Accepted Manuscript

ARBAT: A Flexible Network Architecture for QoE-aware

Communications in 5G Systems

I.F. Akyildiz, A. Kak, E. Khorov, A. Krasilov, A. Kureev

PII: S1389-1286(18)31122-8
DOI: https://doi.org/10.1016/j.comnet.2018.10.016
Reference: COMPNW 6622

To appear in: Computer Networks

Please cite this article as: I.F. Akyildiz, A. Kak, E. Khorov, A. Krasilov, A. Kureev, ARBAT: A Flexible
Network Architecture for QoE-aware Communications in 5G Systems, Computer Networks (2018), doi:
https://doi.org/10.1016/j.comnet.2018.10.016

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
ARBAT: A Flexible Network Architecture for
QoE-aware Communications in 5G Systems
I.F. Akyildiz∗† , A. Kak† , E. Khorov∗ , A. Krasilov∗ , A. Kureev∗
∗ Institute for Information Transmission Problems, Russian Academy of Sciences, Moscow, Russia
Email: {akyildiz, khorov, krasilov, kureev}@iitp.ru
† Broadband Wireless Networking Lab, School of Electrical and Computer Engineering,
Georgia Institute of Technology, Atlanta, USA
Email: {ian, ahan}@ece.gatech.edu
Abstract—The expansion of the service scope of cellular
networks to include a wide variety of services such as mobile
broadband, Internet of Things, and mission-critical machine-type
communications has significantly shaped the evolution towards
5G and beyond systems. All these services impose divergent
and often mutually exclusive requirements in terms of data
rate, latency, and energy efficiency. To satisfy heterogeneous
requirements, 5G systems should have such properties as Quality-
of-Experience awareness, adaptability and flexibility, scalabil-
ity and reliability, support for multiple RATs and backward
US
AN
compatibility, all at a low CAPEX and OPEX. To this end,
software-defined networking and network function virtualization
have been envisioned as key enabling technologies for 5G, and
represent a major paradigm shift for 5G systems. In recent years,
a plethora of software-defined mobile network architectures have
been introduced worldwide, each with their unique features
M
and drawbacks. Within this context, this paper introduces a
new architecture called ARBAT which has been designed to
satisfy and exceed the requirements put forth by 5G. ARBAT
is characterized by many innovative features such as the Uni-
ED
versal Network Device and Unified Cellular Network concepts,
multi-slice modular resource management with the AirHYPE
wireless hypervisor, network-user application interaction through
the xStream platform, and simplified multi-tenant orchestration
through ServiceBRIDGE. The novel features of the ARBAT
PT
infrastructure plane, data plane, control plane and Management
and Orchestration entity are also explained in the paper in detail.
Furthermore, a qualitative evaluation and feature comparison of
ARBAT with other state-of-the-art architectures is conducted to
demonstrate that ARBAT satisfies the aforementioned objectives
CE
of the 5G systems.
Index Terms—Wireless Networking, Cellular Networking, Cel-
lular Communications, Software-defined Networking, Network
Function Virtualization, Architecture, 5G.
AC
I. I NTRODUCTION
W ITH the functional freeze for 3GPP Release 15 taking
place in late 2018 [1], wireless networks have wit-
nessed a sea-change over the past two years, with 5G being
envisioned as a service delivery platform enabling a wide
variety of use-cases, including but not limited to fixed wireless,
enhanced Mobile Broadband (eMBB), massive Machine-type
Communications (mMTC), and Ultra-Reliable Low-latency
Communications (URLLC). This evolution of the cellular
The research has been carried out at IITP RAS and supported by the Russian
Government (Contract No 14.W03.31.0019).
1
T
IP
networking landscape has been driven, in no small part, by
CR
the ever-increasing demand for higher data rates resulting
from the rise in popularity of bandwidth-intensive applications
such as video conferencing, file sharing, social networking,
and streaming services. Mobile data traffic has grown 18-
fold over the past five years and is expected to exceed 49
exabytes per month by 2021 [2]. The highly differentiated
nature of services that are expected to be deployed over or
make use of cellular infrastructure in the near future all pose
different requirements in terms of very high data rates and
minimum latency that are difficult to achieve using existing
cellular networks. More specifically, the monolithic nature and
“one-size-fits-all” approach of 4G make it ill-suited to offer
service-tailored connectivity with a wide variety Quality of
Experience (QoE) guarantees. Accordingly, the 5G cellular
system requirements [3] mandate the following:
peak data rates of up to 20 Gbps,
•
• user experienced data rates of up to 100 Mbps,
• three-fold increase in spectral efficiency compared to
LTE,
• support for up to 10 Mbps/m2 area traffic capacity,
• user-plane latency of less than 1 ms in the Radio Access
Network (RAN), and
• 100-fold increase in network energy efficiency.
To meet these divergent or even mutually exclusive technical
requirements at a reasonable price, a 5G system shall have the
following properties.
• QoE-awareness: The 5G system is not just a pipe for
data transmission. In contrast, it is used to deliver cer-
tain services, each of which imposes Quality of Service
(QoS) requirements. The 5G system shall understand the
requirements for each particular application and serve its
traffic accordingly.
• Adaptability and Flexibility: The 5G system shall suit
a variety of use-cases right from eMBB to URLLC,
and even perform use-case specific reconfiguration if
required. In contrast to 4G systems with rigid Radio
Access Technology (RAT) and inflexible infrastructure,
the 5G system shall be able to adaptively reuse hardware
and radio resources in order to satisfy ever-changing user
demands.
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
• Scalability: The 5G system shall be able to support a
large number of users with different needs. The radio,
computational, fiber resources shall be used in a very
efficient way. For that, the architecture shall support
simultaneous use of various techniques that optimize
spectral efficiency in different scenarios, e.g., simulta-
neous delivery of both massive data flows and light
traffic. Moreover, the topology and the routes in the Core
Network (CN) shall be changed dynamically to balance
the load and avoid congestion.
• Self-healing and Reliability: In case of malfunction
of any devices, including critical ones such as packet
gateways, or controllers, the system shall be able to
reassign their functionality to other devices without any
service degradation.
• Multiple RATs and Backward Compatibility: Due
to the significant investment involved in infrastructure
upgrades, the 5G system will not immediately replace
other types of wireless networks. Additionally, there exist
a wide variety of technologies such as Wi-Fi, Sigfox,
and LoRa that cannot be beaten by 3GPP RATs at
least in the near future. To provide compatibility with
the devices supporting these technologies and to achieve
higher efficiency of radio resource usage, the 5G system
should integrate these RATs.
US
AN
• High Energy Efficiency: In the cellular network, the
RAN accounts for over half of the total power consump-
tion [4]. However, the move towards a disaggregated
RAN provides greater flexibility in resource allocation,
M
and network planning and deployment. For example,
architectural support for techniques such as joint optimal
resource allocation that maximizes system throughput
while minimizing energy consumption, configurable net-
ED
work density, and energy harvesting plays a vital role in
lowering the carbon footprint of the network.
• Low CAPEX and OPEX: Generational upgrades have
a major impact on a communications service provider’s
PT
CAPEX [5]. Therefore, the appeal of any new architecture
is also closely tied to its economic feasibility. In partic-
ular, those architectures that can demonstrate significant
cost savings, when compared to legacy solutions, as well
CE
as the competition, over a period of 5 − 7 years will
see larger deployments. In order to minimize CAPEX,
in contrast to previous generations of cellular systems
with multiple overlapping networks of different operators,
AC
there would be a single 5G infrastructure which will share
resources between Service Providers (SePs). Moreover,
the architecture of the deployed 5G system shall allow
the use of low-cost interchangeable and programmable
devices, while the deployment new services shall require
no hardware upgrades.
To this end, we have previously identified 10 key enabling
technologies that are essential for meeting the requirements put
forth by 5G [6], chief among which are Software-Defined Net-
working (SDN), and Network Function Virtualization (NFV).
In this paper, we introduce the most promising ideas for
5G and beyond systems and propose a new flexible wireless
2
architecture called ARBAT, named after the famous and beau-
tiful street in Moscow. More specifically, we introduce several
innovative features such as the Universal Network Device
(UND), the Unified Cellular Network (UCN) concept, multi-
slice modular resource management with the AirHYPE wire-
less hypervisor, network–user application interaction through
the xStream platform, and simplified multi-tenant orchestration
through ServiceBRIDGE.
The rest of the paper is organized as follows. We continue
our discussion by discussing the state of the art in the domain
of cellular networks based on SDN and NFV in Section II.
Then, in Section III, we present the ARBAT architecture
T
and detail the major components of the system. Further, we
describe the novel features of the ARBAT infrastructure plane,
IP
data plane, control plane and Management and Orchestration
(MANO) entity in Sections IV-VII, respectively. We perform
a qualitative evaluation of the ARBAT and its comparison
CR
to other existing architectures in Section VIII. Finally, we
conclude the paper in Section IX.
II. R ELATED W ORK
In recent years, the domain of Software-Defined Mobile
Networks (SDMNs) has witnessed much traction from both
academia and industry. Early efforts in the SDMN domain
were based on the softwarization of the CN [7], however the
focus has shifted to the RAN [8], [9] in the past years. Today,
there exists an exhaustive body of work on 5G and beyond
wireless network architectures. In this section, we discuss
the key works highlighting their unique features and primary
drawbacks.
While standards pertaining to SDN-based cellular networks
are still in their infancy, the 3GPP Control and User Plane
Separation (CUPS) [10] paradigm introduced in 3GPP Release
14, is partially based on SDN principles. At the same time
it is important to note that CUPS is patently different from
SDN by design, with the 3GPP control (3GPP-C) and user
(3GPP-U) planes being different from the SDN control (SDN-
C) and data (SDN-D) planes. First, CUPS splits the data path
into two paths, namely, the control traffic data path and user
traffic data path. The 3GPP-C includes all functions that deal
with control traffic data path (from the user to the access and
mobility management function), while the 3GPP-D consists of
the functions that process user traffic towards the data network.
In other words, both the control and data plane functions
take part in packet forwarding. This approach is in contrast
to SDN which mandates that all network traffic must pass
through the SDN-D, with the SDN-C having no role in the
physical packets forwarding action. Second, as of the current
release [10] CUPS is limited to the CN only, with a partial
separation of the RAN under study [11], i.e., a large portion
of the RAN still follows the classical coupled control and
user plane paradigm. However, a proper realization of the
SDN concept requires complete separation of the entities that
take the network control decisions from those that apply such
decisions.
A monolithic 5G architecture based on SDN and NFV
concepts is introduced in [12] . The principal advantage of
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
this architecture is the cluster-based RAN concept with local
controllers responsible for each cluster, allowing for efficient
content caching, and inter-cluster scalability. However, we note
that in the presented architecture, each local controller itself
is a physical network function (PNF) with fixed functionality.
Further, while the overall network is scalable at the cluster
level because of cluster independency, scalability within a
cluster is a challenge, due to the potential bottleneck arising
at a given local controller as the cluster size grows.
In a similar manner, SoftNet [13] considers the base station
as a PNF, and proposes a unified RAN and an SDN-based CN.
Additional features include a unified Radio Resource Manage-
ment (RRM) framework and QoS mapping. While the control
plane is distributed, it is primarily deployed within the CN, and
the CN edge with a static distribution, i.e., functions cannot be
moved across network elements, and therefore the flexibility
suffers. Moreover, the unified resource scheduler cannot satisfy
delay requirements for low-latency traffic. FlexRAN [9] is one
of the few RAN solutions to present a functional proof of
concept that includes the RAN controller, and Northbound
(NB) and Southbound (SB) APIs. However, FlexRAN also
considers a monolithic base station which cannot adapt to
different use-cases. Further, the implementation is limited
to LTE-based access and includes only one use-case based
evaluation.
US
AN
A classical SDN architecture using the OpenDaylight
(ODL) platform is introduced in [14]. The prototype is
based on the ODL controller and emulated base stations.
The developed framework is validated with enhanced Inter-
M
Cell Interference Coordination (eICIC) deployed as a control
function. The drawback of this architecture is the lack of
flexibility. The RAN has only one functional split, and the
network functions cannot be moved from one network node
ED
to another.
The Telecom Infra Project’s (TIP) vRAN Fronthaul
group [15] provides a virtualized base station solution that
includes a virtualized Baseband Unit (vBBU), and a Remote
PT
Radio Unit (RRU). The highlight of the vRAN solution is
its ability to function across a variety of fronthaul options
using multiple physical layer functional splits, and fronthaul
bandwidth compression. However, since vRAN caters to phys-
CE
ical layer functional splits only, many network functions are
deployed in a centralized manner at the vBBU. While this
approach is useful for features such as centralized scheduling,
and coordinated multipoint (CoMP), it also renders the ar-
AC
chitecture inflexible in the face of low latency use-cases that
mandate a distributed function deployment close to the end-
users. Further, vRAN provides an NFV solution only and does
not incorporate SDN, i.e., control and data plane separation is
absent.
Similar to TIP, NEC also provides a virtualized C-RAN
Distributed Unit (DU) solution [16], that supports two distinct
functional splits between the DU and Radio Unit (RU) – the
L2 and L3 splits. In the L2 split, physical layer functions
are deployed at the RU, whereas in the L3 split, the RU
additionally performs the MAC and RLC functions. On the
one hand, the L2 split offers greater centralization, while on
the other hand, the L3 split is better suited for low latency use-
3
cases. In that sense, NECs solution is more versatile but less
granular than TIP’s. However, despite the perceived versatility,
NEC’s architecture does not make use of SDN, and is an NFV-
only solution at the moment.
Very recently, the xRAN forum [17] merged with the C-
RAN alliance to form the ORAN alliance. More specifically,
the alliance provides a RAN architecture [18] and a fronthaul
specification [19]. Supporting both SDN and NFV, xRAN
introduces a standalone xRAN controller that interacts with
the Central Unit (CU) and Distributed Unit (DU) to optimize
overall network performance. xRAN also includes a relational
database that reflects the network’s state and is used by the
T
network controller to carry out its operations. However, the
current specification provides support for a single functional
IP
split only, that concentrates much of the RAN functions at the
CU. This can potentially decrease the performance in scenarios
with several types of traffic, especially for low-latency traffic
CR
which requires the function deployment as close to the user
as possible.
Based on the Central Office Re-architected as a Datacenter
(CORD) platform [20], Mobile-CORD (M-CORD) [21] seeks
to implement an SDN and NFV-based cellular network span-
ning both the RAN and CN. M-CORD is a composition of
a virtualization platform, a virtual infrastructure management
platform, an orchestration solution, and a network operating
system (NOS). Network control is exercised through a dis-
tributed implementation of xRAN running as an application
on the NOS, while network slicing is implemented via the
ProgRAN application. In essence, M-CORD can be regarded
as an extension of the xRAN, and, while it improves upon
certain aspects by extending the control and user plane sepa-
ration and bringing control functionality down to the RU level,
it also suffers from similar drawbacks such as limited fronthaul
adaptability, and limited use-case flexibility.
The NGMN Alliance has also put forth their version of the
5G architecture [22], wherein the use of a variety of RATs and
functional splits is suggested for each type of traffic. While
centralized control is implied, the architecture does not delve
on the actual placement of control functions. The METIS-
II project [8] follows up the ideas of the NGMN alliance
by providing a highly-detailed RAN framework based on the
CU-DU concept explained above, with support for several
functional splits and multiple RATs. However, we note the
absence of interface definitions between the control and data
planes.
Similar to METIS-II, the 5G NORMA project [23] provides
a detailed architecture with such features as multi-layer con-
trol, CU-DU functional splits, and support for multi-RAT. In
the proposed architecture, the DU is RAT-specific, and there-
fore it cannot be re-used for different RATs. On the one hand,
it decreases the cost and provides better migration since a
legacy transmitter device is used for each RAT (e.g. an eNB for
LTE, an access point for Wi-Fi). On the other hand, decreases
flexibility. The architecture also includes frameworks for RRM
and mobility management.
Further, the X-HAUL project [24] supplements the ideas
in 5G NORMA with backhaul and fronthaul solutions. X-
HAUL proposes a hierarchical three-tier control plane– the L0
 ACCEPTED MANUSCRIPT

FOR SUBMISSION TO JOURNAL, OCTOBER 2018 4

User Infrastructure Provider SMF

Service
AMF
Equipment PCF Provider
Link

Management and Orchestration (MANO)

Adaptation SRRM Policies and
Control Plane Scheduling xNode
VNFs
Mobility AirHYPE Mobility Transport
Management Management Billing
Session RRC Functions Other Policies
Control Plane
Management Southbound APIs (including P4) Load Balancing
Bearer RLC
Management MAC Native Services
Paging Upper PHY
Measurement VoIP
UPF
Lower PHY TV
Configuration
Security
PDCP Video
RF SMS
Configuration Forwarding
Etc.
Data Plane
Hypervisors

T
Data Plane
3rd Party
PDCP Services
RLC

IP
MAC
PHY
RF External Data
Network
Infrastructure Plane
Unified Cellular Network

CR
Fig. 1: The ARBAT architecture.

Controller which is responsible for a given area, the L1 Con-
troller which exercises control over a set of L0 controllers and
logically performs path setup across areas, and Top Controller
which includes multiple L1 controllers in its domain.
To summarize, we note the following shortcomings in the
US
AN
prior work: (i) physical centralization of control functional-
ities, (ii) virtualization without consideration for SDN, (iii)
static function distribution, (v) absence of radio resource
virtualization solutions, and (iv) a lack of support for multiple
M
Data Plane: Drawing on both CUPS and SDN, the
•
data plane within ARBAT exclusively refers to the user
traffic data path, i.e., the data plane does not contain any
control functions, or carry control traffic. The data plane
is characterized by the presence of network agents that
serve as endpoints for control functions. Moving beyond
OpenFlow, ARBAT makes the use of P4 [27] to allow
for custom data processing pipelines. While the network
functions in the data plane belong to the SeP (Service

RATs. While physical centralization decreases development
and deployment costs, it also leads to poor scalability and
flexibility. To this end, ARBAT has been designed with a view
to overcome the aforementioned drawbacks.
ED
Provider), their operation is governed through the control
plane, by the InP with policy inputs from the SeP.
• Control Plane: It implements network control logic,
through VNFs (Virtual Network Functions) called control

functions. Common examples of control functions include

III. ARBAT A RCHITECTURE OVERVIEW scheduling, mobility management, and link adaptation.
A. System Domains The control plane executes in-network control for net-
work service operations and is managed by the InP.
PT

ARBAT has been designed to meet and exceed the 5G
system requirements identified in Section I. Based on the
concepts of SDN [25] and NFV [26], ARBAT consists of
separate data and control planes, an infrastructure plane and
CE
• MANO Framework: It serves as the bridge between
SePs and InPs, and is responsible for the overall or-
chestration and lifecycle management of both network
services and network functions.

a MANO framework for Virtual Network Function (VNF)

deployment and management, as shown in Fig. 1. Below we
briefly describe the primary components of ARBAT. B. Stakeholders
• Infrastructure Plane: It represents the physical and With a view towards lowering infrastructure costs, we envi-
AC

virtual hardware in the network. It consists of Wireless
Transmission Points (WTPs) that use radio resources to
transmit signals, network nodes that provide computing
resources, interconnecting links that provide connectivity
between nodes, and hypervisors that virtualize these
resources. The infrastructure plane lies under the control
of the Infrastructure Provider (InP), and infrastructure
under a single InP forms an infrastructure domain. With
ARBAT, we introduce the concepts of UND (Universal
Network Device) and UCN (Unified Cellular Network) in
the infrastructure plane. Since ARBAT supports multiple
InPs, there are multiple infrastructure domains in the
network.
sion a single 5G infrastructure that will be utilized by several
SePs. Recognizing that multi-tenancy is a core feature of
next-generation wireless networks, the network should provide
for a clear delineation of the purview of the different stake-
holders associated with network operations – Infrastructure
Providers (InPs) and SePs. The InP deploys and supports the
infrastructure, provides pipes for SePs, and seeks to maximize
resource utilization and operational profit, while the SeP seeks
to maximize profit while operating under QoS requirements.
More specifically, within ARBAT, where the InP and SeP
are the major stakeholders, it is also necessary to outline the
operational ownership of each plane. Following the concept of
the Light Mobile Virtual Network Operator (MVNO), ARBAT
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
incorporates infrastructure, control and data planes that are
entirely under the purview of the InP, and a MANO which is
shared with the SeP. Such a separation is only one of many
possibilities. Another approach could involve the InP having
control over the infrastructure plane only, with the data and
control planes belonging to the SeP.
Generally, SePs “provide” network services to end-users,
while InPs “deliver” the contracted services to SePs. Examples
of end-users include individuals or enterprises that require
data or telephony services, while a cellular provider or an
ISP are examples of SePs. An end-user is the SeP’s customer,
while the SeP is a customer of the InP itself. A SeP may
contract with multiple InPs based on its requirements. More
specifically, the SeP is responsible for providing the VNF
models, VNF-Forwarding Graphs (VNF-FGs), specific poli-
cies such as those relating to restrictions on the placement
of VNFs, preferred routes, etc., and Service Level Agreement
(SLA) parameters such as throughput, latency, and reliability.
The actual instantiation and lifecycle management of network
functions and NFV Infrastructure (NFVI), network operations
and control are all under the purview of the InP. Thus, the
physical entities in the network are owned by the InP, while
the logical entities are the SeP’s property. Further, we note
that within the context of ARBAT, the network slicing [28]
is an infrastructural construct, and not visible to the SeP. The
US
AN
InP may choose to make use of a slice for service delivery
in case several services have common network functions. For
example, if an InP has an active eMBB slice deployed over
its infrastructure, and has to serve eMBB flows from several
M
SePs, then these can be accommodated within a single eMBB
slice. By separating the SeP from the day-to-day network
operations, we can ensure that the InP has complete control
over their infrastructure, and the need for resource isolation
ED
can be avoided, allowing for optimal resource utilization.
In the following, we describe in detail each of the three
planes – the infrastructure plane in Section IV, the data plane
in Section V, the control plane in Section VI, along with
PT
the Management and Orchestration (MANO) framework in
Section VII.
IV. I NFRASTRUCTURE P LANE
CE
The infrastructure plane falls under the purview of the InP,
with different InPs making their infrastructure assets available
to the SePs for deployment of services. Physically, this plane
consists of servers, switches, Remote Radio Heads (RRHs),
AC
and interconnecting links. While, logically the infrastructure
can be modeled as a set of distributed resource groups,
where each resource group is characterized by its physical
location. The primary resources under consideration including
computational, storage, memory, network, and radio resources.
Such multi-dimensional resource abstraction provides great
flexibility to the SePs in the design of network services. To
this end, we model the ARBAT infrastructure plane to consist
of the following entities:
• Universal Network Devices (UNDs): This is the funda-
mental building block of ABRAT, and represents all the
physical devices in the network, i.e., radio units, servers,
switches, and legacy infrastructure.
5
• Transport Links (TLs): They represent the physical
links that connect different UNDs to each other.
In the following subsections, we describe the two entities
in detail.
A. The Universal Network Device (UND)
The primary building block of the ARBAT is the network
node called UND. The physical realization of the network
consists of an interconnection between such UNDs forming
a flat network that connects to other external data networks
through private peering arrangements, or Internet exchange
T
points. A UND can be a Physical Network Function (PNF)
tied to specialized hardware, a Virtualizable Whitebox (VWB)
IP
on which Virtual Network Functions (VNFs) are deployed, or
a combination of the two. The UND is described by a set of
hardware resources and built-in PNFs. Further, we note that
CR
the introduction of the UND does not imply that the network
consists of the same or similar devices. Rather, the UND
can be viewed as a logical construct that helps in describing
and parameterizing the huge variety of networking devices
to achieve optimality in resource utilization, and flexibility in
network design. Moreover, the UND provides the capability to
extend Multi-access Edge Computing (MEC) as close to the
user as possible.
The PNF representation of a UND is motivated by two
major factors. First, PNFs are used to represent legacy infras-
tructure that cannot be virtualized, e.g., LTE eNodeBs, legacy
access points, non-programmable switches, etc. Thus, ARBAT
can use existing LTE infrastructure, simplifying migration
from an old RATs to new ones. Second, we note that the
implementation of functionality associated with WTPs (Wire-
less Transmission Points) such as analog-to-digital/digital-
to-analog conversion, up/down-conversions, and the duplexer
function can be done far more efficiently, and at a lower cost,
in hardware as opposed to software. Therefore, WTPs are
also represented by the PNFs. This is especially important
for many existing RATs, such as Wi-Fi, which access points
can be seamlessly integrated within the ARBAT architecture.
In more general terms, WTPs within ARBAT are functionally
equivalent to RRHs (Remote Radio Heads) commonly found
in existing architectures. Thus, the PNF representation allows
ARBAT to function with and virtualize a variety of front-ends
that are readily commercially available.
The UNDs may also have additional computational re-
sources, such as processing, memory, storage, etc., which
are amenable to virtualization and support the deployment of
VNFs. For example, a UND with the WTP PNF and additional
computing resources may also execute all other data plane
functions for a URLLC service in order to reduce latency.
We can further differentiate between virtualizable hardware
based on the presence of specialized features such as FPGAs,
or support for the P4 language. More specifically, FPGAs
are better suited for functions related to baseband processing
than commodity CPUs, while P4 is a high-level language for
programming protocol-independent packet processors. Unlike
the OpenFlow specification that explicitly specifies protocol
headers on which it operates, P4 suggests that switches should
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
support flexible mechanisms for parsing packets and matching
header fields, allowing control functions to leverage these
capabilities through a common, open interface. In particular,
ARBAT incorporates support for P4 to provide a low-latency
data path for user-traffic. The motivation for that comes from
the fact that the functions performed by certain VNFs, such
as those related to packet forwarding, are not inherently
suited to general purpose computing platforms. In other words,
deploying such VNFs on a general purpose processor (GPP)
will lead to a phenomenon similar to slow path processing [29]
causing the network to experience widespread congestion.
However, this bottleneck could be alleviated if these func-
tions were to be performed by dedicated switching hardware
operating at line speed. Introduced in 2008, OpenFlow was a
major step in providing access to line-rate forwarding, however
OpenFlow presents a fixed-function data plane pipeline, which
is difficult to extend and modify. By enabling custom pipelines
that can be loaded and controlled on-demand, P4 presents a
significant opportunity to realize such data forwarding related
functions of the cellular network over high-speed switching
hardware. For example, a majority of the functions performed
by the User Plane Function (UPF) introduced by 3GPP [30]
are amenable to a P4-based implementation, wherein the UPF
control logic is implemented as a VNF on a GPP, while the
packet processing is done on the forwarding device.
US
AN
B. Transport Links (TL)
The TLs in ARBAT represent the physical interconnections
M
between UNDs, and consist of both wired and wireless links.
More specifically, we consider the following types of TLs.
• Wired: Fiber, coax, and copper.
ED
• Wireless: sub-6 GHz, microwave, and mmWave.
For a more detailed overview of the different solutions that
can be deployed on these links, we refer the reader to [31].
Further, in the context of ARBAT, fronthaul exclusively refers
PT
to the TL that terminates at a UND represented by a WTP
PNF, and makes use of eCPRI split option E [31] which is
suited for time-domain IQ sample transport. While the use of
eCPRI imposes high throughput and low latency requirements
CE
on the fronthaul, we envision such PNFs being either co-
located with, or deployed sufficiently close to VWB UNDs
that implement the PHY layer functions, in addition to the
use of IQ compression [32] where required. For other physical
AC
layer splits that may exist between UNDs, eCPRI options A
through D are used, depending on the use case. For example,
centralization of a baseband processing functions at a single
UND requires a lower physical layer split with high throughput
and low latency demands, on the other hand, performing
the baseband processing in a distributed manner on WVBs
close to WTPs, relaxes the link requirements. The former
approach allows for greater centralization gain by allowing
for centralized scheduling and resource management at the
cost of robust TL requirements, while the latter allows for the
use of non-ideal TLs at the cost of centralization gain. The
trade-off ultimately depends on the nature of service that is
being deployed, and its QoS requirements.
6
C. Resource Virtualization and Abstraction
Thus far we have described the primary building blocks
of the infrastructure plane, namely, UNDs and TLs. From a
network deployment perspective, the infrastructure plane can
be modeled as an undirected graph, where the UNDs represent
the nodes, and the TLs represent the edges. Consequently, the
need for resource virtualization necessitates the use of hyper-
visors. The virtualization of computing resources– processing,
memory, and storage has been investigated a great deal with
several virtualization solutions such as KVM [33], LXC [34],
Xen [35], and Hyper-V [36] being readily available. Similarly,
there exist network hypervisors that provide more than one
T
networking context per physical networking device to allow
for the provisioning of differentiated services– FlowVisor [37]
IP
and its extensions [38] that support OpenFlow; and the re-
cently proposed HyPer4 [39] and HyperV [40] hypervisors
CR
for virtualizing P4-based UNDs. However, the virtualization
of radio resource remains an ongoing challenge [41], [42].
To this end, ARBAT introduces a new wireless hypervisor–
AirHYPE. Described in detail in Sec. VI, AirHYPE is a major
step towards ensuring that: (i) each service or SeP is presented
with a set of virtualized radio resources, and (ii) the InP is
able to achieve optimal resource utilization through the use of
statistical multiplexing while maximizing profits.
UNDs within ARBAT are characterized by an n-tuple
resource abstraction, which represents for each UND: (i)
a set of available hardware units and their characteristics,
and (ii) programming logic executed on this hardware. As
for hardware, the UND may contain the following units:
(i) RF-front ends which are characterized by the operating
frequency bands and maximum supported bandwidth, general
purposes processors or FPGAs with given processing capa-
bilities, (iii) memory and storage, (iv) Application-Specific
Integrated Circuits (ASICs) which execute specific functions,
etc. Based on the available hardware, UND can execute some
specific PNFs which cannot be changed and/or software which
can be changed by the InP. In the latter case, the InP can
virtualize UND resources and use them to run some VNFs.
Depending on VNF requirements (e.g., in terms of memory,
and processing load), the InP can select appropriate UND. For
example, UNDs with FPGA are more suitable for executing
baseband signal processing functions, while control functions
such as scheduling can be executed on UND with a general
purpose processor. More details on how InP maps VNFs to
UNDs are given in Section VII. In a manner similar to UNDs,
the properties of a transport link can be defined by its capacity,
delay, and reliability.
D. Unified Cellular Network
In ARBAT, the role of each UND is determined by the set of
active VNFs deployed on it. Therefore, the UND represents
a different context for each network slice, and there is no
predefined CU or DU. For example, a UND can run only PHY
VNFs for an eMBB slice, and almost all VNFs for URLLC
slice. Consequently, there is no broad distinction between the
CN and RAN, and through ARBAT, we introduce the concept
of the Unified Cellular Network (UCN)
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
UND n Control Plane
Access-agnostic Control
Function UND n
Generic Control Data Plane
Messages Capabilities
UND n Data Plane
Network Agent
UND n
Data Plane Function Network Agent
Fig. 2: Control and data plane interaction via network agents.
More specifically, the support for UCN comes from the
Protocol Data Unit (PDU) session concept introduced by
3GPP [43] and the generic UND model proposed herein.
First, we recognize that a PDU session is a logical connection
established between a UE and a data network, and that the
data exchanged between the two entities is processed by a set
of network functions that would traditionally be classified as
belonging to either the CN or RAN. Second, we note that,
by definition, it is possible to deploy any network function on
the UND provided that the given UND can meet the func-
tion’s resource requirements. In this manner, the PDU session
US
AN
connectivity can be realized as a network function chain [44],
which originates at the UND with WTP functionality, traverses
a set of UNDs and terminates at the UND that peers with the
external data network. The peering UND also serves as the
PDU session anchor, e.g., this could be the UND which hosts
M
the UPF function [45].
The use of UCN within ARBAT provides certain key
advantages:
ED
• Flat Architecture: By eschewing the classical concepts
of CN and RAN, UCN allows for the implementation of
a flat architecture devoid of any hierarchy. A distributed
flat architecture serves to achieve much low user traffic
PT
latency than hierarchical solutions [46]. The reason for
this is that in a hierarchical architecture, user traffic must
be routed through the so-called CN nodes. By introducing
an anchor, hierarchical cellular networks reduce the flexi-
CE
bility with which user traffic can be routed to the external
data network, to the detriment of network latency. On the
other hand, with UCN, ARBAT can set up service chains
that are optimized for traffic delivery to data networks,
AC
without any topological restrictions except those relating
to node and link capacity.
• Network Deployment Flexibility: UCN simplifies the
mapping of network functions on to the infrastructure
plane by not classifying UNDs on the basis of RAN and
CN. Further, UCN also brings flexibility to deployments,
as the same UND may be used for both RAN and
CN functions for different services, thus allowing for
increased infrastructure sharing.
• Reduced CAPEX and OPEX: UCN is expected to
lower costs for both the InP and the SeP. First, from
the InP’s perspective infrastructural constraints associated
with only a subset of the hardware belonging to the core
7
Control Plane
UND n+1 Access-specific
Access-agnostic
UND n+2
Access-agnostic
Data Plane
UND n+1 Network Agent
Network Agent
UND n+2
T
IP
network are now taken care of. Second, for the SeP,
UCN allows universal placement of network functions on
UNDs, thus avoiding the potentially high costs associated
CR
with deploying functions on dedicated virtualized core
network hardware.
V. DATA P LANE
It is responsible for forwarding user data between end-points
(e.g., between a UE and a remote server or between two UEs).
In ARBAT, connectivity between the end-points is realized as a
chain of network functions (both VNFs and PNFs) deployed at
UNDs. As mentioned in Section IV, depending on the UND
capabilities, different network functions can be deployed on
it. Let us describe how the data flow is forwarded from a UE
to a remote server. The first UND on the path shall have an
RF front-end in order to transmit/receive signals to/from UE
over the wireless channel, i.e., this UND serves as the WTP.
Further, this UND and/or neighboring UND(s) shall execute
RAT-specific network functions (e.g., for LTE RAT, these
UNDs can execute baseband signal processing, MAC, RLC,
and PDCP functions). After that, data packets are forwarded to
other UNDs which perform forwarding functions all the way
to the remote server.
As detailed in Section VII, for each service, the MANO
framework configures and deploys a specific network function
chain by taking into account available resources at UNDs
and transport links, and service QoS requirements. In other
words, ARBAT uses dynamic functional split. For example,
for an eMBB service, we can split RAN functions between
several neighboring UNDs in order to implement CoMP
and centralized scheduling solutions which allow increasing
spectral efficiency. In contrast, for a URLLC service, all RAN
functions must be deployed on a single UND which has an
RF front-end because of tight latency requirements.
In multi-RAT scenarios, often it makes operational sense
to have some access-agnostic control functions that issue
generic commands over the SBI (Southbound Interface), in
addition to access-specific control functions. On the other
hand, the data plane functions are access-specific. Since the
access-agnostic control functions must function with a variety
of data plane functions, we introduce network agents. The
access-agnostic control plane functions manage the data plane
functions through network agents via the SBI, i.e., the network
agents serve as the termination point for the SBI as shown in
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
Control Function
RRC Functions
Mobility
Protocol Stack Management
RRC Functional
RLC Decomposition Paging
MAC
PHY
Site
Configuration
System
Measurements
Fig. 3: Functional decomposition of cellular protocol stack.
Fig. 2. Thus, the agents are vital for native multi-RAT support,
and the use of agents leads to a simplified SBI. They are also
responsible for exposing data plane functionality to the control
plane, for example, the PHY can expose PHICH, RACH,
PRACH, PUCCH, and PUSCH parameters via a network
agent to be configured by the controller, while the MAC data
plane function could expose parameters such as MCS, PMI
configuration, RB assignment bitmap, etc. Then, RAT-specific
agents in the data plane convert generic messages to RAT-
specific messages for the corresponding data plane functions.
In this manner, the SBI is kept independent from the RAT in
use, and any change in the supported RATs does not trigger a
change in the SBI.
Taking into consideration a data forwarding function, the
agent in question is the P4 agent which exposes the forwarding
table to the corresponding control function. During the table
update operation, the agent receives updates from the con-
M
troller and modifies the match-and-action tables accordingly.
On the other hand, if the underlying P4-capable hardware
is to interface with an OpenFlow controller, the agent will
perform the additional function of providing an OpenFlow-to-
ED
P4 mapping. Therefore, we see that the use of an agent allows
the independent evolution of the data and control planes. In
the aforementioned example, the same data plane function can
interface with two different control plane functions.
PT
VI. C ONTROL P LANE
It is responsible for network control and performance opti-
CE
mization. Following the SDN paradigm, the network functions
can be categorized into control and data plane functions, where
the decisions are made by the control plane functions, while
the implementation of these decisions is carried out by the
AC
data plane functions. As shown in Fig. 3, the non-exhaustive
list of control plane functions includes:
• Mobility Management: Associate UE to a particular
cell(s).
• Paging: Notify inactive UEs about incoming flows.
• Scheduling: Allocate resource blocks to different
UEs/flows.
• Link Adaptation: Set the transmission scheme (modu-
lation and coding scheme, number of MIMO layers, etc.)
• Power Control: Allocate power for each resource block.
• System Measurements and Reporting: Gather statistics
(e.g., Channel State Information (CSI), buffer status) used
by other functions.
8
• xNode: provide interface between user-applications and
Link Adaption the network as proposed in [47].
Power Control Based on the decisions made by the control plane functions,
the data plane functions at different layers of the protocol stack
Interference
Management process data packets. In other words, they add, remove, or
modify headers, concatenate or segment data packets to create
xNode
a transport block of a given size, encode or decode transport
Scheduling block(s), generate signals with the given power and modulation
scheme, and finally, transmit or receive them over the wireless
channel. As mentioned in Section V, control functions can
either be access-agnostic or access-specific. Access-agnostic
control functions are not tied to a particular RAT and provide
T
generic control functionality. For example, if the InP deploys
IP
multi-RAT joint scheduling over their network, the scheduling
function would be access-agnostic. xNode is another promi-
nent example of a potential access-agnostic control function.
CR
On the other hand, access-specific control functions cater to
specific RATs. Every control function can be designed to be
access-specific, but the design of access-agnostic functions
requires careful consideration of the system features that can
US be handled in a generic manner.
The use of NFV allows for the deployment of control and
data plane functions at different UNDs. The placement of con-
trol functions shall take into account latency constraints. For
AN
example, the decisions made by scheduling and link adaptation
functions must be provided to the corresponding data plane
functions (e.g., MAC/PHY responsible for creating transport
blocks) almost instantly. Such tight latency restrictions imply
that these control functions shall be deployed close to the
corresponding data plane functions.
The functional decomposition of the protocol stack for
different RATs [1], [8] is a generally well-studied topic, and
therefore we do not delve into it in this paper. Instead, we will
focus on the novel features of the control plane introduced in
ARBAT. First, we introduce a novel control plane entity called
xNode that enables communication between applications and
the network. Thanks to this feature, we can easily classify
traffic and obtain specific QoS requirements for each data
flow. Second, we introduce a multi-slice RRM framework that
utilizes information collected by xNode and distribute radio
resources between various flows to satisfy their specific QoS
requirements and optimize QoE for the end users.
A. xNode
To enable QoE-aware resource management, ARBAT incor-
porates the xStream platform introduced in [47]. This platform
provides duplex communication between applications running
at endpoint devices and the network. It is a flexible platform
and can be used to improve performance for various types
of traffic. In particular, in [47], we propose a set of solutions
significantly increasing QoE metrics for web and video traffic.
In [48], we design new radio resource and traffic management
algorithms for URLLC which increase network capacity (i.e.,
the maximum network load at which reliability and latency
requirements are satisfied for more than 99% of UEs) and
makes the network resilient to overloads.
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
ARBAT
xNode
Client
xNode
VNF
VNF
Content
UE Server
Fig. 4: The xStream platform.
Following the idea of xStream, we introduce xNode, which
is a network function providing an interface for control infor-
mation exchange between the network and endpoint devices.
Such communication is organized via a separate TCP con-
nection which is established between an xNode and either a
client application running at the UE or a server application
in the Internet, as shown in Fig. 4. In turn, xNode interacts
with other control plane functions (e.g., functions responsible
for resource allocation) to improve network performance. Note
that multiple instances of xNode can be deployed in order to
balance control traffic. Also, it is possible to use a separate
xNode for each specific service. For example, in the case of
URLLC service, xNode shall be deployed very close to the
data path to provide low latency control information exchange.
US
AN
Communication between applications and the network via
xNode provides the following benefits. First, the application
can directly inform the network about types of generated
data flow (e.g., VoIP, Video, Web), enabling easy and ac-
curate traffic classification without sophisticated deep packet
M
inspection and/or machine learning algorithms used in exist-
ing networks [49]. Second, the application can provide the
network with (i) specific QoS requirements (numeric values
ED
for required latency, throughput, etc) of a particular flow,
(ii) forthcoming traffic characteristics (packet size, packet
intensity) and (iii) the current state of the application (e.g., the
size of the buffer at the video client). With such information,
the network can provide specific service for the data flow
PT
and improve QoE for the end user. Moreover, it can fairly
allocate resources to maximize the overall performance, while
in traditional architectures, the applications always compete
CE
for the resources. For example, the network can temporarily
give more resources for a video client with low buffer to avoid
video playback interruption, if the quality for other users does
not suffer. Third, the network can inform the application about
AC
available resources and expected transmission characteristics,
which can be used by the application to generate traffic ac-
cordingly. For example, adaptive video streaming applications
can select an appropriate video bitrate/resolution by taking
into account the expected link throughput signaled by the
network. Thus xStream allows moving many decision-making
algorithms from applications to the xNode.
B. Multi-slice Radio Resource Management
Radio resource is the most expensive and scarce resource
in a wireless system. Therefore, the fundamental problem is
to design efficient Radio Resource Management (RRM) algo-
rithms. As mentioned in Section I, in 5G systems, this problem
9
Slice Radio Resource Manager
MAC Scheduler
Intra-slice Intra-slice Intra-slice Intra-slice
Scheduler 1 Scheduler 2 Scheduler 3 Scheduler 4
Inter-slice Scheduler
T
AirHYPE
IP
Fig. 5: Multi-slice RRM framework.
CR
is complicated by the need to share radio resource between
various services with substantially different QoS requirements
and also between various SePs. In the existing architectures,
the network slicing typically uses isolated resources [50], [51].
The network is divided into slices, which serve traffic of a par-
ticular service (e.g., eMBB, URLLC), and each slice obtains a
non-conflicting set of radio resources. With this approach, for
each slice we can use a specific RAT and scheduling algorithm
that takes into account slice QoS requirements. For example,
for an eMBB slice, the scheduler can maximize throughput,
while for a URLLC slice, the scheduler shall satisfy tight
latency and reliability requirements consuming the minimal
amount of radio resources. Moreover, such isolated slices
are very favorable for SePs, since they can implement their
own scheduling policies independently from InP. However,
this approach degrades spectral efficiency and reduces overall
performance. Since resource sets are isolated, one slice cannot
use the resources allocated to another slice, even if those
resources are not in use currently. Moreover, resource isolation
diminishes channel diversity gain, i.e., UEs from different
slices cannot use all resources to find resource blocks with
the highest quality, which, in turn, reduces spectral efficiency.
The problems described above can be partially addressed by
using a common multi-slice scheduling algorithm which takes
into account the QoS requirements for each slice. Such an
algorithm would dynamically distribute resource blocks every
TTI between slices and between flows in each slice. For that
it would typically be RAT-specific and deployed at (or close
to) the UND that performs MAC layer functions. However,
in ARBAT, for various slices different functional splits can
be used. In particular, to implement CoMP techniques, for
the eMBB slice, the corresponding scheduler can be deployed
at a UND controlling several UNDs with RF front-ends. On
the contrary, for URLLC slice the whole protocol stack and
the scheduler shall be deployed at a UND with RF front-end
because of tight latency requirements. Thus, we cannot use a
joint scheduler for eMBB and URLLC slices in the considered
scenario.
To enable flexible and efficient resource sharing between
various slices, we propose a multi-slice RRM framework as
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
Data Plane
Function
SRRM
Control Plane UND
Function
eMBB
data
RLC
MAC
MAC scheduler
PHY IQ samples
UND
IQ samples
Front-end
RF
UND
US
Fig. 6: Operation of multi-slice RRM framework in case of eMBB and URLLC resource slicing.
AN
shown in Fig. 5 that utilizes virtualized radio resources, and
consists of the following components: (i) Slice Radio Resource
Manager (SRRM) for long-term resource allocation, (ii) multi-
slice MAC scheduler for short-term allocations which consist
M
of a set of intra-slice schedulers and inter-slice scheduler, and
(iii) a wireless hypervisor (AirHYPE) for resource virtualiza-
tion and conflict resolution. In the following, we describe each
of these components in more detail.
ED
SRRM operates on a long-term timescale and determines
the average amount of radio resources that should be allocated
to each slice. For that, SRRM can use information obtained
from applications via xNode (e.g., number of flows in each
PT
slice, their QoS requirements and characteristics). For each
slice, SRRM determines: (i) appropriate RAT(s), and (ii) radio
resources that can be used to serve this slice. Note that SRRM
can allocate overlapping sets of virtualized radio resources for
CE
several slices. In this case, SRRM shall determine the long-
term share of resources that can be used by each slice. The
decisions made by SRRM and slice QoS requirements are
signaled to the MAC scheduler.
AC
The MAC scheduler operates on a short-term timescale
(i.e., every Time Transmission Interval) and determines which
Resource Blocks (RBs) and at which transmission points (RF
front-ends) shall be used to transmit a particular data flow.
Since in our framework, slices can use an overlapping set
of resources, we allocate RBs to flows belonging to different
slices in two stages. At the first stage, for each slice, the intra-
slice scheduler considers all RBs and pre-allocates them to the
corresponding flows using slice-specific policy. The policy can
take into account slice QoS requirements, and, moreover, it can
be provided/configured by the SeP (e.g., SeP can prioritize
some specific flows). Since slices can use shared resources,
the same RBs can be allocated to flows belonging to different
10
Long-term resource
allocation
URLLC
data
RLC
MAC
scheduler MAC
Short-term resource
T
allocation PHY
IP
IQ samples
AirHYPE
CR
Front-end
RF
UND
slices. Thus, at the second stage, the inter-slice scheduler
resolves such a conflict and selects a single flow to serve in
the given RB. For that, the inter-slice scheduler can use the
following information: spectral efficiency of competing flows
in the considered RB, the average share of resources allocated
to the corresponding slices, QoS/QoE requirements of the
flows, etc. The actual policy used by the inter-slice scheduler
is determined by the InP, but this policy shall ensure that in
the long-term each slice obtains a share of resources given by
SRRM. An example of a multi-slice MAC scheduler can be
found in [47]. In the paper, we consider a scenario with web
and video slices and design intra- and inter-slice scheduling
policies which improve QoE for both types of traffic.
In the proposed RRM framework, MAC schedulers can be
deployed at different UNDs and can use overlapping sets of re-
sources. In particular, this is needed to enable efficient resource
sharing between eMBB and URLLC slices as shown in Fig. 6.
In the considered scenario, due to tight latency constrains all
functions for a URLLC slice (including MAC scheduler) shall
be deployed at the UND with RF front-end (see right part of
Fig. 6). In contrast, for an eMBB slice we can deploy MAC
scheduler at UND controlling several RF front-ends (e.g., in
order to implement CoMP technique). Thus, in the considered
scenario, different slices (and their MAC schedulers) can use
overlapping time-frequency resources that ultimately map to
the same physical RF front-end. To resolve such conflicts
and multiplex different data streams, we introduce a wireless
hypervisor called AirHYPE.
Below we provide an example how to design AirHYPE
for OFDM-based RATs (e.g., LTE, Wi-Fi, and NR). First, to
provide flexibility and RAT-agnostic operation, we recognize
that AirHYPE must be placed as low in the protocol stack
as possible. In particular, by placing AirHYPE between the
 ACCEPTED MANUSCRIPT

FOR SUBMISSION TO JOURNAL, OCTOBER 2018 11

Control Information

Frequency- AirHYPE
eCPRI STAGE 1: domain
Stream 1 Convert to IQ
UND 1 BUFFER
frequency-
domain IQ VRG1
STAGE 3:
Time-domain
STAGE 2: PRG IQ
Convert to
RESOURCE RF
time-domain
Frequency- RE-MAPPER
IQ
eCPRI STAGE 1: domain VRGN
Stream N Convert to IQ
UND N BUFFER
frequency-

T
domain IQ

IP
Control Information
Fig. 7: AirHYPE – The Wireless Hypervisor.

