IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 48, NO.

7, JULY 2013

1565

Co-Integration of an RF Energy Harvester

Into a 2.4 GHz Transceiver
Jens Masuch, Member, IEEE, Manuel Delgado-Restituto, Senior Member, IEEE, Dusan Milosevic, Member, IEEE,
and Peter Baltus, Senior Member, IEEE

AbstractThis paper presents an RF energy harvester embedded in a low-power transceiver (TRX) front-end. Both the
harvester and the TRX use the same antenna and operate at the
same frequency of 2.4 GHz. To decouple the harvester from the
TRX, different concepts are proposed regarding the transmitter
(TX) and receiver (RX). To avoid loading the TX, the harvester
is decoupled with an nMOS switch that can be enabled with a
start-up rectifier. Concerning the RX, the decoupling mechanism
relies on the nonlinear input impedance of the main RF-DC
converter. The harvester also includes a supply management
circuit for over-voltage protection and charging energy storage
devices with a constant current or voltage. The energy harvester
has been co-integrated with the low power TRX in a 130 nm
CMOS process and achieves a measured peak power conversion
efficiency of 15.9%. For input power levels of at least
dBm,
it is able to charge up a supply capacitor to a regulated voltage
of 1.34 V. The impact of the harvester on the TRX performance
is measured with respect to an identical TRX front-end without
harvester, showing little impact on the TRX performance. Both
TX output power and RX noise figure are degraded by less than
0.5 dB. As an additional feature, the start-up rectifier is also used
for demodulation of On-Off-Keying (OOK) signaling, which can
be used as a secondary wake-up channel. Since the required area
for the harvester is only 0.019 mm (
% of the total active
TRX area), it can be added to the TRX at almost no cost.
Index TermsRectifier, RF energy harvesting, RF-to-DC converter, RF switch, wake-up receiver, wireless sensor networks.

I. INTRODUCTION

OWER to the nodes of a Wireless Sensor Network (WSN)

is usually provided through primary batteries. However,
batteries store a finite energy and need replacement when
depleted, what can be cost-intensive or even not feasible. To
alleviate these problems, energy harvesting from the ambient
has emerged as an alternative which aims to make the nodes
energy autonomous [1]. The feasibility of this approach is
favored by the fact that transceivers (TRXs) in WSN protocols
operate in burst-mode and remain idle most of the time; a state
in which small energy resources are consumed. Only during the
Manuscript received November 30, 2012; revised January 16, 2013; accepted
February 15, 2013. Date of publication March 29, 2013; date of current version June 21, 2013. This work has been supported in part by the Spanish Ministry of Science and Innovation under Grant TEC2012-33634 and in part by the
20072013 FEDER Program.
J. Masuch and M. Delgado-Restituto are with the Institute of Microelectronics
of Seville (IMSE-CNM-CSIC), 41092 Seville, Spain (e-mail: masuch@imsecnm.csic.es; mandel@imse-cnm.csic.es).
D. Milosevic and P. Baltus are with the Mixed-Signal Microelectronics Group
of the Eindhoven University of Technology (e-mail: D.Milosevic@tue.nl;
P.G.M.Baltus@tue.nl).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JSSC.2013.2253394

short time slots of the transmission bursts, large peak currents

are required to power the electronics of the TRX. As these
eventual power demands can be hardly covered by scavenged
resources alone, rechargeable batteries or supply capacitors
must be employed. Hence, harvesting techniques for WSNs
must be combined with the appropriate circuitry for charging
these storage elements during the sleep periods of the TRX [2],
[3]. This paper covers both aspects and presents an RF energy
harvester, together with a supply management circuitry, which
have been co-integrated with a 2.4 GHz Bluetooth Low Energy
transceiver.
The energy efficiency of an RF harvester depends on the
available input power levels and the characteristics of the energy storage device attached to the sensor node. The storage element should have a small form factor to fit in the dimensions
of the sensor, a high energy density and low leakage to extend
the discharge cycle life. Nowadays, the most promising solutions for energy storage in WSN nodes, including solid-state
thin-film rechargeable batteries [4], [5] or super-capacitors [6],
[7], require a minimum charging current on the order of a few
tens of As to combat leakage [4], [8], [9]. Hence, considering
a supply voltage of 1 V, charging a rechargeable battery or supercapacitor requires a minimum DC power on the order of 25
W [8]. It means that an incident RF power of about 100 W
(
dBm) is needed for RF energy harvesting if conversion
losses are taken into account [2]. This makes RF harvesting less
effective in dense ad hoc networks where a large area must be
flooded with RF radiation to power many wireless sensor nodes,
further considering the small antenna sizes employed [10], [11].
However, the solution works well where there is a high power
scanner or other sources in near proximity to the wireless device or where the nodes can be detached when not in use (for instance, those in a body area network) and placed nearby a power
source (for instance, a WLAN router which typically emits at 20
dBm transmission power). In both scenarios, the obvious advantage is that the sensor needs no connector for a wired charger.
Energy harvesting by RF-DC conversion may be incorporated into a wireless node by adding a dedicated antenna [2],
[12], [13] or, more elegantly, by re-using the antenna of the
TRX, as shown in Fig. 1. The latter facilitates system miniaturization but requires some kind of decoupling between the
harvesting and communication operations. This can be accomplished by using two different carrier frequencies and a dualband antenna [14]. However, this requires bulky LC matching
networks to make the impedance looking into one functional
unit high at the frequency the other functional unit is tuned to.
Contrarily, in the proposed approach, which was early presented in [15], [16], both the RF-DC conversion stage and the

