-based
[2] CATRENE Working Group on Energy Autonomous Syst., “Energy sensor interface in 130 nm CMOS for wireless sensor networks,” in
autonomous systems: Future trends in devices, technology, and Design, Autom. Test Eur. Conf. Exhibit., Mar. 2013, pp. 1431–1435.
systems,” 2009. [Online]. Available: http://alexandria.tue.nl/reposi- [27] L. Li, S. T. Block, D. E. Duarte, and L. Changzhi, “A 0.45-V MOS-
tory/books/675451.pdf FETs-based temperature sensor front-end in 90 nm CMOS with a non-
[3] H. J. Visser, A. C. F. Reniers, and J. A. C. Theeuwes, “Ambient RF calibrated 3.5 C3 relative inaccuracy from 5.5 C to 105 C,”
energy scavenging: GSM and WLAN power density measurements,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 60, no. 11, pp. 771–775,
in 38th Eur. Microw. Conf., 2008, pp. 721–724. Nov. 2013.
[4] M. M. Tentzeris and Y. Kawahara, “Novel energy harvesting technolo- [28] S. W. Chen, M. H. Chang, W. C. Hsieh, and W. Hwang, “Fully on-chip
gies for ICT applications,” in Int. Appl. and Internet Symp., Turku, Fin- temperature, process, and voltage sensors,” in Proc. IEEE Int. Circuits
land, 2008, pp. 373–376. Syst. Symp., 2010, pp. 897–900.
[5] R. J. Vyas, B. B. Cook, Y. Kawahara, and M. M. Tentzeris, “E-WEHP: [29] B. Li et al., “An antenna co-design dual band RF energy harvester,”
A batteryless embedded sensor-platform wirelessly powered from am- IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 60, no. 12, pp.
bient digital-TV signals,” IEEE Trans. Microw. Theory Techn., vol. 61, 3256–3266, Dec. 2013.
no. 6, pp. 2491–2505, Jun. 2013. [30] K. Niotaki et al., “A compact dual-band rectenna using slot-loaded dual
[6] U. Bergqvist et al., “Mobile telecommunication base stations exposure band folded dipole antenna,” IEEE Antennas Wireless Propag. Lett.,
to electromagnetic fields, report of a short term mission within COST- vol. 12, pp. 1634–1637, 2013.
244bis,” in Proc. COST-244bis Short Term Mission on Base Station [31] A. Collado and A. Georgiadis, “Optimal waveforms for efficient wire-
Exposure, 2000, pp. 1–77. less power transmission,” IEEE Microw. Wireless Compon. Lett., vol.
[7] Agence Nat. Fréq., “Cartoradio,” Mar. 10, 2012. [Online]. Available: 24, no. 5, pp. 354–356, May 2014.
http://www.cartoradio.fr/cartoradio/web [32] A. N. Parks and J. R. Smith, “Sifting through the airwaves: Efficient
[8] W. Huang, B. Zhang, X. Chen, K.-M. Huang, and C.-J. Liu, “Study on and scalable multiband RF harvesting,” in IEEE Int. RFID Conf., Apr.
an S-band rectenna array for wireless microwave power transmission,” 8–10, 2014, pp. 74–81.
Progr. Electromagn. Res., vol. 135, pp. 747–758, 2013.
[9] U. Olgun, C.-C. Chen, and J. L. Volakis, “Design of an efficient ambient
WiFi energy harvesting system,” IET Microw., Antennas, Propag., vol.
6, no. 11, pp. 1200–1206, 2012. Véronique Kuhn received the Engineering Diploma
[10] C. Mikeka, H. Arai, A. Georgiadis, and A. Collado, “DTV band mi- degree in microelectronic from the ENSICAEN
cropower RF energy-harvesting circuit architecture and performance Engineering School, Caen, France, in 2004, and
analysis,” in RFID Technol. Appl., 2011, pp. 561–567. is currently working toward the Ph.D. degree in
[11] D. Pavone, A. Buonanno, M. D’Urso, and F. G. Della Corte, “Design electronics and microwaves at Telecom-Bretagne,
considerations for radio frequency energy harvesting devices,” Progr. Brest, Brittany, France.
Electromagn. Res., vol. 45, pp. 19–35, 2012. From 2004 to 2012, she was a Design Engineer
[12] A. Nimo, D. Grgic, and L. M. Reindl, “Impedance optimization of involved with audio and power management analog
wireless electromagnetic energy harvester for maximum output effi- circuits with ST Microelectronics, Grenoble, France.
ciency at W input power,” in Proc. Active and Passive Smart Struct. Her main research interests are design and optimiza-
Integr. Syst., San Diego, CA, USA, 2012, vol. 8341, pp. 83410W1–14. tion of systems in the energy harvesting context.
1778 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 5, MAY 2015

Cyril Lahuec was born in Orléans, France, in 1972. decoders and related topics, energy harvesting, reliability in nanoscale circuits
He received the B.Sc (Hon.) degree from the Uni- and systems, and implementation of clique-based neural networks.
versity of Central Lancashire, Lancashire, U.K., in
1993, and the M.Eng. and Ph.D. degrees from the
Cork Institute of Technology, Cork, Ireland, in 1999
and 2002, respectively. Christian Person (M’94) received the Ph.D. degree
He was with Parthus Technologies (now Ceva), in electronics from the University of Brest, Brest,
Cork, Ireland, while working on his doctoral France, in 1994.
research and then as a Consultant. In 2002, he Since 1991, he has been an Assistant Professor
joined the Department of Electronic Engineering, with the Microwave Department, Ecole Nationale
Telecom-Bretagne, Brest, Brittany, France, as a Supérieure des Télécommunications de Bretagne,
Full-Time Lecturer. For four months in 2011, he was with the CMOS Sensors Brest, France. In 2003, he became a Professor with
and Systems Group, University of Edinburgh, as a Visiting Researcher. His the Telecom Institute/Telecom Bretagne, Brest,
research interests are frequency synthesis, analog integrated circuit (IC) design, Brittany, Fnance, where he currently conducts
channel decoding, and biomedical applications. research with the Information and Communication
Science and Technology Laboratory (Lab-STICC
UMR CNRS 3192). He is involved in the development of new technologies
for microwave and millimeter-wave applications and systems. His activities
Fabrice Seguin was born in Talence, France, in are especially focused on the design of passive functions (filters, couplers)
1973. He received the Ph.D. degree from the Uni- and antennas, providing original solutions in terms of synthesis techniques,
versite Bordeaux 1, Bordeaux, France, in 2001. His analysis and optimization procedures, as well as technological implementation
doctoral research concerned the current mode design [foam, plastic, low-temperature co-fired ceramic (LTCC)]. He is also concerned
of high-speed current conveyors and applications in with RF integrated front-ends on Si and is currently involved in different
RF circuits. research programs dealing with system-on-chip (SoC)/system-in-package (SiP)
In 2002, he joined the Electronic Engineering De- antennas and reconfigurable structures for smart systems.
partment, Telecom-Bretagne, Brest, Brittany, France,
as a Full-Time Lecturer. With the Pôle de Recherche
Avancée en Communications (PRACOM), he is cur-
rently involved with design issues of analog channel