This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Cellular LTE-A Technologies for the Future

Internet-of-Things:
Physical Layer Features and Challenges
Mahmoud Elsaadany, Student Member, IEEE, Abdelmohsen Ali, Student Member, IEEE and Walaa Hamouda, Senior
Member, IEEE

Abstract—Human-generated information has been the main a result, the speculations about the number of devices expected
interest of the wireless communication technologies designs for to access the Internet in the near future is increasing every
decades. However, we are currently witnessing the emerge of day. This is indeed supported by the Internet-of-Things (IoT)
an entirely different paradigm of communication introduced by
machines, and hence the name Machine Type Communication framework being promoted to allow a tremendous number
(MTC). Such paradigm arises as a result of the new applications of "Things" to generate and communicate information among
included in the Interne-of-Things (IoT) framework. Among the each other without the need of human interaction [2]. The term
enabling technologies of the IoT, Cellular-based communication "Things" can be used to refer to human services, machineries
is the most promising and more efficient. This is justified by (or parts of machines), sensors in smart grids, monitoring
the currently well-developed and mature radio access networks,
along with the large capacities and flexibility of the offered data devices in e-Health applications, smart motors/cars or any
rates to support a large variety of applications. On the other house hold device in a smart home or a smart city [3][4].
hand, several radio-access-network groups put efforts to optimize In terms of M2M communications and IoT features, a
the 3GPP LTE standard to accommodate for the new challenges new paradigm of networks has to respect the requirements
by introducing new communication categories paving the way to of machines, such as power and cost [5]. For instance, a
support the machine-to-machine communication within the IoT
framework. In this paper, we provide a step-by-step tutorial dis- set-and-forget type of application in M2M devices, such as
cussing the development of MTC design across different releases smart meters, require very long battery life where the device
of LTE and the newly introduced user equipment categories has to operate in an ultra low-power mode [6]. Moreover,
namely: MTC Category (CAT-M) and Narrowband IoT Category the future network should allow for low complex and low
(CAT-N). We start by briefly discussing the different physical data rate communication technologies which provide low cost
channels of the legacy LTE. Then we provide a comprehensive
and up-to-date background for the most recent standard activities devices that encourages the large scale of the IoT. The network
to specify CAT-M and CAT-N technologies. We also emphasize on architecture, therefore, needs to be flexible enough to provide
some of necessary concepts used in the new specifications, such these requirements and more.
as the narrowband concept used in CAT-M and the frequency In this regard, a considerable amount of research has
hopping. Finally, we identify and discuss some of the open been directed towards available network technologies such
research challenges related to the implementation of the new
technologies in real life scenarios. as ZigBee (IEEE 802.15.4), Bluetooth (IEEE 802.15.1), or
WiFi (IEEE 802.11b) by interconnecting devices in a form of
Index Terms—IoT, M2M, MTC, Legacy LTE, CAT-M, CAT- large heterogeneous network [7][8]. Furthermore, solutions for
N, LTE-A specifications, Narrowband concept, Frequency re-
tuning, Repetitions transmission, Enhanced coverage, Low cost, the heterogeneous network architecture (connections, routing,
Low power. congestion control, energy-efficient transmission, etc.) have
been presented to suit the new requirements of M2M com-
munications. However, it is still not clear whether these so-
I. I NTRODUCTION
phisticated solutions can be applied to M2M communications
The Internet technology has undergone enormous changes due to constraints on the hardware complexity, coverage, and
since its early stages and it has become an important communi- coordination. Indeed, while WiFi, Bluetooth and ZigBee are
cation infrastructure targeting anywhere, anytime connectivity. widely used nowadays for -more or less- similar applications
Historically, human-to-human (H2H) communication, mainly as M2M communication, the coverage range of these tech-
voice communication, has been the center of importance. nologies is very short [9][10]. Also, operation on unlicensed
Therefore, the current network protocols and infrastructure are spectrum forces such technologies to adopt spectrum sensing
optimized for human-oriented traffic characteristics. Lately, an techniques (listen-before-talk) and may force a restriction on
entirely different paradigm of communication has emerged the transmission duty cycle. Although these reasons do not
with the inclusion of "machines" in the communications kill the chances of local area network (LAN) technologies to
landscape. The exchange of any machine-generated traffic is enable the IoT and MTC frameworks, it would urge the need
known as Machine-to-Machine (M2M) communication [1]. As for a unified standard (or at least coordination and organization
mechanisms) to serve the needs of M2M and IoT [11]. On the
The authors are with the Department of Electrical and Computer Engi-
neering, Concordia University, Montreal, Quebec, H3G 1M8, Canada (e- other hand, Low Power Wide Area (LPWA) networks present a
mail:(m_elsaad, ali_abde, hamouda)@ece.concordia.ca). good candidate to support the aforementioned diverse require-

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

ments of the IoT framework [12][13][14]. A variety of LPWA stations, giving rise to “star-of-stars" topology, and hence
technology candidates can overcome the short range constraint improves the probability of successfully received messages.
of the LAN and still satisfy the power and latency constraints However, this increases the overhead of the network side as the
using either proprietor or cellular technologies (using licensed resulting duplicate receptions are filtered out in the back-end.
or unlicensed spectrum). It seems more efficient to take Some studies reported the performance of LoRa for outdoor
advantage of the currently well developed and mature radio [19][20][21] and [22] for indoor settings.
access networks. With the large coverage and flexible data 3) INGENU RPMA: Unlike LoRa and Sigfox, INGENU
rates offered by cellular systems, research efforts from industry (also known as On-Ramp Wireless) is a proprietary LPWA
have recently been focused on optimizing the existing cellular technology that operates in 2.4 GHz ISM band and takes
networks considering M2M specifications [15]. Among the advantage of more relaxed regulations on the unlicensed
possible solutions, the famous proprietary technologies: Sigfox spectrum use across different regions [23]. INGENU leads
[16] and LoRa [17], along with the new developments of the efforts to standardize the physical layer specifications under
current cellular technologies such as the new categories of IEEE 802.15.4k standard [24]. INGENU uses a patented
LTE-A user equipments are considered. physical access scheme named as Random Phase Multiple Ac-
cess (RPMA) [25] Direct Sequence Spread Spectrum, which
it employs for uplink communication only. Using CDMA,
A. Unlicensed Technologies
RPMA enables multiple transmitters to share a single time
1) Sigfox: Sigfox is a French company works with network slot. However, RPMA increases the duration of time slot of
operators to offer an end-to-end LPWA connectivity solution traditional CDMA and then distributes the channel access
based on its patented technologies. Sigfox Network Operators within this slot by adding a random offset delay for each
deploy the proprietary base stations equipped with cognitive transmitter. By asynchronous access grants, RPMA reduces
software defined radios to operate as a secondary system overlapping between transmitted signals and thus increases
(unlicensed), and connect them to the back-end servers using signal to interference ratio for each individual link [26]. IN-
an IP-based network. The connectivity to the base station is GENU provides bidirectional communication, although with
simplified and uses only Binary Phase Shift Keying (BPSK) a slight link asymmetry. For downlink communication, base
modulation in an ultra narrow bandwidth (100Hz) in the 868 stations spread the signals for individual end devices and then
MHz or 915 MHz ISM band. This way, Sigfox utilizes band- broadcast them using CDMA. Further, the end devices can
width efficiently and promises ultra-low power consumption, adjust their transmit power for reaching closest base station
and inexpensive RF chain designs. However, Sigfox offers a and limiting interference to nearby devices.
throughput of only 100bps rendering it a candidate for low
traffic applications. Further, a Sigfox downlink communication
can only precede uplink communication after which the end B. Licensed Technologies
device should wait to listen for a response from the base Among other solutions, scenarios defined by the 3rd Gener-
station. The number and size of messages over the uplink are ation Partnership Project (3GPP) standardization body emerge
limited to 140 12-byte messages per day to conform to the as the most promising solutions to enable wireless infrastruc-
regional regulations on use of license-free spectrum. ture of M2M communications [27]. Due to the M2M commu-
2) LoRa/LORAWAN: A special interest group constituted nication challenges and the wide range of supported device
from several commercial and industrial partners known as specifications, developing the features for M2M communica-
LoRaTM Alliance proposed LoRaWAN, as an open standard tion, also refers to machine-type-communication (MTC) in the
defining the network architecture and layers above the LoRa context of Long Term Evolution (LTE), started as early as
physical layer. LoRa (short for Long Range), originally devel- release 10 (R10) for the advanced LTE standard [28]. From
oped and commercialized by Semtech Corporation [18], is a the history of M2M communication (in the LTE convention)
physical layer technology that modulates the signals in SUB- development, the first generation of a complete feature MTC
GHz ISM band. Using chirp spread spectrum (CSS) technique, device has emerged in R12. In this release, R12, the 3GPP
a narrow band input signal spread over a wider channel band- committee has defined a new profile referred to as cate-
width. The resulting signal has noise like properties, making gory 0 or CAT-0 for low-cost MTC operation [29]. Also a
it harder to detect or jam and hence, at the receiver, the signal full coverage improvement is guaranteed for all LTE duplex
enjoys an increased resilience to interference and noise. LoRa modes. Indeed, the effort continued to future releases including
supports multiple spreading factors (between 7-12) to decide release 13 (R13) that was released late in 2016. In this front,
the tradeoff between range and data rate. Higher spreading two special categories, namely CAT-M for MTC and CAT-
factors deliver long range at an expense of lower data rates. N for Narrowband-IoT (NB-IoT), have been incorporated by
Also the combination of Forward Error Correction (FEC) with the 3GPP to LTE specifications to support complete M2M
the spread spectrum technique to further increase the receiver and IoT features, respectively. The new categories satisfy the
sensitivity. The data rate ranges from 300bps to 37.5kbps general requirements of MTC and can support the wide range
depending on spreading factor and channel bandwidth. Further, of IoT applications. For example, the capabilities of the new
multiple transmissions using different spreading factors can be categories can support applications in the domain of fleet
received simultaneously by a LoRa base station. The messages management and logistics, which require secured, wide range,
transmitted from end devices are received by multiple base real time and accurate information with typical data rate of

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

hundreds of Kbps at speeds ranging from 10 to 150 Km/h. On as support for user identity confidentiality, entity authenti-
the other hand, they can also support applications with very cation, confidentiality, data integrity, and mobile equipment
low moving speeds (or stationary) and moderate data rates identification. Commercial launches of LTE-MTC networks is
with hours of latency , such as, automation and monitoring expected to take place in 2017. Unlike LTE-MTC, NB-IoT
applications. systems can be deployed as stand-alone systems or can be
Several working groups in radio-access-networks (RAN) adopted to the LTE guard band. With its reduced bandwidth
contribute very actively to the work on MTC and IoT-related of only 180KHz, low data rate devices can leverage the
optimization for 3GPP LTE networks. From day one, the sup- extended coverage, reduced complexity, and reduced power
port for MTC was one of the major concerns for the 3GPP and consumption by employing NB-IoT systems [33]. Indeed, NB-
the development for a robust MTC design was divided across IoT is assumed to be the gate for the future cellular IoT devices
different releases [30]. Since LTE has the ability to support that has out-of-reach issues, while seldom exchanging data
high performance, high throughput devices, the objective was with the network.
to develop high volume, low cost, low complexity, and low Since the introduction of LTE-A, researchers invested their
throughput User Equipment (UE) LTE-based MTC devices. effort in addressing various challenges to realize these systems
This rapid change in the standard spirit causes considerable and studying the issues that may appear from implementation
amount of updates to all protocol stack layers including perspectives. In fact, a countable number of survey papers
Radio Resource Control (RRC) layer, Medium Access Control exist in the literature in the context of legacy LTE covering
(MAC) layer, and physical layer [31]. However, the updates to a wide range of areas such as physical layer resiliency for
RRC are quite thin when compared to the significant amount of LTE as an OFDM-based system [35], resource allocation [36],
changes required for the physical layer [15]. The reason is that emerging applications for LTE in vehicular networks [37],
MTC/IoT features are more related to the implementation side uplink random access techniques [38], scheduling techniques
of the device rather than the procedures for communications. for uplink and downlink streams [39][40], control channel
For example, one of the mandatory features for CAT-N is evolution in LTE systems [34], and M2M based on LTE
to be flexible in deployment such that it can be deployed systems [15][38][41]. However, with the introduction of CAT-
in-band or in the guard band of an LTE carrier, or in the M and CAT-N LTE technologies, it is important to provide
extreme case to be stand-alone system. These requirements a detailed background about the new features and solutions.
put restrictions on the physical layer design so that the final Investigation of realistic solutions towards combating various
NB-IoT system looks like a new system independent of the practical system implementations has become critical towards
legacy LTE system. As a matter of fact, the legacy LTE system the actual system deployment. Hence to the best of our
provides definition to Frequency Division Duplex (FDD) and knowledge, a comprehensive tutorial on MTC and NB-IoT
Time Division Duplex (TDD) modes. However, based on the communications with its focus on LTE systems is not available
current network deployments, it is a fact that a large scale of in the literature. Therefore, the main purpose of this paper is
the deployments employ FDD rather than TDD. Indeed, it is to provide a review on the studies appeared in the standard
China which launches TDD networking early in 2009 [32]. For agreements, helping the readers to understand what has been
this reason, CAT-N has been defined only for FDD as long as investigated (architecture, technologies, requirements, chal-
R13 is concerned. Motivated by this discussion, in this tutorial, lenges, and proposed solutions) and what still remains to be
we present a comprehensive background about the legacy LTE addressed as implementation challenges. In addition, this paper
and the introduced updates to support CAT-M and CAT-N will reveal an evolutionary path of the LTE-MTC and NB-IoT
technologies [33]. Due to its importance, the standard view systems for futuristic research. For instance, open research
will be mainly from the physical layer perspective focusing topics, targeting the efficient algorithmic solutions for initial
on FDD mode. synchronization, cell search, frequency tracking, and channel
estimation, have been considered. Furthermore, highlighting
the potential implementation aspects for reduced cost and
C. Motivation power consumption requirements are assumed.
LTE-MTC standards-based family of technologies supports
several technology categories, such as CAT-0 and CAT-M [34]. D. Contributions
CAT-0 category is now fully commercial and it is already used In this paper, in addition to the brief introduction to the
in many M2M/IoT deployments. The new CAT-M category is legacy LTE, we provide a comprehensive and up-to-date
a low power wide area technology, which supports IoT through background for the most recent standard activities to specify
lower device complexity and provides extended coverage, CAT-M and CAT-N technologies. We also identify and discuss
while allowing the reuse of the LTE installed base [6]. CAT- some of the key open research challenges related to the imple-
M allows an extended battery lifetime for a wide range of mentation side of such technologies. Our main contributions
use cases, with the modem costs reduced to 20-25% of the can be summarized as follow:
current Enhanced General Packet Radio Service (EGPRS) • The development cycle of the LTE standard is discussed
modems [28]. Supported by all major mobile equipment, chip to show the motivation for the new categories based
set and module manufacturers, LTE-MTC networks will co- on the growing features. The target requirements for
exist with 2G, 3G, and 4G mobile networks and benefit from MTC and IoT categories are presented, compared, and
all the security and privacy mobile network features, such analysed.

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

A. Motivation B. Contribution C. Organization 1) DMRS and SRS 3) PUCCH

I. Introduction
2) PUSCH 4) PRACH
A. FDD Frame
Structure
II. LTE Development for
MTC and IoT
1) Ref
B. Uplink Physical 3) PBCH 5) PDCCH
siganls
Channels
III. Physical Layer Features
in Legacy LTE Systems 2) Sync
4) PDCFICH 6) PDSCH
signals
C. Downlink Physical
Channels and
functionalities
1) Narrowband 2) Frequency
3) Repetition
concept hopping

A. MTC Features

IV. Physical Layer Features

1) PRACH 2) PUCCH 3) PUSCH
for LTE-MTC Systems

B. Uplink Channels

Paper 1) MPBCH 2) MPDCCH 3) MPDSCH

Structure
C. Downlink
D. Challenges
Channels

1) Frame structure 2) Sync. signals 3) NRS

A. Modes of
operation
V. Physical Layer Features
4) NPBCH 5) NPDCCH 6) NPDSCH
for NB-IoT Systems
B. Downlink
physical Channels
1) NPRACH 2) NPUSCH

C. Uplink Physical
D. Challenges
Channels

B. Low Cost C. IoT

A. Low Power Support
Support Market
VI. IoT Implementation
Challenges and Future
Directions

Fig. 1. Chart for the paper structure.

• An introduction to the conventional LTE system is pre- • Since NB-IoT system is the most recent technology
sented. The objective of this overview is to familiarize adopted to LTE, the system is described from scratch. The
the reader with the basic features and concepts for the modes of operation are classified and studied. The most
legacy LTE-A. The system structure is discussed with the recent agreements about CAT-N according to R13 are
definition of various physical channels for both uplink considered. In the downlink side, the new synchronization
an downlink sides in FDD mode. In this regard, the signals and reference signals are discussed in addition
frame structure, uplink physical channels, and downlink to the conventional data and control channels. In the
physical channels are reviewed. The functionalities and uplink direction, the full system is presented with enough
brief description about the operation are assumed for each details for fair comparison with other systems. Logical
physical channel. reasoning is followed to criticize the decisions for various
• The specifications for the LTE-MTC system are con- channel designs. The implementation challenges are then
sidered with reasonable explanations for the decisions discussed so that researchers are encouraged to address
and alternatives. The potential updates including the nar- these potential issues.
rowband concept, the new introduced downlink control • IoT general challenges and future directions are captured.
channel, repetitions, and frequency hopping are discussed Low power consumption and reduced complexity are
in details. Also, the implementation challenges brought highlighted in regards to the LTE-MTC and NB-IoT
by these specification are classified and discussed. systems.

