IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 12, NO.

2, APRIL 2016 633

Design, Implementation, and Performance

Evaluation of a Flexible Low-Latency
Nanowatt Wake-Up Radio Receiver
Michele Magno, Member, IEEE, Vana Jelicic, Student Member, IEEE,
Bruno Srbinovski, Student Member, IEEE, Vedran Bilas, Senior Member, IEEE,
Emanuel Popovici, Senior Member, IEEE, and Luca Benini, Fellow, IEEE

Abstract—Wireless sensor networks (WSNs) have I. I NTRODUCTION

received significant attention in recent years and have
found a wide range of applications, including structural
and environmental monitoring, mobile health, home
automation, Internet of Things, and others. As these
W IRELESS sensor networks (WSNs) have been recog-
nized as an enabling technology for a large variety
of applications, including smart homes and cities, agriculture,
systems are generally battery operated, major research transportation, health and fitness, entertainment, and structural
efforts focus on reducing power consumption, especially
for communication, as the radio transceiver is one of the
health monitoring [1]. Strict energy constraints of battery-
most power-hungry components of a WSN. Moreover, with powered wireless sensor nodes have introduced the necessity
the advent of energy-neutral systems, the emphasis has of energy awareness in both software and hardware solutions.
shifted toward research in microwatt (or even nanowatt) Reducing the communication power consumption of WSN
communication protocols or systems. A significant number nodes is important, as the radio transceiver is one of the com-
of wake-up radio receiver (WUR) architectures have been
proposed to reduce the communication power of WSN
ponents with the highest power consumption. Optimizing the
nodes. In this work, we present an optimized ultra-low power consumption of the wireless transceiver can free up
power (nanowatt) wake-up receiver for use in WSNs, power budget to add much more functionality. In addition, since
designed with low-cost off-the-shelf components. The the battery size is decisive in determining the size of battery-
wake-up receiver achieves power consumption of 152 nW operated systems, low-power circuits could enable smaller
(with −32 dBm sensitivity), sensitivity up to −55 dBm (with
maximum power of 1,2 µW), latency from 8 µs, tunable
batteries and lead to miniaturization required by many appli-
frequency, and short commands communication. In addi- cations such as wearable WSNs, medical body area networks
tion, a low power solution, which includes addressing and implantable devices.
capability directly in the wake-up receiver, is proposed. To establish communication, two radios (the receiver and
Experimental results and simulations demonstrate low the transmitter) need to be synchronized as a message can
power consumption, functionality, and benefits of the
design optimization compared with other solutions, as well
be received only if the radio is in its listening state, and
as the benefits of addressing false positive (FP) outcomes idle listening consumes significant power. Hence, a signifi-
reduction. cant design effort is required to alleviate this power waste.
To reduce communication power consumption, several tech-
Index Terms—Nanowatt wake-up radio receiver (WUR),
power optimization, ultra-low power, wireless sensor net- niques have been proposed [2]–[7] for lowering or eliminating
works (WSNs). the power wasted due to idle listening of the transceiver.
Duty cycling is a common technique to reduce the idle mode
Manuscript received August 04, 2014; revised October 09, 2015;
accepted December 16, 2015. Date of publication February 05, 2016;
energy consumption which consists of switching from listen-
date of current version March 29, 2016. This work was supported ing mode to sleep mode [2]. However, while duty cycling
by the project Scientific co-operation between Eastern Europe and helps save power, it can severely limit the network reactiv-
Switzerland (SCOPES) and SWISS NATIONAL FUNDATION (SNF)
under Grant IZ74Z0_160481, Science Foundation Ireland (SFI) Marei
ity as the radios are OFF (or in the sleep state) and they
Centre and Analog Devices at University College Cork (UCC). Paper cannot receive messages. Duty cycling is in general a syn-
no. TII-15-0772. chronous technique, where the radio is woken up for a fixed
M. Magno and L. Benini are with the Department of
Electrical, Electronic, and Information Engineering, University of
or adaptive time interval to listen if there is any relevant
Bologna, Bologna 40126, Italy, and also with the Swiss Federal incoming message. However, there are three types of communi-
Institute of Technology Zurich, Zurich 8092, Switzerland (e-mail: cation techniques: synchronous, pseudo-asynchronous, or pure
michele.magno@iis.ee.ethz.ch; luca.benini@unibo.it).
V. Jelicic and V. Bilas are with the Faculty of Electrical Engineering and
asynchronous [3].
Computing, University of Zagreb, Zagreb HR-10000, Croatia (e-mail: From a power consumption perspective, asynchronous
vana.jelicic@fer.hr; vedran.bilas@fer.hr). schemes are considered by far the most efficient, and the most
B. Srbinovski and E. Popovici are with the Department of Electrical
and Electronics Engineering, University College Cork, Cork, Ireland
effective realization of an asynchronous communication which
(e-mail: e.popovici@ucc.ie). is achieved by reducing or eliminating idle listening using a
Color versions of one or more of the figures in this paper are available wake-up radio receiver (WUR) [7]. Such a device is coupled
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TII.2016.2524982
with the main radio transceiver having the role of listening

1551-3203 © 2016 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.
634 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 12, NO. 2, APRIL 2016

