2554 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 21, NO.

3, THIRD QUARTER 2019

5G Waveforms for IoT Applications

Ivo Bizon Franco de Almeida , Luciano Leonel Mendes , Joel J. P. C. Rodrigues , Senior Member, IEEE,
and Mauro A. A. da Cruz

Abstract—In the future, many objects will communicate with as an integration of wired and wireless communication tech-
each other and with human beings, enabling a vast range of nologies, sensors and actuators that allow users to control and
possible services and applications. However, this scenario comes monitor objects (things) through the Internet, which also coop-
with stringent network requirements. This paper studies and
discusses recent technologies that were proposed to achieve the erate among themselves [4]. IoT devices will be pervasive, i.e.,
requirements of the Internet of Things (IoT) and massive machine they will be present in houses, hospitals, streets, parking lots,
type communications (mMTC). It focuses on their physical layer farms, and factory plants.
characteristics and identifies gaps that 5G networks need to Consequently, the connectivity profile will change drasti-
address to achieve full connectivity in the IoT scenario. Hence, cally from what is now experienced in 4G cellular networks,
a detailed tutorial on the possible radio access technologies
for 5G networks focusing on IoT use-cases is presented. The where most connected devices are smartphones that start new
advanced 5G waveforms discussed are: 1) orthogonal frequency data transfers according to the owner’s profile.
division multiplexing (OFDM); 2) universal filtered multicarrier Forecasts regarding IoT devices state that, by 2020, the con-
(UFMC); 3) filter bank multicarrier (FBMC); and 4) generalized nection density will get up to 106 devices/km2 [5]. Hence, the
frequency division multiplexing (GFDM). The features that each radio access technology must cope with a massive number
radio access technology should present to address the main IoT’s
requirements are emphasized. of heterogeneous devices as well as a high volume of data
consumed by smartphones. The 4G network is not prepared
Index Terms—5G, PHY layer, Internet of Things, FBMC, to accommodate the envisioned IoT services, although efforts
GFDM, OFDM, UFMC, wireless communications.
are being made to introduce this type of service with Long
Term Evolution (LTE) in most recent releases [6]. For instance,
maintaining orthogonality and synchronization among users
I. I NTRODUCTION
and radio base stations (RBS) demands a considerable amount
HE DEPLOYMENT of the fifth generation (5G) mobile
T network is expected to start by 2020 [1]. This new and
omnipresent network will represent a significant change in the
of energy. Consequently, battery-powered devices cannot oper-
ate for long periods of time without constant battery replace-
ment. This issue represents a significant shortcoming for the
industry, business, and people’s life in general. Coverage, data IoT scenario [7]. IoT applications are an important motivation
throughput, latency, and energy efficiency will be enhanced for the development of a new mobile network, and it is also
compared to the current fourth generation (4G) system [2]. one of the major challenges faced by 5G networks.
Therefore, a new range of novel services and applications is Regarding 5G’s physical layer (PHY) standardization, there
yet to emerge. For instance, the Internet of Things (IoT) [3] is an exciting discussion about the best suited waveform for
is a networking paradigm that is promising to change the way achieving the required performance among all the scenarios
technology is experienced in daily life. IoT can be defined that 5G will address. Figure 1 illustrates the main scenarios
and possible applications for 5G. The scenarios include i) ultra
Manuscript received April 16, 2018; revised October 6, 2018,
December 26, 2018, and March 2, 2019; accepted March 20, 2019. Date reliable low latency communications (URLLC), ii) enhanced
of publication April 11, 2019; date of current version August 20, 2019. This mobile broadband (eMBB), iii) extreme remote area commu-
work was supported in part by the Fundação para a Ciência e a Tecnologia nications (eRAC), and iv) massive machine type communi-
under Project UID/EEA/50008/2019, in part by the Brazilian National Council
for Research and Development (CNPq) under Grant 309335/2017−5, in part cations (mMTC). URLLC will enable critical services over
by the Fundo de Apoio ao Desenvolvimento das Comunicações (presiden- mobile networks. Hence, it will require overall latency less
tial decree 264/10, November 26, 2010), Republic of Angola, in part by than 1 ms, ultra accurate device positioning, data rates in the
RNP, with resources from MCTIC, through the Centro de Referência em
Radiocomunicações Project of the National Institute of Telecommunications order of 1 Gbps, and 99.999% reliability [8], [9]. In general
(Inatel), Brazil, under Grant 01250.075413/2018-04, and in part by the RNP lines, eMBB shall provide very large bandwidth allocation
with resources from 5G-RANGE BR-EU Project. (Corresponding author: by operating with frequencies in the range of 6–90 GHz,
Joel J. P. C. Rodrigues.)
I. B. F. de Almeida, L. L. Mendes, and M. A. A. da Cruz are support to device-to-device (D2D) communication, massive
with the PPGT, National Institute of Telecommunications, Santa Rita do multiple input multiple output (MIMO) and data rates up
Sapucaí 37540000, Brazil (e-mail: ivobizon@gee.inatel.br; luciano@inatel.br; to 10 Gbps [10]. eRAC will play an important role, for
maurocruzter@gmail.com).
J. J. P. C. Rodrigues is with the PPGT, National Institute of example, in precision agriculture, forest fire prevention, envi-
Telecommunications, Santa Rita do Sapucaí 37540000, Brazil, also with the ronmental monitoring, and precise weather reports [3]. mMTC
Instituto de Telecomunicações, 1049-001 Lisbon, Portugal, and also with presents a great challenge for 5G design due to its contrast-
the PPGEE, Federal University of Piauí, Teresina 64049-550, Brazil (e-mail:
joeljr@ieee.org). ing requirements. However, IoT represents one of the major
Digital Object Identifier 10.1109/COMST.2019.2910817 design stimuli for 5G networks. Support device density in the
1553-877X  c 2019 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.
DE ALMEIDA et al.: 5G WAVEFORMS FOR IoT APPLICATIONS 2555

that pre-commercial trials did not reach the 5G goals.

However, [15] do not cover the details of the IoT applications
and there is no comparison with other technologies regarding
this specific use case. In [16], the authors bring an enlightening
discussion about the applications of MTC, including its main
challenges and promising solutions. In this tutorial, we list
the main non-3GPP technologies that can provide IoT service,
highlighting their advantages and disadvantages. Additionally,
it describes the main waveforms proposed for 5G, discussing
how each new modulation scheme can contribute with the
IoT over 5G scenario. Therefore, this paper builds a bridge
between the current IoT solutions and the future possibilities
for the 5G networks.
The main contributions of the paper are the following:
i. Study the IoT scenarios to identify critical requirements,
followed by a research on the current technologies that
should provide connectivity for IoT;
ii. Presents a discussion on the gaps left by some IoT solu-
tions that 5G networks must cover to unleash the full
potential of the mMTC applications;
iii. A review of the state of the art regarding 5G PHY
waveform candidates.
Fig. 1. Expected scenarios for 5G networks.
The remaining of the paper is organized as follows.
Section II presents the main IoT use-cases and its require-
ments. Section III compares currently proposed techniques to
order of 1 million devices per km2 , very high energy effi- identify the gaps that need to be filled by the 5G mobile
ciency for allowing battery-powered devices to remain active networks. Section IV reviews the state of the art on multi-
for at least 10 years without battery replacement are the carrier modulation techniques that are 5G PHY candidates.
primary challenges [11], [12]. The 5G standardization is cur- Section V discusses the most suitable technique for the IoT
rently a work in progress that will continuously evolve to scenario based on the studies presented in this paper as well
address all the requirements imposed on the future network. as simulation results. Finally, Section VI concludes the paper
On December 2017, 3GPP approved the non-standalone 5G and suggests further research works.
new radio specifications, where the 4G core is used to provide
connectivity management, while the radio access is provided
by the new PHY. This version will allow operators to launch II. I OT S CENARIO AND R EQUIREMENTS
5G enhanced mobile broadband services faster and with less IoT promises to be a revolution in the telecommunications
investments [13]. On July 2018, 3GPP presented the stan- market by providing Internet connectivity to everyday objects.
dalone 5G new radio specifications, with a 5G core network The vast range of applications and services demand several
capable of providing the new 5G services. With the 5G core, requirements. In general, the requirements are the follow-
other 5G scenarios, such as mMTC and URLLC will become ing: (i) low power consumption that allows battery-powered
possible [13]. Currently, 3GPP Release 16 is under develop- devices to remain in service for roughly 10 years; (ii) low
ment, and it will bring new features for the 5G networks, cost devices, which allows a great number of things to be
including the TV and radio broadcast capabilities, networking purchased without an excessive amount of investment; (iii) ser-
slicing, full IoT support, vehicle-to-everything communication, vice availability, for devices to operate anywhere and anytime;
integration with satellite networks and novel radio techniques. (iv) low device maintenance, for allowing general users to
Release 16 will result in the IMT-2020, and is the major input operate; and v) scalability, i.e., the network should cope with
for the complete 5G standardization. Release 16 is expected the rapid growth in device population [4]. Besides the broad
to be ready by December 2019 [14]. scope of requirements behind the IoT scenario, the 5G PHY
This tutorial aims to clarify the requirements that 5G PHY will also deal with high data throughput and low latency appli-
needs to fulfill so that connectivity and quality of service cations. Since all these competing requirements will not be
(QoS) are achieved in IoT use-cases, i.e., in the mMTC simultaneously required by a given application, adapting and
scenario. For further comprehension, this paper maps these tailoring the 5G PHY to support the demands from the user
requirements as PHY characteristics. on the fly is a key feature of the future mobile network. For
Since 5G and IoT are popular research areas, there are this reason, a flexible PHY becomes a vital part of the whole
important tutorials about these technologies. In [15], the network [7].
authors presents a tutorial about the main techniques that Some standards today can partially address the IoT require-
can trigger the 5G new services. Also, they present the ments, e.g., IEEE 802.15.4 focusing on the industrial perspec-
main challenges for implementing these techniques and show tive and Bluetooth focusing on the domestic use-cases [17].
2556 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 21, NO. 3, THIRD QUARTER 2019