0018-9200/$31.00 2013 IEEE

1566

IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 48, NO. 7, JULY 2013

Fig. 1. Architecture of the wireless sensor node with RF energy harvesting.

wireless link operate at the same carrier frequency and share

the same single-band antenna. A supply management circuit
(SMC) driven by an RF-DC converter is used to regulate the
output voltage of a storage device. The area consumption of the
RF-DC conversion stage and the SMC is only 0.019 mm in a
130 nm CMOS process, so it can be easily co-integrated with a
TRX. Additionally, the harvester can be readily extended to implement a secondary On-Off-Keying (OOK) receiver channel,
which can be used to send a wake-up call to the sensor node. By
using this feature, the transceiver can be asynchronously woken
up only when needed and remain disabled otherwise. The harvester achieves an experimental power conversion efficiency
(PCE) of 15.9% at 0 dBm input power. The efficiency performance is better than the dual-band approach in [14], and the
occupied area is one order of magnitude smaller. Most importantly, the harvester has little impact on the transceiver performance; less than 0.5 dB degradation of output power and noise
figure. This has been experimentally verified with two fabricated instances of a TRX front-end, which differ on the presence or not of the proposed harvester.
The paper is organized as follows. Section II discusses a key
issue for the co-integration of RF harvesters: the decoupling of
the harvesting functionality from the receiver and transmitter
parts of the TRX. Afterward, Section III presents the architecture of the proposed harvester, including the secondary wake-up
receiver, and Section IV gives implementation details of the
main building blocks. Then, Section V presents the measured
results and Section VI concludes the paper.

Fig. 2. Simple decoupling concept with three RF switches and their respective
signal power levels when turned-on (closed).

Fig. 3. Alternatives for a single-ended active-low RF switch to connect the

and
) to the harvester (modeled as
): with LC
antenna (modeled as
-transmission line (b) and with start-up rectifier (c). To
resonator (a), with
illustrate how the de-activated switch impairs the TX-performance, the transand
with an ideal switch
).
mitter is shown in gray (modeled as

II. RF HARVESTER/TRX CO-INTEGRATION

The front-end of the harvesting transceiver has to support
three modes of operation, namely data transmission, data reception and energy harvesting. In the simplest configuration,
the three modes are multiplexed by switches such that only one
is active at a time and the remaining two functional units are
disconnected from the antenna by means of switches, as illustrated in Fig. 2. Considering the signal power ranges at which
the three functional unit operate, we see that the main issue is to
decouple the TX from the RF-DC conversion block because the
power ranges overlap. In other words, the switches avoid that
the RF output power of the TX is converted back to DC by the
harvester. On the other hand, the power ranges of the RX and
RF-DC converter do not overlap. This characteristic allows to
merge the switches
and
and decouple the two functional units using the nonlinear input impedance of the RF-DC
converter, as will be detailed in Section III.
Another important aspect regarding the switches is that different supply voltage conditions apply for them. Activating the
TX- or RX-switch only makes sense when the supply voltage is

sufficiently high to operate TX or RX, respectively. On the other

hand, the switch for the harvester should be turned on even if no
supply voltage is available, i.e., if the energy storage of the wireless sensor is completely discharged.
In standard CMOS technologies the required functionality of
the TX- and RX-switches can be easily implemented by means
of enhancement-mode transistors [17]. The harvesting switch
, however, requires an active-low behavior as exhibited by
depletion-mode nMOS transistors which are able to form conducting channels even with negative gate-source voltages. As
depletion-mode devices are not available in standard CMOS
technologies [18], three different alternatives are discussed in
the following.
A. LC Resonator
One possibility to implement an active-low RF switch with
an inherently active-high device, such as an enhancement-mode
nMOS transistor, is by using resonators [19]. In the simple
model shown in Fig. 3(a) the antenna is represented by a

MASUCH et al.: CO-INTEGRATION OF AN RF ENERGY HARVESTER

1567

voltage source
and a real impedance
. With the digital
control voltage HV_disable being low, the transistor
is an
is connected in series to the
open switch and the load
antenna through
and
, which ideally act as a short-circuit
at resonance. When HV_disable is high, a parallel resonance
and
, providing an ideally infinite
circuit is formed by
input impedance
at resonance.
Let us assume the harvester is active and the switch is characterized by its insertion loss factor ILF, defined as the ratio of
the available power from the antenna to the power delivered to
. Then, assuming matched source and load impedthe load
, the required quality factor of the series
ances
resonance circuit
can be expressed as
(1)
where represents the resonance frequency. The de-activated
switch may be assessed by quantifying how much the power
transfer from the antenna to the RX, or equivalently from the
TX to the antenna, is impaired by the switch. Let us define the
switch-induced degradation factor DGF as the ratio of availwith its impedance
to the
able power from the source
. Then, assuming again matched condipower delivered to
tions
, the required quality factor of the parallel
resonance circuit
can be expressed as
(2)
If we further assume equal quality factors of the serial and parallel resonators, they may be expressed as a function of the
switch requirements.
(3)
As an example, for an insertion loss of 1 dB
and TX/RX-degradation of 0.5 dB
quality factor of the resonators would be 5.5.
B.