continuously to the transmission medium and waking up the receivers only detect activity in the communication channel
main transceiver upon detection of an incoming message. and cannot distinguish a wake-up signal from other RF activ-
There are numerous features that a wake-up radio device ity. Moreover, it is impossible to receive commands or data to
must support to become effective. First of all, the power con- address a node. Another limit of these circuits is a short commu-
sumption of the WUR has to be orders of magnitude lower nication range compared to most WSN applications. Usually,
than that of the main transceiver in the receiving mode. Other they have a sensitivity around −25 dBm which implies the
important features are greater sensitivity, robustness to inter- use of higher transmission power or (where possible) a bigger
ference, selectivity, and low latency. Typically, sensitivity (the antenna to extend the range to tens of meters.
weakest signal the receiver is able to sense) is the most impor- Due to these limitations, they are more suitable for
tant optimization goal with the ultimate aim being to fix the short-range applications, which do not need any address-
sensitivity of the main transceiver at a much lower power. ing mechanism. For example, they can be used in some
The sensitivity is directly related to the communication range: implantable chips, body area networks, near field communi-
the greater the sensitivity, the longer the range. However, cation, and Radio-Frequency identification (RFIC). Despite
improving the receiver sensitivity is typically constrained with that, some of the interesting zero-power building blocks are
increased power consumption. present in a number of passive WUR architectures, such as
In this paper, an optimized architecture for an ultra-low passive rectifiers with interrupt block, with many designs being
power, low-cost wake-up receiver is presented, considering all presented in the literature in recent years [5]. Although the
the constraints and specifications described above. In more zero-power consumption feature is very appealing, we do not
detail, the contributions of this paper are as follows. implement, it in our proposed solution, as we are focused
1) A set of optimization techniques targeting ultra-low on solutions for a long range with sensitivity from −30 to
power wake-up receivers. −55 dBm that would give benefit to a wider range of WSN
2) The design and implementation of a wake-up receiver applications. For this reason, the proposed approach uses a
which has only one ultra-low power comparator as an semi-active receiver with a single-stage rectifier instead of a
active component, and an optional ultra-low power micro- multistage charge pump used in the RFID solutions, as the
controller for addressing capability. The wake-up receiver main goal is maximizing the communication range. In fact,
is built using a minimal number of low-cost, off-the-shelf passive RFID systems are achieving only a few centimeters of
components. range, while our solution can achieve several meters.
3) Experimental validation of the proposed approach, in
terms of power consumption, sensitivity, range, data rate,
functionality, and addressing capability when connected B. Semi-Active Wake-Up Circuits
to a microcontroller. The majority of the proposed design approaches are semi-
4) Comparison with existing wake-up receivers and a active, whereas the minority of the receiver’s components are
media access control (MAC) protocols (with and with- battery powered. The most common approach is based on
out addressing) to evaluate the benefits of the proposed an envelope detector, implemented with passive components
solution. (Schottky diodes, MOSFETs, or ad-hoc Integrated Circuits
This paper is organized as follows. Section II reviews the (ICs) for the radio front end) followed by an active component
related work. In Section III, the architectural considerations are (a comparator) to generate an interrupt. The research in [8]–
discussed, while Section IV presents the submicrowatt wake- [10] presents such architectures for ultra-low power WURs for
up receiver architecture proposed in this work. Finally, Sections WSN devices achieving a reduction of sensor node listening
V and VI present experimental setup and experimental results, activities and drastically reducing the overall network power
respectively, with three different implementations. Section VII consumption. The solutions use a multistage rectifier rather
concludes this paper. than a single-stage without elaborating this choice or methods
for power and sensitivity measurement.
We use a single-stage rectifier combined with an ultra-low
II. R ELATED W ORK offset comparator. In our solution, this is a significant design
Research on wake-up radio use in WSN to reduce power con- difference with respect to other solutions as a combination of a
sumption has been prolific in recent years. With a variety of single-stage rectifier and a comparator with an optimized offset
proposed methods and techniques, three groups of radio fre- voltage is able to achieve the minimal sensitivity possible with
quency (RF) wake-up systems can be identified: fully passive a passive front end (−55 dBm). Moreover, we optimized the
circuits, semi-active circuits, and fully active circuits. reference voltage source of the comparator eliminating the high
delay due to long preamble needed in other solutions [8]–[10].
In [11], a very interesting solution with a 98-nW power con-
A. Fully Passive Wake-Up Circuits
sumption is presented. This solution is similar to the solutions
Wake-up systems with fully passive circuits work with- presented above (using a rectifier and a comparator), but the
out any power supply as the circuit harvests energy from the custom CMOS rectifier was designed to achieve sensitivity of
radio communication and uses it to generate an interrupt. They −41 dBm. This sensitivity is lower (allowing a shorter range)
are mostly realized using charge pump, Schottky diodes, and than the one achieved in our approach. Also, we present an off-
CMOS or MOSFET technology. However, the fully passive the-shelf solution which can reduce cost and implementation
MAGNO et al.: DESIGN, IMPLEMENTATION, AND PERFORMANCE EVALUATION 635

time. A previous work published in [12] a similar architecture.

In this work, more details of the different design parameters are
presented, and a more complete set of evaluations with simu-
lation and in-field measurements has been performed. Finally,
the benefits of the addressing capability have been shown in the
experimental results.

C. Fully Active Wake-Up Circuits

This group of circuits use active components both for the
Fig. 1. Generic block diagram of a wireless node with a separate WUR.
rectifier and the interrupt generator. In [13], a solution with a
rectifier, high-band baseband amplifier, and the wake-up sig-
nal recognition is presented. The sensitivity of −47.2 dBm is
good, similar to our obtained results. However, as for all fully
active solutions, the power consumption of more than 6 µW
is much higher than for our solution. The aim of this paper
is to design a high sensitivity and under µW solution. The
authors in [15] propose an architecture featuring a low-noise
amplifier and fully active solution to achieve sensitivity of
−89 dBm. However, the solution suffers from very high power
consumption of few mW, which offsets the main benefits of
the wake-up receiver. In [16], CMOS chip including envelope
detector, low-noise baseband amplifier, mixed-signal correla-
Fig. 2. Block diagram of the wake-up receiver with the impedance
tion unit, and auxiliaries for stand-alone operation is presented. matching of WUR and the antenna to maximize the power transferred.
The achieved performance is very interesting: −71 dBm of sen-
sitivity at 868 MHz and 2.4-µW power consumption (at 1 V)
with a quite high latency of 7 ms. In this work, we present comparator is also described in detail together with the opti-
an off-the-shelf solution which can be prepared without large mization of the reference voltage source generator. To the best
nonrecurring engineering investments and ramp up to produc- of the authors’ knowledge, this is the first paper which takes into
tion volume of new dedicated chips. Moreover, our solution account all aspects of wake-up receiver design, from the match-
achieves better performance in power consumption, latency, ing network, rectifier, stages of rectifier, interrupt generator, and
and flexibility thanks to the frequency independency. from false wake-ups to the addressing capabilities.

D. Media Access Control With Wake Up Radio III. WUR D ESIGN B LOCKS
A thorough survey of various wake-up schemes and their In this section, we present the WUR blocks and the design
advantages over the wake-on (duty-cycling) schemes is pre- consideration. Fig. 1 shows a generic block diagram of a wire-
sented in [3]. It is shown that the wake-up radio without less node with a WUR. The design of the wake-up radio
addressing presented in [9] has an advantage over the other requires careful consideration of design issues in RF, analog
schemes due to its very low power consumption and low electronics, and digital and system design to carefully evaluate
latency. However, an addressing mechanism is needed to reduce the following tradeoffs:
the power consumption used during network formation, and if 1) wake-up range versus energy consumption;
the wake-up receiver can receive some command, the MAC 2) wake-up range versus delay;
data communication protocol can be simplified for low power 3) same-band versus different-band wake-up radio;
consumption. There are two ways of implementing the address- 4) addressing versus not addressing.
ing mechanism: 1) custom circuit for addressing/address com- Fig. 2 shows the main blocks required to design a WUR.
parison [14]; and 2) using a generic microcontroller [10]. The The wake-up message is received by the WSN node trough the
addressing methodology will have implications on the overall antenna. Then, a matching network is needed to provide max-
power budget and will influence the selectivity of the system. imal power transfer between the antenna (typically with 50 Ω
This confirms the importance of the capability of receiving data, impedance) and the rest of the circuit. The rectifier converts the
which is included in our approach. input radio signal into a dc signal. In general, in order to keep
The approach proposed in this paper significantly extends the the power consumption as low as possible, the rectifier is pas-
state-of-the-art with respect to power consumption, latency, and sive. Once the dc signal is generated, it needs to be “processed”
sensitivity. Moreover, it gives guidelines and insights to take by the interrupt generator to alert further logic, or the microcon-
into account when designing a WUR with off-the-shelf compo- troller. The interrupt generator also consists of a passive/active
nents. The selection of the single-stage rectifier is justified, and filtering or addressing circuitry to activate the interrupt accord-
its advantage over the multistage architecture is demonstrated ing to the address. In the following sections, all the blocks are
with measurements. The importance of the offset voltage of the analyzed in detail.
636 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 12, NO. 2, APRIL 2016