CR
lower physical layer and RF front-end, we can ensure that it VRG1
processes only time- or frequency-domain IQ samples from
different streams, the nature of which does not depend on
the considered RAT. At the same time, we note that in its
current form, ARBAT virtualizes radio resources for each RAT
separately, and therefore AirHYPE does not multiplex streams
across different RATs. Second, we design AirHYPE to make
US VRG2
RESOURCE
PRG
AN
RE-
MAPPER
use of the existing signaling framework provided by eCPRI,
thus eschewing the need for extra control overhead. The goal
of AirHYPE is to multiplex several eCPRI streams into a URLLC
single eCPRI stream, which is then passed to physical RF eMBB
M

front-end. empty
The operation of AirHYPE is illustrated in Fig. 7. First, Fig. 8: Resource re-mapper example.
multiple eCPRI streams (1, . . . , N ) from different UNDs pro-
ED

vide the input data, which consists of serialized time- or

frequency-domain IQ samples. The first stage of AirHYPE any additional synchronization. Further, it does not require any
is to convert each input stream to a frequency-domain IQ modification to the physical layer and follows a plug-and-play
data set, where the ith set represents the Virtual Resource approach.
PT

Grid (VRG) corresponding to the ith input stream, with each

VRG being stored in a buffer. We note that if the incoming VII. M ANAGEMENT AND O RCHESTRATION (MANO)
data stream already consists of frequency-domain samples, The use of NFV within ARBAT requires a robust, yet
then the first stage can be skipped. The second stage is a low complexity, MANO framework as shown in Fig. 9.
CE

resource re-mapper which shall multiplex multiple VRGs into
a single Physical Resource Grid (PRG). The InP can deploy
any multiplexing algorithm at the resource re-mapper. For
example, as shown in Fig. 8, if a URLLC stream is competing
AC
The MANO framework is shared between the InP and SeP,
and is primarily tasked with the management of virtualized
infrastructure, orchestration of network services, and the life-
cycle management of VNFs [52]. It typically consists of

for a subset of the total physical resources required by an
eMBB stream, then the re-mapper, by recognizing the higher
priority of the URLLC stream (the priority is provided with
the control information in eCPRI stream), can preempt the
eMBB transmission and schedule the URLLC transmission.
Another option is that re-mapper can superpose IQ samples
from different streams. In this case, the receiver can use
successive interference cancellation techniques to demultiplex
the original streams. Finally, at the third stage, the obtained
PRG is converted to a time-domain data stream which is
passed on to the RF front-end for transmission.
We should note that AirHYPE takes advantage of the bit
synchronous interface of the eCPRI link and does not require
three major functional blocks: the Network Function Vir-
tualization Orchestrator (NFVO), Virtual Network Function
Manager (VNFM), and the Virtualized Infrastructure Manager
(VIM), along with a Network Catalog for network service
and network function definitions, that are commonly found
in solutions such as the Open Network Automation Platform
(ONAP) [53], and Open Source MANO (OSM) [54]. In
addition, we introduce two new components– the Network
Status Database (NSD) for monitoring the current network
state, and ServiceBRIDGE that acts as an interface between
the SeP and InP domains. Fig. 9 also shows the interactions
between the different functional blocks. Below, we provide
detailed information about each of them.
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
Service Provider
Network Catalog
MANO
InP1 InP2 InPN
Function
Fig. 9: The MANO framework.
A. Network Function Virtualization Orchestrator
As the part of MANO, the NFVO plays a key role in the
system performance, overseeing the global network resources
and allocating resources between network slices. The NFVO
directly interacts with SePs and is responsible for fulfilment
of SePs’ requests for service. The NFVO receives from each
service i the set of network requirements Ri such as maximal
US
packet delay, memory, capacity and etc. If InP cannot fulfil all
requirements for service, it is forced to pay penalty Pi given
by SeP which may be monetary. Using the information about
AN
the current network state from the NSD, and templates from
the catalog, the NFVO decides how to construct a network
service and dynamically modify it, fulfilling requirements Ri
for each service and minimizing the total penalty Ptot = ΣPi .
M
The NFVO manages the lifecycle of each network service, in-
cluding instantiation, scale-out/in, performance measurements,
event correlation, and termination.
ED
B. Virtual Network Function Manager
VNFM receives from the NFVO the set of requirements for
virtual functions that are need to be implemented. Apart from
PT
that, the VNFM is responsible for lifecycle management of
each VNF instance– scaling, changing operations, adding new
resources, and communicating between states of VNFs and
other functional blocks.
CE
C. Virtualized Infrastructure Manager
Operating under the purview of the InP, the VIM is respon-
AC
sible for controlling and managing the Infrastructure Plane
resources– computing, network, and radio; and therefore works
in close cooperation with the computing, network, and wireless
hypervisors. The mapping of physical resources to virtual
entities, and the associated lifecycle management fall under
the purview of the VIM. It is through the VIM’s Northbound
Interface (NBI) that physical and virtualized resources are
made available to the VNFM and the NFVO. Further, the
VIM also organizes virtual links, networks, and ports, and
is responsible for the management of the NSD repository
described in the following section. In ARBAT, we envision
multi-VIM, multi-site deployments that allow SePs to make
the use of infrastructure from multiple InPs.
12
ServiceBRIDGE MANO
Network Status
NFV OrchestratorN
DatabaseN
VNF ManagerN VIMN
VNFs NFV InfrastructureN
SeP InPN Shared
Function Function
T
IP
It is worth noting that such an approach allows imple-
menting dynamical multi-layer functional split in the network.
CR
A static split poses fixed requirements regarding throughput
and latency, which may be difficult to meet with a less-than-
ideal fronthaul. Instead, a dynamic functional split adapts to
fronthaul availability and use-case requirements. A dynamic
functional split plays a major role in enabling use-case adapt-
ability. Since we can map VNFs on different UNDs using
VIM, implementation of such a mechanism becomes possible.
D. Network Catalog
The Network Catalog is a repository of available PNFs,
VNFs, and network services that can be used by other blocks
such as the NFVO and VNFM for service instantiation. The
Network Catalog is a shared entity with its the contents being
owned by both the SeP and the InP. Each PNF and VNF
in the catalog is represented by a template called the PNF
Descriptor (PNFD) and VNF Descriptor (VNFD) respectively,
which captures its deployment and operational behavior. On
the one hand, a PNFD describes the functionality of a PNF,
along with available interfaces. On the other hand, a VNFD
contains information relating to the sub-components and their
dependencies and interconnections, resource allocation crite-
ria, geo-location placement, etc. of a VNF. Similarly, network
service behavior is captured by the Network Service Descriptor
(NSD). The NSD consists of PNFDs and VNFDs associated
with the constituent PNFs and VNFs, the VNF-Forwarding
Graphs (VNF-FGs) that form the service, VNF dependency
requirements, and interconnection link requirements. While
the InP makes use of the Network Catalog to carry out
deployments, the SeP is responsible for the onboarding of its
PNFs, VNFs, and network services.
E. Network Status Database
The NSD holds information about the devices in the net-
work, the links between them, and the deployed services. The
NSD provides real-time information about resource usage and
availability, along with service status. Given the potentially
large and distributed nature of the NSD, it is implemented in
the form of a NoSQL database, with the document store, key-
value store and, graph databases being potential candidates
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
Network Status
Network Catalog ServiceBRIDGE
Database
2. Network Catalog Query
3. Infrastructure Status Query
4. Network Catalog Response
5. Infrastructure Status Response
9. Infrastructure Status Update
14. Service Update
Fig. 10: The service instantiation and delivery procedure.
for implementation. More specifically, the NSD contains the
following information:
• UNDs: UID, special features (P4, FPGA, etc.), geograph-
ical location, resource availability, resource occupancy,
and cost per unit of resource.
• TLs: Endpoints, type (wired or wireless), resource avail-
ability, resource occupancy, latency, and cost per unit of
resource.
• Services: SID, owner (tenant ID), constituent VNFs
(VNF IDs), UNDs (UIDs), logical links (endpoints),
resource utilization, and SLA (latency and throughput
requirements).
M
Further, all information stored in the NSD except for that
relating to services is exposed to the SeP through Service-
BRDIGE, i.e., the SeP can only view information pertaining
to resource availability and pricing.
ED
F. ServiceBRIDGE
Given the strict separation of InP and SeP domains in
ARBAT, ServiceBRIDGE is intended as an interface between
PT
these two stakeholders. More specifically, in ARBAT, the
infrastructure and lifecycle management of network functions
are the responsibility of the InP in order to achieve the optimal
resource utilization. For example, the InP is free to place VNFs
CE
on UNDs under its purview, so long as the SLA requirements
from the SeP are met. However, this poses a challenge in a
multi-InP environment, where the SeP may require a service
that spans multiple InP domains. Since it is impractical to
AC
expect different InPs to converge to a common provisioning
decision, we introduce ServiceBRIDGE as shown in Fig. 9.
ServiceBRIDGE interfaces with the Network Catalog from
the SeP, and with the NSD and NFVO from each InP. From the
Network Catalog it obtains the VNF-FG associated with the
service, and from the NSD it receives information about the
resource availability in each InP’s domain, along with the per-
unit resource cost. The primary function of ServiceBRIDGE
is to partition a large service chain covering multiple InP
domains, into multiple small sub-chains that cover a single InP
domain. Each of these sub-chains has their own set of SLAs
derived from the SLA of the parent service chain. These sub-
chains are delivered to the corresponding NFVOs, which then
13
NFV Orchestrator VNF Manager VIM
1. Sub-Chain
6. Embedding and
Chaining
7. Resource Reservation Request
8. Resource Reservation Response
10. Instantiation Request
11. VNF Configuration
T
13. Acknowledgement 12. Acknowledgement
IP
CR
perform service orchestration based on additional information
from the Network Catalog. The metric used for service chain
partitioning is left for the SeP to decide. For example, the SeP
could choose profit maximization as the basis for partitioning,
US or could choose to partition for lower operating costs by
allowing flexibility in the SLA requirements of the sub-chains.
On the other hand, since the chain partitioning takes place at a
level above the NFVO, the InP is not aware of this procedure.
AN
From the InP’s perspective, the NFVO component receives
multiple sub-chains from different SePs simultaneously, and
interacts with the Network Catalog, VNFM, and VIM to
embed these chains on to the underlying infrastructure in a
manner that achieves optimal resource utilization. Thus, the
InP is only responsible for delivering the SLA associated
with the sub-chain. To summarize, ServiceBRIDGE allows for
provisioning of network services across multiple infrastructure
domains, without the complexity associated with inter-domain
interaction.
Within ARBAT, the Service Instantiation and Delivery
procedure is responsible for provisioning network service
requests from the SeP. In Fig. 10, it is assumed that the
service chain partitioning procedure has been completed by
ServiceBRDIGE. In Step 1, ServiceBRIDGE delivers the sub-
chain to the NFVO, which in turn requests the Network
Catalog for details about the constituent VNFs, and the NSD
for current network status, in Steps 2–5 respectively. Using
the information received, the NFVO determines the optimal
placement of VNFs over the virtualized infrastructure, and
their interconnections as shown in Step 6. Step 7 involves
the NFVO sending a resource reservation request to the VIM.
Once the VIM performs the resource allocation, it sends an
acknowledgment back to the NFVO, which in turn updates
the NSD in Steps 8-9. Then, an instantiation request is sent to
the VNFM in Step 10, along with the information relating to
instantiation parameters such as the resource requirements of
VNFs, interconnecting link bandwidth requirements, scalabil-
ity parameters, etc. The VNFM interacts with the VIM in Step
11, and configures the VNFs along with the interconnecting
links over the virtualized infrastructure. Steps 12–13 repre-
sent the successive acknowledgements that are sent from the
VNFM to the NFVO, and from the NFVO to ServiceBRIDGE.
 ACCEPTED MANUSCRIPT

FOR SUBMISSION TO JOURNAL, OCTOBER 2018 14

TABLE I: Main properties of ARBAT in comparison with key existing SDMN architectures.
(a) Infrastructure, Control Plane, Scalability, and Modularity.
System Architecture Infrastructure Control Plane Scalability Modularity
Limited flexibility Limited
- Base station: Monolithic
SoftNet [13] in control function (control logic NA
- CN–RAN: Hierarchical
distribution concentrated in CN)
- Base station: Partially disaggregated No control & data Limited Limited
TIP vRAN [15]
- CN–RAN: Hierarchical plane separation (absence of SDN) (can add RRUs only)
- Base station:
Limited
Fully disaggregated No control & data Limited
NEC NFV C-RAN [16] (can add RUs
- CN–RAN: plane separation (absence of SDN)
with L2 functions)
Hierarchical
Mixed
M-CORD - Base station: Partially disaggregated Dynamic distribution
High (Fully modular CN and
(xRAN + ProgRAN) [20] - CN–RAN: Hierarchical of control functions

T
partially modular RAN)
High
- Base station: Fully disaggregated Dynamic distribution
METIS-II [8] High (Fully modular RAN,

IP
- CN–RAN: Hierarchical of control functions
not focused on CN)
- Base station: Fully disaggregated Dynamic distribution High
ARBAT High
- CN–RAN: Unified of control functions (UNDs and UCN)

CR
(b) Fronthaul Adaptability, Multi-RAT Support, Network Slicing, and Special MANO Features.
System Architecture Fronthaul Adaptability Multi-RAT Network Slicing Special MANO Features
Limited Supported
SoftNet [13] No specific solution No specific solution
(single functional split) (Unified RAN)
TIP vRAN [15]

NEC NFV C-RAN [16]

M-CORD
Limited
(PHY layer split only)
High
(flexible functional splits)
Limited
US
Not Supported

Not Supported
Limited Support
End-to-End Network Slicing
(E2E-NS)
No specific solution
No specific solution

No specific solution
AN
Through ProgRAN No specific solution
(xRAN + ProgRAN) [20] (single functional split) (LTE and Wi-Fi)
Supported
High Spectrum Assignment
METIS-II [8] (AIV-specific and AIV-agnostic slicing
(flexible functional splits) Coordination
AIV-agnostic functions)
Supported
High
ARBAT (UNDs and Dynamic ServiceBRIDGE
M

(flexible functional splits)

network agents)
Note: “NA” implies that the parameter is not available in published sources.
ED

Finally, in Step 14, ServiceBRIDGE updates the NSD to can consist of either monolithic or disaggregated base
reflect the newly provisioned service, and service delivery is stations. Further, the disaggregation can either be partial
complete. or full. In partial disaggregation only a limited part of
the base station functionality is disaggregated, while in
PT

VIII. Q UALITATIVE E VALUATION full disaggregation, there is complete flexibility in the

distribution of base station functions. From a scalability
Currently, we are in the process of implementing the entire and cost standpoint, a fully disaggregated base station,
ARBAT architecture and carrying out a quantitative evaluation as used in ARBAT, would be the preferred option. At
CE

and comparison of ARBAT with existing architectures. This
task is complicated by the fact that although the existing
architectures are supposed to be open systems, unfortunately
that is not the case. Consequently, in this section, we perform
AC
the same time, instead of using a hierarchical CN–RAN,
ARBAT uses a UCN which provides design flexibility,
and simplifies reconfiguration of the network and the
deployment of new services.

a qualitative comparison of ARBAT with the state of the art
SDMN solutions described in Section II. Our comparison is
based on the following properties.
• Infrastructure: The SDMN solutions we discussed in
Section II employ SDN, NFV or a combination of both.
While neither is necessary for the other to exist, a
combination of the two allows for the implementation of
a wide variety of features– policy-based control, network
slicing, network automation, etc. and is the preferred
solution. ARBAT implements both SDN and NFV at
the grassroots level by introducing the concept of UND,
which supports PNFs, and allows for resource virtualiza-
tion that supports the deployment of VNFs. The system
• Control Plane: There are two aspects of the control plane
design. First, the control plane can be either be physically
centralized or physically distributed, with the latter being
preferred from a system scalability perspective. Second,
the control plane function distribution can either be static
or dynamic. While a static function distribution is easy to
implement, a dynamic distribution allows the system to
adapt to a variety of use-cases. ARBAT has a physically
distributed control plane with dynamic distribution of
control functions, which makes the 5G system scalable
and flexible. Further, ARBAT also introduces a novel
user application–network interface, namely xNode, which
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
significantly enhance traffic engineering.
• Scalability: It is measured by how well the network
responds to increase in traffic, and is characterized by
the absence of links or network components that are
prone to congestion. A major benefit of ARBAT in this
regard is that it scales well in response to an increase
in the number of users, and the amount of traffic, before
requiring provisioning of new hardware and additional
capacity. The disaggregated approach to network design,
decentralization of control functionalities, and UCN– all
play a major role in enhancing the scalability. Further,
the use of UNDs adds an additional layer of reliability to
the network, as network functions can be migrated easily
to the nearest UND in the case of device failure.
• Modularity: There are two aspects to network
modularity– hardware and software. First, from the hard-
ware perspective, the addition of new components must
follow a plug-and-play approach. To this end, the con-
cepts of UND and UCN, that allow for rapid provision-
ing of resources and practically unlimited flexibility in
network design, are two significant enablers towards a
modular network. Second, from the software perspective,
the set of VNFs deployed across UNDs can be modified
easily using the MANO framework as has been described
in Section VII. Thus, ARBAT exhibits significant modu-
US
AN
larity, and can seamlessly adapt to changing use-cases.
• Fronthaul Adaptability: The system should have provi-
sion for functioning with a variety of use-cases, includ-
ing eMBB, mMTC and URLLC. To optimize spectral
M
efficiency and to satisfy strict requirements for different
slices, different functional splits are needed. In other
words, use-case adaptability is tied to a dynamic and
distributed control plane that allows changes in network
ED
function distribution, which also allows the architecture
to function with a variety of fronthaul options. ARBAT
supports flexible functional splits, which results in high
fronthaul adaptability.
PT
• Multi-RAT: In contrast to many other existing archi-
tectures, ARBAT is RAT-agnostic and supports multiple
RATs via the same generalized architecture. Owing to
the use of UNDs and network agents, ARBAT can easily
CE
integrate devices supporting existing and emerging RATs.
• Network Slicing: The use of network slicing allows
operators to offer differentiated services over the same
infrastructure while optimizing network resource utiliza-
AC
tion, and as such has been recognized by 3GPP as a
key feature [10]. Therefore, support for slicing is a key
required feature. While many existing architectures leave
slicing out of the consideration, or proposed a semi-static
isolated slicing, ARBAT is the first-ever architecture with
native support for dynamic slicing. The latter is especially
important for RRM, since typically radio resource be-
come a bottleneck for the entire system. The designed
modular multi-slice RRM allows both achieving high
spectral efficiency and fulfillment of QoS requirements
for different slices.
• Special MANO features: To manage the virtualized
infrastructure and orchestrate network services, ARBAT
15
uses a rich MANO framework. In ARBAT, MANO is
complicated because of the necessity to support the strict
separation of InP and SeP domains. To address this issue,
it uses ServiceBRIDGE as an interface between these two
stakeholders.
To this end, we have summarized the main features of AR-
BAT in comparison with other system architectures in Table I.
Qualitatively, we consider ARBAT to be the most feature-
complete SDMN architecture based on its infrastructure and
control plane design, multi-layer RRM framework, ability to
adapt to a variety of use-cases and fronthaul options, high
scalibility, multi-RAT support and modularity.
T
IP
IX. C ONCLUSION
In this paper, we have introduced a novel architecture
CR
for 5G and beyond wireless systems, called ARBAT. AR-
BAT has many innovative features which aim at providing
highly efficient QoE-aware communications in heterogeneous
environments with low CAPEX and OPEX. Specifically, by
following the virtualization paradigm and replacing the hierar-
chical CN and RAN with a UCN consisting of UNDs, we can
easily bring the external data network closer to users, thereby
reducing latency and enabling the URLLC use-case. At the
same time, for highly efficient spectrum usage and operation
with massive antennas, control functions can be deployed at
a fewer number of central UNDs with higher computational
capabilities. Moreover, the concept of UNDs allows ARBAT
to both integrate legacy devices with hardware-defined PNFs,
as well as make use of infrastructure resources for running
various network functions efficiently.
Tight communication between applications and the network
through the xStream platform combined with the original
multi-slice modular resource management aims at maximiz-
ing network capacity with respect to the provided QoE for
different slices. xStream synchronizes network capabilities
and application demands excluding computational-heavy non-
optimal and error-prone machine learning-based decisions on
traffic characteristics and network capabilities. Non-isolated
slicing paradigm avoids wasting of channel resources and al-
lows maximizing user-perceived spectral efficiency, while the
modular design of resource allocation allows running MNVO-
defined radio-resource schedulers for particular slices. The
latter together with enhanced MANO with ServiceBRIDGE
simplifies multi-tenant orchestration. We are in the process
of fully implementing the ARBAT architecture and making it
publicly available soon.
R EFERENCES
[1] The 3rd Generation Partnership Project (3GPP). (2018) Release 15.
[Online]. Available: http://www.3gpp.org/release-15
[2] “Cisco Visual Networking Index,” Cisco Systems, Inc., Tech. Rep., 2017.
[3] “Framework and overall objectives of the future development of IMT
for 2020 and beyond,” International Telecommunication Union (ITU),
Tech. Rep. RM.2083-0, Sep. 2015.
[4] A. Sorsaniemi, “5G and Energy Efficiency,” DG Connect, EC, Tech.
Rep., Nov. 2017.
[5] L. Duan, J. Huang, and J. Walrand, “Economic analysis of 4G network
upgrade,” in 2013 Proceedings IEEE INFOCOM, Apr. 2013.
 ACCEPTED MANUSCRIPT
FOR SUBMISSION TO JOURNAL, OCTOBER 2018
[6] I. F. Akyildiz, S. Nie, S.-C. Lin, and M. Chandrasekaran, “5G roadmap:
10 key enabling technologies,” Computer Networks, vol. 106, pp. 17–48,
Sep. 2016.
[7] X. Jin, L. E. Li, L. Vanbever, and J. Rexford, “SoftCell,” in Proceedings
of the ninth ACM conference on Emerging networking experiments and
technologies - CoNEXT ’13, 2013.
[8] H. Tullberg, P. Popovski, Z. Li, M. A. Uusitalo, A. Hoglund, O. Bulakci,
M. Fallgren, and J. F. Monserrat del Rı́o, “The METIS 5G system
concept: Meeting the 5G requirements,” in IEEE Communications Mag-
azine, vol. 54, no. 12. Institute of Electrical and Electronics Engineers
(IEEE), 2016, pp. 132–139.
[9] X. Foukas, N. Nikaein, M. M. Kassem, M. K. Marina, and K. Kon-
tovasilis, “FlexRAN,” in Proceedings of the 12th International on
Conference on emerging Networking EXperiments and Technologies -
CoNEXT 16, 2016.
[10] The 3rd Generation Partnership Project (3GPP). (2018) Release 14.
[Online]. Available: http://www.3gpp.org/release-14
[11] TSG-Radio Access Network, “Study of separation of NR Control Plane
(CP) and User Plane (UP) for split option 2,” The 3rd Generation
Partnership Project (3GPP), Tech. Rep. TR 38.806, 2018, version 15.0.
[12] V. G. Vassilakis, I. D. Moscholios, B. A. Alzahrani, and M. D.
Logothetis, “A software-defined architecture for next-generation cellular
networks,” in 2016 IEEE International Conference on Communications
(ICC), May 2016.
[13] H. Wang, S. Chen, H. Xu, M. Ai, and Y. Shi, “SoftNet: A software
defined decentralized mobile network architecture toward 5G,” IEEE
Network, vol. 29, no. 2, pp. 16–22, 2015.
[14] A. Gopalasingham, L. Roullet, N. Trabelsi, C. S. Chen, A. Hebbar, and
E. Bizouarn, “Generalized software defined network platform for radio
access networks,” in IEEE Consumer Communications and Networking
Conference (CCNC), 2016.
[15] “Creating an ecosystem for vRANs supporting non-ideal fronthaul,”
US
AN
Telecom Infra Project, Tech. Rep., 2017.
[16] “NFV C-RAN for Efficient RAN Resource Allocation,” NEC Corpora-
tion, Tech. Rep., 2016.
[17] The xRAN Alliance. (2018) xRAN Forum. [Online]. Available:
http://www.xran.org
[18] S. Singh and A. Raja, “xRAN Reference Controller Design Specifica-
M
tion,” The Open Networking Foundation (ONF), Tech. Rep., 2017.
[19] xRAN Fronthaul Working Group, “Control, User and Synchronization
Plane Specification,” xRAN.org, Tech. Rep. XRAN-FH.CUS.0, 2018,
version 2.0.
[20] “Central Office Rearchitected as a Datacenter (CORD),” Open Network-
ED
ing Foundation (ONF), Tech. Rep., 2016.
[21] “M-CORD as an Open Reference Solution for 5G Enablement,” Open
Networking Foundation (ONF), Tech. Rep., 2017.
[22] NGMN Alliance, “5G End-to-end Architecture Framework,” Tech. Rep.,
2017.
PT
[23] A. Banchs, M. Breitbach, X. Costa, U. Doetsch, S. Redana, C. Sartori,
and H. Schotten, “A Novel Radio Multiservice Adaptive Network
Architecture for 5G Networks,” in 2015 IEEE 81st Vehicular Technology
Conference (VTC Spring), May 2015.
[24] J. Gutiérrez, N. Maletic, D. Camps-Mur, E. Garcı́a, I. Berberana,
CE
M. Anastasopoulos, A. Tzanakaki, V. Kalokidou, P. Flegkas, D. Syrivelis
et al., “5G-XHaul: a converged optical and wireless solution for 5G
transport networks,” Transactions on Emerging Telecommunications
Technologies, vol. 27, no. 9, pp. 1187–1195, 2016.
[25] N. Feamster, J. Rexford, and E. Zegura, “The road to SDN,” ACM
SIGCOMM Computer Communication Review, vol. 44, no. 2, pp. 87–98,
AC
Apr. 2014.
[26] “Network Functions Virtualisation,” European Telecommunications
Standards Institute (ETSI), Tech. Rep., 2012.
[27] P. Bosshart, G. Varghese, D. Walker, D. Daly, G. Gibb, M. Izzard,
N. McKeown, J. Rexford, C. Schlesinger, D. Talayco, and A. Vahdat,
“P4: programming protocol-independent packet processors,” ACM SIG-
COMM Computer Communication Review, vol. 44, no. 3, pp. 87–95,
Jul. 2014.
[28] TSG-Services and System Aspects, “Study on management and or-
chestration of network slicing for next generation network,” The 3rd
Generation Partnership Project (3GPP), Tech. Rep. TR 28.801, 2018,
version 15.1.
[29] J. Aweya, “IP router architectures: an overview,” International Journal
of Communication Systems, vol. 14, no. 5, pp. 447–475, 2001.
[30] TSG– Services and System Aspects, “System Architecture for the 5G
System,” The 3rd Generation Partnership Project (3GPP), Tech. Rep. TS
23.501, 2018, version 15.2.
16
[31] M. Jaber, M. A. Imran, R. Tafazolli, and A. Tukmanov, “5G Backhaul
Challenges and Emerging Research Directions: A Survey,” IEEE Access,
vol. 4, pp. 1743–1766, 2016.
[32] B. Guo, W. Cao, A. Tao, and D. Samardzija, “LTE/LTE-a signal com-
pression on the CPRI interface,” Bell Labs Technical Journal, vol. 18,
no. 2, pp. 117–133, Sep. 2013.
[33] A. Kivity, Y. Kamay, D. Laor, U. Lublin, and A. Liguori, “KVM: the
Linux Virtual Machine Monitor,” in In Proceedings of the 2007 Ottawa
Linux Symposium (OLS-07), 2007.
[34] M. Helsley, “LXC: Linux container tools,” IBM devloperWorks Technical
Library, vol. 11, 2009.
[35] P. Barham, B. Dragovic, K. Fraser, S. Hand, T. Harris, A. Ho, R. Neuge-
bauer, I. Pratt, and A. Warfield, “Xen and the art of virtualization,” in
Proceedings of the nineteenth ACM symposium on Operating systems
principles - SOSP ’03, 2003.
[36] J. A. Kappel, A. Velte, and T. Velte, Microsoft Virtualization with Hyper-
T
V: Manage Your Datacenter with Hyper-V, Virtual PC, Virtual Server,
and Application Virtualization. McGraw-Hill, Inc., 2009.
IP
[37] R. Sherwood, G. Gibb, K.-K. Yap, G. Appenzeller, M. Casado, N. Mck-
eown, and G. Parulkar, “FlowVisor: A network virtualization layer,”
OpenFlow Consortium, Tech. Rep., 2009.
[38] A. Blenk, A. Basta, M. Reisslein, and W. Kellerer, “Survey on Network
CR
Virtualization Hypervisors for Software Defined Networking,” IEEE
Communications Surveys & Tutorials, vol. 18, no. 1, pp. 655–685, 2016.
[39] D. Hancock and J. van der Merwe, “HyPer4,” in Proceedings of the 12th
International on Conference on emerging Networking EXperiments and
Technologies - CoNEXT’16, 2016.
[40] C. Zhang, J. Bi, Y. Zhou, A. B. Dogar, and J. Wu, “HyperV: A High
Performance Hypervisor for Virtualization of the Programmable Data
Plane,” in 2017 26th International Conference on Computer Communi-
cation and Networks (ICCCN), Jul. 2017.
[41] M. Kalil, M. Youssef, A. Shami, A. Al-Dweik, and S. Ali, “Wireless re-
source virtualization: opportunities, challenges, and solutions,” Wireless
Communications and Mobile Computing, vol. 16, no. 16, pp. 2690–2699,
Jul. 2016.
[42] X. Foukas, G. Patounas, A. Elmokashfi, and M. K. Marina, “Network
Slicing in 5G: Survey and Challenges,” IEEE Communications Maga-
zine, vol. 55, no. 5, pp. 94–100, May 2017.
[43] “PDU Session User Plane protocol,” The 3rd Generation Partnership
Project (3GPP), Tech. Rep. 38.415, 2018, version 15.1.0.
[44] G. Cheng, H. Chen, H. Hu, Z. Wang, and J. Lan, “Enabling network
function combination via service chain instantiation,” Computer Net-
works, vol. 92, pp. 396–407, Dec. 2015.
[45] TSG-Services and System Aspects, “Study on charging aspects of 5G
system architecture phase 1,” The 3rd Generation Partnership Project
(3GPP), Tech. Rep. TR 32.899, 2018, version 15.0.
[46] 5G Innovation Center, “The Flat Distributed Cloud (FDC) 5G Architec-
ture Revolution,” Institute for Communication Systems, University of
Surrey, Tech. Rep., Jan. 2016.
[47] I. F. Akyildiz, E. Khorov, A. Kiryanov, D. Kovkov, A. Krasilov, M. Li-
ubogoshchev, D. Shmelkin, and S. Tang, “xStream: A New Platform
Enabling Communication between Applications and the 5G Network.”
in 2018 IEEE Globecom Workshops (GC Wkshps), Dec. 2018.
[48] E. Khorov, A. Krasilov, and A. Malyshev, “Reliable low latency com-
munications in LTE networks,” in 2017 IEEE International Black Sea
Conference on Communications and Networking (BlackSeaCom), Jun.
2017.
[49] H. Kim, K. C. Claffy, M. Fomenkov, D. Barman, M. Faloutsos, and
K. Lee, “Internet traffic classification demystified: myths, caveats, and
the best practices,” in Proceedings of the 2008 ACM CoNEXT confer-
ence, 2008, p. 11.
[50] O. Sallent, J. Perez-Romero, R. Ferrus, and R. Agusti, “On Radio Access
Network Slicing from a Radio Resource Management Perspective,”
IEEE Wireless Communications, vol. 24, no. 5, pp. 166–174, Oct. 2017.
[51] M. Richart, J. Baliosian, J. Serrat, and J.-L. Gorricho, “Resource Slicing
in Virtual Wireless Networks: A Survey,” IEEE Transactions on Network
and Service Management, vol. 13, no. 3, pp. 462–476, Sep. 2016.
[52] “Network Functions Virtualisation (NFV); Management and Orchestra-
tion,” European Telecommunications Standards Institute (ETSI), Tech.
Rep. ETSI GS NFV-MAN 001, 2014, version 1.1.1.
[53] “Architecture Overview,” Open Network Automation Platform (ONAP),
Tech. Rep., 2018.
[54] Open Source MANO, “OSM Release FOUR Technical Overview,”
European Telecommunications Standards Institute (ETSI), Tech. Rep.,
2018.
 ACCEPTED MANUSCRIPT

FOR SUBMISSION TO JOURNAL, OCTOBER 2018 17

Ian F. Akyildiz is currently the Ken Byers Chair Professor
in Telecommunications with the School of Electrical and
Computer Engineering, Director of the Broadband Wireless
Networking Laboratory, and Chair of the Telecommunication
Group at Georgia Institute of Technology, Atlanta, USA.
Since 2011, he serves as a Consulting Chair Professor with
the Department of Information Technology, King Abdulaziz
University, Jeddah, Saudi Arabia, and with the Computer Engi-
neering Department at the University of Cyprus since January
2017. He is a Megagrant Research Leader with the Institute for
Information Transmission Problems at the Russian Academy
of Sciences, in Moscow, Russia, since May 2018. His current
sioning in wireless networks, development of analytical and
simulation tools for performance evaluation and optimization
of different wireless access technologies. He has co-authored
over 20 research papers.
Aleksey Kureev received the B.S degree and M.S. degree
in applied mathematics and physics from the Moscow Institute
of Physics and Technology (MIPT) in 2015 and 2017, respec-
tively. He is a researcher in Wireless Network Laboratory in
the Institute for Information Transmission Problems, Russian
Academy of Sciences. Currently, he is working toward the
PhD degree in telecommunications under the supervision of
Evgeny Khorov. His research interests includesoftware define

T
research interests are in 5G wireless systems, nanonetworks, networking, network function virtualization, massive machine-
Terahertz band communications, and wireless sensor networks to-machine communication, ultra-dense networks and SDR

IP
in challenged environments. He is an IEEE Fellow (1996) and prototyping of various radio access technologies.
an ACM Fellow (1997). He received numerous awards from
the IEEE and the ACM, and many other organizations. His h-

CR
index is 115, and the total number of citations is above 105K
as per Google scholar as of October 2018.
Ahan Kak received the B.S. degree in Electrical Engineer-
ing from VJTI, University of Mumbai in 2016. Currently, he
is working toward the PhD degree in electrical and computer
engineering under the supervision of Prof. Ian F. Akyildiz.
His research interests include software-defined networking,
Internet of things, and cellular communications.
US
AN
Evgeny Khorov is the Head of the Wireless Networks
Lab, Institute for Information Transmission Problems, Russian
Academy of Sciences. In 2015, he was a Visiting Research
Fellow with Kings College London. For breakthrough results
M

of the joint project, in 2015 and 2017 Huawei RRC awarded

him as the Best Cooperation Project Leader. Being a voting
member of IEEE 802.11, he has contributed to 802.11ax
standard with many proposals. He has authored over 70 papers.
ED

His main research interests are related to 5G and beyond

wireless systems, next generation Wi-Fi, protocol design and
cross-layer optimization. He has led dozens of national and
international projects sponsored by academia funds and in-
PT

dustry. He was a recipient of the Moscow Prize for Young

Scientists in 2013 and Russian Government Award in Science
and Technology for Young Scientists, the Best Paper Award
from IEEE ISWCS in 2012, and the Best Cited Review Paper
CE

Award from Elsevier Computer Communications in 2018. He

gives keynotes and tutorials and participates in panels at large
conferences, including IEEE Globecom in 2017, IEEE PIMRC
in 2017, IEEE ICC in 2016, ISWCS in 2014, and NEW2AN
AC

in 2018. He chairs TPC of the IEEE Globecom 2018 CA5GS

Workshop and IEEE BlackSeaCom 2019. He also serves as
an Editor for Ad Hoc Networks.
Artem Krasilov received his MS degree and PhD degree
in telecommunications from Moscow Institute of Physics and
Technology in 2010 and 2013, respectively. Currently he is
a senior researcher in the Institute for Information Transmis-
sion Problems, Russian Academy of Sciences. He has been
involved in various national and international research projects
supported by both academic foundations (FP7 programme,
Russian Government) and industrial partners. His professional
interests are related to SDN and NFV based architectures for
the next generation wireless networks, 5G, QoE/QoS provi-
 ACCEPTED MANUSCRIPT

FOR SUBMISSION TO JOURNAL, OCTOBER 2018 18

T
IP
CR
US
AN
M
ED
PT
CE
AC
 ACCEPTED MANUSCRIPT

FOR SUBMISSION TO JOURNAL, OCTOBER 2018 19

T
IP
CR
US
AN
M
ED
PT
CE
AC
 ACCEPTED MANUSCRIPT

FOR SUBMISSION TO JOURNAL, OCTOBER 2018 20

T
IP
CR
US
AN
M
ED
PT
CE
AC
 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/ACCESS.2018.2883788, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI

User Influence on Mobile Terminal

Antennas: A Review of Challenges and
Potential Solutions for 5G Antennas
RIZWAN KHAN1 , (Student Member, IEEE), AZREMI ABDULLAH AL-HADI1 , (SENIOR
MEMBER, IEEE), PING JACK SOH1 , (SENIOR MEMBER, IEEE), MUHAMMAD RAMLEE
KAMARUDIN2 , (SENIOR MEMBER, IEEE), MOHD TARMIZI ALI3 , (SENIOR MEMBER, IEEE)
AND OWAIS.4
1
School of Computer and Communication Engineering, Universiti Malaysia Perlis (UniMAP), 02600 Arau, Malaysia.
2
Centre for Electronic Warfare, Information and Cyber, Cranfield Defence and Security, Cranfield University, Defence Academy of the U.K., Shrivenham, SN6
8LA, U.K.
3
Antenna Research Centre, Fakulti Kejuruteraan Elektrik, Universiti Teknologi Mara, Shah Alam, Selangor.
4
Electrical Engineering Department, Comsats University, Abbottabad, Pakistan.
Corresponding author: Muhammad Ramlee Kamarudin (e-mail: Ramlee.Kamarudin@cranfield.ac.uk),
Azremi Abdullah Al-Hadi (e-mail: azremi@unimap.edu.my)
“This research was supported financially by the Ministry of Science, Technology and Innovation under Sciencefund (Grant no:
01-01-015-SF0258).”

ABSTRACT This article presents a comprehensive review of mobile terminal antenna researches per-
formed in the past seven years and the current challenges related to the user’s influence on the performance
of fifth generation (5G) terminal antennas. The main challenges for designing of mobile terminal antennas
are to meet the compact size requirements of built-in structures and their multiband capabilities. The antenna
design techniques that are used to achieve broader operating bandwidths with smaller antenna dimensions
will be first discussed. This is followed by the effects of user interactions with the head/hand for mobile
antennas, in terms of radiation efficiency and consequently, the correlation of Multiple Input Multiple
Output (MIMO) antenna systems. The ultimate aims of this survey article are as follows: 1) to highlight the
different frequencies of mobile terminal antennas for different applications, 2) to highlight mobile terminal
antennas which have been developed for 5G application, 3) to study and discuss the effects of user’s hand
on 5G mobile terminal antennas, 4) to discuss the research gap, issues and challenges in the field of user’s
effects on mobile terminal antennas for 5G applications. In addition to that, an investigation of the users’
hand effects on two MIMO mobile terminal antennas operational in the sub-6GHz 5G band is presented.
This investigation, performed using two MIMO antennas is an attempt to formulate guidelines on efficient
mobile terminal antenna design in the presence of user’s hand in C Band (from 3.4 to 3.6 GHz) and LTE-U
Band 46 (from 5.15 to 5.925 GHz).

INDEX TERMS Electromagnetic influence of users, mobile terminal antennas, multiband antennas, MIMO
antennas.

I. INTRODUCTION design of antennas. The rapid growth of mobile systems

VER the last three and a half decades, advancements
O in cellular communication standards have significantly
affected the development of terminal antennas. Two major
factors have been involved in this evolution. Firstly, is the
users’ demand, which is mainly composed of aesthetical and
ergonomic considerations, and secondly, the introduction of
new spectrums in line with the evolving regulatory standards.
Therefore, one of the most challenging requirements for
designers of mobile wireless communication systems is the
towards 5G requires antennas to be wideband, multiband
or even ultra wideband to enable coverage interoperability
of mobile services and to reduce system complexity. Other
challenges for mobile terminal antennas include size com-
pactness, ease of integration into the handheld chassis and
coexistence with and the support for multi-antenna operation
and MIMO systems. MIMO is one of the major enablers of
5G technology in achieving up to 100 times of bandwidth
relative to 4G and LTE-Advance (LTE-A) systems. This

VOLUME 4, 2016 1

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

List of Abbreviation.
3G 3rd Generation ME Multiplexing efficiency
4G 4th Generation MIMO Multiple input multiple output
5G 5th Generation MITT Ministry of Industry and Information Technology
AD1 Antenna Design 1 PCB Printed Circuit Board
AD2 Antenna Design 2 PCS Personal Communications Service
AE1 Antenna Element 1 PIFA Planar Inverted F Antenna
AE2 Antenna Element 2 RE Radiation Efficiency
CCE Capacitive Coupling Element RH Right Hand
DCS Digital Cellular System RR Radio Regulation
ECC Envelop Correlation Coefficient SAR Specific Absorption Rate
FS Free Space SISO Single Input Single Output
GSM Global System for Mobile communications UB Upper Band
IM Impedance Matching UMTS Universal Mobile Telecommunications System
ITU International Telecommunication Union UHF Ultra High Frequency
LB Lower Band WC With Casing
LH Left Hand WiMAX Worldwide Interoperability for Microwave Access
LTE-A Long Term Evolution-Advance WLAN Wireless Local Area Network
MC Mutual Coupling WRC World Radio Conference

tion activities to identify common usable spectrums across

FIGURE 1: Skin properties according to different frequen-
cies, (a) relative permittivity and (b) conductivity [12].
technology offers increased system capacity with additional
number of antenna elements, without the need for additional
power or frequency spectrum. A well performing MIMO
system requires high efficiency for each element and a low
correlation between them [1], [2]. However, this requires
sufficient inter-element spacing, which is difficult to find in
mobile terminals as they are ideally designed to be compact
and wideband. This is to ensure that new smart phones are
backward compatible, supporting technologies prior to the
fourth generation (4G) for global coverage and roaming using
the same device [3].
Planar inverted-F antenna (PIFA) is one of the most com-
monly used internal antennas in the handheld segment (such
as mobile phones) due to its ease of fabrication, low profile
and potentially low cost [4]. Another most commonly used
antenna type in mobile phones is the monopole antenna. Such
antennas started to gain popularity upon the introduction of
the new 3G, 4G and 5G frequency bands, as they are able to
offer broader bandwidth relative to PIFAs [4]. Additional par-
asitic elements such as capacitive coupling elements (CCE)
are typically used in mobile terminal antennas to excite in-
herently non-resonant structures. Their structure and location
are carefully designed to excite particular wave modes on the
ground plane in a systematic manner [5], [6].
Furthermore, there are significant on-going standardiza-
2 VOLUME 4, 2016
many countries to enable global roaming and economy of
scale. The World Radio Conference 2015 (WRC-15) revised
the Radio Regulations (RR) and the International Telecom-
munication Union (ITU) treaty for radio spectrum [7] iden-
tified and deployed 4G spectrum in many parts of the world.
As 5G standards and technologies continues to mature, bands
already in use for 4G will also evolve as 4G to 5G transition
bands [8]. One of such realistic bands announced is the C-
band between 3.4 and 3.6 GHz [7]. In early 2015, LTE band
42 (from 3.4 to 3.6 GHz) was initially defined as the band
of interest for China [9]. This was revised in November 2017
by China’s Ministry of Industry and Information Technology
(MITT), which officially reserved the 3.3 to 3.6 GHz and 4.8
to 5 GHz for the country’s 5G service, with the 3.3 to 3.6 GHz
band being intensively investigated as a pioneering band for
realizing 5G MIMO [10], Europe (from 3.4 to 3.8 GHz) [11],
and Korea (from 3.4 to 3.7 GHz) [12]. Besides that, another
potential sub-6 GHz frequency band for 5G is the LTE band
46 (from 5.15 to 5.925 GHz), also known as the unlicensed
LTE band (LTE-U).
Besides designing operational MIMO antennas in these
sub-6 GHz bands, an additional step in ensuring operation
in the design procedure is to study the users’ influence on
the performance of mobile terminal antennas. The antenna
design process for mobile terminals must also account for
the effects of its housing for optimized MIMO performance.
Besides that, upcoming regulatory standards are expected to
incorporate performance evaluation of these terminals with
hand models by manufacturers. This has propelled research
interests in quantifying the interaction between radiating
elements in wireless terminals and biological tissues of users.
Such concern is obvious given the fact that the parts of the
body (head, hand or body) located in proximity of the termi-
nal causes increase in absorption losses, impedance variation,
radiation pattern deterioration and detuning of resonance
frequency. Moreover, the resulting absorption losses varies
depending on the frequency due to the different permittivity

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 1: Summary of SISO Antennas [14]-[21].

Ref Antenna Design Bandwidth [GHz] Ref Antenna Design Bandwidth [GHz]

[14] 0.82 - 3 [18] 0.88 - 0.96

and 1.85 - 1.99

[15] 0.3 - 0.7 [19] 2.39 - 2.49

and
5.07 - 5.88

[16] 1.33 - 3.5 [20] 2.27 - 2.52

and
5.29 - 5.53

[17] 0.8 - 0.97 [21] 0.7 - 0.96

and
1.5 - 5.9

of user’s hand with frequency, besides their hand grips, as
shown in Fig. 1 [13]. This is due to the different penetration
depths resulting from the use of different frequency bands.
This paper is organized as follows. An overview of mobile
terminal antennas operating at different frequency bands is
first discussed. This is followed by a critical review of the
recent publications on mobile terminal for 5G application
over last seven years. From this review, current research
issues and challenges related to the user’s hand effects within
the sub-6 GHz 5G bands will be identified and discussed
in detail. Finally, a potential solution is presented via an
evaluation work on two different antennas operating in the
sub-6 GHz bands is performed in close vicinity of user’s hand
and presented. The concluding remarks presented at the end
of this work include future strategies to alleviate these effects
throughout the design process.
II. SISO ANTENNAS IN MOBILE TERMINALS
Throughout the years, mobile phones have undergone sig-
nificant changes in terms of size, shape and functionalities.
VOLUME 4, 2016
With each development, the smartphones are generally aimed
with more compact sizes and lighter weights for improved
mobility. Such specification requires antennas to be designed
to be broad- and/or multiband within a smaller available
volume. The following section will present the recent liter-
ature regarding wide-band and dual band Single Input Single
Output (SISO) antenna for mobile terminal.
A. WIDE BAND AND DUAL BAND MOBILE TERMINAL
ANTENNAS
Due to lack of wide band and dual band SISO antennas in the
past seven years, several designs will be summarized in this
section. It is also worthy to mention that a previous review
on mobile terminal antennas was presented in [3]. Wide band
antennas for mobile terminal have been studied in [14]-[21]
since 2011. Instead of implementing contacting feeds used
in conventional PIFAs, a coupled feed structure is used to
improve the bandwidth in [14], resulting in an operation from
0.82 to 3 GHz. Next, a monopole sleeve antenna is proposed
in [15].
3

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
An enhanced operational bandwidth from 0.3 GHz to
0.7GHz is achieved in this design by adding a conical sleeve
and helical structure on the top of the antenna. Meanwhile,
a wideband planar monopole antenna for mobile terminals
is proposed in [16]. The increase in bandwidth is achieved
by connecting two arms of a U-shape radiator using two
parallel radiating strips. Next, the wideband operation of a
PIFA antenna is enabled in [17] by introducing a slot on
the top of radiating element. Further down the frequency
spectrum, an internal broadband antenna is designed for the
lower Ultra High Frequency (UHF) band in [18], resulting in
its operation across all lower frequency cellular bands, from
0.7 to 0.96GHz. To enable dual band operation, combining
the operation of two radiators with different electrical lengths
is a typical method used. An example is provided in [19] for a
dual-band PIFA implementing a longer radiating element for
lower frequency resonance and a shorter element for higher
resonance. Besides that, a dual-band parasitic element is also
designed to reduce the effects of ground plane, this leads
to the reduction in the electromagnetic energy deposited in
the hand phantom and hence lower SAR. Next, a contour
integrated dual band compact antenna elements for low
profile mobile terminals is presented in [20]. This design
utilizes a multiple resonating structure to result in a dual
band operation at 2.4 and 5.5GHz. Additional resonance for
antennas may also be introduced by implementing coupling
strips, as seen in [21]. A dual-band antenna with shorted
coupling strips is presented for mobile terminals by using
this technique. This additional resonance combined with the
existing operating band to enable a wideband characteristic
for the proposed antenna. Antennas from [14]-[21] and their
bandwidths are summarized in Table-I
B. MULTI BAND MOBILE TERMINAL ANTENNAS
Next, recent published investigations on multiband mobile
terminal antennas for different bands are presented and dis-
cussed in this section [22]-[31]. Note that these antennas
are being integrated internally in the housing of the mobile
terminals, and thus can be used as a resonant cavity to
improve radiation. Moreover, another important trend is that
mobile terminals are increasingly integrated with multimedia
capabilities since the late 1990s [3]. The first antenna in [22]
combines different wave modes of the antenna and chassis
to achieve multiband antenna without any matching circuits
or magneto-dielectric materials. Another use of modes for
this purpose is a dual zeroth-order resonance by using com-
posite right and left handed metamaterial line [23]. Next, an
internal quad band printed monopole antenna for an oval-
shaped mobile phone is presented in [24]. Three different
slits are integrated onto the oval-shaped radiating elements
to obtain operation in the four mobile communication bands.
Besides the concept of slots and slits, the combined use of
the electric and magnetic monopole antenna is also used to
enable multiband operation for mobile phone application.
This technique is proposed in [25], where a new magneto-
electric monopole antenna is designed using two horizontal
4 VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
slot monopoles and a vertical monopole with a shorting strip
to allow a dual-band and wideband operation. Besides that,
the use of simple parasitic elements for a printed antenna is
shown to be able to add to the number of resonance in [26].
This is done by using two simple metal stubs to separate three
resonances to result in operation across six different cellular
bands. Meanwhile, a method of hybrid loop on a printed PIFA
for mobile terminals is presented in [27] to operate in a multi-
band characteristic. Another concept of utilizing additional
strip with the radiating element multiband operation can be
achieved as shown in [28]. This additional strip is etched
from the no-ground area located near the radiating element.
Similarly, the multiband behavior of the antenna in [29] can
also be introduced by etching circular slots on the radiating
element itself. Another example of this method is by using
multiple slots and meandered strips as shown in [30]. Besides
that, several resonant branches were designed in [31] to
obtain multiband feature, whereas a coupling element is used
to miniaturize the size of the antenna. The designs and their
operating frequencies proposed in [22]-[31] are summarized
in Table II.
From this section, it is observed that to enable wideband
operation to fulfill cellular requirements, the useful tech-
niques includes etching slots on their radiating elements.
On the other hand, others have proposed the insertion of
parasitic elements to result in additional bands. However, a
drawback of this technique is the added design complexity
and larger sizes, which may cause interference with other
circuitry located on the mobile terminal chassis.
III. MIMO ANTENNAS FOR MOBILE TERMINALS
The deployment of MIMO systems ideally requires the em-
ployment of multiple antennas at both the base stations and
mobile terminals. This enables a linear increase in channel
capacity with the additional number of antennas. A recent re-
view on MIMO antennas for mobile terminals was presented
in [2], whereas the following section discusses more details
of the literature not covered in [2].
A. WIDEBAND MOBILE TERMINAL ANTENNAS
MIMO antennas are considered as an integral part of LTE
and LTE-A systems. Moreover, the trend of integrated mobile
terminals in recent years increases the number of operating
frequency bands to operate across different wireless appli-
cations. One way to reduce the number of antenna elements
and to provide coverage for different wireless standards is to
use wide band antennas. Different wide bands MIMO have
been published in last seven years. While a part of these
literature are found in [2], the remaining papers [32]-[47] will
be further discussed.
Wideband diversity antenna for mobile terminals are pre-
sented, operating from 1.8 to 4.6 GHz in [32], and from
1.85 to 2.31GHz in [32]. In [32] diversity is achieved by
utilizing two orthogonally oriented quasi-complementary an-
tennas formed using a combined electric dipole and a mag-
netic slot. They are located at the separate ends of a mobile
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 2: Summary of Multiband SISO Antennas [22]-[31].