, the required

-Transmission Line

A second possibility to implement the active-low RF switch

with an nMOS transistor is based on a
-transmission line
as shown in Fig. 3(b). With HV_disable=0 the load impedance
is connected to the antenna through an ideally loss-less transequal to load
mission line with a characteristic impedance
. When HV_disand source impedance
grounds the output of the
able is pulled high, the transistor
transmission line which rotates this short-circuit one semi-circle
across the Smith chart to an open-circuit. Hence,
would
ideally be infinite but in practice depends on the on-resistance
of the transistor [20].
(4)

C. Start-Up Rectifier
A third option for the active-low RF switch is to use an nMOS
transistor as a pass device and generate the required positive
gate-source voltage by means of a start-up rectifier, as shown in
Fig. 3(c). The start-up rectifier can have very small dimensions
and does not require an outstanding efficiency, because it is
loaded only capacitively by the gate of the pass transistor
.
However, in this concept the input power has to exceed a certain
level given by the required turn-on voltage swing of the start-up
rectifier. In order to de-activate the RF switch, the input of the
rectifier is tied to ground by the transistor
and the antenna
is mainly loaded capacitively by the coupling capacitor
.
D. Topology Selection
For selecting the most suitable topology, the transceiver
front-end to be used for co-integrating the RF-DC conversion
has to be taken into account. In this paper, we will assume the
transceiver is characterized by a high internal RF impedance.
This is a conventional solution in low-power TRXs [21],
[22] which allows for an efficient power amplifier at low
output levels. Additionally, the passive voltage gain due to
the impedance up-transformation, not only helps to relax the
specifications of the front-end low-noise amplifier or even
suppress it altogether, but it is also expected to allow for an
efficient RF-DC conversion since the rectifier performance
usually improves for higher input voltage swings [23].
Assuming the RF impedance level is about 1 k , the LC resonator approach has to be excluded for a fully-integrated solution as the required inductance would be approximately 90 nH
according to (2) for
and
GHz. Similarly, the transmission line approach can be hardly
integrated in silicon because the
-transmission line would
be about 16 mm long [20]. For an integrated solution to be feasible, the TRX operating frequency would have to be at least one
order of magnitude larger, where RF energy harvesting becomes
less meaningful due to the much higher path loss (proportional
to ).
The solution with the start-up rectifier can be very area-efficient as no inductors or transmission lines are needed. However, it does not work for small signal levels because rectification relies on the non-linear large-signal characteristics of a
diode. Anyhow, this is still a valid option for the given application if we keep in mind that we want to harvest RF energy only
if the incoming power level is at least
dBm. This yields
an input amplitude of at least 400 mV (for
k ) which
is large enough to drive the start-up rectifier [2]. Therefore, the
start-up rectifier solution is well suited for co-integration with
TRX front-ends exhibiting high internal RF impedance.
For lower RF impedances, the start-up rectifier option is still
a working solution, but it requires higher incoming power levels
to harvest RF energy. For instance, if we assume
and
an amplitude of 200 mV to drive the rectifier, the required input
power raises to
dBm and, hence, shorter distances to the RF
energy emitter are needed for harvesting. Nonetheless, this approach still requires much lower chip area than the LC resonator
solution (it would require a 4.5 nH inductor under similar conditions) and, therefore, the impact on the overall floor-planning
of the TRX is smaller.

1568

IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 48, NO. 7, JULY 2013

Fig. 4. Front-end of the low-power transceiver with the new parts for energy harvesting highlighted with gray background and the additional OOK detector in a
dashed box.
TABLE I
DECOUPLING STRATEGY BETWEEN THE MIXER AND THE RECTIFIER

III. RF HARVESTER ARCHITECTURE

Fig. 4 shows the co-integration of a start-up rectifier based RF
energy harvester and a low-power 2.4 GHz transceiver front-end
previously reported in [24]. The non-highlighted parts depict the
relevant blocks of this front-end, which uses an on-chip transformer to up-convert the antenna impedance to about 1 k . Note
that the front-end already comprises the nMOS RF switches
and
to connect the power amplifier or the passive mixer to the antenna, respectively.
In the modified front-end, the already existing RX-switch
is re-used as the switch for the RF energy harvester.
This is possible because the mutual interaction between the
passive mixer and the main rectifier, which performs the actual
RF-to-DC conversion, can be kept low. A key aspect toward
this goal is that the rectifier is an intrinsically nonlinear block
whose parallel input resistance
strongly depends on
the input power. Accordingly, the rectifier is designed such
that it shows high input resistance at low input power level
(
dBm) when the RX may be active and the mixer
is matched to the transformer output at about 1 k . Contrarily,
the rectifier exhibits a matched resistance
when
is around
dBm and the RX is disabled. Note
that the reactive parts
and
are less significant in
these considerations because they do not dissipate power and
can be tuned out by the transformer. Table I summarizes both
operating conditions and shows that only one device is matched
while the other presents high impedance. This approach also
guarantees that the noise figure and sensitivity of the RX, both
measured at low signal levels, are hardly affected by the RF-DC
converter.
In order to activate the RX-switch for energy harvesting
without any external supply voltage, the start-up rectifier is directly connected to the internal transformer port
. Because
of the small load that has to be charged by the start-up rectifier,