A. Frequency, Sensitivity, and Antenna

The selection of an operating frequency band for the wake-
up receiver is crucial since it will affect the size of the receiving
antenna, the operating range of the system and the matching
network selection, and availability of the required passive com-
ponents. To get an idea how the frequency affects these features
of the wake-up receiver, the Friis equation is used
Fig. 3. Envelope detector architecture with two diodes. The input of the
Pt · Gr · Gt · λ2 envelope is the RF signal while the output is the envelope of the signal
Pr = 2 . (1) in the dc domain.
(4π) · dn
Pr is the power received from an antenna, Pt is the trans- C. Matching Network
mission power, Gr and Gt are the gains of the receiving and
transmitting antenna, respectively, d is the transmission dis- As mentioned earlier, the maximum power of the received
tance, λ is the wavelength of the frequency used, and n is radio signal available on the antenna (1) is transferred to the
the path loss exponent. Pr is usually expressed in dBm and rest of the circuit when the impedance of the receiver circuit
the sensitivity of the receiving circuits equals the lowest value (Zcirt. ) equals the conjugate of the antenna impedance (Zant ).
of Pr at which the circuit can receive the wake-up message Since, in general, Zcirt is not equal to Zant ∗, a matching network
reliably. Thus, from (1), considering Pt = +10 dBm, Gr and circuit is required in between to achieve this important goal and
Gt = 1 dBi and frequency of 868 MHz, the wake-up circuit not to waste precious received power.
with sensitivity −50 dBm can achieve the maximal theoreti- The effect of mismatches is highly visible at frequencies
cal range of 34 m. Accordingly, lower frequencies allow longer over 300 MHz when significant amount of the RF power can
ranges. be reflected when it reaches an RF circuit. The most common
However, decreasing the frequency increases the antenna measure of mismatch is return loss expressed in dB or by S-
length. Depending on the application scenario, it may be impos- parameters. If an RF signal is incident on the input side of the
sible to use a large size antenna (i.e., for body area networks) circuit, some of the signal is reflected and some is transmitted
so migration to higher frequency has to be considered. If through the circuit. The ratio of the reflected electromagnetic
lower frequencies are part of the system specification, then field to the incident field is the reflection coefficient (S11).
one has to consider that antenna miniaturization at these fre- The main goal of impedance matching is the reduction or
quencies would induce significant losses which are reflected in elimination of the reflected RF signal, which involves measur-
reduced communication range or higher transmission power. ing the reflected power and designing the matching network.
Moreover, the application scenarios often force usage of the It is important to note that the matching network is strictly
same antenna/frequency as for the radio used in the network. linked with the selected working frequency. Thus, although
In fact, in many applications, to achieve low cost and small a receiver could work in different frequency domains, the
form factor, the wake-up signal has to be generated from the matching network limits this freedom.
main radio. The most commonly used frequencies in WSNs
are 868 MHz and 2.4 GHz which are the most popular for D. Rectifier
the wake-up communications as well, although examples using
other frequencies can be found in [5]–[9]. To recover the information content from the received mod-
ulated carrier waveform, a rectifier is needed. As mentioned
before, using OOK modulation enables building simple and
B. Modulation passive demodulation circuit. The circuit function is to remove
The WUR is a simple radio receiver with ultra-low power the residual carrier signal while preserving the modulating
consumption which calls for a very simple architecture. This (envelope) waveform. A simple form of rectifier consists of a
design goal also imposes constraints on the modulation used diodes detector connected to a capacitor from the output of the
for the wake-up radio. If the modulation used is highly com- circuit to the ground (Fig. 3) which works as an envelope detec-
plex [e.g., phase shift keying (PSK), quadrature phase shift key tor. The recovered envelope signal has lower amplitude and the
(QPSK), and others] the wake-up receiver demodulation cir- same form as the original signal. The peak of the incoming ac
cuits will be more complex and will require more power. The signal must be higher than the bias voltage of the diode to allow
most popular modulation scheme used in wake-up receivers is the diodes to rectify the signal. Since the envelope detector cir-
the on-off keying (OOK) [3]. OOK denotes the simplest form of cuit is basically a half-wave rectifier, there will be a dc level
amplitude-shift keying (ASK) modulation that represents digi- developed across capacitors.
tal data with the presence or absence of a carrier wave. This One of the crucial requirements for WUR receiver circuit is
modulation works as a Morse code, where, to transmit a digital to operate with weak input RF signals to achieve long range.
signal (1 and 0), a carrier wave is switched ON and OFF, respec- For a typical WSN application operating at 2.4 GHz, at dis-
tively. This simple modulation allows a drastic simplification of tances of 4–5 m between nodes, the RF signal power received
the WUR circuitry and provides power reduction opportunities on the antenna is already −30 dBm or less. As the peak volt-
as will be presented in the following sections. In the rest of this age of the ac signal obtained at the antenna is very low, diodes
paper, we assume that OOK modulation is used. with the lowest possible bias voltage are preferable. To have a
MAGNO et al.: DESIGN, IMPLEMENTATION, AND PERFORMANCE EVALUATION 637

high sensitivity WUR, zero-bias diodes should be implemented.

Rectifiers implemented with a custom IC solution are present in
the literature [11], but these are not commercially available and
are out of the goals of this paper. In this work, we will use an
envelope detector as a rectifier using commercial-of-the-shelf
zero-bias diodes able to achieve sensitivity of −56 dBm.
Rectifier’s number of stages: For the ideal electronic compo-
nents, the output voltage of the envelope detector Vout rises with
the number of the stages N

Vout = 2N (Vc − Vd ) (2)