Ref Antenna Design Bandwidth [GHz] Ref Antenna Design Bandwidth [GHz]

[22] 0.46 - 785, [27] 0.834 - 0.907,

880 - 980 1.985 - 2.845
and and
1.45 - 1.49 3.253 - 3.593

[23] 0.824 - 0.96 [28] 0.70 - 0.96,

and 1.75 - 2.3
1.710 - 2170 and
3.1 - 4.38

[24] 0.824 - 0. 96 [29] 2.08 - 2.17,

and 3.55 - 3.65,
1.71 - 1.99 4.88 - 4.94
and
5.68 - 5.76

[25] 0.704 - 1.04 [30] 0.45 - 0.474,

and 0.86 - 1.04,
1.56 - 2.46 1.7 - 2.43
and
2.5 - 2.71

[26] 0.89 - 0.96, [31] 0.87 - 0.96,

and 1.7 - 2.2
1.4 - 2.7 and
2.49 - 2.72

terminal chassis. Meanwhile, a concept of two symmetric
slot-monopole-hybrid elements was used to achieve diversity
performance of the antenna in [33]. A similar concept of
combining the electric dipole and a square magnetic slot a
special diversity is achieved with in the bandwidth between
2.0 and 5.6 GHz in [34]. Finally, an antenna with EM-
coupled feed is also used to obtain dual band operation [35].
Several decoupling method have also been introduced for
multiple wide band antennas in [36]-[47]. Neutralization line
technique is used to decouple these antennas in [36]-[40]. A
thin neutralization line is inserted between two antennas to
reduce mutual coupling. Another method for this purpose is
by inserting a parasitic element between the antenna elements
[41], [42]. Ground current modification is also an alterna-
VOLUME 4, 2016
tive way to decrease the mutual coupling between antenna
elements [43]-[45]. One of the most recent techniques is by
designing a metamaterial band decoupling structure to min-
imize mutual coupling. The decoupling component is made
of sub-wavelength metal-air layers, which can be treated as
a singular medium operating over a broad frequency band
[46]. Finally, another new method to decouple antennas is by
designing a balanced and unbalanced antenna to achieve high
level of isolations [47].
B. DUAL BAND AND MULTIBAND MOBILE TERMINAL
ANTENNAS
An effective method to minimize the number of antenna
elements in a mobile terminal is to design dual- or multiband
5

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 3: Summary of MIMO Wideband Antenna [32]-[47]

Ref Antenna Design Bandwidth [GHz] Ref Antenna Design Bandwidth [GHz]

[32] 1.8 - 4.6 [40] 1.7 - 2.87

[33] 1.85 - 2.31 [41] 1.79 - 3.77

[34] 2.0 - 5.6 [42] 1.7 - 3.5

[35] 1.7 - 2.75 [43] 1.62 - 2.92

[36] 0.74 - 0.96 [44] 1.71 - 2.72

[37] 0.87 - 2.82 [45] 1.68 - 2.74

[38] 3.4 - 3.6 [46] 4.55 - 4.75

[39] 1.67 - 2.76 [47] 2.2 - 2.6

6 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
antennas. This is for the same purpose, which is to provide
coverage of across different wireless applications. Investiga-
tions performed on dual- and multiband antennas includes
the mutual coupling reduction techniques proposed in [48]-
[54]. An impedance transformer based on a T-shaped slot is
used for to minimize mutual coupling between two different
antennas operating in dual and wideband modes in [48].
Besides that, defected ground structure is also an effective
way for the same purpose, as seen in the dual band MIMO
in [48]. Meanwhile, ground slots and the introduction of
parasitic elements can also enable mutual coupling reduction
in dual band antennas [49]-[51]. To ensure compactness
of a dual band MIMO antenna system, differently shaped
antenna elements are designed in [52], [53], resulting in the
antenna miniaturization of up to 50 %. The next method in
enabling size compactness is by using contour integration on
a compact two-element MIMO antenna in [54] for dual band
operation. Another method is by wrapping the antenna on a
spacer, which is located at the edge of the mobile chassis [55].
In this work, a single antenna element is fed by dual port to
enable dual band MIMO operation.
On the other hand, several multiband antennas have also
been proposed for the purpose of operating in more than two
cellular bands [56]-[60]. The first is a multiband diversity
antenna designed using branches of a folded monopole. At
least one of the branch is terminated with a rectangular patch,
resulting in a multiband behavior [56]. Besides diversity,
multiband antennas can also be designed by introducing
different branches in the radiating elements, or by using
stub elements, loop elements and meander lines [57]-[60]. In
decoupling multiband antennas, ground current modification
and decoupling slots are more often used [61], [62]. A
complete review on mutual coupling techniques for MIMO
antennas can also be found in [63].
In this section, several wide, dual- and multiband antennas
for MIMO have been studied, including methods to decouple
multiple radiating elements in these antennas. While most of
the reviewed techniques have been proven effective, it can
be concluded that the use of neutralization line can only be
effective for single band antennas. However, an introduction
of t-shaped transformer on the ground plane is useful for to
decouple both single and dual band antennas. For the case of
multiband antenna ground perturbation and decoupling slot
lines are useful to reduce the mutual coupling.
IV. METAL-RIMMED MOBILE TERMINAL ANTENNAS
In recent years, the antenna and mobile phone industry are
increasingly designing smartphones with metal-rims and big
display screens. Such metal-rimmed smartphones offers ex-
cellent mechanical robustness and good aesthetic appearance.
Despite that, the existence of such structures poses a signif-
icant challenges for engineers due its undesirable impact on
antenna performance, especially when the continuous metal-
rim type is selected for use. Due to the significance of this
issue, research on metal-rimmed antenna is increasingly in-
vestigated and reported, which may warrant a separate review
VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
on this topic. In this survey, the main focus on the effects
of the user’s hand is maintained, and relevant literature on
this aspect for metal-rimmed antennas is discussed in the
following paragraphs.
Recent studies the user’s effect on the performance of
metal-rimmed mobile terminal antenna and have been pre-
sented in [64]-[71]. Thus far, there very limited number
of studies on the effects of the users for metal-rimmed
mobile terminal antennas, and existing studies are limited
to designs operating in the cellular band below 3 GHz,
specifically, for the GSM 850/900, DCS, PCS, UMTS 2100,
LTE 2300/2500 bands. In [64], [65] a multiband, dual loop
antenna surrounded by a continuous metal rim is presented.
Three different user’s hand grips were used in this study, and
it was found that efficiency of the proposed antenna degraded
from 75% to between 23% and 32% due to the proximity of
the users’ hand. On the other hand, the work in [65] further
considered an additional scenario where the user held the
antenna with two hands. In case of two hands, efficiency
degraded approximately 15% and 25%. Next, the researchers
in [66] presented a parallel dual-loop antenna embedded on
the system’s ground. A L-shaped coupled feed line with
addition matching networks is used to excite a multiple loop
resonant modes. A maximum efficiency reduction of below
20% and 40% is observed for LB and UB in three user’s hand
grips on the proposed antenna. Meanwhile, a narrow frame
antenna with hybrid multimode for a metal rimmed smart
phone is presented in [67]. A three grounded patches with
a small gap were used to connect the metal to the system’s
ground plane. A SAR values were analyzed for the proposed
antenna using the human head model instead of hands. It is
observed that the SAR values for the proposed antenna are
well below the regulated limits and efficiency values are also
within the limits for practical use of the antenna. In [68], an
inverted-F mobile terminal antenna surrounded by a metal
rim with two slots is proposed. To obtain a high radiation
performance, the metal rim is cut into three parts, resulting
in the proposed antenna’s operation in a multiband mode.
However, due to direct influence of the user’s hand, its effi-
ciency degraded to about 40 % in all operating bands. Finally
a reconfigurable multiband antenna is proposed in a narrow-
framed, metal-rimmed smartphone in [69]. In this design,
reconfiguration is achieved by using a PIN diode to enable
the antenna operation in two states. Multiple user’s scenarios
are also used to investigate its effects on antenna perfor-
mance, and it is noticed that antenna efficiency degraded to
an average of 30% when held in the user’s hand scenario.
Despite that, the SAR values maintained below the allowable
limits, indicating that the antenna can be used in practice.
Besides this, metal frames can also be designed to act as
the antenna radiating element. This is done in [70], where
full LTE band operation is achieved by integrating inductors,
band pass filter and an additional matching network. The
antenna performance is then analyzed in talk mode, and this
operation resulted in the antenna efficiency degradation to
between 10.3% and 17.5% in the lower band (0.698 to 0.96)
7
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 4: Summary of MIMO Dual and Multiband Antenna. [48]-[62]

Ref Antenna Design Bandwidth [GHz] Ref Antenna Design Bandwidth [GHz]

[48] 2.4 - 2.48, [56] 0.746 - 0.798,

and and
3.4 - 3.6 1.71 - 2.7

[49] 2.35 - 2.6 [57] 1.924 - 2.2,

and 2.836 - 3.235
5.1 - 5.3 and
5.629 - 5.743

[50] 1.93 - 2.05 [58] 0.68 - 0.912,

and 1.92 - 2.203
3.99 - 4.23 and
2.38 - 3.603

[51] 2.6 - 2.8 [59] 0.88 - 0.96,

and 1.75 - 1.87,
3.4 - 3.6 2.3 - 2.4
and
2.4 - 2.5

[52] 0.58 - 0.92 [60] 0.826 - 1.005,

and 1.527 - 2.480,
2.955 - 3.13 3.436 - 3.690
and
5.34 - 5.725

[53] 0.803 - 0.823 [61] 0.824 - 0.960

and and
2.44 - 2.90 1.71 - 2.690

[54] 2.34 - 2.5 [62] 0.704 - 0.787,

and 0.88 - 0.96
5.18 - 5.5 and
1.8 - 2.69

[55] 2.4
and
5.0

8 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 5: Summary of Metal-Rimmed Antennas [64]-[71].

Ref Antenna Design Bandwidth [GHz] Ref Antenna Design Bandwidth [GHz]

[64] 0.824 - 0.96 [68] 0.824 - 0.96

and and
1.71 - 2.69 1.71 - 2.69

[65] 0.798 - 0.968 [69] 0.824 - 0.96

and and
1.44 - 2.95 1.71 - 2.69

[66] 0.824 - 0.96 [70] 0.698 - 0.96

and and
1.71 - 2.69 1.71 - 2.69

[67] 0.822 - 0.965 [71] 0.698 - 0.96

and and
1.59 - 2.91 1.71 - 2.69

and between 14 % and 30% for upper band (1.71 to 2.69.
Besides matching circuits, varactor diodes can also be used
to miniaturize the size of antenna as demonstrated in [71].
In this work, a reconfigurable open slot antenna is placed at
the bottom chassis of the mobile terminal and is integrated
with a metal rim. The structure’s performance is studied
using both user’s hand and a head phantom. A significant
total efficiency reduction is observed for the antenna, which
was 18% initially, degrading to between 6% and 10% (in the
lower band) from 0.698 to 0.96 GHz and between 6% and
18% (in the upper band) from 1.71 to 2.69 GHz when the
user’s finger touches the metal rim slot. However, SAR values
produced are still below the regulated limits.
From this section it is noticed that the use of metal-rimmed
antennas is a promising solution to achieve a multiband
operation. This can be simply implemented by introducing
several ground patches and gaps onto the outer metal rim.
Besides that, another method is to use the entire metal
rim as a radiating element, which further facilitates antenna
miniaturization. Another creative method to make the most
VOLUME 4, 2016
of metal rims is separate them into individual parts, such as
using slits to form multiple radiators (such as IFAs) to excite
multiple resonant modes. Besides that, the implementation of
coupling strips may also contribute to such aim. In summary,
researches thus far have been focused on the design of the
metal-rimmed antenna and in capitalizing such structure.
However, the main challenge remains for such structure,
which is to address the performance deterioration caused by
the user’s hand.
V. MOBILE TERMINAL ANTENNA FOR 5G APPLICATION
The rapid demand increase of mobile data presented by
the wide use of smartphones has been compounded by the
bandwidth spectrum shortage in recent years [72]. Wireless
devices can typically operate between 700 MHz and 2.6
GHz [73]. Due to this, the implementation of the 4G cellular
networks, and research and standardization activities on the
5G technologies are rapidly progressing both in academia
and industry [74]. Due to the near-standardization of 5G fre-
quencies, literature on terminal antenna designs are limited,
and available ones will be discussed in this section [75].
9

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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
TABLE 6: Summary for 5G Mobile Terminal Antennas [78]-[83].
Ref Antenna Design
[78]
[79]
[80]
The first is a substrate integrated magneto-electric dipole
antenna proposed in [75]. This design operated from 4.98 to
6.01 for 5G. A H-shaped ground plane is also introduced to
miniature the antenna in its operating band. Future MIMO
applications in mobile terminals are likely to favor wideband
antennas, as demonstrated in [76], [77]. In [76], a wideband
printed MIMO antenna operational from 3 to 9 GHz band
for 5G is proposed. Meanwhile, a circular polarized antenna
for 5G is presented in [77] with an operational bandwidth of
11.08 %, ranging from 3.58 to 4 GHz. Next, a dual polarized
hybrid eight-antenna array in the 2.6 GHz band (from 2.55 to
2.65) for 5G MIMO is proposed in [78], [79]. Meanwhile,
in [80], an eight element printed MIMO antenna system
for a 5G mobile terminal operating between 2.54 and 2.62
GHz was investigated. The compatibility between 4G and 5G
hardware is the main feature of the antenna module presented
in [81]. On the other hand, a compact four port MIMO mobile
terminal antenna is investigated in [82] for 5G applications.
The most recent work proposed a tri-polarized 12-element
MIMO array antenna for 5G application, where the antenna is
operational from 3.4 to 3.6 GHz [83]. Finally, the researchers
in [81]-[83] have indicated that the operational bandwidth for
5G is expected to be between 3.4 and 3.6 GHz.
10
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
10.1109/ACCESS.2018.2883788, IEEE Access
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
Bandwidth [GHz] Ref Antenna Design Bandwidth [GHz]
2.55 - 2.65 [81] 0.824 - 0.96
1.71 - 2.69
and
3.4 - 3.6
2.55 - 2.65 [82] 3.4 - 3.6
1.8 - 1.92, [83] 3.4 - 3.6
2.3 - 2.4
and
2.54 - 2.62
VI. EFFECTS OF USER’S ON 5G MOBILE TERMINAL
ANTENNAS
The upcoming 5G technology launch by 2020 is also cap-
italized by developers for active array systems and chipset
manufacturers to meet the needs for complete deployment
[84]. The frequency band lower than 6 GHz known as sub-6
GHz 5G band for mobile terminal antenna is band of interest
for many countries. The antenna design for mobile terminals
must also account for the the effects of its housing and
location of the radiator for optimized MIMO performance
[85]. While the interaction of terminal with the user’s hand
for GSM, LTE, WLAN, WiMAX and another cellular band
below 3.5 GHz mobile terminals have been investigated in
previous literature [86], there is a need for such research
in the 5G bands. Such research are still limited, especially
for antennas operating in the vicinity of the 5G sub-6GHz
band [87]-[91]. One of such research is a 12-port 5G massive
MIMO antenna array in the sub-6GHz for mobile terminals
investigated considering the user’s hand [87]. However, the
placement of the user’s hand seen in [87] is inaccurate, with
the hand inserted inside the chassis of the mobile terminal.
The hand is also placed opposite the antenna, at locations
where there is no direct interaction with the antenna’s radiat-
ing areas with more than 50 % of radiated power. Meanwhile,
a study in terms of specific absorption rate is presented
in[88], performed on a two-element quad band antenna array
placed on different locations on the chassis. It was concluded
VOLUME 4, 2016
 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/ACCESS.2018.2883788, IEEE Access
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
TABLE 7: Summary for 5G Mobile Terminal Antennas with User’s Influence [87]-[91].
Ref Antenna Design Bandwidth [GHz]
[87] 3.4 - 3.8
and
5.15 - 5.925
[88] 1.565 - 1.585,
2.4 - 2.484,
2.5 - 2.57,
2.62 - 2.69,
3.3 - 3.4
and
5.15 - 5.35
[89] 4.5 - 5.5
that the SAR values for the antenna located at the bottom of
the chassis resulted in lower SAR due to the larger distance
of this antenna from the human head. Next, a simple study
of the index finger’s influence on a PIFA antenna perfor-
mance is discussed in [89]. The effect of a user’s finger
is studied on sixteen finger locations, indicating radiation
losses and matching efficiency of up to -1.6 dB and -0.9
dB, respectively, in the 5 GHz band. Finally, the user’s effect
on a mobile terminal phased array was investigated in [90],
[91]. In [90] it was observed that the phased array antenna
designed on the metal frame used in data mode suffered
losses of up to 3 dB, a loss of at least 5 dB is observed at 3.5
GHz band when two hands is used to hold the terminal. The
users’ effects are also investigated further up the frequency,
from 5 to 6 GHz, and is presented in [91]. It is shown that the
operation of the antenna in the mobile terminal is degraded
in terms of gain by up to 2 dB and 7 dB, when used near a
user’s hand and head, respectively.
Due to the limited available research in the 5G sub 6 GHz
band, it is difficult to conclude accurately how the user’s
effects may affect antenna parameters. It is known that the
user’s hand affects the antenna performance differently when
the operated in the lower (below 3 GHz) and higher (above 3
GHz) frequencies. This is a cause for concern as the two main
frequencies expected to be used for the 5G sub-6 GHz band
VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
Ref Antenna Design Bandwidth [GHz]
[90] 3.4 - 3.6
[91] 5.0 - 6.0
are from 3.4 to 3.6 GHz (Band C) and from 5.15 to 5.925
GHz (LTE-U band 46).
VII. EFFECTS OF USER’S HAND ON 5G SUB-6GHZ
BANDS
The detailed review of mobile terminal antennas provided in
the previous sections indicated that there is a need for new
antennas operational within the sub-6 GHz 5G frequencies
besides supporting existing mobile frequencies. This work
is attempted to investigate the effects of the user’s hand in
two sub-6 GHz 5G bands. In this work, two antennas were
designed: the first is operational in LTE band 46 (AD1),
whereas the second, is a dual band antenna operating in the
Band C as its lower band (LB) and LTE Band 46 as its
upper band (UB) (denoted as AD2). They are both placed
orthogonally on a same mobile terminal chassis before their
performance are being investigated in free space and in
proximity of two hand phantoms.
A. WIDEBAND (AD1) AND DUAL BAND (AD2) MOBILE
TERMINAL ANTENNA
The detailed dimensions for AD1 and AD2 are illustrated
in Figure and Figure 2. The two identical PIFAs are placed
orthogonally relative to each other on the edges of a 1.575
thick Rogers RT/Duroid-5880 substrate with a surface area
11
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 2: Geometry of the proposed AD1, in (a) 3D view,

(b) side view, (c) bottom view and (d) prototype in free space FIGURE 4: Simulated and measured S-parameters in free
(FS), with casing (WC), and with hand (WH). space for AD1.

FIGURE 3: Geometry of proposed (AD2), in (a) 3D view, FIGURE 5: Simulated and measured S-parameters in free
(b) top view, (c) bottom view and (d) prototype in free space space for AD2.
(FS), with casing (WC), and with hand (WH).

parison between the measured and simulated S-parameters

of 55 × 110 mm2 for both AD1 and AD2, see Figure 3(a).
This is to maximize isolation between them. The dielectric
permittivity of the substrate is 2.2 and its loss tangent is
0.0009. The copper metallization on the substrate is 0.035
mm thick (i.e., yellow region in Figures 2 and 3) on its reverse
side. The copper plate forming the PIFA element is 0.291
mm thick. Each PIFA is fed using a 50 Ω coaxial probe, at
an optimized location. To fine tune impedance bandwidth,
a 3.5 mm folded extension is introduced at the top edge of
between two resonant lengths in the top plate (See Figure
3). Upon optimization of AD1 and AD2, both antennas are
fabricated as shown in Figures 2(d) (for AD1) and Figures
3(d) (for AD2). Different form AD1, a slot is etched onto
each antenna element for AD2 to facilitate the separation of
two bands. The simulated and measured S-parameters for the
proposed AD1 MIMO antenna presented in Figure 4 indicate
good agreements. The measured -6dB impedance bandwidth
is 2.68GHz (from 4 to 6.68GHz) for antenna element 1 (AE1)
and 2.82GHz (from 3.92 to 6.74GHz) for antenna element 2
(AE2). This fulfills the frequency band requirement for LTE
Band 46. The isolation within the 2.68 GHz and 2.82GHz
band is between 16dB and 21dB. On the other hand, a com-
12
for AD2 is shown in Figure 5.
Measurements indicate that AD2 operates from 3.35 to
3.59 GHz (in the C Band) and from 5.15 to 6.8 GHz which
fulfills the requirement of LTE Band 46 (with at least -6 dB
of reflection coefficient). Within the operational bands, the
isolation is -15 dB and -18 dB in the lower band and upper
band, respectively. In general, the simulated and measured
results agreed well, with minor discrepancies caused by
fabrication inaccuracies.
B. USER’S IMPACT ON THE PERFORMANCE OF AD1
Since AD1 operates well throughout the whole frequency
range for LTE Band 46, the effects of the user’s hands on the
antenna’s impedance matching (IM), mutual coupling (MC)
and radiation efficiency (RE) are evaluated next. This study
is setup by inserting the antenna prototype into a 1 mm thick
casing dimensioned at 112 × 57 × 7.7 mm3 ; see Figure
2(d)). This casing is made using poly (methyl methacrylate)
with a permittivity, εr of 2.8. For a realistic representation
of mobile terminal casings in practice, and to emulate the
worst-case interaction, the antenna is positioned within the
box to achieve a minimal separation of 1 mm to the user’s
VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
FIGURE 6: Simulation of AD1 with user’s (a) LH and RH,
and (b) index finger positions for analysis of user effect.
finger. Figure 6(a) illustrates simulation of AD1 with the
bounding box and the right hand (RH) and left hand (LH)
of a SHO3TO6-V3 phantom [12] (with εr = 22.5 between
3 and 6 GHz). Both user hands were accounted for during
performance evaluations due to the orthogonal placement of
the AEs on the PCB. When holding a mobile phone, the
closest user’s component to the AEs is the finger, thus it is
imperative that their interactions with the radiated waves be
characterized.
The index finger is positioned at five different locations
(P1, P2, P3, P4, and P5) in the vicinity of the AEs, as
illustrated in Figure 6(b), while keeping the hand position
constant. P1, P2 and P3 are located on top of the radiating
element, whereas P4 and P5 are positioned on the bottom
and side of the AEs, respectively. Since most of the antenna
radiation originates from the top surface, P1, P2 and P3 are
locations representing direct interference by the finger on the
AEs. On the contrary, P4 and P5 are finger locations which
are non-directly interfering with the antenna. Evaluation for
all locations in this study is performed using the center
frequency of AD1 at 5.41 GHz. This choice of the center
frequency is expected to be most representative of its perfor-
mance throughout the entire operating bandwidth. Moreover,
for evaluations with users using the phantom, evaluation at
the center frequency also ensure that the simulations are
performed using the most average properties of the user’s
hand throughout the operating frequency band, which is most
reasonable, especially when the operating bandwidth is large.
1) User’s Impact on IM and MC
First, the effects of the user’s right and left hands on the
IM, MC, and bandwidth with different index finger locations
near the AEs are simulated, see Figure 7. The following
observations can be made.
• The impedance matching of the AD1 in free space
(FS) is -11 dB at its center frequency. The position-
ing of hand as well as with casing (WC) decreases
VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
FIGURE 7: Effect on AD1 MIMO showing (a) effect of RH
on AE1 IM and MC, and (b) effect of LH on AE2 IM and
MC.
the IM of antenna by up to 4 dB due to the dielectric
properties of the casing and the user’s hand.
• In all studied cases, the index finger placed at posi-
tions P1, P2 and P3 affected all assessed parameters
more significantly compared to P4 and P5 due to
the fingers’ location in the way of the antenna’s
radiation. This is validated by observing the worst
IM, which is when the finger is placed on P3, i.e.
on top of the highly radiative area of the antenna.
In Figure 2 it can been seen that antenna patch
is directly fed by a coaxial probe. It is generally
observed that the current distribution changes as the
finger placement is varied on top of the antenna.
This then affected the antenna performance in terms
of IM, and is worst when the finger is placed at
the coaxial probe feed (P3), where the current is
highest. This is mainly due to the finger’s more sig-
nificant interaction with the antenna’s impedance at
this high current position (P3) as compared to the
other two (P1 and P2) positions.
• When the hand is closely located to the AEs, the
MC varied differently for both AEs for all selected
locations: up to 4 dB when the antenna is held in
the right hand and 6 dB in the left hand, see Figure
7.
• Impedance bandwidth variation is also observed.
This is due to the lengthening of the electrical
length of the chassis due to the dielectric loading
by the casing and phantom. Table VII summarizes
the impedance bandwidth for both AEs at different
locations.
2) User’s Impact on RE
Firstly, the RE of the antenna are simulated in FS and WC
(see Figure 8) to obtain their reference efficiency. These
values are then compared with the RE of the antennas evalu-
ated in the vicinity of the SHO3TO6-V3 left and right-hand
phantoms, as illustrated in Figure 8(a) and (b). These results
indicate that the hands decrease the RE at least by 2.5 dB, and
up to 4.8 dB in the worst case when held in the RH, relative to
the RE in free space. Similarly, the RE for the antenna when
13
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 8: Simulated Bandwidth for the Wideband MIMO Antenna (AD1) With Right and Left Hand
FS WC P1 P2 P3 P4 P5
Bandwidth
[GHz] [GHz] [GHz] [GHz] [GHz] [GHz] [GHz]
(3.92 (3.70 (3.82 (3.62 (3.62 (3.88 (3.76
RH AE1
-6.74) -6.515) -6.73) -6.40) -6.43) -6.89) -7.10)
(4.06 (3.79 (3.95 (3.90 (3.87 (3.88 (3.76
AE2
-6.70) -6.62) -6.73) -6.91) -6.60) -6.89) -7.10)
(3.92 (3.70 (3.92 (3.94 (3.95 (3.90 (3.88
AE1
-6.74) -6.515) -6.57) -6.46) -6.43) -6.49) -6.50)
LH
(4.06 (3.79 (3.67 (3.69 (3.69 (3.75 (3.85
AE2
-6.70) -6.62) -7.01) -7.15) -5.78) -6.96) -6.84)

shift of 220 MHz compared to in FS. For the case held

FIGURE 8: Effect on AD1 MIMO showing (a) effect of RH
on RE (left) AE1 and (right) AE2 and (b) effect of LH on RE
(left) AE1 and (right) AE2.
held in the LH is observed as 3 dB, and in worst case up to 5.8
dB. It is clear from the figures that P1, P2 and P3 significantly
affects the RE of both antenna elements, especially when
located at P2 and P3, when the index figure is placed exactly
or near to the feed point of the antenna. Besides the fact being
radiating through that particular region, the high radiation
quality factor of the AEs lead to relatively high resistive
losses compared to the structures with lower radiation.
C. EXPERIMENTAL ASSESSMENT OF AD1
The fabricated prototype of the proposed AD1 is presented
in Figure 2(d). Simulated and measured S-parameters, effi-
ciency and MIMO performance parameters (envelop corre-
lation coefficient (ECC) and multiplexing efficiency (ME))
when placed in the casing and held in user’s hands are shown
in Figures 9 and 10. These figures indicate good simulation-
measurement agreements. The measured reflection coeffi-
cient (S11) of AD1 with casing resulted in a 6dB bandwidth
of 2.81 GHz (from 3.70 to 6.515 GHz) with a downward
14
in user’s hand, the measured S-parameter shown in Figure
9(b) indicates a downwards shift of 300 MHz, with about
4 dB of additional impedance mismatch relative to FS. The
impedance mismatch and shift in resonant frequency is due
to dielectric loading of casing and user’s hand on the mobile
terminal. The original operating bands of AD1 are shaded
in Figure 9(a) and (b) to facilitate the identification of the
resonance shift.
The RE of the antennas is then measured with one antenna
excited while the other antenna ports are terminated using
a 50 Ω load before every evaluation. Figure 10(a) and (b)
shows the simulated and measured antenna RE in FS, WC,
and with user (WU), indicating a maximum RE of -0.8 dB
for both AEs in FS and WC. Meanwhile, up to 4.5 dB of
loss can be observed from Figure 10(b) when the antenna
is in close proximity to the user. Figure 10(c) shows the
measured ECC of all scenarios (FS, WC, and WU) and it
is seen that the ECC in all cases is not significantly affected.
Thus, this evaluation is extended to study ME, as illustrated
in Figure 10(d) in all scenarios. Notice that the losses in
terms of ME is around 5 dB in WU compared to the cases
in FS or WC. Figure 10(c) shows the presence of user hand
increases the ECC between ports. However, even when in
the presence of the user’s hand, the correlations between the
ports are still low, with a maximum measured ECC value of
0.12 at 4 GHz. This evaluation is then extended to study ME
in all scenarios, as illustrated in Figure 10(d). Notice that
the losses in terms of ME is quantified to be about 5 dB in
WU compared to the cases in FS or WC, which is caused
by the differences in radiation efficiencies in the presence
of user’s hand. The antenna with degraded total efficiency
caused reduction in the signal-to-noise (SNR) of the received
signals, and consequently affected ME, as this parameter is
dependent on both the ECC and total efficiency [92].
D. USER IMPACT ON THE PERFORMANCE OF AD2
Next, AD2 is assessed in the presence of the same hand
phantoms as was performed for AD1. Due to their same
sizes, AD2 can be fairly compared to AD1 in terms of IM,
MC, RE. The same casing is also used for AD2 as discussed
VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 10: Simulated and measured results for AD1 (a) RE in FS and WC, (b) RE in WU scenario, (c) ECC in all scenarios
(FS, WC and WU) and (d) ME in all scenarios (FS, WC and WU).

FIGURE 9: Simulated and measured S-parameters of AD1

(a) WC and (b) WU.

FIGURE 12: User effects on AD2: (a) effect of RH on AE1

IM and MC, (left) LB and (right) HB and (b) effect of LH on
AE2 IM and MC, (left) LB and (right) HB.

1) User’s Impact on IM and MC

FIGURE 11: Simulation setup of AD2 with user’s (a) LH and
RH and (b) index finger positions.
before, and its evaluation setup with SHO3TO6-V3 left and
right hands is illustrated in Figure 11(a). Similarly, five
different locations of index finger are chosen (P1, P2, P3, P4,
and P5) placed in the radiating elements, see Figure 11(b).
Evaluations are mainly performed at the center frequency of
the LB (3.58 GHz) and UB (5.885 GHz) for this dual band
MIMO antenna.
VOLUME 4, 2016
Due to the contribution of the slot in enabling the additional
resonance in AD2, it is important to account its behavior
when placed in vicinity of user. To ensure consistency, the
palm-to-handset gap is being kept constant as during evalua-
tions for AD2. Figure 12 illustrates the IM and MC for user’s
LH and RH at five different index finger locations (P1, P2,
P3, P4, and P5). The IM in free space for LB and UB is -
16 dB, whereas the IM variation in presence of user hand
for AE1 is observed up to 8 dB for LB and 4 dB in the
UB for the RH phantom. Similarly, for AE2, 4 dB variations
are observed in both LB and UB with user’s LH, see Figure
12(b). The MC is also evaluated for the dual band case in all
five locations. Due to the orthogonal placement of the two
AEs, the resulting MC caused by the LH and RH are slightly
different: with about 4 dB of variation when held using the
LH and 2 dB for RH, depending further on the location of
user’s finger around AEs. Upon evaluation of IM and MC,
the impedance bandwidth changes in both bands for all index
finger locations is studied, and is summarized in Table-VIII.
15

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

TABLE 9: Simulated Bandwidth for Dual Band MIMO Antenna (AD2) With Right and Left Hand
FS WC P1 P2 P3 P4 P5
Bandwidth Band
[GHz] [GHz] [GHz] [GHz] [GHz] [GHz] [GHz]
(3.38 (3.28 (3.25 (3.24 (3.16 (3.22 (3.23
LB
-3.78) -3.68) -3.67) -3.65) -3.74) -3.77) -3.72)
AE1
(5.15 (5.00 (5.02 (5.17 (5.05 (4.99 (5.26
RH HB
-6.62) -6.48) -6.57) -6.87) -6.98) -6.85) -6.93)
(3.39 (3.33 (3.36 (3.36 (3.35 (3.34 (3.33
LB
-3.78) -3.77) -3.80) -3.79) -3.77) -3.77) -3.77)
AE2
(5.05 (4.97 (4.95 (4.93 (4.95 (4.97 (5.03
HB
-6.62) -6.72) -6.69) -6.70) -6.68) -6.81) -6.65)
(3.38 (3.34 (3.26 (3.26 (3.25 (3.26 (3.27
LB
-3.78) -3.76) -3.63) -3.63) -3.63) -3.63) -3.62)
AE1
(5.15 (5.00 (4.91 (4.91 (4.89 (4.79 (4.79
HB
-6.62) -6.48) -6.44) -6.3) -6.32) -6.43) -6.45)
LH
(3.39 (3.33 (3.23 (3.23 (3.19 (3.17 (3.15
LB
-3.78) -3.77) -3.68) -3.67) -3.68) -3.63) -3.62)
AE2
(5.05 (4.97 (4.98 (4.97 (4.7 (4.94 (4.93
HB
-6.62) -6.72) -7.00) -7.15) -6.89) -6.86) -6.78)

2) User’s Impact on RE
Figure 13 illustrates the effects of the finger placement on
RE with casing, relative to its performance in FS. This is
followed by studying the effects of the user’s hand on RE.
As expected, lower values of RE are obtained when a hand is
present in the vicinity of AD2, especially when index finger
is positioned right on top of the antennas (P1, P2 and P3).
RE starts decreasing up to 4.6 dB and 5.2 in the LB and
UB, respectively, in the presence of the LH. Likewise, RE
in LB and UB varies between 4 dB and 6 dB, respectively
for RH. It is shown from Figure 12 that the most sensitive
part of proposed AD2 for RE is when finger is closest to the
most radiative part of the antenna (P2 and P3). Besides the
decrease in RE, these locations also resulted in degradation
of the impedance matching. It is concluded that the user’s
hand affected AD2 differently for the two AEs, where AE2
exhibit less loss in RE as compared to AE1.
FIGURE 13: User’s effect on AD2: (a) effect of RH on RE
(left) AE1 and (right) AE2 and (b) effect of LH on RE (left)
E. EXPERIMENTAL ASSESSMENT OF AD2
AE1 and (right) AE2.
The proposed AD2 is then fabricated and tested, as shown
in Figure 3(d), and its feeding points are kept same as AD1.
The measured S-parameters in both WC and WU scenarios
are shown in Figure 14, indicating a reasonable agreement then changes to be from 3.120 to 3.62 GHz (in the LB) and
with simulations. A downward shift of 100 MHz in the LB from 4.889 to 6.624 GHz (in the UB), with simultaneous
and 150 MHz in the UB are observed due to the dielectric impedance mismatch of 11 dB and 4 dB in the LB and UB,
loading of the casing. Meanwhile, 140 MHz of downwards respectively.
shifting in the LB and 220 MHz in the UB are observed for The antenna RE for AD2 is also measured in sequence for
AE1 in proximity of a user’s hand. The measured bandwidth all scenarios (FS, WC and WU), with the adjacent (not as-
with casing is from 3.44 to 3.99 GHz (in the LB) and from sessed) antenna terminated using a 50 Îl’ load. The measured
5.00 to 6.725 GHz (in the UB). On the other hand, the most RE in the C band and LTE Band 46 varied approximated
affected element for AD2 is AE1 due to the location of the between 4.6 dB and 5.2 dB in the LB and UB, respectively,
index finger placed directly over it. The antenna bandwidth see Figure 15. Measured ECC is then calculated to assess
16 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
FIGURE 14: Simulated and measured S-parameters of AD2:
(a) with casing and (b) with user’s hand.
the MIMO performance of AD2 using 3D field radiation
patterns, assuming a uniform 3D angular power spectrum.
The ECC in Figure 15(c) is less than 0.08 and 0.12 in the
LB and UB, respectively. This indicates satisfactory MIMO
performance for both operating bands of AD2. From Figure
15(d), it is also observed that the ME of AD2 is also affected
by the presence of the user’s hand, with losses of around 4
dB and 3.5 dB in the LB and UB, respectively, relative to
FS. This is intuitive as antenna with degraded efficiency will
reduce signal-to-noise ratio of the received signal
VIII. CONCLUSION
In this paper, single and MIMO terminal antennas for current
future 5G wireless communication bands considering user’s
interaction have been reviewed. It is shown that the evolution
of the spectrums of mobile technologies towards 5G have
mainly influenced antennas’ operational requirements, and
consequently their development trends, from functioning in
a single band to multiband, followed by MIMO to massive
MIMO. In this context, this review aims to provide, discuss
and summarize researches relevant to the design and alle-
viation of users’ effects on current mobile antennas in the
design process. Recent investigations on mobile terminal an-
tennas for different dual-, multi- and wideband characteristic
indicate that the implementation of different broadbanding
and multibanding techniques are needed to meet the more
challenging hardware requirements. It also imperative for
mobile terminal antennas to be built in their casings, and
designed to be more compact in size to meet the demands of
multi-wireless standards operation using a single device. The
findings of this review can be categorized into three major
aspects.
1) Literature Survey of MIMO Mobile Terminal Antennas:
Recent MIMO terminal antennas for different frequencies,
including different decoupling techniques are presented and
discussed. They include concept of neutralization line, use of
parasitic element, ground perturbations, decoupling slot line,
orthogonal polarization, a T-shaped transformer and the use
of metamaterials. The choice of the different techniques used
for MIMO antenna decoupling is mainly dependent on the
VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
operating frequency bands. It is also observed that a neutral-
izing line is useful when the antenna is operating in a single
band, whereas a T-shaped transformer and ground current
modification facilitate decoupling for dual band antennas.
Finally, multiband antennas are most effectively decoupled
using ground perturbation and decoupling slots lines. The
newly defined MIMO antenna multiplexing efficiency was
also described, showing the relation between the MIMO per-
formance parameters and efficiencies versus the correlation
of the antenna elements for 5G bands.
The future focus of mobile terminal antenna is to utilize
the 5G frequency bands along with other cellular bands. As
the 5G technology is maturing, research efforts are being
intensified in many countries. Due to this factor, several
different frequencies have been investigated for 5G mobile
terminal, which includes the band from 2.5 to 2.65 GHz, from
3.58 to 4 GHz and from 4.98 to 6.01 GHz. Nonetheless, it is
noticed that most researched band below 6 GHz is focused in
the bands between 3.4 and 3.6 GHz, and from 5.15 to 5.925
GHz. This indicates that the operational bandwidth for 5G is
expected to be either one of these two bands, and these bands
can be categorized as the sub-6 GHz 5G bands.
2) Analysis of User’s Hand on 5G Mobile Terminal Antennas:
However, antennas for terminals must also account for the
effects of users’ proximity during operation. Due to the
different characteristic of user’s skin properties at frequen-
cies above 3 GHz and the limited available research on
this aspect, it is difficult to accurately predict the effects of
user’s hand on antenna parameters by examining previous
research on lower frequency bands. To fulfill this research
gap, several antennas have been designed and investigated
with the influence of users for wide and dual band sub 6GHz
frequencies in this work. The two antennas, AD1 and AD2,
are sized similarly, fed using the same connectors and built
using the same materials. AD1 is designed for wideband
operation in LTE Band 46, whereas AD2 is designed to
operate in dual band mode in the C band and LTE Band
46. Evaluations in proximity of user’s hand are performed
using the right and left-hand phantoms, both in simulations
and measurements. This is to examine the changes in terms
of IM, MC and impedance bandwidth, RE, ECC and ME.
From the simulations and measurements, it is observed that
the presence of the hand affected the matching more sig-
nificantly in AD2 compared to AD1, especially in the LB.
This is due to strong current distribution around the slot in
AD2, which is critical in splitting the wideband operation
into two to enable the dual band operation of AD2. The
change in terms of impedance matching when the user’s
finger is placed on a high current area near to the slot
indicates its sensitivity as the LB operation requires a longer
electrical length. Nonetheless, the detuning of the impedance
bandwidth for both antennas indicated consistency as they
are both generally shifted downwards due to the dielectric
loading of the casing and hand. Moreover, variation of MC
between the two antenna elements is constant, besides the
17
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

FIGURE 15: Simulated and measured results for AD2 (a) RE in FS and WC, (b) RE in WU scenario, (c) ECC in all scenarios
(FS, WC and WU) and (d) ME in all scenarios (FS, WC and WU).

consistent decrease of the RE. Placement of the user’s hand ACKNOWLEDGMENT

directly on top of the antenna element (at locations P1, P2
and P3) influenced their performance significantly. On the
contrary, locating the hand at a distance from the antenna
elements can in some cases even improve the MC of the
antennas. This is due to the radiated power being coupled
to the user’s index finger, which means that the smaller
E-field strength results in less MC in the presence of the
hand in comparison to the free space. Besides that, other
antenna parameter remains unaffected. The finding from this
investigation provides a good understanding on the effects of
user’s hands in designing future 5G mobile terminals in the
sub-6 GHz band.
3) Future Strategies to Improve MIMO Mobile Terminal
Antennas:
Several practical issues on how the human body impacts the
antenna operation and efficiency were studied using single
and multi-element antennas, besides reviewing the related
recent and ongoing researches. Due to its small size require-
ments, the need of the integration of multiband and multi-
wireless standard using a single hardware and mitigating
the human body effects, antenna design for mobile phone is
always the art of compromising between the size, the phone
appearance, and the performance.
Smartphones with full metal casing is also gaining signif-
icant interest in recent years in the mobile phone industry.
This is due to their improved mechanical strength, attractive
appearance, and better thermal conductivity. A number of
such designs have been proposed, especially for frequencies
below 3 GHz, but there is no study reporting the reduction
in performance due to the effects of user’s thus far. This
is possibly another research aspect which is of interest for
future investigations, especially methods in overcoming per-
formance variations and the effects of users due to the use of
continuous metal rims in smartphone antennas. Upcomming
researches are foreseen to be focused on the optimization
of the performance-enhanced multiband and reconfigurable
multi-antenna and massive MIMO systems for future wire-
less generations, with the aim of reducing the impact of the
human body.
This research was supported financially by the Ministry of
Science, Technology and Innovation under eScience fund
(Grant no: 01-01-015-SF0258)
REFERENCES
[1] S. Zhang, K. Zhao, Z. Ying, and S. He. “Body-effect-adaptive Compact
Wideband LTE MIMO Antenna Array with Quad Elements for Mobile
Terminals," PIERS Proceedings, Stock. Sweden, vol. 1, pp. 1858–1861,
2013.
[2] S. Zhang and Z. Ying, “MIMO Antennas for Mobile Terminals," FER-
MAT, 2016.
[3] Z. Ying, “Antennas in cellular phones for mobile communications," in
Proceedings of the IEEE, 2012, vol. 100, no. 7, pp. 2286–2296.
[4] J. Hirokawa and M. Zhang, “Handbook of Antenna Technologies", 2015.
[5] R. Valkonen, J. Ilvonen, C. Icheln, and P. Vainikainen, “Inherently non-
resonant multi-band mobile terminal antenna," Electron. Lett., vol. 49, no.
1, pp. 11–12, 2013.
[6] A. Cihangir, F. Ferrero, G. Jacquemod, P. Brachat, and C. Luxey, “Inte-
gration of Resonant and Non-Resonant Antennas for Coverage of 4G LTE
Bands in Handheld Terminals," Forum Electromagn. Res. Methods Appl.
Technol., vol. 3, no. 5, pp. 1–12, 2014.
[7] WRC-15 Press Release, “World Radio Communication
Conference Allocates Spectrum for Future Innovation," Available:
http://www.itu.int/net/pressoffice/press_releases/2015/56.aspx.
[8] M. Eroglu S., Toprak S., Urgan O, MD, Ozge E. Onur, MD, Arzu
Denizbasi, MD, Haldun Akoglu, MD, Cigdem Ozpolat, MD, Ebru Akoglu,
“5G Spectrum Recommendations," Saudi Med J, vol. 33, pp. 3–8, 2015.
[9] MT-2020 (5G) Promotion Group, “White Paper on 5G Concept,"
http://www.imt-020.org.cn/zh/documents/download/4, 2015.
[10] http://www.atimes.com/article/china-reserves-spectrum-5g-services/
[11] Qualcomm, “Making the Best Use of Licensed and Unlicensed Spectrum,"
https://www.qualcomm.com/media/documents/files/making-the-best-
use-of-unlicensed-spectrum-presentation.pdf, 2015.
[12] SK Telecom, “SK Telecom 5G White Paper,"
http://www.sktelecom.com/img/pds/press/SKT-
5G%20White%20Paper-V1.0-Eng.pdf, 2015.
[13] S. and P. E. A. 2010-2016, “Speag,"
https://www.speag.com/products/em phantom/hands/3to6ghz-hands-
2/, 2016.
[14] M. Kim, W. Lee, and Y. J. Yoon, “Wideband antenna for mobile terminals
using a coupled feeding structure," 2011 IEEE Int. Symp. Antennas
Propag., pp. 1910–1913, 2011.
[15] Z. Linl, H. Liu, and C. Liu, “Design of Sleeve Broadband Antenna for
Mobile Terminals," in Progress In Electromagnetic Research Symposium
(PIERS)., 2016, no. 88, pp. 8–11.
[16] R. Rahman, K. M. Morshed, S. Sabrin, and M. Rahman, “Wideband planar
monopole antenna for LTE, GSM, Bluetooth, WiMAX, DCS, PCS, and
GPS mobile terminals," 2nd Int. Conf. Electr. Inf. Commun. Technol.
EICT 2015, pp. 309–313, 2016.
[17] R. Gomez-Villanueva, H. Jardon-Aguilar, and R. Linares-Y-miranda,
“Broadband PIFA antenna for mobile communications terminals," 2014