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Table I
D EFINITIONS OF ACRONYMS AND N OTATIONS

Acronym Definition Acronym Definition

3GPP Third Generation Partnership Project NCCE Narrowband Control Channel Elements
ADC Analog-to-digital conversion NPRACH Narrowband Physical Random Access Channel
AWGN Additive White Gaussian Noise NPSS Narrowband Primary Synchronization Signal
BLER Block Error Rate NPUSCH Narrowband Physical Uplink Shared Channel
CCE Control channel elements NSSS Narrowband Secondary Synchronization Signal
CFO Carrier frequency offset PAPR Peak to Average Power Ratio
CQI Channel Quality Indicator PBCH Physical Downlink Broadcast Channel
CRC Cyclic Redundancy Check PDCCH Physical Downlink Control Channel
CRS Cell-Specific Reference Signals PDCFICH Physical Downlink Control Format Indicator Channel
CSI Channel State Information PDSCH Physical Downlink Shared Channel
CSS Common Search Space PHY Physical Layer
DAC Digital-to-analog conversion PMI Precoding Matrix Information
DCI Downlink Control Information PRACH Physical Random Access Channel
DCI Downlink Control Information PRB physical resource block
DMRS Demodulation Reference Signals PSS Primary Synchronization Signal
DRX Discontinuous reception PUCCH Physical Uplink Control Channel
ECCE Enhanced control channel elements PUSCH Physical Uplink Shared Channel
EGPRS Enhanced General Packet Radio Service RAN Radio-access-networks
EPDCCH Enhanced physical downlink control channel RAR Random access response
FDD Frequency Division Duplex RE Resource elements
H2H Human-to-human REG Resource element group
HARQ Hybrid Automatic Repeat Request RI Rank Indication
IEEE Institute of Electrical and Electronics Engineers RLC Radio link control
LAN Local Area Network RNTI Radio Network Temporary Identifier
IoT Internet-of-Things RRC Radio Resource Control
LTE Long Term Evolution RSRP Reference Signal Received Power
LPWA Low Power Wide Area SFBC Space Frequency Block Coding
M2M Machine-to-Machine SRS Sounding Reference Signal
MAC Medium Access Control SSS Secondary Synchronization Signal
MCL Mutual Coupling Loss TDD Time Division Duplex
MIMO Multiple-Input Multiple-Output ToA Time of Arrival
MPDCCH MTC Physical Downlink Control Channel UCI Uplink control information
MTC Machine Type Communication UE User Equipment
NB Narrowband USS User Specific Search Space

E. Organization definitions of the main concepts for this category are not only
The paper is organized as follows. Section II introduces the presented but the main differences among various categories
development cycle of the LTE standard to provide the fun- are assumed as well. In Sections VI-C, the implementation
damental targets for both LTE-MTC and NB-IoT categories. challenges are demonstrated for the NB-IoT system. It is of
A summary of the main physical layer features of various great interest that low power and low complexity requirements
MTC/IoT categories is introduced to highlight the main dif- are potential aspects for the cellular based IoT network. The
ferences. Based on the presented category classification, the advances to these requirements in the scope of LTE-MTC and
legacy LTE fundamentals are presented in Section III. In NB-IoT systems are also provided in Section VI-C. Finally,
addition to the general features, both uplink and downlink conclusions are drawn in Section VIII. The overall structure
physical channels are discussed in Sections III-B and III-C, of the paper is shown in Fig. 1, and Table I lists the acronyms
respectively. The second item in this classification is the LTE- and notations used in the paper.
MTC development which is considered in Section IV. The
most important features for LTE-MTC are presented in Sec- II. LTE D EVELOPMENT FOR MTC AND I OT
tions IV-A. Again, the uplink and downlink physical channels
are investigated and the main differences are highlighted when Although data transmission has been on the rise in the
compared to the legacy LTE system. cellular networks for human-involved applications in the last
The up-to-date agreements about the NB-IoT system includ- decade, cellular networks are mainly optimized for H2H
ing the modes of operations, the uplink physical channels, and communication. However, characteristics of M2M traffic are
downlink physical channels are considered in Section V. The different from the human-generated traffic in the cellular

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

networks. The main differences between H2H and M2M traffic The 3GPP enhancement for LTE to accommodate the re-
can be expressed as [42][43]: quirements of MTC and IoT are summarized next:
• In contrast to human-traffic, in M2M devices, the uplink
traffic is higher than downlink. A. CAT-0 in Release-12
• While human-traffic is mostly concentrated during day- Although Category 1 (CAT-1) was the lowest among all
light and evening, M2M traffic is more uniformly gener- LTE UE categories in R11 in terms of transmission capability
ated by the M2M devices throughout the day. (10 Mbps peak downlink transmission rate and 5 Mbps for
• In some applications (e.g., involving metering devices), the uplink), it was concluded that a new category will be
M2M traffic is periodic. needed to support the new requirements of MTC and IoT.
• In some monitoring applications, the volume of traffic Category 0 is the new standardized category for this purpose
increases sharply after the detection of events (burst in R12. CAT-0 UEs have a reduced transmission rate of 1
traffic). Mbps peak rates for both uplink and downlink. CAT-0 UEs
• For many classes of M2M devices, M2M devices have enjoy a reduced complexity by up to 50% compared to CAT-
a much lower mobility than human devices. However, 1. The new features of the new category include the use of
for the health care devices, and for accessories such as only one receiver antenna with a maximum receive bandwidth
Google glass and Apple watch, the mobility is the same of 20 MHz, which eliminates the use of dual receiver chains.
as the human devices. Also the support of FDD half-duplex operation with relaxed
• The quality-of-service requirements of M2M and human switching time eliminating the need for duplex filters, which
devices may be vastly different. help the manufacturers to significantly reduce the modem cost
While an M2M device typically sends/receives a small compared to more advanced UE categories.
packet of data at each transmission, the extremely large
number of M2M devices may cause severe problems in both B. CAT-M or LTE-MTC in Release-13
access channel and traffic channel of a radio access network
For further complexity reduction techniques, on top of the
and congestion in the core network [44][45]. For these reasons,
ones introduced for CAT-0, a new category, namely Category
from the network access perspective, M2M access requests
MTC (CAT-M), is proposed in the recent work of R13 [52].
are classified by 3GPP into two groups of uncoordinated/non-
The aim of the LTE-MTC (CAT-M) Task Force was to
synchronized and coordinated/synchronized traffic [46]. Coor-
provide a market representation to accelerate the wide-spread
dinated/synchronized traffic is a type of traffic generated by
adoption of 3GPP-based LTE-MTC technology. LTE-MTC is
many similar M2M devices in reaction to an event, whereas
addressing the low-power wide-area IoT market opportunity
uncoordinated/non-synchronized traffic is a result of indepen-
using licensed spectrum with the intent to launch commercial
dent reporting of data. Accordingly, the medium access for
solutions 2017. The main objectives of the LTE-MTC (CAT-
M2M devices would require a special attention and hence the
M) are:
medium access techniques have to be revised to address such
• Facilitate demonstrations and proof of concept trials
challenges . For instance, the access channel overhead has been
carefully investigated where many approaches, including the which strengthen the LTE-MTC solution to meet the low
classical Back-off techniques, the access class barring scheme, power requirements.
• Lead industry partners to build a strong end-to-end in-
slotted access, and others, are assumed [47]. The reader is
encouraged to refer to [48][49][50] [51] for more details about dustry chain for LTE-MTC growth, development and
the higher layer challenges and solution for M2M devices from deployment.
• Further reduction in complexity of LTE-MTC devices.
3GPP perspective.
From the physical layer perspective, the 3GPP standardiza- Recent studies indicate that CAT-M features a complexity
tion community has provisioned suitability of LTE to allow reductions up to 75-80% compared to CAT-1. The most
MTC communication and connectivity over LTE network. A important additional feature is the possibility to implement
lot of studies have been conducted to optimize the radio access the UE transmitter and receiver parts with reduced bandwidth
related technologies and mechanisms. The main task of this compared to legacy LTE UEs operating with 20 MHz band-
effort was to: width. specifically, a CAT-M UE will operate with a maximum
channel bandwidth limited to 1.4 MHz. Another differentiating
• Improve the support of low-cost and low-complexity
feature in CAT-M is the coverage enhancements of more than
device types to match low performance requirements
15 dB (i.e. the received SNR ≈ - 15 dB), enabling the reach
(low data rates and delay tolerance) of certain MTC
the UEs behind the thickest walls or under the ground.
applications.
• Provide extended coverage for MTC devices in challeng-
ing locations. C. CAT-N or NB-IoT LTE in Release-13
• Prolong long battery life of the UEs by enabling very low Since the core IoT devices or massive MTC devices typi-
energy consumptions. cally send small amounts of data and require extended cov-
• Optimize signalling of small data transmission to increase erage, a special category, namely NB-IoT, has been incor-
the cell/network capacity to serve very large numbers of porated to LTE specifications to support IoT features [53].
devices. The design targets for this special category require reduced

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Table II for transmission of a particular transport channel. Each trans-

S UMMARY OF THE BASIC REQUIREMENTS FOR CAT-0, CAT-M, AND port channel is mapped to a corresponding physical channel. In
CAT-N SYSTEMS
addition to the physical channels with corresponding transport
Specification R12 R13 R13 channels, there are also physical channels without corre-
CAT-0 CAT-M CAT-N sponding transport channels. These channels, known as L1/L2
control channels, are used for downlink control information
Downlink peak rate (Mbps) 1 1 0.2
(DCI), providing the terminal with the information required
Uplink peak rate (Mbps) 1 1 0.144
for proper reception and decoding of the downlink data
Number of UE receive antennas 1 1 1 transmission. As for uplink control information (UCI), they
Duplex mode Half Half Half are used to provide the scheduler and the Hybrid Automatic
duplex duplex duplex Repeat Request (HARQ) protocol with information about the
UE receive Bandwidth (MHz) 20 1.4 0.2 situation at the terminal. The relationship between the logical
Maximum UE Transmit power 23 23 20 channels, transport channels, and physical channels in LTE
(dBm) differs in downlink versus uplink transmissions. Next, we will
describe -in some details- various physical channels used in
the uplink and downlink as well as some of the important
complexity, promote battery longevity, and enhanced coverage. signalling used within each of them.
Furthermore, the need to support high data rates seldom
applies to massive MTC. The link budget of NB-IoT has
a 20dB improvement over conventional LTE-A [54]. These A. FDD Frame Structure
requirements have been realized by utilizing a single receive For the LTE to feature a high spectrum flexibility, the
antenna system, supporting only QPSK modulation in the frequency spectra are formed as concatenation of physical
downlink side, and employing extended discontinuous recep- resource blocks (PRBs) each consists of 12 subcarriers. Sub-
tion cycles to reduce the power consumption in deep sleep carriers are separated by 15 KHz, hence, the total bandwidth
modes. Moreover, signal repetition is considered as the key of PRB is 180 KHz. This enables configurations for transmis-
factor to provide performance gain [55]. sion bandwidth form 1.4 MHz with 6 PRBs to a maximum
A summary of the main physical layer features of varies bandwidth of 20 MHz consisting of 110 PRBs. The available
LTE developments is presented in Table II. Furthermore, the channel bandwidths are (1.4. 3, 5, 10, 15, 20 MHz), with
main features for various physical layer channels included transmission bandwidth occupies 90% of all channels, expect
in different technologies are summarized in Table III and for the 1.4 MHz channel which has only 77% efficiency. The
Table IV. The objective of these tables is to list the main rest of the channel bandwidth is used as a guard band to reduce
differences between different technologies which will be ad- the unwanted emissions outside the neighbouring bands.
dressed in the rest of this manuscript. LTE specifies two downlink frame structures. A type 1
frame applies to an FDD deployment and a type 2 frame is
used for a TDD deployment. Each frame is composed of 10
III. P HYSICAL L AYER F EATURES IN L EGACY LTE subframes and each subframe consists of two time slots. Each
S YSTEMS time slot is 0.5 msec, thus a radio frame is 10 msec. The three
Among the objectives of the LTE standard is to create a components of a resource grid are used for, user data, control
more efficient and streamlined protocol stack and architecture. channels, and reference and synchronization signals.
Many dedicated channels specified in previous 3GPP standards Figure 2 shows FDD downlink radio frame structure. The
have been replaced by shared channels and the total number of duration of each frame is 10ms, composed of ten 1 ms
physical channels has been reduced. Logical channels repre- subframes denoted by indices ranging from 0 to 9. Each
sent the data transfers and connections between the radio link subframe is subdivided into two slots of 0.5ms duration.
control (RLC) layer and the MAC layer. In LTE, two types of Each slot is composed of seven or six OFDM, depending
logical channel are defined: the traffic channels and the control on whether a normal or an extended cyclic prefix is used.
channels. While the traffic logical channel is used to transfer The DCI is placed within the first slot of each subframe. The
the data of users, the control logical channels, communicate DCI carries the content of the PDCCH, PCFICH, and PHICH,
the necessary signalling to sustain the connectivity. and together they occupy up to the first three OFDM symbols
Transport channels connect the MAC layer to the phys- in each subframe. The PBCH containing the MIB is located
ical layer, and the physical channels are processed by the within subframe 0 and the PSS and SSS are located within
transceiver at the physical layer. Each physical channel is subframes 0 and 5. The uplink subframe structure is similar
mapped on the resource grid to a set of resource elements to the downlink frame. It is composed of 1ms subframes
(REs) that carry information from higher layers of the protocol divided into two 0.5 ms slots. Each slot is composed of either
stack for eventual transmission on the air interface. seven or six single carrier frequency division multiplexing
The most important transport channel types are the down- (SC-FDM) symbols, depending on whether a normal or an
link shared channel and uplink shared channel, which are used extended cyclic prefix is used. The inner-band (towards the
for data transmission in the downlink and uplink respectively. center) resource blocks are reserved for data resource elements
A physical channel carries the time-frequency resources used (PUSCH) in order to reduce out-of-band emissions, while

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Table III
S UMMARY FOR THE SUPPORTED FEATURES AND FUNCTIONALITIES FOR VARIOUS DOWNLINK PHYSICAL CHANNELS FOR LEGACY LTE, CAT-M, AND
CAT-N DEVICES .

Channel Legacy LTE LTE CAT-M LTE CAT-N

- PSS and SSS signals
- NPSS and NSSS signals
- Used for cell differentiation and
- NPSS is used for timing acquisition
frame timing acquisition
and Initial synchronization
- Also, used for initial synchronization
- NSSS carries the cell information
Sync. - Sent periodically every 5 msec - Same as legacy
- Each spans a complete subframe
Signals - Span the central 72 subcarriers
- Each has a periodicity of 10msec
- PSS is Zadoff-Chu sequence
- Generation for both sequences is
- SSS is a concatenated binary m-
based on Zadoff-Chu sequence
sequence
- A pseudo random sequence with cell
specific initializations used.
- The sequence is modulated through
QPSK
Same as legacy with the exceptions:
- The mapping is a function of the cell
- The mapping process uses different
ID and the CP type
Cell-Specific - Same as legacy OFDM symbols when compared to
- Used to assist channel estimation for
Reference legacy
coherent detection
Signals - Only two antenna ports are supported
- Also, used for channel quality indi-
cation
- Time-frequency orthogonal mapping
is utilized for different antenna ports
- Carries Master information block
which has the essential cell parame- - A new channel is defined
ters such as the bandwidth and frame - The master information block con-
timing. tents has been updated
- it is always allocated in the central - The message size has been reduced
- Same as legacy with the addition
72 subcarriers to 1600 bits instead of 1920 bits
that repetition is possible to enable
Physical - Uses QPSK as the basic modulation - The message is segmented into 8
coverage enhancement and frequency
Broadcast - The message is segmented over four segments where each segment is iden-
tracking under low SNR regimes.
Channel data chunks, each is carried in a sepa- tically repeated 8 times.
rate radio frame. - Each repetition is carried over a
- Blind decode is essential to deter- complete subframe with a periodicity
mine: the radio frame numbering and of one complete radio frame.
the number of Tx antennas.
- Used to carry the Downlink Control
- A new channel, namely MPDCCH,
Information (DCI)
is defined
- Also, used to define the paging op-
- New search spaces are defined with
portunities
reduced blind decode capabilities (up - A new channel, namely NPDCCH, is
- Different formats are defined for dif-
to 20 candidates) defined
ferent DCI messages.
- New formats are defined to carry a - Only up to four candidates are re-
- Mapped to the first couple of OFDM
new set of DCIs quired for blind decode
symbols in a subframe.
- Repetition is supported to enhance - New formats are defined to carry a
- Aggregation is defined to reduce the
the decoding new set of DCIs
Physical coding rate
- Frequency hopping is supported - Repetition is supported to enhance
Control - Uses QPSK modulation
- Mapping spans the whole OFDM the decoding
Channel - Uses convolutional codes
symbols per PRB - Mapping spans the whole OFDM
- Message is scrambled by the user
- Aggregation levels up to 24 control symbols per PRB
identity
elements can be used - Aggregation levels of up to two con-
- Blind decode is required for the UE
- Localized and distributed mapping trol elements can be used
to extract the control information, if
are supported
any.
- Control DCI is meant for future
- Search spaces are defined to enable
downlink assignment
UE better allocates its DCI
Same as legacy with the following - A new channel, namely NPDSCH, is
- Carries physical data with high data
exceptions: defined
rates
- Modulation orders of only QPSK and - Convolutional coding is used
- Uses Turbo coding
16-QAM are supported - Only QPSK is supported
Physical - Higher order modulation up to 256
- A maximum transport block size of - A maximum transport block size of
Shared QAM is supported
1Kb is required 680bits is required.
Channel - Transmit diversity and spatial multi-
- Repetition is supported - Fragmentation is applied
plexing are required
- Frequency hopping is supported - Repetition is supported

edges are reserved as a control region. The reference signals B. Uplink Physical Channels
necessary for data demodulation are interspersed throughout LTE has three uplink physical channels namely, the Physical
the data and control channels. Uplink Shared Channel (PUSCH), Physical Uplink Control
Channel (PUCCH), and Physical Random Access Channel
(PRACH). The PUSCH carries the user data transmitted from
the user terminal. while the PRACH is used for initial access

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Table IV
S UMMARY FOR THE SUPPORTED FEATURES AND FUNCTIONALITIES FOR VARIOUS UPLINK PHYSICAL CHANNELS FOR LEGACY LTE, CAT-M, AND
CAT-N DEVICES .