where Vd is the bias voltage of the diodes and VC is the volt-

age at the input of the envelope detector. Equation (2) confirms Fig. 4. VC as function of the equivalent resistance REQ for constant
the requirement of zero-bias diodes. In reality, one could not Pr = −30 dBm. Increasing REQ by decreasing the numbers of stages N
continue adding as many stages as required to convert the input will increase Vc.
RF signal into a desired output voltage Vout , because (2) is an
approximate equation which does not consider the effect of
power wasted by adding diodes, capacitors, and Printed Circuits
Board (PCB) lines as explained in [17]. Thus, increasing the
number of stages N reduces the whole efficiency of the envelope
detector. This is particularly important for the wake-up receiver
where the RF signal is very weak and Vd becomes comparable
with VC .
Another important effect to be considered is that VC depends
on the number of stages as well. From (3) and (4), as more
stages are added, equivalent input resistance of the enve-
lope detector (REQ ) becomes smaller, the conductance (XEQ )
becomes higher and VC decreases
VC 2 
Pr = V C I = VC = Pr REQ (3) Fig. 5. Nanopower high-sensitivity WUR architecture.
REQ
1
REQ ∝ ; XEQ ∝ N. (4) and trigger an IC directly, as presented in previous sections,
N
this is unfeasible for the weak RF signals typical for the WUR
Fig. 4 depicts the input voltage of the envelope detector VC receivers. In fact, as Fig. 4 shows, with a received signal of
relation to the equivalent resistance REQ for a constant received −30 dBm or lower, the input voltage of the rectifier is only a few
power Pr = −30 dBm. Fig. 4 shows that VC increases with an mV and adding stages to the envelope detector will decrease
increase of the REQ , according to (3), and the measured value the input voltage wasting the multiplier effect. For this reason,
of REQ for different stages. Therefore, in order to maximize the this block needs to be supplied with battery as it generates an
range of the WUR receiver, we need to maximize the VC that is interrupt with higher voltage (at least 1.5 V) able to wake up
detected by the comparator, which depends also from the mini- a microcontroller or an IC. The simplest and low power way
mal offset voltage of the comparator self (see next section). So, to do this (the base to the most WURs) is using an ultra-low
it is very crucial to utilize a single-stage envelope detector. This power comparator [8]–[10]. From the signal generated by the
is also the main difference between our architecture and previ- envelope detector, it is necessary to create both the reference
ous works (i.e., [9]) where a double-stage rectifier has been used voltage and the signal at the input of the comparator. If the
without any justification and resulting in reduced sensitivity. low power consumption is the main constraint, this reference
source can be passive using a simple Resistor–Capacitor (RC)
low-pass filter. In the same way, it is possible to decouple the
E. Interrupt Generator
comparator input, filtering the output of the comparator with
The function of this block is to generate an interrupt to a another RC filter. Fig. 5 depicts a wake-up radio architecture
microcontroller with the shortest latency and minimal influ- where a semi-passive interrupt generator with the comparator as
ences of interferences. This is the first block of the wake-up the only active component is present. This architecture is sim-
receiver operating with a dc signal and the enveloped data ple, consumes negligible power, and although it is not possible
(Fig. 2). There is also an option to add the addressing capa- to perform addresses parsing, the data are preserved for further
bility to this block to selectively wake up at the specific address blocks with addressing capability.
and to generate digital data for the microcontroller or an IC. Fig. 6 shows the output voltage of an interrupt generator built
Despite the case of the RF energy harvesting [18], where multi- with a comparator compared to the RF signal arriving to the
stage envelope detectors can provide a voltage higher than 1 V antenna. The interrupt generator gives a digital envelope of the
638 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 12, NO. 2, APRIL 2016

address. There are different implementations possible for the

addressing generator; however, the majority uses an ad-hoc
designed ASIC or an ultra-low power microcontroller. It is
important to notice when an addressing circuit is present,
the final wake up signal is generated only after the address
is received. This increases the overall wake up latency but
can reduce the number of false positives (FPs) and, in turn,
the power consumption at the system level. The wake-up
time with addressing is proportional to the length of the
address and baud rate with Wakeup Time = Nbits ∗ Tbit where
Tbit = 1/Baud rate, and Nbits is the number of the bits used for
the address.

Fig. 6. OOK signal on the antenna of the WUR (plot 1 down /yellow) and
IV. P ROPOSED U LTRA -L OW P OWER WUR
data out from the interrupt generator (plot 2 up/blue). This section presents the proposed design of a nanowatt
WUR which uses only a comparator as an active component to
original OOK message at the output. This feature is very inter-
evaluate and demonstrate the proposed analysis in a real devel-
esting for developing further integrated or discrete circuits, to
opment. Moreover, this section will present a method to use the
automatically detect the addressing or use the microcontroller,
WUR by adding addressing capability, keeping the power low.
which is woken up through the interrupt generator to process
All the blocks presented in previous sections are described in
the data and make the decision to wake up the radio or else go
detail.
back to sleep.
The proposed architecture’s (Fig. 5) comparator is the only
Due to low-pass filtering at the comparator output, the inter-
component powered at all times. The main goals when design-
rupt is generated only if the preamble (Fig. 6) is longer than
ing this receiver are as follows:
400 µs (that can be tuneable by the Resistor–Capacitor-Resistor
1) ultra-low power, around or less than 1 µW;
(RCR) filter). This will protect the circuits from interference but
2) high sensitivity, with at least −35 dBm;
will create a delay shown in Fig. 6. This is another tradeoff of
3) flexibility adaptable to different frequencies in ISM band
the wake-up receiver which requires careful design. The dura-
with OOK modulation;
tion of the delay can be set by changing the value of R2 , C6 ,
4) reactivity less than 100 µs for the first interrupt;
and R3 in Fig. 5.
5) preservation of the received data, for further processing to
The main criteria for selecting the comparator circuit are:
perform addressing or receive commands;
1) power consumption; 2) input offset voltage VIO ; and 3) prop-
6) optional ultra-low power and low-cost addressing man-
agation delay.
agement.
Bearing in mind that the goal is to build a wake-up receiver
circuit that will have very low power consumption, the quies-
cent current consumption of the comparator must be in the order A. Rectifier
of few nano-Amps. We built a single-stage rectifier with HSMS-285C diodes
The input offset voltage is another important parameter of [22]. HSMS-285C is used in the proposed solution, as the
the comparator. It is defined as the differential input voltage requirements are to cover frequencies below 1 GHz; however,
to apply in order for the comparator to be at a toggling level the same schematic we are presenting can be used for frequen-
[19]. As further discussed in [19], the input offset voltage also cies above 1.5 GHz (HSMS 282x diodes). These diodes are
limits the resolution of a comparator. Therefore, for very small optimized for incoming power lower than −20 dBm and offer
signals (in the same order as the input offset voltage–VIO ), the sensitivity of −57 dBm that makes them the best solution on
comparator toggles at an undesired rate or does not toggle at all. the market at the time this paper is written. The diodes are
To increase the sensitivity of the wake-up receiver and avoid the successfully used in several RF energy harvesting and wake-
possible issues discussed above, the input offset voltage of the up applications [18], which confirm their performance in weak
comparator has to be lower than the minimal expected output RF signal environments.
voltage of the envelope detector circuit, VC .
B. Comparator
F. Addressing Management Circuit
Three different comparators were used in the presented study
The addressing management circuit is an optional subsystem in order to evaluate the tradeoff of power and sensitivity. We
in charge to decode the data received by WUR and generate selected TLV3691 from Texas instruments (TI) with ultra-low
a further trigger only when the address is recognized. If it is current consumption and 2.5-mV input offset, AS1976 from
present, this is the first digital part which is woken up from Austria Microsystems with 1.8-mV input offset and LPV7215
the interrupt generation. The addressing circuit provides a from TI with 0.5-mV input offset. Table I shows the perfor-
digital interface for the main microcontroller to be set with the mance of the comparators in terms of power, sensitivity, and
desirable address. When the addressing circuit is not present, range. The input offset voltage of each selected comparator was
the main microcontroller can process the data and evaluate its measured using the method presented in [19]. The results in the
MAGNO et al.: DESIGN, IMPLEMENTATION, AND PERFORMANCE EVALUATION 639