it can be designed with much smaller dimensions than the main

rectifier. However, in order to prevent the start-up rectifier from
sinking the output power of the PA, its operation is controlled
by an external signal, TX enable, in such a way that when the
PA is active, signal rectification is disabled and vice versa.
Note that the start-up rectifier comprises 5 cascaded stages to
guarantee that the RX-switch is turned-on whenever there is
significant RF input power to harvest, i.e.,
dBm.
By only adding an extra start-up rectifier stage to the
front-end, the harvester can also implement an On-Off-Keying
(OOK) detector. Hence, this additional feature comes with
almost no implementation cost and offers an additional receive-channel aside from the main TRX. This second channel
can be used, for example, to send a wake-up signal to the sensor
node. Such wake-up channels are frequently implemented in
wireless sensor nodes to save battery power and keep the main
TRX completely disabled until a wake-up signal is received
[14], [25][28]. Here, OOK-detection is achieved by comparing the output
of the additional rectifier stage to the
output
of the first stage in the start-up rectifier chain, as
shown in Fig. 5. As none of these rectifier stages is loaded by
a DC current and have their DC-inputs to ground, they provide
the same DC output voltage and only differ in their transient
behavior. As will be detailed in Section IV.C, the time constant
of
is much larger than the one of
, so that, whereas
follows the OOK-envelope,
tracks its average. By
exploiting this fact, the OOK detection is simply performed by
a comparator.
Finally, the supply management circuit (SMC) transfers the
DC output power of the main rectifier to the energy storage
through the node
. An internal regulator limits this voltage
to a maximum of 1.34 V to prevent oxide breakdown in the
TRX. For a fast harvester start-up, the SMC is supplied with
a small standby current
through the node
to avoid
the settling transients of the internal references, provided the energy storage is not empty. Nevertheless, with an empty energy
storage, the harvester is still able to charge the load because
is much smaller than
under normal conditions. Also for
speeding up the start-up, the SMC provides an internal supply
to pre-charge the RX-switch. The transistors

MASUCH et al.: CO-INTEGRATION OF AN RF ENERGY HARVESTER

1569

Fig. 5. Concept of the implemented OOK detection.

Fig. 7. Simulated parallel input resistance

of the main rectifier.

Fig. 6. Schematic of one stage of the main rectifier (gate dimensions in m).

and
ensure the supply currents to flow in the desired directions only, indicated by arrows in Fig. 4. In this prototype,
the nodes
and
, which could be simply tied together
to the load, use separate pins to allow for more flexibility and
testability.
IV. CIRCUIT DESIGN
A. Main Rectifier
The main rectifier is implemented as a cascade of 4 AC-coupled rectifier cells, each using the conventional 4-transistor configuration of Fig. 6 [23], [29], [30]. Note, that each stage sits in
its own p-well, connected to
, which is isolated by a deep
n-well tied to the output voltage
. For input amplitudes
below the threshold voltages, no current is flowing through the
cell transistors, thus leading to a high parallel input resistance,
. As the input voltage increases, a periodic current starts
flowing through the cell transistors and the input resistance of
the structure diminishes. Fig. 7 shows
of the complete
rectifier as a function of the available input power from a 1 k
source, which closely follows the desired behavior expressed in
Table I. The main rectifier has been designed to match the 1 k
RF impedance at
dBm input power by choosing the transistor widths accordingly. To match the rectifier at lower input
power levels, larger widths would be needed to reduce the input
impedance.
B. Start-Up Rectifier
For the start-up rectifier, completely different requirements
apply. Since it is only capacitively loaded by the gates of the

Fig. 8. Schematic of one stage of the start-up rectifier stage (gate dimensions
in m).

RX-switch, its efficiency is less important. It is much more critical to ensure that the start-up rectifier does not sink power from
the PA when the TX is active. Accordingly, the 4-transistor rectifier cell is modified with an additional switch
to disable the
start-up rectifier during transmission, as shown in Fig. 8. Pulling
disable high effectively shorts the two internal nodes and creates
a virtual ground. In this mode, the RF input impedance of the
modified rectifier cell is contributed by the series connection of
the two coupling capacitors
and the on-resistance of transistor
. In order to reduce power dissipation in TX-mode,
is made large enough so that the resistive part of the input
impedance is small. On the other hand, when disable is low,
increases the parasitic load on the internal nodes of the rectifier leading to reduced efficiency. Accordingly, this concept is
only suitable for the start-up rectifier but not for the main rectifier. Apart from
, the remaining devices are close to their
respective minimum size in order to keep the capacitive loading
of the RF nodes as low as possible. As for the main rectifier, each
start-up rectifier cell sits in a p-well connected to
which
is isolated by a deep n-well.
C. OOK-Detector
The OOK-detector employs the same rectifying stage as in
Fig. 8. The design principle is based on the comparison between
provided by this stage and the avthe envelope voltage
erage voltage
obtained from the first stage of the start-up
rectifier. As mentioned before, both voltages are at the same DC
level and only differ in their transient response. To analyze the

1570

IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 48, NO. 7, JULY 2013

Fig. 9. Schematic of the asymmetric OOK-comparator (gate dimensions in

m).