18 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 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/ACCESS.2018.2883788, IEEE Access
Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS
11th Int. Conf. Electr. Eng. Comput. Sci. Autom. Control. CCE 2014,
2014.
[18] J. Holopainen, J. Ilvonen, R. Valkonen, A. A. H. Azremi, and P.
Vainikainen, “Study on the minimum required size of the low-band cel-
lular antenna in variable-sized mobile terminals," in Proceedings of 6th
European Conference on Antennas and Propagation, EuCAP 2012, 2012,
pp. 2754âĂŞ2758.
[19] M. R. I. and M. Ali, “Ground current modification of mobile terminal
antennas and its effects," IEEE Antennas Wirel. Propag. Lett., vol. 10, pp.
438–441, 2011.
[20] K. Wang, R. A. M. Mauermayer, and T. F. Eibert, “Contour-Integrated
Dual-Band Compact Antenna Elements and Arrays for Low-Profile Mo-
bile Terminals," IEEE Trans. Antennas Propag., vol. 63, no. 7, pp. 3305–
3311, 2015.
[21] J. Ahn, S. Kim, M. J. Lee, and Y. S. Kim, “A compact printed dual-band
WLAN antenna with a shorted coupling strip for mobile terminals," Asia-
Pacific Microw. Conf. Proceedings, APMC, pp. 313–315, 2012.
[22] J.-H. Kim, Y.-B. Chae, J.-H. Lim, and T.-Y. Yun, “Printed internal an-
tenna for mobile broadcasting (DVB-H/T-DMB) and communications
(GSM900)," IET Microwaves, Antennas Propag., vol. 6, no. 6, p. 680,
2012.
[23] D. C. S. Mobile, “A Novel Compact Multiband Antenna Employing,"
IEEE Antennas Wirel. Propag. Lett., vol. 12, pp. 1688–1691, 2014.
[24] Y. Du and A. Zhao, “An internal quad-band printed monopole antenna
for oval-shaped mobile terminals," IEEE Trans. Magn., vol. 48, no. 2, pp.
683–686, 2012.
[25] J. Guo, L. Zhou, B. Sun, and Y. Zou, “Magneto-electric monopole antenna
for terminal multiband applications," Electron. Lett., vol. 48, no. 20, p.
1249, 2012.
[26] C. K. Lin, C. H. Lin, Y. C. Lin “Simple printed multiband antenna with
novel parasitic-element design for multistandard mobile phone applica-
tions’" IEEE Trans. Antennas Propag., vol. 61, no. 1, pp. 488–491, 2013.
[27] I. Ahmed, I. Shoaib, N. Shoaib, A. Rasheed, and S. Shoaib, “A Printed
Hybrid Loop Planar Inverted-F Antenna for Next Generation Handheld
Terminals," in European Conference on Antennas and Propagation (Eu-
CAP) A, 2013, pp. 2044–2047.
[28] T. N. Mai, A. C. Lepage, B. Huyart, and Y. Pinto, “Compact multi-band
planar monopole antenna for LTE terminals," 8th Eur. Conf. Antennas
Propagation, EuCAP 2014, no. EuCAP, pp. 3278–3280, 2014.
[29] M. Ur-rehman and S. Haxha, “Design of A Slotted-Patch Microstrip
Antenna for Mobile Terminals," in IEEE International Symposium on An-
tennas and Propagation & USNC/URSI National Radio Science Meeting,
2015, pp. 1716–1717.
[30] W. Chen and X. C. , Yuan Yao , Junsheng Yu , Xiaoming Liu, “Design
of a novel multiband antenna for mobile terminals," in In International
Wireless Symposium (IWS), 2015, pp. 1–4.
[31] Y. Guo, H. Liu, X. Dai, and Z. Lin, “Design of Multi-band Antenna for 4G
Mobile Terminals," in Progress In Electromagnetic Research Symposium
(PIERS), 2016, pp. 491–493.
[32] M. Sonkki, E. Antonino-Daviu, M. Ferrando-Bataller, and E. T. Salonen,
“Planar wideband polarization diversity antenna for mobile terminals,"
IEEE Antennas Wirel. Propag. Lett., vol. 10, no. c, pp. 939–942, 2011.
[33] G. Kang, Z. Du, and K. Gong, “Compact broadband printed slot-
monopole-hybrid diversity antenna for mobile terminals," IEEE Antennas
Wirel. Propag. Lett., vol. 10, pp. 159–162, 2011.
[34] M. Sonkki, E. Antonino-Daviu, M. Cabedo-Fabres, M. Ferrando-Bataller,
and E. T. Salonen, “Improved planar wideband antenna element and its
usage in a mobile MIMO system," IEEE Antennas Wirel. Propag. Lett.,
vol. 11, no. c, pp. 826–829, 2012.
[35] S. Sun, M. Cheng, S. Lu, and J. Lin, “Compact MIMO PIFA for LTE/
WWAN operation in the mobile application," Proc. 3rd Asia-Pacific Conf.
Antennas Propagation, APCAP 2014, no. 36, pp. 26–28, 2014.
[36] A. Cihangir, F. Ferrero, G. Jacquemod, P. Brachat, and C. Luxey, “Neu-
tralized coupling elements for MIMO operation in 4G mobile terminals,"
IEEE Antennas Wirel. Propag. Lett., vol. 13, pp. 141–144, 2014.
[37] W. Li, Z. Tu, and Q. Chu, “Design of Planar Wideband MIMO Antenna
for Mobile Phones’" Wirel. Symp. (IWS), 2015 IEEE Int., pp. 1–4, 2015.
[38] M. Abdullah, Y.-L. Ban, K. Kang, M.-Y. Li, and M. Amin, “Eight-
Element Antenna Array At 3.5 Ghz for Mimo Wireless Application," Prog.
Electromagn. Res. C, vol. 78, no. August, pp. 209–216, 2017.
[39] Y. Wang and Z. Du, “A Wideband Printed Dual-Antenna System With a
Novel Neutralization Line for Mobile Terminals," Antennas Wirel. Propag.
Lett., vol. 12, pp. 1428–1431, 2013.
VOLUME 4, 2016
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
[40] Y. Wang and Z. Du, “A wideband printed dual-antenna with three neutral-
ization lines for mobile terminals," IEEE Trans. Antennas Propag., vol. 62,
no. 3, pp. 1495–1500, 2014.
[41] A. Akdagli and A. Toktas, “Wideband MIMO antenna with enhanced
isolation for LTE, WiMAX and WLAN mobile handsets," Electron. Lett.,
vol. 50, no. 10, pp. 723–724, 2014.
[42] S. S. Alja’Afreh, Y. Huang, Q. Xu, L. Xing, and O. A. Saraereh,
“MIMO antenna system of a compact 4-element PILA for 4G handset
applications," 2016 Loughbrgh. Antennas Propag. Conf. LAPC 2016, no.
November, 2017.
[43] H. Huang and J. Wu, “Decoupled Dual-Antenna with Three Slots and a
Connecting Line for Mobile Terminals," IEEE Antennas Wirel. Propag.
Lett., vol. 14, no. c, pp. 1730–1733, 2015.
[44] Y. Wang and Z. Du, “A printed dual-antenna system operating in the
GSM1800/GSM1900/UMTS/ LTE2300/LTE2500/2.4-GHz WLAN bands
for mobile terminals," IEEE Antennas Wirel. Propag. Lett., vol. 13, pp.
233–236, 2014.
[45] Y. Wang and Z. Du, “A Wideband Printed Dual-Antenna with a Protruded
Ground for Mobile Terminals Yan," IEEE Trans. Antennas Propag., vol.
62, no. 3, pp. 1495–1500, 2014.
[46] S. Xu, M. Zhang, H. Wen, and J. Wang, “Deep-subwavelength Decoupling
for MIMO Antennas in Mobile Handsets with Singular Medium," Sci.
Rep., vol. 7, no. 1, p. 12162, 2017.
[47] E. Foroozanfard, E. De Carvalho, and G. F. Pedersen, “Design and Evalua-
tion of Full-Duplex Terminal Antennas in Realistic User Scenarios," IEEE
Antennas Wirel. Propag. Lett., vol. 16, no. c, pp. 1851–1854, 2017.
[48] S. Zhang, B. K. Lau, Y. Tan, Z. Ying, and S. He, “Mutual coupling
reduction of two PIFAs with a T-shape slot impedance transformer for
MIMO mobile terminals," IEEE Trans. Antennas Propag., vol. 60, no. 3,
pp. 1521–1531, 2012.
[49] J. N. Lee, K. C. Lee, N. H. Park, and J. K. Park, “Design of dual-band
MIMO antenna with high isolation for WLAN mobile terminal," ETRI J.,
vol. 35, no. 2, pp. 177–187, 2013.
[50] M. Ikram, R. Hussain, and M. S. Sharawi, “A 4G MIMO Antenna System
with Dual Function Ground Slots," in 5th Asia-Pacific Conference on
Antennas and Propagation (APCAP), 2016, vol. 3, pp. 199–200.
[51] G. Li, H. Zhai, Z. Ma, C. Liang, R. Yu, and S. Liu, “Isolation-improved
dual-band MIMO antenna array for LTE/WiMAX mobile terminals," IEEE
Antennas Wirel. Propag. Lett., vol. 13, no. c, pp. 1128–1131, 2014.
[52] M. A. Jan, D. N. Aloi, and M. S. Sharawi, “A 2×1 compact dual band
MIMO antenna system for wireless handheld terminals," RWW 2012 -
Proc. IEEE Radio Wirel. Symp. RWS 2012, pp. 23–26, 2012.
[53] M. S. Sharawi, A. B. Numan, M. U. Khan, and D. N. Aloi, “A dual-
element dual-band MIMO antenna system with enhanced isolation for
mobile terminals," IEEE Antennas Wirel. Propag. Lett., vol. 11, pp. 1006–
1009, 2012.
[54] K. Wang and T. F. Eibert, “Low profile dual band WLAN antenna array
for mobile terminals," IEEE Antennas Propag. Soc. AP-S Int. Symp., vol.
2015-Octob, pp. 378-379, 2015.
[55] C. Y. Chiu, S. Shen, and R. D. Murch, “Dual-band antenna pair for MIMO
WiFi compact mobile terminals," 2016 IEEE Antennas Propag. Soc. Int.
Symp. APSURSI 2016 - Proc., pp. 65–66, 2016.
[56] D. Wu, S. W. Cheung, T. I. Yuk, and X. L. Sun, “A MIMO antenna for
mobile applications," in International Workshop on Antenna Technology
(iWAT), 2013, no. 1, pp. 87–90.
[57] W. J. Krzysztofik, “Space Diversity Parameters of MIMO Systems Small
Antenna Array for Mobile Terminal," in 10th European Conference on
Antennas and Propagation (EuCAP), 2016, pp. 3–6.
[58] S. Shoaib, I. Shoaib, X. Chen, and C. G. Parini, “MIMO antennas for next
generation mobile terminals," 2016 10th Eur. Conf. Antennas Propagation,
EuCAP 2016, pp. 4–7, 2016.
[59] X. Zhao and J. Choi, “Multiband MIMO antenna for 4G mobile terminal,"
Asia-Pacific Microw. Conf. Proceedings, APMC, no. c, pp. 49–51, 2013.
[60] S. Shoaib, I. Shoaib, N. Shoaib, X. Chen, and C. G. Parini, “Design
and Performance Study of a Dual-Element Multiband Printed Monopole
Antenna Array for MIMO Terminals," IEEE Antennas Wirel. Propag.
Lett., vol. 13, no. c, pp. 329–332, 2014.
[61] J.-F. W. HuiFen Huang, “Decoupled Hepta-Band Antenna Artray with
Three Slots for WWAN/LTE Mobile Terminals," in Asia-Pacific Mi-
crowave Conference Proceedings, APMC, 2015, vol. 7, pp. 6–8.
[62] Q. Sun, B. Sun, L. Sun, W. Huang, and Q. Ren, “Broadband Two-
Element Array with Hybrid Decoupling Structures for Multimode Mobile
Terminals," IEEE Antennas Wirel. Propag. Lett., vol. 14, no. c, pp. 1431–
1434, 2015.
19
 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/ACCESS.2018.2883788, IEEE Access

Author et al.: Preparation of Papers for IEEE TRANSACTIONS and JOURNALS

[63] X. Chen, S. Zhang, and Q. Li, “A Review of Mutual Coupling in MIMO [86] R. Khan, A. A. Al-Hadi, and P. J. Soh, “Recent Advancements in User
Systems," IEEE Access, vol. XX, 2017. Effect Mitigation for Mobile Terminal Antennas: A Review," IEEE Trans.
[64] Y. L. Ban, Y. F. Qiang, Z. Chen, K. Kang, and J. H. Guo, “A dual-loop Electromagn. Compat., pp. 1–9, 2018.
antenna design for hepta-band WWAN/LTE metal-rimmed smartphone [87] Y. Li, C.-Y.-D. Sim, Y. Luo, and G. Yang, “12-Port 5G Massive MIMO
applications," IEEE Trans. Antennas Propag., vol. 63, no. 1, pp. 48–58, Antenna Array in Sub-6GHz Mobile Handset for LTE Bands 42/43/46
2015. Applications," IEEE Access, vol. 3536, no. c, pp. 1–1, 2017.
[65] Y. Yan, Y.-L. Ban, G. Wu, and C.-Y.-D. Sim, “Dual-loop antenna with [88] H. S. Singh and M. K. Meshram, “Effect of User Proximity on Inter-
band-stop matching circuit for WWAN/LTE full metal-rimmed smart- nal Quad Band Mobile Phone MIMO/Diversity Antenna Performances,"
phone application," IET Microwaves, Antennas Propag., vol. 10, no. 15, Wirel. Pers. Commun., vol. 95, no. 2, pp. 1417–1431, 2017.
pp. 1715–1720, 2016. [89] C. M. N. C. Isa, A. A. Al-Hadi, S. N. Azemi, A. M. Ezanuddin, H. Lago,
[66] L. W. Zhang, Y. L. Ban, C. Y. D. Sim, J. Guo, and Z. F. Yu, “Parallel Dual- and M. F. Jamlos, “Effects of hand on the performance of 5 GHz two-
Loop Antenna for WWAN/LTE Metal-Rimmed Smartphone," IEEE Trans. port terminal antennas," 2016 IEEE Asia-Pacific Conf. Appl. Electromagn.
Antennas Propag., vol. 66, no. 3, pp. 1217–1226, 2018. APACE 2016, no. December, pp. 207–210, 2016.
[67] J. Lian et al., “Hybrid Multi-Mode Narrow-Frame Antenna for WWAN / [90] I. Syrytsin, S. Zhang, and G. F. Pedersen. “Performance Investigation of
LTE Metal-Rimmed Smartphone Applications," IEEE Access, vol. 4, pp. a Mobile Terminal Phased Array with User Effects at 3.5 GHz for LTE
3991–3998, 2016. Advanced," IEEE Antennas Wirel. Propag. Lett., vol. 16, no. c, pp. 226–
[68] Y. Liu, Y. M. Zhou, G. F. Liu, and S. X. Gong, “Heptaband Inverted-F 229, 2017.
Antenna for Metal-Rimmed Mobile Phone Applications," IEEE Antennas [91] C. Di Paola et al., “Investigation of User Effects on Mobile Phased
Wirel. Propag. Lett., vol. 15, no. c, pp. 996–999, 2016. Antenna Array from 5 to 6 GHz Investigation of User Effects on Mobile
[69] Y.-F. Qiang, G. Wu, Y.-L. Ban, H. Wang, and K.-L. Wong, “Reconfigurable Phased Antenna Array from 5 to 6 GHz," in 12th European Conference on
narrow-frame antenna for LTE/WWAN metal-rimmed smartphone appli- Antenna and Propagation (EuCAP), 2018, no. April.
cations," IET Microwaves, Antennas Propag., vol. 10, no. 10, pp. 1092– [92] R. Tian and B. K. Lau, “Multiplexing Efficiency of MIMO Antennas with
1100, 2016. User Effects," Int. Symp. Antennas Propag. Soc., pp. 1–2, 2012.
[70] H. Chen and A. Zhao, “LTE Antenna Design for Mobile Phone With Metal
Frame," IEEE Antennas Wirel. Propag. Lett., vol. 15, no. c, pp. 1462–
1465, 2016.
[71] M. Stanley, Y. Huang, H. Wang, H. Zhou, Z. Tian, and Q. Xu, “A Novel
Reconfigurable Metal Rim Integrated Open Slot Antenna for Octa-Band
Smartphone Applications," IEEE Trans. Antennas Propag., vol. 65, no. 7,
pp. 3352–3363, 2017.
[72] T. S. Rappaport, J. N. Murdock, and F. Gutierrez, “State of the art in 60-
GHz integrated circuits and systems for wireless communications," Proc.
IEEE, vol. 99, no. 8, pp. 1390–1436, 2011.
[73] T. S. Rappaport., S. Sun., R. Mayzus., H. Zhao., Y. Azar., K. Wang., G.
N. Wong., F. Schulz., K. Jocelyn., M. Samimi, and Gutierrez, “Millimeter
wave mobile communications for 5G cellular: It will work!," IEEE Access,
vol. 1, pp. 335–349, 2013.
[74] I. Chih-Lin, C. Rowell, S. Han, Z. Xu, G. Li, and Z. Pan, “Toward green
and soft: A 5G perspective," IEEE Commun. Mag., vol. 52, no. 2, pp. 66–
73, 2014.
[75] H. W. Lai and H. Wong, “Substrate integrated magneto-electric dipole
antenna for 5G Wi-Fi," IEEE Trans. Antennas Propag., vol. 63, no. 2, pp.
870–874, 2015.
[76] ul Haq, M. Aziz, M. Arif Khan, and M. R. Islam. “MIMO Antenna
Design for Future 5G Wireless Communication Systems," Softw. Eng.
Artif. Intell. Netw. Parallel/Distributed Comput., pp. 175–183, 2016.
[77] K. M. Mak, H. W. Lai, K. M. Luk, and C. H. Chan. “Circularly polarized
patch antenna for future 5G mobile phones," IEEE Access, vol. 2, pp.
1521–1529, 2014.
[78] M. Y. Li et al, “Eight-Port Orthogonally Dual-Polarized Antenna Array for
5G Smartphone Applications," IEEE Trans. Antennas Propag., vol. 64, no.
9, pp. 3820–3830, 2016.
[79] M. Y. Li, Y. L. Ban, Z. Q. Xu., and G. Wu. “Eight-port orthogonally dual-
polarised MIMO antennas using loop structures for 5G smartphone," IET
Microwaves, Antennas Propag., vol. 11, no. 12, pp. 1810–1816, 2017.
[80] Z. Qin, W. Geyi, M. Zhang. and J. Wang. “Printed eight-element MIMO
system for compact and thin 5G mobile handest," Electron. Lett., vol. 52,
no. 6, pp. 5–6, 2016.
[81] Y. L. Ban, C. Li, C. Y. D. Sim, G. Wu, and K.L. Wong. “4G/5G Mul-
tiple Antennas for Future Multi-Mode Smartphone Applications," IEEE
Access, vol. 4, no. c, pp. 2981–2988, 2016.
[82] M. Abdullah, Y. L. Ban, K. Kang, O. K. Kings Ford Sarkodie, and M. Y.
Li, “Compact 4-port MIMO antenna system for 5G mobile terminal," 2017
Int. Appl. Comput. Electromagn. Soc. Symp. - Italy, ACES 2017, vol. 1,
pp. 4–5, 2017.
[83] M. Y. Li, Y. L. Ban, Z. Q. Xu, J. Guo, and Z. F. Yu, “Tri-Polarized 12-
Antenna MIMO Array for Future 5G Smartphone Applications," IEEE
Access, vol. 6, pp. 6160–6170, 2018.
[84] Y. Qi, S. Member, G. Yang, L. Liu, and S. Member, “5G Over-the-Air
Measurement Challenges: Overview," Trans. Electromagn. Compat., vol.
59, no. 6, pp. 1661–1670, 2017
[85] P.-D. V. D. and L.-P. Raidel, “Design of a Dual-Band PIFA for Handset
Devices and it SAR Evaluation," Ing. Investig. y Tecnol., vol. 17, no. 2,
pp. 169–178, 2016.

20 VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/.
 A Review of Patch Antenna Design for 5G
Abirami M
Department of Electronics and Telecommunication Engineering
Karpagam College of Engineering
Coimbatore, India
abiramimanoharansona@gmail.com
Abstract— Due to rapid increase of mobile user’s, demands
occur for mobile communication. Mobile users need more
features on their mobile phones such as high date rate, efficient
communication, reduced traffic, comfort to use various
applications etc. Service providers are in need to satisfy the needs
of mobile users which can be done with help of 5G technology.
5G technology provides very high bandwidth, reduced latency
better Quality of Service, optimum capacity, wide band of
spectrum availability. 5G is operated at mm wave band in that
we can provide high frequency range with large amount of
bandwidth. This paper is based on the study of suitable antenna
design for 5G technology for the use of mobile communication.
Measurements of various parameters of an antenna design is also
studied which is to check whether the microstrip patch antenna
design is suitable for 5G mobile communication.
Keywords—5G technology, Mobile communication, Microstrip
patch antenna, Millimeter wave spectrum
I. INTRODUCTION
Due to shortage of global bandwidth wireless service
providers are in need to provide better performances for all of
their customers and also they have to introduce new features
for the requirements of mobile users. Since wireless
communication is developed upto 4G, in future to handle the
traffic rate, scarcity of bandwidth and quality of service 5G is
introduced. Design a suitable antenna for 5G technology is
very important while developing 5G mobile phones. Various
parameters of an antenna have to be measured to check the
suitability for 5G technology. Antenna is designed to operate
at mm wave spectrum.
CMOS technology is used with this mm wave antenna
design. Some other techniques like MIMO and Adaptive
beamforming are also used to improve the performances of
mobile phones form 4G to 5G [11].
In this paper, introduction of 5G and evaluation of 5G are
discussed in secion 1. In section 2 various 5G related
microstrip patch antenna design papers are discussed. In
section 3 review papers are comparitively analyzed. Section 4
is ended with conclusion.
II. LITERATURE REVIEW
A. Circularly polarized antenna for 5G
Circularly polarized patch antenna is designed for 5G
technology. Miniaturization of patch antenna and beamwidth
enhancement is mainly focus on this paper. 5G mobile phones
used for the application of satellite communication, cellular
networks and also used for safety communications. 5G
provides accurate global positioning, wide range of
bandwidth, good coverage and high quality of service. To
design a suitable antenna for 5G networks certain parameters
are to be considered such as operating frequency, antenna size,
polarization, manufacturing cost, , bandwidth. Mobile
communications requires that the radiation pattern of new
antenna design should able to cover complete azimuth angles
and maximum of elevation angles. Directional antennas are
preferred which have good beam tracking ability for satellite
communication. Circular shaped folded type antenna with 4
and 8 slots are introduced to reduce the size of an antenna. To
enhance the beamwidth of the patch antenna two techniques
are followed. One is dielectric substrate is surrounded by
patch antenna and another method is metallic block is added at
back side of the antenna. [6]
B. Directional Antenna with mm wave spectrum
Due to rapid use of mobile users, challenges occur for
service provider to increase the band of spectrum to avoid
shortage of bandwidth and also to provide efficient
communication. In this paper directional antenna is used with
the frequency of 2.8 GHz. For the need of less weight, small
size, simple design microstrip patch antenna is used. 5G
mobile phones offer effective communication, low latency,
and massive connectivity. [1]
C. Wideband antenna for 5G
Wideband antenna of rectangular shaped antenna with
microstrip line feed is used for the 5G technology. Operating
frequency of this antenna is at 6GHz. It results gain as 3.7dB
with directivity of 6.62 dB and 500MHz bandwidth. Antenna
parameters are measured to satisfy the needs of 5G technology
and also some parameters like atmospheric absorption of
waves due to rain fall or wind which may cause losses of
information. Far field radiation pattern is used for this antenna
design. [8]
D. Steerable directional antenna for 5G
In this paper steerable directional antennas are used in mm
wave mobile communication. Antenna is operated with the
frequency of 28 and 38GHz. Antenna design includes two
rectangular patch antennas with single element of RT/ Duroid
5880 substrate. Various parameters are measured to check microstrip loss. Coaxial feed is used with this design, inner
whether the antenna could able to operate with 5G technology connecter contacts with patch antenna and outer connecter
to satisfy the needs of mobile users and also service provider. contacts with dielectric constant. Resonance frequency of this
Some effective approaches are followed in this design such as, antenna is at 59.5GHz which produce return loss as -44.99dB.
designing an antenna which should operate with multiple [7]
resonances, Optimization of impedance matching; increase the
thickness of substrate and reducing effective permittivity of III. COMPARATIVE ANALYSIS OF DIFFERENT ANTENNA DESIGN
the substrate. Radiation losses can be reduced by designing FOR 5G
thin and high dielectric constant of the substrate. It provides Different types of antennas are designed for 5G mobile
gain 9.0dB and efficiency as 83%. [3] phones are analyzed based on type of antenna, techniques and
materials used for antenna design and their advantages are
E. CPW feed design for 5G cellular tabulated in table1.
In this paper antenna is designed with CPW feed which
can be suitable for future 5G technology. Operating TABLE I. COMPARATIVE ANALYSIS OF DIFFERENT ANTENNA DESIGN
FOR 5G
frequencies of an antenna are 3.73GHz, 5.56GHz and 8.4GHz
which is suitable for WLAN, WSN, Wi-Fi/Wi-Max and Type of antenna
Features/
HyperLAN. Microstrip patch antenna is preferred because of Paper used or materials Technology used
advantages
used
its cost, size, weight, flexibility etc. Fractal technology is used Rectangular
with these designs which provide good impedance matching Simultaneous use
shaped microstrip
of mobile phones,
and it could operate with multiband of frequencies [7] patch antenna, CMOS technology
Avoid shortage of
simultaneously [9]. FR4 substrate and
bandwidth
Coaxial feed
F. Array antenna for 5G mobile phones Wideband slotted
microstrip patch
In this paper antenna array is modified with sub array antenna, Roger
which is placed along the mobile phone to cover wide area. [4] RT/ Duroid 5880 MIMO technology To improve gain
This technique will avoid traffic rate. Linear phased array substrate and
antenna with omni directional radiation pattern is used with microstrip line
feed
this design. Antenna is operated with 18-28 GHz frequency. Steerable sub
To cover wide space in 5G mobile phones beam steering is array patch Coverage of
[2] MIMO
proposed. Three identical sub arrays are used which is placed antenna with space is improved
by the side of mobile phones where high gain achieved. identical shape
Microstrip patch
Coaxial or probe feed is used to design this antenna. [2] antenna, CPW
Better impedance
matching,
[9] feed is used with Fractal technology
G. Wideband slotted patch antenna for 5G Roger RT/ Duroid
multiple band
spectrum
Rectangular wideband slotted microstrip patch antenna is 5880 substrate
designed for 5G technology. It is operated with the frequency steerable Optimum
directional impedance
of 5GHz. MIMO technology is implemented to increase the [3] antennas, Roger
Uses mm wave
matching,
quality of service, gain. Antenna design includes RT5880 spectrum
RT/ Duroid 5880 Multiple
substrate with the thickness of 0.6mm and 2.2 as dielectric substrate resonance
constant. This antenna design is suited well for 5G cellular Environmental
Wideband Antenna is parameters are
mobile phones which provide reflection coefficient as -
antenna with designed with Far analyzed to
36.54dB and bandwidth as 300MHz [4]. [8]
rectangular field radiation improve the
shaped antenna pattern overall
H. Antenna solution for future 5G performances
For the application of video and multimedia, mobile users [1]
Directional
Millimeter Wave
Improve
require high quality and low latency transmission which Antenna bandwidth
creates new challenge for service provider to satisfy the need Miniaturization
of customers. In order to provide new features and also to [6]
Circularly Miniaturization of patch antenna
avoid shortage of spectrum 5G technology is introduced. Due polarized antenna technique is used and beamwidth
enhancement
to the scarcity of bandwidth mm wave spectrum is used for 5G
technology. Mm wave spectrum could support simultaneous
use of mobile phones. CMOS technology is integrated with
5G technology. Instead of increasing large bandwidth spatial IV. CONCLUSION
reuse technique is introduced in these designs which provide In this survey, designs of various patch antennas are
flexibility, reduced cost with effective communication. discussed for the application of new 5G mobile phones. By
Atmospheric absorption of radiating waves due to rain fall are using 5G mobile phones, can track earth quake and tsunami
measured. While design of an antenna with low loss tangent prior to mobile phones, can ON and OFF home appliances,
substrate will increase antenna efficiency and will reduce can fold mobile as per user requirement, easy to identify a
stolen mobile within a Nano sec, can access office desktop
from home. Among various techniques, millimeter antenna
with 28GHz frequency is suitable for 5G mobile phones, this
antenna design is tested in real time environment that produce
better signal outage reduced path loss and return loss. Antenna
array can be introduced to improve overall performance of the
antenna.

References

[1] Ankita P. Manekar, Dr. S. W. Varade. IJARCCE ISSN (Online)

2278-1021 ISSN (Print) 2319 5940 (2016, June). “Design and
Simulation of Directional Antenna for Millimeter Wave Mobile
Communication”, International Journal of Advanced Research in
Computer and Communication Engineering Vol. 5, Issue 6
[2] Atima Agarwal, Sweta Agarwal, September-October 2016),
“Simulation and Analysis of 5G Mobile Phones Antenna”,
(International Journal of Electronics and Communication
Engineering and Technology (IJECET) Volume 7, Issue 5, pp. 07–
12, Article ID: IJECET_07_05_002, ISSN Print: 0976-6464 and
ISSN Online: 0976-6472
[3] Chauhan, Brajlata, Sandip Vijay, and S. C. Gupta. “Millimeter-
Wave Mobile Communications Microstrip Antenna for 5G-A
Future Antenna." International Journal of Computer
Applications 99.19 (2014): 15-18.
[4] Chong Ming Sam, Mastanesh Mokayef, (2016, July). “A
Wideband Slotted Microstrip Patch Antenna for Future 5G”, EPH-
International Journal Of Science And Engineering, ISSN:2454-
2016, Vol.2,Issue:7
[5] Gupta, Akhil, and Rakesh Kumar Jha. "A survey of 5G network:
Architecture and emerging technologies." IEEE access 3 (2015):
1206-1232.
[6] Mak, Ka Ming, et al. "Circularly polarized patch antenna for future
5G mobile phones” IEEE Access 2 (2014): 1521-1529.
[7] Mamunur Rashid. MD., Saddam Hossain, (2014, November).
“Antenna Solution for Millimeter Wave Mobile Communication
(MWMC):5G”. International Journal of Scientific Research
Engineering & Technology (IJSRET), ISSN 2278 – 0882, Volume
3, Issue 8
[8] MHD Amen Summakieh, Mastaneh Mokayef , (2016). “Single
Wideband Microstrip Patch Antenna for 5G Wireless
Communication Applications”. In Journal of Electrical and
Electronics Engineering (IJRDO) Volume-1, Issue-4, Paper-2
[9] Mohan, Gaikwad Pooja, and Mrs SR Chougale. "CPW Feed
Microstrip Patch Antenna Design for Future 5G Communication."
[10] Outerelo, David Alvarez, et al. "Microstrip antenna for 5G
broadband communications: Overview of design issues." Antennas
and Propagation & USNC/URSI National Radio Science Meeting,
2015 IEEE International Symposium on. IEEE, 2015.
[11] Rappaport, Theodore S., et al. "Millimeter wave mobile
communications for 5G cellular: It will work!" IEEE access 1
(2013): 335-349.
 The 2015 International Workshop on Antenna Technology
Antennas and Propagation for Millimeter Wave
Radio Systems for 5G Mobile Networks
- the Tokyo Tech Wireless Fiber Project-
M. Ando, J. Hirokawa, M. Zhang, K. Toyosaki, M. Ali
Dept. of Electrical and Electronic Engineering
Tokyo Institute of Technology
Tokyo, Japan
mando@antenna.ee.titech.ac.jp
Abstract—The future mobile network called “5G” is
becoming a hot topic. The heterogeneous networks using multi-
band “HetNet” is one of the promising candidates which provide
more than 1000 times system capacity. Millimeter-wave is
expected to be utilized for the pico cell coverage as well as the
fronthaul to connect these enhanced RRHs. This paper reviews
the antennas and propagation research works developed in
“Tokyo Tech wireless fiber project”, where many of the project
outputs are directly applicable for realization of multi-band
HetNet utilizing millimeter wave. The 60GHz compact range
communication is focused upon.
Keywords—compact range communication; millimeter wave;
5G networks; modified edge representation (MER)
I. INTRODUCTION
The future mobile network called “5G” is becoming a hot
topic. The heterogeneous networks using multi-band “HetNet”
is one of the promising candidates which provide more than
1000 times system capacity [1][2]. Millimeter-wave is
expected to be utilized for the pico cell coverage as well as the
fronthaul to connect these enhanced RRHs. This paper reviews
the antennas and propagation research works developed in
“Tokyo Tech wireless fiber project” [3][4]. It started in 2007
and will continue until 2016. In the first part of the project (-
2012), we developed transceiver IC, small and high gain
antennas as well as the propagation analyses. Two main
objectives have been 40GHz band outdoor mid-rage (up to 1
km) communication equipment with 2Gbps (1Gbps x 2) [5]
and 60GHz short range (up to tens of meters) communication
equipment with 6Gbps [6]-[8], both with low power
consumption. In the second part of the project (2012-), in
addition to further enhancement of the device-level
performances, the application of these to mobile cloud
networks has been intensively considered. For example, the
gigabit access transponder equipment (GATE) has been
proposed as the stress-free millimeter wave access point [8]
while the link aggregation and/or the diversity is studied for
overcoming the degradation of reliability of millimeter wave
links due to rain[9]. Many of the project outputs already or
having been developed, are directly applicable for realization
of pico cell in HetNet with millimeter wave in 5G, whose
concept is now under standardization[10]. Here after
978-1-4799-7717-8 ©2015 IEEE
K. Sakaguchi
Dep. of Information and Communications Technology
Osaka University
Osaka, Japan
sakaguchi@comm.eng.osaka-u.ac.jp
summarizing the HetNet, the measured performance of 60GHz
compact range communication system together with the fast
algorithm for shadowing analysis by modified edge
representation (MER) is presented [11][12].
II. MULTI-BAND HETEROGENIOUS NETWORK FOR 5G
Fig.1 illustrates one of the candidates for the future mobile
network. We need to enhance the network system capacity by
1000 times in 5 to 10 years. Capacity increases with the
number of cells but the interferences between cells become
serious and prevent the increase. Most effective way out of
this difficulty is to introduce multi-band HetNet and to utilize
the millimeter-wave frequencies for the pico cell coverage.
Millimeter wave has many advantages such as, more than 100
times wider bandwidth, possible high isolation between cells
and associated flexibility in cell design.
Macro BS
Enhanced RRH
based 60G pico BS
Mm-wave fronthaul 60G pico BSs are
is introduced to connected to C-
reduce deployment RAN as RRHs to
cost integrate into
3GPP standards
Fig.1 Multi-band HetNet with millimeter wave pico cells connected to C-
RAN with millimeter wave fronthaul.
System rate (network capacity) was estimated for a model
presented in Fig.2, where the macro cell with the size 500m is
covered by 2GHz (10MHz BW and 46dBm) while the pico
cells are covered by 60GHz with the channel identical to
48
 The 2015 International Workshop on Antenna Technology

existing unlicensed one (2.16GHz BW and 10dBm). It target

target usage
usage image
image
demonstrated 1000 times increase of system rate would be Outdoor Mobile Network Backhaul/Fronthaul
realized by adopting 100 pico cells in 1 macro cell, which
satisfies the requirement in next 10 years.

5 years later
Many
small cell BSs 2+60GHz band Radio Entrance

1Gbps x 2 =2Gbps @ 40GHz

Macro䋺 Fixed Wireless Access

Center freq: 2GHz File Transfer᧤ 0ᨺ᧭0ᨩ᧥Gate/Mobile terminal Indoor

BW: 10MHz
Tx power: 46dBm
Macro BS 6Gbps(user rate) @ 60GHz
Small-cell䋺 Peer-to-peer
Kiosk download
Center freq: 60GHz
BW: 2.16GHz
Tx power: 10dBm Fig.3 Two target usage images in the project.
Small-cell BSs
Fig.4 summarizes the key activities in 60GHz indoor systems
in the second phase of the project. They include;
Fig.2 Simulation model for system rate increase with the Multi-band HetNet A. 60GHz Systems
consisting of 2GHz macro cell and 60 GHz pico cell.
• Transceiver IC consisting of RF-CMOS, (Highly
These system studies now suggest the advantage of Intelligent RF-CMOS and BB chip, 16QAM, 6Gbps
millimeter wave for constructing the pico cell in HetNet. The user rate, 0.5W, LDPC(1440,1344) )
above discussion has been motivated due to the spectrum • Gigabit access transponder equipment (GATE) system
congestion in mobile network designers but the feasibility has as the AP for the network. Compact range
never been confirmed from the view point of the hardware communication up to tens of meters. A smart phone size
technology by the AP, MTT and ED communities where the mobile terminal.
millimeter wave devices, antennas and propagation have been
studied in decades. Therefore, it is urgent for both the network • High efficiency modulation/demodulation (Signal
and hardware researchers to exchange their knowledge for processing for compensation of CMOS phase noise, IQ-
realizing the 5G network. This talk introduces the millimeter imbalance, multi-path effects aiming 10Gbps Ethernet
wave research activities conducted in the “wireless fiber option.)
Hz
project” of Tokyo Institute of Technology, which include the Gigabit Access Trans. Equip. 㧔GATE㧕 WiLAN
40
䋨వⴕធ⛯㹢ᄢኈ㊂䊂䊷䉺᝼ฃḰ஻䋩
G
10 0 0
5dBi
Antenna
1000
45dBi
Antenna
Hz
900 900

antennas and propagation study as well as transceiver IC for

38 dB
60G 800 34
800dB
Hz
60G 700 30 dB
700

millimeter wave as the promising elements supporting the 5G

26 dB
600

distance [mm]
distance [mm]

600
22 dB
500
18
500dB

networks. 14 dB
400
400
300
10 dB
6300
dB
200
60GHz 60GHz 2200
dB
ធ⛯䈏ᜬ⛯䈚䉇䈜䈇 100 -2 dB
100
-6 dB
-2 00 -10 0 0 1 00 2 00 -200 -100 0 100 200
h o r i z o n t a l p o s i ti o n [ m m ] horizontal position [mm]

III. TOKYO TECH WIRELESS-FIBER PROJECT Highly Intelligent CMOS 60GHz RF Frontend and BB Chip
䂾 㜞ᐲ䉟䊮䊁䊥䉳䉢䊮䉴ᕈ䉕஻䈋䈢㪍㪇㪞㪟㫑Ꮺ㪩㪝䊐䊨䊮䊃䉣䊮䊄䋯䊔䊷䉵䊋䊮䊄㪚㪤㪦㪪㓸Ⓧ࿁〝
Process ⅣႺ᷷ᐲ䊶ශട Noise 㪩㪝㩷䊐䊨䊮䊃䉣䊮䊄 BB
䊔䊷䉴䊋䊮䊄
䊒䊨䉶䉴 Temperature, voltage 㔀㖸ାภ
The 5 year (2007-2012) project “RF Coexisting
RF FE chip B/B chip
㔚࿶ᄌൻ
change
ᄌേ Interference
㊀⇥ 60GHz in
RX A out D out
ABB DBB

Technology on High Speed Baseband IC for Millimeter Wave

೙ᓮㇱ

CTL
೙ᓮㇱ

LO CTL

Adaptive caliblation, self-

⥄ᓞᕈ⢻⵬ఘᯏ⢻䇮䉶䊦䊐䊁䉴䊃 60GHz out
TX A in
ABB DBB
D in

ᯏ⢻╬䈮䉋䉎ή✢․ᕈ䈱ะ਄
Radio Systems” was supported by Government (MIC) as well test
ታⵝ

as Industry’s participation. The objective was to develop RF ៞タ

៤Ꮺ┵ᧃ ή✢䉝䉪䉶䉴䉭䊷䊃
coexisting technology on high speed baseband CMOS for P2P FDD 10Gbps x 2, Ethernet
Phone GATE

millimeter wave radio systems. Outdoor and indoor beyond optional

6Gbps, 0.5W, GATE
Gbps were designed, the former of which was installed in
Tokyo Tech campus. Based upon its success, the 4 year (2012-
2016) project is now underway which focuses upon the High efficiency modulation/demodulation 16QAM –CMOS Phase noise, IQ-imb., Propagation

application to the millimeter-wave gigabit network for mobile
cloud services. Fig.3 indicates two systems developed in the
project; one is the 40GHz outdoor system, while another is the
60GHz indoor system. Both are for P2P services and could be
applicable for the entrance, backhaul or fronthaul connecting
pico cells up to 1 km and for the short range communication
up to tens of meters.
Fig.4 Research topics in 60GHz short range communication in the latter half
of the project. 6Gbps user rate file transfer systems.
B. 40GHz Systems
Fig.5 summarizes the key activities in 40 GHz outdoor
systems in the second phase of the project.

978-1-4799-7717-8 ©2015 IEEE 49

 The 2015 International Workshop on Antenna Technology

• 40GHz, 64 QAM DDD System with 2Gbps (1Gbps x 2)

up to 1km. 15cm x 30cm. (12bit 800Msps 0.15W AD Large Aperture Array/Reflector
converter for cancellation dynamic range of 36dB. Compact range
service area
Paired high gain antennas with isolation more than
80dB. ) Mobile
terminal

• Proactive control of the network for small scale

diversity and link aggregation against the heavy rain.
• Network traffic control for big files transfer in GATE.
(Cooperation of WiFi and Millimeter wave)
Network
150 operation, Proactive routing against Rain 40GHz 64QAM DDD System
140 September 12, 2009

D
Estimation by Rx Level by 4 paths
130
120
Tipping-bucket Rain Gauge at W8 1Gbps x 2
110 Tipping-bucket Rain Gauge at I6
Rain rate (mm/h)

100
90
80
70
60
50
40
Different Rain Rates for Strong rain
Compact range
30 Similar Rain Rates for Weak rain
20

D = 30cm䇭 䇭 䌀60GHz 䇭 䇭 R1 = 1.5m䇭 䇭 R2 = 36m

10
0
19:00 19:30 20:00 20:30 21:00

D = 2m䇭 䇭 䌀10GHz 䇭 䇭 R1 = 10m䇭 䇭 R2 = 300m

Time

Fig.6 Concept of Compact range communication in 60 GHz as the access

Throughput reduction < 10% point to the network.
Network operation, Gate data handling in the network

1Gbps x2, 260MHz

Trunk
Medium
5 [UVGO QP% JKRHQT) ) DRU5 [UVGO

B. Bit Error Rate (BER) Measurement

Speed

6Gbps, Network Data < 1.5 times & WCNDKV# & % 

+O RTQXGFXGTUKQP
The feasibility of the concept is demonstrated by the
measurement. Large slot arrays with various sizes (2 to 4000
.QY ,KVVGT2..

elements) are fabricated and BER distribution is measured over

,
High
Speed AD Cov. 12bit 800Msps for DDD
System 1.6wÆ0.15w the compact ranges. Fig. 7 presents an example result for 32 x
GATE #n GATE #2 GATE #1 32 array. This realizes the error-free communication range up
to more than 10m in length and the width comparable to that of
Fig.5 Research topics in 40GHz medium range DDD communication in the
latter half of the project.
the antenna aperture. Multi-path free environment is confirmed
by comparing BER with and without absorber on the nearby
side walls. The BER degradation in quite near distance to the
IV. COMPACT RANGE COMMUNICATION FOR THE ACCESS POINT array aperture (<1m) is observed and is now under study in
TO THE NETWORKS. terms of inter symbol interference (ISI). 

A. Compact range communication x

A new concept named as “Compact range communication” y

has been proposed to extend and define the service area of Tx:
z

60GHz short range communication. If the antenna gain is high

various ANTs

Image of
enough and the aperture is large, we could clearly define and reception zone Rx:
open-ended WG

utilize the radiative near-field (Fresnel region; between the

reactive near-field and the far-field) as the user service area.
This is analogous to the compact range measurement where the
plane wave region is synthesized as is depicted in Fig.6. It
demonstrates the large slot array in corporate feed with 64 x 64
elements and 29cm x 29cm in size. Transfer jet type of file
transfer for the user in line of sight is assumed. Some
advantages in this concept are,
• Field amplitude is almost constant in the range and the
design of mobile receivers may be easy. Fig.7 Measured BER distribution for the compact range communication
for 32 x 32 array in 60 GHz as the access point to the network.
• Field outside of the compact range decays rapidly and
multi-paths from the environment are drastically C. Shadowing Fields Estimation Argorithm by Modified Edge
attenuated; quasi-multipath-free zone is realized over Representation (MER)
the service area. This, at the same time, realizes very
low interference between adjacent ranges or cells. Multi-paths or reflected waves are not reliable in millimeter
wave frequency and the direct ray contribution is quite
• Field strength is weak everywhere (even on the dominant in comparison to lower frequencies. Therefore, most
aperture) as it is inversely proportional to the aperture important difficulty in millimeter wave communication is the
area. This is good in terms of EM human exposure. shadowing (blocking) effects of the obstacles. In stead of

978-1-4799-7717-8 ©2015 IEEE 50

 The 2015 International Workshop on Antenna Technology

conventional statistical propagation model, deterministic V. CONCLUDING REMARKS.

algorithm, especially the high frequency diffraction techniques This paper reviews the millimeter wave technology
such as the ray tracing has been applied. Since the wavelength
developed in “wireless fiber project” 2007-2016 in Tokyo
is much smaller than the structure, computationally light
algorithm is highly recommended. Authors have been Institute of Technology. The outputs are promising candidates
developing a unique method MER based upon GTD equivalent for the millimeter wave pico cells and back- and front-haul
edge current method[11]. Only the periphery should be defined links in 5G multi-band heterogeneous networks. The 60GHz
and fields including diffraction in shadow region is accurately compact range communication using GATE is focused upon.
calculated. Fig.8 presents the definition of modified edge (τ). ACKNOWLEDGMENT
The simulation model as well as the field is presented in Fig.9
for 2GHz [12]. The fields are compared with the MoM results This work was conducted in part as “the Research and
(Wipl-D) and high accuracy is observed. In high frequency of Development for Expansion of Radio Wave Resources” under
60GHz, the computation time of MER would be about 1/200 of the contract of the Ministry of Internal Affairs and
MoM. Communications, Japan. Authors are indebted to project
source observer
members for providing great support in the research works and
n̂
rˆi rˆo many information for this talk.
tˆ tˆ rˆ n̂
τˆ o REFERENCES
n̂ rˆi n̂ rˆo rˆi
Jo ,M o rˆi τˆ
[1] K. Sakaguchi, S. Sampei, H. Shimodaira, R. Rezagah, G.K. Tran, K.
rˆo Araki, "Cloud Cooperated Heterogeneous Cellular Networks," in Proc.
IEEE ISPACS 2013, pp.787-791, Nov. 2013.
τˆ rˆi n̂ rˆo τˆ
J MER ,M MER [2] EU-Japan MiWEBA project, <www.miweba.eu>
[3] M Ando, “Tokyo Tech wireless-fiber project for millimeter wave
tˆ τˆ tˆ systems”, The International Workshop on Millimeter Wave Wireless
Fig.8 Modified Edge Representation in diffraction analysis Technology and Applications -Tokyo Institute of Tech., pp.8-13,
December 6, 2010.
[4] Makoto Ando, "Planar Waveguide Arrays for Millimeter Wave
Systems," IEICE Trans. Commun., Vol.E93-B, No.10, pp.2504-2513,
Oct. 2010. (invited).
[5] Miao Zhang, Jiro Hirokawa and Makoto Ando, "Double-Layer Plate-
Laminated Waveguide Slot Array Antennas for a 39 GHz Band Fixed
Wireless Access System," IEICE Trans. Commun., Vol.E97-B, No.1,
pp.122-128, Jan. 2014.
[6] M. Zhang, J. Hirokawa and M. Ando, "Receptionable Area Enlargement
in MMW Short Range Communication Using Waveguide Slot Antennas
with Large Number of Elements," IEEE AP-S International Symposium
(USNC/URSI National Radio Science Meeting), Session: 454.6,
Chicago, IL, USA, July 8-14, 2012.
[7] M. Ando, M. Zhang, J. Hirokawa, T. Taniguchi, "A 60GHz Gigabit
Access Transponder Equipment for Short Range and Short-time File
Transfer," Proc. of URSI Commission B International Symposium on
Electromagnetic Theory (EMTS), paper no.: 21PM1E-06, pp.215-217,
Hiroshima, Japan, May 20-24, 2013. (invited)
[8] M. Zhang, K. Toyosaki, J. Hirokawa and M. Ando,”Evaluation of a 60
GHz-Band Compact-range Gigabit Wireless Access System using a
Large Array Antenna” IEEE AP-S International Symposium, Session:
107.1, Memphis, TN, USA, July 2014.
[9] Hung V. LE, Hasan Md. MOHIBUL, Takuichi HIRANO, Toru
TANIGUCHI, Akira YAMAGUCHI, Jiro HIROKAWA, Makoto
ANDO,"Millimeter-Wave Propagation Characteristics and Localized
Rain Effects in a Small-Scale University Campus Network in Tokyo,"
IEICE Trans. Commun., Vol.E97-B, No.5, pp.1012-1021, May 2014.
[10] MiWEBA International Workshop on Millimeter-wave for 5G in
CEATEC2014 in cooperation with IEICE-CS, Oct. 8, 2014
<www.miweba.eu/?page_id=251>
[11] Tsutomu MURASAKI, Masahide SATO, Yoshio INASAWA and
Fig.9 Shadowing effects of two obstacles in 2 GHz. The fields by MER Makoto ANDO, "Equivalent Edge Currents for Modified Edge
and MoM is compared. Representation of Flat Plates: Fringe Wave Components," IEICE Trans.
Electronics, Vol.E76-C, No.9, pp.1412-1419, Sep. 1993.
D. Network off-lloading by burst-type big file transfer [12] M. Ali and M. Ando, "Fast estimation of Millimeter –wave Shadowing
for Channel Modeling in Compact Range Communication by
For full use of the high speed of 60GHz GATE, data Modified Edge Representation (MER) ", Tech. Committee Report
handling (contents delivery) in the network as well as the EMT-14-133, IEICE, Nov. 22, 2014.
cooperation with WiFi is important.