TABLE I
WAKE -U P R ADIO P ERFORMANCE FOR D IFFERENT C OMPARATORS AT
2 V S UPPLY VOLTAGE

Fig. 7. Nano-power wake up radio with addressing capability provided

table show that the current consumption is higher for the com- by the PIC12LF ultra low power microcontroller.
parator with a lower voltage offset. However, as explained in
the previous section, the voltage offset affects the sensitivity of network using advance design system (ADS). In the proposed
the WUR. implementation, 868 MHz was selected which has a good trade-
The proposed design (Fig. 5) uses an adaptive threshold off between antenna size and range, and it is commonly used in
mechanism that keeps the inverted input of the comparator at WSNs. An equivalent model was built in ADS using SPICE
half of the input signal level. With this approach, the power parameters of the HSMS-285c diode.
consumption of the circuit is reduced since, instead of a volt- The matching network evaluated by ADS consists of an LC
age divider, the signal from the antenna is used for generating circuit, L = 74 nH and C = 1.2 pF. This matching has to be
the threshold. The adaptive threshold mechanism has been taken into account as initial evaluation. The exact value of the
designed using a simple RC circuit (R1 − C5 ) connected to the L and C has to be measured directly on the board using the
inverted input of the comparator while the noninverted input S11 parameter measurements to have optimal matching. This
is directly connected to the signal from the envelope detector concept will be described in more details in the next section.
allowing the comparator to detect both weak and strong signals
effectively.
E. Ultra-Low Power Addressing
C. Preamble Detector In this section, we propose a solution to add addressing and
Following the comparator, there is the preamble detector create a Nanopower WUR with addressing capability to evalu-
which generates the interrupt and the digital data. We used ate the benefits on the overall reduction in power consumption.
a specific preamble in the wake-up message which can be Fig. 7 shows the main blocks of the proposed solution. The
detected by this passive circuit. The preamble is an OOK- first block is WUR presented in the previous section with
modulated signal (Fig. 6) with a specific length (number of sensitivity from −32 to −55 dBm, depending on the selected
bytes) sent at a specific bit rate (frequency fpa ). The pream- comparator. The addressing block is implemented to further
ble detector is responsible for generating a wake-up interrupt avoid FP messages and wake up the node only when intended.
for a microcontroller (or other IC) only if a preamble with the In fact, this second block generates the final wake-up signal for
defined frequency has been received. To reject signals with the the main microcontroller only if the first byte received is the
frequency lower than fpa , a low-pass filter with a cut-off fre- correct address. The main microcontroller can set its address
quency lower or equal to fpa can be implemented, as shown in dynamically during run time through the SPI port and also
Fig. 5 (R2 and C6 ). At the rising edge of the comparator, C6 send and receive commands. In the presented architecture, this
will gradually start to charge. At the falling edge of the com- task is done by an 8-bit ultra-low power microcontroller from
parator, for signals with frequency f < fpa , capacitor C6 will microchip (PIC12LF1552) which consumes only 40 nW at 2 V
be completely discharged only through resistor R3 by the time of sleep power. So, the overall power consumption in listening
the next rising edge occurs on the output of the comparator. In mode is increased for only 40 nW with respect to the power
this way, the signals with frequencies f < fpa will not be able consumption of Table I. The PIC can be also bypassed, and the
to charge the capacitor C6 to a level that can trigger an interrupt interrupt and data after the first block can be sent directly to the
for a microcontroller (0.66 VCC ). The Schottky diode (D5 ) on main microcontroller.
the output of the comparator does not allow the capacitor C6
to discharge through R2 and the comparator output when the
V. E XPERIMENTAL S ETUP
output of the comparator is grounded (0 V). As in Fig. 6, this
mechanism is able to generate the data from the WUR which To evaluate the functionality and to verify the performance of
can be detected by a microcontroller or an IC. the proposed WUR in terms of power consumption, sensitivity,
latency, and data addressing capabilities, the wake-up receiver
was developed, and measurements were carried out for both
D. Matching Network versions with and without addressing capability. The prototype
Once the components are selected, the working frequency platform was developed using the proposed WUR connected
has to be defined to design needed impedance matching to the PCB with CC430F5147 system-on-chip (SoC) from TI.
640 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 12, NO. 2, APRIL 2016

Fig. 8. First prototyped platform (WUR and CC430F5147).

The SoC is an ultra-low power microcontroller with an inte- Fig. 9. Wake-up signal and data (blue) generated from the WUR, and
grated RF transceiver core able to communicate at 868 MHz sample period of the microcontroller to acquire the data.
with OOK modulation. The prototype platform is presented in
Fig. 8. It has two separate antennas, one for the main transceiver
and one for the wake-up receiver. Although the receiver and the
Latency is also a key feature for the transmission dura-
transmitter on chip are tuned to 868 MHz, in a future devel-
tion and the overall consumption of the transmitting node. To
opment, the prototype can share a single antenna. This choice
evaluate this power consumption, we measured the transmis-
provided a more accurate measurement of the wake-up receiver
sion power of the CC430F5147 microcontroller. The average
which will not be affected by the antenna switching (RF mul-
power consumption measured when transmitting in OOK with
tiplexer). For the wake up with addressing capability (Fig. 7),
+10 dBm output power was 105 mW. This is a considerable
we developed a new prototype including the wake-up receiver
transmission power especially if WUR architectures need a
and the PIC microcontroller in a single PCB. The CC430F5147
long preamble of hundreds of milliseconds as in [9] and [24].
was used as a transmitter. As we mentioned before, the trans-
Address processing: To evaluate the capability to receive data
mitter was tuned to send an OOK modulated signal at 868 MHz
and use this data for addressing or commands, firmware on the
and 10 kbit/s or fpa = 1 kHz. A 2-bit preamble has been used
microcontroller was developed to sample the data after receiv-
beforehand to send the command/address. The values of the
ing the wake-up interrupt. Fig. 9 shows the waveform of the
resistors and capacitors (Fig. 5) used for the interferences avoid-
wake-up receiver in blue (bottom plot) and the sample time of
ance circuit were as follows: R2 = 1 kΩ; R3 = 100 kΩ; and
the microcontroller in yellow (top plot). After the interrupt has
C6 = 200 nF. The values of the RC filter for the comparator ref-
been generated and the microcontroller wakes up, it starts to
erence are C3 = 22 pF C5 = 1 nF. All the three configurations
acquire the data until the wake-up interrupt becomes low again.
with the three proposed comparators have been implemented
In this case, a 16-bit packet was sent and correctly received by
and tested, and the experimental results are presented in next
the microcontroller.
section.
Impedance matching: A network analyzer was used to mea-
sure the S11 parameter of the previously determined matching
VI. E XPERIMENTAL V ERIFICATION circuit using ASD tool. Measurements of the S11 parameter
showed that the matching is not optimized for 868 MHz but
In this section, we present the experimental results and the for 3.1 GHz. The reason for this error is due to imperfect model
simulations, in order to evaluate the benefits of the proposed of the diodes and comparator, the PCB effect, and other side
solution compared with previous works. Moreover, we evaluate effects when developing board (such as the antenna connectors
the two versions, with and without on-board addressing. and soldering). Based on these measurements, we changed the
C = 7 pF and L = 22 nH, and the matching was optimized for
868 MHz with S11 of −29 dB (Fig. 10).
A. Wake-Up Radio Features Measurements Power consumption and sensitivity: As we mentioned before,
Minimal latency evaluation: Due to the interference avoid- the comparator used affects both the power consumption and
ance filter (Fig. 5) and the value of the RC filter selected in the sensitivity. Measurements of both the parameters were
our experimental setup, the WUR generates wake-up inter- carried. Table I shows: 1) the measured sensitivity of three dif-
rupts only if the preamble is detected for 60 µs that is also the ferent versions of the WUR; 2) the power consumption; and
latency of the wake-up signal. We also measured the minimal 3) range in open air at +10-dBm transmission power and a 3-
latency achieved by eliminating the interferences avoidance cir- dBi antenna. The maximal range achieved with the LPV7215
cuit which is only around 8 µs. It is clear that in that condition, comparator is 50 m (with −55 dBm sensitivity) which is an
the wake-up radio can provide a high number of FPs due to the interesting value for a wide range of outdoor applications. The
interference, so it is convenient to keep the interference avoid- range of 7 m obtained with TLV3691-based version can be used
ance circuit and tune the RC filter according to the desirable in a body area network scenario and where power consumption
latency. is critical. To evaluate the power consumption of the WUR, we
MAGNO et al.: DESIGN, IMPLEMENTATION, AND PERFORMANCE EVALUATION 641