Channel Legacy LTE LTE CAT-M LTE CAT-N

- Used for channel estimation by the
eNodeB- Sent over complete OFDM - Definition differs from single-tone
symbols every slot operation to multi-tone operation
Demodulation - Zadoff-Chu sequences are utilized - Same as legacy but only for 1.4MHz - Multi-tone operation utilizes the
Reference for the generation legacy Zadoff-Chu sequences with an
Signals - Applied with different density for introduced frequency shift.
data and control channels
- Used to send the preamble sequence
- A new channel is defined, namely
during the initial call setup
NPRACH
- Occupies only 1.4MHz
- To increase the coverage distance, a
- Different preamble configurations Same as legacy with the following
single tone sequence is required with
are supported additions:
a new defined subcarrier spacing of
- Preamble sequence is based on - Repetition is allowed
Random Access 3.75KHz
Zadoff-Chu generation - Repetition is a function of the CE
Channel - Preamble is sent in a form of symbol
- Different cyclic shifts are defined to level
group
reduce the collision possibilities. - Frequency hopping is mandatory
- Frequency hopping is mandatory
- Different preamble formats are sup-
within the preamble group
ported to mitigate various channel
- Repetition is supported
conditions
- Carries the uplink control informa-
Same as legacy with the following
tion
exceptions:
- Different uplink formats (1 to 5) are
- Format 1/1A and 2/2A are the only
defined to carry uplink data including
supported formats
CQI, PMI, RI, and uplink scheduling
Physical Control - ACK/NACK and CQI are the only There is no dedicated control channel
requests.
Channel possible paylaods defined
- Uses Zadoff-Chu sequence modula-
- Frequency hopping is required be-
tion to transmit the payload
tween the two RBs at the band edges
- Mapped to two PRBs at the band
- Repetition is supported
edges
- A new channel is defined, namely
NPUSCH
- NPUSCH has two formats: one for
control and one for data transmission
- Carries the uplink paylaod in addi- - Only ACK/NACK is assumed for
- Same as legacy with the following
tion to control information in a pig- control transmission
exceptions:
gyback fashion, if required. - NPUSCH can be sent over either
- Only QPSK and 16-QAM are sup-
Physical Shared - Utilizes SC-FDMA transmission single tone or multi-tone
ported
Channel - Uses QPSK, 16-QAM, or 64QAM - Packet segmentation is required
- Repetition is required
for modulation - Turbo coding is assumed
- Frequency hopping is mandatory
- Employs Turbo encoding - Repetition is mandatory
- Only BPSK and QPSK are sup-
ported
- Rotated constellation is achieved
through signal generation

of a UE to the network through transmission of random access prefix is used, DMRS signals are located on the fourth OFDM
preambles. The PUCCH carries the UCI, including scheduling symbol of each slot and extend across all the resource blocks.
requests, acknowledgments of transmission success or failure In the case of PUCCH, the location of DMRS will depend on
(ACKs/NACKs), and reports of downlink channel measure- the format of the control channel.
ments including the channel quality Indicator (CQI), Precoding 2) Sounding Reference Signals (SRS): SRSs are transmitted
Matrix Information (PMI), and Rank Indication (RI). There are on the uplink in order to enable the base station to estimate
two types of uplink reference signals in the LTE standard: the the uplink channel response at different frequencies. These
Demodulation Reference Signals (DMRS) and the Sounding channel-state estimates may be used for uplink channel-
Reference Signal (SRS). Both uplink reference signals are dependent scheduling. This means the scheduler can allocate
based on Zadoff-Chu sequences [56]. user data to portions of the uplink bandwidth where the chan-
Zadoff-Chu sequences are also used in generating downlink nel responses are favorable. SRS transmissions are used for
Primary Synchronization Signals (PSSs) and uplink preambles. timing estimation and control of downlink channel conditions
Reference signals for different UEs are derived from different when downlink and uplink channels are reciprocal or identical
cyclic shift parameters of the base sequence. (TDD mode).
1) Demodulation Reference Signals (DMRS): DMRSs are 3) Physical Uplink Control Channel (PUCCH): The
transmitted by UE as part of the uplink resource grid. They PUCCH carries three types of control signaling information:
are used by the uplink channel estimation to equalize and ACK/NACK signals for downlink transmission, scheduling
demodulate the uplink control (PUCCH) and data (PUSCH) requests (SR) indicator sent by UE when it wants to transmit
information. In the case of PUSCH, when a normal cyclic uplink data on PUSCH, and finally the feedback from the

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Legacy control 1 Radio Frame (10ms)

symbols
99
98
97
96
95

rs
94

rie
93

car
92

sub
72
19
20MHz

18
17
16
15 1 PRB

12 subcarriers
14
13
12 Cell specific RSs
11
10
9 Legacy control REs
8
7
6
5
4
Freq 3 1 sub-frame (1ms)
2
1
14 OFDM symbols
0
Slot 6 Slot 7
Time
0.5 ms 0.5 ms

Fig. 2. Frame structure of legacy LTE systems for FDD with Normal CP type.

downlink channel information, including the Channel Quality prevent collisions of the symbols at the base station as the
Indicator (CQI), the Precoding Matrix Indicator (PMI), and the PRACH is sent without timing advance. The preamble format
Rank Indicator (RI). The last three indicators are representing defines the time duration of each field. There are 3 formats
the channel state information (CSI). Furthermore, the feedback depending on the size of the cell and the signal strength.
of the downlink channel information relates to MIMO modes 5) Physical Uplink Shared Channel (PUSCH): When the
in downlink. In order to ensure the correct choice of the MIMO UE receives an uplink scheduling grant, the PUSCH carries
transmission schemes in downlink, each terminal must perform uplink user data and signalling transport blocks arriving from
measurements on the quality of the radio link and report the MAC layer to the physical layer. The UE sends one
the channel characteristic to the base station. This essentially transport block at a time, where CRC is attached to it to help
describes the channel quality functions of the UCI as contained the base station in the error detection process. The resulting
in the PUCCH. block is turbo coded with rate 1/3 and sent through rate
matching. Then the UE reassembles the coded transport blocks
4) Physical Random Access Channel (PRACH): If a UE in the form of codewords, where codewords from the data
would like to transmit on the PUSCH but does not have block are multiplexed with the control signals. Following this
resource on this channel, it should send a scheduling request process, the transport block is passed to the physical processor
on the PUCCH. However, in order to initiate access to the where each codeword is scrambled and the modulation mapper
PUCCH, the UE shall initiate the random access procedure. It groups the codeword bits into modulation symbols. These
uses the Physical Random Access Channel (PRACH) to trans- modulated symbols go through a forward FFT then mapped
mit a preamble to begin such procedure. Since this corresponds to the physical resources using the resource element mapper.
to the first communication from the UE to the base station, Finally, time domain uplink signal is generated using the single
the system does not know the type or specifications of the carrier frequency division multiple access (SC-FDMA). On
UE device. After exchanging messages with the UE, the base the resource grid, the PUSCH occupies a contiguous set of
station sends the UE receives resource grants on the PUSCH resource blocks around the center of the uplink band, and the
and the required timing advance. Various transmission modes, edges of the band are reserved for the PUCCH. Each subframe
such as Cyclic Delay Diversity (CDD) and Precoding Vector contains six PUSCH symbols and one demodulation reference
Switching (PVS), provide a transparent way of decoding the symbol.
preamble information. The PRACH is transmitted on 6 RBs
for the duration up to one subframe long. The exact length and
the frequency offset of the PRACH is advertised by the base C. Downlink Physical Channels and Functionalities
station using SIB2. A PRACH transmission has a cyclic prefix, The Legacy LTE has a single type of traffic logical chan-
a preamble, and a guard period. The preamble sequences have nel which is the Dedicated Traffic Channel (DTCH), and
one or two symbols of 800 microseconds, and are generated four types of control logical channel: the Broadcast Control
from Zadoff-Chu sequences. The guard period is used to Channel (BCCH), the Paging Control Channel (PCCH), the

10

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Common Control Channel (CCCH), and the Dedicated Control The interference between the reference signals is mitigated
Channel (DCCH). The dedicated logical traffic channel and all by generating mutually orthogonal patterns for each pair of
the logical control channels, except for PCCH, are multiplexed consecutive reference symbols.
to form a transport channel known as the Downlink Shared
Channel. The Paging Control Channel (PCCH) is mapped to CSI Reference Signals
the Paging Channel (PCH) and combined with the DL-SCH CSI-RSs are designed for cases where we have between
to form the Physical Downlink Shared Channel (PDSCH). four and eight antennas. CSI-RSs were first introduced in
The PDSCH and four other physical channels (PDCCH, LTE R10. They are designed to perform a complementary
Physical Downlink Control Channel; PHICH, Physical Hybrid function to the DMRS in LTE transmission mode 9. While the
Automatic Repeat Request Indicator Channel; PCFICH, Phys- DMRS supports channel estimation functionality, a CSI-RS
ical Control Format Indicator Channel; and PBCH, Physical acquires CSI. To reduce the overhead resulting from having
Broadcast Channel) provide all the user data, control infor- two types of reference signal within the resource grid, the
mation, and system information needed in the unicast mode, temporal resolution of CSI-RSs is reduced. This makes the
which are delivered from higher layers. system incapable of tracking rapid changes in the channel
1) Reference Signals: Downlink reference signals support condition. Since CSI-RSs are only used with four to eight
the channel estimation functionality needed to equalize and MIMO antenna configurations, and this configuration is only
demodulate the control and data information. They are used active with low mobility, the low temporal resolution of
in CSI measurements (such as RI, CQI, and PMI) needed CSI-RSs does not pose a problem.
for channel quality feedback. LTE specifies five types of
reference signal for downlink transmission, namely: Cell- 2) Synchronization Signals: In addition to reference sig-
Specific Reference Signals (CSR), UE-Specific Reference nals, LTE also defines synchronization signals. Downlink
Signals, Channel-State Information Reference Signal (CSI- synchronization signals are used in a variety of procedures,
RS), MBMS reference signals, and Positioning reference including the detection of frame boundaries, determination
signals. MBMS reference signals are used in the coherent of the number of antennas, initial cell search, neighbor cell
demodulation employed in multicast/broadcast services, and search, and handover. Two synchronization signals are avail-
the positioning reference signal is first introduced in R9 to able in the LTE: the Primary Synchronization Signal (PSS) and
provide measurements on multiple cells helping in estimating the Secondary Synchronization Signal (SSS). Synchronization
the position of a given terminal. signals are related to the PHY cell identity. There are 504
cell identities defined in the LTE, organized into 168 groups,
Cell-Specific Reference Signals each of which contains three unique identities. The PSS carries
Cell-Specific Reference Signals (CSRs) are common to all the unique identities 0, 1, or 2, whereas the SSS carries
users in a certain cell, and are transmitted in every downlink the group identity with values 0 to 167. Thus the physical
subframe and in every resource block in the frequency layer cell identity Ncell 1 2
ID = 3NID + NID is uniquely defined by
1
domain, thus cover the entire cell bandwidth. The CRSs can a number NID in the range of 0 to 167, representing the
be used by the terminal for channel estimation for coherent cell ID group, and a number N2ID in the range of 0 to 2,
demodulation of any downlink physical channel except representing the sector ID within the group. Both the PSS
PMCH and PDSCH in the case of transmission modes 7, 8, and the SSS are mapped onto the central 62 subcarriers with
or 9, corresponding to non-codebook-based precoding. The another 10 subcarriers on the boundaries padded with zeros,
CRSs can also be used by the UE to acquire CSI. Also the UE forming the central 6 RBs (72 subcarriers located around
measurements such as CQI, RI, and PMI performed on CRSs the DC subcarrier). Using this structure, a UE can receive
are used as the basis for cell selection and handover decisions. both synchronization signals without prior knowledge of the
downlink bandwidth. In an FDD frame, they are positioned
UE-Specific Reference Signals in subframes 0 and 5, next to each other with the PSS and
UE-specific reference signals, also known as demodulation the SSS placed in the last two OFDM symbols of slots 0
reference signals (DMRS), are only used in downlink trans- and 10 respectively. The PSS determines one of the three
mission (modes 7, 8, or 9), where CSRs are not used for possible values of the cell identity within a group. To do so,
channel estimation. DMRSs first introduced in LTE R8 in each cell ID uses one of three Zadoff-Chu root sequences of
order to support a single layer, and later in R9 to support up to length 63. The process of the PSS is done by 5 ms monitoring
two layers. Furthermore, an extended specification introduced and comparing it to find the used root sequence making use
in R10 aimed to support up to eight simultaneous reference of the good cross-correlation properties of the Zadoff-Chu
signals. sequences. Hence, the UE can measure the time at which the
When only one DMRS is used, 12 reference symbols are PSS arrived and extract a cell identify within a group. Then,
inserted in resource blocks pair. CSRs require spectral nulls the UE uses this timing information to receive the SSS which
or unused resource elements on all other antenna ports when uses an interleaved concatenation of two binary m-sequences,
a resource element on any given antenna is transmitting a each of length 31 known as Gold sequences to identify the
reference signal. This is a major difference between CSR and cell group. The combination of these two sequences differs
DMRS. When two DMRSs are used on two antennas, all between subframes 0 and 5. The concatenated sequence is
12 reference symbols are transmitted on both antenna ports. scrambled with a scrambling sequence given by the primary

11

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

synchronization signal. Using cross correlation, the UE will antenna instead of 4 antennas, then 48 REs will be having
be able to identify the CP type (Normal or extended) and the CRSs and the remaining elements (out of 144) will be Null.
duplex mode (FDD or TDD) as well as exact timing within
the frame. 4) Physical Downlink Control Format Indicator Channel
3) Physical Downlink Broadcast Channel (PBCH): The (PDCFICH): The PDCFICH is used to define the number of
PBCH carries the Master Information Block (MIB), which OFDM symbols that carry the Downlink Control Information
contains the basic PHY system information and cell-specific (DCI) in a subframe. The PDCFICH information is mapped
information needed during the cell search. After the mobile to specific resource elements belonging to the first OFDM
terminal correctly acquires the MIB, it can then read the symbol in each subframe. The possible values for PDCFICH
downlink control and data channels and perform necessary (one, two, three, or four) depend on the bandwidth, frame
operations to access the system. The MIB contains four fields structure, and subframe index. For the 1.4 MHz bandwidth,
of information. The first two fields hold information regarding since the number of resource blocks is quite small, PDCFICH
downlink system bandwidth and PHICH configuration. The may need up to four symbols for control signaling. However,
downlink system bandwidth is communicated as one of six for the larger bandwidths, the number of can take up to three
values for the number of resource blocks in downlink (6, 15, OFDM symbols.
25, 50, 75, or 100). Those values for the number of resource
blocks map directly to bandwidths of 1.4, 3, 5, 10, 15, and 20 5) Physical Downlink Control Channel (PDCCH): In order
MHz, respectively. The PHICH configuration field of the MIB to start communication between the base station and the
specifies the duration and amount of the PHICH. The PBCH is mobile terminal (UE), a PDCCH is defined for each Physical
always confined to the first four OFDM symbols found in the Downlink Shared Channel (PDSCH) channel. PDCCH mainly
first slot of the first subframe of every radio frame. The base contains the scheduling decisions that each terminal requires
station maps the MIB on the PBCH across 40 ms periods (four in order to successfully receive, equalize, demodulate, and
radio frames), with portions transmitted in the first subframe decode the data packets. Since the PDCCH information must
of every frame. When using normal CP, the PBCH occupies be read and decoded before decoding of PDSCH begins, in
72 subcarriers (6 RBs) centered on the DC subcarrier using a downlink PDCCH occupies the first few OFDM symbols
the first four symbols of slot one. of each subframe. The exact number of OFDM symbols at
The BCH data arrives to the coding unit in the form of the beginning of each subframe occupied by the PDCCH
a maximum of one transport block every transmission time (typically one, two, three, or four) depends on various factors,
interval of 40 ms. Generally there are 14 information bits + including the bandwidth, the subframe index, and the use of
10 spare bits (set to all zeroes currently), makes total 24 bits. unicast or multicast services.
From these information bits, 16 CRC parity bits are computed. The control information carried on the PDCCH is known
The eNodeB can use 1, 2 or 4 antennas for transmission. The as Downlink Control Information (DCI). Depending on the
CRC bits are scrambled based on the 1, 2 or 4 antenna used format of the DCI, the number of resource elements (i.e., the
in the transmitter. Hence, the total number of bits becomes 14 number of OFDM symbols needed to carry them) varies. There
+ 10 + 16 = 40 bits. After the convolutional encoder, the total are 10 different possible DCI formats specified by the LTE
number of bits (for normal CP) becomes = 40 × 3 = 120. standard (known as 0,1,1A,1B,1C,1D,2,2A,2B,2C,2D,3,3A,4),
Then, 24 × 3 NULLs are appended to these 120 bits to make with each format contains a specific set of information and
192 for sub-block interleaving and inter-column permutations. certain purpose. The information carried by DCI format are:
These bits are repeated 16 times by discarding the appended resource allocation information, such as resource block size
Nulls resulting in 120 × 16 = 1920 bits (or 1728 bits in case and resource assignment duration; transport information, such
of Extended CP), then QPSK mapped so that the total number as multi-antenna configuration, modulation type, coding rate,
of QPSK symbols become = 1920/2 = 960 symbols (or 864 and transport block payload size; and information related to
symbols in case of Extended CP). the HARQ, including its process number, the redundancy
These 960 symbols are segmented into 4 equal sized version, and the indicator signaling availability of new data.
self-decodable units or segments. These symbols are then The scheduling messages transmitted over the PDCCH is
placed in PBCH Resource Elements in the second slot of the addressed to a certain Radio Network Temporary Identifier
first subframe (slot 1). That is the first subframe’s second (RNTI). A variety of RNTI types are used in LTE to define the
slot contains 960/4 = 240 symbols (or 216 symbols in case identity of the intended UE which should read the scheduling
of Extended CP) and then inserted in the OFDM resource messages, and the type of these messages. The type of the
elements. PBCH is restricted to the 72 subcarriers around the RNTI is used to generate the CRC code that will be used by
DC in the resource grid irrespective of the UE bandwidth. the transport layer to encode the DCI information.
The PBCH is transmitted in the first four OFDM symbols The resource element mapping of the PDCCH is done using
of the second slot of the first subframe in every radio frame. control channel elements (CCE), each contains 9 resource
For example, in a subframe which contains PBCH (e.g. first element groups (REG). Each REG consists of 4 REs. Based
subframe of a frame) there are total 72 × 7 × 2 = 1008 REs on the length of the DCI message, the base station maps the
(for normal CP). Out of that, currently in that subframe, the PDCCH onto 1,2,4 or 8 consecutive CCE. This is called the
total number of CRS (Cell Reference Signal) = (4 × 6) × 6= aggregation level. In other words, based on the length of the
144 REs (for 4 antennas system). If the system is using single DCI, PDCCH can be scheduled onto 36, 72, 144, or 288

12

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

resource elements. 1 PRB Slot Slot Slot Slot Slot Slot

Further, the CCEs are organized into common search spaces #0 #1 # 10 # 11 # 18 # 19
and UE-specific search spaces. The first type is available to all NBn+2 0.5ms
UE in the cell and have predefined location in the downlink
region. However, the second type is assigned to groups of UEs
NBn+1
and its location depends on the RNTI type of the UE.
As the aggregation level of a UE changes from a subframe to
another, and the length of DCI changes the aggregation level, NBn
the UE locates its PDCCH and find the DCI format using blind
detection. In other words, a UE attempts to decode using all
NBn-1 SSS PSS
DCI formats with candidate locations in the search space, and
Symbol Symbol
the correct combination is the one that checks the CRC. The
reliability of the PDCCH is enhanced by means of transmit NBn-2
1 ms
diversity using space-frequency block codes, and is protected
Sub-frame Sub-frame Sub-frame
against inter-cell interference using a cell-specific scrambling #0 #5 #9