978-1-4799-7717-8 ©2015 IEEE 51

 PHYSICA L R EVIEW LET T ERS week ending
VOLUME 93, N UMBER 19 5 NOVEMBER 2004

Babinet Principle Applied to the Design of Metasurfaces and Metamaterials

F. Falcone,1,* T. Lopetegi,1 M. A. G. Laso,1 J. D. Baena,2 J. Bonache,3 M. Beruete,1
R. Marqués,2 F. Martı́n,3 and M. Sorolla1
1
Departamento de Ingenierı́a Eléctrica y Electrónica, Universidad Pública de Navarra, 31006-Pamplona, Spain
2
Departamento de Electrónica y Electromagnetismo, Universidad de Sevilla, 41012-Sevilla, Spain
3
Departament d’Enginyeria Electrónica, Universitat Autónoma de Barcelona, 08193-Barcelona, Spain
(Received 16 February 2004; published 1 November 2004)
The electromagnetic theory of diffraction and the Babinet principle are applied to the design of
artificial metasurfaces and metamaterials. A new particle, the complementary split rings resonator, is
proposed for the design of metasurfaces with high frequency selectivity and planar metamaterials with
a negative dielectric permittivity. Applications in the fields of frequency selective surfaces and
polarizers, as well as in microwave antennas and filter design, can be envisaged. The tunability of
all these devices by an applied dc voltage is also achievable if these particles are etched on the
appropriate substrate.

DOI: 10.1103/PhysRevLett.93.197401 PACS numbers: 78.20.Ci, 41.20.Jb, 42.25.Fx, 84.40.–x

Artificial metamaterials and metasurfaces with special According to the electromagnetic theory of diffraction
electromagnetic properties have been a subject of grow- [11], the field in the shadowed region (z > 0) is the field
ing interest in recent years [1,2]. Most proposed meta- scattered by the CSRR, E0c ; B0c . For z < 0, the total field is
materials make use of split ring resonators (SRRs) [3], or given by [11]
similar geometries, to achieve a negative effective per-
meability in a certain frequency range. The negative E c  E0c  E0;r 0
c  Ec ; Bc  B0c  B0;r 0
c  Bc ; (2)

permittivity has been commonly obtained from an array
of metallic wires or plates [2,4]. No particles acting as
point electric dipoles with negative polarizability have
been proposed to the date. In addition to these bulk
metamaterial designs, one- and two-dimensional planar
microwave circuits which show a left-handed behavior
have been recently proposed [5–7], some of them making
use of the SRR concept [7]. More recently, the applica-
tion of these concepts to the design of artificial surfaces
with special electromagnetic properties has been consid-
ered [8].
In this Letter we present a new approach for the design
of planar metamaterials and metasurfaces, which is based
on the Babinet principle. The key element of this new ap-
proach is the complementary split ring resonator (CSRR),
the complementary screen of the SRR (see Fig. 1). As a
first step in our analysis the behavior of a perfectly con-
ducting and infinitely thin SRR in an external electro-
magnetic field E0 ; B0 [see Fig. 2(a)] is considered. The
scattered field E0 ; B0 is approximately given by the field
produced by a resonant magnetic dipole [3]
m  0 !20 =!2  11 B0  z^ z^ ; (1)
where !0 is the frequency of resonance of the SRR and 0
is a geometrical factor. This approximation neglects
higher order multipolar fields [2,3]. It also neglects
cross-polarization effects [9,10] (these effects are consid-
ered later in this Letter). Let us now consider the behavior
of the CSRR when it is illuminated from z < 0 by an
external electromagnetic field E0c ; B0c [see Fig. 2(b)].
where E0;r 0;r
c ; Bc is the field that would be reflected by the
metallic screen without the CSRRs etched on it. The
scattered fields, E0c ; B0c and E0 ; B0 , must fulfill some sym-
metries that arise from the fact that they are produced by
currents which are confined in the z  0 plane: the com-
ponents B0z , E0x , and E0y must be even functions of z, while
E0z , B0x , and B0y must be odd functions of the same vari-
able [11].
According to the Babinet principle, if a screen with
apertures (the CSRR) is illuminated from z < 0 by an
incident field E0c ; B0c and its complementary screen (the
SRR) is illuminated by some complementary incident
fields E0  cB0c , B0  1=cE0c , then at the shadowed
side (z > 0) the total fields must satisfy [11]
E c  cB  E0c ; cBc  E  B0c : (3)
Assuming that the SRR scattered field is described by (1),
y
c
rext
z x
d
SRR
CSRR
FIG. 1. Geometries of the SRR and the CSRR.

197401-1 0031-9007=04=93(19)=197401(4)$22.50  2004 The American Physical Society 197401-1

 PHYSICA L R EVIEW LET T ERS
VOLUME 93, N UMBER 19
E ∼ E + E′ Ec ∼ Ec + Ec + Ec′ Ec ∼ E′c
0 0 0, r
B ∼ B + B′ + Bc′ Bc ∼ B′c
0 0 0, r
Bc ∼ Bc + Bc
p m
-p
m p
-m
z<0 z>0 z<0 z>0
(a) (b)
FIG. 2. Illustration of the behavior of a SRR (a) and a
CSRR (b) when they are illuminated by an external field
coming from z < 0. Big arrows account for the main excitation
mechanism [3] given by (1) and (5). Small arrows stands for the
cross-polarization effect [9,10].
it can be easily verified that in order to satisfy (3), the
fields scattered by the CSRR at z > 0, E0c ; B0c should be
those produced by an electric dipole p  1=cm, or
1 excited. This fact would prevent the excitation of a mul-
p  0 !20 =!2  11 E0c  z^ z^ :
c2
In the nonshadowed region (z < 0) the sign of this dipole
must change, in order to produce the aforementioned
symmetry properties of the scattered fields. Thus, for
z < 0 we finally obtain
p  0;c !20 =!2  11 E0c  z^ z^ ; 0;c  1=c2 0 ;
where c is the velocity of light in vacuum. For lossy and/
or thick screens, as well as in conventional circuit boards,
the previous equations should be considered only as an
approximation. Since E0c  E0;r 0 ^ ^
c  2Ec  z z at z  0 , it
may be convenient, for some applications, to substitute (5)
by
p  0 !20 =!2  11 Eext ;
where 0  0;c =2 and Eext  E0c  E0;r c is the total ex-
ternal field produced by the sources and the metallic
screen without the CSRR.
The aforementioned results can be directly applied to
the design of artificial metasurfaces. In fact, the results
for a single CSRR can be extended to a system of many
CSRRs with a density of N CSRRs per square meter.
Since the CSRRs are electrically small [3], the distance
between them can be made much smaller than the inci-
dent radiation wavelength. Thus we are in the long wave-
length limit, and the considered metallic surface can be
seen—from the source side — as an electric dipolar sheet
of magnitude Ps  12 Np on top of a flat metallic screen.
From the opposite side, a dipolar sheet of the same
intensity but of opposite sign is seen [see Fig. 2(b)]. Let
us consider the incidence of a plane wave on a CSRR
metasurface. Assuming that the angle of incidence and
the polarization of this wave allows for the CSRRs exci-
197401-2
week ending
5 NOVEMBER 2004
tation, the induced dipolar sheet also produces an elec-
tromagnetic wave which interferes with the plane wave
reflected at the metallic screen. The effect of this inter-
ference can be dramatic near the resonance !0 . In par-
ticular, the CSRRs screen could be potentially tailored in
order to destroy the reflected wave. At this frequency all
the electromagnetic power will be transmitted through
the screen. This behavior is just the dual of that expected
for a surface of N SRRs per square meter, illuminated by
the complementary wave. In this case, a magnetic dipolar
sheet is produced which, eventually, cancels the trans-
mitted wave, thus reflecting all the incident power.
Up to now, the cross-polarization effects in the SRR
[9,10] have been neglected. In the frame of such an
approximation, the aforementioned effects in a multiple
CSRR (SRR) screen would appear only if there is a
normal component of the electric (magnetic) incident
field, because only in this case the CSRRs (SRRs) are
(4) tiple SRR (CSRR) metasurface by a normally incident
plane wave, thus limiting its usefulness as frequency
selective surfaces. However, when cross-polarization ef-
fects are considered [9,10], it becomes apparent that a
similar excitation can be observed for incident fields with
a nonvanishing component of the electric field along the
y axis of the SRRs (see Fig. 1). From duality, a similar
behavior is expected for CSRRs illuminated by an exter-
(5) nal magnetic field polarized along the same axis.
Electromagnetic simulations (using the CST Microwave
Studio electromagnetic solver) have confirmed this hy-
pothesis. An experimental setup has been also designed
and built up to prove such effects. The experimental
device consists of a pair of CSRRs and SRRs metasurfa-
ces which were illuminated from one side by a normally
incident plane wave. The transmitted field was measured
(6) at the opposite side (two horn antennas were used for
these purposes). Both metasurfaces were etched on a
commercial low loss microwave substrate (Arlon 250
LX-0193-43-11, whose parameters are shown in the cap-
tion of Fig. 3). The measured transmission coefficients for
the appropriate polarization of the incident wave — mag-
netic (electric) field along the y axis for the CSRR (SRR)
metasurface — are shown in Fig. 3. A sharp transmission
peak can be observed for the CSRR metasurface at the
resonance, whereas a sharp null can be also observed in
the SRR metasurface at a similar frequency. These results
confirm the proposed theory (the frequency of resonance
of the SRRs predicted by the model reported in [12] is
!0 ’ 24:41 109 s1 ). The shift between the fre-
quency of resonance of the SRR and the CSRR metasur-
faces can be mainly attributed to the effect of the
dielectric substrate, which affects in a different way the
frequency of resonance of the SRRs and the CSRRs. The
limited height of the transmission peak for the CSRR
metasurface can be attributed to the presence of the
197401-2
 PHYSICA L R EVIEW LET T ERS
VOLUME 93, N UMBER 19
1 Thus, the effect of the CSRRs is to change the effective
0.8
0.6
|S21|
0.4 SRR
CSRR
0.2
0
3 3.5 4 4.5 5 5.5 6 6.5
Frequency (GHz)
FIG. 3. Measured transmission coefficient through a CSRR
(solid line) and a SRR (dashed line) frequency selective sur-
face. The CSRRs or SRRs are etched on a commercial low loss
microwave board with permittivity   2:430 , thickness h 
0:49 mm, and metallizations of copper with a thickness t 
35 m. The CSRR and SRR parameters are (see Fig. 1) rext 
3:5 mm, c  0:4 mm, and d  0:4 mm. The CSRRs and SRRs
are arranged in a square lattice with periodicity a  8 mm.
The characteristics of the incident waves are described in the
text.
substrate, the finite size of the metasurface, and Ohmic
losses (simulations in the absence of the substrate and
neglecting losses show a 100% of transmission). As ex-
pected, no effect was observed for the orthogonal polar-
ization, thus showing that a multiple CSRR or SRR
metasurface also acts as a polarizer. Subsequent simula-
tions of transmission through thick (up to 5 mm) per-
fectly conducting metallic screens have shown that this
thickness significantly affects the location of the trans-
mission maxima but does not destroy the reported quali-
tative effect. Similar simulations show that the location of
the maxima is not significantly affected by the angle of
incidence, nor by the CSRR periodicity. This indepen-
dence of the maxima location with the angle of incidence
and CSRR periodicity is in complete agreement with the
proposed theory: the resonance is mainly a property of
the individual CSRRs, not a characteristic of the array.
Another application of the reported concepts is in the
design of planar metamaterials. Most of present high
frequency passive circuits are designed in planar tech-
nology, so that the electromagnetic field is confined be-
tween a metallic ground plane and some metallic strips or
patches, with a dielectric substrate filling the interspac-
ing. Let us suppose that some CSRRs are etched on the
ground plane at regions where the electromagnetic field is
significant. The effect of these CSRRs is to produce an
additional polarization in this region. According to (6),
this polarization is simply proportional to the external
field, and highly negative (that is, antiparallel to this
field) around and above the frequency of resonance.
197401-3
week ending
5 NOVEMBER 2004
permittivity of the substrate from positive to negative
values in some frequency region near the frequency of
resonance of the CSRRs. Therefore, a substrate region of
effective negative permittivity can be created by this
procedure.
In order to show this effect by an experiment, the
planar structure schematized in Fig. 4 has been designed
and manufactured. This is a modification of the well-
known microstrip transmission line, incorporating some
CSRRs etched on the ground plane, underneath the upper
strip. Capacitive gaps can be also present on the strip.
According to the previous theory, a negative effective
dielectric permittivity will appear in the substrate region
of interest, near the CSRRs resonance. Thus, microwave
propagation will be inhibited in this frequency range. In
order to switch this stop band behavior to a bandpass one
(i.e., microwave propagation only at those frequencies
where eff < 0), we can resort to a strategy well known
by electrical engineers: by periodically loading the mi-
crostrip line with capacitors, the inductive series imped-
ance of the strip line is changed to a capacitive one, thus
allowing propagation on negative  substrates. This is just
the expected effect of the capacitive gaps etched on the
microstrip. It is also worth noting that the resulting trans-
mission line has a backward-wave behavior and can be
considered as one dimensional analogous of a left-handed
medium [13].
The practical device was manufactured on a com-
mercial low loss microwave dielectric board (Rogers
RO-3010), and the device parameters are shown in the
caption of Fig. 5. The strip line was chosen in order to
obtain the usual characteristic impedance of 50  for the
host line. Finally, the capacitive gaps —when present —
have been designed in order to introduce a per unit length
series capacitance which dominates, in the frequency
range of interest, over the line inductance. Figure 5 shows
the simulated and measured transmission coefficient for
the four-stage microstrip transmission line shown in
Fig. 4, with and without capacitive gaps etched on the
metallic strip. A good agreement can be found between
the experiment and the simulations, and discrepancies can
be attributed to Ohmic losses in the CSRRs (losses were
neglected in the simulations). A passband appears for the
FIG. 4. A microstrip line (in black) over a ground plane (in
gray) with some CSRRs etched on it. Capacitive gaps have been
etched on the strip in order to obtain a left-handed passband.
197401-3
 PHYSICA L R EVIEW LET T ERS
VOLUME 93, N UMBER 19
0 for making negative- metamaterials, based on the use of
-10
|S21| (dB)
-20
-30
-40
2.5 3 3.5 4 4.5
Frequency (GHz)
FIG. 5. Simulated results (using CST Microwave Studio) for
the transmission coefficient in the structure of Fig. 4 with
(dashed line) and without (dotted line) capacitive gaps. The
experimental results for the transmission coefficient with ca-
pacitive gaps are also shown (solid line). Substrate parameters
are   10:20 , h  1:27 mm, with a copper layer of 35 m
thick. The CSRR parameters are rext  2:5 mm and c  d 
3 mm. The strip width is w  1:2 mm.
line with CSRRs and capacitive gaps at the same frequen-
cies where a stop band can be observed for the line
without the gaps. We interpret these results as a conse-
quence of the aforementioned effective negative dielec-
tric permittivity of the substrate in such frequency range.
This interpretation is confirmed by the dispersion relation
of the line periodically loaded with CSRRs and capacitive
gaps, which has been computed from the transmission
coefficient of a single unit cell. The results of these
simulations (not explicitly shown due to the limited space
of this Letter) evidence a backward-wave behavior, in
agreement with the proposed theory. In contrast to SRR-
based structures, the CSRR structures have a connected
metallic ground plane which extend over all the structure.
This last characteristic makes easy the tuning of the
structure if dielectric substrates with a significant depen-
dence of its permittivity on an external electrostatic field
[14] are used. For instance, in the modified transmission
line shown in Fig. 4, it is enough to apply an external dc
voltage between the strip line and the ground to obtain
this tuning. Simulations have shown a high sensitivity of
the location of the transmission peak of Fig. 5 to varia-
tions of the substrate permittivity.
In summary, the Babinet principle has been applied to
the design of artificial surfaces and materials with ex-
traordinary electromagnetic properties. A new particle,
the CSRR, which acts as an electric point dipole with
negative polarizability, has been proposed following this
approach. This concept goes beyond the usual approach
197401-4
week ending
5 NOVEMBER 2004
metallic wires or plates. It has been shown that by etching
CSRRs in a common metallic plate, artificial metasurfa-
ces with special electromagnetic transmission properties
can be designed. Artificial planar metamaterials with
negative  can be also designed by a similar procedure.
By properly combining these designs with other elements
providing an effective negative magnetic permeability, a
left-handed behavior is achieved. Practical application of
these concepts in the design of frequency selective sur-
faces and polarizers, as well as antennas, filters, and other
microwave devices, is envisaged. These applications will
take advantage of the small electrical size of the CSRRs
(and SRRs) at resonance, thus resulting in a significant
miniaturization.
This work has been supported by the Spanish Ministry
of Science and Technology and FEDER funds by Project
Contracts No. TEC2004-04249-C02-01, No. TEC2004-
04249-C02-02, and No. TIC2002-04528-C02-01.
*Electronic address: falcone_fj@tsm.es
[1] L. Martin-Moreno, F. J. Garcia-Vidal, H. J. Lezec, K. M.
Pellerin, T. Thio, J. B. Pendry, and T.W. Ebbesen, Phys.
Rev. Lett. 86, 1114 (2001).
[2] D. R. Smith, W. J. Padilla, D. C. Vier, S. C. Nemat- Nasser,
and S. Schultz, Phys. Rev. Lett. 84, 4184 (2000).
[3] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J.
Stewart, IEEE Trans. Microwave Theory Tech. 47,
2075 (1999).
[4] R. Marqués, J. Martel, F. Mesa, and F. Medina,
Microwave Opt. Tech. Lett. 35, 405 (2002).
[5] G.V. Eleftheriades, A. K. Iyer, and P. C. Kremer, IEEE
Trans. Microwave Theory Tech. 50, 2702 (2002).
[6] A. Grbic and G.V. Eleftheriades, J. Appl. Phys. 92, 5930
(2002).
[7] F. Martı́n, J. Bonache, F. Falcone, M. Sorolla, and
R. Marqués, Appl. Phys. Lett. 83, 4652 (2003).
[8] T. J. Yen, W. J. Padilla, N. Fang, D. C. Vier, D. R. Smith,
J. B. Pendry, D. N. Basov, and X. Zhang, Science 303,
1494 (2004).
[9] R. Marqués, F. Medina, and R. Rafii-El-Idrissi, Phys.
Rev. B 65, 144440 (2002).
[10] Ph. Gay-Balmaz, and O. J. F. Martin, J. Appl. Phys. 92,
2929 (2002).
[11] J. D. Jackson, Classical Electrodynamics (Wiley,
New York, 1999), 3rd ed.
[12] R. Marqués, F. Mesa, J. Martel, and F. Medina, IEEE
Trans. Antennas Propag. 51, 2572 (2003).
[13] G.V. Eleftheriades, O. Siddiqui, and A. K. Iyer, IEEE
Microwave Wireless Compon. Lett. 13, 51 (2003).
[14] L. D. Landau, and E. M. Lifshitz, Electrodynamics of
Continuous Media (Pergamon Press, Oxford, 1984),
2nd ed.
197401-4
326 IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 10, 2011

Directivity Enhancement to Vivaldi Antennas Using

Compactly Anisotropic Zero-Index Metamaterials
Bin Zhou and Tie Jun Cui, Senior Member, IEEE

Abstract—The traditional Vivaldi antenna has an ultrawide

bandwidth, but low directivity. To enhance the directivity, we pro-
pose a high-gain Vivaldi antenna based on compactly anisotropic
zero-index metamaterials (ZIM). Such anisotropic ZIM are de-
signed and fabricated using resonant meander-line structures,
which are integrated with the Vivaldi antenna smoothly and hence
have compact size. Measurement results show that the directivity
and gain of the Vivaldi antenna have been enhanced significantly
in the designed bandwidth of anisotropic ZIM (9.5–10.5 GHz),
but not affected in other frequency bands (2.5–9.5 GHz and
10.5–13.5 GHz).
Index Terms—High directivity, metamaterials, Vivaldi antennas,
zero index of refraction.

I. INTRODUCTION
N THE past 10 years, great interests have been focused
I on metamaterials, and lots of fantastic and interesting phe-
nomena resulting from metamaterials have been demonstrated,
such as the negative refraction [1], invisibility cloaks [2]–[7],
and zero index of refraction [8]. Recently, the anisotropic
zero-index metamaterials (ZIM) have been proposed theo-
retically [9] and demonstrated experimentally [10]. Different
from the isotropic ZIM, the anisotropic ZIM with only one
component of the permittivity or permeability being zero have
high efficiency to radiate and receive electromagnetic waves
due to the impedance match [9], [10].
The Vivaldi antenna was first proposed by Gibson in
1979 [11], and has since been widely used in many problems, Fig. 1. Designed unit cell of the meander-line structure on a dielectric substrate,
such as ultrawideband (UWB) systems, microwave imaging, in which a = a =4 mm, w =02 : mm, m = 3 77
: mm, m =27 : mm.
The thickness of the substrate along the z -axis is 0.5 mm, and the dielectric
phase array systems, etc. The Vivaldi is a traveling-wave slot substrate is chosen as F4B with the permittivity of 2.65 and the tangent loss of
antenna with exponentially tapered shape having the good 0.001. (a) The electric polarization is along y . (b) The electric polarization is
feature of ultrawide bandwidth. However, the directivity of along x. (c) Retrieved results of the permittivity and permeability.
the original Vivaldi antenna is low. To improve the directivity,
some approaches have been developed, for example, using
high-directivity Vivaldi antenna. It is more important that the
the array of Vivaldi and using the photonic band-gap (PBG)
anisotropic ZIM structures can be embedded into the original
substrate [12]. However, such approaches are complicated and
Vivaldi antenna smoothly and easily in our design. Hence, the
costly.
designed anisotropic ZIM are not a burden to the Vivaldi an-
In this letter, we combine the good features of anisotropic
tenna, which can enhance the directivity and gain significantly
ZIM and traditional Vivaldi antenna together to develop a
in the bandwidth of anisotropic ZIM.

Manuscript received February 06, 2011; revised March 22, 2011; accepted
April 06, 2011. Date of publication April 15, 2011; date of current version April II. ANTENNA DESIGN AND RESULTS
28, 2011. This work was supported in part by a Major Project of the National
Science Foundation of China under Grants 60990320 and 60990324, the 111 Metamaterials are made up of many periodic element cells,
Project under Grant 111-2-05, and the National Science Foundation of China
under Grants 60871016, 60901011, and 60921063.
which can be either resonant cells such as the split-ring res-
The authors are with the State Key Laboratory of Millimeter Waves, De- onators (SRRs) and electric inductance–capacitance (ELC) res-
partment of Radio Engineering, Southeast University, Nanjing 210096, China onators [3], or nonresonant cells such as the I-shaped structures
(e-mail: binchouseu@gmail.com; tjcui@seu.edu.cn). [6]. Different kinds of element cells have their specific attri-
Color versions of one or more of the figures in this letter are available online
at http://ieeexplore.ieee.org. butions and applications. In the design of anisotropic ZIM, we
Digital Object Identifier 10.1109/LAWP.2011.2142170 choose an electrically resonant unit cell called the meander-line
1536-1225/$26.00 © 2011 IEEE
ZHOU AND CUI: DIRECTIVITY ENHANCEMENT TO VIVALDI ANTENNAS
Fig. 2. Traditional Vivaldi antenna and the directivity enhancement via AZIM.
(a) Traditional Vivaldi antenna model. (b) Experimental sample. (c) Simulation
results of near fields with and without AZIM (" = 10 and " =1 ) at
10 GHz. (d) Simulation results of far-field radiation patterns with and without
AZIM.
resonator [13], which can be cooperated with the Vivaldi an-
tenna smoothly and efficiently. The designed unit cell of the
327
Fig. 3. Design and fabricated high-gain Vivaldi antennas based on AZIM.
(a) Design model and (b) fabricated sample of the Vivaldi antenna with
single-layer AZIM. (c) Designed model and (d) fabricated sample of the
Vivaldi antenna with multilayer AZIM.
meander-line structure on a dielectric substrate is shown in
Fig. 1(a) and (b), in which mm, mm,
mm, mm. The thickness of the substrate
along the -axis is 0.5 mm, and the dielectric substrate is
chosen as F4B with the permittivity of 2.65 and the tangent
loss of 0.001. In Fig. 1(a), the electrical polarization is along
, resulting in no electrical resonance along this direction.
However, along , electrical resonance can be obtained in
328 IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 10, 2011

Fig. 4. Measurement results of Vivaldi antennas based on AZIM. (a) Return loss. (b) Radiation patterns at 9.5 GHz. (c) Radiation patterns at 10 GHz. (d) Radiation
patterns at 10.5 GHz. (e) Gain across the whole passband from 3 to 12 GHz.

Fig. 1(b). This can be used to construct anisotropic ZIM as procedure [14], the effective permittivity and permeability are
shown in Fig. 2(c). With the help of the standard retrieval obtained, as illustrated in Fig. 1(c). From Fig. 1(c), the zero
ZHOU AND CUI: DIRECTIVITY ENHANCEMENT TO VIVALDI ANTENNAS
TABLE I
MEASURED GAIN AND HPBW OF THE AZIM-BASED VIVALDI ANTENNA
permittivity point is observed in the frequency at around 10
GHz, in which the variable equals 2.7 mm. While is
equal to zero, and are other constants that can be designed
to match with the impedance in free space. Hence, this kind of
anisotropic metamaterial has great advantage in designing the
high-gain Vivaldi antenna.
The traditional Vivaldi antenna is a planar tapered slot an-
tenna, and the microstrip line is used to feed the antenna. We
have designed and fabricated such an antenna working in the
frequency range from 3 to 14 GHz, as shown in Fig. 2(a) and
(b). To verify the ability of anisotropic ZIM to enhance the
directivity of the Vivaldi antenna, we make full-wave simu-
lations of the Vivaldi antennas with and without a thin-layer
anisotropic ZIM in the front region of the antenna using the
COMSOL at 10 GHz, as demonstrated in Fig. 2(c) and (d). Here,
the anisotropic ZIM have relative permittivity components
and . Since the magnetic fields are vertical to
the substrate in the xoy plan, we use transverse magnetic (TM)
source to excite the wave propagation in the COMSOL mod-
eling. From the near-field distributions shown in Fig. 2(c), we
clearly observe that the Vivaldi antenna without the anisotropic
ZIM layer generates spherical-like waves, which result in low
directivity, while the Vivaldi antenna with the anisotropic ZIM
layer generates plane-like waves, which produce high direc-
tivity. This has been verified by the far-field radiation patterns
shown in Fig. 2(d). A significant enhancement of the directivity
is obviously observed with the thin anisotropic ZIM layer.
Following the above numerical analysis and the design of
metamaterial unit cells, we construct the new Vivaldi antennas
based on the anisotropic ZIM, where two designs with a single
layer and multiple layers in the vertical direction ( -direction)
are presented, as illustrated in Fig. 3. From Fig. 3(a) and (b), the
single anisotropic ZIM layer is embedded in the front region of
the Vivaldi antenna, having the same substrate. Hence, the fabri-
cation of the single-ZIM-layer Vivaldi antenna is easy, making
a smooth mixture of the metamaterial with antenna. The mul-
tiple-layer ZIM need several layers of dielectric substrates with
meander-line patterns that are the same as the single-layer ones,
and the foam whose height was equal to 4 mm is used to con-
nect each layer, as shown in Fig. 3(c) and (d). In this case, the
329
antenna performance can be improved significantly, but the fab-
rication is still easy, and the size is still compact. To enhance the
directivity of the Vivaldi antenna at other frequencies, we only
need to change the arm dimension of the meander line, , to
adjust the zero-index point.
The measurement results of Vivaldi antennas based on the
proposed anisotropic ZIM are illustrated in Fig. 4, in which
the traditional Vivaldi antenna is also measured to compare the
performance. From the measured return loss ( ) shown in
Fig. 4(a), it is clear that the ultrawideband feature of the Vivaldi
antenna is well kept after using the anisotropic ZIM: The re-
turn loss is less than 10 dB from 2.5 to 13.5 GHz. How-
ever, the gain and directivity of the antenna have been greatly
enhanced by the anisotropic ZIM in the frequency band from
9.5 to 10.5 GHz, as demonstrated in Fig. 4(b)–(d). In other
frequency bands (2.5–9.5 GHz and 10.5–13.5 GHz), the an-
tenna performance is not affected, as demonstrated in Fig. 4(e).
The detailed comparison of antenna gain and half-power band-
width (HPBW) is given in Table I. Obviously, more significant
improvements of gain and HPBW are observed using the mul-
tiple-layer anisotropic ZIM.
III. CONCLUSION
In this letter, the compact single-layer and multiple-layer
planar anisotropic ZIM are proposed to enhance the gain and
directivity of the Vivaldi antenna. The measured results verify
the significant enhancements. The designed meander-line
structures are well integrated to the antenna, and hence the new
antenna can find broad applications.
REFERENCES
[1] R. A. Shelby, D. R. Smith, and S. Schultz, “Experimental verification
of a negative index of refraction,” Science, vol. 292, pp. 77–79, 2001.
[2] J. B. Pendry, D. Schurig, and D. R. Smith, “Controlling electromagnetic
fields,” Science, vol. 312, pp. 1780–1783, 2006.
[3] D. Schurig, J. J. Mock, B. J. Justice, S. A. Cummer, J. B. Pendry, A.
F. Starr, and D. R. Smith, “Metamaterial electromagnetic cloak at mi-
crowave frequencies,” Science, vol. 314, pp. 977–980, 2006.
[4] U. Leonhardt and T. Tyc, “Broadband invisibility by non-euclidean
cloaking,” Science, vol. 323, pp. 110–112, 2009.
[5] J. Li and J. B. Pendry, “Hiding under the carpet: A new strategy for
cloaking,” Phys. Rev. Lett., vol. 101, p. 203901, 2008.
[6] R. Liu, C. Ji, J. J. Mock, J. Y. Chin, T. J. Cui, and D. R. Smith, “Broad-
band ground-plane cloak,” Science, vol. 323, pp. 366–369, 2009.
[7] H. F. Ma and T. J. Cui, “Three-dimensional broadband ground-plane
cloak made of metamaterials,” Nature Commun., vol. 1, p. 21, 2010,
DOI: 10.1038/ncomms 1023.
[8] M. Silveirinha and N. Engheta, “Design of matched zero-index meta-
materials using nonmagnetic inclusions in epsilon-near-zero media,”
Phys. Rev. B, vol. 75, p. 075119, 2007.
[9] Y. G. Ma, P. Wang, X. Chen, and C. K. Ong, “Near-field plane-wave-
like beam emitting antenna fabricated by anisotropic metamaterial,”
Appl. Phys. Lett., vol. 94, p. 044107, 2009.
[10] Q. Cheng, W. X. Jiang, and T. J. Cui, “Radiation of planar electromag-
netic waves by a line source in anisotropic metamaterials,” J. Phys. D,
Appl. Phys., vol. 43, p. 335406, 2010.
[11] P. J. Gibson, “The Vivaldi aerial,” in Proc. 9th Eur. Microw. Conf.,
1979, pp. 101–105.
[12] T. J. Ellis and G. M. Rebeiz, “MM-wave tapered slot antennas on mi-
cromachined photonic bandgap dielectrics,” in IEEE MTT-S Int. Mi-
crow. Symp. Dig., 1996, pp. 1157–1160.
[13] W. X. Tang, H. Zhao, X. Y. Zhou, J. Y. Chin, and T. J. Cui, “Negative
index material composed of meander lines and SRRs,” Prog. Electro-
magn. Res. B, vol. 8, pp. 103–114, 2008.
[14] D. R. Smith, S. Schultz, P. Markos, and C. M. Soukoulis, “Deter-
mination of effective permittivity and permeability of metamaterials
from reflection and transmission coefficients,” Phys. Rev. B, vol. 65,
p. 195104, 2002.
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 11, NOVEMBER 1999
Magnetism from Conductors and Enhanced
Nonlinear Phenomena
J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, Member, IEEE
Abstract— We show that microstructures built from nonmag-
netic conducting sheets exhibit an effective magnetic permeability
e , which can be tuned to values not accessible in naturally
occurring materials, including large imaginary components of
e . The microstructure is on a scale much less than the wave-
length of radiation, is not resolved by incident microwaves, and
uses a very low density of metal so that structures can be
extremely lightweight. Most of the structures are resonant due to
internal capacitance and inductance, and resonant enhancement
combined with compression of electrical energy into a very small
volume greatly enhances the energy density at critical locations
in the structure, easily by factors of a million and possibly
by much more. Weakly nonlinear materials placed at these
critical locations will show greatly enhanced effects raising the
possibility of manufacturing active structures whose properties
can be switched at will between many states.
Index Terms— Effective permeability, nonlinearity, photonic
crystals.
I. INTRODUCTION
I N A SENSE, every material is a composite, even if the
individual ingredients consist of atoms and molecules. The
original objective in defining a permittivity and permeability
was to present an homogeneous view of the electromagnetic
properties of a medium. Therefore, it is only a small step to
replace the atoms of the original concept with structure on a
larger scale. We shall consider periodic structures defined by
a unit cell of characteristic dimensions . The contents of the
cell will define the effective response of the system as a whole.
Clearly, there must be some restrictions on the dimensions
of the cell. If we are concerned about the response of the sys-
tem to electromagnetic radiation of frequency , the conditions
are easy to define as follows:
(1)
If this condition were not obeyed, there would be the
possibility that internal structure of the medium could diffract
as well as refract radiation giving the game away immediately.
Long wavelength radiation is too myopic to detect internal
structure and, in this limit, an effective permittivity and
permeability is a valid concept. In Section II, we shall discuss
how the microstructure can be related to , .
In an earlier paper [1], we showed how a structure consisting
of very thin infinitely long metal wires arranged in a three-
Manuscript received March 8, 1999; revised July 9, 1999.
J. B. Pendry is with The Blackett Laboratory, Imperial College, London
SW7 2BZ, U.K.
A. J. Holden, D. J. Robbins, and W. J. Stewart are with GEC-Marconi
Materials Technology Ltd, Caswell, Towcester, NN12 8EQ, U.K.
Publisher Item Identifier S 0018-9480(99)08781-5.
0018–9480/99$10.00  1999 IEEE
2075
dimensional (3-D) cubic lattice could model the response of a
dilute plasma, giving a negative below a plasma frequency
somewhere in the gigahertz range. Theoretical analysis of this
structure has been confirmed by experiment [2]. Sievenpiper
et al. have also investigated plasma-like effects in metallic
structures [3], [4].
Ideally, we should like to proceed in the magnetic case by
finding the magnetic analogue of a good electrical conductor:
unfortunately, there is not one. Nevertheless, we can find some
alternatives that we believe do give rise to interesting magnetic
effects.
Why should we go to the trouble of microstructuring a
material simply to generate a particular ? The answer
is that atoms and molecules prove to be a rather restrictive
set of elements from which to build a magnetic material.
This is particularly true at frequencies in the gigahertz range
where the magnetic response of most materials is beginning
to tail off. Those materials, such as the ferrites, that remain
moderately active are often heavy, and may not have very
desirable mechanical properties. In contrast, we shall show,
microstructured materials can be designed with considerable
magnetic activity, both diamagnetic and paramagnetic, and
can, if desired, be made extremely light.
There is another quite different motivation. We shall see
that strong magnetic activity implies strongly inhomogeneous
fields inside the material. In some instances, this may result
in local field strengths many orders of magnitude larger
than in free space. Doping the composite with nonlinear
material at the critical locations of field concentration gives
enhanced nonlinearity, reducing power requirements by the
field enhancement factor. This is not an option available in a
conventional magnetic material.
We show first how to calculate for a system, then we
propose some model structures that have magnetic activity
and give some numbers for these systems. Finally, we show
how electrostatic energy can be strongly concentrated in these
structures and, hence, demonstrate the potential for enhancing
nonlinear effects.
II. DEFINING AN EFFECTIVE PERMEABILITY
We are seeking to build structures with effective epsilon
and as follows:
(2)
where we assume that the structure is on a scale much shorter
than the wavelength of any radiation so that we can sensibly
2076 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 11, NOVEMBER 1999
Fig. 1. Unit cell of a periodic structure. We assume that the unit cell
dimensions are much smaller that the wavelength of radiation, and average
over local variations of the fields. In the case of the B -field, we average over
the faces of the cell and in the case of the H -field, over one of the edges.
speak of an average value for all the fields. A key question is
“how do the averages differ?” Clearly, if the structure is made
of thin wires or sheets of metal, then if the averages were
taken over the same regions of space, , would always
be unity. However, we observe that Maxwell’s equations
(3)
may be applied in the integral form
(4)
where the line integral is taken over a loop “ ,” which encloses
an area “ .”
This form of the equations immediately suggests a prescrip-
tion for averaging the fields. For simplicity, we shall assume
that the periodic structure is described by a unit cell whose
axes are orthogonal, as shown in Fig. 1. Some of the arguments
used in this section are similar to those we used in deriving a
finite-difference model of Maxwell’s equations [5].
We choose to define the components of by averaging
the -field along each of the three axes of the unit cell. If we
assume a simple cubic system
(5)
There is only one caveat concerning the definition of the
unit cell: its edges must not intersect with any of the structures
contained within the unit cell. This leaves us free to cut the
structure into a whole number of unit cells when we come
to create a surface and ensures that the parallel component
of is continuous across the surface as required in a
consistent theory of an effective medium.
To define , we average the -field over each of the
three faces of the unit cell, defined as follows:
surface defined by the vectors , ;
surface defined by the vectors , ;
Fig. 2. Model A consists of a square array of metallic cylinders designed to
have magnetic properties in the direction parallel to the axes of the cylinders.
surface defined by the vectors , .
Hence, we define
(6)
The ratio defines the effective epsilon and from (2)
(7)
Thus, if we seek a large effect, we must try to create fields
that are as inhomogeneous as possible.
We shall explore various configurations of thin sheets of
metal, derive , and discuss the results with a view to
making the effect as large as possible.
III. EXAMPLES OF MAGNETIC MICROSTRUCTURES
A. An Array of Cylinders
We start with a very simple structure for the purposes of
illustration, i.e., “model A” shown in Fig. 2. Let us apply an
external field , which we shall take to be parallel to the
cylinders. We assume that the cylinders have a conducting
surface so that a current per unit length flows. The field
inside the cylinders is
(8)
where the second term on the right-hand side is the field caused
directly by the current, and the third term is the result of
the depolarizing fields with sources at the remote ends of the
cylinders. If the cylinders are very long, the depolarizing field
will be uniformly spread over the unit cell, but will have the
same number of lines of force in it as the direct field inside
the cylinders. We now calculate the total electromotive force
PENDRY et al.: MAGNETISM FROM CONDUCTORS AND ENHANCED NONLINEAR PHENOMENA
(emf) around the circumference of a cylinder as follows:
emf
(9)
where is the resistance of the cylinder surface per unit area.
The net emf must balance and, therefore,
(10)
We are now in a position to calculate the relevant averages.
The average of the -field over the entire unit cell is
(11)
However, if we average the -field over a line lying entirely
outside the cylinders
(12)
Hence, we define
(13)
For an infinitely conducting cylinder or in the high fre-
quency limit, is reduced by the ratio of the cylinder
volume to the cell volume. This ratio of volumes will turn
out to be the key factor in determining the strength of the
effect in all our models. Evidently, in the present model,
can never be less than zero or greater than unity. It should
also be mentioned that to maximize the effect, we could have
replaced the metallic cylinders with prisms of square cross
section to maximize the volume enclosed within the prism.
If the resistivity of the sheets is high, then the additional
contribution to is imaginary, but always less than unity
(14)
A further point that should be noted is that all the structures
we discuss have electrical as well as magnetic properties. In
2077
Fig. 3. Model B consists of a square array of cylinders as for model A, but
with the difference that the cylinders now have internal structure. The sheets
are divided into a “split ring” structure and separated from each other by a
distance d. In any one sheet, there is a gap that prevents current from flowing
around that ring.
Fig. 4. When a magnetic field parallel to the cylinder is switched on it
induces currents in the “split rings,” as shown here. The greater the capacitance
between the sheets, the greater the current.
this particular case, we can crudely estimate for electric fields
perpendicular to the cylinders
(15)
where is the fraction of the structure not internal to a
cylinder. In deriving (15), we assume that the cylinder is
a perfect conductor and neglect depolarizing fields arising
from interaction between cylinders. Inclusion of in our
calculations removes one difficulty by ensuring that
(16)
Evidently, without the velocity of light in the effective
medium would have exceeded that in free space. Most of the
structures discussed in this paper have a similar .
B. Capacitative Array of Sheets Wound on Cylinders
The previous structure showed a limited magnetic effect.
We now show how to extend the range of magnetic properties
available to us by introducing capacitative elements into the
structure. We take the same structure of cylinders as before,
except that the cylinders are now built in a “split ring”
configuration, as shown in Fig. 3.
The important point is that there is a gap that prevents
current from flowing around any one ring. However, there
is a considerable capacitance between the two rings, which
enables current to flow (see Fig. 4).
Detailed calculations give
(17)
2078 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 11, NOVEMBER 1999

Fig. 5. The effective magnetic permeability for model B shows a resonant

structure dictated by the capacitance between the sheets and magnetic in-
Fig. 6. Generic dispersion relationship for resonant structures with a e .
ductance of the cylinder. We sketch the typical form of a highly conducting
sample,   0. Below the resonant frequency, e is enhanced, but above The solid lines represent twofold degenerate transverse modes and the dashed
resonance, e is less than unity and may be negatively close to the resonance.
line a single longitudinal magnetic plasmon mode.

where is the fractional volume of the cell occupied by the

of the plasma modes of a gas of free electrical charges [6],
interior of the cylinder
[7]. Of course, we have no free magnetic poles, only the
appearance of such as currents around the cylinders make the
(18)
cylinder ends appear to support free magnetic poles in the
and is the capacitance per unit area between the two sheets fashion of a bar magnet.
Together with , given in (15), which is also applicable
(19) here, we can illustrate a generic dispersion relationship, as
shown in Fig. 6.
Hence, The relevant points to note are as follows.
1) Wherever is negative there is a gap in the dispersion
relationship, i.e., for
(20)
(25)

2) A longitudinal magnetic plasma mode, dispersionless in

Since we now have capacitance in the system that can
this approximation, is seen at .
balance the inductance present, has a resonant form, which
3) The dispersion relation converges asymptotically to the
is sketched in Fig. 5.
free-space light cone, as discussed above. In fact, metal-
Fig. 5 illustrates the generic form of for all the struc-
lic structures in general represent a fresh approach to the
tures we present here.
photonic insulator concept introduced independently by
We define to be the frequency at which diverges
Yablonovitch [8], [9] and John [10].
as follows:
If we take the following values:
(21) m
and to be the “magnetic plasma frequency” m
m (26)
(22) we get
Hz (27)
Note that the separation between and , which is
a measure of the range of frequencies over which we see a Hz (28)
strong effect, is determined by
Note that the frequency at which the structure is active
(23) corresponds to a free-space wavelength of 10 cm, much greater
that the 0.5-cm separation between cylinders. This will be
the fraction of the structure not internal to a cylinder. As for typical of these capacitative structures and implies that the
the case , the simple cylinder, the high frequency limit is effective medium approximation will be excellent.
given by
C “Swiss Roll” Capacitor
(24)
We take the same arrangement of cylinders on a square
We mention in passing that the system sustains longitudinal lattice as before, except that the cylinders are now build
magnetic modes at the magnetic plasma frequency, the analog as shown in Fig. 7. The important point is again that no
PENDRY et al.: MAGNETISM FROM CONDUCTORS AND ENHANCED NONLINEAR PHENOMENA 2079

Fig. 7. In model C , a metallic sheet is wound around each cylinder in a coil.

Each turn of the coil is spaced by a distance d from the previous sheet.

Fig. 9. Dispersion with frequency of e for a Swiss roll structure, cal-

culated for the parameters shown in (36), assuming that the metal has zero
resistivity.

If we take the values we used before in (26)

m
m
Fig. 8. When a magnetic field parallel to the cylinder is switched on, it m
reduces currents in the coiled sheets, as shown here. Capacitance between the (33)
first and last turns of the coil enables the current to flow.
we get
current can flow around the coil, except by virtue of the self Hz (34)
capacitance (see Fig. 8).
In this instance, we find for the effective permeability (35)
i.e., there is much more capacitance in this model and the range
of active frequencies is an order of magnitude lower than it
was in model , which used only two overlapping sheets.
Choosing an even smaller scale and reducing the number
of turns in order to drive up the frequencies to our range of
(29) interest
m
where is as before the fraction of the structure not internal m
to a cylinder, and the capacitance per unit area between the m
first and the last of the coils is (36)
(30) we get

The critical frequencies are Hz (37)

Hz (38)
Note that the free-space wavelength at the plasma frequency
is around 3 cm, and compare this to the very much smaller
spacing between cylinders of 0.05 cm.
(31)
We shall now calculate the dispersion of for various
parameters. First let us take the parameters given in (36). The
resulting dispersion of is shown in Fig. 9.
We next enquire “what is the effect of making the sheets
resistive?” Below we present a series of calculations for
(32) various values of the resistivity given in
In Fig. 10, we increase the resistivity from 0.1 to 10.0 .
Note the broadening of the resonance, the complementary
2080 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 11, NOVEMBER 1999

Fig. 10. Dispersion with frequency of e for a Swiss roll structure, calculated for the parameters shown in (36), for various values of the resistivity
of the sheets: 0.1, 2.0, 5.0, 10.0
.