TABLE II
WAKE -U P R ADIO FP/FN P ERFORMANCE

V2) Without any addressing, but with the interference avoid-

Fig. 10. S11 parameter evaluated for the impedance matching with ance circuits, and tuned to generate an interrupt only if an
L and C optimized experimentally. The new matching obtains S11 OOK message long at least 60 µs is detected.
−29 dBm at 868 MHz.
V3) With both addressing and interference avoidance circuits.
So, the first bit activates the PIC microcontroller after
60 µs, and then the PIC processes the data and wakes up
used a PCB with only the wake-up receiver soldered and sup- the main microcontroller only if the address matches the
plied with 1.8 V. We used a 100-Ω shunt resistor to check the WUR address.
current flowing in the resistor. The static power consumption Table II shows the performance in terms of FPs and FNs for
(when listening to a channel) of the presented WUR archi- the three implementations with respect to the distance. There is
tecture is determined by the static power consumption of the a high number of FPs on both configurations without address-
comparator and can achieve only 152 nW for −32 dBm and ing. In fact, the WUR without addressing wakes up the main
1196 nW for the −55 dBm version. microcontroller also for the messages intended to wake up other
The power consumption of wake-up receiver with the nodes. Moreover, it is possible to notice the influence on the
addressing capability directly on board through the PIC micro- interferences in the number of FPs. On the other hand, the WUR
controller has been measured both in sleep mode and in active without addressing has zero FNs. This is due to the very high
mode. The PIC microcontroller increases the overall power probability that at least one of the transmitted 8 bits for the
consumption for 40 nW when no data are received and for addresses is received and generates an interrupt. The test on
63 µW during the active mode and processing the data. The the WURs with addressing improved the FP performance, espe-
power consumed by the CC430 in order to process the data cially for short ranges where the data received are more reliable
is 600 µW. This confirms the benefits of using the addressing due to the quality of the signal. The number of FPs and FNs
capability with the PIC over the external addressing (done by increases when the range increases due to weaker signals and
the main microcontroller), even when the microcontroller is an the corruption of the data.
ultra-low power microcontroller like the CC430.
FP and false negative (FN) analysis: To evaluate the benefits
of the addressing and interferences avoidance circuits to reduce
FPs, we tested the WUR communication in our lab. The setup B. Comparison With Existing WUR Solutions and MAC
has been done as follows. Protocols
1) One WUR unit with a +3 dBi antenna and the LPV7125 Fig. 11 presents the three proposed WUR solutions (with
comparator has been assembled and deployed. and without addressing capabilities) compared to recent state
2) One transmitter using the CC430 SoC has been placed at of the art of the WUR solutions. The solution with the lowest
different distances (1, 5, 10, 15, 20, 25 m), transmitting power consumption (Roberts [11], 98 nW) has a lower sensitiv-
1000 activating messages of 8 bits representing wake-up ity, −41 dBm, and it does not support addressing, which makes
or addressing message. its effective energy consumption significantly higher in real
3) We sent four different addresses to wake up four different deployments (as shown in the following results). The solution
WURs. with the highest (−81 dBm) sensitivity (Milosiu [25]) also has
So, only 250 messages contained the address for the tested a high power consumption (3 µW) and especially high latency
wake-up receiver and 750 messages will contain addresses for (484 ms). Our solution without addressing support achieves the
other three WURs not evaluated. lowest latency (8 µs), but its consumption in real-world appli-
We tested three different versions of the WUR. cation will be increased due to excessive unnecessary wake-ups.
V1) Without any addressing capability and interference avoid- Our solution with addressing support shows the best tradeoff
ance circuits, so the WUR generates an interrupt for the in the power-latency-sensitivity space, taking into account that
main microcontroller in 8 µs whenever an OOK signal is the WUR described by Roberts [11] does not support address-
detected. ing. Moreover, as explained throughout this paper, our proposed
642 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 12, NO. 2, APRIL 2016

Fig. 13. Lifetime increment of the nodes with WURs against the WOR
(0.1% duty-cycle), considering the addressing possibilities. Our pro-
posed WURs are with addressing capabilities.