Freq.
pattern.
Time
Although the described PDCCH functions work fine for
most of the situations, it suffers from some limitations as Fig. 3. Frame structure of LTE-MTC systems for FDD with Normal CP
the requirements of the LTE network evolves. First, limiting type showing the synchronization signals. The concept of narrowband is
highlighted.
the channel resources to the first 3-4 symbols of slot 0 has
a direct limitation on the cell capacity (number of users
could be scheduled). Second, PDCCH is transmitted only scheme according to channel quality observed at the mobile
in a distributed manner, hence, it cannot benefit from the terminal. The measurements made at the terminal must be
beamforming using MIMO while the base station enjoys sent back to the base station in order to help the scheduling
multiple antennas. Third, with the PDCCH transmissions of decisions. At each subframe, the mobile terminal needs to be
a certain user spread over the entire frequency band, it does notified about the scheduling from the base station for each
not benefit form frequency selective scheduling nor inter-cell transmitted resource block. Among the information that must
interference coordination. be communicated; are the number of resource blocks allocated
The aforementioned reasons urged the evolution towards the to a user, the transport block size, the type of modulation,
Enhanced physical downlink control channel (EPDCCH) in the coding rate, and the type of MIMO mode used per each
R11. The EPDCCH still carries the same information of PD- subframe.
CCH except that it shares the resources with the traffic PDSCH
to increase the capacity. Within each subframe, a resource
block pair is either assigned to PDSCH or EPDCCH, which IV. P HYSICAL L AYER F EATURES FOR LTE-MTC S YSTEMS
makes the capacity of the new control channel adjustable and The behaviour of the MTC terminals is different from the
also makes use of interference coordination. EPDCCH is trans- legacy LTE users, for which the LTE was optimized. There-
mitted on newly defined four antenna ports (AP 107 - 110), fore, in order to accommodate the new requirements, the CAT-
which are associated with RS (occupying the same REs as AP M UE category is recently added for MTC communication in
7-10). EPDCCH and its RS are preceded with a user-specific R13. A lot of effort was put to specify a new UE for MTC
preceding matrix, and also supports multiuser MIMO (MU- operation in LTE which allows for enhanced coverage com-
MIMO)with up to 4 layers (i.e. supporting simultaneously 4 pared to existing LTE networks and low power consumption.
different users). In details, the LTE-MTC specifies a new R13 low complexity
6) Physical Downlink Shared Channel (PDSCH): After UE category for MTC operation in LTE half duplex FDD mode
the base station sends the UE a scheduling command, it based on the R12 low complexity UE category supporting
transmits the data of the DL-SCH using the scheduling com- additional capabilities. Among these capabilities, is a reduced
mands defined. The PDSCH carries downlink user data and UE bandwidth of 1.4 MHz for the downlink and uplink with
signalling transport blocks arriving from the MAC layer to the ability to operate within any system bandwidth. Also, the
the PHY. Specifically, transport blocks are transmitted one frequency multiplexing of bandwidth reduced UEs and non-
at a time in each subframe. The base station adds 24-bit MTC UEs should be accommodated. For the new UE category
CRC to each DL-SCH transport block which is used by the to have a tangible complexity reduction, the maximum trans-
UE for error detection. Following adaptive modulation and mit power of the new UE power should be reduced such that
coding, the modulated symbols are mapped onto multiple an integrated power amplifier (PA) implementation is possible.
time-frequency resource grids, which are eventually mapped Also, power consumption reduction is a must for the LTE-
to multiple transmit antennas for transmission. The type of MTC UE target ultra-long battery life. The most important
MIMO technique used in each subframe can be adapted based feature of the new category, is to improve the LTE coverage
on the received SNIR (which indicates the channel conditions). corresponding to almost 15 dB for FDD.
It should be mentioned that, the PDSCH and the PUSCH are A variety of techniques is considered to achieve such
the only physical channels that can adapt their modulation requirements, including: subframe bundling techniques with

13

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

HARQ for physical data channels (PDSCH, PUSCH), repe- in order of increasing physical resource-block number. The
tition techniques for control channels (e.g. PBCH, PRACH, narrowband concept and numbering is highlighted in Fig. 3.
EPDCCH), uplink PSD boosting with smaller granularity
than 1 PRB, resource allocation using EPDCCH with cross-
C. Modes Of Operation
subframe scheduling and repetition, new physical channel
formats with repetition for SIB/RAR/Paging, new SIB for 1) Frequency Hopping: With the small bandwidth offered
bandwidth reduced and coverage enhanced UEs, and increased to MTC UE and the use of single receiver chain, the frequency
reference symbol density and frequency hopping techniques. and spatial diversity are taken away. To retrieve some of
the lost frequency diversity, the frequency hopping concept
is added to the MTC LTE system. In other words, MTC
A. MTC Features transmissions will hop from one narrowband to another to
More constrains are added on the MTC UE to reduce its make use of transmitting over different channels, and hence,
cost and complexity. The features of the physical downlink provides frequency diversity. However, frequency hopping
control channel for MTC are summarized as: shall introduce new challenges. One of these issues is that, the
• The design of the physical downlink control channel for UE will have to re-tune its RF chain every time it hops. This
MTC is based on (E)PDCCH. re-tuning takes time which might affect the overall throughput.
• The introduction of new DCI messages to R13 for low According to R13 specifications, two OFDM symbols worth of
complexity UEs. time is needed for re-tuning. Further, re-tuning will take place
• The use of a narrowband (within 6 PRBs) control channel. during the legacy control channel symbols are to be discarded
• Its usage for other UEs in enhanced coverage. anyway by the MTC UE. This choice of re-tune time and
• The demodulation of the control channel shall be based period reduces the chance of loosing a considerable amount
on CRS and/or DMRS. of network throughput. Another issue with frequency hopping
is that the UE and eNodeB both need to know the hopping
pattern. This needs to be signalled or must be determinable
B. Narrowband Concept from certain system parameters.
The bandwidth choice of a certain base station is configured For MPDCCH, the first NB is determined without any new
once and remains unchanged during operation, hance, it would higher layer configuration involved (i.e. it is specified in the
be a good choice to use a bandwidth unit that can be a system information messages). The other NB are determined
common divisor of the available bandwidth options in the using a single configurable offset. The offset, signalled in
legacy LTE. Then the choices become limited to 6 RBs or the system information, is cell-specifically configured and
one RB. However, in order for the MTC devices to capture applicable to all CE levels. In CE mode A, the hopping is
the signature of the LTE signal, it will have to receive the turned on or off dynamically by higher-layer signalling in the
synchronization signals PSS and SSS. These signals occupy DCI if the hopping is enabled. The number of NB for hopping
the central 6 RBs of the bandwidth of the base station, which is either 2 or 4. This parameter is signalled to the UE and it is
makes it desirable for the MTC devices to take 6 RBs as the cell-specific. In reality, one problem with frequency hopping
basic unit of bandwidth. Not only that, but also as mentioned is that channel estimation across subframes has been the main
before, the legacy PDCCH is mapped to the full occupied implementation scheme to enhance the estimation accuracy.
bandwidth, which makes decoding PDCCH is not possible in However, with frequency hopping, the channel estimation will
MTC applications, since the UE is limited to bandwidth of need to be restarted whenever hopping is scheduled. Based
only 1.4 MHz. However, the enhanced version EPDCCH, uses on the specifications, the MPDCCH transmission will stay
one PRB pair as the basic resource unit, which makes it a good on one NB for Y subframes, where Y is in the range of
candidate to be used to control MTC UE. Furthermore, the use 2-8 subframes to allow for some benefit of cross-subframes
of only one PRB for the MTC control channels based on EPD- channel estimation.
CCH was reported to be insufficient, as it could provide the This value Y is called the frequency hopping interval and
required converge enhancement (CE) [57]. It was concluded it is cell-specific. There are 4 values signalled to specify
in [58] that, increasing the bandwidth of the basic EPDCCH the frequency hopping interval for all channels in a specific
to 6 PRBs in the 1.4 MHz bandwidth on the MTC UE along channel direction and UE mode, namely mode A DL, mode
with repetition will be sufficient to have a good coverage of B DL, mode A UL, mode B UL. For FDD, and mode A, the
-14 dB. Furthermore, coverage enhancement can be achieved values that this parameter may take are among the set 1, 2, 4,
by employing EPDCCH that supports beamforming which 8. For mode B, they take values in the set 2, 4, 8, 16.
increases coverage by directing the power of the base station 2) Repetitions: As mentioned before, the limitations forced
towards the UE. For these reasons, it is agreed on to use the on the MTC LTE directly affects the downlink control channel,
6RBs (or one narrowband (NB)) as the basic bandwidth unit which in turn, has a direct impact on the performance of the
for MTC. A narrowband is defined as six non-overlapping downlink. For the downlink to have sufficient link budget to
consecutive physical resource blocks in the frequency domain. support the prospected coverage enhancement, an enhanced
The total number of downlink narrowbands in the downlink version of the distributed EPDCCH is reused as the base
transmission bandwidth configured in the cell is given by control signal for MTC. Thus, enhancement is obtained by
DL
NRB
B = ⌈ 6 ⌋ and are numbered nN B = 0, ..., NN B − 1
DL DL
NN transiting repeated copies of the same signal over time. As

14

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

0
10
R=1
For each coverage enhancement level, UEs with different
EPA−5
AWGN geographic locations may have different radio conditions and
R=16 R=8 R=4
R=2 propagation delay. Thus, different PRACH repetition levels
could apply to different preamble formats to adapt to different
R=64
propagation delay and compensate for different propagation
BLER

−1
10
losses. For each PRACH coverage enhancement level, there
is a PRACH configuration done at higher layers with a
PRACH configuration index values from 0 to 63, a PRACH
frequency offset, and a number of PRACH repetitions per
P RACH
−2
10
R=32
R=32 attempt Nrep . The UE selects the repetition level to use
−16 −14 −12 −10 −8 −6 −4 −2 0 2 for the initial transmission based on DL measurements, but
SNR in dB this is only acceptable if sufficient DL measurement accuracy
Fig. 4. Effect of repetition on the BLER versus Average SNR for aggregation
can be achieved within a reasonable DL measurement time.
level L=2 in MPDCCH decode. If sufficient accuracy cannot be achieved, the UE will start at
the lowest configured PRACH repetition level. The number of
repetitions as well as the starting subframe are also configured
a direct result of the use of such repetition code, the link by higher layer signalling. It has been proposed that, the UE
performance can be enhanced through time diversity and boost should remember what PRACH repetition level is used last
the control signal energy. On the other hand, the repetition has time and use this information when setting the starting point
the effect of increased decoding time (more latency), which for the next access. PRACH frequency hopping may provide
requires more wake-up time of the MTC device. frequency diversity gain and reduce the number of repetition.
In order to ideally show the effect of repetition, the con- From this point of view, the power consumption of enhanced
trol channel performance has been considered. For instance, PRACH transmission could be saved due to reduced active
Additive White Gaussian Noise (AWGN) and EPA-5 [59] time.
were simulated. EPA-5 is a standard LTE channel with large 4) MTC Physical Downlink Control Channel (MPDCCH):
coherence time (maximum Doppler spread is 5Hz). The perfor- EPDCCH has been chosen to be the starting point towards
mance is evaluated under perfect synchronization conditions in designing the new MTC control channel MPDCCH. However,
addition to the assumption of perfect knowledge of the channel special requirements have been adapted to best suite the new
state information. The objective is to evaluate the performance MTC platform and to support the required coverage with
loss due to the fading channel without incorporating any reasonable complexity and power consumption. The new set
implementation or complexity loss. That way, the EPA-5 effect of features for MPDCCH require defining new set of downlink
is obvious when compared to AWGN. Fig. 4 shows the control formats, adding a possibility for a common UE to
performance, represented by the Block Error Rate (BLER), access the control channel by introducing the new common
for decoding a single MPDCCH candidate that corresponds search space, and enhancing the control channel assignment
to aggregation level 2. Different repetition levels are assumed procedure to support the new MTC features such as repetition
for both channels. It is very clear that the gain of repetition and frequency hopping. A summary of the introduced features
appears significantly at very low SNR values. From Fig. 4, to design the new MPDCCH is shown in Fig. 5.
one can notice that enhanced coverage LTE-MTC UE oper- 5) MTC Search Spaces: In LTE systems, the downlink
ating at SNR=-15dBs requires repetitions in the order of 64 control region is shared by all UEs in one cell. Each UE
and higher for aggregation level 2 to achieve BLER of 1% should monitor the control region and perform blind decoding
and less. Indeed, the conventional EPDCCH performance is to detect whether or not there is control information for itself.
characterized by the case in which R = 1. It is obvious that In order to reduce the number of blind decoding trials per UE,
without repetition, classical EPDCCH has no ability to decode each UE has a defined search space area of the control region
the control channel correctly at very low SNR values for this to monitor, rather than having to monitor the whole control
aggregation level. region. The search space is defined on the basis of enhanced
3) MTC Physical Random Access Channel (MPRACH): control channel elements (ECCEs). Control information may
The legacy design of the PRACH channel is limited to 6 RBs occupy 1, 2, 4, or 8 CCEs aggregated together depending on
only, which makes it suitable for the MTC use as it signi- the size of the control information and the channel quality of
fies the narrowband constraint. However, the legacy PRACH the UE.
needs some modifications to support the extra path-loss due The search space starting point for a UE is determined by a
to the extended coverage. Hence, repetitions and frequency hash function and the search space size is determined by the
hopping is exploited to provide the necessary diversity to PDCCH aggregation level. The hash function randomizes the
the MPRACH. Like the legacy PRACH, The physical layer search space locations of different UEs and effectively reduces
random access preamble consists of a cyclic prefix of length the blocking probability. The MPDCCH has two broad classes
TCP and a sequence part of length TSEQ with different 5 of search spaces: UE-specific search space (USS) for messages
preamble formats (0,1,2,3,4). The values of these parameters directed specifically to the UE and Common Search Space
depend on the frame structure and the random access configu- (CSS) for messages directed to multiple users or for messages
ration and the preamble format is controlled by higher layers. to a specific UE before the USS has been configured. It is

15

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

UE Specific Search Type-0 CSS introduced to the system. The motivation is to skip the
Space (SSS) unused parameters such as the number of codewords since
Search Spaces
(SS) Type-1 CSS
LTE-MTC is defined for a single codeword. Also, the DCI
Common Search messages for both uplink and downlink grants were designed
Space (CSS)
to fit the same number of bits with some indication bits to
Type-2 CSS differentiate between the formats. This is mainly to reduce the
blind decoding iterations and hence, reduced UE complexity.
Format 6-0A Furthermore, with the introduction of enhanced coverage,
repetition, and frequency hopping, the UE will mainly require
Format 6-0 Format 6-0B new information to facilitate the use of repetition, narrowband
resource assignment, frequency hopping flag. For example,
MPDCCH DCI Formats
Format 6-1 Format 6-1A downlink DCI assignment should indicate the MPDCCH
repetition factor so that UE can expect when the message
Format 6-1B
is complete. For these reasons and others, three new DCI
Format 6-2
formats, namely format 6-0, 6-1, and 6-2 have been defined
for LTE-MTC system for uplink grant, downlink scheduling,
Aggregation Levels Localized and paging, respectively. To best suite the operating mode,
whether being Mode A or Mode B, there are two versions for
Aggregation Types Distributed each format where the number of information bits differ.
MPDCCH
Assignment 1 Set
PRB Sets
(2 or 4 PRBs) V. P HYSICAL L AYER F EATURES FOR NB-I OT S YSTEMS
A. Deployment Scenarios and Modes of Operation
2 Sets
Repetition (2+4 PRBs) As a finite and scarce natural resource, spectrum needs to
be used as efficiently as possible. Thus, technologies that use
Frequency Hopping spectrum tend to be designed to minimize usage [60]. To
achieve spectrum efficiency, NB-IoT has been designed with
Fig. 5. Summary for the introduced features for MPDCCH. a number of deployment options for GSM, WCDMA, or LTE
spectrum. There are three deployment scenarios.
1) In-band Operation: An NB-IoT carrier is a self-
worth to mention that the legacy EPDCCH has only USS since contained network element that uses a single PRB. For
PDCCH is used for the CSS. In MPDCCH, the UE will have in-band deployments with no IoT traffic present, the
to blindly decode both USS and CSS. The search spaces will PRB can be used by LTE for other purposes, as the
differ depending on the CE mode whether mode A or mode infrastructure and spectrum usage of LTE and NB-IoT are
B. fully integrated. The base station scheduler multiplexes
6) MTC DCI Formats: Generally, the eNodeB employs NB-IoT and LTE traffic onto the same spectrum, which
the downlink control information messages to send downlink minimizes the total cost of operation for IoT services.
scheduling commands, uplink scheduling grants, and uplink To support full flexible design, the specifications define
power control commands to the UE. The DCI can be written two modes for the in-band operation. The first mode,
using several different formats. Indeed, conventional LTE- namely Same-PCI mode, assumes that the NB-IoT carrier
A systems up to R12 support 14 different formats, namely has identical cell parameters (i.e., cell ID and number of
format 0, 1/1A/1B/1C/1D, 2/2A/2B/2C, 3/3A, 4, and 5. Each Tx antennas) as the donor legacy cell. The other mode
format contains a specific set of information and has a specific considers some flexibility of having a different cell ID
purpose. However, some formats have been grouped under and different number of Tx antennas.
common name which implies that the main functionality of 2) Stand-alone Operation: This mode of operation is
these downlink messages would have something in common mainly intended to replace a GSM carrier with an NB-IoT
but the details will be different. For instance, the main theme carrier. By steering some GSM traffic to the WCDMA
for Format 1 and its group is to mainly provide the downlink or LTE network, one or more of the GSM carriers can
scheduling information such as the PRB allocation, modula- be used to carry IoT traffic. As GSM operates mainly
tion and coding, and single or multiple users support. The in the 900 MHz and 1,800 MHz bands (spectrum that
differences from one format to another in the same group is present in all markets), this approach accelerates time
are to which transmission mode and antenna system to use. to market, and maximizes the benefits of a global-scale
Also, the basic purpose for Format 2 is to set configuration infrastructure.
for MIMO systems. However, they differ on how the MIMO 3) Guard band Operation: this can be applied either in
system is configured on being used for open loop, closed loop, WCDMA or LTE. To operate in a guard band without
beamforming, and multi-user MIMO configurations. Although causing interference, NB-IoT and LTE need to coexist.
there have been multiple formats to send various scheduling The physical NB-IoT layer is designed with the require-
commands, in LTE-MTC systems, new formats have been ments of LTE guard band coexistence specifically taken