Fig. 10. Schematic of the supply management (gate dimension in m).

time constants, let us consider one rectifier stage as a pulsed

voltage source with some nonlinear source resistance
.
Then, the time constant on
is simply given by
(5)
where
is the input capacitance of the OOK comparator.
On the other side, the start-up rectifier comprises 5 identical
stages connected in series. Each stage can be modeled by the
same pulsed source and
as before. This is possible because each stage sits in its own isolated p-well and so all stages
operate at the same condition. Accordingly, the output voltage
of the chain
is approximately 5 times
, and both signals have approximately the same time constant
(6)
where the capacitance of the OOK comparator has been neglected. So, as long as
is smaller than
pF, the
time constant of
is much larger than the one of the envelope tracking signal
.
To detect notches in the RF envelope, the comparator shown
in Fig. 9 is implemented. It outputs a logical 1 for
, where
is the offset voltage of the comparator, and logical 0 otherwise. In order to prevent the detection
of false notches when
, the comparator employs
an asymmetric pMOS input pair with different multiplicity-factors to guarantee a positive offset voltage
, also taking into
account device mismatch. The comparator is supplied with the
internal supply voltage
and biased with a reference current of 30 nA, which are both obtained from the supply management circuit.
D. Supply Management Circuit (SMC)
Fig. 10 shows the schematic of the SMC. At its input, a regulator limits the rectified signal
to a DC voltage of 1.34 V,
which is three times the value of the internal reference voltage
. The output stage of the SMC contains a pMOS pass transistor
which is on for
, where
is the offset of the controlling amplifier
. Similarly to the
OOK comparator, this amplifier is designed such that its offset
can only take negative values in order to prevent a reverse current through the pass device. When the control signal
is high
(constant current mode), the pass device is switched off and the
current mirror
outputs a constant current of about 70 A.

Fig. 11. Microphotograph of the assembled TRX with energy harvester,

showing the OOK, start-up and main rectifier in the inset. The total die size is
1.6 mm by 1.3 mm.

The switch
prevents current reversing from
to
through the source-bulk diode of
.
The internal references are obtained from a CMOS-only reference generator without resistors [31] which only consumes
25 nA and starts operating for a supply voltage of 0.7 V. It is
supplied by the internal supply voltage
which is obtained either from
or
, depending on which is larger.
The selection is accomplished by the comparator CMP and the
switches
. The blocks connected to
also define the
standby current
drawn from the load, which add up to
about 60 nA. The typical start-up time for the reference generator and regulator upon an input voltage step on
is always less than 2 ms for
V. In pulsed operation with
pulse widths smaller than 2 ms, the gate of the pass transistor
is kept low by a 200 fF-capacitor, and nodes
and
can be still charged up. In this case,
is clamped at
one pMOS threshold voltage above ground.
V. EXPERIMENTAL RESULTS
The energy harvester, including the supply management circuit, has been integrated together with the 2.4 GHz low-power
TRX front-end [24] in a 130 nm standard CMOS technology (Fig. 11). The area occupied by the energy harvester
is 0.019 mm , which breaks down to 2100 m , 4400 m and
12500 m for the start-up rectifier, the main rectifier and the
supply management circuit, respectively. This amounts about

MASUCH et al.: CO-INTEGRATION OF AN RF ENERGY HARVESTER

1571

Fig. 13. Charging scenario with 10 F load capacitance connected to

and
(bottom curve) with two different RF pulses sequences (
dBm,
GHz): (a) with a pulsewidth of 0.2 ms and a period of 2 ms, (b)
with a pulsewidth of 5.6 ms and a period of 100 ms.

Fig. 12. Measured performance of the energy harvester: (a) open circuit
in constant current mode
output voltage (10 M load), (b) output current
, (c) power conversion efficiency PCE without current limitation
for different load conditions (
tied to
).

2% of the total active area of the TRX. The chip has been assembled in a QFN-36 package and soldered onto a test board. In
order to perform RF measurements, a surface-mount balun with
2:1 impedance ratio [32] has been used to convert the differential
100 antenna interface of the chip into the single-ended 50
standard impedance of the laboratory equipment.
Fig. 12(a) shows the experimental open-load output voltage
of the harvester for different supply voltage conditions and
regulator settings. Note that with an external supply present
(
V), the RX-switch is always turned-on and the
harvester generates higher output voltages at low input power
levels (
dBm). If
V, the start-up rectifier
turns on the RX-switch for
dBm. For input power
levels above
dBm, the output voltage becomes limited

to 1.34 V when the regulator is active. Otherwise the output

voltage is soft-limited by the ESD diodes to about 1.9 V.
Fig. 12(b) shows the output current of the harvester in constant current mode
for different load voltages. For
dBm, the harvester outputs a constant current of
about 65 A. Note that, for high load voltages, the current mirror
leaves the saturation region and the output current is reduced.
Finally, Fig. 12(c) shows the measured PCE for different load
voltages without current regulation
. The peak efficiencies are obtained at about 0 dBm with a maximum of 15.9%
for a load voltage of 1.2 V. It must be emphasized, the proposed
harvester is matched to a real antenna impedance by a transformer and it requires a switch in the RF path for decoupling.
These two elements entail losses of about 2 dB and 1.5 dB, respectively, before the RF power reaches the rectifier. This makes
the PCE obtained by isolated rectifiers [30] hardly achievable in
co-integration scenarios. Nonetheless, according to post-layout
simulations of the main rectifier alone, the peak PCE is approximately 50% at
dBm input power and, therefore, comparable
to standalone converters. Some slight improvements on the PCE
performance of the co-integrated harvester may be possible by
allowing a larger DC output voltage of the rectifier, though.
Fig. 13 shows the charging behavior of the harvester with a
load capacitor of 10 F connected to both
and
.

1572

Fig. 14. Measured 10%90% rise time for charging a 10 F storage capacitor
to the final voltage of 1.34 V.

IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 48, NO. 7, JULY 2013

Fig. 16. Return loss

in RX-mode.

Fig. 17. Measured noise figure (NF) and output power

mented TRX front-end with and without energy harvester.