More recent technologies, such as low power wide area A. Bluetooth Low Energy
(LPWA) networks, are becoming attractive due to their avail- The Bluetooth technology is commonly found in smart-
ability and low power consumption [17]. Main examples phones, cars, wireless mouses, keyboards and earphones. This
are SigFox and Long Range (LoRa) [18], [19]. However, communication protocol characterizes a personal area network
the ubiquitous status is still hard to reach with these tech- (PAN), and it is designed to provide short distance (approxi-
niques, since they are not so well established as cellular mately 10m) connectivity among devices. However, to expand
networks [20], [21]. Nevertheless, cellular networks are expe- its territory into the IoT scenario, the standard needed to
riencing a huge increase in the signaling traffic due to the large reduce energy consumption and, therefore, be able to operate
number of connected devices [22]. Under these circumstances, with coin-sized batteries [32]. Hence, the Bluetooth Special
the possibility of connecting devices that are roughly synchro- Interest Group (SIG) added a low energy (BLE) configuration
nized by a non-orthogonal waveform decreases the signaling when they released Bluetooth version 4.0 for targeting this
burden and becomes very attractive. market segment [32], [33]. Differently from the conventional
IoT devices are classified as controlling or monitoring Bluetooth, BLE is optimized for transmitting short packets of
devices [23], [24]. IoT use-cases can be divided into domes- data sporadically.
tic, industrial, and mission-critical. Domestic objects focus One of the main differences between conventional Bluetooth
on general users and include commercial products. Even and its low energy configuration lies in the PHY. While the
though commercial products are still scarce, they are gener- conventional has 79 channels with 1 MHz bandwidth, BLE
ally associated with smart houses, enhanced learning, energy presents 40 channels with 2 MHz bandwidth. In both con-
monitoring, personal applications, and e-health [4]. Its most ventional Bluetooth and BLE these RF channels are divided
relevant requirements are low cost, low maintenance, and bat- in two types: advertising and data channels. The advertising
tery life span. In addition, due to the quantified self paradigm, channels are used for device discovery, broadcast, and con-
security and privacy are also a major concern [25]. Industrial nection establishment, whereas data channels are used for
IoT and mission-critical IoT present similar specific require- data transmission between the connected devices [26], [33].
ments. On one hand, industrial IoT is focused on providing Also, both operate in the unlicensed industrial scientific med-
automation solutions for industries and further increasing ical (ISM) band (2.4 GHz). In the conventional Bluetooth,
workers’ security in hazardous factory plants. On the other the modulation technique can vary from Gaussian frequency
hand, mission-critical IoT can be used for monitoring essen- shift keying (GFSK) to four phase shift keying (4-PSK) and
tial services or hazardous locations, e.g., nuclear power plants, 8-PSK, whereas in BLE only GFSK is used. GFSK presents
offshore petroleum extraction platforms, and health-related very low peak-to-average power ratio (PAPR), which translates
applications such as inpatient monitoring. These applications to low energy consumption, since high efficiency power ampli-
demand a higher level of network reliability with packet loss in fiers operating close to the 1 dB compression point can be
the order of 10−9 , latency in the order at 1 ms, and availability employed [34].
of 99.999% of the time [5]. Conventional Bluetooth 4.0 operates in a master-slave con-
figuration, and the network can be configured as a mesh,
point-to-point, or broadcast topology. The network formed
III. N ON 5G T ECHNOLOGIES T HAT A DDRESS with a master device and one or more slaves is called a piconet.
I OT R EQUIREMENTS All the data transfer happens upon the establishment of the
Several communication technologies have emerged for con- piconet. A scatternet is formed by conjunction of one or more
necting IoT devices, some are novel and designed for IoT piconets. As a result, network scalability can be achieved.
use-cases, and some are already well employed in other sce- However, latency might be in the order of a few seconds due
narios. Such technologies are, namely, Bluetooth 5.0 Low to the mesh network configuration. In BLE 4.0, a slave node
Energy [26], IEEE 802.15.4 [27], LoRa [20], Sigfox [28], cannot connect with more than one master, so only star and
WiFi HaLow [29], Narrow band IoT (NB-IoT) [30], [31]. point-to-point topologies are possible, which limits its ability
Regardless, these technologies are not mature enough or do to scale.
not attend to all requirements to be used on a enormous The latest version, called Bluetooth 5, was proposed by the
scale [12], [17]. Bluetooth SIG in 2016 [26]. The focus for improvements in
The PHY of these technologies is discussed, emphasizing this version was the BLE configuration, while conventional
the advantages and drawbacks of each technology. Table I Bluetooth remains roughly the same as previous versions [35].
summarizes the main PHY characteristics of the aforemen- This newer BLE version presents a two-fold increase in trans-
tioned technologies. Such comparison can be constructive for mission rate, from 1 Mbps to 2 Mbps, and the possibility
identifying the technological gaps that 5G networks need to increase range with the introduced coded operation mode
to fulfill for the IoT paradigm to become a reality. 5G that operates at 500 kbps [35], [36]. Despite the improve-
will not completely replace the technologies presented in ments, the BLE 5 still can only operate in star topology, which
this section. Instead, 5G should be an important compo- limits its ability to scale [26]. Nevertheless, efforts by either
nent of the IoT connectivity scenario. In other words, 5G academia and industry have been made to enable a mesh topol-
will complement these technologies in future applications and ogy in BLE and, therefore, increasing its changes to be largely
services. employed in the IoT market [37].
DE ALMEIDA et al.: 5G WAVEFORMS FOR IoT APPLICATIONS 2557

TABLE I
PHY C HARACTERISTICS OF THE M AIN T ECHNOLOGIES T HAT A DDRESS I OT R EQUIREMENTS

B. IEEE 802.15.4 IEEE 802.15.4 has seven different operation modes defined.
IEEE 802.15.4 defines the medium access control (MAC) From IoT perspective, the modes that lead to low power
and PHY layers [38]. The physical layer operates in different consumption are offset quadrature phase shift keying with
ISM bands according to the region where it is deployed. The direct sequence spread spectrum (OQPSK-DSSS), differen-
2.4 GHz band is universal, other frequency bands are, e.g., tial QPSK with chirp spread spectrum (DQPSK-CSS), and
Gaussian frequency shift keying (GFSK) with no spread spec-
868 MHz in Europe and 915 MHz in North America.
trum technique employed. The maximum data rate for these
IEEE 802.15.4 is designed for PANs, and it is mostly
modes are respectively 250 kbps, 1 Mbps, and 100 kbps [38].
employed in embedded systems for agricultural, environmen-
ZigBee operates with low transmission power and duty cycle.
tal, and industrial monitoring [39]. Differently from the IEEE
However, to the best of authors’ knowledge, a practical esti-
802.11 family, IEEE 802.15.4 is not focused on high data mation of battery consumption in these modes is not available
rates, and differently from Bluetooth, it is not focused on in the literature.
connecting personal devices. It was proposed to be a low cost IEEE 802.15.4 operates in the unlicensed ISM band
and energy efficient wireless protocol for resource-constrained and robustness to interference becomes a challenging issue.
sensor networks in large geographical areas. However, the spread spectrum techniques used cause low
For example, ZigBee employed IEEE 802.15.4 successfully. interference in other systems, and also are more immune to
The major advantage over the others is that ZigBee is able to interference caused by others [41].
operate in a multi-hop configuration. Hence, allowing network The addition of the Internet protocol (IP) to IEEE 802.15.4
scalability [40]. However, as aforementioned, latency may not networks can be accomplished by the 6LoWPAN protocol,
achieve the desired values due to the network topology. which is a short term for IPv6 over low power wireless
2558 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 21, NO. 3, THIRD QUARTER 2019