16

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

into consideration. Again, NB-IoT uses OFDM in the all of them are only indicated by the NSSS. The NSSS, S(k),
downlink and SC-FDMA in the uplink. The design of is generated according to,
NB-IoT has fully adopted LTE numerology, using 15kHz πuk(k+1)

subcarriers in the uplink and downlink, with an additional S(k) = C̄q (k ′ ) e−j2πθf k e−j 131 , 0 ≤ k < 132 (1)
option for 3.75kHz subcarriers in the uplink to provide where k ′ = k mod 128, the root sequence, u, is related to the
capacity in signal-strength-limited scenarios. N cell N cell
cell ID, NID , by u = (NID mod 126) + 3, and the cyclic
It should be noted that, NB-IoT supports operation with shift, θf , is related to the System Frame Number (SFN), nf ,
31
only one or two Tx antenna ports. For operation with two such that θf = 132 (nf /2) mod 4. The modulated sequence,
Tx antenna ports, NB-IoT uses the conventional Space Fre- C̄q (k ), is given by C̄q (k ′ ) = 2Cq (k ′ ) − 1, where q is a cell
′

quency Block Coding (SFBC) employing the Alamouti map- specific parameter that is given by q = ⌊NID N cell
/126⌋ and Cq
ping. Unlike other LTE-based systems, NB-IoT utilizes the forms four complementary 128-bits binary sequences.
same transmission scheme for all physical channels including In conventional LTE, primary and secondary synchroniza-
Narrowband Physical Downlink Control Channel (NPDCCH), tion signals (i.e., PSS and SSS, respectively) are mapped
Narrowband Physical Broadcast Channel (NPBCH), and Nar- to two consecutive OFDM symbols in the same slot with a
rowband Physical Downlink Shared Channel (NPDSCH). periodicity of 5msec. However, NPSS is mapped to subframe
5 of every radio frame. NSSS is mapped to the last 11
ODFM symbols of subframe 9 in radio frames having nf mod
B. Downlink Physical Channels 2 = 0. Sequences are mapped to frequency sub-carriers in an
1) Frame Structure: With a carrier bandwidth of just increasing order, then applied across time as shown in Fig. 6.
200KHz, an NB-IoT carrier can be deployed within an LTE 3) Narrowband Reference Symbols (NRS): Due to the lack
carrier as one PRB. Fig. 6 shows a 3MHz LTE carrier in of fully adoptable LTE signal structure in case of guard and
which a single PRB is assigned to NB-IoT. An operating stand-alone modes, new reference signals or pilots, namely
NB-IoT band is defined as a contiguous set of 12 sub- NRS, are inserted within the transmitted signal to assist the
carriers forming one PRB. A single radio frame is 10ms channel estimation process which is required for coherent
which consists of 10 subframes with equal duration. Each detection at the UE side. Similar to legacy LTE, NRS uses
subframe is divided into two slots with equal periods. Unlike a cell-specific frequency shift derived as the modulo division
conventional LTE which defines two CP types with different of the NB-IoT cell ID by 6. The conventional LTE CRS
CP patterns, NB-IoT in R13 supports only the normal CP type, sequence is reused for NRS generation where the centre of
where a slot is composed of 7 OFDM symbols. According LTE CRS sequence is employed as NRS sequence for all
to the specification [54], if the signal is sampled at 1.92 PRBs. The NRS is mapped to the last two OFDM symbols of
MSamples/sec, similar to LTE-MTC, the CP length of the the slot for both antenna ports in case of transmit diversity as
first symbol in each slot is 10 samples and those of the other shown in Fig. 6. As NPSS/NSSS occupy the last 11 OFDM
symbols are 9 samples long. Also, in this case, the OFDM symbols within the subframes transmitting NPSS/NSSS for
symbol spans N = 128 sub-carriers. normal CP, NRS are not mapped to these subframes. To
2) Synchronization Signals: NB-IoT intends to occupy a ensure the demodulation and/or measurement performance,
narrow bandwidth of only 200KHz, which is not backward it has been agreed to transmit NRS in all valid subframes
compatible to the supported bandwidths by the legacy LTE. except the NPSS/NSSS subframes regardless of whether there
Therefore, NB-IoT redefines the cell attach procedure includ- is downlink transmission in these valid subframes or not.
ing cell search and initial synchronization [61][60]. During In in-band operation mode, the LTE CRS REs should be
initial synchronization, CFO is estimated and compensated to reserved from NB-IoT to avoid pollution to the LTE channel
enable proper signal detection. The UE acquires the physical estimation and measurement. However, as a special case,
cell identification by employing the cell search procedure. To NPBCH decoding shall not rely on LTE CRS due to the lack of
cope with these changes, NB-IoT employs new set of synchro- knowledge of the legacy PRB index information. This means
nization signals, namely Narrowband Primary Synchronization that the new reference signal (i.e. NRS) alone should be able to
Signal (NPSS) and Narrowband Secondary Synchronization provide sufficient channel estimation performance for NPBCH
Signal (NSSS) [54]. The new sequences have different band- decoding. The NPDCCH and NPDSCH decoding performance
width, mapping, periodicity, and generation when compared to can be ensured also based on NRS only. In this sense, LTE
the legacy LTE synchronization signals. Unlike conventional CRS is not a must in-band operation mode in terms of DL
LTE, cell ID is encapsulated only in the secondary sequence channel decoding. However, the standardization is designed to
without involving the primary sequence. provide the flexibility to UE implementation for the use of LTE
NPSS and NSSS sequences are constructed from a fre- CRS although the performance requirements do not rely on it.
quency domain Zadoff-Chu sequence where NPSS length is 11 On the other hand, the extra signalling overhead to support the
samples while the NSSS consists of 132 samples. The NPSS, use of LTE CRS would be limited since the resource mapping
Pl (n), is generated such that Pl (k) = Q(l)e−jπuk(k+1)/11 , information of the LTE CRS needs be anyhow indicated to
where 0 ≤ k < 11, 3 ≤ l < 14 is the OFDM symbol index, the realize the rate matching around the LTE CRS REs. LTE
sequence root u = 5, and Q(l) is a modulation sequence given CRS may also benefit measurement accuracy when available.
by {1, 1, 1, 1, −1, −1, 1, 1, 1, −1, 1}, respectively. In NB-IoT For these reasons, the agreements allow two modes for in-
system, there are still 504 unique physical cell IDs. However, band operation. The first assumes that the full information

17

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Legacy control 1 sub-frame (1ms) 1 sub-frame (1ms) 1 sub-frame (1ms) P13(10) S(11)
S(131)
symbols 14 OFDM symbols 14 OFDM symbols P3(10) 14 OFDM symbols Slot #18 Slot #19

14

12 subcarriers
13
12
11
10
9
3MHz

8
7
6
5
4
3 NRS Subframe #0 Subframe #1 Subframe #5 P13(0) S(0) Subframe #9 S(120)
2 P3(0)
1
NSSS Subframe
Freq.

0 NPBCH Subframe NPDCCH/NPDSCH NPSS Subframe

Subframe Inband CRS
Time
Fig. 6. Radio frame structure for NB-IoT systems. The allocated RB is expanded in time to show the NPSS/NSSS symbols mapping in addition to the
broadcast channel and one data/control subframe.

of the legacy cell in terms of the cell ID and number of Tx of only 120 bits. Furthermore, it was agreed that NPBCH
antennas is identical to the NB-IoT carrier. In this case, the consists of 8 independently decodable blocks. Thus, after CRC
UE is free to utilize the CRS for its own demodulation and/or attachment and channel coding, the NB-MIB is rate matched
measurements. On the other hand, the other mode assumes to 1600 bits instead of 1200 bits for Normal CP type.
that the UE is not aware of the legacy cell ID, but it has The rate matched bits are scrambled by the conventional
the information of the number of its Tx antennas which helps LTE scrambler that is initialized with the NB-IoT Physical Cell
determining the rate matching around the CRS. Identifier (PCI) in each radio frame fulfilling nf mod 64 = 0,
4) Narrowband Physical Downlink Broadcast Channel where nf is the SFN. This simply refers to the fact that a single
(NPBCH): Since NB-IoT has various deployment scenarios NB-MIB transmission spans 64 radio frames. After the channel
that require different system configurations, some Narrowband interleaver, the bits are QPSK modulated. The modulated
Master Information Block (NB-MIB) fields are required to be bits are mapped to resource elements in a frequency then
operation mode dependent. For example, the UE would require time fashion. During the mapping, the 800 QPSK modulated
to puncture the CRS in case of in-band operation. Therefore, symbols are segmented into 8 consecutive segments such that
a field representing the number of legacy Tx antennas is each segment is repeated in 8 consecutive radio frames. In
beneficial. In the same mode, the PRB index is needed to other words, identical symbols carrying the same segment are
validate the working assumption regarding the potential usage transmitted with 80ms duration. It is essential to enable the
of CRS for demodulation. In addition, there should be an NPBCH decoding without any prior information about the
indication for the raster frequency offset for the in-band operation mode. Indeed, only cell ID and proper frame timing
operation. It is understood that, during start, the UE would are required through the conventional cell search procedure.
scan the frequency spectrum searching for a valid NB-IoT Therefore, it has been agreed that NB-PBCH is transmitted in
carrier. In legacy LTE, the raster frequency is assumed to be subframe 0 in every radio frame, where the first 3 symbols
100KHz starting from the center of the LTE band. However, in a subframe are not utilized independent of the operational
since the NB-IoT in-band deployment assumes that the NB- mode. In addition, for rate matching purposes, the CRS are
IoT carrier would be allocated as a single PRB with 180KHz punctured assuming 4 Tx antenna ports even for guard and
width, the raster frequency to the NB-IoT carrier would not be stand-alone modes. Moreover, the number of NRS ports is
multiples of 100KHz. Indeed, for odd LTE bandwidths, there considered based on two transmit antennas independent of
is a frequency offset of ±7.5KHz for the raster frequency to the actual configuration. However, the number of NRS ports
be multiple of 100KHz. In fact, the synchronization signal (i.e., whether 1 or 2) is indicated by NB-PBCH CRC masking
design and mapping are carefully studied so that a UE can similar to the conventional system.
lock to the NB-IoT carrier with an ambiguity of ±7.5KHz. It 5) Narrowband Physical Downlink Control Channel
is clear that a new field has to be added to the NB-MIB in (NPDCCH): The NB-IoT has its own control channel with
order to differentiate between the various frequency offsets as customized features and definitions. Although some features
the bandwidth would not be known. have been adopted from LTE-MTC system, there are couple of
In all cases, similar to LTE-MTC, NB-MIB indicates the restrictions that mandates redefinitions for some of the control
scheduling information for the system information messages channel concepts. First of all, the NPDCCH transmission
by defining the TBS size and repetition filed for the first SIB becomes packet-based as the system only supports half-duplex
message (i.e., NB-SIB1). For these reasons, it was beneficial to mode at least for R13 version. There is no overlapping between
design new NB-MIB fields not only to include the introduced the data channel and the control channel within the same
set of fields but also to interpret the fields depending on subframe. The UE will monitor the NPDCCH when it expects
the operation mode. The decision is to extend the NB-MIB a random access response (RAR), a new DL assignment,
length to 34 bits before CRC attachment. This means that the paging, or uplink grant. In case of RAR or paging, the UE
code word for the broadcast channel becomes 150 bits instead has to monitor a common search space. Therefore, similar to

18

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

UE Specific Search To simplify the NPDCCH mapping, the concept of en-

Space (SSS) Type-1 CSS
hanced resource element group (EREG) has not been defined
Search Spaces
(SS) for NPDCCH. The control REs are directly mapped to the
Common Search Narrowband Control Channel Elements (NCCEs). Only two
Type-2 CSS
Space (CSS)
NCCEs are available in one subframe where one NCCE con-
sists of 6 subcarriers per OFDM symbol in a subframe. Aside
Format N0 from the expected performance mismatch between the two
NCCEs due to the frequency selectivity of the channel, it has
NPDCCH DCI Formats Format N1 been agreed that NCCE mapping simply employs Frequency
Division Multiplexing (FDM) structure in which the first 6
consecutive subcarriers are assigned to one NCCE and the
Format N2
other 6 subcarriers to the other NCCE. To support transmit
diversity and gain some spatial diversity for NPDCCH, REs
NPDCCH
Assignment
Aggregation Levels should be paired with shortest possible distance between REs
in each pair. In this way, the instantaneous channel response
for both REs within one pair are almost the same and highly
Repetition
correlated, which is necessary for achieving high diversity
gain. However, RE-pairs are wrapped around the NRS REs
Fig. 7. Summary for the introduced features for NPDCCH. when applicable. Due to the availability of two NCCEs only,
two aggregation levels are only supported. In all cases except
UE search space with repetition one and two, all candidates
LTE-MTC and unlike the conventional EPDCCH, two search are expected to use aggregation level 2 to provide better
spaces have been defined for NB-IoT. Second, due to the performance under deep coverage.
reduced set of information carried by the DCI messages, three An important aspect of physical layer design for NB-IoT
new DCI formats, namely N0, N1, and N2, have been defined system is to keep the UE power consumption low. Conse-
only for NB-IoT in R13. Third, NPDCCH assignment relies quently, the UE should turn on its receiver for as short periods
on repetition to support unconditional coverage enhancement of time and in as few occasions as possible during the day.
with relatively small aggregation levels again for complexity One way to keep this period short is to reduce the DCI size in
reduction. Fig. 7 shows the essential differences between order to achieve NPDCCH transmissions with small payload
NPDCCH and other control channels in various LTE systems. and short transmission time. In the DCI design for MTC,
Two broad classes of search spaces have been defined for this was already considered through various simplifications
NB-IoT, namely USS for messages directed specifically to to reduce the number of DCI bits. The same principles are
the UE and CSS for messages directed to multiple users, applied to the coverage modes in NB-IoT. Furthermore to
paging, or for messages to a specific UE before the USS has reduce the complexity, it is agreed to target the same DCI
been configured (i.e., RAR message). Similar to LTE-MTC format for uplink grant and downlink assignment, as well
with repetitions enabled, NB-IoT provides the definition for as different coverage levels. Thus, DCI messages can be
the search space as joint combination of aggregation level carried by NPDCCH using payload size no more than 23 bits
and repetition factor. By varying the repetition number, a for both UL grant/DL assignments which in turn defines a
Type-1 CSS is defined for paging with more possibilities for new set of DCI formats. DCI format N0 aims to scheduling
repetitions (a maximum of 2048) and another type, namely of NPUSCH in one UL cell with the essential information
Type-2 CSS, is specifically defined for RAR with a maximum about repetition, resource assignment, coding and modulation,
of 8 repetitions. To avoid blocking, each UE has search space and HARQ handling. Format N1, which has the same size
with different values for each parameter (i.e., aggregation as format N0, is used for the scheduling of one NPDSCH
level and number of repetitions) so that the scheduler has the codeword in one cell and random access procedure initiated
ability to avoid blocking and schedule many UEs. For UEs in by a NPDCCH order. Last, Format N2 is employed for paging.
enhanced coverage, where it is efficient to allocate the whole Here the contents only spans 14 bits with a too short but
bandwidth to one NPDCCH, this search space can be based significantly repeated messages to respect the reachability of
on different repetition levels only, whereas for UEs in normal all UEs under all coverage conditions.
coverage which there may be a lesser need for repetition, it Finally, DL transmissions to UEs in poor coverage may
could be based on fewer repetition levels but more aggregation require many repetitions and thereby block the DL for other
levels. When the number of candidates is concerned, NB- UEs. One way to mitigate the risk for blocking is to introduce
IoT is designed to monitor the minimum number of blind intermediate transmission gaps at well-defined time instants
decodes to reduce complexity and power consumption. Indeed, during long DL transmissions. During these gaps the scheduler
a maximum of 3 candidates can be monitored every subframe can transmit control and data information to other UEs. In
in case of no repetition and only 4 candidates are targeted our view, using a combination of transmission gaps and time
when repetition is employed. In addition, during the NPDCCH division multiplexing between data and control scheduling
monitoring, the UE is not required to simultaneously process allows NB-IoT to keep the basic DL scheduling unit in
any of the UL channels and/or other DL channels. frequency as full band (i.e., 12 subcarriers).