Fig. 15. Operation of the OOK-demodulator without external supply and at an

dBm: at the top, the modulation used in the transmitted
input RF power of
and
connected
OOK-signal; below, the self-generated supply with
(using the common load capacitance of 10 F) and, at the bottom, the detected
OOK-notches ook detect.

Although this load capacitance is too small to serve as a real

energy storage, it still allows for verification of the dynamic
harvesting performance when pulsed RF energy is received as,
for instance, provided by a WLAN router. In Fig. 13 the results
for two different RF pulse widths are shown, representing the
minimum and maximum WLAN packet lengths of 0.2 ms and
5.6 ms, respectively. In both scenarios,
starts from 0 V,
i.e., a completely discharged energy storage. These measurements were also carried out at different input power levels.
Fig. 14 shows that for both pulse sequences, the harvester is
able to progressively charge the 10 F load until the regulator
limits the voltage to 1.34 V, for input power levels of at least
dBm.
Fig. 15 shows the operation of the OOK-detector without
external supply voltage. In this example, bursts of 10 notches
with 25 s width are transmitted and the RF input power is
dBm, which is the lowest power level that allows for correct notch detection. Note that due to the OOK-modulation, the
rectified supply voltage
drops to about 630 mV,
which eventually limits the notch detection. This is because the

of the imple-

comparator stops working correctly for lower supply voltages.

The detector has been tested successfully for notch widths from
5 s to 80 s, which allows for data rates of up to 100 kb/s using
pulse-interval encoding. With an external supply available, i.e.,
an energy storage which has been charged previously to 1.0 V,
the minimum RF input level for correct OOK-detection drops
to
dBm.
Fig. 16 shows the input matching of the transceiver, both in
energy harvesting mode and in RX-mode, measured through
the balun with a network analyzer. Note that in both modes the
transceiver is well matched in the 2.4 GHz ISM band, i.e., the
return loss is below
dB and
dB in harvest-mode and
RX-mode, respectively.
Finally, the impact of the harvester on the TRX performance
is measured by comparing two TRX front-ends, which only
differ on the presence or not of the proposed energy harvester.
The two TRX front-ends have been fabricated in different MPW
runs. Fig. 17 shows the measured noise figure (NF) and output
power
for both chips. Due to process variations between
the MPW runs, the TRX with harvester achieves its best performance at a slightly higher frequency. For the same reason, the
sample with harvester also shows a slightly higher peak output
power although the presence of the disabled start-up rectifier
theoretically decreases the output power due to its not purely
capacitive input impedance. Regarding NF, the TRX is only degraded by about 0.2 dB with the harvester co-integrated. These
results are in good agreement with post-layout simulations of
the two TRX front-ends using the same technology parameters that predict a degradation of 0.4 dB and 0.2 dB for NF

MASUCH et al.: CO-INTEGRATION OF AN RF ENERGY HARVESTER

1573

TABLE II
COMPARISON WITH RECENT RF ENERGY HARVESTER CO-INTEGRATED WITH TRX

and
, respectively. In this case, post-layout simulation results are more meaningful because process parameters are equal
in simulation but not in the measurements due to the different
MPW runs.
No impact on the linearity of the RX front-end due to the
presence of the harvester is observed. With and without harvester, the linearity is limited by the first baseband amplifier
at the output of the passive mixer, which provides a 1 dB
compression point of about
dBm and a 3rd order intercept point of about
dBm, both referred to the antenna
input. At the power levels where these points are measured
(
dBm), the input impedance of the main rectifier can
be still regarded linear without a significant impact on the
overall RX linearity.
Table II summarizes the performance of the energy harvester
and compares it to recently published RF energy harvesters that
share a common antenna with a transceiver or receiver. Note that
the harvester of Ishizaki et al. [33] requires a supply voltage for
harvesting in order to tune the input matching network to the RF
signal to be harvested. Concerning the efficiency, the proposed
harvester achieves similar levels as the two other solutions. Also
in terms of the degradation of the transceiver and the sensitivity
of the wake-up channel similar results are obtained compared
to Lerdsitsomboon, et al. [14]. However, the main advantage of
the proposed solution is that it occupies a much lower die area
and does not need any bond wire inductor.
VI. CONCLUSION
An energy harvester has been presented that can be easily
integrated into a low-power transceiver front-end, where both
the harvester and the TRX operate at the same frequency of
2.4 GHz. Decoupling between the TX and the harvester is
achieved by an RF-switch which is turned on in passive state
by using a start-up rectifier. The decoupling between the RX
and the harvester exploits the different input impedances of
these blocks. Further, the harvester can be easily extended to
implement a secondary wake-up receiver which can be used to
asynchronously turn on the transceiver on demand. The harvester achieves a measured peak PCE of 15.9% and degrades
the TX- and RX-performance by less than 0.5 dB. Since the
required additional area for the harvester is only 0.019 mm ,
which amounts about 2% of the total active TRX area, it can be
added to the TRX at almost no cost.