personal networks. 6LoWPAN is a mid-layer protocol placed quality of service. Sigfox has created a communication pro-
between the network and MAC layer, i.e., under IPv6 protocol tocol based entirely on IoT requirements, and it is the one
and the IEEE 802.15.4. The 6LoWPAN protocol compresses that has given one step further to realize the envisioned IoT
the overhead necessary to transmit IP packets, thus allowing paradigm.
an efficient transmission and energy savings [39], [42].
E. WiFi HaLow
C. LoRa WiFi HaLow is part of the WiFi family and it is specified
by the standard IEEE 802.11ah. Similarly to the aforemen-
LoRa is classified as Low Power Wide Area Network tioned technologies, it offers a long range, low power, and
(LPWAN) technology developed by the Semtech [18], [43]. low rate solution to connect a huge number of devices. In
It was proposed in 2015, so it is more recent than ZigBee contrast, IEEE 802.11ah presents a more complex PHY. It
and Bluetooth. Similarly to them, it operates at a low rate supports modulation indexes up to 256-QAM that are transmit-
and offers low power consumption. However, differently from ted using orthogonal frequency division multiplexing (OFDM),
the previously mentioned technologies, LoRa (and LPWANs and multiple input multiple output (MIMO) is present in both
in general) focuses on wide area coverage. LPWANs have downlink and uplink [48], [49]. Therefore, the maximum data
coverage much larger than PANs but smaller than cellular rate is 78 Mbps when the channel bandwidth is set to 16 MHz
networks. The LoRa coverage is around tens of kilome- and the modulation index is set to 256. Since it operates at
ters [19], [44]. This feature makes LoRa an interesting option unlicensed frequencies below 1 GHz, the coverage ranges from
for applications related to smart cities and precision agri- 100 m to 1 km [50].
culture. Applications that demand higher network throughput
might be set aside, since LoRa provides, at maximum, only F. Narrow Band IoT
50 kbps [19]. Furthermore, practical results have shown that
a few million devices can be attended by a single base sta- Different from all mentioned techniques, Narrow Band IoT
tion [44]. However, the coverage area decreases exponentially (NB-IoT) is the only solution based on cellular networks. For
as the number of devices increase. this reason, it presents a large coverage area and a higher level
At the physical layer, LoRa uses a proprietary modula- of reliability, due to the dedicated spectrum band. NB-IoT is
tion technique based on Chirp Spread Spectrum (CSS) in one of the systems that are called pre-5G or 4.5G. For instance,
conjunction with GFSK. The CSS allows a simpler time it is totally dedicated to connect IoT devices on a large
and frequency synchronization, so less expensive components scale, and it represents a major step towards the conjunction
can be employed. In addition, the CSS based modulation between cellular technologies and IoT. It was proposed in LTE
presents inherent robustness to doubly dispersive chan- Release 13 by the 3rd Generation Partnership Project (3GPP)
nels [45]. Another strong point of LoRa’s PHY is that the in 2015 [51]. NB-IoT can be employed in three different con-
modulation schemes used present constant envelope. Hence, figurations: (i) stand-alone, where it can transmit data over one
high-efficiency power amplifiers can be used, and energy con- GSM channel with 200 kHz bandwidth, (ii) guard-band, where
sumption is optimized. However, the 50 kbps rate can only be carriers are placed inwards the LTE guard band occupying the
achieved with GFSK. The maximum data rate with the CSS bandwidth of one Resource Block (RB) (180 kHz), and (iii) in-
modulation is 11 kbps [46]. band, where one RB is assigned for the IoT device to operate
within the LTE bandwidth [52]. As NB-IoT stand-alone mode
will operate in the GSM bandwidth, an efficient re-farming
D. Sigfox of GSM’s band is possible, and this ensures coexistence with
Sigfox is also classified as a LPWAN and it operates in an legacy and present systems [30].
ultra narrow band configuration. In contrast with LoRa, Sigfox NB-IoT proposes improved indoor coverage and supports
is a proprietary scheme and, therefore, with closed documen- up to 52547 devices per cell [53], and battery life span of
tation. Consequently, very little information is available to the approximately 10 years with a 5 Wh battery if the devices
community. Nonetheless, some information about its PHY can transmit every 2 hours a 200 bytes message with 33 dBm
be found in [28], [47]. output power [51]. As an illustration, a 5 Wh battery means
The operating frequency band is 868 MHz in Europe and that a device should operate with approximately 6 CR2032
915 MHz in the United States. The multiple access tech- batteries [54]. The latency can get up to 10s [55]. Although 5G
nique is asynchronous, so there is no energy expenditure with requirements are more strict, NB-IoT is also one step forward
synchronization between device and RBS. For establishing a in providing connectivity for IoT devices.
connection, the device transmits three consecutive messages.
Each message is transmitted over different time and frequency G. Limitations of the Available IoT Connectivity
slots that are randomly chosen. This is an example of paradigm Technologies
change in the multiple access technique, and it has made The 5G cellular network will not completely replace the
Sigfox an interesting solution for IoT connectivity. Another technologies presented in this section. Instead, they should be
key point is that the reception is based on a cooperative an important part of the IoT connectivity scenario. In other
scheme. Any RBS can receive messages from the transmit- words, 5G will supplement these technologies for making real
ting devices. This reception diversity translates in to a better the future applications and services.
DE ALMEIDA et al.: 5G WAVEFORMS FOR IoT APPLICATIONS 2559

TABLE II
WAVEFORM ’ S C HARACTERISTICS AND C ORRESPONDENT I OT B ENEFITS networks, some OFDM characteristics hinder the use of this
waveform in future mobile networks [58]. Consequently, new
multicarrier techniques, as well as improvements to OFDM,
are being recently studied.
So far, CP-OFDM will remain active in 5G non-standalone
systems, which will be operating in the current 4G frequency
bands [59]. This part of the 5G will enable the EMBB case,
giving users a glimpse of what is yet to come in the future
standalone 5G. OFDM lacks performance to be deployed as
the only 5G radio access technology that can address all the
5G scenarios. The research trend related to other scenarios
is to leave behind the strict synchronism and find a trade-off
between interference, complexity, and cost [1], [7], [60].
In the OFDM system, parallel streams of data are transmit-
ted by orthogonal subcarriers, so no intercarrier interference
(ICI) is observed. Therefore, if proper synchronism is
Technologies that are already available present interesting achieved, a user remains orthogonal to others. The modula-
solutions for IoT connectivity. However, they present draw- tion procedure is carried out through an inverse fast Fourier
backs that hinder IoT full connectivity. In particular, Bluetooth transform (IFFT) at the transmitter side. Demodulation is then
and IEEE 802.15.4 employ modulation techniques that are accomplished using the FFT. For this reason, zero forcing
known for the low energy consumption a fairly good data rate equalization can be easily done in the frequency domain since
for IoT standards, but they lack in coverage area. Wifi HaLow it will be a simple scalar inversion [61]. Hence, OFDM is a
presents a complex PHY structure compared to its competi- simple choice when compared to recently proposed waveforms
tors, which enables 1 km coverage area and the highest data for 5G. Figure 2 illustrates the OFDM transceiver diagram.
rate. However, the OFDM employment in the PHY demands To protect the OFDM symbol against multipath propaga-
high energy consumption due to the strict synchronism needed tion impairments, a copy of part of the end of the symbol
and the high PAPR. Likewise, NB-IoT suffers from the same is placed at its beginning. This copy is called a cyclic prefix
OFDM shortcomings. LPWANs have presented an attractive (CP). By doing so, the channel does not ruin the orthogo-
approach, trading data rate for coverage comparable to cellular nality among subcarriers, and a simple equalization procedure
networks and low energy consumption. These technologies fall can be accomplished. Even though CP protects the OFDM
short in terms of reliability since they operate in the unlicensed symbol, it also causes a reduction in the spectral efficiency
ISM bands. since it does not contain useful user information. In the per-
Given these points, one can notice that some 5G should spective of short packet transmission, i.e., low latency, the
provide enough coverage for addressing IoT smart farming CP means low efficiency in resource utilization since the size
use-cases while employing techniques that are energy effi- of an OFDM symbol could be the same as the CP size [62].
cient. In addition, the scalability and reliability of the licensed Nonetheless, the CP size is proportional to the channel’s delay
spectrum is necessary. profile, and in applications such as smart farming, where the
distance between devices and base stations is tens of kilo-
IV. WAVEFORM C ANDIDATES FOR 5G PHY meters, the CP can become larger than the useful information.
This section presents an overview of the main candidates Figure 3 exemplifies a hypothetical scenario where the OFDM
for 5G PHY, highlighting its drawbacks and improvements symbol is the same length as the CP.
over the current 4G technique. Comparisons will be performed Interference among users, i.e., ICI, is not observed if orthog-
in terms of OOB emission, energy efficiency, complexity, onality is maintained through strict synchronism. Any user
efficiency considering short packet transmission, and latency. using different subcarriers remains orthogonal to others if
Equally, the study compares the PHY from the technologies proper synchronism is achieved. If synchronism is not sat-
presented in Section III with the 5G candidates. Table II sum- isfied, devices will suffer from self interference, as well
marizes the desired characteristics of the 5G PHY waveform as interference with other users. Thus, from IoT’s perspec-
and their relation with the IoT connectivity scenario. tive, lots of energy will be spent until the synchronization
process is finished, and 10 years battery life span will be
impracticable. In addition, OFDM shows high peak-to-average
A. Orthogonal Frequency Division Multiplexing ratio (PAPR), which means that low-efficiency amplifiers
OFDM is the current waveform employed in LTE PHY. must be used [63]. From IoT’s perspective, this leads to
It has been extensively studied and widely deployed in high battery consumption. However, when the synchronization
wired and wireless communication systems, e.g., IEEE process is finished a simple zero-forcing frequency domain
802.11n (WiFi) [56] and Asynchronous Digital Subscriber equalizer can be used for canceling the multipath channel
Line (ADSL) [57]. Indeed, OFDM is a robust multicarrier effect [61].
technique that has become attractive because of its low com- Furthermore, data symbols are shaped with a rectangular
plexity. However, due to the wide range of requirements of 5G filter in the time domain, so subcarriers will be sinc shaped in
2560 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 21, NO. 3, THIRD QUARTER 2019