19

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

6) Narrowband Physical Downlink Shared Channel Rate CRC TBS

TBCC
(NPDSCH): To provide backward compatibility for in-band Matching attach Generation
NB-IoT operation, downlink transmission with 15 kHz subcar-
rier spacing is required for all the modes of operation. For this
reason, the NB-IoT downlink symbol duration, slot duration, QPSK Layer
Scrambler Repetition
and subframe duration are reused from legacy LTE. However, Modulation mapping
the NPDSCH processing has many updates to account for the
reduced complexity, low power, and extended coverage targets.
RF CP OFDM RE
Fig. 8 shows the downlink transmitter processing for NB-IoT
Processing insertion Generation Mapping
systems. There are couple of significant differences here. This
includes the maximum supported TBS size, utilizing only Tail-
Fig. 8. Block diagram for the shared channel processing in NB-IoT systems.
Biting Convolutional Coding (TBCC) for channel coding, no
HARQ processing, reduced modulation order, and applying
repetition with the proper scrambling. be achieved by adaptive transmissions. In addition, if RV is
Indeed, due to the reduced bandwidth, the maximum TBS introduced, the rate matching for TBCC for LTE will have to
size for downlink NB-IoT is not expected to be greater than be changed specifically for NB-IoT, significantly reducing the
that in MTC (i.e. 1000 bits). If no fragmentation is allowed at synergy with LTE. This may add to the difficulty in supporting
physical layer, the stand-alone system in case of one Tx an- NB-IoT in some legacy LTE base stations.
tenna with QPSK modulation can carry 2×(12×14−2×4) = Repetitions are proposed to be applied at subframe level. A
320 bits in a single subframe which is too far from a realistic maximum of 2048 repetitions are defined to support a target
supported data rate when coding is considered. Also, many SNR of -12.6dB. Each repetition has a unique scrambling
fragments would be required from the upper layers which sequence based on the time stamp of the carrying subframe.
introduce significantly low utilization due to the overhead. For This ensures a fully flexible support for mobility in addition
these reason, a sensible option is to allow transmission of the to the peak to average power variations. Similar to NPDCCH,
data with fragmentation at physical layer. After rate matching, the NPDSCH is mapped to entire subframes (i.e., multiple
it was agreed that the transmitted block size is divided into NPDSCHs are multiplexed in a Time Division Multiplexing
NF subframes which can be as maximum as 10. This enables (TDM) manner). Contiguous and non-contiguous resource
the extension of the TBS size to be 680 bits as maximum for allocations can be considered to trade-off latency with inter-
NB-IoT. Further, mapping a single transport block to multiple user blocking especially in the extreme coverage case. DL gaps
subframes would avoid a very high initial code rate. are defined for NPDSCH to reduce the blocking probability
Tail-Biting Convolutional Coding (TBCC) has replaced for UE users in deep coverage on the account of the user in
Turbo codes for simplicity. The main issue brought by this good coverage environments.
decision is the concern regarding any optimization towards the
support of high code rates. The TBCC is known to provide less
performance gain when compared to Turbo coding for high C. Uplink Physical Channels
code rates. This is a non-essential feature whose introduction Similar to MTC systems, in NB-IoT systems, a UE initially
would negatively impact the reuse of existing LTE procedures estimates its coverage level based on a path-loss estimate and
thus incurring increased development cost and time-to-market. use an NPRACH preamble according to its estimated coverage
In fact, utilizing TBCC agrees with the understanding that level. NPUSCH is initially set-up according to the UE’s
NB-IoT payloads are generally small and in addition, for estimated coverage level. However, if the network determines
IoT applications requiring larger transport block sizes, LTE that the UE’s path-loss estimate is poor, the network can recon-
already provisions the MTC enhancements. Moreover, QPSK figure a UE to a different NPUSCH coverage level based on
has been fully evaluated for coverage, capacity, latency and the eNodeB measurements of the uplink signal characteristics.
energy consumption, link level coexistence. It has been shown This is actually a fundamental update happening to the NB-
to fulfill the performance targets at low coverage modes. On IoT when compared to MTC [62][63]. To reduce complexity
the other hand, 16QAM has never been assumed for NB- and power consumptions, the UE is no longer asked to
IoT downlink to achieve the target SNR requirements with report local channel quality indication. For this reason, the
TBCC. For these reasons, QPSK has been selected to be the functionality of the uplink control channel becomes very thin
only supported modulation scheme for downlink NB-IoT and and it actually reduces to only sending ACK/NACK. Indeed,
enhanced coverage MTC. However, it was agreed to leave this task can be easily handled by the usual data channel
16QAM for further study in future releases. and hence no dedicated control channel is defined for NB-
To significantly reduce the complexity, only one HARQ IoT. Two uplink physical channels are only supported for
process is supported for NB-IoT with no Redundancy Version NB-IoT, namely the Narrowband Physical Random Access
(RV) reported. Actually, NB-IoT UEs are very likely to use Channel (NPRACH) and the Narrowband Physical Uplink
low code rates in both the uplink and the downlink in order to Shared Channel (NPUSCH).
support coverage extension, but RV is not shown to provide Since UEs in extreme coverage are power limited and
clear coding gains for low code rates (i.e., rate ≤1/3). Thus, for their performance is sensitive to power amplifier efficiency,
a single HARQ process, the benefits of RV (if any) can already a single tone transmission is desirable to achieve a Peak-to-

20

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Average Power Ratio (PAPR) that is close to 0dB irrespective Initial SC

index
of their coupling loss. Indeed, the 3GPP standardization body Symbol
Frequency
has established various studies from independent companies Group Repetition
Hopping
Generation
to evaluate and validate the NB-IoT system numerology. A
NPRACH
summary of the evaluation is presented in [64]. The results Frequency
Preamble Format
have considered the performance aspects, the capacity, the Offset
modes of operation, the UE complexity, the UE battery life
Fig. 9. Block diagram for the uplink random access channel processing in
or power consumption, the coverage, and the cell size. After NB-IoT systems. The transmission for NPRACH is applied only to 3.75KHz
analysing the tradeoff, it has been agreed that a single tone subcarrier spacing mode.
transmission is supported for uplink NB-IoT systems whether
by utilizing 15KHz or 3.75KHz as a subcarrier spacing. This is Inner hopping pattern Pseudo random outer hopping pattern

an interesting feature for NB-IoT in the uplink side that would

11
require significant revisions to the uplink frame structure and 10
various channel processing. It is worth to mention that the NB-

Relative subcarrier index in

NPRACH band of interest
9
IoT UE support for single carrier is mandatory. However, the 8
7
support of the legacy multi-tone transmission can be indicated
6
by the UE during the random access procedure. 5
1) Narrowband Physical Random Access Channel 4
(NPRACH): The random access procedure starts with random 3
2
access preamble transmission from the UE to eNodeB [65]. To 1
achieve enhanced coverage in NB-IoT system, the LTE-MTC 0
procedure has been partially adopted. The main difference is 0 1 2 3 4 5 6 7 8 9 10 11
the information about the time/frequency resources allocated Group index

for NPRACH. Thus, in general, a small number of PRACH

Fig. 10. Illustration for frequency hopping pattern in NPRACH. fixed inner
coverage levels is defined and the PRACH coverage enhance- hopping pattern within 4 groups is applied while pseudo random shift is
ment is achieved through repetition. Different coverage levels applied as an outer hopping sequence.
correspond to different NPRACH resources (i.e., different time
resources, different frequency resources, preamble set, and
repetition factors) and the available resources are signalled only a single cyclic prefix is added [54]. The symbols and
in system information messages [66]. The UE selects the the CP constitute the new introduced concept of symbol
initial NPRACH coverage level and associated NPRACH groups. The preamble consists of one symbol in time and
resource set based on its own DL measurement such as the occupies only one tone in the frequency domain with 3.75KHz
Reference Signal Received Power (RSRP) measurement [67]. subcarrier spacing. However, the NPRACH transmission dy-
The eNodeB can get a rough estimate of the UE coverage level namically hops from one tone to another within 12 adjacent
by observing the NPRACH coverage level for the successful subcarriers. Within the available 48 subcarriers, a subset of
NPRACH preamble transmission. If the UE fails to access the these subcarriers NNPRACH
sc is defined for a given cell to
network after the maximum number of attempts at the highest be the active NPRACH frequency resources. In addition, an
NPRACH repetition level, the UE should stop further attempts offset parameter is introduced to point to the first subcarrier
at the highest NPRACH repetition level and report the failure allowed for this cell. To select the initial subcarrier index
to higher layers. prior to hopping, the UE randomly selects a subcarrier index
Indeed, there exist various design tradeoffs in choosing for NPRACH within NNPRACHsc , adds the initial cell specific
the subcarrier spacing for the single tone frequency hopping offset to the selected number, and then use the resulting
NPRACH. For example, for a given configured NPRACH modulo-sum as the starting subcarrier index for the hopping
bandwidth, the larger the subcarrier spacing, the smaller the process. Actually, the number of allowed NPRACH frequency
number of subcarriers. Since different single tone frequency resources (i.e., NNPRACH
sc ) has been partitioned into two sets,
hopping NPRACH preambles are effectively separated in the where the subcarrier indices falling within the second set
frequency domain, using smaller subcarrier spacing implies indicate that the UE supports multi-tone transmission for the
a higher number of available preambles. On the other hand, RAR in addition to the default single-tone mode. Thus, the UE
utilizing larger subcarrier spacing may help improve the time- has to be careful with the subcarrier set from which the random
of-arrival estimation performance at the eNodeB and is less number is chosen based on its transmission capabilities. Fig. 9
sensitive to carrier frequency offset in the uplink. Since the shows the block diagram of the NPRACH transmitter at the UE
later issue is attached to eNodeB complexity not the UE side, side with the three main blocks: the hopping support, the group
it has been agreed to utilize the smallest possible subcarrier symbol generation, and repetition. Two preamble formats with
spacing, namely 3.75KHz, as the only mode of operation different sequence lengths are provided for NB-IoT system. As
for NPRACH [68]. With the limited 180KHz bandwidth, usual, the preamble format is selected based on the channel
this choice results in up to 48 different preambles from the conditions and coverage level.
frequency allocation perspective. Frequency hopping is essential for NPRACH in order to
To reduce the overhead, four symbols are combined and facilitate the Time of Arrival (ToA) estimation at the eN-

21

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

odeB from which a time advance is reported to the UE Table V

so that synchronization is achieved. To provide large ToA NPUSCH PARAMETERS IN NB-I OT SYSTEMS
range (i.e., large cell sizes) and accurate ToA estimation,
NPUSCH ∆f NRU
sc NRU
slots NRU Qm DMRS
a two layer hopping pattern is utilized for NPRACH [69]. Format indices
The inner hopping is chosen to be fixed for simplicity and 3.75KHz 1 16 4
small to provide large ToA range. Indeed, hopping is applied {1,2}
1 16 {1,2,3,
among four symbol groups with a fixed pattern to ensure a
1 3 8 4,5,6,
ToA estimation range of about 8km cell size when utilizing 15KHz 2 3
3.75KHz subcarrier spacing. On the other hand, an output 6 4 7,8,10}
hopping layer is employed to enhance the ToA estimation 12 2
accuracy. For this layer, a pseudo random hopping pattern 3.75KHz 1 4 0,1,2
2 1 1
in reused from the legacy PUSCH hopping. A cell specific 15KHz 1 4 2,3,4
random pattern is utilized to mitigate the cell interference.
Without pseudo random hopping, the NPRACH transmissions
in one cell may cause persistent interference to the NPRACH be active during NPUSCH format 1 transmission. Indeed, the
and/or NPUSCH transmissions in the neighbouring cells. In concept of subframe can be still used in this case because
addition, an NPRACH capacity reduction is expected due to actually, in this mode, nothing has changed when compared
the fact that neighbouring cells configure different frequency to legacy LTE.
resources for NPRACH to avoid inter-cell interference. This To maintain compatibility and flexibility in the description
reduction can be somehow compensated if random hopping is for both 3.75KHz or 15KHz modes, the concept of Resource
applied. Last, the pseudo random nature of the outer hopping Unit (RU) is developed for NB-IoT NPUSCH to provide the
introduces flexibility to the NPRACH assignments. Indeed, essential framework for resource mapping [70]. A resource
a single cell may configure different NPRACH bandwidths unit is defined to be Nsymb NRU slots consecutive SC-FDMA
not just a single number of active subcarriers, NNPRACH
sc . symbols in the time domain and NRU sc consecutive subcarriers
NPRACH transmission with one-level fixed hopping plus in the frequency domain, where Nsymb = 7 is the number
additional pseudo random hopping can be readily scaled as of SC-FDMA symbols per slot. Similar to the downlink,
the bandwidth increases [63]. Fig. 10 shows one example fragmentation is allowed at the physical layer to segment the
to illustrate the frequency hopping feature for the NPRACH data packet into multiple units, namely NRU . As a special
channel. case, NPUSCH Format 2 always employs the smallest possible
2) Narrowband Physical Uplink Shared Channel transmission unit to carry the ACK/NACK message (i.e., a
(NPUSCH): Due to the introduction of the single carrier single RU is utilized or NRU =1). Table V summarizes the
3.75KHz mode, the symbol period has been expanded to different parameters for NPUSCH under different supported
four times its original value. Accordingly, the frame structure modes.
for uplink NB-IoT system has been revised. Indeed, for In case of Format 1, the transport block size is selected
NPUSCH in 3.75KHz mode, the radio frame consists of only based on modulation, coding, and the assigned number of sub-
5 slots numbered from 0 to 4 where each slot spans 2msec carriers per symbol whenever multi-tone transmission is con-
period. The subframe concept is dropped in this case and sidered. After Cyclic Redundancy Check (CRC) attachment,
only slot handling is considered. The slot still consists of the code block is encoded. Turbo coding and rate matching
7 SC-FDMA symbols such that the OFDM symbol length schemes in legacy LTE have been reused for NPUSCH [55],
becomes 512 samples when sampled at 1.92MHz. Unlike the because it provides significantly higher coding gain than
classical LTE system for Normal CP type, the cyclic prefix convolutional coding. Although the decoding complexity of
length for NPUSCH in 3.75KHz mode has a uniform period turbo coding is higher than that of convolutional coding,
of 16 samples [62]. Thus, the remaining 144 samples within the decoding is performed at the eNodeB side, where high
the slot, when compared to the legacy 960 samples slot complexity is considered acceptable. During rate matching,
period, are left as guard time to accommodate for a possible the number of available RUs is assumed. One HARQ pro-
collision with any sounding reference signal from the legacy cess for dedicated transmissions is supported for uplink. To
LTE system. minimize the downlink signalling overhead, it has been ap-
The NPUSCH channel has two formats, namely Format 1 proved that only two RVs, namely 0 and 2, are supported
and Format 2 [63]. The first format is the conventional data for NPUSCH. Due to the wide variation of the transmission
transmission mode of the NPUSCH channel. However, the duration between different coverage levels, it is hard to fix the
second format is dedicated to carry the control information time interval between first transmission and re-transmission.
represented in only ACK/NACK. It is mandatory for a UE Consequently, asynchronous HARQ is utilized. However, after
to transmit both formats through a single-tone configuration. Midentical codeword transmissions, the RV index is toggled
However, format 1 may be also transmitted through a multi- within the same packet transmission if repetition is enabled,
tone structure only when conventional 15KHz spacing is where Midentical = min(4, NRep ) and NRep is the repetition
utilized and when the UE capabilities permit. In multi-tone count [63].
scenario, there are three possible configurations to allocate After modulation, repetitions are applied at slot/subframe
the frequency subcarriers. Only 3, 6, or 12 subcarriers can level. The number of supported repetitions for NPUSCH are

22

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

Payload Resource
CRC Channel Symbol Transform SC-FDMA
Scrambler Element
Attach interleaver Repetition Precoding Generation
Mapping

Turbo Rate BPSK/ Reference

Encoder & Matching & QPSK Symbols Multi-Tone
interleaver HARQ Mapper Generation transmission
(15KHz mode)

ACK/
NACK Resource Symbol
Repetition BPSK Symbol
Scrambler Element Phase
Encoder Mapper Repetition
Mapping Insertion

Single-Tone
Reference
transmission SC-FDMA
Symbols
(3.75KHz or Generation
Generation
15KHz modes)

Fig. 11. Block diagram for the uplink shared channel processing in NB-IoT systems.

{1, 2, 4, 8, 16, 32, 64, 128}, where the repetition count through the DCI format N1. For both subcarrier spacings, the
is indicated in the DCI [70]. The repeated slots are then DMRS density is adopted from the legacy control channel
transmitted over a single-tone chain or a multi-tone chain. It Format 1/1a, which uses three DMRS symbols per slot.
was agreed that QPSK and BPSK are supported for single-
tone NPUSCH transmission where phase rotation is introduced VI. I MPLEMENTATION C HALLENGES
to reduce PAPR and out-of-band emission. To obtain the A. Low Power Support
benefits, phase rotation should be implemented in a contiguous
way at symbol level including both data symbol and DMRS Battery longevity depends on how efficiently a device can
symbols. On the contrary, the multi-tone processing follows utilize various idle and sleep modes that allow large parts of
the conventional process of employing a transform precoder to the device to be powered down for extended periods. The NB-
support SC-FDM, a resource element mapper to fill the time- IoT specification addresses the physical layer technology and
frequency grid with the proper resource elements, and finally idling aspects of the system. Like LTE, NB-IoT uses two main
synthesize the time domain baseband signal by utilizing the protocol states: IDLE and CONNECTED. In IDLE mode,
SC-FDMA generation block. The detailed block diagram for devices save power, and resources that would be used to send
the NPUSCH is shown in Fig. 11. measurement reports and uplink reference signals are freed
up. In CONNECTED mode, devices can receive or send data
In [71], simulations were performed to investigate the link directly. Discontinuous reception (DRX) is the process through
level performance of NPUSCH assuming different DMRS which networks and devices negotiate when devices can sleep
densities. Results show that by increasing the DMRS density and can be applied in both IDLE and CONNECTED modes.
from 1 symbol per slot (i.e. the same as in legacy LTE) to For CONNECTED mode, the application of DRX reduces the
2 symbols per slot, the block error rate performance is only number of measurement reports devices send and the number
improved by 0.5 dB for low mobility channel. The gain cannot of times downlink control channels are monitored, leading to
even compensate for the loss due to the decrease of data battery savings. 3GPP R12 supports a maximum DRX cycle
symbol density from 6 to 5 per slot. In fact, the target Mutual of 2.56 seconds, which has been extended to 10.24 seconds in
Coupling Loss (MCL) of 164dB [33] is comfortably met for R13 (eDRX). However, any further lengthening of this period
NPUSCH by reusing the legacy LTE DMRS density. Hence, is as yet not feasible, as it would negatively impact a number of
it was agreed that NPUSCH reuses the LTE uplink DMRS RAN functions including mobility and accuracy of the system
pattern for the 15 kHz subcarrier spacing. For the 3.75 KHz information. In IDLE mode, devices track area updates and
subcarrier spacing, a similar pattern as legacy LTE DMRS can listen to paging messages. To set up a connection with an
be employed for DMRS. The little difference from legacy LTE idle device, the network pages it. Power consumption is much
DMRS is to avoid the possible collision between DMRS and lower for idle devices than for connected ones, as listening for
LTE sounding signal. pages does not need to be performed as often as monitoring
In case of NPUSCH Format 2, the number of subcarriers the downlink control channel.
available per RU is always 16. Therefore, the single bit Compared with legacy LTE, the link budget of NB-IoT
representing ACK/NACK is encoded by a repetition code has a 20dB margin, and use cases tend to operate with
to 16 bits which are modulated by BPSK modulation [55]. lower data rates. The coverage target of NB-IoT has a link
When repetition is enabled, the symbols are mapped to the budget of 164dB, whereas the current LTE is 142.7 dB.
available consecutive NAN AN
Rep resource units, where NRep is The 20dB improvement corresponds to a sevenfold increase
the repetition factor for the ACK/NACK message. Of course, in coverage area for an open environment, or roughly the
the ACK/NACK resource assignment is signalled to the UE loss that occurs when a signal penetrates the outer wall of

23

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

a building. Standardization activities in 3GPP have shown 1

that NB-IoT meets the link budget target of 164dB, while 0.9
simultaneously meeting the M2M application requirements for SNR=−6dB
0.8
data rate, latency, and battery life.
0.7 SNR=−15dB