(a) (b)
Fig. 11. Dispersion with frequency of e for a Swiss roll structure. (a) Calculated for the parameters shown in (36), except that the resistivity of the sheets
is now 2.0
, and the radius of the cylinders has been reduced from 2.0 1004 to 12.6 1004 m, thus raising the resonant frequency by a factor of two.
2 2
(b): d, the spacing between the sheets, has been reduced to 0.25 21005 m, bringing the resonant frequency back to the original value.

behavior of and , dictated by Kramers Kronig, and
how resistivity limits the maximum effect achieved.
We next explore the dependence on the radius of the
cylinders. In Fig. 11, the radius of the cylinders is decreased,
reducing the volume fraction occupied by the cylinders, and
raising the resonant frequency by a factor of two. We also
decrease , the spacing between the sheets, increasing the
capacitance in the system and bringing the resonant frequency
back down to its original value.
Using capacitative cylindrical structures such as the Swiss
roll structure, we can adjust the magnetic permeability typ-
ically by a factor of two and, in addition, if we desire,
introduce an imaginary component of the order of unity. The
latter implies that an electromagnetic wave moving in such
PENDRY et al.: MAGNETISM FROM CONDUCTORS AND ENHANCED NONLINEAR PHENOMENA
(a) (b)
Fig. 12. (a) Plan view of a split ring showing definitions of distances. (b)
Sequence of split rings shown in their stacking sequence. Each split ring
comprises two thin sheets of metal. The ring shown is a scaled-up version,
defined by the parameters shown in Fig. 13.
a material would decay to half its intensity within a single
wavelength. This presumes that we are seeking broad-band
effects that persist over the greater part of the 2–20-GHz
region. However, if we are prepared to settle for an effect
over a narrow range of frequencies, spectacular enhancements
of the magnetic permeability can be achieved, limited only be
the resistivity of the sheets and by how narrow a band we are
willing to tolerate.
IV. AN ISOTROPIC MAGNETIC MATERIAL
The structures shown above give magnetic properties when
the field is aligned along the axes of the cylinders, but
have essentially zero magnetic response in other directions.
They suffer from another potential problem: if the alternate
polarization is considered where the electric field is not parallel
to the cylinders, the system responds like an effective metal
because current is free to flow along the length of the cylinders.
For some applications, this highly anisotropic behavior may be
undesirable. Therefore, we redesign the system with a view to
restoring isotropy and minimizing purely electrical effects.
To this end, we need a basic unit that is more easily packed
into arrays than is a cylinder and that avoids the continuous
electrical path provided by a metal cylinder. We propose an
adaptation of the “split ring” structure, in which the cylinder
is replaced by a series of flat disks each of which retains
the “split ring” configuration, but in slightly modified form
(see Fig. 12). First, we shall calculate the properties of disks
stacked in a square array, as shown if Fig. 13. This structure
is still anisotropic, a problem we shall address in a moment,
but by eliminating the continuous conducting path that the
cylinders provided, it eliminates most of the electrical activity
along this direction.
The two-dimensional square array of Fig. 13 can be made
by printing with metallic inks. If each printed sheet is then
fixed to a solid block of inert material with thickness , the
blocks can be stacked to give columns of rings. This would
establish magnetic activity along the direction of stacking, i.e.,
the -axis. The unit cell of this structure is shown in Fig. 14
on the left-hand side.
How do we make a symmetrical structure? Start from the
structure just described, comprising successive layers of rings
2081
Fig. 13. Plain view of a split ring structure in a square array (lattice spacing
a).
Fig. 14. Building 3-D symmetry: each successive restacking of the structure
adds a ring to another side of the unit cell.
stacked along the -axis. Next, cut up the structure into a
series of slabs thickness , make incisions in the – -plane,
and be careful to avoid slicing through any of the rings. Each
of the new slabs contains a layer of rings, but now each ring is
perpendicular to the plane of the slab and is embedded within.
Print onto the surface of each slab another layer of rings and
stack the slabs back together again. The unit cell of this second
structure is shown in the middle of Fig. 14.
In the next step, a third set of slabs is produced by cutting
in the – -plane, printing on the surface of the slabs, and
reassembling. Finally, we now have a structure with cubic
symmetry whose unit cell is shown in the right-hand side of
Fig. 14.
Of course, an alternate method of manufacturing this struc-
ture would be to start from a set of cubes of the inert material
and laboriously stick rings to their sides before assembling
the cubes into a lattice. The cut-and-paste method we suggest
above is much more efficient.
Now, let us calculate the effective permeability. First, we
need to calculate the capacitance between the two elements of
the split ring. We shall assume
(39)
(40)
(41)
Under these conditions, we can calculate the capacitance
between unit length of two parallel sections of the metallic
strips
(42)
2082 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 11, NOVEMBER 1999

(a) (b)
Fig. 15. Plot of e for the cubic split ring structure calculated using the chosen parameters. (a) For copper rings, 1 = 200:0. (b) For more
resistive rings, 1 = 2000:0.

The effective magnetic permeability we calculate, on the
assumption that the rings are sufficiently close together and
that the magnetic lines of force are due to currents in the
stacked rings, are essentially the same as those in a continuous
cylinder. This can only be true if the radius of the rings is of
the same order as the unit cell side. We arrive at
If we choose to manufacture the split rings from a layer of
copper, it is easily possible to achieve . Evidently,
from Fig. 15, this produces a highly resonant structure.
In order to see a substantial effect, we have to increase the
resistance either by increasing the resistivity of the material of
which the rings are made or by making them thinner.
The scaling of frequency with size can be deduced from
(45), in which we see that the resonant frequency scales
uniformly with size: if we double the size of all elements
in a given structure, the resonant frequency halves. Nearly all
the critical properties are determined by this frequency.

(43) V. ENHANCED NONLINEAR EFFECTS

We have seen how the addition of capacitance to the struc-
ture gives a far richer variety of magnetic behavior. Typically,
this happens through a resonant interaction between the natural
where is the resistance of unit length of the sheets measured
inductance of the structure and the capacitative elements and,
around the circumference.
at the resonant frequency, electromagnetic energy is shared
To give some examples, let us choose a convenient set of
between the magnetic fields and the electrostatic fields within
parameters
the capacitative structure. To put this more explicitly, take the
m split ring structure described in Figs. 12 and 13, most of the
electrostatic energy of the capacitor is located in the tiny gap
m
between the rings. Concentrating most of the electromagnetic
m energy in this very small volume will results in an enormously
m enhanced energy density.
m (44) If we wish to enhance the nonlinear behavior of a given
compound, we locate a small amount of the substance in the
Figs. 12 and 13 show the rings drawn to scale. These gap where the strong electrostatic fields are located. Since
parameters do not quite satisfy all the inequalities, which the response scales as the cube of the field amplitude, we
is difficult to do with reasonable numbers, but note that the can expect enhancements of the order of the energy density
inequalities are only important to the accuracy of our formulas, enhancement squared. Furthermore, not only does the structure
not to the functioning of the structure. The resonant frequency enhance the nonlinearity, it does so in a manner that is very
at which diverges is given by economical with the material: less that 1% of the structure
need be filled with the nonlinear substance.
(45) Note that there is a symmetry between, on the one hand,
the present structures designed to generate a magnetic perme-
ability and within which we find enhanced electrostatic fields
or and, on the other hand, the earlier thin-wire structures [1],
[2] designed to generate a negative electrical permittivity, and
GHz (46) within which we find enhanced magnetic fields.
PENDRY et al.: MAGNETISM FROM CONDUCTORS AND ENHANCED NONLINEAR PHENOMENA 2083

Fig. 16. The emf acting around one of the sheets of the split ring in Fig. 12
as a function of the distance s around the ring. Vin denotes the emf on the
inner ring, and Vout that on the outer ring. Note that this ring is cut at s = 0
so that the emf is discontinuous.

Fig. 17. Enhancement of the energy density of the electric field within the
We shall now calculate the energy density in the capacitance gap between the split rings (see Figs. 12 and 13) for two different values of
between the two split rings in Figs. 12 and 13. First, we the resistivity of the metal sheet. The corresponding values of e are shown
calculate the voltage between the two rings as a function of in Fig. 15.
the incident magnetic field . The electric field between the
two halves of the ring is shown in Fig. 16 and is of the order Hence,
(47)
(53)
We calculate that
A more detailed picture of enhancement as a function of
frequency is shown in Fig. 17.
For example, a beam of microwaves at 13.41 GHz with
(48) power flux of 10 W m has an electric field strength of the
Hence, on substituting from (42) and (47) into (48) order of 2 10 V m in vacuo. If this beam were incident
on, and entirely transmitted into, our magnetic structure, it
would generate a field strength of the order of 10 V m
in the space between the split rings, or of the order of 10
V between the edges of the two rings: more than enough
to cause electrical breakdown in air. It is evident that these
(49) structures have considerable potential for enhancing nonlinear
We now argue that the electrostatic energy density in the phenomena. Furthermore, the nonlinear medium need only
incident electromagnetic field is equal to the magnetic energy be present in the small volume within which the energy is
density, which, in turn, can be related to the electrostatic concentrated, opening the possibility of using small quantities
energy density in the ring. Hence, of expensive material, and reducing any requirements of
mechanical integrity that a larger structure would impose.
In passing, we draw an analogy with surface-enhanced
Raman scattering (SERS), observed on rough metallic
surfaces—typically silver surfaces. The Raman signal from
molecules absorbed on these surfaces may be enhanced by
(50) factors of the order of 10 over that seen on insulating surfaces.
The Raman effect is proportional to the second power of the
If we evaluate this formula on resonance, we get a much electromagnetic-mode density at the surface, and it is known
simplified formula as follows: that roughness can enhance the local-mode density by factors
resonant enhancement of up to 10 –10 , hence, the spectacular Raman enhancement
(see [11] for further details and references). A very similar
(51) local enhancement takes place in our system and, we expect,
can be exploited in an analogous fashion.
In conclusion, we have shown how to design structures
Let us take as an example the following parameters used made from nonmagnetic thin sheets of metal, which respond
to calculate Fig. 15: to microwave radiation as if they had an effective magnetic
permeability. A wide range of permeabilties can be achieved
m
by varying the parameters of the structures. Since the active
m ingredient in the structure, the metal film, comprises a very
m small fraction of the volume, typically 1:10 , the structures
may be very light, and reinforced with strong insulating mate-
rial to ensure mechanical strength, without adversely affecting
(52) their magnetic properties. It is likely that the structures will
2084 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 47, NO. 11, NOVEMBER 1999

be exploited for their ability to concentrate the electromagnetic [12] P. M. Bell, J. B. Pendry, L. Martøn-Moreno, and A. J. Ward, “A program
energy in a very small volume, increasing its density by a huge for calculating photonic band structures and transmission coefficients of
complex structures,” Comput. Phys. Commun., vol. 85, p. 306, 1995.
factor, and greatly enhancing any nonlinear effects present. [13] E. Yablonovitch, T. J. Gmitter, and K. M. Leung, “Photonic band
structure: The face-centered-cubic case employing nonspherical atoms,”
REFERENCES Phys. Rev. Lett., vol. 67, pp. 2295–2298, 1991.
[14] J. B. Pendry, “Calculating photonic band structure,” J. Phys. Condens.
[1] J. B. Pendry, A. J. Holden, W. J. Stewart, and I. Youngs, “Extremely Matter, vol. 8, pp. 1085–1108, 1996.
low frequency plasmons in metallic meso structures,” Phys. Rev. Lett., [15] R. H. Ritchie, “Plasma losses by fast electrons,” Phys. Rev., vol. 106,
vol. 76, pp. 4773–4776, 1996. pp. 874–881, 1957.
[2] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, “Low [16] J. B. Pendry and A. MacKinnon, “Calculation of photon dispersion
frequency plasmons in thin wire structures,” J. Phys. Condens. Matter, relationships,” Phys. Rev. Lett., vol. 69, pp. 2772–2775, 1992.
vol. 10, pp. 4785–4809, 1998.
[3] D. F. Sievenpiper, M. E. Sickmiller, and E. Yablonovitch, “3D wire
mesh photonic crystals” Phys. Rev. Lett., vol. 76, pp. 2480–2483, 1996.
[4] D. F. Sievenpiper, E. Yablonovitch, J. N. Winn, S. Fan, P. R. Villeneuve,
and J. D. Joannopoulos, “3D metallo-dielectric photonic crystals with J. B. Pendry, photograph and biography not available at the time of
strong capacitive coupling between metallic islands,” Phys. Rev. Lett., publication.
vol. 80, pp. 2829–2832, 1998.
[5] J. B. Pendry “Calculating photonic band structure,” J. Phys. Condens.
Matter, vol. 8, pp. 1085–1108, 1996.
[6] D. Pines and D. Bohm, “A collective description of electron interactions:
II collective versus individual particle aspects of the interactions,” Phys. A. J. Holden, photograph and biography not available at the time of
Rev., vol. 85, pp. 338–353, 1952. publication.
[7] D. Bohm and D. Pines, “A collective description of electron interactions:
III Coulomb interactions in a degenerate electron gas,” Phys. Rev., vol.
92, pp. 609–625, 1953.
[8] E. Yablonovitch, “Inhibited spontaneous emission in solid state physics
and electronics,” Phys. Rev. Lett., vol. 58, pp. 2059–2062, 1987. D. J. Robbins, photograph and biography not available at the time of
[9] , “Photonic band gap crystals,” J. Phys. Condens. Matter, vol. 5, publication.
pp. 2443–2460, 1993.
[10] S. John, “Strong localization of photons in certain disordered lattices,”
Phys. Rev. Lett., vol. 58, pp. 2486–2489, 1987.
[11] F. J. Garcia Vidal and J. B. Pendry, “Collective theory for surface
enhanced Raman scattering,” Phys. Rev. Lett., vol. 77, pp. 1163–1166, W. J. Stewart (M’88), photograph and biography not available at the time
1996. of publication.
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/LAWP.2017.2690973, IEEE
Antennas and Wireless Propagation Letters
Miniaturized Planar Yagi Antenna Utilizing
Capacitively-Coupled Folded Reflector
S. Ahdi Rezaeieh, Member, IEEE, M.A. Antoniades, Senior Member, IEEE, A.M. Abbosh, Senior Member IEEE
 driven-element that represents a folded dipole structure [5], the
Abstract— The analysis and design of a unidirectional and
wideband printed Yagi-antenna with miniaturized size using a
capacitively-coupled reflector is presented. Reflectors have been
widely used for cancelling back-radiation of Yagi antennas,
however their application to reduce the size of printed Yagi
antennas while maintaining their radiation characteristics has not
been previously investigated. This is important for applications
operating at low microwave frequencies requiring compact
antennas, such as telemetry and microwave imaging. To that end,
the driven element, which is a bowtie-dipole, is loaded with a pair
of capacitive gaps. Moreover, the reflector is folded towards the
bowtie driver, capacitively coupling them together. This
modification excites a patch mode that is considerably lower than
the main resonance of the bowtie. To enhance the directivity of the
antenna, a half-bowtie director is added near the driven bowtie.
Thus, the structure is miniaturized by more than 60% compared
to existing wideband printed Yagi antennas. Moreover, the
antenna attains a wide fractional bandwidth of 48% at
0.69-1.12 GHz with peak front-to-back ratio and gain values of 10
dB and 5.5 dBi, respectively.
Index Terms— Yagi-antenna, capacitive
unidirectional antenna, wideband antenna.
I. INTRODUCTION
Considering their stable gain, unidirectional radiation and
wide operating bandwidths, printed Yagi and quasi-Yagi
antennas are among the most desired antennas for many
applications, such as mobile systems, medical telemetry and
medical microwave imaging [1], [2]. However, considering the
inverse relation between the operating frequency of an antenna
and its size, the size of printed Yagi antennas operating at low
frequencies, e.g. at the lower end of microwave or ultra-high
frequencies (UHF), would be extremely large and not practical
if conventional design methods are used. This is because the
driven element of a Yagi-antenna consists of a half-wave dipole
antenna accompanied by a longer reflector element or a large
ground plane in the case of quasi-Yagi structures [3].
Several methods were suggested to address the large size of
conventional Yagi-antennas. The reported methods can be
summarized as a meandering technique for reducing the size of
the dipole [4], employing a pseudo-monopole antenna as a
The paper is submitted March 7, 2017.
S. Ahdi Rezaeieh and A.M. Abbosh are with the School of Information
Technology and Electrical Engineering, The University of Queensland, 4072,
Brisbane, Australia. (E-mail: s.ahdirezaeieh@uq.edu.au).
M. A. Antoniades is with the Department of Electrical and Computer
Engineering, University of Cyprus, 1678 Nicosia, Cyprus (e-mail:
mantonia@ucy.ac.cy). He is also an Adjunct Fellow in the School of
Information Technology and Electrical Engineering, The University of
Queensland, 4072 Brisbane, Australia (Email: m.antoniades@uq.edu.au)
1536-1225 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
1
addition of extended strip-line conductors to the ground plane
[6], the design of an arc-shaped radiator on a flexible substrate
[7], utilizing a reflector shorter than the driver by creating a
stepped slot on the reflector [8], using double dipole structures
[9], and increasing the electrical length of the antenna by
inductively loading elements of the Yagi antenna [10].
Investigating these structures reveals that the size of the
antenna is mainly reduced at the cost of multiband or
narrowband operation in [1]-[5], [7]-[8] or the need for
additional balanced-to–unbalanced (balun) structures to
achieve wider operating bandwidths as well as higher
front-to-back ratios (FBR) [3]. Thus, there is still a lack of a
comprehensive approach that can provide a compact structure
with a wide operating bandwidth and a moderate FBR as
required for modern short-range applications, such as wireless
local areas networks and microwave imaging [11].
A planar Yagi-antenna with a tightly coupled strip patch that
performs as both a reflector and a resonator is presented. To
coupling, lower the resonance of the antenna, a pair of capacitive gaps are
also created on the driven element, which is a bowtie antenna.
To enhance the limited capacitance created by the gaps and to
excite an additional lower resonance, the reflector (strip patch)
is located in the close vicinity of the dipole. It is folded to both
reduce the side length of the antenna and to excite the patch
mode resonance through increased capacitive coupling between
the bowtie and the reflector. To enhance the front-to-back ratio
(FBR), a director is placed at a close distance to the driven
element. It is designed to have a half-bowtie shape to both
occupy less area and to achieve maximum coupling with the
driven bowtie. The abovementioned structure achieves a
compact area of 0.28λ×0.23λ, where λ is the wavelength at the
lowest frequency. The proposed design is 60% smaller than
other wideband Yagi-antennas, and has a wide measured
bandwidth of 48%, and peak gain and FBR values of 5.5 dBi
and 10 dB, respectively.
II. ANTENNA DESIGN
Considering that a half-wavelength dipole is typically the
driven element of a Yagi-antenna, the bowtie structure that is
shown in Fig. 1(a) is selected as the main radiator of the
antenna. The bowtie structure is selected to provide a wider
operating bandwidth than conventional dipole structures [12].
Its size is adjusted to resonate at about the half-wavelength
mode centered at 1 GHz. It is printed on an FR4 epoxy substrate
with dielectric constant of 4.4, loss tangent of 0.02 and
thickness of 0.8 mm. To eliminate the need for large ground
planes and to avoid feeding the antenna directly by a coaxial
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/LAWP.2017.2690973, IEEE
Antennas and Wireless Propagation Letters
cable, a coplanar waveguide (CPW) feeding structure is
designed to feed the antenna. The antenna is fed using a
subminiature version A (SMA) connector that is inserted
through via holes and is connected to the corresponding arms of
the dipole (Fig. 1(a)).
To reduce the main resonance of the antenna, the bowtie is
loaded with series capacitors. The most straightforward method
of creating a series capacitance on the bowtie is to etch gaps on
both the arms. However, considering that on a dipole antenna,
strong currents are present around the feeding point at the
center of the structure, the gaps should be added far from the
center to avoid disturbing the surface currents, and hence
enable wider bandwidths to be achieved. Nevertheless, the
further the gaps are positioned, the loading capacitance value
reduces due to weaker currents existing towards the edge of the
bowtie. Thus, to achieve antenna miniaturization, the
capacitance of the gaps should be either increased using chip
capacitors, or an additional low-frequency resonance should be
excited.
While higher value chip capacitors are available, they are
avoided due to their lower self-resonant frequencies. Thus, to
avoid this shortfall, and to excite a lower resonance and reduce
the size of the antenna, the configuration that is shown in
Fig. 1(b) is proposed. As seen, the reflector is folded by 90
degrees with respect to and towards the center of the design and
is located in the vicinity of the bowtie dipole. The resulting
reflection coefficient obtained using ANSYS Electronic
Desktop (AEDT) 2016 is depicted in Fig. 1(c). This structure is
novel in the sense that conventionally reflectors have been used
to cancel back radiation, however, in the proposed design, the
reflector is used to resonate at its patch mode as well. The total
length of the reflector is designed to be slightly larger than half
a wavelength at the excited resonance of 0.73 GHz (coupled
patch resonance).
In the design of Yagi-Uda antennas, the reflector is generally
spaced at a 0.1 λ-0.25 λ distance from the driving element to
provide the required phase difference for back radiation
cancellation. However, since it is used in the proposed design
as a resonating element as well, it is folded and positioned at a
0.05 λ distance from the center of the dipole (λ is the
wavelength at the resonance of the dipole that is at around
1 GHz). With the folding configuration, two main goals are
achieved; firstly, the lateral size of the antenna is reduced by
folding the two ends of the reflector to 0.28λ , which is around
50% smaller than the side length of a conventional
Yagi-antenna. Secondly, the required energy to excite the lower
resonance is created by coupling the bowtie with the upright
arms of the reflector without the need for any connection or
external feeding.
(a)
1536-1225 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
2
(b)
(c)
Fig. 1. Proposed capacitively-loaded (a) bowtie antenna, (b) Yagi-antenna, and
(c) reflection coefficient performance from AEDT. L1 = 65, L2 = 37, L3 = 15, L4
= 7.4, L5 = 17.6, L6 = 5, L7 = 20, Ld = 44, Ldi = 60, g1 = 5.95, g2 = 17, gc = 8, W1 =
35, W2 = 50, W3 = 35, W4 = 5, W5 = 11, W6 = 50, W7 = 5, Lf = 8, g = 1, W= 120, Wd
= 54, L = 100, l1 = 43, l2 = 3, (Unit: mm)
The director’s structure is also designed in a
half-bowtie-shape to achieve better mutual coupling with the
driven element, the bowtie antenna. Its size is selected to be
around half a wavelength at 1 GHz. The capacitive loading
created on the bowtie and between the bowtie and the reflector
excites a new patch mode. Thus, the size of the antenna is
significantly reduced, enabling the antenna to achieve a wide
operating band between 0.7-1.06 GHz (fractional bandwidth of
41.5%), as seen in Fig. 1(c).
A. Impedance Matching Mechanism
The reflector of a Yagi-antenna has an inductive reactance
and its reactance value can be adjusted by either increasing its
length or modifying its width [8]. Therefore, it can be used to
adjust the input impedance of the antenna, and consequently
improve the operating bandwidth. Thus, the effect of either one
of these parameters on the input impedance are studied here. It
should be noted that at each stage only one of the parameters is
varied and the other parameters are kept fixed.
As mentioned before, the coupling between the bowtie and
the reflector is utilized to excite the patch resonance at
0.73 GHz. In the proposed design, both length and width of the
reflector are designed to provide maximum coupling with the
edges of the bowtie and adjust the input impedance of the
antenna for a wide operating bandwidth. This approach is
devised to limit the required lateral size for high capacitance
values. In this method, the width of the reflector is increased
towards the edges of the reflector, g1, to increase the capacitive
coupling between the vertical arms of the reflector and the
loaded bowtie. This phenomenon is evident from the surface
current at the edges of the reflector and the bowtie. As seen, the
level of the coupling at patch resonance of 0.73 GHz (see Fig. 2
(a)) is around 50% stronger compared to the resonance at 1
GHz (Fig. 2 (b)). Because of this strong coupling, the patch
(reflector) mode is generated, and the antenna resonates at
0.73 GHz. The resonance of the patch can be controlled by
increasing the length of the folded part of the reflector, Ld, or
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/LAWP.2017.2690973, IEEE
reducing the gap between the horizontal section of the reflector
and the dipole, g1 (Fig. 2(c)). The same scenario applies to
reducing the gap between the folded arm of the reflector and the
bowtie, gc, yet, these results are not presented for brevity. To
emphasize the importance of the capacitive coupling in the
excitation of the patch resonance, the reflection coefficient of
the antenna with a reduced length of the reflector’s upright
arms (Ld = 24 mm) is presented in Fig. 2(c). It is evident that no
patch resonance is excited due to the limited capacitance
created merely by the bowtie capacitive gaps .
(a)
(c)
Fig. 2. (a) Surface current magnitude at the coupled patch resonance of 0.73
GHz and (b) the dipole resonance of 1 GHz. (c) Reflection coefficient
variations with the length of the reflector and the gap between the reflector and
the capacitively-loaded bowtie.
(a)
(b)
(c)
Fig. 3. (a) Reflection coefficient, and (b) real and (c) imaginary part of the input
impedance variations with a change in the position of the director and the gap
between the central part of the reflector and the bowtie dipole.
While increasing the coupling between the dipole and the
reflector lowers the patch resonance, it also results in a
reduction of the operating bandwidth if the coupling is achieved
using all the sections between the dipole and the reflector. This
is because excessive coupling disturbs the strong surface
currents at the center of the dipole (see Fig. 2(a)), and hence
1536-1225 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Antennas and Wireless Propagation Letters
3
reduces the operating bandwidth of the antenna. Therefore, the
gap between the reflector and the bowtie dipole on the x-axis,
g2, is increased towards the center of the design as can be seen
from Fig. 1(b).
The aforementioned issue applies to the position of the
director as well. The reflection coefficient of the antenna with
changes in the position of the director is presented in Fig. 3(a).
As seen, by reducing the gap between the director and the
dipole, L5, the operating bandwidth gradually reduces. To
clearly show the reason behind this phenomenon, the real and
imaginary parts of the input impedance are depicted in Fig. 3(b)
and (c). By closing the gap, the reactance of the input
impedance becomes negative at lower frequencies due to the
excessive capacitive coupling between the director and the
dipole.
(b)
III. ANTENNA MEASUREMENTS
To verify the simulation results, a prototype of the proposed
design was fabricated and is presented in Fig. 4 (a). As shown
in Fig. 4 (b)-(c), to feed the antenna the lower legs of the SMA
were removed and the pins in the same row were used for
soldering. The central pin was soldered to the left arm while the
side pins were soldered to the right arm of the bowtie. The
excess parts of the pins were removed after the soldering
process. To minimize the spurious radiation using conventional
balanced-to-unbalanced (balun) structures, Richco© ring split
core clamp ferrite beads were used throughout the
measurements to block leakage currents on the feed cable. It is
noted that in applications where the antenna is used inside the
circuitry of a complete system, such as WLAN modems or
mobile phones, the need for the balun is automatically
eliminated and thus the achieved miniaturization is not
sacrificed using the vertical SMA and balun. The reflection
coefficient of the antenna is presented in Fig. 5 (a). There is a
reasonable agreement between the simulated and measured
results and a wide measured fractional bandwidth of 48% is
achieved, covering the band of 0.69-1.12 GHz. The shift at the
higher resonance is due to the disturbance of the fields at the
center of the bowtie caused by the soldering of the SMA and the
proximity effect of the coaxial cable. The gain of the antenna
was also measured in an anechoic chamber at the maximum
radiation direction (x-direction). As shown in Fig. 5 (b), the
antenna achieves a peak gain of 5.5 dBi at 1.05 GHz with a
stable average gain of more than 3 dBi across the operating
band. This is the main advantage of the proposed antenna
compared to the most recent wideband design [14], which has
very low gain at the central and higher frequencies to achieve
the same miniaturization levels as the proposed antenna.
The radiation patterns of the antenna were measured at the
two principle x-y and x-z planes at the three sample frequencies
of 0.75 GHz, 0.85 GHz and 1 GHz. These results are co-plotted
with the simulated radiation patterns and presented in Fig. 6. As
seen, there is good agreement between the measured and
simulated results. The antenna achieves an average front-to
back-ratio (FBR) value of 8 dB and 9 dB at 0.75 GHz and
0.85 GHz, respectively, with a peak FBR value of 10 dB at
1 GHz. The slightly lower FBR values at the lower frequencies
are due to the closely located reflector with respect to the
driver, as needed to create the required patch resonance.
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/LAWP.2017.2690973, IEEE
Antennas and Wireless Propagation Letters
4

Consequently, the resultant phase difference between the TABLE I

COMPARISON OF THE PROPOSED WITH RECENT PLANAR YAGI-ANTENNAS
driven element and the reflector is not optimum. Nevertheless, Centre Freq. Min FBR
the antenna radiates most of its accepted power in the Ref. Size (λ) BW (%) Area (λ02)
(GHz) (dB)
x-direction. Moreover, the co-polarization levels are 25 dB [4] 0.34×0.27 1.6 & 2.6 4 & 6.5 8 0.09
[5] 0.5×0.45 2.4 8 15 0.23
higher than the cross-polarization levels in the x-direction at all [6] 0.36×0.41 7 100 10 0.15
measured frequencies. [7] 0.17×0.43 1.58 5 10 0.07
To highlight the success of the proposed design, its [8] 0.46×0.38 2.4 8 15 0.17
Prop. 0.28 × 0.23 0.9 48 8 0.06
performance is compared to the most recent designs in the
IV. CONCLUSION
literature in Table I. As can be seen, the proposed antenna is
60% smaller than existing wideband planar Yagi-antenna A compact and wideband planar Yagi antenna has been
designs [6]. By carefully studying the tabulated criteria, it can presented. To create a lower resonance without increasing the
be concluded that the proposed design fulfills the gap between antenna’s physical size, the reflector is capacitively coupled to
achieving a reasonably wideband, unidirectional and yet very the driven bowtie to resonate at its patch mode. Furthermore,
compact antenna that can be used for many short-range series capacitive loading is added to the bowtie, and the
applications operating in the UHF band. reflector is folded at both ends towards the bowtie, resulting in
antenna miniaturization. The antenna has a compact size of
0.28λ × 0.23λ. It has a wide measured fractional bandwidth of
48% (0.69-1.12 GHz) with peak gain and FBR values of 5.5
dBi and 10 dB, respectively. Owing to its compact structure
(a) (b) (c) with a wide operating bandwidth and reasonably directional
Fig. 4. (a) Fabricated capacitively-loaded Yagi antenna. Feeding mechanism, radiation characteristics, it can be widely used for short-range
(b) front view, and (c) rear view.
wireless applications.
REFERENCES
[1] Y. Liu, Y. Chen, H. Lin and F. Juwono, "A novel differentially fed
compact dual-band implantable antenna for biotelemetry applications,"
IEEE Antennas Wireless Propag. Lett., vol. 15, pp. 1791-1794, 2016
[2] Z. G. Liu and Y. X. Guo, "Compact low-profile dual band metamaterial
antenna for body centric communications," IEEE Antennas Wireless
Propag. Lett., vol. 14, pp. 863-866, Dec. 2015.
[3] B. K. Tehrani, B. S. Cook and M. M. Tentzeris, "Inkjet Printing of
Multilayer Millimeter-Wave Yagi-Uda Antennas on Flexible Substrates,"
(a) IEEE Antennas Wireless Propag. Lett.,vol. 15, pp. 143-146, 2016.
[4] H. C. Huang, J. C. Lu and P. Hsu, "A Compact dual-band printed
Yagi-Uda antenna for GNSS and CMMB applications," IEEE Trans.
Antennas Propag., vol. 63, no. 5, pp. 2342-2348, May 2015.
[5] R. Bhattacharya, R. Garg and T. K. Bhattacharyya, "A compact Yagi-Uda
type pattern diversity antenna driven by CPW-fed Pseudomonopole,"
IEEE Trans. Antennas Propag., vol. 64, no. 1, pp. 25-32, Jan. 2016.
[6] J. Wu, Z. Zhao, Z. Nie and Q.-H. Liu, "Bandwidth enhancement of a
planar printed quasi-Yagi antenna with size reduction," IEEE Trans.
Antennas Propag., vol. 62, no. 1, pp. 463-467, Jan. 2014.
(b) [7] M.-Ch. Tang, T. Shi and R. W. Ziolkowski, "Flexible efficient quasi-Yagi
Fig. 5. Simulated vs. measured (a) reflection coefficient and (b) gain of the printed uniplanar antenna," IEEE Trans. Antennas Propag., vol. 63,
capacitively-loaded Yagi antenna. no. 12, pp. 5343-5350, Dec. 2015.
[8] Y. Luo and Q. X. Chu, "A Yagi-Uda antenna with a stepped-width
reflector shorter than the driven element," IEEE Antennas Wireless
Propag. Lett., vol. 15, pp. 564-567, 2016.
[9] B. K. Tan, S. Withington and G. Yassin, "A compact microstrip-fed
planar dual-dipole antenna for broadband applications," IEEE Antennas
Wireless Propag. Lett., vol. 15, pp. 593-596, 2016.
[10] J. A. Tirado-Mendez, H. Jardon-Aguilar, R. Flores-Leal, M. Reyes-Ayala
and F. Iturbide-Sanchez, "Inductively-loaded Yagi-Uda antenna with
cylindrical cover for size reduction at VHF-UHF bands," IEEE Trans
Antennas Prop, vol. 59, no. 2, pp. 357-362, Feb. 2011.
[11] S.A. Rezaeieh, A. Zamani, K.S. Bialkowski, A. Mahmoud and A.M.
Abbosh, “Feasibility of using wideband microwave system for
non-invasive detection and monitoring of pulmonary oedema,” Sci. Rep.,
vol. 5, no. 14047, 2015.
[12] S. W. Qu, J. L. Li and Q. Xue, "Bowtie dipole antenna with wide
beamwidth for base station application," IEEE Antennas Wireless
(a) (b) (c) Propag. Lett., vol. 6, pp. 293-295, 2007.
Fig. 6. Measured vs. simulated radiation patterns of the antenna in the x-y plane [13] S. A. Schelkunoff, Electromagnetic Waves. New York: Van Nostrand,
(top) and x-z plane (bottom) at (a) 0.75 GHz, (b) 0.85 GHz and (c) 1 GHz. Ch. 11, 1943.
Measured data are presented by dashed lines whereas simulated data are [14] S. Ahdi Rezaeieh, K. S. Bialkowski, A. Zamani and A. M. Abbosh,
denoted by solid lines. The outer curves represent co-polarization and the "Loop-dipole composite antenna for wideband microwave-based medical
curves at the center represent cross-polarization. diagnostic systems with verification on pulmonary edema detection,"
IEEE Antennas Wireless Propag. Lett., vol. 15, pp. 838-841, Dec. 2016.

1536-1225 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
On the Design of Millimetre-Wave Antennas for 50
Menna El Shorbagyl, Raed M. Shubair2 ,3, Mohamed I. AIHajri3 , Nazih Khaddaj Mallat4
1 Information Engineering Technology Department, German University in Cairo, Egypt
2 Electrical & Computer Engineering Department, Khalifa University, UAE
3 Research Laboratory of Electronics, Massachusetts Institute of Technology, USA
4 Al Ain University of Science and Technology, UAE
Email: rshubair@kustar.ac.ae; rshubair@mit.edu

Abstract-This paper addresses the millimeter-wave antenna
design aspect of the future 5G wireless systems. The paper
reviews the objectives and requirements of millimeter-wave
antennas for 5G. Recent advances in mm-wave antenna are
reported and design guidelines are discussed. In particular, four
different designs are identified from the recent literature based
on their attractive characteristics that support 5G requirements
and applications. The first design employs a dual-band slotted
patch antenna operating at 28 GHz and 38 GHz. The antenna has
circular polarization and is excited by a single-feed microstrip
line. The present design is desirable for high-gain antenna array
implementation in the mm-wave band, in order to compensate
for the mm-wave propagation loss. The second design that is
presented employs a compact planar inverted-F antenna (PIFA)
with single layer dielectric load of a superstrate to enhance
the gain and achieve a wide impedance bandwidth resulting in
high efficiency. The third design that operates in the mm-wave
band is a T-Shaped patch antenna. The proposed antenna a
wideband range from (26.5 GHz - 40 GHz) of the Ka band.
The PET substrate was used as it offers some advantages;
low cost, high flexibility, harmless to human body and resistive
towards environmental effects. The last mm-wave antenna design
presented employs two MIMO arrays each composed of 2x2
antenna elements. The two MIMO array configurations are
spatially orthogonal to each other which results in polarization
diversity.
Index Terms-Millimeter-wave antenna, 5G.
I. INTRODUCTION
There have been dynamic research activities around the
world in advancing the next-generation 5G wireless networks.
More than five billion devices demand wireless connections
that run voice, data, and other applications in todays wire-
less networks. The amount of mobile data has expanded
dramatically throughout the years due to the availability of
smart handheld devices, which support broadband wireless
applications such as multimedia and interactive gaming. The
efficient deployment of the 5G systems requires the design
of compact yet efficient antennas. There has been much
interest within the antenna research community to develop
efficient antenna designs for the future 5G specifically design
to operate in two 5G-frequency bands: 28 GHz band and
38 GHz band. Antenna design for mm-wave 5G must take
into account the high propagation loss due to atmospheric
absorption at millimetre waves [1]. Moreover, mm-wave an-
tennas must be compact in size to fit into handheld devices.
978-1-5090-2586-2/16/$31.00 @2016 IEEE
Hence, various miniaturization techniques must be employed
including slotting, stacking, grounding, and material properties
[2],[3],[4],[5]. This paper describes and compares four antenna
designs in recent literature for mm-wave 5G applications.
II. 5G TECHNOLOGY
Up till now 5G standards are not available for us. However,
some researchers have started to put the base for the technol-
ogy that will provide these standards. This technology mostly
consists of wireless access systems, frequency utilization,
power consumption, antenna and propagation [6].
A. 5G Spectrum
There is an enormous expanse in the millimeter wave
spectrum, more specifically 28 GHz and beyond that is largely
overlooked until now. On October 22nd 2015, FCC proposed
new rules (FCC 15138) for wireless broadband frequencies of
28 GHz, 37 GHz, 39 GHz and 64 - 71 GHz bands. Researchers
are targeting these frequencies for 5G applications. Internet
of Things (loT). loT is a system of physical items to gather
and exchange information. For a more direct incorporation
between the physical world and computer-based frameworks,
loT enables these devices to be detected and controlled re-
motely to enhance the efficiency. The terms loT and 5G are
used conversely because of the shared traits in the intents of
both.
B. 5G Objectives
The fundamental goals of 5G will be focused on towards
enhancing the limit of the systems with better coverage at a
lower cost. With a more prominent push for greener advances
every year, diminished force utilization will likewise be an
essential goal of 5G. The most essential and exceptionally
basic target of all is the "capacity" as it straightforwardly
identifies with the developing client interest for speedier and
higher information rates. The innovation is being researched to
meet these high information rate targets is the massive MIMO.
III. DUAL-BAND MM-WAVE ANTENNA
A dual-band slotted patch antenna shown in Fig. 1 was
presented in [7] which operates at 28 GHz and 38 GHz. The
antenna has circular polarization and is excited by a single-
feed microstrip line. The presented design is desirable for
 TABLE I
DESIGN PARAMETERS OF PROPOSED ANTENNA (MM)

Ll L2 tl t2 t3
3.1 2.5 0.1 0.7 0.4
t4 ts LJ WJ DJ
0.4 0.5 1.5 0.2 0.9

high-gain antenna array implementation in the mm-wave band,

in order to compensate for the mm-wave propagation loss. As
shown in Fig. 1, the antenna design is based on a squared
patch that comprises L-shaped slots close to the patch edges.
This arrangement adds capacitive and inductive effects, which
result in desirable two distinct resonant mm-wave frequencies.
The antenna was fabricated using optical photolithography on Fig. 2. Measured and simulated axial ratio of the CP patch antenna at the
bore sight
an RT/Duroid 5880 with a thickness of 254 um, dielectric
constant of 2.2, and loss tangent of 0.0009. The optimized
parameters for the proposed antenna are shown in Table. 1.
The antenna has been prototyped and the measured results
show that the antenna can simultaneously provide circular
polarization in the two bands (28 GHz and 38 GHz). The
left hand circular polarization gain is 15 dB less than the right
hand circular polarization gain in the bore sight direction; the
proposed antenna operates at right hand circular polarization
in two bands [7]. The measured and simulated axial ratio is
shown in Fig. 2 while the measured reflection coefficient is
shown in Fig. 3.

Fig. 3. Measured reflection coefficient of the dual-band antenna

thickness of 0.035mm. The proposed antenna intended to

Fig. 1. Geometry (top and side view) and main geometrical parameters of
the proposed mm-Wave patch antenna
IV. COMPACT PLANAR INVERTED-F ANTENNA
Another successful application that utilizes mm-wave an-
tennas was presented in [8]. The proposed design is based on
a compact planar inverted-F antenna (PIFA) with single layer
dielectric load of a superstrate to enhance the gain and achieve
a wide impedance bandwidth which has the advantage of high
efficiency was used in [8]. It covers the 28 GHz millimeter
wave frequency bands with size of 15mm x 15mm. The
antenna, which is shown in Fig. 4, is printed on a Rogers5870
substrate with a thickness of 0.635mm and copper-cladding
be integrated with the front end of a mobile phone using a
GSG probe or a 1.85 mm K-connector. Antenna dimensions
are optimized using parametric analyses in CST Microwave
Studio. The antenna has an impedance-matched bandwidth of
980 MHz (27.47 - 28.45 GHz) with the presence of a battery,
as shown in Fig. 5. The obtained minimum and maximum
value of the radiation efficiency, with the presence of the
battery, within the bandwidth of the antenna are 97% and
99%, at 27.47 GHz and 28.45 GHz, respectively, and the total
efficiencies are 88% and 96%, at 27.47 GHz and 28 GHz,
respectively. Both efficiencies of the antenna are unaffected
by the presence of the battery, while maximum and minimum
realized gains are 8.8 dBi and 8.25 dBi, at 28 GHz and 28.5
GHz, respectively.
V. T-SHAPED PATCH ANTENNA
The third design that operates in the mm-wave band is a
T-Shaped patch antenna that was presented in [9] as shown
in Fig. 6. The proposed antenna a wideband range from (26.5
GHz - 40 GHz) of the Ka band. The PET substrate was used as
 Fig. 6. Prototype of proposed printed antenna on PET film: (a) simulated
prototype; (b) fabricated prototype; (c) microscopic image of fabricated
prototype by DMP

Fig. 4. The top view of the radiating element without dielectric load

Fig. 7. 18111 plot of the proposed printed PET film antenna

of two x two antenna elements are used. Fig. 8 shows the two
antenna arrays embedded inside a mobile handset.
Fig. 9 shows the reflection coefficient of the patch antenna
designed for the 29 GHz 5G frequency band. The array
Fig. 5. Input reflection coefficient of the proposed antenna configurations are orthogonal to each other. This orthogonal
orientation leads to spatial diversity which in turn results in
polarization diversity. The ability to place multiple antenna
it offers some advantages; low cost, high flexibility, harmless arrays inside the mobile handset also gives the flexibility of
to human body and resistive towards environmental effects. operating in several configurations, like, exciting each arrays
The T-shaped antenna has a length of 4.7mm and is printed at a time or exciting both the arrays at the same time [6].
on 16mm x 16mm PET substrate. Feeding is achieved using
a coplanar waveguide and aperture in the ground plane. Table
2. provides the details of the optimized dimensions of the
TABLE II
proposed antenna. Fig. 7 shows the simulated and measured OPTIMIZED DIMENSIONS OF PROPOSED PRINTED ANTENNA
S11 plots of the proposed design, taking into account the effect
Dimensions of the proposed antenna
of fabric and human body on the measurements. The peak gain
Parameters Dimensions
of antenna is 8.2dBi at 38.75GHz and the gain is above 4dBi W Width of slot of ground plane 12.8 mm
in the complete range of operation L Length of slot of ground plane 7.5 mm
Wp Width of radiating patch 12 mm
VI. MIMO ANTENNA ARRAY FOR MM-WAVE Lp Length of radiating patch 4.7 mm
G Gap between patch and ground 2.4 mm
The authors in [6] presented an antenna designs that high- Lg Length of CPW ground 6.3 mm
lights the concept of massive MIMO which utilizes antenna Ls Length of the slot arrangement 2.9 mm
arrays and beamforming. Two MIMO arrays each composed Gs Gap between two symmetrical slot arrangements 5.2 mm
 VII. CHALLENGES OF MM- WAVE ANTENNA
DESIGN FOR 5G
Generating and receiving millimetre waves is a challenge,
but the biggest and most challenging factor with these high
frequencies is the traveling media. The biggest challenges
are atmospheric and free-space path loss. To combat severe
propagation loss, directional antennas are employed at both
transmitter and receiver to achieve a high antenna gain [10].
[10] Y. Niu, Y. Li, D. Jin, L. Su, and A. V. Vasilakos, ''A survey of millimeter
[5] M. I. Hussein, A. Hakam, M. Ouda, and R. M. Shubair, "Compact low-
profile planar elliptical antenna for UWB applications," in 2016 10th
European Conference on Antennas and Propagation (EuCAP), pp. 1-2.
[6] G. Gampala and C. J. Reddy, "Design of millimeter wave antenna arrays
for 5g cellular applications using FEKO," in 2016 IEEEIACES Inter-
national Conference on Wireless Information Technology and Systems
(ICWITS) and Applied Computational Electromagnetics (ACES), pp. 1-
2.
[7] H. Aliakbari, A. Abdipour, R. Mirzavand, A. Costanzo, and P. Mousavi,
''A single feed dual-band circularly polarized millimeter-wave antenna
for 5g communication," in 2016 10th European Conference on Antennas
and Propagation (Eu CAP), pp. 1-5.
[8] K. M. Morshed, K. P. Esselle, and M. Heimlich, "Dielectric loaded
planar inverted-f antenna for millimeter-wave 5g hand held devices,"
in 2016 10th European Conference on Antennas and Propagation
(EuCAP), pp. 1-3.
[9] S. F. Jilani and A. Alomainy, "Planar millimeter-wave antenna on low-
cost flexible PET substrate for 5g applications," in 2016 10th European
Conference on Antennas and Propagation (Eu CAP), pp. 1-3.
wave (mmWave) communications for 5g: Opportunities and challenges."

Fig. 8. Mobile handset with two 2x2 antenna arrays

Fig. 9. Patch antenna operating at 29 GHz

VIII. CONCLUSION

This paper focused on the design of mID-wave antennas

for 5G wireless systems. The objectives and requirements of
mID-wave antennas for 5G were reviewed. Recent advances in
mID-wave antenna design were reported and design guidelines
were discussed. In particular, four different designs that were
reported recently in the literature have been identified based
on their attractive characteristics that support 5G requirements
and applications.

REFERENCES

[1] R. M. Shubair, A. M. AIShamsi, K. Khalaf, and A. Kiourti, "Novel

miniature wearable microstrip antennas for ISM-band biomedical
telemetry," in Antennas Propagation Conference (!APC), 2015 Lough-
borough, pp. 1--4.
[2] A. Omar and R. Shubair, "UWB coplanar waveguide-fed-coplanar strips
spiral antenna," in 2016 10th European Conference on Antennas and
Propagation (EuCAP), pp. 1-2.
[3] M. Y. EISalamouny and R. M. Shubair, "Novel design of compact
low-profile multi-band microstrip antennas for medical applications," in
Antennas Propagation Conference (!APC), 2015 Loughborough, pp. 1-
4.
[4] A. Hakam, M. I. Hussein, M. Ouda, R. Shubair, and E. Serria, "Novel
circular antenna with elliptical rings for ultra-wide-band," in 2016 10th
European Conference on Antennas and Propagation (EuCAP), pp. 1--4.
1658 IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 12, 2013

Single-Feed Slotted Bowtie Antenna

for Triband Applications
HaiWen Liu, Member, IEEE, Hao Jiang, Xuehui Guan, JiuHuai Lei, and Shen Li

Abstract—In this letter, a triband bowtie antenna for

3.5/4.5/5.8-GHz applications using slot technique is presented. The
proposed microstrip-fed antenna forms operating frequencies by
inserting two pairs of slots with different lengths on the isosceles
triangle microstrip patch without increasing the overall antenna
area. The size of the proposed antenna is determined by the
middle resonant frequency , and thus it is compact in nature.
The geometry of the bowtie antenna is symmetrical to the direction
of the feedline. The proposed antenna is designed, analyzed, and
verified by simulations and measurements.
Index Terms—Antenna, bowtie, microstrip-fed, multifrequency,
slot, triband.

I. INTRODUCTION

W ITH the development of wireless communication sys-

tems, electronic devices often require several antennas
to handle a variety of signals. The growth of the information in-
dustry has resulted in the miniaturization of the electronic prod-
ucts, so that an antenna with multifrequency operation capa-
bility, compact size, low cost, and high efficiency has become
an important component for the integration wireless communi-
cation systems [1]–[3].
An antenna with multiple functions can simplify the com-
munication system. It is well known that the microstrip bowtie
patch antenna has excellent characteristics, such as low cost,
light weight, easy fabrication, and ability to print on the same
board as transceiver or receiver [4]. Therefore, to achieve
the above factors, the microstrip patch bowtie antenna is a
good candidate. The main methods achieving multifrequency Fig. 1. (a) Geometry of the single-band bowtie antenna without slot (proposed
parameters: mm, mm, mm, mm).
performances are as follows: 1) using several different resonant (b) Its simulated reflection coefficient .
modes of a single microstrip patch; 2) changing surface instan-
taneous current distributions of resonance modes by loading
or etching slot on a single patch [5]–[7]; 3) utilizing multiple with wider tuning range. Its size is relatively large since it
microstrip patches on the single-layer substrate. In [8], a novel is mainly determined by its lower operating frequency. In
design for the modified bowtie slot antenna with a rectangular [10], a coplanar waveguide (CPW)-fed bowtie slot antenna is
tuning stub for 2.4/5.2/5.8-GHz triple-band wireless applica- proposed to be capable of triband operation with satisfactorily
tions is presented. In [9], the proposed slot dipole antenna like low cross-polarization levels within each band.
other multifrequency antennas is capable of triband operation In this letter, a microstrip-fed bowtie antenna is proposed
to be capable of triband operation frequencies for WiMAX
(3.47–3.51 GHz), INSAT (4.54–4.60 GHz), and WLAN
Manuscript received November 01, 2013; accepted December 06, 2013. Date
of publication December 11, 2013; date of current version January 17, 2014. (5.75–5.81 GHz), independent of allocating different operating
This work was supported by the National Science Foundation of China under frequencies. Integrating multiple functions on an antenna can
Grant No. 61061001 and.the International Cooperation Funds and Science and
reduce the antenna’s size and weight and increase the elec-
Technology Innovation Team of Jiangxi Province of China under Grants No.
20121BDH80015 and No. 20122BCB24025. trical properties of the antenna. In order to simplify the feed
The authors are with East China of Jiaotong University, Nanchang 330013, networks, the geometry of the bowtie antenna is symmetrical
China (e-mail: liuhaiwen@gmail.com).
to the direction of the feedline. Details of the antenna design
Color versions of one or more of the figures in this letter are available online
at http://ieeexplore.ieee.org. are presented. Simulation and measurement results are also
Digital Object Identifier 10.1109/LAWP.2013.2294751 presented and discussed.