Fig. 2. Illustration of the OFDM transceiver diagram.

This is a key attribute of FBMC, since relaxing orthogo-

nality constrains leads to better time-frequency localization
and diminishes the interference caused by transmissions from
asynchronous users [67].
In order to solve this neighboring ICI problem, OQAM
comes into play. The imaginary and real parts of the data sym-
bols are shifted by T/2, where T represents the data symbol
time, and a π/2 phase rotation between adjacent subcarriers
Fig. 3. Low efficiency in short packet transmission due to CP in the OFDM
frame structure. is also introduced. Thus, the subcarriers with even index carry
the real part, and odd indexed subcarriers carry the imaginary
part of the data symbol. No interference is observed in time or
the frequency domain. As a result, undesired levels of out-of- frequency domain, and the maximum spectral efficiency can
band (OOB) emission are observed [64]. be achieved. This arrangement is called FBMC-OQAM [69].
A possible implementation of the FBMC-OQAM
transceiver structure is shown in Fig. 4. For the FBMC
B. Filter Bank Multicarrier notation, let K represent the total number of subcarriers,
FBMC was first proposed by Saltzberg [65] and Chang [66], dk ,m the complex QAM data symbol that is transmitted
in 1967, and it was rediscovered as an alternative PHY for over the kth subcarrier in the mth time-slot and gk ,m [n] the
future radio access by Bellanger [67]. It is a multicarrier synthesis/analysis filters. The discrete-time FBMC-OQAM
modulation technique where orthogonality between all the sub- transmit signal is obtained by
carriers is not strictly maintained. There are different schemes
+∞
 K
 −1
to construct the pair modulator demodulator. In this paper, it is   (I)
x [n] = R dk ,m gk ,m [n]
presented the FBMC-STM (Staggered Modulated Multi-tone)
m=−∞ k =0
with the use of offset QAM (OQAM). This scheme uses a pro-
+∞
 K −1
totype digital FIR (Finite Impulse Response) filter to obtain   (Q)
+j I dk ,m gk ,m [n], (1)
the filter bank structure. Each filter is shifted in frequency,
m=−∞ k =0
allowing subcarriers to occupy the available bandwidth. Half
Nyquist prototype filters with good frequency localization are (I) (Q)
where gk ,m [n] and gk ,m [n] are respectively given by
preferred since they do not cause ISI at the sampling instant
 
and minimize the intrinsic ICI between subcarriers [68], [69]. (I) k π
A pulse with good time-frequency localization means that it gk ,m [n] = p[n − mK ] exp j 2π n + j k (2)
K 2
is well contained both in time and frequency domains.      
(Q) 1 k π
Figure 5 shows a set of subcarriers in the frequency domain, gk ,m [n] = p n − m + K exp j 2π n + j k , (3)
2 K 2
and it is possible to see that ICI only happens between neigh-
boring subcarriers due to good frequency localization of the where p[n] is the prototype filter impulse response. The
prototype filter. Hence, orthogonality between neighboring received signal passes through the analysis filter bank, and
subcarriers must be ensured. For the sake of comparison, the estimated data symbols d̂k ,m are obtained. The complex
OFDM must ensure orthogonality among all subcarriers. conjugate operation is represented by (·)∗ .
DE ALMEIDA et al.: 5G WAVEFORMS FOR IoT APPLICATIONS 2561

Fig. 4. Illustration of the FBMC-OQAM transceiver diagram.

represented by a matrix multiplication yielding to

x = Ad, (4)

where x represents the transmit vector before the CP addi-

tion, d represents the complex-valued data symbol vector with
N elements, and A represents the transmit matrix with size
N × N. The transmit matrix, or modulation matrix can be
obtained by circularly shifting the prototype filter to the kth
subcarrier and to the mth subsymbol. Thus, A is composed by
N versions of the prototype filter which yields to

A = g0,0 g1,0 · · · gK −1,0 · · · g0,M −1 · · · gK −1,M −1 , (5)

Fig. 5. Set of FBMC subcarriers in frequency domain emphasizing the
interference between neighboring subcarriers.
where gk ,m represents the vector which contains samples from
the circularly shifted prototype filter. Figure 7 shows the abso-
lute value of the GFDM modulation matrix. As one can see a
The main advantages of FBMC-OQAM schemes are spec- GFDM frame carries N = KM complex valued data symbols,
tral efficiency and low OOB emission. The spectral efficiency where K represents the total number of subcarriers and M the
can be considered 1 if a long stream of data is to be trans- total number of subsymbols. The GFDM transceiver diagram
mitted. However, FBMC-OQAM employs a linear filtering is described by Fig. 6 [62].
procedure for shaping data symbols. Consequently, for short In principle, GFDM subcarriers are not orthogonal to each
packet transmission, the ramp-up and ramp-down of the fil- other. As a consequence, they may exhibit intrinsic ICI and
ters leads to lower spectral efficiency since during that time ISI depending on the prototype filter and on the demodulation
no useful data is transmitted. Hence, FBMC-OQAM might method [71]. At the receiver side, transmitted data symbols are
not be the best option for sporadic sensor data transmis- retrieved from the received vector also by a matrix operation
sion that transmits small packets of collected data. On the given by
other hand, the linear filtering also leads to very low OOB
emissions, so it leads to lower interference levels even when d̂ = Byeq , (6)
roughly synchronized devices are sharing subcarriers in oper-
where d̂ represents the estimated data symbols, B the demod-
ation. Therefore, devices can save energy be completing a
ulation matrix and yeq the equalized received vector.
simpler synchronization procedure.
Three demodulation techniques are described in the litera-
ture. The zero forcing solution is able to cancel intrinsic ICI
and ISI from non-orthogonal prototype filters. However, it gen-
C. Generalized Frequency Division Multiplexing erates the noise enhancement effect due to the non-contained
GFDM is somewhat similar to FBMC in the sense that it is frequency response of this demodulation matrix. The matched
also based in filter bank theory, and it is built upon a prototype filtering solution is able to cancel intrinsic ISI if the proto-
filter response. However, GFDM implements circular filtering type filter is half-Nyquist, and it maximizes the signal to noise
to shape the data symbols [70]. As the waveform presents a ratio at the sampling instants. However, it is not able to can-
circular behavior, CP can be used to protect the multicarrier cel intrinsic ICI. The minimum mean square solution is able
symbol against multipath propagation impairments. to cancel intrinsic interferences while keeping a low level of
At the GFDM transmitter, data symbols modulate the fil- noise enhancement since it takes into consideration the noise
ters across a time-frequency lattice. This operation can be statistics. Additionally, it takes into consideration the channel
2562 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 21, NO. 3, THIRD QUARTER 2019

Fig. 6. Illustration of the GFDM transceiver diagram.

mission-critical IoT applications where latency is an important

metric. Moreover, it can be employed in coexistence with 4G
signals [76].