SSS Detection Rate

0.6
B. Low Cost Support
0.5
NB-IoT devices support reduced peak physical layer data
rates: in the range of 100-200kbps or significantly lower for 0.4

single-tone devices. To facilitate low-complexity decoding in 0.3 Coherent SSS AWGN

devices, turbo codes are replaced with convolutional codes for 0.2
Coherent SSS EPA−5
Diff. SSS AWGN
downlink transmissions, and limits are placed on maximum Diff. SSS EPA−5
0.1
transport block size (which is 680 bits for DL and 1000
0
bits for UL). The performance requirements set for NB-IoT 0 50 100 150
make it possible to employ a single receiver antenna. As Processing Time in msec

a result, the radio and baseband demodulator parts of the Fig. 12. Detection probability versus processing time for SSS under different
device need only a single receiver chain. By operating NB- detection techniques. Different channel conditions are considered with SNR=-
IoT devices in half duplex so that they cannot be scheduled to 6dB and -15dBs [79].
send and receive data simultaneously, the duplex filter in the
device can be replaced by a simple switch, and only a single
number of antenna ports and the operating SNR, conventional
local oscillator for frequency generation is required. In fact,
cell search and initial synchronization techniques have to be
these optimizations reduce cost and power consumption. At
revised. One fundamental approach is defined in [79] by apply-
200kHz, the bandwidth of NB-IoT is substantially narrower
ing time-averaging over different decision statistics to enhance
than other access technologies including LTE-MTC systems.
the accuracy of the legacy techniques. For example, Fig. 12
The benefit of a narrowband technology lies in the reduced
shows the coherent SSS detection (where the channel state
complexity of analog-to-digital conversion (ADC), digital-to-
information is assumed to be known) versus the differential
analog conversion (DAC), reduced number of HARQ pro-
SSS detection for the legacy LTE under different averaging
cesses, and subframe buffering. Since all physical channels
periods. It is clear that detection accuracy can be enhanced by
utilize a unified transmission mode, channel estimation can be
increasing the averaging period. However, high Doppler effects
adopted for all subframes unlike LTE-MTC in which coherent
have not been addressed. Research efforts are encouraged in
decoding for the data channel requires a stand-alone channel
this direction to provide innovative low-cost solutions for this
estimation rather than the control channel one. All these
harsh environment.
features and others bring benefits in terms of low cost and
2) Frequency Tracking in CAT-M: Even after the initial
low power consumption. NB-IoT brings about a significant
synchronization stage, frequency tolerance is always present
design change in terms of the placement of the device’s power
due to the Doppler shift and uncompensated residual errors,
amplifier (PA). Integrating this element directly onto the chip,
hence frequency tracking is required. As an OFDM system,
instead of it being an external component, enables the cheaper
legacy LTE systems can employ the conventional Maximum-
single-chip modem implementations.
Likelihood estimation and compensation loops for the fre-
quency tracking [80][81][82]. Two issues arise with the in-
C. Challenges and Implementation Aspects troduction of LTE-MTC: (1) During EDRX cycles, frequency
1) Initial Synchronization in CAT-M: From CAT-M per- tracking is deactivated and DRX wake-up procedure requires
spective, in literature, there are various techniques presented to re-synchronization. (2) Enhanced coverage UEs are required
perform initial synchronization, cell search, frequency track- to track frequency errors in very low SNR regimes. A new
ing, and typical chain decoding. During initial synchronization challenge is then introduced to keep tracking loops with high
and cell search, all techniques [72][73][74][75][76][77] share accuracy at these low SNRs. In this environment, the legacy
the same procedure in the following order: (1) A coarse sym- techniques can be inefficient and hence new approaches are
bol timing has to be obtained first so that the received signal also encouraged. As an initial solution, the authors in [83] have
can be converted from time domain to frequency domain. defined a frequency tracking approach through the repetitive
At this stage, there are algorithms to estimate the fractional nature of the broadcast channel mapping which is utilized ba-
part of CFO as well [72][73]. (2) PSS (or sector ID) will be sically for the enhanced coverage support. However, the most
detected in the second step. (3) SSS (or cell ID group) will recent specifications left it open for the broadcast repetition to
be found next [74][75][78]. (4) The detection of the integer be cell-specific feature. That is, broadcast channel repetition
part of CFO can be fulfilled. Some algorithms have been may not be employed at least for 1.4MHz cell.
presented to enable this estimation within either Step 2 or Step 3) Channel Estimation in CAT-M: When it comes to vari-
3 [76]. Other algorithms are based on different time domain ous chain processing, channel estimation and equalization are
approaches [77]. the main challenges for LTE-MTC systems. Conventionally,
Since LTE-MTC introduces new challenges to the system in LTE systems, reference signals are inserted within the
represented in the limited degrees of freedom regarding the transmitted signal to assist the channel estimation process

24

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

which is required for coherent detection. Generally, several synchronization, carrier frequency offset (CFO) is estimated
channel estimation schemes that vary in their complexity and and compensated to enable proper signal detection. The UE
performance have been presented [84], [85]. A 2-D Minimum- acquires the cell ID by employing the cell search procedure.
Mean Squared Error (MMSE) channel estimation technique To cope with these changes, NB-IoT employs new set of
has been introduced in [86]. However the proposed channel synchronization signals, namely NPSS and NSSS. The new
estimation technique depends on the knowledge of the channel sequences have different bandwidth, mapping, periodicity, and
statistics and the operating SNR which are usually unknown. generation when compared to the legacy LTE synchronization
A robust MMSE channel estimation technique has been in- signals. Unlike conventional LTE, cell ID is encapsulated
troduced in [85], [86] in order to remove the dependency only in the secondary sequence without involving the primary
on the exact channel statistics. In this case, filter coefficients sequence. One of the challenges is to acquire the initial
are calculated based on worst-case channel conditions and timing, frequency acquisition, and efficiently search for the
are used in the estimation process at the expense of minor serving cell ID. Although it looks straightforward to apply
performance degradation. Another simple estimation scheme cross-correlation with the known reference for these detection
can be used for systems that have uniformly distributed hypothesis, the detection performance and complexity are the
pilots, and where pilot separation satisfies the sampling theory main challenges for such algorithms especially under low SNR
conditions. In this case, an ideal low-pass filter can be used in and various channel conditions [91]. To highlight the issue,
order to reconstruct the original signal from its samples [87]. we have simulated the NB-IoT system to detect the cell ID
Again, the performance measures for these techniques have under perfect synchronization environment. Differential cross-
been designed and optimized for the legacy operating SNR correlation is employed to reduce the effect of the channel.
values and under the assumption that high speed channels are However, Fig. 13 shows a significant degradation in the
totally supported. Conversely, LTE-MTC would require special detection performance when the Channel State Information
treatment to maintain enough performance at low complexity (CSI) is available, even with decision statistic averaging across
and deep coverage conditions. M windows or radio frames.
It is to be noted that, special reference signal is attached One more challenge is to maintain the UE in synchroniza-
to the new introduced MPDCCH channel to help pilot-aided tion with the eNodeB at very low SNR regimes. There are two
channel estimation techniques. In [88], the channel power- aspects for this issue. First, the basic cell search may encounter
delay profile is approximated to be only described by the a loss in detection. Since the NSSS sequence, that carries
mean delay and the root-mean-square delay spread. A Linear the Cell ID information, spreads across consecutive OFDM
MMSE filter is introduced to estimate those parameters for symbols, the effect of the fractional frequency offset has to be
the LTE Multi-Input Multi-Output (MIMO) system where carefully investigated. For example, with only 1KHz frequency
the user specific pilot pattern (i.e., the same pilot structure offset and subcarrier spacing of 15KHz, the phase over one slot
utilized by MPDCCH) is employed. The main issue with this (i.e., 960 samples at 1.92MSamples/sec rate) would change by
approach is the complexity. For the same pilot structure, an 2π×960×(1/15)/128 = π, meaning that all NSSS subcarriers
iterative 2-D MMSE channel estimation method is considered contained by the second slot will experience a flipped sign
in a multi-user environment [89]. The approach aims not when compared to the original transmitted sequence. This will
only to cancel the interference effect, but also to enhance certainly reduce the detection accuracy. Therefore, a special
the channel estimates iteratively by introducing a mean-square attention has to be paid for the synchronization assumptions
error criterion after each iteration. Again, the technique was for cell search and data decoding. Second, when compared to
not designed for a single user downlink reception purpose legacy LTE, the number of reference symbols for DL NB-IoT
and it is somehow complex. To the best of the authors is insufficient to utilize the conventional pilot-based frequency
knowledge, practical channel estimators for MPDCCH have tracking mechanisms [82]. Furthermore, the repetition struc-
not been addressed. One can guess that such process has ture for the mapped symbols are performed on the subframe
been addressed before and no revisions are required since the level unlike MTC in which the broadcast channel mapping has
DMRS structure for MPDCCH is inherited from the classical a repetitive structure. This MTC PBCH repetitive structure
EPDCCH channel [88][90]. However, there are couple of enables the differential phase to track the frequency offset
challenges introduced to MPDCCH that would require this and hence maintain the UE in synchronization [83]. On top
revision. (1) The operating SNR for MPDCCH has been of that, the NB-IoT system supports a multi-carrier operation
reduced to -15dBs. (2) Repetition is supported for MPDCCH in which the eNodeB can dynamically switch NB-IoT UE to
and hence channel estimators quality can be enhanced. (3) another band in which neither NPSS nor NSSS is present. In
The channel estimator has to consider frequency hopping. (4) this case, reference signals are the only data aided symbols to
Complexity is a real issue for MTC framework, thus reduced be utilized for synchronization. Research effort is encouraged
complexity is essential for the channel estimators. For these to investigate such an issue.
reasons, channel estimation for MPDCCH requires a revision 5) Channel Estimation in NB-IoT: Although the main
and new proposals. theme of the NB-IoT devices to be stationary due to the
4) Synchronization and Frequency Tracking of NB-IoT: significant reduced MCL requirements, the specifications left
NB-IoT intends to occupy a narrow bandwidth of only it open for the vendors to support mobility with high Doppler
200KHz, which is not backward compatible to the sup- spread [63]. It is unforgotten that even with good coverage
ported bandwidths by the legacy LTE [61]. During initial NB-IoT UE from the power perspective, the UE losses both

25

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

1
UEs induces a huge amount of information exchange over-
head and affects the UE power consumption and complexity.
0.8 The efficiency of multicasting with massive MIMO, in non-
Cell ID Detection Rate

M=1 cooperative setups, is also a limiting factor [99]. As the

0.6 number of base station antennas increases the number of the
M=10 channel estimates also increases for each terminal, which in
0.4 turn needed hundred times more uplink slots to feedback the
Diff. Corr. EPA channel responses to the base station [100].
Diff. Corr. AWGN
0.2
Known CSI EPA
M=20 B. 5G enabling Technologies
Known CSI AWGN
0 Despite the spent effort to enhance and optimize the current
−16 −14 −12 −10 −8 −6 −4 −2 0
SNR in dBs networks and their interfaces to accommodate the increas-
ing demand of IoT applications, other directions should be
Fig. 13. Simulation for the cell detection for NB-IoT system for two scenarios
and under different fading conditions. A differential cross-correlation is
explored to provide alternative technologies to enable the
compared to a direct correlation with known CSI. Various averaging windows 5G. For example, all the current communication systems
are utilized to show the effect of averaging on the decision statistics. use bi-directional communications without time or frequency
duplex which limits the capabilities of the devices. Full-Duplex
breaks such barrier by transmitting and receiving at the same
frequency and spatial diversities due to the reduced cost and time and on the same frequency, introducing the potential
reduced bandwidth features. Therefore, channel estimation al- of doubling the system capacity and reducing the system
gorithms have to be revised to provide reasonable performance delay. Also, the current multiple access techniques schedule
at very low SNR values while supporting mobility. Indeed, it the users on orthogonal channels i.e. different time and/or
is always the tradeoff between performance and complexity frequency which greatly limits the system capacity. A newly
to design such channel estimation algorithm. However, in proposed technique that can overcome this limitation is the
NB-IoT, reduced complexity is a key target and hence the nonorthogonal multiple access (NOMA). Such new techniques
compromise has to be carefully investigated and analysed. can benefit the development process of new radio access for
the next generation networks.
VII. I OT AND F UTURE D IRECTIONS
VIII. C ONCLUSION
A. Air Interface Scalability The new applications introduced by the IoT framework
As MTC and IoT require supporting tens of thousands of forced new challenges on the existing cellular technologies.
connected devices in a single cell, LTE cellular networks While such technologies are initially designed to support
should explore new avenues of research and development, such human-generated traffic, new designs are proposed to ac-
as massive MIMO [92][93] to help the network air interface commodate for the emerging machine-generated traffic which
to accommodate such increased number of connections. In has different characteristics. In this paper, we addressed in
essence, massive MIMO is an evolving technology resulting details the development on the LTE to support MTC and IoT.
from up-scaling the traditional MIMO systems. In massive New UE categories has to be introduced to support the new
MIMO system, large-size arrays can serve a massive number requirements of each system, namely, CAT-M and CAT-N.
of user terminals using spatial multiplexing [94]. As one of Each category has a distinct set of specifications and limi-
the most promising ingredients of the emerging 5G technology, tations. While traditional coverage enhancements techniques
massive MIMO is a commercially attractive solution since a are not applicable in the new categories, the new alternatives
hundred-fold higher capacity is possible with the same number have been discussed to extract diversity gain through time
of base stations. Such gain is made possible due the recent repetitions and frequency hopping. Some challenges to the
advances in directional beamforming with low power, and implementation to support low cost and low power operations
flexible beam adjusting with low peak-to-average power ratio are also discussed.
in MIMO-OFDM systems [95][96][97][98].
However, some challenges are associated with such increase R EFERENCES
in number of antennas is case of MTC. For example, when [1] A. Ali, W. Hamouda, and M. Uysal, “Next generation M2M cellular
large number of UEs are very close to each other, the sys- networks: challenges and practical considerations,” IEEE Commun.
Mag., vol. 53, no. 9, pp. 18–24, September 2015.
tem cannot render such highly correlated channel vectors as [2] V. Gazis, “A Survey of Standards for Machine to Machine (M2M) and
orthogonal which will affect the massive MIMO performance, the Internet of Things (IoT),” IEEE Commun. Surveys Tuts., vol. PP,
therefore, user-scheduling algorithms should be aware of such no. 99, pp. 1–1, 2016.
[3] Z. Fan, R. J. Haines, and P. Kulkarni, “M2M communications for E-
fact. health and smart grid: an industry and standard perspective,” IEEE
Moreover, inexpensive and low-power components still need Wireless Commun., vol. 21, no. 1, pp. 62–69, February 2014.
more investigations to build such massive MIMO [92]. Chan- [4] L. Atzori, A. Iera, and G. Morabito, “The Internet
of Things: A survey,” Computer Networks, vol. 54,
nel state information, associated with large number of antennas no. 15, pp. 2787 – 2805, 2010. [Online]. Available:
for massive MIMO systems and coordination among different http://www.sciencedirect.com/science/article/pii/S1389128610001568