REFERENCES
[1] Y. Manoli, Energy harvestingFrom devices to systems, in Proc.
Eur. Solid-State Circuits Conf., 2010, pp. 2736.
[2] T. Le, K. Mayaram, and T. Fiez, Efficient far-field radio frequency
energy harvesting for passively powered sensor networks, IEEE J.
Solid-State Circuits, vol. 43, no. 5, pp. 12871302, 2008.
[3] P. Li and R. Bashirullah, A wireless power interface for rechargeable
battery operated medical implants, IEEE Trans. Circuits Syst. II, Exp.
Briefs, vol. 54, no. 10, pp. 912916, 2007.
[4] N. J. Dudney, Solid-state thin-film rechargeable batteries, Mater. Sci.
Eng. B, vol. 116, no. 3, pp. 245249, 2005.
[5] Cymbet Corporation, EnerChip CBC050 Datasheet Rev B, Rechargeable Solid State Energy Storage. 2010.
[6] GW109/GW209 Supercapacitor Datasheet Rev 1.0. Cap-XX Pty. Ltd.,
Jun. 2011.
[7] S. H. Lin, K. C. Chiang, A. Chin, and F. S. Yeh, High-density and lowleakage-current MIM capacitor using stacked TiO /ZrO insulators,
IEEE Electron Device Lett., vol. 30, no. 7, pp. 715717, 2009.
[8] D. Costinett, E. Falkenstein, R. Zane, and Z. Popovic, RF-powered
variable duty cycle wireless sensor, in Proc. Eur. Microw. Conf., 2010,
pp. 4144.
[9] Printed Thin Film Battery, Fraunhofer ENAS, 2010.
[10] H. J. Visser, A. C. F. Reniers, and J. A. C. Theeuwes, Ambient RF
energy scavenging: GSM and WLAN power density measurements,
in Proc. Eur. Microw. Conf., 2008, pp. 721724.
[11] R. J. M. Vullers, R. V. Schaijk, H. J. Visser, J. Penders, and C. V. Hoof,
Energy harvesting for autonomous wireless sensor networks, IEEE
Solid-State Circuits Mag., vol. 2, no. 2, pp. 2938, 2010.
[12] G. Haizheng and R. Sobot, RF power harvesting analog front-end circuit for implants, in Proc. Can. Conf. Electr. Comp. Eng., 2009, pp.
907910.
[13] G. Papotto, F. Carrara, A. Finocchiaro, and G. Palmisano, A 90
nm CMOS 5 Mb/s crystal-less RF transceiver for RF-powered WSN
nodes, in IEEE Int. Solid-State Circuits Conf. Dig. Tech. Papers,
2012, pp. 452454.
[14] W. Lerdsitsomboon and K. O. Kenneth, Technique for integration of
a wireless switch in a 2.4 GHz single chip radio, IEEE J. Solid-State
Circuits, vol. 46, no. 2, pp. 368377, 2011.
[15] J. Masuch, D. Milosevic, M. Delgado-Restituto, and P. Baltus, An
RF-to-DC energy harvester for co-integration in a low-power 2.4 GHz
transceiver frontend, in Proc. IEEE Int. Symp. Circuits Syst., 2012,
pp. 680683.
[16] J. Masuch, M. Delgado-Restituto, D. Milosevic, and P. Baltus, An RF
energy harvester with supply management for co-integration into a 2.4
GHz transceiver, in Proc. Eur. Solid-State Circuits Conf., 2012, pp.
113116.
[17] Y. P. Zhang, L. Qiang, F. Wei, A. C. Hoe, and L. He, A differential
CMOS T/R switch for multistandard applications, IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 53, no. 8, pp. 782786, 2006.
[18] T. Ohnakado, S. Yamakawa, T. Murakami, A. Furukawa, E. Taniguchi,
H. Ueda, N. Suematsu, and T. Oomori, 21.5-dBm power-handling
5-GHz transmit/receive CMOS switch realized by voltage division effect of stacked transistor configuration with depletion-layer-extended
transistors (DETs), IEEE J. Solid-State Circuits, vol. 39, no. 4, pp.
577584, 2004.
[19] J.-H. Wang, H.-H. Hsieh, and L.-H. Lu, A 5.2-GHz CMOS T/R switch
for ultra-low-voltage operations, IEEE Trans. Microw. Theory Tech.,
vol. 56, no. 8, pp. 17741782, 2008.