D. Universal Filtered Multicarrier and Filtered OFDM

UFMC and F-OFDM can be thought as a compromise
between OFDM and FBMC. In particular, FBMC uses a
well located filter to filter each subcarrier, whereas OFDM
implicitly applies a rectangular to filter each subcarrier. It is
important to point out that although rectangular filters have
good localization in time domain, they present poor localiza-
tion in frequency domain, i.e., strong OOB emission. UFMC
and F-OFDM apply the filtering over a block of subcarriers,
i.e., subband wise. The size of this block can vary or can be
set to a specific value, e.g., each block is composed by 12
subcarriers.
Differently from UFMC, F-OFDM filters the entire band
Fig. 7. Absolute value of the modulation matrix for M = 7 subsymbols, occupied by a OFDM signal. Therefore, the main difference
K = 5 subcarriers and Dirichlet pulse as prototype filter. between UFMC and F-OFDM lies in the size of the filter
band. UFMC employs a filter with a tail size comparable to
the channel’s delay profile. It is important to point out that
statistics, so no previous equalization is necessary for esti-
the UFMC symbols do not cause ISI due to the filter tails.
mating the transmitted data symbols from the received signal.
Therefore, the filter tail serves the same function as the CP in
These demodulation matrices are respectively given by
OFDM systems.
BZF = A−1 (7) On the other hand, F-OFDM employs filters that overlap
BMF = AH (8) in time domain, so ISI among multicarrier symbols occurs
 −1 and a CP is necessary for avoiding harmful ISI. This filters
BMMSE = Rw + AH HH HA AH HH , (9) lead to lower OOB emissions when compared to UFMC [72].
However, CP must be inserted for ensuring interference free
where (·)H represents the Hermitian conjugate transpose oper- transmission [77]. In both systems, the OOB emission is atten-
ation and Rw = σ 2 IN represents the covariance matrix of the uated when compared to OFDM. However, not as much as
noise vector whose variance is represented by σ 2 . IN is an FBMC since it uses subcarrier filtering.
identity matrix of size N × N. Figure 8 illustrates the UFMC transceiver block diagram.
OQAM can be used in combination with GFDM for At the transmitter, a set of complex-valued data symbols rep-
dealing with intrinsic ICI, and further enhance GFDM’s flex- resented by sk are divided into B blocks, and inserted at a
ibility [73]. Even though OOB emission is reduced when N-point IFFT. Then, each subband is independently filtered.
compared to OFDM due to more a contained filter in frequency The length of the filter can be chosen based on a trade-off, i.e.,
domain, abrupt transitions between GFDM frames lead to longer filters will lead to lower levels of OOB emission and
elevated OOB emissions. However, as demonstrated in [74] robustness to frequency misalignment. On the other hand, short
windowing can be used to achieve lower OOB emissions. filters will lead to shorter multicarrier symbols, and therefore
Furthermore, the same low OOB emission of FBMC-OQAM smaller latency [78].
can be achieved with GFDM by zero padding the prototype
filter as shown in [75]. Therefore, GFDM presents the same
benefits of FBMC-OQAM when it comes to low interference. V. WAVEFORM E VALUATION FOR THE I OT S CENARIO
Low latency scenarios can benefit from the GFDM frame The possibility of operating with relaxed orthogonality with-
structure since different subsymbols can be addressed to dif- out major performance degradation is the main advantage that
ferent devices. Hence, it becomes an attractive alternative for FBMC brings to IoT scenario. For instance, this characteristic
DE ALMEIDA et al.: 5G WAVEFORMS FOR IoT APPLICATIONS 2563

Fig. 8. Illustration of the UFMC transceiver diagram [72].

allows rough synchronized users/devices to transmit without

causing excessive interference to others [79]. However, FBMC
falls short in terms of efficiency in the transmission of short
data packets. Due to the linear filtering process involved, there
is a ramp-up and ramp-down time of the filters, so in a small
period of time it is not transmitting any useful data, and the
transmission efficiency is reduced. If packets occupy more or
less the same time as the ramp-up plus ramp-down, efficiency
falls down to around 50% [80]. Since the FBMC waveform
is generated by a linear filtering procedure, there is no reason Fig. 9. GFDM symbol structure to achieve efficient low latency transmis-
sions.
to put a CP at the beginning of a symbol for protecting the
transmit signal against interference caused multipath propaga-
tion. Therefore, this technique shows a more efficient use of
time resources than OFDM if a long of stream data is trans-
mitted [81]. It is also important to point out that due to the
low out of band emission, FBMC suits the requirements to
occupy spectrum holes, or television whites spaces (TVWS).
Thus, allowing an efficient allocation of the fragmented spec-
trum [67]. Another drawback, unfortunately common to all the
candidates, is the high PAPR [58].
Conversely, the GFDM circular filtering needs the CP at the
beginning of the GFDM symbol to protect it against multipath
propagation impairments. However, the CP that would protect
K data symbols in OFDM will protect KM data symbols in
GFDM. Meaning that GFDM presents a more efficient way
to allocate time-frequency resources than FBMC considering
short packet transmission, and OFDM [74]. GFDM shows a
great deal of flexibility by being able to change its parameters Fig. 10. Power spectral density comparison between 5G waveform candi-
according to the requirements. The core of 5G’s flexibility dates.
could be a software-defined waveform that will change its
configurations depending on a software defined base station. are possible to be obtained by GFDM. Since each GFDM
GFDM waveform can be deployed as a framework to obtain subsymbol corresponds to one OFDM symbol, the subsymbol
the performance of other multicarrier techniques. For instance, can be smaller than an OFDM symbol, and still have the same
GFDM can be used to generate OFDM and FBMC variations spectral efficiency. Hence, as the latency at the physical layer
by changing its configuration parameters, such as the number is dependent on the frame size, the system’s latency can be
of subsymbols (M), subcarriers (K), and prototype filter. As an improved [74]. Fig. 9 shows a GFDM symbol configuration
example, GFDM’s modulation matrix (Fig. 7) has a circular where better spectral efficiency and very low latency can be
filtering behavior. If the modulation matrix is changed to have achieved when compared to OFDM.
a linear filtering behavior, the performance is then compara- Fig. 10 shows the power spectral density comparison
ble to FBMC. Therefore, all the benefits that FBMC shows between OFDM, FBMC, GFDM, and UFMC. FBMC presents
2564 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 21, NO. 3, THIRD QUARTER 2019

TABLE III
C OMPARISON A MONG WAVEFORM C ANDIDATES FOR 5G

the lowest OOB emission, and therefore it is the candidate for can change according to the user’s need, i.e., if the user needs
roughly synchronized users since it will cause the least amount several RBS, then, the filtering process occurs within this allo-
of interference in adjacent bands, i.e., other users. To compare cated band. As a result, the user’s computational processing
the waveforms, we have used the root raised cosine filter with burden is lightened. Therefore, energy consumption can be
β = 0.1, M = 9 subsymbols, and K = 128 subcarriers for reduced at the user’s terminal. In addition, the reduced filter
GFDM. For FBMC, the PHYDYAS [68] filter was used as a length leads to a lower latency when more than one RB is
prototype filter, and the number of subcarriers was set to 128 allocated.
as well. For Windowed GFDM, a cosine window was applied. Given these points, Table III presents a summary of the main
OFDM and UFMC also have 128 subcarriers. characteristics of the waveform candidates [77], [85], [86]. The
In UFMC, a post-sub-band filtering aims to reduce the row that compares the OOB emission of the waveforms can
pronounced OOB emission of OFDM [82]. The number of be further comprehended with the help of Fig. 10.
subcarriers within a sub-band can be chosen according to the
requirements. The main advantage upon other candidates is the
low complexity and easy system migration, since it is based VI. F INAL R EMARKS AND F UTURE R ESEARCH
on the current LTE radio access technology. As an example, The availability and reliability required by some critical
for LTE compatibility it can be set equivalently to the smallest IoT applications go against the desired 10 years battery life
possible resource allocation of LTE’s resource grid, a resource span of other more relaxed IoT applications. Thus, defining
block (RB), which is composed by 12 OFDM subcarriers [83]. an approach for the PHY that fits all the IoT scope becomes
When compared to FBMC in terms of latency, UFMC shows challenging. This paper has shown that recent technologies
advantages. Since UFMC deploys the filtering operation in a present attractive solutions to address the IoT requirements.
group of subcarriers, the filter is broader in frequency, and As shown, these technologies will not be totally replaced by
shorter in time. As a result, the block size is shorter, and the 5G cellular network since some of the current technolo-
latency is reduced [64]. In comparison to GFDM, UFMC does gies employ drastically different PHY techniques from the 5G
not show intrinsic interference within one sub-band allowing PHY candidates. Moreover, this paper presented an overview
a simpler receiver design. In general, the UFMC transceiver of 5G PHY capabilities to enable the IoT connectivity through
design is less costly in terms of computational complexity than the future 5G mobile network. The range of requirements that
GFDM [72]. IoT scenario imposes goes, for example, against the require-
An improvement to UFMC has been recently proposed ments of high throughput scenarios. Consequently, flexibility
in [84]. In the proposed scheme, the filtered sub-band size at the physical layer is a vital characteristic to address such
DE ALMEIDA et al.: 5G WAVEFORMS FOR IoT APPLICATIONS 2565