26

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

[5] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and [26] T. Myers, D. Werner, K. Sinsuan, J. Wilson, S. Reuland, P. Singler,
M. Ayyash, “Internet of Things: A Survey on Enabling Technologies, and M. Huovila, “Light monitoring system using a random phase
Protocols, and Applications,” IEEE Commun. Surveys Tuts., vol. 17, multiple access system,” Jul. 2 2013, uS Patent 8,477,830. [Online].
no. 4, pp. 2347–2376, Fourthquarter 2015. Available: https://www.google.com/patents/US8477830
[6] B. Panigrahi, H. K. Rath, R. Ramamohan, and A. Simha, “Energy and [27] S.-Y. Lien, K.-C. Chen, and Y. Lin, “Toward ubiquitous massive ac-
spectral efficient direct Machine-to-Machine (M2M) communication cesses in 3GPP machine-to-machine communications,” IEEE Commun.
for cellular Internet of Things (IoT) networks,” in Int. Conference on Mag., vol. 49, no. 4, pp. 66–74, 2011.
Internet of Things and Applications, Jan 2016, pp. 337–342. [28] E. Soltanmohammadi, K. Ghavami, and M. Naraghi-Pour, “A Survey
[7] A. Damnjanovic, J. Montojo, Y. Wei, T. Ji, T. Luo, M. Vajapeyam, of Traffic Issues in Machine-to-Machine Communications over LTE,”
T. Yoo, O. Song, and D. Malladi, “A survey on 3GPP heterogeneous IEEE Internet Things J., vol. PP, no. 99, pp. 1–1, 2016.
networks,” IEEE Wireless Commun., vol. 18, no. 3, pp. 10–21, June [29] A. Rico-Alvarino, M. Vajapeyam, H. Xu, X. Wang, Y. Blankenship,
2011. J. Bergman, T. Tirronen, and E. Yavuz, “An overview of 3GPP en-
[8] M. Levesque, F. Aurzada, M. Maier, and G. Joos, “Coexistence Anal- hancements on machine to machine communications,” IEEE Commun.
ysis of H2H and M2M Traffic in WiFi Smart Grid Communications Mag., vol. 54, no. 6, pp. 14–21, June 2016.
Infrastructures Based on Multi-Tier Business Models,” IEEE Trans. [30] 3GPP, “Standarization of Machine-type Communications,” 3rd Gener-
Commun., vol. 62, no. 11, pp. 3931–3942, Nov 2014. ation Partnership Project, Tech. Rep. V0.2.4, June 2014.
[9] T. Adame, A. Bel, B. Bellalta, J. Barcelo, and M. Oliver, “IEEE [31] A. Rajandekar and B. Sikdar, “A Survey of MAC Layer Issues and
802.11AH: the WiFi approach for M2M communications,” IEEE Protocols for Machine-to-Machine Communications,” IEEE Internet
Wireless Commun., vol. 21, no. 6, pp. 144–152, December 2014. Things J., vol. 2, no. 2, pp. 175–186, April 2015.
[10] T. Taleb and A. Kunz, “Machine type communications in 3GPP [32] S. Chen, S. Sun, Y. Wang, G. Xiao, and R. Tamrakar, “A comprehen-
networks: potential, challenges, and solutions,” IEEE Commun. Mag., sive survey of TDD-based mobile communication systems from TD-
vol. 50, no. 3, pp. 178–184, March 2012. SCDMA 3G to TD-LTE(A) 4G and 5G directions,” China Commun.,
[11] X. Xiong, K. Zheng, R. Xu, W. Xiang, and P. Chatzimisios, “Low vol. 12, no. 2, pp. 40–60, Feb 2015.
power wide area machine-to-machine networks: key techniques and [33] R. Ratasuk, B. Vejlgaard, N. Mangalvedhe, and A. Ghosh, “NB-IoT
prototype,” IEEE Commun. Mag., vol. 53, no. 9, pp. 64–71, September system for M2M communication,” in IEEE Wireless Commun. and
2015. Networking Conference Workshops, April 2016, pp. 428–432.
[12] U. Raza, P. Kulkarni, and M. Sooriyabandara, “Low power wide area [34] M. Wang, J. Zhang, B. Ren, W. Yang, J. Zou, M. Hua, and X. You, “The
networks: An overview,” IEEE Commun. Surveys Tuts., vol. PP, no. 99, Evolution of LTE Physical Layer Control Channels,” IEEE Commun.
pp. 1–1, 2017. Surveys Tuts., vol. 18, no. 2, pp. 1336–1354, Secondquarter 2016.
[13] W. Guibene, K. E. Nolan, and M. Y. Kelly, “Survey on Clean Slate [35] C. Shahriar, M. L. Pan, M. Lichtman, T. C. Clancy, R. McGwier,
Cellular-IoT Standard Proposals,” in 2015 IEEE Int. Conf. on Computer R. Tandon, S. Sodagari, and J. H. Reed, “PHY-Layer Resiliency in
and Information Technology; Ubiquitous Computing and Commun.; OFDM Communications: A Tutorial,” IEEE Commun. Surveys Tuts.,
Dependable, Autonomic and Secure Computing; Pervasive Intelligence vol. 17, no. 1, pp. 292–314, Firstquarter 2015.
and Computing, Oct 2015, pp. 1596–1599. [36] G. Ku and J. M. Walsh, “Resource Allocation and Link Adaptation in
LTE and LTE Advanced: A Tutorial,” IEEE Commun. Surveys Tuts.,
[14] J. P. Bardyn, T. Melly, O. Seller, and N. Sornin, “IoT: The era of
vol. 17, no. 3, pp. 1605–1633, thirdquarter 2015.
LPWAN is starting now,” in ESSCIRC Conf. 2016: 42nd European
[37] G. Araniti, C. Campolo, M. Condoluci, A. Iera, and A. Molinaro, “LTE
Solid-State Circuits Conf., Sept 2016, pp. 25–30.
for vehicular networking: a survey,” IEEE Commun. Mag., vol. 51,
[15] F. Ghavimi and H. H. Chen, “M2M communications in 3GPP
no. 5, pp. 148–157, May 2013.
LTE/LTE-A Networks: Architectures, Service Requirements, Chal-
[38] A. Laya, L. Alonso, and J. Alonso-Zarate, “Is the Random Access
lenges, and Applications,” IEEE Commun. Surveys Tuts., vol. 17, no. 2,
Channel of LTE and LTE-A Suitable for M2M Communications? A
pp. 525–549, Secondquarter 2015.
Survey of Alternatives,” IEEE Commun. Surveys Tuts., vol. 16, no. 1,
[16] Sigfox. (2017). [Online]. Available: http://www.sigfox.com/
pp. 4–16, First 2014.
[17] LoRa. (2017). [Online]. Available: https://www.lora-alliance.org/What- [39] N. Abu-Ali, A. E. M. Taha, M. Salah, and H. Hassanein, “Uplink
Is-LoRa/Technology scheduling in LTE and LTE-Advanced: Tutorial, Survey and Evaluation
[18] Semtech. (2017). [Online]. Available: http://www.semtech.com/ Framework,” IEEE Commun. Surveys Tuts., vol. 16, no. 3, pp. 1239–
[19] J. Petajajarvi, K. Mikhaylov, A. Roivainen, T. Hanninen, and M. Pet- 1265, Third 2014.
tissalo, “On the coverage of LPWANs: range evaluation and channel [40] F. Capozzi, G. Piro, L. A. Grieco, G. Boggia, and P. Camarda,
attenuation model for lora technology,” in 2015 14th Int. Conf. on ITS “Downlink Packet Scheduling in LTE Cellular Networks: Key Design
Telecommunications (ITST), Dec 2015, pp. 55–59. Issues and a Survey,” IEEE Commun. Surveys Tuts., vol. 15, no. 2, pp.
[20] K. E. Nolan, W. Guibene, and M. Y. Kelly, “An evaluation of low 678–700, Second 2013.
power wide area network technologies for the Internet of Things,” in [41] M. A. Mehaseb, Y. Gadallah, A. Elhamy, and H. Elhennawy, “Clas-
2016 Int. Wireless Commun. and Mobile Computing Conf. (IWCMC), sification of LTE Uplink Scheduling Techniques: An M2M Perspec-
Sept 2016, pp. 439–444. tive,” IEEE Commun. Surveys Tuts., vol. 18, no. 2, pp. 1310–1335,
[21] T. PetriÄĞ, M. Goessens, L. Nuaymi, L. Toutain, and A. Pelov, Secondquarter 2016.
“Measurements, performance and analysis of LoRa FABIAN, a real- [42] M. Z. Shafiq, L. Ji, A. X. Liu, J. Pang, and J. Wang, “Large-
world implementation of LPWAN,” in 2016 IEEE 27th Annual Int. scale measurement and characterization of cellular machine-to-machine
Symp. on Personal, Indoor, and Mobile Radio Commun. (PIMRC), Sept traffic,” IEEE/ACM Trans. Netw., vol. 21, no. 6, pp. 1960–1973, Dec
2016, pp. 1–7. 2013.
[22] P. Neumann, J. Montavont, and T. NoÃńl, “Indoor deployment of [43] X. Jian, X. Zeng, Y. Jia, L. Zhang, and Y. He, “Beta/m/1 model for
low-power wide area networks (LPWAN): A LoRaWAN case study,” machine type communication,” IEEE Commun. Lett., vol. 17, no. 3,
in 2016 IEEE 12th Int. Conf. on Wireless and Mobile Computing, pp. 584–587, March 2013.
Networking and Commun. (WiMob), Oct 2016, pp. 1–8. [44] M. S. Ali, E. Hossain, and D. I. Kim, “LTE/LTE-A random access for
[23] Ingenu Tech. (2017) RPMA technology for massive machine-type communications in smart cities,” IEEE Commun.
the Internet of Things. [Online]. Available: Mag., vol. 55, no. 1, pp. 76–83, January 2017.
http://theinternetofthings.report/Resources/Whitepapers/4cbc5e5e- [45] M. Laner, P. Svoboda, N. Nikaein, and M. Rupp, “Traffic models for
6ef84455b8cdf6e3888624cb_RPMA%20Technology.pdf machine type communications,” in ISWCS 2013; 10th Int. Symp. on
[24] IEEE, “IEEE standard for local and metropolitan area networks– part Wireless Communication Systems, Aug 2013, pp. 1–5.
15.4: Low-rate wireless personal area networks (lr-wpans)–amendment [46] 3GPP-TR, “Study on ran improvements for machine-type communica-
5: Physical layer specifications for low energy, critical infrastructure tions (release 11),” Technical specification group radio access network,
monitoring networks.” IEEE Std 802.15.4k-2013 (Amendment to IEEE Tech. Rep. 37.868, 2011.
Std 802.15.4-2011 as amended by IEEE Std 802.15.4e-2012, IEEE Std [47] A. Laya, C. Kalalas, F. Vazquez-Gallego, L. Alonso, and J. Alonso-
802.15.4f-2012, IEEE Std 802.15.4g-2012, and IEEE Std 802.15.4j- Zarate, “Goodbye, aloha!” IEEE Access, vol. 4, pp. 2029–2044, 2016.
2013), pp. 1–149, Aug 2013. [48] A. Laya, L. Alonso, and J. Alonso-Zarate, “Is the random access
[25] M. Centenaro, L. Vangelista, A. Zanella, and M. Zorzi, “Long-Range channel of lte and lte-a suitable for m2m communications? a survey of
Communications in Unlicensed Bands: the Rising Stars in the IoT and alternatives,” IEEE Commun. Surveys Tuts., vol. 16, no. 1, pp. 4–16,
Smart City Scenarios,” IEEE Wireless Commun., vol. 23, October 2016. First 2014.

27

1553-877X (c) 2017 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2017.2728013, IEEE
Communications Surveys & Tutorials

[49] J. Choi, “On the adaptive determination of the number of preambles [78] C.-C. Liao, P.-Y. Tsai, and T.-D. Chiueh, “Low-Complexity Cell Search
in rach for mtc,” IEEE Commun. Lett., vol. 20, no. 7, pp. 1385–1388, Algorithm for Interleaved Concatenation ML-Sequences in 3GPP-LTE
July 2016. Systems,” IEEE Wireless Commun. Lett., vol. 1, no. 4, pp. 280–283,
[50] S. Dama, T. V. Pasca, V. Sathya, and K. Kuchi, “A novel rach 2012.
mechanism for dense cellular-iot deployments,” in 2016 IEEE Wireless [79] A. Ali and W. Hamouda, “Cell search evaluation: A step towards the
Commun. and Networking Conf., April 2016, pp. 1–6. next generation LTE-MTC systems,” in 2016 IEEE Wireless Commun.
[51] M. Vilgelm, H. M. GÃijrsu, W. Kellerer, and M. Reisslein, “Latmapa: and Networking Conf., April 2016, pp. 1–6.
Load-adaptive throughput- maximizing preamble allocation for prior- [80] W. Xu and K. Manolakis, “Robust Synchronization for 3GPP LTE
itization in 5g random access,” IEEE Access, vol. 5, pp. 1103–1116, system,” in IEEE Global TeleCommun. Conf., 2010, pp. 1–5.
2017. [81] K. Guo, W. Xu, and G. Zhou, “Differential Carrier Frequency Offset
[52] D. Flore. (2015, Feb) Evolution of LTE in Release 13. [Online]. and Sampling Frequency Offset Estimation for 3GPP LTE,” in IEEE
Available: http://www.3gpp.org/news-events/3gpp-news/1628-rel13 Veh. Tech. conf., May 2011, pp. 1–5.
[53] J. Gozalvez, “New 3GPP Standard for IoT [Mobile Radio],” IEEE Veh. [82] F. Classen and H. Meyr, “Frequency synchronization algorithms for
Technol. Mag., vol. 11, no. 1, pp. 14–20, March 2016. OFDM systems suitable for communication over frequency selective
[54] Ericsson, “Introduction of NB-IoT,” 3GPP TSG RAN WG1 Meeting- fading channels,” in IEEE Veh. Tech. conf., Jun 1994, pp. 1655–1659
85, Tech. Rep. R1-165971, May 2016. vol.3.
[55] Huawei and HiSilicon, “Introduction of Rel-13 feature of NB-IoT in [83] A. Ali and W. Hamouda, “Employing Broadcast Channel for Frequency
36.212,” 3GPP TSG RAN WG1 Meeting-85, Tech. Rep. R1-166045, Tracking in LTE-MTC Systems,” IEEE Wireless Commun. Lett., vol. 5,
May 2016. no. 4, pp. 436–439, Aug 2016.
[56] M. Hua, M. Wang, K. W. Yang, and K. J. Zou, “Analysis of the [84] J. J. van de Beek, O. Edfors, M. Sandell, S. K. Wilson, and P. O.
frequency offset effect on zadoff-chu sequence timing performance,” Borjesson, “On channel estimation in OFDM systems,” in IEEE Veh.
IEEE Trans. Commun., vol. 62, no. 11, pp. 4024–4039, Nov 2014. Tech. conf., vol. 2, Jul 1995, pp. 815–819.
[57] J. Hu, M. Wang, K. J. Zou, K. W. Yang, M. Hua, and J. Zhang, [85] G. Auer and E. Karipidis, “Pilot aided channel estimation for OFDM:
“Enhanced lte physical downlink control channel design for machine- a separated approach for smoothing and interpolation,” in IEEE Int.
type communications,” in 2015 7th Int. Conf. on New Technologies, Conf. on Commun.,, vol. 4, May 2005, pp. 2173–2178.
Mobility and Security (NTMS), July 2015, pp. 1–5. [86] Y. Li, L. J. Cimini, and N. R. Sollenberger, “Robust channel estimation
[58] M. Wang, W. Yang, J. Zou, B. Ren, M. Hua, J. Zhang, and X. You, for OFDM systems with rapid dispersive fading channels,” IEEE Trans.
“Cellular machine-type communications: physical challenges and so- Commun., vol. 46, no. 7, pp. 902–915, Jul 1998.
lutions,” IEEE Wireless Commun., vol. 23, no. 2, pp. 126–135, April [87] E. A. Ahmed and M. M. Khairy, “Semi-adaptive channel estimation
2016. technique for LTE systems,” in IEEE Symp. on Computers and Com-
[59] J. Lorca, “Increasing coverage and maximum CFO in DFT-s-OFDM mun., June 2011, pp. 457–482.
for Machine-Type Communications,” in Int. Conf. on Commun., 2015, [88] C. Y. Hsieh, D. W. Lin, and C. Ma, “LMMSE-Based Channel Esti-
pp. 982–987. mation for LTE-Advanced MIMO Downlink Employing UE-Specific
[60] Y. E. Wang et al., “A Primer on 3GPP Narrowband Internet of Things Reference Signals,” in IEEE Veh. Tech. conf., May 2015, pp. 1–5.
(NB-IoT),” CoRR, vol. abs/1606.04171, 2016. [89] W. J. Hwang, J. H. Jang, and H. J. Choi, “An enhanced channel es-
[61] Motorola Mobility, “Introduction of NB-IoT,” 3GPP TSG RAN WG1 timation with partial interference cancelation for MU-MIMO system,”
Meeting-85, Tech. Rep. R1-165111, May 2016. IEEE Commun. Lett., vol. 16, no. 8, pp. 1232–1235, August 2012.
[62] MCC Support, “Final Report of 3GPP TSG RAN WG1-84 v2.0.0,” [90] T. Kamenosono, M. Kaneko, K. Hayashi, and M. Sakai, “Compressed
3GPP TSG RAN WG1 Meeting-84, Tech. Rep. R1-163406, April 2016. sensing-based channel estimation methods for LTE-Advanced multi-
[63] ——, “Final Report of 3GPP TSG RAN WG1-85 v1.0.0,” 3GPP TSG user downlink MIMO system,” in IEEE Veh. Tech. conf., May 2015,
RAN WG1 Meeting-85, Tech. Rep. R1-166056, Aug 2016. pp. 1–5.
[64] Huawei and HiSilicon, “Summary of NB-IoT evaluation results,” 3GPP [91] Intel Corporation, “Receiver algorithms and complexity analyses for
TSG RAN WG1 Meeting-83, Tech. Rep. R1-157741, Nov 2015. NB-IoT synchronization,” 3GPP TSG RAN WG1 Ad-Hoc Meeting,
[65] Nokia Networks, Alcatel-Lucent, and Alcatel-Lucent Shanghai Bell, Tech. Rep. R1-161897, March 2016.
“Random Access Procedure for NB-IoT,” 3GPP TSG RAN WG1 [92] E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzetta, “Massive
Meeting-84, Tech. Rep. R1-160458, Feb 2016. MIMO for Next Generation Wireless Systems,” IEEE Commun. Mag.,
[66] Sony, “NB-PRACH Coverage Levels,” 3GPP TSG RAN WG1 vol. 52, no. 2, pp. 186–195, February 2014.
Meeting-84, Tech. Rep. R1-160768, Feb 2016. [93] L. Lu, G. Y. Li, A. L. Swindlehurst, A. Ashikhmin, and R. Zhang, “An
[67] Ericsson, “Introduction of NB-IoT,” 3GPP TSG RAN WG1 Meeting- Overview of Massive MIMO: Benefits and Challenges,” IEEE J. Sel.
85, Tech. Rep. R1-165973, May 2016. Topics Signal Process., vol. 8, no. 5, pp. 742–758, Oct 2014.
[68] Intel Corporation, “NB-PRACH and random access procedure,” 3GPP [94] J. Nam, J. Y. Ahn, A. Adhikary, and G. Caire, “Joint spatial division
TSG RAN WG1 Meeting-84, Tech. Rep. R1-160415, Feb 2016. and multiplexing: Realizing massive MIMO gains with limited channel
[69] Huawei, HiSilicon, and Neul, “NB-PRACH design,” 3GPP TSG RAN state information,” in 2012 46th Annual Conf. on Information Sciences
WG1 Meeting-84, Tech. Rep. R1-160316, Feb 2016. and Systems (CISS), March 2012, pp. 1–6.
[70] Intel Corporation, “NB-PUSCH design and scheduling,” 3GPP TSG [95] A. Pitarokoilis, S. K. Mohammed, and E. G. Larsson, “On the optimal-
RAN WG1 Meeting-84, Tech. Rep. R1-160414, Feb 2016. ity of single-carrier transmission in large-scale antenna systems,” IEEE
[71] Huawei and HiSilicon, “NB-PUSCH design,” 3GPP TSG RAN WG1 Wireless Commun. Lett., vol. 1, no. 4, pp. 276–279, August 2012.
Meeting-84, Tech. Rep. R1-160325, Feb 2016. [96] C. Studer and E. G. Larsson, “PAR-Aware Large-Scale Multi-User
[72] J.-J. van de Beek, M. Sandell, and P. Borjesson, “ML estimation of time MIMO-OFDM Downlink,” IEEE J. Sel. Areas Commun., vol. 31, no. 2,
and frequency offset in OFDM systems,” IEEE Trans. Signal Process., pp. 303–313, February 2013.
vol. 45, no. 7, pp. 1800–1805, 1997. [97] S. K. Mohammed and E. G. Larsson, “Per-Antenna Constant Enve-
[73] S. Ma, X. Pan, G.-H. Yang, and T.-S. Ng, “Blind Symbol Synchro- lope Precoding for Large Multi-User MIMO Systems,” IEEE Trans.
nization Based on Cyclic Prefix for OFDM Systems,” IEEE Trans. Commun., vol. 61, no. 3, pp. 1059–1071, March 2013.
Veh. Technol., vol. 58, no. 4, pp. 1746–1751, 2009. [98] Y. Zeng, R. Zhang, and Z. N. Chen, “Electromagnetic Lens-Focusing
[74] I. Kim, Y. Han, and H. K. Chung, “An efficient synchronization signal Antenna Enabled Massive MIMO: Performance Improvement and Cost
structure for OFDM-based cellular systems,” IEEE Trans. Wireless Reduction,” IEEE J. Sel. Areas Commun., vol. 32, no. 6, pp. 1194–
Commun., vol. 9, no. 1, pp. 99–105, 2010. 1206, June 2014.
[75] Y. Tsai, G. Zhang, D. Grieco, and F. Ozluturk, “Cell search in 3GPP [99] Z. Xiang, M. Tao, and X. Wang, “Massive MIMO Multicasting in
long term evolution systems,” IEEE Veh. Technol. Mag., vol. 2, no. 2, Noncooperative Cellular Networks,” IEEE J. Sel. Areas Commun.,
pp. 23–29, 2007. vol. 32, no. 6, pp. 1180–1193, June 2014.
[76] K. Manolakis, D. Gutierrez Estevez, V. Jungnickel, W. Xu, and [100] R. Deng, S. Zhou, and Z. Niu, “Scalable Non-Orthogonal Pilot Design
C. Drewes, “A Closed Concept for Synchronization and Cell Search for Massive MIMO Systems with Massive Connectivity,” in 2016 IEEE
in 3GPP LTE Systems,” in IEEE Wireless Commun. and Networking Globecom Workshops (GC Wkshps), Dec 2016, pp. 1–6.
Conf., 2009, pp. 1–6.
[77] H. Setiawan and H. Ochi, “A low complexity physical-layer identity
detection for 3GPP Long Term Evolution,” in Int. Conf. on Advanced
Communication Technology, vol. 1, 2010, pp. 8–13.

28

1553-877X (c) 2017 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.