number of FPs increases with the WUR range and number of

transmitters, so this number can be also higher in real deploy-
Fig. 11. Comparison of the proposed WURs (WUR 1, WUR 2, and
WUR 3) with and without addressing support, with the recently published ment. Thus, the increase in design complexity and in power
WURs, in terms of power consumption, sensitivity, and latency. consumption of the WUR introduced by the addressing decod-
ing circuitry proves to be beneficial for the overall energy
consumption of the WUR. WUR solutions are compared with
the duty-cycling solution (WOR) with 0.1% duty cycle, where
the main radio turns ON and OFF periodically regardless of the
message reception. WOR is a capability featured by some TI
radios that enables periodically waking up from radio sleep
mode and listening for incoming packets without microcon-
troller interaction [26]. This feature is very common in several
commercial radios to save energy. For 100 messages per hour,
our WUR 3 (higher power consumption) has power levels 10
times lower, compared to the duty-cycled solution, while our
least power consuming WUR consumes 12.7 times less power.
Fig. 12. Average power consumption of the communication with pro-
posed WURs and recently published WURs, compared to the duty- For 10 messages per hour, power consumption values are much
cycled radio (WOR) with 0.1% duty cycle, taking into account that bigger (28.6 times and 84.2 times, respectively). The increase
Roberts WUR and Oller13 WUR do not have addressing possibilities. in average power consumption with more received messages is
due to the fact that the major part of energy consumption is due
architecture is flexible and fully adaptable for covering a wide to the activation of the main radio for receiving and transmitting
range of frequencies and power/latency/sensitivity tradeoff. the messages. For the duty-cycled solution, the average power
To evaluate the benefits of using the wake-up radio over consumption increases 7% when changed from 10 messages per
WSN protocol and for a further comparison with the state of the hour to 100 messages per hour, due to the significant influence
art, simulations using measured data were performed. Fig. 12 of radio transmission, regardless of the received messages and
shows the average power consumption of different WUR solu- processing.
tions, for different numbers of received wake-up messages per Finally, we evaluated a real-world scenario where a node
hour (10, 50, and 100 messages per hour). When the wake- receives a wake-up message, sends an acknowledgment, acti-
up message arrives, the WUR processes it (reads the address vates its sensor, acquires the data, processes the data, and sends
if there is any) and turns on the main transceiver to send the the information in a short message to another node. Fig. 13
acknowledgment and to receive the main message. The rest of shows how many times each WUR solution outperforms the
the time the system is in inactive mode. If WUR does not have duty-cycled solution in terms of node’s battery lifetime. As the
addressing capabilities, each message received by the wake-up worst case (proposed WUR 3 with addressing support, for 100
circuitry wakes up the node. We suppose the effective number messages per hour), our solution achieves an 8 times longer
of received messages for such WURs (our solutions without lifetime.
addressing, as well as Roberts [11] and Oller [8]) to be three
times higher, due to FPs. The influence of the increased WUR
receiver activity without the addressing support to the incre- VII. C ONCLUSION
ment of their average power consumption can be seen. The Numerous design considerations for an ultra-low power,
average power consumption of our solutions without addressing high-sensitivity, and fast reaction WUR for WSN have been
is three to four times higher than the average power consump- proposed and exemplified through the development of a com-
tion of our solutions with addressing support (depending on plete wake-up receiver subsystem with off-the-shelf com-
the actual WUR and the number of messages per hour). The ponents. We presented circuit topologies and optimization
MAGNO et al.: DESIGN, IMPLEMENTATION, AND PERFORMANCE EVALUATION 643

methodologies for impedance-matched passive rectifier design, [14] G. U. Gamm et al., “Low-power sensor node with addressable wake-up
nanopower interrupt generator, and addressing capability. The on-demand capability,” Int. J. Sens. Netw., vol. 11, no. 1, pp. 48–56, 2012.
[15] S. Drago, D. M. W. Leenaerts, F. Sebastiano, L. J. Breems,
design techniques are supported by mathematical models, K. A. A. Makinwa, and B. Nauta, “A 2.4 GHz 830 pJ/bit duty-cycled
thorough simulations, and experiments on three different wake-up receiver with -82 dBm sensitivity for crystal-less wireless sen-
developments to better define the achievable tradeoffs. As sor nodes,” in Proc. IEEE Int. Conf. Solid-State Circuits (ISSCC’10). San
Francisco, CA, USA, Feb. 7–11, 2010, pp. 224–225.
many applications require addressing capabilities, the proposed [16] C. Hambeck, S. Mahlknecht, and T. Herndl, “A 2.4 µW wake-up receiver
approach is able to send the received data to a microcon- for wireless sensor nodes with–71dBm sensitivity,” in Proc. IEEE Int.
troller or other IC to perform address detection. The proposed Symp. Circuits Syst. (ISCAS), May 15–18, 2011, pp. 534–537.
[17] A. W. Scott and R. Frobenius, RF Measurements for Cellular Phones and
approach can significantly reduce power consumption of the Wireless Data Systems. Hoboken, NJ, USA: Wiley, 2011.
main radio and then of the node and network. Experimental [18] P. Nintanavongsa, U. Muncuk, D. R. Lewis, and K. R. Chowdhury,
results indicate that the proposed WUR outperforms the state “Design optimization and implementation for RF energy harvesting cir-
cuits,” IEEE J. Emerg. Sel. Topics Circuits Syst., vol. 2, no. 1, pp. 24–33,
of the art bringing a flexible solution with ultra-low power Mar. 2012.
consumption, high sensitivity, very low latency, and address- [19] R. Smat. (2012 Oct.) Introduction to Comparators, Their Parameters and
ing capability. Field measurements confirmed that the proposed Basic Applications, AN4071 Application note [Online]. Available: http://
www.st.com/st-web-ui/static/active/en/resource/technical/document/
WUR consumes only 152 nW in lower power version with application_note/DM00050759.pdf.
−32 dBm, and achieves up to −55 dBm of maximum sensitiv- [20] Austrian Microsystems. AS1976, AS1977 Ultra Low Current, 1.8V
ity and 1.2 µW power consumption. Moreover, the benefits of Comparators, 1997–2007, AS1976 Datasheet [Online]. Available:
http://www.ams.com/eng/Products/Power-Management/Comparators/
the addressing over no addressing are presented measuring the AS1976.
FP and FP rates. Finally, simulations of a real-world scenario, [21] Texas Instruments. (2013 Sep.) M SP 430T M SoC With RF Core,
where implementation of different WUR solutions is compared CC430F5137 Datasheet [Online]. Available: http://www.ti.com/lit/ds/
symlink/cc430f5137.pdf.
with other wake-up solutions and duty-cycled radio protocol [22] Avago Technologies US. (2009 May 29). HSMS285x-Series Datasheet
for WSNs, show that with our solutions, a battery lifetime [Online]. Available: http://www.avagotech.com/docs/AV02-1377EN.
prolongation of about eight times (worst-case) is achieved. [23] V. Jelicic, M. Magno, D. Brunelli, V. Bilas, and L. Benini, “Benefits of
wake-up radio in energy-efficient multimodal surveillance wireless sensor
network,” IEEE Sensors J., vol. 14, no. 9, pp. 3210–3220, Sep. 2014.
R EFERENCES [24] J. Oller, I. Demirkol, J. Casademont, J. Paradells, G. Gamm, and
L. Reindl, “Performance evaluation and comparative analysis of sub-
[1] A. Willig, “Recent and emerging topic in wireless industrial communica- carrier modulation wake-up radio systems for energy-efficient wireless
tions: A selection,” IEEE Trans. Ind. Informat., vol. 4, no. 2, pp. 102–124, sensor networks,” Sensors, vol. 14, pp. 22–51, 2014.
May 2008. [25] H. Milosiu et al., “A 3-µW 868-MHz wake-up receiver with -83 dBm sen-
[2] I. Demirkol, C. Ersoy, and E. Onur, “Wake-up receivers for wireless sen- sitivity and scalable data rate,” in Proc. ESSCIRC, Bucharest, Romania,
sor networks: Benefits and challenges,” IEEE Wireless Commun., vol. 16, Sep. 16–20, 2013, pp. 387–390.
no. 4, pp. 88–96, Aug. 2009. [26] D. Dang, “CC430 wake-on-radio functionality,” Texas Instrum., Dallas,
[3] V. Jelicic, M. Magno, D. Brunelli, V. Bilas, and L. Benini, “Analytic TX, USA, Appl. Rep. SLAA459A, 2012.
comparison of wake-up receivers for WSNs and benefits over the wake-
on radio scheme,” in Proc. 7th ACM Workshop Perform. Monit. Meas.
Heterogen. Wireless Wired Netw., Oct. 2012, pp. 99–106. Michele Magno (M’13–SM’16) received the
[4] M. Magno, D. Boyle, D. Brunelli, E. Popovici, and L. Benini, “Ensuring M.S. and Ph.D. degrees in electronic engineer-
survivability of resource-intensive sensor networks through ultra-low ing from the University of Bologna, Bologna,
power overlays,” IEEE Trans. Ind. Informat., vol. 10, no. 2, pp. 946–956, Italy, in 2004 and 2010, respectively.
May 2014. Currently, he is a Postdoctoral Researcher
[5] H. Ba et al., “Passive RFID-based wake-up radios for wireless sensor with Swiss Federal Institute of Technology (ETH)
networks,” in Wirelessly Powered Sensor Networks and Computational Zurich, Zurich, Switzerland, and a Research
RFID. New York, NY, USA: Springer, 2013, pp. 113–129. Fellow with the University of Bologna. He
[6] A. Castagnetti, A. Pegatoquet, T. N. Le, and M. Auguin, “A joint duty- has collaborated with several universities and
cycle and transmission power management for energy harvesting WSN,” research centers, such as the University College
IEEE Trans. Ind. Informat., vol. 10, no. 2, pp. 928–936, May 2014. Cork and the Tyndall Institute, Cork, Ireland;
[7] D. Spenza, M. Magno, S. Basagni, L. Benini, M. Paoli, and C. Petrioli, Imperial College London, London, U.K.; the University of Trento,
“Beyond duty cycling: Wake-up radio with selective awakenings for long- Trento, Italy; Politecnico di Torino, Torino, Italy; the University of British
lived wireless sensing systems,” in Proc. IEEE Conf. Comput. Commun. Columbia, Vancouver, BC, Canada; etc. He has authored more than 60
(INFOCOM), 2015, pp. 522–530. papers in international journals and conferences. His research interests
[8] J. Oller et al., “Design, development, and performance evaluation of a include wireless sensor networks, power management techniques, and
low-cost, low-power wake-up radio system for wireless sensor networks,” extension of lifetime of batteries-operating devices and embedded video
ACM Trans. Sens. Netw., vol. 10, no. 1, p. 11, 2013. surveillance.
[9] S. Marinkovic and E. Popovici, “Nano-power wireless wake up receiver
with serial peripheral interface,” IEEE J. Sel. Areas Commun., vol. 29,
no. 8, pp. 1641–1647, Sep. 2011. Vana Jelicic (S’09) received the Ph.D. and
[10] B. Alírio Soares and N. Borges Carvalho, “A low-power wakeup radio M.S. degrees in electrical engineering from
for application in WSN-based indoor location systems,” Int. J. Wireless the Faculty of Electrical Engineering and
Inf. Netw., vol. 20, no. 1, pp. 67–73, 2013. Computing, University of Zagreb, Zagreb,
[11] N. E. Roberts and D. D. Wentzloff, “A 98nW wake-up radio for wireless Croatia, in 2014 and 2009, respectively.
body area networks,” in Proc. IEEE Radio Freq. Integr. Circuits Symp. She was a Visiting Researcher at the
(RFIC), Jun. 17–19, 2012, pp. 373–376. Department of Electrical Engineering and
[12] M. Magno and L. Benini, “An ultra low power high sensitivity wake- Computer Science, University of Bologna,
up radio receiver with addressing capability,” in Proc. IEEE 10th Int. Bologna, Italy, from 2010 to 2012. In 2015, she
Conf. Wireless Mobile Comput. Netw. Commun. (WiMob), Oct. 8, 2014, was a Postdoctoral Researcher with the Faculty
pp. 92–99. of Electrical Engineering and Computing,
[13] K. Takahagi et al., “Low-power wake-up receiver with subthreshold University of Zagreb. Her research interests include smart sensors and
CMOS circuits for wireless sensor networks,” Analog Integr. Circuits wireless sensor networks (WSNs), with focus on power management in
Signal Process., vol. 75, no. 2, pp. 199–205, 2013. WSNs with high-consuming sensors.
644 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 12, NO. 2, APRIL 2016