diversity. In such a manner, IoT power constrained devices [11] Ericsson. 5G Radio Access—Capabilities and Technologies. Accessed:
and high data consuming smartphones can operate under the Mar. 6, 2018. [Online]. Available: https://www.ericsson.com/assets/
local/publications/white-papers/wp-5g.pdf
same network. [12] M. R. Palattella et al., “Internet of Things in the 5G era: Enablers,
Alternative PHY techniques for 5G networks have been architecture, and business models,” IEEE J. Sel. Areas Commun., vol. 34,
extensively studied in recent literature. UFMC presents an no. 3, pp. 510–527, Mar. 2016.
[13] (2019). 3rd Generation Partnership Project, Release 15. [Online].
attractive ease evolution from the current system and also Available: http://www.3gpp.org/release-15
some benefits regarding roughly synchronized devices. FBMC [14] (2019). 3rd Generation Partnership Project, Release 16. [Online].
shows interesting attributes regarding connectivity of rough Available: http://www.3gpp.org/release-16
[15] M. Shafi et al., “5G: A tutorial overview of standards, trials, challenges,
synchronized devices allowing low battery consumption. Also, deployment, and practice,” IEEE J. Sel. Areas Commun., vol. 35, no. 6,
FBMC’s low out of band emission allows operation in frag- pp. 1201–1221, Jun. 2017.
mented spectrum. GFDM’s main advantages are flexibility, [16] Y. Cao, T. Jiang, and Z. Han, “A survey of emerging M2M systems:
Context, task, and objective,” IEEE Internet Things J., vol. 3, no. 6,
the most efficient use of the CP in short packet transmis- pp. 1246–1258, Dec. 2016.
sions, and low latency. In addition, if prototype filters with [17] S. Andreev et al., “Understanding the IoT connectivity landscape: A
good frequency localization are used, the synchronization exi- contemporary M2M radio technology roadmap,” IEEE Commun. Mag.,
gency is lighted which translates to battery savings. GFDM vol. 53, no. 9, pp. 32–40, Sep. 2015.
[18] J. de Carvalho Silva, J. J. P. C. Rodrigues, A. M. Alberti, P. Solic, and
can also present the benefits of other multicarrier techniques A. L. L. Aquino, “LoRaWAN—A low power WAN protocol for Internet
at the expense of transmitter/receiver complexity increase. of Things: A review and opportunities,” in Proc. 2nd Int. Multidiscipl.
As shown, the requirements for different 5G scenarios go Conf. Comput. Energy Sci. (SpliTech), Jul. 2017, pp. 1–6.
[19] U. Raza, P. Kulkarni, and M. Sooriyabandara, “Low power wide area
against each other in some cases, so flexibility at the PHY networks: An overview,” IEEE Commun. Surveys Tuts., vol. 19, no. 2,
becomes essential. Accordingly, finding approaches to increase pp. 855–873, 2nd Quart., 2017.
energy efficiency and to diminish PAPR in multicarrier wave- [20] L. Vangelista, A. Zanella, and M. Zorzi, “Long-range IoT technologies:
The dawn of LoRaTM ,” in Proc. 1st Int. Conf. FABULOUS, Ohrid,
forms are important requirements that need to be satisfied North Macedonia, Sep. 2015, pp. 51–58.
for successfully employ 5G networks as the main gateway [21] O. Georgiou and U. Raza, “Low power wide area network analysis: Can
for IoT applications. For instance, predistortion techniques LoRa scale?” IEEE Wireless Commun. Lett., vol. 6, no. 2, pp. 162–165,
Apr. 2017.
that aim to increase the linearity of power amplifiers lead [22] A. Biral, M. Centenaro, A. Zanella, L. Vangelista, and M. Zorzi, “The
to energy savings. Likewise, the complexity for transmitting challenges of M2M massive access in wireless cellular networks,” Digit.
and especially receiving such complex waveforms needs to Commun. Netw., vol. 1, no. 1, pp. 1–19, 2015.
[23] L. D. Xu, W. He, and S. Li, “Internet of Things in industries: A survey,”
be addressed for simple 5G transceiver design. To elaborate IEEE Trans. Ind. Informat., vol. 10, no. 4, pp. 2233–2243, Nov. 2014.
a flexible PHY that is able to modify itself for optimizing [24] Q. Zhang and F. H. P. Fitzek, “Mission critical IoT communication
predetermined requirements remains an open issue. Research in 5G,” in Proc. 1st Int. Conf. FABULOUS, Ohrid, North Macedonia,
Sep. 2015, pp. 35–41.
in real-life IoT environments should be conducted to con- [25] M. Swan, “The quantified self: Fundamental disruption in big data sci-
firm the best waveform choices for 5G. Thereupon, IoT ence and biological discovery,” Big Data, vol. 1, no. 2, pp. 85–99,
objects should be connected through these novel waveforms, 2013.
how and which one will be determined in the future 5G [26] Bluetooth Core Specification Version 5.0, Bluetooth SIG, Kirkland, WA,
USA, 2016.
standard. [27] G. Lu, B. Krishnamachari, and C. S. Raghavendra, “Performance eval-
uation of the IEEE 802.15.4 MAC for low-rate low-power wireless
networks,” in Proc. IEEE Int. Conf. Perform. Comput. Commun., 2004,
pp. 701–706.
R EFERENCES [28] Sigfox. Radio Technology Keypoints. Accessed: Feb. 2019. [Online].
Available: https://www.sigfox.com/en/sigfox-iot-radio-technology
[1] M. Agiwal, A. Roy, and N. Saxena, “Next generation 5G wireless
[29] Wi-FI Alliance Introduces Low Power, Long Range Wi-Fi HaLow.
networks: A comprehensive survey,” IEEE Commun. Surveys Tuts.,
Accessed: Feb. 2019. [Online]. Available: https://www.wi-fi.org/news-
vol. 18, no. 3, pp. 1617–1655, 3rd Quart., 2016.
events/newsroom/wi-fi-alliance-introduces-low-power-long-range-wi-fi-
[2] J. G. Andrews et al., “What will 5G be?” IEEE J. Sel. Areas Commun., halow
vol. 32, no. 6, pp. 1065–1082, Jun. 2014.
[30] R. Ratasuk, B. Vejlgaard, N. Mangalvedhe, and A. Ghosh, “NB-IoT
[3] P. Sethi and S. R. Sarangi, “Internet of Things: Architectures, proto- system for M2M communication,” in Proc. IEEE Wireless Commun.
cols, and applications,” J. Elect. Comput. Eng., vol. 2017, Jan. 2017, Netw. Conf., Apr. 2016, pp. 1–5.
Art. no. 9324035. doi: 10.1155/2017/9324035. [31] D. Flore, 3GPP Standards for the Internet-of-Things, vol. 25, GSMA
[4] L. Atzori, A. Iera, and G. Morabito, “The Internet of Things: A survey,” MIoT, London, U.K., 2016.
Comput. Netw., vol. 54, no. 15, pp. 2787–2805, 2010. [32] K.-H. Chang, “Bluetooth: A viable solution for IoT? [Industry perspec-
[5] “5G systems,” Stockholm, Sweden, Ericsson, White Paper, 2017. tives],” IEEE Wireless Commun., vol. 21, no. 6, pp. 6–7, Dec. 2014.
[6] (2017). 3rd Generation Partnership Project, Release 14. [Online]. [33] Bluetooth Core Specification Version 4.0, Bluetooth SIG, Kirkland, WA,
Available: http://www.3gpp.org/release-14 USA, 2010.
[7] G. Wunder et al., “5GNOW: Non-orthogonal, asynchronous waveforms [34] J. M. Rabaey et al., “PicoRadios for wireless sensor networks: The next
for future mobile applications,” IEEE Commun. Mag., vol. 52, no. 2, challenge in ultra-low power design,” in IEEE Int. Solid-State Circuits
pp. 97–105, Feb. 2014. Conf. Dig. Tech. Papers, vol. 1, Feb. 2002, pp. 200–201.
[8] G. P. Fettweis, “The tactile Internet: Applications and challenges,” IEEE [35] M. Collotta, G. Pau, T. Talty, and O. K. Tonguz, “Bluetooth 5: A con-
Veh. Technol. Mag., vol. 9, no. 1, pp. 64–70, Mar. 2014. crete step forward toward the IoT,” IEEE Commun. Mag., vol. 56, no. 7,
[9] S. Zhang, X. Xu, Y. Wu, and L. Lu, “5G: Towards energy-efficient, low- pp. 125–131, Jul. 2018.
latency and high-reliable communications networks,” in Proc. IEEE Int. [36] H. Snellman, M. Savolainen, J. Knaappila, and P. Rahikkala,
Conf. Commun. Syst., Nov. 2014, pp. 197–201. “Bluetooth 5, refined for the IoT,” Austin, TX, USA, Silicon Labs, White
[10] Y. Niu, Y. Li, D. Jin, L. Su, and A. V. Vasilakos, “A survey of millimeter Paper, 2017, pp. 1–12.
wave communications (mmWave) for 5G: Opportunities and challenges,” [37] S. M. Darroudi and C. Gomez, “Bluetooth low energy mesh networks:
Wireless Netw., vol. 21, no. 8, pp. 2657–2676, Nov. 2015. [Online]. A survey,” Sensors, vol. 17, no. 7, p. E1467, 2017. [Online]. Available:
Available: https://doi.org/10.1007/s11276-015-0942-z http://www.mdpi.com/1424-8220/17/7/1467
2566 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 21, NO. 3, THIRD QUARTER 2019