1574

[20] A. Shirvani, D. K. Su, and B. A. Wooley, A CMOS RF power amplifier with parallel amplification for efficient power control, IEEE J.
Solid-State Circuits, vol. 37, no. 6, pp. 684693, 2002.
[21] I. Bhatti, R. Roufoogaran, and J. Castaneda, A fully integrated transformer-based front-end architecture for wireless transceivers, in IEEE
Int. Solid-State Circuits Conf. Dig. Tech. Papers, 2005, vol. 1, pp.
106587.
[22] B. W. Cook, A. Berny, A. Molnar, S. Lanzisera, and K. S. J. Pister,
Low-power 2.4-GHz transceiver with passive RX front-end and
400-mV supply, IEEE J. Solid-State Circuits, vol. 41, no. 12, pp.
27572766, 2006.
[23] S. Mandal and R. Sarpeshkar, Low-power CMOS rectifier design for
RFID applications, IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 54,
no. 6, pp. 11771188, 2007.
[24] J. Masuch, M. Delgado-Restituto, and A. Rodriguez-Vazquez, Transformer based front-end for a low power 2.4 GHz transceiver, in Proc.
IEEE Asia-Pacific Conf. Circuits Syst., 2010, pp. 288291.
[25] N. M. Pletcher, S. Gambini, and J. Rabaey, A 52 W wake-up receiver
dBm sensitivity using an uncertain-IF architecture, IEEE J.
with
Solid-State Circuits, vol. 44, no. 1, pp. 269280, 2009.
[26] H. Xiongchuan, S. Rampu, W. Xiaoyan, G. Dolmans, and H. de Groot,
A 2.4 GHz/915 MHz 51 W wake-up receiver with offset and noise
suppression, in Proc. IEEE Int. Solid-State Circuits Conf. Dig. Tech.
Papers, 2010, pp. 222223.
[27] C. Hambeck, S. Mahlknecht, and T. Herndl, A 2.4 W wake-up redBm sensitivity, in Proc.
ceiver for wireless sensor nodes with
IEEE Int. Symp. Circuits Syst., 2011, pp. 534537.
[28] Y.-I. Kwon, J. Byun, H.-J. Jo, and H.-Y. Lee, An ultra low power
CMOS transceiver having a short range wake-up for lR-WPAN applications, in Proc. Eur. Microw. Integr. Circuits Conf., 2009, pp. 262265.
[29] A. Facen and A. Boni, Power supply generation in CMOS passive
UHF RFID tags, in Proc. Conf. Res. Microelectron. Electron., 2006,
pp. 3336.
[30] K. Kotani, A. Sasaki, and T. Ito, High-efficiency differential-drive
CMOS rectifier for UHF RFIDs, IEEE J. Solid-State Circuits, vol.
44, no. 11, pp. 30113018, 2009.
[31] H. J. Oguey and D. Aebischer, CMOS current reference without resistance, IEEE J. Solid-State Circuits, vol. 32, no. 7, pp. 11321135, 1997.
[32] Chip-Balun WE-BAL 748421245 Specification for Release, Oct. 2004,
Wrth Elektronik.
[33] H. Ishizaki and M. Mizuno, A 0.2 mm , 27 Mbps 3 mW ADC/FFTless FDM BAN receiver with energy exploitation capability, IEEE J.
Solid-State Circuits, vol. 45, no. 4, pp. 921927, 2010.

Jens Masuch (S09-M12) was born in Rostock,

Germany, in 1978. He received the B.S. degree
(high honors) in electrical engineering from the Milwaukee School of Engineering, Milwaukee, WI, in
2002, and the M.S. degree in information technology
from the Mannheim University of Applied Sciences,
Mannheim, Germany, in 2004. In 2012, he received
the Ph.D. degree in physics-microelectronics from
the University of Seville, Seville, Spain.
From 2004 to 2008, he was with Atmel Germany
GmbH, Heilbronn, Germany, where he worked on
several analog building blocks for UHF and HF RFID transponder ICs. From
2008 to 2012 he worked at the Institute of Microelectronics of Seville, Seville,
Spain, on ultra-low power wireless transceivers. Since 2012, he is with Dialog Semiconductor, Munich, Germany, where he works on power management
ICs. His research interests are low-power wireless transceivers, RF energy harvesting and power management.

IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 48, NO. 7, JULY 2013

Manuel Delgado-Restituto (M96SM12) received the M.S. and Ph.D. degrees (honors) in
physics-electronics from the University of Seville,
Seville, Spain, in 1990 and 1996, respectively.
He is currently a Senior Researcher with the
Institute of Microelectronics of Seville, Centro
Nacional de Microelectronica, Consejo Superior de
Investigaciones Cientificas, Seville, where he heads
a research group on low-power medical microelectronics. He has authored/edited three books, around
20 chapters in contributed books, and some 150
articles in peer-reviewed specialized publications. His research interests are in
the design of silicon microsystems to understand biological neural systems,
the development of neural prostheses and brain-machine interfaces, and the
implementation of wireless BAN transceivers and RFID systems.
Dr. Delgado-Restituto has served or is serving as an Associate Editor and
a Guest Editor for different IEEE and non-IEEE journals (including the IEEE
TRANSACTIONS ON CIRCUITS AND SYSTEMS I AND II). He is in the committee
of different international conferences, is a member of the IEEE Biomedical Circuits and Systems Technical Committee, and has served as the Technical Program Chair of different international IEEE conferences. He is the deputy Editor-in-Chief of the IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN
CIRCUITS AND SYSTEMS (in 20112012).

Dusan Milosevic was born in Nis, Serbia, in 1974.

He received the M.S. degree in electronics and
telecommunications engineering from the University
of Nis, Serbia, in 2001, and the Ph.D. degree in
electrical engineering from Eindhoven University of
Technology (TUE), The Netherlands, in 2009.
Since 2001 he has been with TUE, first as a
Ph.D. student and later on as a research scientist.
During his Ph.D. research, he was focusing on
switched-mode RF power amplifiers and techniques
for high-efficiency linear amplification. Since 2010,
he has been an Assistant Professor in the Mixed-signal Microelectronics group
at TUE. His research interests include RF and microwave power amplifiers and
ultra-low power RF front ends.

Peter G. M. Baltus (M08SM11) was born in

Sittard, The Netherlands, on July 5th, 1960. He
received the M.S. degree in electrical engineering
and the Ph.D. degree from Eindhoven University of
Technology, Eindhoven, The Netherlands, in 1985
and 2004, respectively.
He worked for 22 years at Philips and later NXP
in Eindhoven, Nijmegen, Tokyo, and Sunnyvale in
various functions, including Research Scientist, Program Manager, Architect, Domain Manager, Group
Leader and Fellow. In 2007, he started his current job
at the Eindhoven University of Technology as a Professor in high-frequency
electronics and Director of the Centre for Wireless Technology, Eindhoven. He
holds 16 patents and coauthored more than 70 papers.