1536-1225 © 2013 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.
LIU et al.: SINGLE-FEED SLOTTED BOWTIE ANTENNA FOR TRIBAND APPLICATIONS 1659

Fig. 2. (a) Geometry of the proposed dual-band bowtie antenna with a slot
(proposed parameters: mm, mm, mm). (b) Its
simulated reflection coefficient .

Fig. 4. Effect of varying slot parameters on reflection coefficient . (a) Slot

length . (b) Slot length . (c) Microstrip patch length .

chosen as a reference antenna and is symmetrical to the -direc-

tion along the center plane of the feedline. The antenna is fabri-
cated on a Taconic RF-35A2 substrate with a relative dielectric
constant , loss tangent , and thickness
mm. Reflection coefficient of the single-band
Fig. 3. Geometry of the proposed triband bowtie antenna with two slots (pro-
posed parameters: mm, mm, mm). antenna is simulated by Ansoft HFSS ver. 10. Its dimension pa-
rameters and results are given in Fig. 1(b). Simulated impedance
bandwidth dB of the frequency band for the
II. ANTENNA STRUCTURE AND DESIGN single-band antenna is 4.62–4.69 GHz (1.5%).
It is well known that slot perturbation on the patch resonator
A single-band bowtie antenna without any slot on the patch can be used to change the surface current distributions to form
resonator is shown in Fig. 1(a), where a pair of isosceles triangle multiband performances [11], [12]. Thus, one slot is etched on
microstrip patches fed by a microstrip line with input impedance the isosceles triangle microstrip patch to excite extra resonant
of 50- is depicted. The microstrip patch antenna has a length frequency at GHz. The slot has a length of with
of with width . The feedline has a length of and width width , and the distance between the slot and the middle
. The ground plane is located on the reverse side of the sub- plane of the feedline is indicated as . Configuration of the
strate in the shape of a rectangle, covering the entire back. It is proposed dual-band bowtie antenna with one slot is presented
1660 IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 12, 2013

Fig. 6. Photograph of the fabricated triband bowtie antenna with two slots.

Fig. 7. Reflection coefficients of the fabricated triband bowtie antenna.

The simulated reflection coefficients with varied slot

Fig. 5. Simulated surface current distributions at (a) , (b) , and (c)
in Fig. 2(a), and its simulated reflection coefficient is
shown in Fig. 2(b). The simulated impedance bandwidths
dB of the frequency bands for the dual-band an-
tenna are 3.57–3.61 GHz (1.11%) and 4.74–4.81 GHz (1.46%),
respectively.
Similarly, the other slot as a perturbation segment changes the
surface current distributions to form the third frequency band.
Hence, a shorter slot with a length of and width is
etched on the microstrip patch to excite another extra resonant
frequency at , resulting in triband performances. The dis-
tance between the slot and the middle plane of the feedline is
indicated as . Geometry of the proposed triband antenna with
two slots is shown in Fig. 3.
It is concluded that the size of the proposed antenna is de-
termined by the resonant frequency of the patch resonator,
which also forms the middle frequency band. The upper oper-
ating frequencies and the lower operating frequencies
are obtained by inserting two pairs of slots with different lengths
on the isosceles triangle microstrip patch without increasing the
overall antenna area.
lengths and isosceles triangle microstrip patch length
are shown in Fig. 4. It is clear from the figures that when the
slot length (or ) is increased, the operating frequency
. (or ) shifts to a lower frequency. The bandwidth is almost un-
affected by changing the slot length. Therefore, the slot lengths
( and ) can be used as a frequency tuner for the oper-
ating frequency ( and ). Meanwhile, the isosceles triangle
microstrip patch length is increased so that the operating fre-
quencies shift to the lower frequencies.
Fig. 5 illustrates the surface current distributions of the pro-
posed triband antenna at different resonant frequencies. Ana-
lyzing surface current distributions can confirm the resonance
modes at the different operating frequencies. In Fig. 5(b), at the
middle resonant frequency , it is noted that the surface current
is almost distributed on the overall isosceles triangle microstrip
patch. It justifies that the middle frequency is generated by the
overall isosceles triangle microstrip patch. Meanwhile, the sur-
face current at the lower frequency is mainly distributed at
four base angles of the isosceles triangle microstrip patch, as
shown in Fig. 5(a). Therefore, the long slot mainly generates
the lower frequency. Similarly in Fig. 5(c), the surface current
at the upper frequency centralizes in the region nearby the
middle and two angles of the isosceles triangle microstrip patch,
so the short slot generates the upper frequency. According to the
above discussion, two pairs of slots are used to change surface
current distributions of resonance modes to achieve triband.
LIU et al.: SINGLE-FEED SLOTTED BOWTIE ANTENNA FOR TRIBAND APPLICATIONS
Fig. 8. Radiation patterns of the proposed antenna at different operating
frequencies. Black solid line: simulated E-plane copolarization. Magenta solid
line: simulated H-plane copolarization. Black short-dash line: measured E-plane
copolarization. Magenta short-dash line: measured H-plane copolarization.
Black short-dash-dot line: E-plane cross polarization. Magenta short-dash-dot
line: H-plane cross polarization. (a) GHz. (b)
(c) GHz.
III. RESULTS
The proposed triband bowtie antenna is fabricated by
adopting two pairs of slots with different lengths. Its photo-
graph is given in Fig. 6. The measured and simulated reflection
coefficients of the prototype antenna are plotted in Fig. 7.
The measured bandwidths dB of the lower,
middle, and upper frequency bands for the prototype antenna
are 1.14% (3.47–3.51 GHz), 2.2% (4.5–4.60 GHz), and 1.03%
(5.75–5.81 GHz), respectively. Measured results agree well
with the simulated ones, proving the validity of the introduced
design principles.
The measured and simulated radiation patterns of the pro-
posed antenna in - (E-) and - (H-) planes are illustrated in
1661
Fig. 8. It is obvious that good broadside radiation is obtained and
the three frequencies almost have the same polarization planes.
The cross-polarization levels in E-plane are less than 20 dB,
and relatively smaller cross-polarization levels in H-plane are
observed. Thus, the cross-polarization levels can be acceptable
in the principle planes. Moreover, the obtained gains of the an-
tenna at 3.5, 4.5, and 5.8 GHz are about 1.1, 3.34, and 5.1 dBi,
respectively.
IV. CONCLUSION
The compact triband bowtie antenna fed by a microstrip line
has been presented with satisfactory characteristics and nearly
independent allocation of the three operating frequencies. Two
pairs of slots are etched on the isosceles triangle microstrip
patch resonator to change the surface current distributions re-
sulting in triband without increasing the antenna size, thus it is
compact in nature. The antenna geometry is simple and easy to
fabricate and implement. Consequently, the proposed antenna is
suitable for multifrequency applications of wireless communi-
cation systems.
REFERENCES
[1] A. M. Montaser, K. R. Mahmoud, and H. A. , “Tri-band slotted bow-tie
antenna design for RFID reader using hybrid CFO-NM algorithm,” in
Proc. 29th Nat. Radio Sci. Conf., 2012, pp. 119–126.
[2] Y. Lin and S. Tsai, “Coplanar waveguide-fed uniplanar bow-tie an-
tenna,” IEEE Trans. Antennas Propag., vol. 45, no. 2, pp. 305–306,
Feb. 1997.
[3] C. M. Li, K. Wang, and C. K. Chen, “Small tri-band monopole antenna
for WiMAX/WLAN applications,” J. Electromagn. Waves Appl., vol.
25, pp. 1297–1307, 2011.
[4] S. Hong and K. Chang, “Single-feed triple-frequency rectangular mi-
crostrip patch antenna with pairs of spur-lines,” Electron. Lett., vol. 42,
no. 12, pp. 673–674, Jun. 2006.
[5] J.-S. Chen, “Dual-frequency annular-ring slot antennas fed by CPW
feed and microstrip line feed,” IEEE Trans. Antennas Propag., vol. 53,
no. 1, pp. 569–571, Jan. 2005.
[6] W.-S. Chen and K.-L. Wong, “A coplanar waveguide-fed printed slot
GHz. antenna for dual-frequency operation,” in Proc. IEEE AP-S Int. Symp.,
Jul. 2001, vol. 2, pp. 140–143.
[7] J.-S. Chen, “Triple-frequency annular-ring slot antennas fed by CPW
and microstrip line,” in Proc. IEEE AP-S Int. Symp., Jun. 2003, vol. 2,
pp. 557–560.
[8] J. H. Yoon and Y. C. Lee, “Modified bow-tie slot antenna for the
2.4/5.2/5.8 GHz WLAN bands with a rectangular tuning stub,”
Microw. Opt. Technol. Lett., vol. 52, no. 1, pp. 126–130, Jan. 2011.
[9] S.-Y. Chen, Y.-C. Chen, and P. Hsu, “CPW-fed aperture-coupled
slot dipole antenna for tri-band operation,” IEEE Antennas Wireless
Propog. Lett., vol. 7, pp. 535–537, 2008.
[10] Y.-C. Chen, S.-Y. Chen, and P. Hsu, “A compact triband bow-tie
slot antenna fed by a coplanar waveguide,” IEEE Antennas Wireless
Propog. Lett., vol. 9, pp. 1205–1208, 2010.
[11] Y.-Y. Chen, Y.-C. Jiao, G. Zhao, F. Zhang, Z.-L. Liao, and Y. Tian,
“Dual-band dual-sense circularly polarized slot antenna with a
C-shaped grounded strip,” IEEE Antennas Wireless Propog. Lett., vol.
10, pp. 915–918, 2011.
[12] M. S. Ghaffarian and G. Moradi, “A novel harmonic suppressed
coplanar waveguide (CPW)-fed slot antenna,” IEEE Antennas Wire-
less Propog. Lett., vol. 10, pp. 788–791, 2011.
 IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
What Will 5G Be?
Jeffrey G. Andrews, Fellow, IEEE, Stefano Buzzi, Senior Member, IEEE, Wan Choi, Senior Member, IEEE,
Stephen Hanly, Member, IEEE, Angel Lozano, Fellow, IEEE, Anthony C.K. Soong, Fellow, IEEE,
Jianzhong Charlie Zhang, Senior Member, IEEE
Abstract—What will 5G be? What it will not be is an in-
cremental advance on 4G. The previous four generations of
cellular technology have each been a major paradigm shift
that has broken backwards compatibility. And indeed, 5G will
need to be a paradigm shift that includes very high carrier
frequencies with massive bandwidths, extreme base station and
device densities and unprecedented numbers of antennas. But
arXiv:1405.2957v1 [cs.IT] 12 May 2014
unlike the previous four generations, it will also be highly
integrative: tying any new 5G air interface and spectrum together
with LTE and WiFi to provide universal high-rate coverage and
a seamless user experience. To support this, the core network
will also have to reach unprecedented levels of flexibility and
intelligence, spectrum regulation will need to be rethought and
improved, and energy and cost efficiencies will become even more
critical considerations. This paper discusses all of these topics,
identifying key challenges for future research and preliminary
5G standardization activities, while providing a comprehensive
overview of the current literature, and in particular of the papers
appearing in this special issue.
I. I NTRODUCTION
A. The Road to 5G
In just the past year, preliminary interest and discussions
about a possible 5G standard have evolved into a full-fledged
conversation that has captured the attention and imagination
of researchers and engineers around the world. As the long-
term evolution (LTE) system embodying 4G has now been
deployed and is reaching maturity, where only incremental
improvements and small amounts of new spectrum can be
expected, it is natural for researchers to ponder “what’s next?”
[1]. However, this is not a mere intellectual exercise. Thanks
largely the annual visual network index (VNI) reports released
by Cisco, we have quantitative evidence that the wireless
data explosion is real and will continue. Driven largely by
smartphones, tablets, and video streaming, the most recent
(Feb. 2014) VNI report [2] and forecast makes plain that
an incremental approach will not come close to meeting the
demands that networks will face by 2020.
In just a decade, the amount of IP data handled by wireless
networks will have increased by well over a factor of 100:
from under 3 exabytes in 2010 to over 190 exabytes by 2018,
on pace to exceed 500 exabytes by 2020. This deluge of data
J. G. Andrews (jandrews@ece.utexas.edu) is with the University of Texas
at Austin, USA.
S. Buzzi (buzzi@unicas.it) is with University of Cassino and Southern
Lazio, Italy, and with CNIT, Italy.
W. Choi (wchoi@kaist.edu) is with Korea Advanced Institute of Science
and Technology (KAIST), Daejeon, South Korea.
S. Hanly (stephen.hanly@mq.edu.au) is with Macquarie University, Sydney,
Australia.
A. Lozano (angel.lozano@upf.edu) is with Universitat Pompeu Fabra
(UPF), Barcelona, Spain.
A. C. K. Soong (anthony.soong@huawei.com) is with Huawei Technologies,
Plano, Texas, USA.
J. C. Zhang (jianzhong.z@samsung.com) is with Samsung Electronics,
Richardson, Texas, USA.
Article last revised: May 14, 2014
1
has been driven chiefly by video thus far, but new unforeseen
applications can reasonably be expected to materialize by
2020. In addition to the sheer volume of data, the number of
devices and the data rates will continue to grow exponentially.
The number of devices could reach the tens or even hundreds
of billions by the time 5G comes to fruition, due to many new
applications beyond personal communications [3]–[5]. It is our
duty as engineers to meet these intense demands via innovative
new technologies that are smart and efficient yet grounded in
reality. Academia is engaging in large collaborative projects
such as METIS [6] and 5GNOW [7], while the industry
is driving preliminary 5G standardization activities (cf. Sec.
IV-B). To further strengthen these activities, the public-private
partnership for 5G infrastructure recently constituted in Europe
will funnel massive amounts of funds into related research [8].
This article is an attempt to summarize and overview many
of these exciting developments, including the papers in this
special issue. In addition to the highly visible demand for
ever more network capacity, there are a number of other
factors that make 5G interesting, including the potentially
disruptive move to millimeter wave (mmWave) spectrum, new
market-driven ways of allocating and re-allocating bandwidth,
a major ongoing virtualization in the core network that might
progressively spread to the edges, the possibility of an “Internet
of Things” comprised of billions of miscellaneous devices,
and the increasing integration of past and current cellular and
WiFi standards to provide an ubiquitous high-rate, low-latency
experience for network users.
This editorial commences with our view of the “big three”
5G technologies: ultra-densification, mmWave, and massive
multiple-input multiple-output (MIMO). Then, we consider
important issues concerning the basic transmission waveform,
the increasing virtualization of the network infrastructure, and
the need for greatly increased energy efficiency. Finally, we
provide a comprehensive discussion of the equally important
regulatory and standardization issues that will need to be
addressed for 5G, with a particular focus on needed innovation
in spectrum regulation.
B. Engineering Requirements for 5G
In order to more concretely understand the engineering chal-
lenges facing 5G, and to plan to meet them, it is necessary to
first identify the requirements for a 5G system. The following
items are requirements in each key dimension, but it should be
stressed that not all of these need to be satisfied simultaneously.
Different applications will place different requirements on
the performance, and peak requirements that will need to
be satisfied in certain configurations are mentioned below.
For example, very-high-rate applications such as streaming
high-definition video may have relaxed latency and reliability
requirements compared to driverless cars or public safety
applications, where latency and reliability are paramount but
lower data rates can be tolerated.
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
1) Data Rate: The need to support the mobile data traffic
explosion is unquestionably the main driver behind 5G. Data
rate can be measured in several different ways, and there will
be a 5G goal target for each such metric:
a) Aggregate data rate refers to the total amount of data
the network can serve, characterized in units of bits/s/area. The
general consensus is that this quantity will need to increase by
roughly 1000x from 4G to 5G.
b) Edge rate, or 5% rate, is the worst data rate that a
user can reasonably expect to receive when in range of the
network, and so is an important metric and has a concrete
engineering meaning. Goals for the 5G edge rate range from
100 Mbps (easily enough to support high-definition streaming)
to as much as 1 Gbps. Meeting 100 Mbps for 95% of users will
be extraordinarily challenging, even with major technological
advances. This requires about a 100x advance since current
4G systems have a typical 5% rate of about 1 Mbps, although
the precise number varies quite widely depending on the load,
cell size, and other factors.
c) Peak rate is the best-case data rate that a user can hope
to achieve under any conceivable network configuration. The
peak rate is a marketing number, devoid of much meaning to
engineers, but in any case it will likely be in the range of tens
of Gbps.
Meeting the requirements in (a)-(b), which are about 1000x
and 100x current 4G technology, respectively, are the main
focus of this paper.
2) Latency: Current 4G roundtrip latencies are on the order
of about 15 ms, and are based on the 1 ms subframe time
with necessary overheads for resource allocation and access.
Although this latency is sufficient for most current services,
anticipated 5G applications include two-way gaming, novel
cloud-based technologies such as those that may be touch-
screen activated (the “tactile Internet” [9]), and virtual and
enhanced reality (e.g., Google glass or other wearable comput-
ing devices). As a result, 5G will need to be able to support
a roundtrip latency of about 1 ms, an order of magnitude
faster than 4G. In addition to shrinking down the subframe
structure, such severe latency constraints may have important
implications on design choices at several layers of the protocol
stack and the core network (cf. Sect. III).
3) Energy and Cost: As we move to 5G, costs and energy
consumption will, ideally, decrease, but at least they should
not increase on a per-link basis. Since the per-link data rates
being offered will be increasing by about 100x, this means
that the Joules per bit and cost per bit will need to fall by at
least 100x. In this article, we do not address energy and cost
in a quantitative fashion, but we are intentionally advocating
technological solutions that promise reasonable cost and power
scaling. For example, mmWave spectrum should be 10-100x
cheaper per Hz than the 3G and 4G spectrum below 3 GHz.
Similarly, small cells should be 10-100x cheaper and more
power efficient than macrocells. A major cost consideration
for 5G, even more so than in 4G due to the new BS densities
and increased bandwidth, is the backhaul from the network
edges into the core. We address backhaul and other economic
considerations in Section IV-C. As for energy efficiency, we
address this more substantially in Section III-C.
C. Device Types and Quantities.
5G will need to be able to efficiently support a much larger
and more diverse set of devices. With the expected rise of
machine-to-machine communication, a single macrocell may
2
need to support 10,000 or more low-rate devices, along with its
traditional high-rate mobile users. This will require wholesale
changes to the control plane and network management relative
to 4G, whose overhead channels and state machines are not
designed for such a diverse and large subscriber base.
II. K EY T ECHNOLOGIES TO G ET TO 1000 X DATA R ATE
Of the requirements outlined in Sect. I-B, certainly the one
that gets the most attention is the need for radically higher data
rates across the board. Our view is that the required 1000x will,
for the most part, be achieved through combined gains in three
categories:
a) Extreme densification and offloading to improve the area
spectral efficiency. Put differently, more active nodes per
unit area and Hz.
b) Increased bandwidth, primarily by moving towards and
into mmWave spectrum but also by making better use
of WiFi’s unlicensed spectrum in the 5 GHz band.
Altogether, more Hz.
c) Increased spectral efficiency, primarily through advances
in MIMO, to support more bits/s/Hz per node.
The combination of more nodes per unit area and Hz, more
Hz, and more bits/s/Hz per node, will compound into many
more bits/s per unit area. Other ideas not in the above cate-
gories, e.g., interference management through BS cooperation
[10]–[23] may also contribute improvements, but the lion’s
share of the surge in capacity should come from ideas in the
above categories. In the remainder of this section, these are
distilled in some detail.
A. Extreme Densification and Offloading
A straightforward but extremely effective way to increase
the network capacity is to make the cells smaller. This ap-
proach has been demonstrated over several cellular generations
[24], [25]. The first such generation, in the early 1980s, had
cell sizes on the order of hundreds of square kms. Since then,
those sizes have been progressively shrinking and by now they
are often fractions of a square km in urban areas. In Japan,
for instance, the spacing between BSs can be as small as two
hundred meters, giving a coverage area well under a tenth of a
square km. Networks are now rapidly evolving [26] to include
nested small cells such as picocells (range under 100 meters)
and femtocells (WiFi-like range) [27], as well as distributed
antenna systems [28] that are functionally similar to picocells
from a capacity and coverage standpoint but have all their
baseband processing at a central site and share cell IDs.
Cell shrinking has numerous benefits, the most important
being the reuse of spectrum across a geographic area and
the ensuing reduction in the number of users competing for
resources at each BS. Contrary to widespread belief, as long
as power-law pathloss models hold the signal-to-interference
ratio (SIR) is preserved as the network densifies [29].1 Thus,
in principle, cells can shrunk almost indefinitely without a
sacrifice in SIR, until nearly every BS serves a single user
(or is idle). This allows each BS to devote its resources, as
well as its backhaul connection, to an ever-smaller number of
users.
As the densification becomes extreme, some challenges
arise:
1 The power-law pathloss model ceases to apply in the near field, very close
to the transmitter [30].
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
• Preserving the expected cell-splitting gains as each BS
becomes more lightly loaded, particularly low-power
nodes.
• Determining appropriate associations between users and
BSs across multiple radio access technologies (RATs),
which is crucial for optimizing the edge rate.
• Supporting mobility through such a highly heteroge-
neous network.
• Affording the rising costs of installation, maintenance
and backhaul.
We next briefly discuss these challenges, particularly in view
of the other technologies raised in this article.
1) Base Station Densification Gains: We define the BS
densification gain ρ > 0 as the effective increase in data rate
relative to the increase in network density, which is a proxy
here for cost. Specifically, if we achieve a data rate R1 (could
be any measure thereof, e.g., edge rate or aggregate) when the
BS density is λ1 BSs/km2 and then we consider a higher BS
density λ2 , with corresponding rate R2 , then the corresponding
densification gain is
R2 λ1
ρ= . (1)
R1 λ2
For example, if the network density is doubled, and the
aggregate data rate also doubles, then the densification gain is
ρ = 1: the increase in BS density has an exactly proportional
payoff in terms of achieved rates in this case.
In an interference-limited network with full buffers, the
signal-to-interference-plus-noise ratio (SINR) is essentially
equal to the SIR and, because the SIR distribution remains
approximately constant as the network densifies, the best case
scenario is ρ ≈ 1. In reality, buffers are not always full, and
small cells tend to become more lightly loaded than macrocells
as the network densifies. Therefore, the SINR usually increases
with density: in a noise-limited network by increasing the
received signal power, and in interference-limited networks
because the lightly loaded small cells generate less interfer-
ence (while still providing an option for connectivity) [31].
Nevertheless, at microwave frequencies the gain in SINR is not
enough to keep up with the decrease in small cell utilization,
and thus ρ < 1. In an extreme case, consider λ1 and R1 held
fixed with λ2 → ∞. In this asymptotic setting, the small cells
compete for a finite pool of UEs, becoming ever more lightly
loaded, and thus ρ → 0.
Empirically and theoretically, we also observe that ρ im-
proves and can approach 1 with macro-BS muting (eICIC in
3GPP) vs. the macrocells transmitting all the time and thus
interfering with the small cells all the time. This observation
is relevant because the result is not obvious given that the
macrocells are the network bottleneck.
An intriguing aspect of mmWave frequencies is that den-
sification gains ρ  1 may be possible. This is because, as
discussed in Section II-B, at these frequencies communication
is largely noise-limited and increasing the density not only
splits the cell resources and lightens the load, but it may
increase the SINR dramatically. As a striking example of this,
it was recently showed that, under a plausible urban grid-based
deployment, increasing the BS count in a given area from 36
to 96—which decreased the inter-BS distance from 170 meters
down to 85 meters—increased the 5% cell-edge rate from 24.5
Mbps up to 1396 Mbps, giving [32]
1396 · 36
ρ= = 21.3. (2)
24.5 · 96
3
Fig. 1: User association in a multi-RAT network over many frequency
bands is complex. In this simplified scenario, a mobile user in turn
associates with different BSs based on a tradeoff between the gain to
that BS and the traffic load (congestion) that it is experiencing.
While conceding that this massive densification gain corre-
sponds to a particular setup and model, it is nevertheless
remarkable.
In general, quantifying and optimizing the densification
gains in a wide variety of deployment scenarios and network
models is a key area for continued small cell research.
2) Multi-RAT Association: Networks will continue to be-
come increasingly heterogeneous as we move towards 5G.
A key feature therein will be increased integration between
different RATs, with a typical 5G-enabled device having radios
capable of supporting not only a potentially new 5G standard
(e.g., at mmWave frequencies), but also 3G, numerous releases
of 4G LTE including possibly LTE-Unlicensed [33], several
types of WiFi, and perhaps direct device-to-device (D2D)
communication, all across a great many spectral bands. Hence,
determining which standard(s) and spectrum to utilize and
which BS(s) or users to associate with will be a truly complex
task for the network [34].
Determining the optimal user association can be a massive
combinatorial optimization problem that depends on the SINR
from every user to every BS, the instantaneous load at each BS,
the choices of other users in the network, and possibly other
constraints such as the requirement to utilize the same BS and
standard in both uplink and downlink (to facilitate functioning
control channels for resource allocation and feedback) [35],
[36]. Therefore, simplified procedures must be adopted [37],
an example of which appears in this special issue [38]. Even
a simple, seemingly highly suboptimal association approach
based on aggressive but static biasing towards small cells and
blanking about half of the macrocell transmissions has been
shown to increase edge rates by as much as 500% [39].
The joint problem of user association and resource allocation
in two-tier heterogeneous networks (HetNets), with adaptive
tuning of the biasing and blanking in each cell, is considered
in [35], [36], [40]–[45]. An interesting model of hotspot traffic
is considered in [41]–[43] where it is shown that, under
various network utility metrics, the optimal cell association
is determined by rate ratio bias, rather than power level bias.
It will be interesting to extend these models to more general
scenarios, including more than two tiers. A dynamic model of
cell range expansion is considered in [46], where traffic arrives
as a Poisson process in time and the feasible arrival rates, for
which a stabilizing scheduling policy exists, are characterized.
User association and load balancing in a HetNet, with massive
MIMO at the BSs, is considered in [47]. The problem of
determining the optimal associations when there are multiple
RATS, operating at different frequencies and using different
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS 4