[38] IEEE Computer Society LAN/MAN Standards Committee, IEEE [63] M. Engels, Wireless OFDM Systems: How to Make Them Work?
Standard for Local and Metropolitan Area Networks–Part 15.4: New York, NY, USA: Springer, 2002.
Low-Rate Wireless Personal Area Networks (LR-WPANs), IEEE [64] F. Schaich and T. Wild, “Waveform contenders for 5G—OFDM vs.
Standard 802.15.4-2011, 2011. FBMC vs. UFMC,” in Proc. 6th Int. Symp. Commun. Control Signal
[39] D. Culler and S. Chakrabarti, “6LoWPAN: Incorporating IEEE 802.15.4 Process. (ISCCSP), May 2014, pp. 457–460.
into the IP architecture,” San Diego, CA, USA, IPSO Alliance, White [65] B. Saltzberg, “Performance of an efficient parallel data transmis-
Paper, 2009. sion system,” IEEE Trans. Commun. Technol., vol. COM-15, no. 6,
[40] P. Castro, J. L. Afonso, and J. A. Afonso, “A low-cost ZigBee-based pp. 805–811, Dec. 1967.
wireless industrial automation system,” in Proc. CONTROLO, 2017, [66] R. Chang, “High-speed multichannel data transmission with bandlimited
pp. 739–749. orthogonal signals,” Bell Sys. Tech. J, vol. 45, no. 10, pp. 1775–1796,
[41] E. Callaway et al., “Home networking with IEEE 802.15.4: A developing 1966.
standard for low-rate wireless personal area networks,” IEEE Commun. [67] M. Bellanger, “Physical layer for future broadband radio systems,” in
Mag., vol. 40, no. 8, pp. 70–77, Aug. 2002. Proc. IEEE Radio Wireless Symp. (RWS), Jan. 2010, pp. 436–439.
[42] G. Mulligan, “The 6LoWPAN architecture,” in Proc. 4th Workshop [68] M. Bellanger et al., “FBMC physical layer: A primer,” PHYDYAS,
Embedded Netw. Sensors, 2007, pp. 78–82. Eur. Union, Brussels, Belgium, Rep., Jun. 2010. [Online].
[43] LoRaWAN 1.1 Specification, LoRa Alliance Tech. Committee, Fremont, Available: http://www.ict-phydyas.org/teamspace/internal-folder/FBMC-
CA, USA, 2017. Primer_06-2010.pdf
[44] J. Petäjäjärvi, K. Mikhaylov, M. Pettissalo, J. Janhunen, and J. Iinatti, [69] M. Schellmann et al., “FBMC-based air interface for 5G mobile:
“Performance of a low-power wide-area network based on LoRa tech- Challenges and proposed solutions,” in Proc. 9th Int. Conf. Cogn.
nology: Doppler robustness, scalability, and coverage,” Int. J. Distrib. Radio Orient. Wireless Netw. Commun. (CROWNCOM), Jun. 2014,
Sensor Netw., vol. 13, no. 3, pp. 1–16, 2017. pp. 102–107.
[45] LoRa Modulation Basics AN1200.22, Semtech Corporat., Camarillo, [70] G. Fettweis, M. Krondorf, and S. Bittner, “GFDM—Generalized
CA, USA, 2015. frequency division multiplexing,” in Proc. IEEE 69th Veh. Technol. Conf.
[46] LoRaWAN 1.1 Regional Parameters, LoRa Alliance Tech. Committee, VTC Spring, Apr. 2009, pp. 1–4.
Fremont, CA, USA, 2017. [71] M. Matthé, L. L. Mendes, and G. Fettweis, “Generalized frequency divi-
[47] M. Centenaro, L. Vangelista, A. Zanella, and M. Zorzi, “Long-range sion multiplexing in a Gabor transform setting,” IEEE Commun. Lett.,
communications in unlicensed bands: The rising stars in the IoT and vol. 18, no. 8, pp. 1379–1382, Aug. 2014.
smart city scenarios,” IEEE Wireless Commun., vol. 23, no. 5, pp. 60–67, [72] T. Wild, F. Schaich, and Y. Chen, “5G air interface design based on
Oct. 2016. universal filtered (UF-)OFDM,” in Proc. 19th Int. Conf. Digit. Signal
[48] T. Adame, A. Bel, B. Bellalta, J. Barcelo, and M. Oliver, “IEEE Process., Aug. 2014, pp. 699–704.
802.11AH: The WiFi approach for M2M communications,” IEEE [73] I. Gaspar et al., “Frequency-shift offset-QAM for GFDM,” IEEE
Wireless Commun., vol. 21, no. 6, pp. 144–152, Dec. 2014. Commun. Lett., vol. 19, no. 8, pp. 1454–1457, Aug. 2015.
[74] N. Michailow et al., “Generalized frequency division multiplexing for
[49] IEEE Computer Society LAN/MAN Standards Committee, IEEE
5th generation cellular networks,” IEEE Trans. Commun., vol. 62, no. 9,
Standard for Information Technology—Telecommunications and
pp. 3045–3061, Sep. 2014.
Information Exchange Between Systems—Local and Metropolitan Area
[75] I. B. F. de Almeida and L. L. Mendes, “Linear GFDM: A low out-
Networks—Specific Requirements—Part 11: Wireless LAN Medium
of-band emission configuration for 5G air interface,” in Proc. IEEE
Access Control (MAC) and Physical Layer (PHY) Specifications
5G World Forum (5GWF) (WF-5G), Santa Clara, CA, USA, Jul. 2018,
Amendment 2: Sub 1 GHz License Exempt Operation, IEEE Standard
pp. 311–316.
802.11ah-2016, 2016.
[76] I. Gaspar et al., “LTE-compatible 5G PHY based on generalized
[50] E. Khorov, A. Lyakhov, A. Krotov, and A. Guschin, “A survey on
frequency division multiplexing,” in Proc. 11th Int. Symp. Wireless
ieee 802.11 ah: An enabling networking technology for smart cities,”
Commun. Syst. (ISWCS), Aug. 2014, pp. 209–213.
Comput. Commun., vol. 58, pp. 53–69, Mar. 2015.
[77] X. Zhang, M. Jia, L. Chen, J. Ma, and J. Qiu, “Filtered-OFDM—Enabler
[51] “Cellular system support for ultra low complexity and low through-
for flexible waveform in the 5th generation cellular networks,” in Proc.
put Internet of Things, v2.1.0,” 3rd Gener. Partnership Project, Sophia
IEEE Glob. Commun. Conf. (GLOBECOM), San Diego, CA, USA, 2015,
Antipolis, France, Rep. TR 45.820, 2015.
pp. 1–6.
[52] J. Schlienz and D. Raddino, “Narrowband Internet of Things whitepa- [78] G. Bochechka, V. Tikhvinskiy, I. Vorozhishchev, A. Aitmagambetov,
per,” Munich, Germany, Rohde & Schwarz, White Paper, pp. 1–42, and B. Nurgozhin, “Comparative analysis of UFMC technology in 5G
2016. networks,” in Proc. Int. Siberian Conf. Control Commun. (SIBCON),
[53] M. Chen, Y. Miao, Y. Hao, and K. Hwang, “Narrow band Internet of Jun. 2017, pp. 1–6.
Things,” IEEE Access, vol. 5, pp. 20557–20577, 2017. [79] D. Mattera, M. Tanda, and M. Bellanger, “Analysis of an FBMC/OQAM
[54] CR2032 Datasheet and Information Site. Accessed: Feb. 2019. [Online]. scheme for asynchronous access in wireless communications,” EURASIP
Available: http://cr2032.co/ J. Adv. Signal Process., vol. 2015, no. 1, p. 23, 2015. [Online]. Available:
[55] A. D. Zayas and P. Merino, “The 3GPP NB-IoT system architecture for https://doi.org/10.1186/s13634-015-0191-4
the Internet of Things,” in Proc. IEEE Int. Conf. Commun. Workshops [80] F. Schaich, T. Wild, and Y. Chen, “Waveform contenders for 5G—
(ICC Workshops), May 2017, pp. 277–282. Suitability for short packet and low latency transmissions,” in Proc.
[56] F. Peng, J. Zhang, and W. E. Ryan, “Adaptive modulation and cod- IEEE 79th Veh. Technol. Conf. (VTC Spring), May 2014, pp. 1–5.
ing for IEEE 802.11n,” in Proc. IEEE Wireless Commun. Netw. Conf., [81] B. Farhang-Boroujeny, “OFDM versus filter bank multicarrier,” IEEE
Mar. 2007, pp. 656–661. Signal Process. Mag., vol. 28, no. 3, pp. 92–112, May 2011.
[57] R. Katiyar and K. V. Padmaja, “Performance analysis of time domain [82] V. Vakilian, T. Wild, F. Schaich, S. ten Brink, and J.-F. Frigon,
and frequency domain equalizer for ADSL,” in Proc. Int. Conf. Commun. “Universal-filtered multi-carrier technique for wireless systems beyond
Signal Process. (ICCSP), Apr. 2016, pp. 513–516. LTE,” in Proc. IEEE Globecom Workshops (GC Wkshps), Atlanta, GA,
[58] Q. He and A. Schmeink, “Comparison and evaluation between FBMC USA, Dec. 2013, pp. 223–228.
and OFDM systems,” in Proc. 19th Int. ITG Workshop Smart Antennas [83] R. Gerzaguet et al., “The 5G candidate waveform race: A com-
(WSA), Mar. 2015, pp. 1–7. parison of complexity and performance,” EURASIP J. Wireless
[59] R. Nissel, S. Schwarz, and M. Rupp, “Filter bank multicarrier modu- Commun. Netw., vol. 2017, no. 1, p. 13, 2017. [Online]. Available:
lation schemes for future mobile communications,” IEEE J. Sel. Areas https://doi.org/10.1186/s13638-016-0792-0
Commun., vol. 35, no. 8, pp. 1768–1782, Aug. 2017. [84] H. Kim, J. Bang, S. Choi, and D. Hong, “Resource block man-
[60] P. Pirinen, “A brief overview of 5G research activities,” in agement for uplink UFMC systems,” in Proc. IEEE Wireless
Proc. 1st Int. Conf. 5G Ubiquitous Connectivity, Nov. 2014, Commun. Netw. Conf. Workshops (WCNCW), Apr. 2016,
pp. 17–22. pp. 477–480.
[61] Y. S. Cho, J. Kim, W. Y. Yang, and C. G. Kang, MIMO- [85] J. Abdoli, M. Jia, and J. Ma, “Filtered OFDM: A new wave-
OFDM Wireless Communications With MATLAB. Singapore: Wiley, form for future wireless systems,” in Proc. IEEE 16th Int.
2010. Workshop Signal Process. Adv. Wireless Commun. (SPAWC),
[62] L. Mendes et al., “GFDM: Providing flexibility for the 5G physical 2015, pp. 66–70.
layer,” in Opportunities in 5G Networks: A Research and Development [86] A. Roessler, “5G waveform candidates application note,” Rohde &
Perspective. Boca Raton, FL, USA: CRC Press, 2016, pp. 325–390. Schwarz, Munich, Germany, Rep. 1MA271, 2016.
DE ALMEIDA et al.: 5G WAVEFORMS FOR IoT APPLICATIONS 2567