Bruno Srbinovski (S’14) received the M.S. Luca Benini (S’94–M’97–SM’04–F’07) received
degree in electrical engineering and information the Ph.D. degree in electrical engineering from
technologies from the Faculty of Electrical Stanford University, Stanford, CA, USA, in 1997.
Engineering and Information Technologies, He is a Full Professor with the University
Skopje, Macedonia, in 2011, and the M.S. of Bologna, Bologna, Italy, and he is the Chair
degree in electronic engineering from University of Digital Circuits and Systems with Swiss
College Cork, Cork, Ireland, in 2013. He is Federal Institute of Technology (ETHZ), Zurich,
currently working toward the Ph.D. degree at the Switzerland. He served as Chief Architect
University College Cork. for the Platform2012/STHORM project in
His research interests include the area of STmicroelectronics, Grenoble, France, in the
wind energy harvesting, wireless sensor net- period 2009–2013. He has held Visiting and
works, power management, sensor interface, and digital signal. Consulting Researcher positions at École polytechnique fédérale de
Lausanne (EPFL), Lausanne, Switzerland; IMEC, Leuven, Belgium;
Hewlett-Packard Laboratories; and Stanford University, Stanford,
CA, USA. He has authored more than 700 papers in peer-reviewed
Vedran Bilas (M’98–SM’10) received the Ph.D. international journals and conferences, four books, and several book
degree in electrical engineering from the chapters. His research interests include energy-efficient system design,
University of Zagreb, Zagreb, Croatia, in 1999. multicore state of charge design, energy-efficient smart sensors, and
He is a Professor heading the Laboratory sensor networks for biomedical and ambient intelligence applications.
for Intelligent Sensor Systems with the Faculty Dr. Benini is a Member of Academia Europaea.
of Electrical Engineering and Computing,
University of Zagreb. He has over 20 years of
research, development, and technology transfer
experience in the area of sensors and electronic
systems. His research interests include the field
of energy efficient intelligent and networked
sensors in various application domains.

Emanuel Popovici (S’00–M’03–M’08) received

the Dipl. Ing. degree in computer engi-
neering from Politehnica University Timisoara,
Timisoara, Romania, in 1997, and the Ph.D.
degree in microelectronic engineering from the
National University of Ireland, Cork, Ireland, in
2002.
He has been a Senior Lecturer with
the Department of Electrical and Electronic
Engineering, National University of Ireland, since
2002. His research interests include embedded
system design for low power, reliable, and secure computing and com-
munications with applications in smart LED lighting systems, biomedical,
agricultural, structural health, energy control and optimization, learning,
and entertainment domains. He has published widely in these areas,
with more than 10 co-authored papers being distinguished at national
and international levels.
Dr. Popovici co-supervised three interdisciplinary teams who
achieved top awards from the IEEE/International Business Machines
(IBM) Smarter Planet Challenge in 2011, 2013, and 2014.