Ivo Bizon Franco de Almeida received the five-year Joel J. P. C. Rodrigues (S’01–M’06–SM’06)
B.Sc. and M.Sc. degrees in electrical engineering is a Professor with the National Institute of
from the Instituto Nacional de Telecomunicações, Telecommunications (Inatel), Brazil; a Senior
Brazil, in 2016 and 2018, respectively. He is Researcher with the Instituto de Telecomunicações,
currently pursuing the Ph.D. degree with the Portugal; and a Visiting Professor with the Federal
Technische Universiät Dresden, Germany, where University of Piauí, Brazil. He has authored or coau-
he is a Research Associate with the Vodafone thored over 700 papers in refereed international
Chair Mobile Communications Systems. His cur- journals and conferences, 3 books, and 2 patents.
rent research interests include signal processing, He was a recipient of the several Outstanding
machine/deep learning, and multicarrier modulation Leadership and Outstanding Service Awards by
techniques for future wireless systems. IEEE Communications Society and several best
papers awards. He is the Leader of the Internet of Things Research Group
(CNPq), the Director for Conference Development, IEEE ComSoc Board
of Governors, an IEEE Distinguished Lecturer, the Technical Activities
Committee Chair of the IEEE ComSoc Latin America Region Board,
the President of the scientific council at ParkUrbis, Covilhã Science and
Technology Park, the Past-Chair of the IEEE ComSoc Technical Committee
on eHealth and the IEEE ComSoc Technical Committee on Communications
Software, a Steering Committee Member of the IEEE Life Sciences Technical
Community, and the Publication Co-Chair and a Member Representative of
the IEEE Communications Society on the IEEE Biometrics Council. He is
the Editor-in-Chief of the International Journal on E-Health and Medical
Communications and an Editorial Board Member of several high-reputed jour-
Luciano Leonel Mendes received the B.Sc. nals. He has been the general chair and the TPC chair of many international
and M.Sc. degrees in electrical engineering from conferences, including IEEE ICC, IEEE GLOBECOM, IEEE HEALTHCOM,
National Institute of Telecommunications (Inatel), and IEEE LatinCom. He is a member of the Internet Society and a Senior
Brazil, in 2001 and 2003, respectively, and the Member ACM.
Doctoral degree in electrical engineering from
Unicamp, Brazil, in 2007. Since 2001, he has been
a Professor with Inatel, where he has coordinated
the Master Program from 2010 to 2013. He has
led several research projects funded by FAPEMIG, Mauro A. A. da Cruz received the five-year B.Sc.
FINEP, and BNDES. From 2013 to 2015, he was a degree (licentiate) in informatics engineering from
Visiting Researcher with the Vodafone Chair Mobile the Universidade Católica de Angola, Angola, and
Communications Systems, Technical University of Dresden, where he has the M.Sc. degree from Inatel. He is currently pur-
developed his post-doctoral program sponsored by CNPq. Since 2015, he has suing the Ph.D. degree with an international joint
been the Research Coordinator of the Radiocommunication Reference Center, supervision between University Haute-Alsace and
Inatel. His main area of interest is solutions for 5G PHY, a research field where National Institute of Telecommunications (Inatel).
he has published several papers in the last years. He has also been a CNPq His research interests include Internet of Things
level 2 research fellow since 2016. In 2017, he was an Elected Research (IoT), middleware for IoT, and mobile computing.
Coordinator of the 5G Brasil Project.