protocols, has not yet received much attention. However, an

interesting game theoretic approach is taken in [48] to the RAT-
selection problem, where convergence to Nash equilibria and
the Pareto-efficiency of these equilibria are studied. A related
paper in this special issue [49] explores the interaction between
cellular operators and WiFi network owners.
Adding mmWave into the picture adds significant additional
complexity, since even the notion of a cell boundary is blurry
at mmWave frequencies given the strong impact of blockages,
which often result in nearby BSs being bypassed in favor of
farther ones that are unblocked (cf. Fig. 2). On the positive
side, interference is much less important in mmWave (cf.
Section II-B) and thus the need for blanking is reduced.
In summary, there is a great deal of scope for modeling,
analyzing and optimizing BS-user associations in 5G.
3) Mobility Support: Clearly, the continued network densi-
fication and increased heterogeneity poses challenges for the
support of mobility. Although a hefty share of data is served
to stationary indoor users, the support of mobility and always-
on connectivity is arguably the single most important feature
of cellular networks relative to WiFi. Because modeling and
analyzing the effect of mobility on network performance is
difficult, we expect to see somewhat ad hoc solutions such as in
LTE Rel-11 [50] where user-specific virtual cells are defined to
distinguish the physical cell from a broader area where the user
can roam without the need for handoff, communicating with
any BS or subset of BSs in that area. Or in mmWave, restricting
highly mobile users to macrocells and microwave frequencies,
thereby forcing them to tolerate lower rates. Handoffs will be
particularly challenging at mmWave frequencies since transmit
and receive beams must be aligned to communicate. Indeed,
the entire paradigm of a handoff initiated and managed at
layer 3 by the core network will likely not exist in 5G;
instead, handoffs may be opportunistic, based on mmWave
beam alignments, or indistinguishable from PHY/MAC inter-
ference management techniques whereby users communicate
with multiple coordinated BSs, as exemplified by [51] in this Fig. 2: Calculated mmWave BS associations with real building
locations [57]. The shaded regions correspond to association with the
special issue.
BS centered at that shade. Blocking, LOS vs. non-LOS propagation,
4) Cost: Evolving to ever-smaller cells requires ever-
and beam directionality render our usual notion of cell boundaries
smaller, lower-power and cheaper BSs, and there is no fun- obsolete.
damental reason a BS needs to be more expensive than a
user device or a WiFi node [26]. Nevertheless, obtaining
permits, ensuring fast and reliable backhaul connections, and
paying large monthly site rental fees for operator-controlled quo is highly inefficient and the cost for 5G would be
small-cell placements have proven a major hindrance to the greatly reduced in an open-access paradigm for small cells.
growth of picocell, distributed antennas, and other enterprise- One preliminary but successful example is Fon, which as
quality small cell deployments. Of these, only the backhaul of press time boasts over 13 million shared WiFi access points.
is primarily a technical challenge. Regulatory reforms and
infrastructure sharing (cf. Section IV-C) may help address the 5G and all networks beyond it will be extremely dense
other challenges. and heterogeneous, which introduces many new challenges for
Turning to end-user-deployed femtocells and WiFi access network modeling, analysis, design and optimization. We fur-
points, these are certainly much more cost-effective both from ther discuss some of the nonobvious intersections of extreme
a capital and operating expense perspective [24]. However, densification with mmWave and massive MIMO, respectively,
major concerns exist here too. These include the coordination in the next two sections. Before proceeding, however, we
and management of the network to provide enterprise-grade briefly mention that besides cell shrinking a second approach
service, which given the scale of the deployments requires to densification exists in the form of direct D2D communica-
automated self-organization [52]. A further challenge is that tion. This allows users in close proximity to establish direct
these end-user deployments utilize the end-user’s backhaul communication, replacing two long hops via the BS with a
connection and access point, both of which the end-user has single shorter hop. Provided there is sufficient spatial locality
a vested interest in not sharing, and in some countries a legal in the wireless traffic, this can bring about reduced power
requirement not to. Anecdotally, all readers of this article consumption and/or higher data rates, and a diminished latency
are familiar with the scenario where a dozen WiFi access [53]–[55]. Reference [56] in this special issue proposes a novel
points are within range, but all are secured and inaccessible. way of scheduling concurrent D2D transmissions so as to
From an engineering perspective, this closed-access status densify while offering interference protection guarantees.
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
B. Millimeter Wave
Terrestrial wireless communication systems have largely
restricted their operation to the relatively slim range of mi-
crowave frequencies that extends from several hundred MHz
to a few GHz and corresponds to wavelengths in the range
of a few centimeters up to about a meter. By now though,
this spectral band—often called “beachfront spectrum”—has
become nearly fully occupied, in particular at peak times and
in peak markets. Regardless of the efficacy of densification
and offloading, much more bandwidth is needed [58], [59].
Although beachfront bandwidth allocations can be made
significantly more efficient by modernizing regulatory and
allocation procedures, as discussed in Section IV-A, to put
large amounts of new bandwidth into play there is only one
way to go: up in frequency. Fortunately, vast amounts of
relatively idle spectrum do exist in the mmWave range of 30–
300 GHz, where wavelengths are 1–10 mm. There are also
several GHz of plausible spectrum in the 20–30 GHz range.
The main reason that mmWave spectrum lies idle is that,
until recently, it had been deemed unsuitable for mobile com-
munications because of rather hostile propagation qualities,
including strong pathloss, atmospheric and rain absorption, low
diffraction around obstacles and penetration through objects,
and, further, because of strong phase noise and exorbitant
equipment costs. The dominant perception had therefore been
that such frequencies, and in particular the large unlicensed
band around 60 GHz [60], were suitable mainly for very-short-
range transmission [61]–[63]. Thus, the focus had been on
WiFi (with the WiGiG standard in the 60-GHz band) and also
on fixed-wireless applications in the 28, 38, 71–76 and 81–
86 GHz. However, semiconductors are maturing, their costs
and power consumption rapidly falling—largely thanks to the
progress of the aforementioned short-range standards—and
the other obstacles related to propagation are now considered
increasingly surmountable given time and focused effort [64]–
[69].
1) Propagation Issues: Concerning mmWave propagation
for 5G cellular communication, the main issues under investi-
gation are:
Pathloss. If the electrical size of the antennas (i.e., their
size measured by the wavelength λ = c/fc where fc is the
carrier frequency) is kept constant, as the frequency increases
the antennas shrink and their effective aperture scales with
λ2
4π ; then, the free-space pathloss between a transmit and a
receive antenna grows with fc2 . Thus, increasing fc by an
order of magnitude, say from 3 to 30 GHz, adds 20 dB
of power loss regardless of the transmit-receive distance.
However, if the antenna aperture at one end of the link is
kept constant as the frequency increases, then the free-space
pathloss remains unchanged. Further, if both the transmit and
receive antenna apertures are held constant, then the free-space
pathloss actually diminishes with fc2 : a power gain that would
help counter the higher noise floor associated with broader
signal bandwidths.
Although preserving the electrical size of the antennas is
desirable for a number of reasons, maintaining at the same
time the aperture is possible utilizing arrays, which aggregate
the individual antenna apertures: as the antennas shrink with
frequency, progressively more of them must be added in the
original area. The main challenge becomes cophasing these
antennas so that they steer and/or collect energy productively.
This challenge becomes more pronounced when the channel
changes rapidly, for example due to mobility and the higher
5
Doppler shifts at mmWave frequencies or due to rapid alter-
ations in the physical orientation of the devices.
Blocking. MmWave signals exhibit reduced diffraction and
a more specular propagation than their microwave counter-
parts, and hence they are much more susceptible to blockages.
This results in a nearly bimodal channel depending on the
presence or absence of Line-of-Sight (LoS). According to re-
cent measurements [67], [69], as the transmit-receive distance
grows the pathloss accrues close to the free-space value of 20
dB/decade under LoS propagation, but drops to 40 dB/decade
plus an additional blocking loss of 15–40 dB otherwise.
Because of the sensitivity to blockages, a given link can rapidly
transition from usable to unusable and, unlike small-scale
fading, large-scale obstructions cannot be circumvented with
standard small-scale diversity countermeasures. New channel
models capturing these effects are much needed, and in fact
currently being developed [67], [70], [71] and applied to
system-level analysis [57], [72]–[74] and simulation studies
such as [75] and [76] in this special issue.
Atmospheric and rain absorption. The absorption due to
air and rain is noticeable, especially the 15 dB/km oxygen
absorption within the 60-GHz band (which is in fact why this
band is unlicensed), but it is inconsequential for the urban
cellular deployments currently envisioned [64], [66] where
BS spacings might be on the order of 200 m. In fact, such
absorption is beneficial since it further attenuates background
interference from more distant BSs, effectively increasing the
isolation of each cell.
The main conclusion is that the propagation losses for
mmWave frequencies are surmountable, but require large an-
tenna arrays to steer the beam energy and collect it coher-
ently. While physically feasible, the notion of narrow-beam
communication is new to cellular communications and poses
difficulties, which we next discuss.
2) Large arrays, narrow beams: Building a cellular system
out of narrow and focused beams is highly nontrivial and
changes many traditional aspects of cellular system design.
MmWave beams are highly directional, almost like flashlights,
which completely changes the interference behavior as well as
the sensitivity to misaligned beams. The interference adopts an
on/off behavior where most beams do not interfere, but strong
interference does occur intermittently. Overall, interference is
de-emphasized and mmWave cellular links may often be noise-
limited, which is a major reversal from 4G. Indeed, even the
notion of a “cell” is likely to be very different in a mmWave
system since, rather than distance, blocking is often the first-
order effect on the received signal power. This is illustrated in
Fig. 2.
Link acquisition. A key challenge for narrow beams is
the difficulty in establishing associations between users and
BSs, both for initial access and for handoff. To find each
other, a user and a BS may need to scan lots of angular
positions where a narrow beam could possibly be found, or
deploy extremely large coding/spreading gains over a wider
beam that is successively narrowed in a multistage acquisition
procedure. Developing solutions to this problem, particularly
in the context of high mobility, is an important research
challenge.
Leveraging the legacy 4G network. A concurrent utiliza-
tion of microwave and mmWave frequencies could go a long
way towards overcoming some of the above hurdles. An inter-
esting proposal in that respect is the notion of “phantom cells”
(relabeled “soft cells” within 3GPP) [77], where mmWave
frequencies would be employed for payload data transmission
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
Signalling Signalling
Microwave
BS
Data Data & signalling
mmWave BS
Fig. 3: MmWave-enabled network with phantom cells.
from small-cell BSs while the control plane would operate
at microwave frequencies from macro BSs (cf. Fig. 3). This
would ensure stable and reliable control connections, based
on which blazing fast data transmissions could be arranged
over short-range mmWave links [78]. Sporadic interruptions
of these mmWave links would then be far less consequential,
as control links would remain in place and lost data could be
recovered through retransmissions.
Novel transceiver architectures needed. Despite the
progress made in WiFi mmWave systems, nontrivial hardware
issues remain, and in some cases will directly affect how the
communication aspects are designed. Chief among these is the
still-exorbitant power consumption of particularly the analog-
to-digital (A/D) but also the digital-to-analog (D/A) converters
needed for large bandwidths. A main consequence is that,
although large antenna arrays and high receiver sensitivities
are needed to deal with the pathloss, having customary fully
digital beamformers for each antenna appears to be unfeasible.
More likely are structures based on old-fashioned analog phase
shifters or, perhaps, hybrid structures where groups of antennas
share a single A/D and D/A [79]–[82]. On the flip side, offering
some relief from these difficulties, the channels are sparser and
thus the acquisition of channel-state information is facilitated;
in particular, channel estimation and beamforming techniques
exploiting sparsity in the framework of compressed sensing
are being explored [83], [84].
C. Massive MIMO
Stemming from research that blossomed in the late 1990s
[85], [86], MIMO communication was introduced into WiFi
systems around 2006 and into 3G cellular shortly thereafter.
In essence, MIMO embodies the spatial dimension of the
communication that arises once a multiplicity of antennas are
available at base stations and mobile units. If the entries of
the channel matrix that ensues exhibit—by virtue of spac-
ing, cross-polarization and/or angular disposition—sufficient
statistical independence, multiple spatial dimensions become
available for signaling and the spectral efficiency multiplies
accordingly [87], [88].
In single-user MIMO (SU-MIMO), the dimensions are lim-
ited by the number of antennas that can be accommodated on
a portable device. However, by having each BS communicate
with several users concurrently, the multiuser version of MIMO
(MU-MIMO) can effectively pull together the antennas at
those users and overcome this bottleneck. Then, the signaling
dimensions are given by the smallest between the aggregate
number of antennas at those users and the number of antennas
at the BS.
Furthermore, in what is now known as coordinated multi-
point (CoMP) transmission/reception, multiple BSs can coop-
erate and act as a single effective MIMO transceiver thereby
turning some of the interference in the system into useful
6
signals; this concept in fact underpins many of the approaches
to interference and mobility management mentioned earlier in
this section.
Data Well-established by the time LTE was developed, MIMO
mmWave BS was a native ingredient thereof with two-to-four antennas per
mobile unit and as many as eight per base station sector, and
it appeared that, because of form factors and other apparent
limitations, such was the extent to which MIMO could be
leveraged. Marzetta was instrumental in articulating a vision
in which the number of antennas increased by more than an
order of magnitude, first in a 2007 presentation [89] with the
details formalized in a landmark paper [90]. The proposal was
to equip BSs with a number of antennas much larger than the
number of active users per time-frequency signaling resource,
and given that under reasonable time-frequency selectivities
accurate channel estimation can be conducted for at most some
tens of users per resource, this condition puts the number of
antennas per base station into the hundreds. This bold idea,
initially termed “large-scale antenna systems” but now more
popularly known as “massive MIMO”, offers enticing benefits:
• Enormous enhancements in spectral efficiency with-
out the need for increased BS densification, with the
possibility—as is always the case—of trading some of
those enhancements off for power efficiency improve-
ments [91], [92].
• Smoothed out channel responses because of the vast
spatial diversity, which brings about the favorable action
of the law of large numbers. In essence, all small-scale
randomness abates as the number of channel observa-
tions grows.
• Simple transmit/receive structures because of the quasi-
orthogonal nature of the channels between each BS
and the set of active users sharing the same signaling
resource. For a given number of active users, such
orthogonality sharpens as the number of BS antennas
grows and simple linear transceivers, even plain single-
user beamforming, perform close-to-optimally.
The promise of these benefits has elevated massive MIMO
to a central position in preliminary discussions about 5G [93],
with a foreseen role of providing a high-capacity umbrella
of ubiquitous coverage in support of underlying tiers of small
cells. However, for massive MIMO to become a reality, several
major challenges must first be overcome, and the remainder
of this section is devoted to their dissection. For very recent
contributions on these and other aspects, the reader is referred
to a companion special issue on massive MIMO [94]. The
present special issue contains further new contributions, men-
tioned throughout the discussion that follows, plus reference
[95] dealing with the massification of MIMO multicasting [96],
[97].
1) Pilot Contamination and Overhead Reduction: Pilot
transmissions can be made orthogonal among same-cell users,
to facilitate cleaner channel estimates [98], [99], but must
be reused across cells—for otherwise all available resources
would end up consumed by pilots. This inevitably causes
interference among pilots in different cells and hence puts a
floor on the quality of the channel estimates. This interference,
so-called “pilot contamination,” does not vanish as the number
of BS antennas grows large, and so is the one impairment
that remains asymptotically. However, pilot contamination is
a relatively secondary factor for all but colossal numbers
of antennas [100]. Furthermore, various methods to reduce
and even eliminate pilot contamination via low-intensity BS
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
TABLE I: FD-MIMO system-level downlink simulation results at 2.5
GHz. Half-wavelength antenna spacings in both the horizontal and
vertical dimensions at the BSs, 2 antennas per user, 30% overhead.
The baseline is SU-MIMO with 4 antennas per BS and the FD-MIMO
results (average and edge data rates) are for MU-MIMO with 16 and
64 antennas, respectively corresponding to 4 × 4 and 8 × 8 planar
arrays per BS sector.
SU-MIMO FD-MIMO 16 FD-MIMO 64
Aggregate Data Rate (b/s/Hz/cell) 2.32 3.28 6.37
Edge Data Rate (b/s/Hz) 0.063 0.1 0.4
coordination have already been formulated [101], [102]. Still,
a careful design of the pilot structures is required to avoid an
explosion in overhead. The ideas being considered to reign in
pilot overheads include exploiting spatial correlations, so as
to share pilot symbols among antennas, and also segregating
the pilots into classes (e.g., channel strength gauging for link
adaptation v. data detection) such that each class can be
transmitted at the necessary rate, and no faster.
2) Architectural Challenges: A more serious challenge to
the realization of the massive MIMO vision has to do with
its architecture. The vision requires radically different BS
structures where, in lieu of a few high-power amplifiers feeding
a handful of sector antennas, we would have a myriad of
tiny antennas fed by correspondingly low-power amplifiers;
most likely each antenna would have to be integrated with
its own amplifier. Scalability, antenna correlations and mutual
couplings, and cost, are some of the issues that must be
sorted out. At the same time, opportunities arise for innovative
topologies such as conformal arrays along rooftops or on
building facades, and we next dwell on a specific topological
aspect in which innovation is taking place.
Within this special issue, [103] explores alternative and
highly innovative antenna designs based on the utilization of
an electromagnetic lens-focusing antenna.
3) Full-Dimension MIMO and Elevation Beamforming:
Existing BSs mostly feature linear horizontal arrays, which
in tower structures can only accommodate a limited number
of antennas, due to form factors, and which only exploit the
azimuth angle dimension. By adopting planar 2D arrays similar
to Fig. 3 and further exploiting the elevation angle, so-called
full-dimension MIMO (FD-MIMO) can house many more
antennas with the same form factor [104]. As a side benefit,
tailored vertical beams increase the signal power and reduce
interference to users in neighboring cells. Some preliminary
cell average and edge data rates obtained from Samsung’s
network simulator are listed in Table I where, with numbers of
antennas still modest for what massive MIMO is envisioned
to be, multiple-fold improvements are observed.
4) Channel Models: Parallel to the architectural issues run
those related to channel models, which to be sound require
extensive field measurements. Antenna correlations and cou-
plings for massive arrays with relevant topologies must be
determined, and a proper modeling of their impact must be
established; in particular, the degree of actual channel orthog-
onalization in the face of such nonidealities must be verified.
And, for FD-MIMO, besides azimuth, the modeling needs to
incorporate elevation [104]–[106], which is a dimension on
which far less data exists concerning power spectra and angle
spreads. A 3D channel modelling study currently under way
within 3GPP is expected to shed light on these various issues
[107]. References [106], [108] in this special issue also deal
with this subject.
7
5) Coexistence with Small Cells: As mentioned earlier,
massive MIMO BSs would most likely have to coexist with
tiers of small cells, which would not be equipped with massive
MIMO due to their smaller form factor. Although the simplest
alternative is to segregate the corresponding transmissions in
frequency, the large number of excess antennas at massive
MIMO BSs may offer the opportunity of spatial nulling
and interference avoidance with relative simplicity and little
penalty. To confirm the feasibility of this idea, put forth in
[109] and further developed in [110] within this special issue,
comprehensive channel models are again needed.
As networks become dense and more traffic is offloaded to
small cells, the number of active users per cell will diminish
and the need for massive MIMO may decrease. Aspects such
as cost and backhaul will ultimately determine the balance
between these complementary ideas.
6) Coexistence with mmWave: As discussed in Sec. II-B,
mmWave communication requires many antennas for beam-
steering. The antennas are much smaller at these frequen-
cies and thus very large numbers thereof can conceivably fit
into portable devices, and these antennas can indeed provide
beamforming power gain but also MIMO opportunities as
considered in [111] within this special issue. Any application
of massive MIMO at mmWave frequencies would have to find
the correct balance between power gain/interference reduction
and parallelization.
III. D ESIGN I SSUES FOR 5G
In addition to supporting 1000x higher throughput, 5G
cellular networks must decrease latencies, lower energy con-
sumption, lower costs, and support many low-rate connections.
In this section, we discuss important ongoing research areas
that support these requirements. We begin with the most
fundamental aspect of the physical layer—the waveform—and
then consider the evolution of cloud-based and virtualized net-
work architectures, latency and control signaling, and energy
efficiency.
A. The Waveform: Signaling and Multiple Access
The signaling and multiple access formats, i.e., the wave-
form design, have changed significantly at each cellular gen-
eration and to a large extent they have been each generation’s
defining technical feature. They have also often been the
subject of fierce intellectual and industrial disputes, which
have played out in the wider media. The 1G approach, based
on analog frequency modulation with FDMA, transformed
into a digital format for 2G and, although it employed both
FDMA and TDMA for multiple access, was generally known
as “TDMA” due to the novelty of time-multiplexing. Mean-
while, a niche spread spectrum/CDMA standard that was
developed by Qualcomm to compete for 2G [112] became
the dominant approach to all global 3G standards. Once the
limitations of CDMA for high-speed data became inescapable,
there was a discreet but unmistakable retreat back towards
TDMA, with minimal spectrum spreading retained and with
the important addition of channel-aware scheduling [113].
Due to the increasing signal bandwidths needed to support
data applications, orthogonal frequency-division multiplexing
(OFDM) was unanimously adopted for 4G in conjunction with
scheduled FDMA/TDMA as the virtues of orthogonality were
viewed with renewed appreciation.
In light of this history, it is natural to ponder the possibility
that the transition to 5G could involve yet another major
change in the signaling and multiple access formats.
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
1) OFDM and OFDMA: The Default Approach: OFDM has
become the dominant signaling format for high-speed wireless
communication, forming the basis of all current WiFi standards
and of LTE, and further of wireline technologies such as digital
subscriber lines, digital TV, and commercial radio. Its qualities
include:
• A natural way to cope with frequency selectivity.
• Computationally efficient implementation via FFT/IFFT
blocks and simple frequency-domain equalizers.
• An excellent pairing for MIMO, since OFDM allows for
the spatial interference from multiantenna transmission
to be dealt with at a subcarrier level, without the added
complication of intersymbol interference.
From a multiple access vantage point, OFDM invites dy-
namic fine-grained resource allocation schemes in the digital
domain, and the term OFDMA is employed to denote orthog-
onal multiple access at a subcarrier level. In combination with
TDMA, this parcels the time-frequency grid into small units
known as resource blocks that can be easily discriminated
through digital filtering [114]. Being able to do frequency
and time slot allocation digitally also enables more adaptive
and sophisticated interference management techniques such
as fractional frequency reuse or spectrum partitions between
small cells and macrocells. Finally, given its near-universal
adoption, industry has by now a great deal of experience
with its implementation, and tricky aspects of OFDM such
as frequency offset correction and synchronization have been
essentially conquered.
2) Drawbacks of OFDM: Given this impressive list of
qualities, and the large amount of inertia in its favor, OFDM is
the unquestionable frontrunner for 5G. However, some weak
points do exist that could possibly become more pronounced
in 5G networks.
First, the peak-to-average-power ratio (PAPR) is higher in
OFDM than in other formats since the envelope samples are
nearly Gaussian due to the summation of uncorrelated inputs in
the IFFT. Although a Gaussian signal distribution is capacity-
achieving under an average power constraint [115], in the
face of an actual power amplifier a high PAPR sets up an
unattractive tradeoff between the linearity of the transmitted
signal and the cost of the amplifier. This problem can be largely
overcome by precoding the OFDM signals at the cost of a more
involved equalization process at the receiver and a slight power
penalty; indeed, this is already being done in the LTE uplink
[116].
Second, OFDM’s spectral efficiency is satisfactory, but could
perhaps be further improved upon if the requirements of strict
orthogonality were relaxed and if the cyclic prefixes (CPs)
that prevent interblock interference were smaller or discarded.
The paper [117] in this special issue, instead, proposes the
use of a novel OFDMA-based modulation scheme named fre-
quency and quadrature amplitude modulation (FQAM), which
is shown to improve the downlink throughput for cell-edge
users.
Perhaps the main source of concerns, or at least of open
questions, is the applicability of OFDM to mmWave spectrum
given the enormous bandwidths therein and the difficulty of
developing efficient power amplifiers at those frequencies. For
example, a paper in this special issue proposes a single-carrier
signaling with null cyclic prefix as an alternative to OFDM at
mmWave frequencies [75].
3) Potential Alternatives to OFDM: To address OFDM’s
weaknesses, we now overview some alternative approaches
8
being actively investigated. Most of these, however, can be
considered incremental departures from OFDM rather than
the step-function changes that took place in previous cellular
generations.
Time-frequency packing. Time-frequency packing [118]
and faster-than-Nyquist signaling [119]–[121] have been re-
cently proposed to circumvent the limitations of strict orthog-
onality and CP. In contrast to OFDM, where the product of
the symbol interval and the subcarrier spacing equals 1, in
faster-than-Nyquist signaling products smaller than 1 can be
accommodated and spectral efficiency improvements on the
order of 25% have been claimed.
Nonorthogonal signals. There is a growing interest in
multicarrier formats, such as filterbank multicarrier [122], that
are natively nonorthogonal and thus do not require prior syn-
chronization of distributed transmitters. A new format termed
universal filtered multiCarrier (UFMC) has been proposed
whereby, starting with an OFDM signal, filtering is performed
on groups of adjacent subcarriers with the aim of reducing
sidelobe levels and intercarrier interference resulting from poor
time/frequency synchronization [123], [124].
Filterbank multicarrier. To address the drawbacks of rect-
angular time windowing in OFDM, namely the need for large
guard bands, [125] shows that the use of filterbank multicarrier
permits a robust estimation of very large propagation delays
and of arbitrarily high carrier frequency offsets, whereas
OFDM would have required a very long CP to attain the same
performance levels.
Generalized frequency division multiplexing. GFDM is
a multicarrier technique that adopts a shortened CP through
the tail biting technique and is particularly well suited for
noncontiguous frequency bands [126], [127], which makes it
attractive for spectrum sharing where frequency-domain holes
may have to be adaptively filled.
Single carrier. Single-carrier transmission has also been
attracting renewed interest, chiefly due to the development
of low-complexity nonlinear equalizers implemented in the
frequency domain [128]. This may be of particular interest
for mmWave as discussed in this same special issue [75].
Tunable OFDM. We conclude with our own opinion that
OFDM could be well adapted to different 5G requirements
by allowing some of its parameters to be tunable, rather than
designed for essentially the worst-case multipath delay spread.
In particular, given the increasingly software-defined nature of
radios, the FFT block size, the subcarrier spacing and the CP
length could change with the channel conditions: in scenarios
with small delay spreads—notably dense urban/small cells and
mmWave channels—the subcarrier spacing could grow and the
FFT size and the CP could be significantly shortened to lower
the latency, the PAPR, the CP’s power and bandwidth penalty,
and the computational complexity; in channels with longer
delay spreads, that could revert to narrower subcarriers, longer
FFT blocks, and a longer CP.
B. Cloud-based Networking
Although this special issue is mainly focused on the air
interface, for the sake of completeness we briefly touch on
the exciting changes taking place at the network level. In
that respect, the most relevant event is the movement of data
to the cloud so that it can be accessed from anywhere and
via a variety of platforms. This fundamentally redefines the
endpoints and the time frame for which network services
are provisioned. It requires that the network be much more
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
OFDM
0
−10
−20
dB
−30
−40
−50
frequency
Frequency−packed OFDM with frequency spacing set to 0.8/T
0 will emerge: the provision of user-controlled management in
−10
−20
dB
−30
−40
−50
frequency
Filtered OFDM
0 tween the entities in the control and data planes, as well as
−10
−20
dB
−30
−40
−50
frequency
Fig. 4: Frequency-domain magnitude responses of some adjacent
waveforms for OFDM, frequency-packed OFDM, and filtered OFDM.
The two signaling formats alternative to OFDM trade subcarrier
orthogonality for either better spectral efficiency (frequency-packed
OFDM) or lower out-of-band emissions (filtered OFDM).
nimble, flexible and scalable. As such, two technology trends
will become paramount in the future: network function vir-
tualization (NFV) and software defined networking (SDN).
Together, these trends represent the biggest advance in mobile
communication networking in the last 20 years, bound to
fundamentally change the way network services are provided.
Although the move towards virtualization is thus far taking
place only within the core network, this trend might eventually
expand towards the edges. In fact, the term cloud-RAN is
already being utilized, but for now largely to refer to schemes
whereby multiple BSs are allowed to cooperate [129]. If and
when the Bss themselves become virtualized—down to the
MAc and PHY—-this term will be thoroughly justified [130].
1) Network Function Virtualization: NFV enables network
functions that were traditionally tied to hardware appliances
to run on cloud computing infrastructure in a data center. It
should be noted that this does not imply that the NFV infras-
tructure will be equivalent to commercial cloud or enterprise
cloud. What is expected is that there will be a high degree of
reuse of what commercial cloud offers.
It is natural to expect that some requirements of mobile
networks such as the separation of the data plane, control
plane and management plane, will not be feasible within
9
the commercial cloud. Nevertheless, the separation of the
network functions from the hardware infrastructure will be the
cornerstone of future architectures. The key benefit will be
the ability to elastically support network functional demands.
Furthermore, this new architecture will allow for significant
nimbleness through the creation of virtual networks and of
new types of network services [131]. A detailed description of
the NFV architecture is beyond the scope of this paper, and
interested readers can consult [131]–[133] and the references
therein.
As virtualization of the communication network gains trac-
tion in the industry, an old concept, dating back to the 1990s,
network elements. Advances in computing technology have
reached a level where this vision can become a reality, with
the ensuring architecture having recently been termed software
defined networking (SDN).
2) Software Defined Networking: SDN is an architectural
framework for creating intelligent programmable networks.
Specifically, it is defined as an architecture where the control
and data planes are decoupled, network intelligence and state
are logically centralized, and the underlying network infras-
tructure is abstracted from the application [134].
The key ingredients of SDN are an open interface be-
programmability of the network entities by external applica-
tions. The main benefits of this architecture are the logical
decoupling of the network intelligence to separate software-
based controllers, exposing the network capabilities through
an application program interface, and enabling the application
to request and manipulate services provided by the network
[135].
From a wireless core network point of view, NFV and SDN
should be viewed as tools for provisioning the next generation
of core networks with many issues still open in terms of
scalability, migration from current structures, management and
automation, and security.
C. Energy efficiency
As specified in our stated requirements for 5G, the energy
efficiency of the communication chain—typically measured
in either Joules/bit or bits/Joule—will need to improve by
about the same amount as the data rate just to maintain the
power consumption. And by more if such consumption is to be
reduces. This implies a several-order-of-magnitude increase in
energy efficiency, which is extremely challenging. Unsurpris-
ingly, in recent years there has been a surge of interest in the
topic of energy efficient communications, as can be seen from
the number of recent special issues, conferences and research
projects devoted to “green communications” [136]–[138]. In
addition to laudable environmental concerns, it is simply not
viable from a logistical, cost or battery-technology point of
view to continually increase power consumption.
Due to the rapidly increasing network density (cf. Sect.
II-A), the access network consumes the largest share of the
energy [139]. Research has focused on the following areas.
1) Resource allocation: The literature is rich in contribu-
tions dealing with the design of resource allocation strategies
aimed at the optimization of the system energy efficiency
[140]–[146]; the common message of these papers is that, by
accepting a moderate reduction in the data rates that could
otherwise be achieved, large energy savings can be attained.
Within this special issue, [147] introduces an energy-efficient
coordinated beamforming design for HetNets.
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
2) Network Planning: Energy-efficient network planning
strategies include techniques for minimizing the number of
BSs for a coverage target [148] and the design of adaptive
BS sleep/wake algorithms for energy savings [149]–[152]. The
underlying philosophy of these papers is that, since networks
have been designed to meet peak-hour traffic, energy can be
saved by (partially) switching off BSs when they have no active
users or simply very low traffic. Of course, there are different
degrees of hibernation available for a BS2 and attention must
be paid in order to avoid unpleasant coverage holes; this is
usually accomplished through an increase of the transmitted
power from nearby BSs.
3) Renewable energy: Another intriguing possibility is that
of BSs powered by renewable energy sources such as solar
power [153]. This is of urgent interest in developing countries
lacking a reliable and ubiquitous power grid, but it is also
intriguing more broadly as it allows “drop and play” small
cell deployment (if wireless backhaul is available) rather than
“plug and play”. A recent paper showed that in a dense HetNet,
plausible per-BS traffic loads can actually be served solely
by energy harvesting BSs [154]. A more relaxed scenario
is considered in [155], where the resource allocation makes
efficient use of both renewable and traditional energy sources.
4) Hardware solutions: Finally, much of the power
consumption issues will be dealt with by hardware engineers,
with recent work in low-loss antennas, antenna muting, and
adaptive sectorization according to traffic requirements (see,
e.g., [156]).
In summary, energy efficiency will be a major research
theme for 5G, spanning many of the other topics in this article:
• True cloud-RAN could provide an additional opportunity
for energy efficiency since the centralization of the
baseband processing might save energy [157], especially
if advances on green data centers are leveraged [158].
• The tradeoff between having many small cells or fewer
macrocells given their very different power consump-
tions is also of considerable interest [159].
• A complete characterization of the energy consumed
by the circuitry needed for massive MIMO is currently
lacking.
• MmWave energy efficiency will be particularly crucial
given the unprecedented bandwidths [160].
IV. S PECTRUM , R EGULATION AND S TANDARDIZATION
FOR 5G
Departing from strictly technical issues, we now turn our
attention to the crucial intersections that 5G technologies will
encounter with public policy, industry standardization, and
economic considerations.
A. Spectrum Policy and Allocation
As discussed in Section II-B, the beachfront microwave
spectrum is already saturated in peak markets at peak times
while large amounts of idle spectrum do exist in the mmWave
realm. Due to the different propagation characteristics, and
recalling the concept of phantom cells, future systems will
need to integrate a broad range of frequencies: low frequencies
for wide coverage, mobility support, and control, and high
frequencies for small cells. This will require new approaches
2 As an example, a BS serving few users may choose to operate on a reduced
set of subcarriers, or it may switch off some of its sectors.
10
to spectrum policy and allocation methods. Topics such as
massive MIMO and small cells, which address the efficient
use of spectrum, must also be considered important issues
in spectrum policy. Needless to say, spectrum allocation and
policy is an essential topic for 5G, so this section considers the
pros and cons of different approaches to spectrum regulation
in that context.
1) Exclusive Licenses: The traditional approach to spectrum
policy is for the regulator to award an exclusive license to a
particular band for a particular purpose, subject to limitations
(e.g., power levels or geographic coverage). Exclusive access
gives full interference management control to the licensee
and provides an incentive for investments in infrastructure,
allowing for quality-of-service guarantees. Downsides include
high entry barriers because of elevated sunk costs, both in the
spectrum itself and in infrastructure, and that such allocations
are inherently inefficient since they occur over very long time
scales—typically decades—and thus the spectrum is rarely
allocated to the party able to make the best economic use of
it.
To address these inefficiencies, market-based approaches
have been propounded [161]. Attempting to implement this
idea, spectrum auctions have been conducted recently to refarm
spectrum, a process whereby long-held commercial radio and
TV allocations are moved to different (smaller) bands releasing
precious spectrum for wireless communications; a prime exam-
ple of this is the so-called “digital dividend” auctions arising
from the digitization of radio and TV. However, there are
claims that spectrum markets have thus far not been successful
in providing efficient allocations because such markets are not
sufficiently fluid due to the high cost of the infrastructure [162].
According to these claims, spectrum and infrastructure cannot
be easily decoupled.
2) Unlicensed Spectrum: At the other extreme, regulators
can designate a band to be “open access”, meaning that there is
no spectrum license and thus users can share the band provided
their devices are certified (by class licenses). Examples are
the industrial, scientific and medical (ISM) bands, which are
utilized by many devices including microwave ovens, medical
devices, sensor networks, cordless phones and especially by
WiFi. With open access, barriers to entry are much lower and
there is enhanced competition and innovation, as the incredible
success of WiFi and other ISM-band applications makes plain.
The downside of open access is potentially unmanageable
interference, no quality-of-service guarantees, and, possibly,
the “tragedy of the commons,” where no one achieves a desired
outcome. Still, it is useful to consider the possibility of open
access for bands utilized in small cells as future networks
may involve multiple players and lower entry barriers may be
needed to secure the emergence of small-cell infrastructures.
Although interference is indeed a significant problem in
current open access networks, it is interesting to note that
cellular operators nevertheless rely heavily on WiFi offloading:
currently about half of all cellular data traffic is proactively
offloaded through unlicensed spectrum [2]. WiFi hotspots are
nothing but small cells that spatially reuse ISM frequencies. At
mmWave frequencies, the main issue is signal strength rather
than interference, and it is therefore plausible that mmWave
bands be unlicensed, or at a minimum several licensees will
share a given band under certain new regulations. This question
is of pressing interest for 5G.
3) Spectrum Sharing: Options do exist halfway between ex-
clusive licenses and open access, such as the opportunistic use
of TV white space. While the potential of reusing this spectrum
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
is enticing, it is not crystal clear that reliable communication
services can be delivered that way. Alternatively, Authorized
Shared Access [163] and Licensed Shared Access [164] are
regulatory frameworks that allow spectrum sharing by a lim-
ited number of parties each having a license under carefully
specified conditions. Users agree on how the spectrum is to
be shared, seeking interference protection from each other,
thereby increasing the predictability and reliability of their
services.
4) Market-Based Approaches to Spectrum Allocation: Given
the advantages of exclusive licenses for ensuring quality of
service, it is likely that most beachfront spectrum will continue
to be allocated that way. Nevertheless, better utilization could
likely be obtained if spectrum markets could become more
fluid [161]. To that end, liberal licenses do not, in principle,
preclude trading and reallocation on a fast time scale, rendering
spectrum allocations much more dynamic. Close attention must
be paid to the definition of spectrum assets, which have a space
as well as a time scale, and the smaller the scales, the more
fluid the market [165].
In small cells, traffic is much more volatile than in macro-
cells and operators may find it beneficial to enter into sharing
arrangements for both spectrum and infrastructure. Dynamic
spectrum markets may emerge, managed by brokers, allowing
licenses to spectrum assets to be bought and sold—or leased—
on time scales of hours, minutes or even ms [166]. Along
these lines, an interesting possibility is for a decoupling of
infrastructure, spectrum and services [166]. In particular, there
may be a separation between spectrum owners and operators.
Various entities may own and/or share a network of BSs,
and buy and sell spectrum assets from spectrum owners,
via brokers. These network owners may offer capacity to
operators, which in turn would serve the end customers with
performance guarantees. All of this, however, would require
very adaptable and frequency agile radios.
We conclude this discussion by noting that offloading onto
unlicensed spectrum such as TV whitespace or mmWave
bands could have unexpected results. In particular, adding an
unlicensed shared band to an environment where a set of
operators have exclusive bands can lead to an overall decrease
in the total welfare (Braess’ paradox) [167]. This is because
operators might have an incentive to offload traffic even when
this runs counter to the overall social welfare, defined as the
total profit of the operators and the utilities of the users, minus
the costs. An operator might have an incentive to increase
prices so that some traffic is diverted to the unlicensed band,
where the cost of interference is shared with other operators,
and this price increase more than offsets the operator’s benefits.
Further, while unlicensed spectrum generally lowers barriers
to entry and increases competition, the opposite could occur
and in some circumstances a single monopoly operator could
emerge [168] within the unlicensed bands.
B. Regulation and Standardization
1) 5G Standardization Status: Several regional forums and
projects have been established to shape the 5G vision and to
study its key enabling technologies [6], [169]–[171]. For ex-
ample, the aforementioned EU project METIS has already re-
leased documents on scenarios and requirements [172], [173].
Meanwhile, 5G has been increasingly referred to as “IMT-
2020” in many industry forums and international telecommu-
nications union (ITU) working groups [174] with the goal,
11
as the name suggests, of beginning commercial deployments
around 2020.
To explore 5G user requirements and to elaborate a standards
agenda to be driven by them, the ETSI held a future mobile
summit [175] in Nov. 2013. The summit concluded, in line
with the thesis of this paper, that an evolution of LTE may
not be sufficient to meet the anticipated 5G requirements.
That conclusion notwithstanding, 5G standardization has not
yet formally started within 3GPP, which is currently finalizing
LTE Rel-12 (the third release for the LTE-Advanced family of
4G standards). The timing of 5G standardization has not even
been agreed upon, although it is not expected to start until
later Rel-14 or Rel-15, likely around 2016–2017. However,
many ongoing and proposed study items for Rel-12 are already
closely related to 5G candidate technologies covered in this
paper (e.g., massive MIMO) and thus, in that sense, the seeds
of 5G are being planted in 3GPP. Whether an entirely new
standards body will emerge for 5G as envisioned in this paper
is unclear; the ongoing success of 3GPP relative to its erstwhile
competitors (3GPP2 and the WiMAX Forum) certainly gives
it an advantage, although a name change to 5GPP would seem
to be a minimal step.
2) 5G Spectrum Standardization: Spectrum standardization
and harmonization efforts for 5G have begun within the ITU.
Studies are under way on the feasibility of bands above
6 GHz [176], including technical aspects such as channel mod-
elling, semiconductor readiness, coverage, mobility support,
potential deployment scenarios and coexistence with existing
networks.
To be available for 5G, mmWave spectrum has to be
repurposed by national regulators for mobile applications and
agreement must be reached in ITU world radiocommunication
conferences (WRC) on the global bands for mmWave commu-
nications. These processes tend to be tedious and lengthy, and
there are many hurdles to clear before the spectrum can indeed
be available. On the ITU side, WRC-18 is shaping up as the
time and venue to agree on mmWave spectrum allocations for
5G.
In addition to the ITU, many national regulators have also
started their own studies on mmWave spectrum for mobile
communications. In the USA, the technological advisory coun-
cil of the federal communications committee (FCC) has carried
out extensive investigations on mmWave technology in the
last few years and it is possible that FCC will issue a notice
of inquiry in 2014, which is always the first step in FCC’s
rulemaking process for allocation of any new frequency bands.
As discussed above, it is also unclear how such bands will
be allocated or even how they should be allocated, and the
technical community should actively engage the FCC to make
sure they are allocated in a manner conducive to meeting
5G requirements. Historically, other national regulators have
tended to follow the FCC’s lead on spectrum policy.
C. Economic Considerations
The economic costs involved in moving to 5G are substan-
tial. Even if spectrum costs can be greatly reduced through
the approaches discussed above, it is still a major challenge
for carriers to densify their networks to the extent needed to
meet our stated 5G requirements. Two major challenges are
that BS sites are currently expensive to rent, and so is the
backhaul needed to connect them to the core network.
1) Infrastructure Sharing: One possible new business model
could be based on infrastructure sharing, where the owners of
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
infrastructure and the operators are different. There are several
ways in which infrastructure could be shared.
Passive sharing. The passive elements of a network include
the sites (physical space, rooftops, towers, masts and pylons),
the backhaul connection, power supplies, and air-conditioning.
Operators could cover larger geographical areas at a lower cost
and with less power consumption if they shared sites, and this
might be of particular importance in dense 5G networks [177].
Regulation could be required to force major operators to share
their sites and improve competition.
Active sharing. Active infrastructure sharing would involve
antennas, BSs, radio access networks and even core networks.
BS and/or radio access network sharing may be particularly
attractive when rolling out small-cell networks [178]. This
type of sharing could lead to collusion, with anticompetitive
agreements on prices and services [177]. Regulations are
required to prevent such collusion, but on the positive side
are the economies of scale.
Mobile virtual network operators. A small cell may be
operated by a mobile virtual network operator that does not
own any spectrum but has entered into an agreement with
another operator to gain access to its spectrum within the small
cell. The small cell may provide coverage to an enterprise or
business such that, when a user leaves the enterprise, it roams
onto the other operator’s network.
Offloading. Roaming is traditionally used to increase cover-
age in scenarios when service providers’ geographical reaches
are limited. However, in 5G, and as discussed above, traffic
may be offloaded for a different reason: spatial and temporal
demand fluctuations. Such fluctuations will be greater in small-
cell networks. Recent papers consider the incentive for invest-
ment under various revenue-sharing contracts [179], [180]. It
is shown in [179] that sharing increases investment, and the
incentive is greater if the owner of the infrastructure gets the
larger fraction of the revenue when overflow traffic is carried.
A bargaining approach for data offloading from a cellular
network onto a collection of WiFi or femtocell networks is
considered in [49] in this special issue.
2) Backhaul: A major consideration that has been consid-
ered in several places throughout the paper is backhaul, which
will be more challenging to provide for hyper-dense ultra-fast
networks. However, we find optimism in three directions.
• Fiber deployments worldwide continue to mature and
reach farther and farther into urban corridors.
• Wireless backhaul solutions are improving by leaps and
bounds, with considerable startup activity driving inno-
vation and competition. Further, mmWave frequencies
could be utilized for much of the small-cell backhauling
due to their ambivalence to interference. This may in
fact be the first serious deployment of non-LoS mmWave
with massive beamforming gains given that the backhaul
connection is quite static and outdoors-to-outdoors, and
thus more amenable to precise beam alignment.
• Backhaul optimization is becoming a pressing concern,
given its new status as a performance-limiting factor,
and this is addressed in [181], [182] in this special
issue. The problem of jointly optimizing resources in
the radio network and across the backhaul is considered
in [181]. Compression techniques for uplink cloud-
RAN are developed in [182]. Another approach is the
proactive caching of high bandwidth content like popular
video [183].
12
V. C ONCLUSIONS
It is an exciting time in the wireless industry and for
wireless research at large. Daunting new requirements for 5G
are already unleashing a flurry of creative thinking and a sense
of urgency in bringing innovative new technologies into reality.
Even just two years ago, a mmWave cellular system was
considered something of a fantasy; now it is almost considered
an inevitability. As this article has highlighted, it is a long
road ahead to truly disruptive 5G networks. Many technical
challenges remain spanning all layers of the protocol stack
and their implementation, as well as many intersections with
regulatory, policy, and business considerations. We hope that
this article and those in this special issue will help to move us
forward along this road.
ACKNOWLEDGMENTS
The authors thank Arunabha Ghosh (AT&T Labs), Robert
W. Heath Jr. (UT Austin), and Federico Boccardi (Vodaphone)
for very helpful feedback and suggestions on the paper.
R EFERENCES
[1] B. Clerckx, A. Lozano, S. Sesia, C. van Rensburg, and C. B. Papa-
dias, “3GPP LTE and LTE-advanced,” EURASIP Journal on Wireless
Communications and Networking, vol. 2009, 2009.
[2] Cisco, “Visual networking index,” white paper at Cisco.com, Feb.
2014.
[3] M. Corson, R. Laroia, J. Li, V. Park, T. Richardson, and G. Tsirtis,
“Towards proximity-aware internetworking,” IEEE Wireless Commu-
nications Magazine, vol. 17, no. 6, pp. 26–33, Dec. 2010.
[4] A. Maeder, P. Rost, and D. Staehle, “The challenge of m2m commu-
nications for the cellular radio access network,” in Proc., Würzburg
Workshop on IP: Joint ITG and Euro-NF Workshop” Visions of Future
Generation Networks”(EuroView2011), Aug. 2011.
[5] S. Hilton, “Machine-to-machine device connections: worldwide fore-
cast 2010–2020,” Analysys Mason Report, 2010.
[6] “FP7 European Project 317669 METIS (Mobile and wireless com-
munications Enablers for the Twenty-twenty Information Society),”
https://www.metis2020.com/, 2012.
[7] “FP7 European Project 318555 5G NOW (5th Generation
Non-Orthogonal Waveforms for asynchronous signalling),”
http://www.5gnow.eu/, 2012.
[8] “5G-Infrastructure Public-Private Partnership,”
http://5g-ppp.eu/, 2013.
[9] G. P. Fettweis, “The tactile internet: Applications and challenges,”
IEEE Vehicular Technology Magazine, vol. 9, no. 1, pp. 64–70, Mar.
2014.
[10] A. Sendonaris, E. Erkip, and B. Aazhang, “User cooperation diversity.
Part I. System description,” IEEE Trans. on Communications, vol. 51,
no. 11, pp. 1927 – 1938, Nov. 2003.
[11] N. Laneman, D. Tse, and G. Wornell, “Cooperative diversity in
wireless networks: Efficient protocols and outage behavior,” IEEE
Trans. on Info. Theory, vol. 50, no. 12, pp. 3062– 3080, Dec. 2004.
[12] G. J. Foschini, K. Karakayali, and R. A. Valenzuela, “Enormous
spectral efficiency of isolated multiple antenna links emerges in a
coordinated cellular network,” IEE Proceedings - Communications,
vol. 153, no. 4, pp. 548–55, Aug. 2006.
[13] S. Venkatesan, A. Lozano, and R. Valenzuela, “Network MIMO:
Overcoming intercell interference in indoor wireless systems,” in
Proc., Asilomar Conference on Signals, Systems and Computers, Nov.
2007, pp. 83–87.
[14] V. R. Cadambe and S. A. Jafar, “Interference alignment and the degrees
of freedom for the K user interference channel,” IEEE Trans. Inform.
Theory, vol. 54, no. 8, pp. 3425–3441, Aug. 2008.
[15] M. Maddah-Ali, A. Motahari, and A. Khandani, “Communication
over MIMO X channels: Interference alignment, decomposition, and
performance analysis,” IEEE Trans. Inform. Theory, vol. 54, no. 8, pp.
3457–3470, Aug. 2008.
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
[16] S. Venkatesan, H. Huang, A. Lozano, and R. Valenzuela, “A WiMAX-
based implementation of network MIMO for indoor wireless systems,”
EURASIP Journal on Advances in Signal Processing, vol. 2009, Feb.
2009.
[17] H. Huang, M. Trivellato, A. Hottinen, M. Shafi, P. Smith, and
R. Valenzuela, “Increasing downlink cellular throughput with limited
network MIMO coordination,” IEEE Trans. on Wireless Communica-
tions, vol. 8, no. 6, pp. 2983–2989, Jun. 2009.
[18] S. W. Peters and R. W. H. Jr., “Interference alignment via alternating
minimization,” in Proc., IEEE Intl. Conf. on Acoustics, Speech, and
Sig. Proc. (ICASSP), 2009, pp. 2445–2448.
[19] D. Gesbert, S. Hanly, H. Huang, S. Shamai, O. Simeone, and W. Yu,
“Multi-cell MIMO cooperative networks: A new look at interference,”
IEEE Journal on Sel. Areas in Communications, vol. 28, no. 9, pp.
1380 –1408, Dec. 2010.
[20] C. Suh, M. Ho, and D. N. C. Tse, “Downlink interference alignment,”
IEEE Trans. on Communications, vol. 59, no. 9, pp. 2616–2626, Sep.
2011.
[21] P. Wang, H. Wang, L. Ping, and X. Lin, “On the capacity of MIMO
cellular systems with base station cooperation,” IEEE Trans. on
Wireless Communications, vol. 10, no. 11, pp. 3720–3731, Nov. 2011.
[22] O. Simeone, N. Levy, A. Sanderovich, O. Somekh, B. M. Zaidel, H. V.
Poor, and S. Shamai, “Information theoretic considerations for wireless
cellular systems: The impact of cooperation,” Foundations and Trends
in Communications and Information Theory, vol. 7, 2012.
[23] A. Lozano, R. W. Heath, and J. G. Andrews, “Fundamental limits of
cooperation,” IEEE Trans. Inform. Theory, vol. 59, no. 9, pp. 5213–
5226, Sep. 2013.
[24] V. Chandrasekhar, J. G. Andrews, and A. Gatherer, “Femtocell net-
works: a survey,” IEEE Communications Magazine, vol. 46, no. 9, pp.
59–67, Sep. 2008.
[25] M. Dohler, R. W. Heath, A. Lozano, C. B. Papadias, and R. A. Valen-
zuela, “Is the PHY layer dead?” IEEE Communications Magazine,
vol. 49, no. 4, pp. 159–165, Apr. 2011.
[26] J. G. Andrews, “Seven ways that HetNets are a cellular paradigm
shift,” IEEE Communications Magazine, vol. 51, no. 3, pp. 136–144,
Mar. 2013.
[27] J. G. Andrews, H. Claussen, M. Dohler, S. Rangan, and M. C. Reed,
“Femtocells: Past, present, and future,” IEEE Journal on Sel. Areas in
Comm., vol. 30, no. 3, pp. 497–508, Apr. 2012.
[28] R. W. Heath, S. W. Peters, Y. Wang, and J. Zhang, “A current
perspective on distributed antenna systems for the downlink of cellular
systems,” IEEE Communications Magazine, vol. 51, no. 4, pp. 161–
167, Apr. 2013.
[29] H. S. Dhillon, R. K. Ganti, F. Baccelli, and J. G. Andrews, “Modeling
and analysis of K-tier downlink heterogeneous cellular networks,”
IEEE Journal on Sel. Areas in Comm., vol. 30, no. 3, pp. 550–560,
Apr. 2012.
[30] D. Ramasamy, R. Ganti, and U. Madhow, “On the capacity of
picocellular networks,” in Proc., IEEE Intl. Symposium on Information
Theory, July 2013, pp. 241–245.
[31] H. S. Dhillon, R. K. Ganti, and J. G. Andrews, “Load-aware modeling
and analysis of heterogeneous cellular networks,” IEEE Trans. on
Wireless Communications, vol. 12, no. 4, pp. 1666–1677, Apr. 2013.
[32] S. Larew, T. Thomas, and A. Ghosh, “Air interface design and ray
tracing study for 5G millimeter wave communications,” in Proc., IEEE
Globecom B4G Workshop, Dec. 2013.
[33] Qualcomm, “Extending LTE advanced to unli-
censed spectrum,” Feb. 2014. [Online]. Avail-
able: http://www.slideshare.net/qualcommwirelessevolution/
extending-lte-advanced-to-unlicensed-spectrum-31732634
[34] O. Galinina, S. Andreev, M. Gerasimenko, Y. Koucheryavy, N. Hi-
mayat, S. ping Yeh, and S. Talwar, “Capturing spatial randomness
of heterogeneous cellular/WLAN deployments with dynamic traffic,”
IEEE Journal on Sel. Areas in Communications, Sep. 2014.
[35] Q. Ye, B. Rong, Y. Chen, M. Shalash, C. Caramanis, and J. G. An-
drews, “User association for load balancing in heterogeneous cellular
networks,” IEEE Trans. on Wireless Communications, vol. 12, no. 6,
pp. 2706–2716, Jun. 2013.
[36] S. Corroy, L. Falconetti, and R. Mathar, “Cell association in small het-
erogeneous networks: Downlink sum rate and min rate maximization,”
13
in Proc., IEEE Wireless Networking and Comm. Conf., Mar. 2012, pp.
898–902.
[37] J. G. Andrews, S. Singh, Q. Ye, X. Lin, and H. S. Dhillon, “An
overview of load balancing in HetNets: Old myths and open problems,”
IEEE Communications Magazine, vol. 52, no. 4, Apr. 2014.
[38] K. Shen and W. Yu, “Distributed pricing-based user association for
downlink heterogeneous cellular networks,” IEEE Journal on Sel.
Areas in Communications, Sep. 2014.
[39] S. Singh and J. G. Andrews, “Joint resource partitioning and offloading
in heterogeneous cellular networks,” IEEE Trans. on Wireless Com-
munications, vol. 13, no. 2, pp. 888–901, Feb. 2014.
[40] D. Fooladivanda and C. Rosenberg, “Joint resource allocation and user
association for heterogeneous wireless cellular networks,” IEEE Trans.
on Communications, vol. 12, no. 1, pp. 248–257, Jan. 2013.
[41] A. Bedekar and R. Agrawal, “Optimal muting and load balancing for
eICIC,” in Proc., International Symposium on Modeling & Optimiza-
tion in Mobile, Ad Hoc & Wireless Networks (WiOpt), May 2013, pp.
280–287.
[42] S. Borst, S. Hanly, and P. Whiting, “Throughput utility optimization in
HetNets,” in Proc., IEEE Veh. Technology Conf., Jun. 2013, pp. 1–5.
[43] ——, “Optimal resource allocation in HetNets,” in Proc., IEEE Intl.
Conf. on Communications, Jun. 2013, pp. 5437–5441.
[44] S. Deb, P. Monogioudis, J. Miernik, and J. P. Seymour, “Algorithms for
enhanced inter-cell interference coordination (eICIC) in LTE HetNets,”
to appear, IEEE/ACM Transactions on Networking.
[45] B. Soret, K. Pedersen, T. E. Kolding, H. Kröner, and I. Maniatis,
“Fast muting adaptation for LTE-A HetNets with remote radio heads,”
in Proc., IEEE Globecom, Dec. 2013.
[46] S. Hanly and P. Whiting, “On the capacity of HetNets,” in Proc.,
Information Theory and its Applications (ITA), Feb. 2014.
[47] D. Bethanabhotla, O. Bursalioglu, H. C. Papadopoulos, and G. Caire,
“User association and load balancing for cellular massive MIMO,” in
Proc., Information Theory and its Applications (ITA), Feb. 2014.
[48] E. Aryafar, M. W. A. Keshavarz-Haddad, and M. Chiang, “RAT
selection games in HetNets,” in Proc., IEEE INFOCOM, Apr. 2013,
pp. 998–1006.
[49] L. Gao, G. Iosifidis, J. Huang, L. Tassiulas, and D. Li, “Bargaining-
based mobile data offloading,” IEEE Journal on Sel. Areas in Com-
munications, Sep. 2014.
[50] J. Lee, Y. Kim, H. Lee, B. L. Ng, D. Mazzarese, J. Liu, W. Xiao, and
Y. Zhou, “Coordinated multipoint transmission and reception in LTE-
advanced systems,” IEEE Communications Magazine, vol. 50, no. 11,
pp. 44–50, 2012.
[51] J. Kim, H.-W. Lee, and S. Chong, “Virtual cell beamforming in co-
operative networks,” IEEE Journal on Sel. Areas in Communications,
Sep. 2014.
[52] O. Aliu, A. Imran, M. Imran, and B. Evans, “A survey of self organi-
sation in future cellular networks,” IEEE Communications Surveys and
Tutorials, vol. 15, no. 1, pp. 336–361, Jan. 2013.
[53] B. Kaufman and B. Aazhang, “Cellular networks with an overlaid
device to device network,” in Asilomar Conference on Signals, Systems
and Computers, Pacific Grove, CA, Oct. 2008.
[54] C. Yu, K. Doppler, C. B. Ribeiro, and O. Tirkkonen, “Resource sharing
optimization for device-to-device communication underlaying cellular
networks,” IEEE Wireless Communications, vol. 10, no. 8, pp. 2752–
2763, Aug. 2011.
[55] “Special issue on internet of things,” IEEE Wireless Communications
Magazine, no. 6, Dec. 2010.
[56] N. NaderiAlizadeh and S. Avestimehr, “ITLinQ: a new approach for
spectrum sharing in device-to-device communication systems,” IEEE
Journal on Sel. Areas in Communications, Sep. 2014.
[57] S. Singh, M. Kulkarni, and J. G. Andrews, “A tractable model for rate
in noise limited mmwave cellular networks,” in Proc., IEEE Asilomar,
Nov. 2014.
[58] Connecting America: The National Broadband Plan (Chapter
5), http://download.broadband.gov/plan/national-broadband-plan.pdf,
2010.
[59] Report M.2078, ”Estimated spectrum bandwidth requirements for the
future development of IMT-2000 and IMT-Advanced”, ITU-R, 2006.
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
[60] R. C. Daniels and R. W. Heath, “60 GHz wireless communications:
emerging requirements and design recommendations,” IEEE Vehicular
Techn. Magazine, vol. 2, no. 3, pp. 41–50, Sep. 2007.
[61] S. J. Vaughan-Nicholos, “Gigabit Wi-Fi is on its way,” IEEE Com-
puter, Nov. 2010.
[62] T. Baykas, C.-S. Sum, Z. Lan, and J. Wang, “IEEE 802.15.3c:
The first IEEE wireless standard for data rates over 1 Gb/s,” IEEE
Communications Magazine, vol. 49, no. 7, pp. 114–121, Jul. 2011.
[63] T. S. Rappaport, J. N. Murdock, and F. Gutierrez, “State of the art in
60-GHz integrated circuits and systems for wireless communications,”
Proceedings of the IEEE, vol. 99, no. 8, pp. 1390–1436, Aug. 2011.
[64] M. Marcus and B. Pattan, “Millimeter wave propagation; spectrum
management implications,” IEEE Microwave Magazine, vol. 6, no. 2,
pp. 54–62, Jun. 2005.
[65] A. Alejos, M. G. Sanchez, and I. Cuinas, “Measurement and analysis
of propagation mechanisms at 40 GHz: Viability of site shielding
forced by obstacles,” IEEE Trans. on Veh. Technology, vol. 57, no. 6,
pp. 3369–3380, Nov. 2008.
[66] Z. Pi and F. Khan, “An introduction to millimeter-wave mobile
broadband systems,” IEEE Communications Magazine, vol. 49, no. 6,
pp. 101–107, Jun. 2011.
[67] T. S. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, G. N.
Wong, J. K. Schulz, M. Samimi, and F. Gutierrez, “Millimeter wave
mobile communications for 5G cellular: It will work!” IEEE Access,
vol. 1, no. 1, pp. 335–349, Aug. 2013.
[68] W. Roh et al., “Millimeter-wave beamforming as an enabling tech-
nology for 5G cellular communications: theoretical feasibility and
prototype results,” IEEE Communications Magazine, vol. 52, no. 2,
pp. 106–113, Feb. 2014.
[69] S. Rangan, T. Rappaport, and E. Erkip, “Millimeter-wave cellular
wireless networks: Potentials and challenges,” Proceedings of the
IEEE, vol. 102, no. 3, pp. 366–385, Mar. 2014.
[70] T. S. Rappaport, F. Gutierrez, E. Ben-Dor, J. Murdock, Y. Qiao, and
J. I. Tamir, “Broadband millimeter-wave propagation measurements
and models using adaptive-beam antennas for outdoor urban cellular
communications,” IEEE Trans. on Antennas and Propagation, vol. 61,
no. 4, pp. 1850–1859, Apr. 2013.
[71] M. Kulkarni, S. Singh, and J. G. Andrews, “Coverage and rate trends
in dense urban mmwave cellular networks,” in Submitted to IEEE
Globecom, Dec. 2014.
[72] T. Bai, R. Vaze, and R. W. Heath, “Analysis of blockage effects
on urban cellular networks,” Submitted to IEEE Trans. on Wireless
Communications, arXiv preprint arXiv:1309.4141, 2013.
[73] T. Bai and R. W. Heath Jr, “Coverage analysis for millimeter wave
cellular networks with blockage effects,” in Proc., IEEE Global Signal
and Information Processing Conference, Dec. 2013.
[74] T. Bai, R. Vaze, and R. W. Heath Jr., “Using random shape theory to
model blockage in random cellular networks,” in Proc. of Int. Conf.
on Signal Processing and Communications (SPCOM), Jul. 2012, pp.
1 –5.
[75] A. Ghosh, T. A. Thomas, M. Cudak, R. Ratasuk, P. Moorut, F. W.
Vook, T. Rappaport, J. George R MacCartney, S. Sun, and S. Nie,
“Millimeter wave enhanced local area systems: A high data rate
approach for future wireless networks,” IEEE Journal on Sel. Areas
in Communications, Sep. 2014.
[76] M. R. Akdeniz, Y. Liu, S. Sun, S. Rangan, T. S. Rappaport, and
E. Erkip, “Millimeter wave channel modeling and cellular capacity
evaluation,” IEEE Journal on Sel. Areas in Communications, Sep.
2014.
[77] H. Ishii, Y. Kishiyama, and H. Takahashi, “Novel architecture for
LTE-B: C-plane/U-plane split and phantom cell concept,” in Proc.,
Globecom 2012 Workshop on Emerging Technologies for LTE-A and
B4G, Dec. 2012.
[78] T. Bai and R. W. Heath, “Analysis of millimeter wave cellular networks
with overlaid microwave base stations,” in Asilomar Conference on
Signals, Systems and Computers (ASILOMAR), Pacific Grove, CA,
Nov. 2014.
[79] X. Zhang, A. Molish, and S. Kung, “Variable-phase-shift-based RF-
baseband codesign for MIMO antenna selection,” IEEE Trans. on
Signal Processing, vol. 53, no. 11, pp. 4091–4103, Nov. 2005.
[80] J. Wang, Z. Lan, C. Pyo, T. Baykas, C.-S. Sum, M. A. Rahman, J. Gao,
R. Funada, F. Kojima, H. Harada, and S. Kato, “Beam codebook
14
based beamforming protocol for multi-Gbps millimeter-wave WPAN
systems,” IEEE Journal on Sel. Areas in Communications, vol. 27,
no. 8, pp. 1390–1399, Oct. 2009.
[81] V. Venkateswaran and A.-J. van der Veen, “Analog beamforming in
MIMO communications with phase shift networks and online channel
estimation,” IEEE Trans. on Signal Processing, vol. 58, no. 8, pp.
4131–4143, Aug. 2010.
[82] S. Hur, T. Kim, D. Love, J. Krogmeier, T. Thomas, and A. Ghosh,
“Millimeter wave beamforming for wireless backhaul and access in
small cell networks,” IEEE Trans. on Communications, vol. 61, no. 10,
pp. 4391–4403, Oct. 2013.
[83] O. E. Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. W. H. Jr,
“Spatially sparse precoding in millimeter wave MIMO systems,” to
appear, IEEE Trans. on Wireless Communications.
[84] A. Alkhateeb, O. E. Ayach, G. Leus, and R. W. H. Jr, “Channel esti-
mation and hybrid precoding for millimeter wave cellular systemss,”
Submitted to IEEE Journal of Sel. Topics in Signal Processing, arXiv
preprint arXiv:1401.7426, 2014.
[85] G. Foschini and M. J. Gans, “On limits of wireless communications
in a fading environment when using multiple antennas,” Wireless
Personal Communications, vol. 6, pp. 311–335, Mar. 1998.
[86] E. Telatar, “Capacity of multi-antenna Gaussian channels,” European
Trans. Telecommun., vol. 6, pp. 585–95, Nov-Dec. 1999.
[87] F. R. Farrokhi, A. Lozano, G. J. Foschini, and R. A. Valenzuela,
“Spectral efficiency of FDMA/TDMA wireless systems with transmit
and receive antenna arrays,” IEEE Trans. on Wireless Communications,
vol. 1, no. 4, pp. 591–599, Jan. 2002.
[88] A. Lozano and A. M. Tulino, “Capacity of multiple-transmit multiple-
receive antenna architectures,” IEEE Trans. on Inform. Theory, vol. 48,
no. 12, pp. 3117–3128, 2002.
[89] T. L. Marzetta, “The case for MANY (greater than 16) antennas as
the base station,” in Proc., Information Theory and its Applications
(ITA), San Diego, CA, Jan. 2007.
[90] T. Marzetta, “Noncooperative cellular wireless with unlimited numbers
of base station antennas,” IEEE Trans. on Wireless Communications,
vol. 9, no. 11, pp. 3590–3600, Sep. 2010.
[91] J. Hoydis, S. ten Brink, and M. Debbah, “Massive MIMO in the
UL/DL of cellular networks: How many antennas do we need?” IEEE
Journal on Sel. Areas in Communications, vol. 31, no. 2, pp. 160–171,
Feb. 2013.
[92] F. Rusek, D. Persson, B. K. Lau, E. Larsson, T. Marzetta, O. Edfors,
and F. Tufvesson, “Scaling up MIMO: Opportunities and challenges
with very large arrays,” IEEE Signal Processing Magazine, vol. 30,
no. 1, pp. 40–60, Jan. 2013.
[93] F. Boccardi, R. W. Heath, A. Lozano, T. L. Marzetta, and P. Popovski,
“Five disruptive technology directions for 5G,” IEEE Communications
Magazine, vol. 52, pp. 74–80, Feb. 2014.
[94] M. Matthaiou, G. K. Karagiannidis, E. G. Larsson, T. L. Marzetta, and
R. Schober, “Large-scale multiple antenna wireless systems.” IEEE
Journal on Sel. Areas in Communications, vol. 31, no. 2, pp. 113–
116, Feb. 2013.
[95] Z. Xiang, M. Tao, and X. Wang, “Massive MIMO multicasting in
noncooperative cellular networks,” IEEE Journal on Sel. Areas in
Communications, Sep. 2014.
[96] N. D. Sidiropoulos, T. N. Davidson, and Z.-Q. Luo, “Transmit
beamforming for physical-layer multicasting,” IEEE Trans. on Signal
Processing, vol. 54, no. 6, pp. 2239–2251, 2006.
[97] A. Lozano, “Long-term transmit beamforming for wireless multicast-
ing,” in IEEE Int’l Conf. on Acoustics, Speech and Signal Processing
(ICASSP’07), vol. 3. IEEE, 2007, pp. III–417.
[98] B. Hassibi and B. Hochwald, “How much training is needed in
multiple-antenna wireless links?” IEEE Trans. on Info. Theory, vol. 49,
no. 4, pp. 951–963, Apr. 2003.
[99] N. Jindal and A. Lozano, “A unified treatment of optimum pilot over-
head in multipath fading channels,” IEEE Trans. on Communications,
vol. 58, no. 10, pp. 2939–2948, Oct. 2010.
[100] H. Huh, A. M. Tulino, and G. Caire, “Network MIMO with linear
zero-forcing beamforming: Large system analysis, impact of channel
estimation, and reduced-complexity scheduling,” IEEE Trans. on Info.
Theory, vol. 58, no. 5, pp. 2911–2934, May 2012.
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
[101] A. Ashikhmin and T. L. Marzetta, “Pilot contamination precoding in
multi-cell large scale antenna systems,” in Proc., IEEE Intl. Symposium
on Information Theory, Jul. 2012, pp. 1137–1141.
[102] H. Yin, D. Gesbert, M. Filippou, and Y. Liu, “A coordinated approach
to channel estimation in large-scale multiple-antenna systems,” IEEE
Journal on Sel. Areas in Communications, vol. 31, no. 2, pp. 264–273,
Feb. 2013.
[103] Y. Zeng, R. Zhang, and Z. N. Chen, “Electromagnetic lens-focusing
antenna enabled massive MIMO: performance improvement and cost
reduction,” IEEE Journal on Sel. Areas in Communications, Sep. 2014.
[104] Y.-H. Nam, B. L. Ng, K. Sayana, Y. Li, J. Zhang, Y. Kim, and J. Lee,
“Full-dimension MIMO (FD-MIMO) for next generation cellular tech-
nology,” IEEE Communications Magazine, vol. 51, no. 6, pp. 172–179,
Jun. 2013.
[105] X. Lu, A. Tolli, O. Piirainen, M. Juntti, and W. Li, “Comparison of
antenna arrays in a 3-D multiuser multicell network,” in Proc., IEEE
Intl. Conf. on Communications, Jun. 2011, pp. 1–6.
[106] S. Wu, C. Wang, H. Aggoune, M. Alwakeel, and Y. He, “A non-
stationary 3D wideband twin-cluster model for 5G massive MIMO
channels,” IEEE Journal on Sel. Areas in Communications, Sep. 2014.
[107] 3GPP TSG RAN Plenary No. 58, “Study on 3D-channel model for
elevation beamforming and FD-MIMO studies for LTE,” Dec. 2012.
[108] A. Kammoun, H. Khanfir, Z. Altman, M. Debbah, and M. Kamoun,
“Preliminary results on 3D channel modeling: From theory to stan-
dardization,” IEEE Journal on Sel. Areas in Communications, Sep.
2014.
[109] A. Adhikary, E. A. Safadi, and G. Caire, “Massive MIMO and inter-
tier interference coordination,” in Proc., Information Theory and its
Applications (ITA), Feb. 2014.
[110] J. Chen and V. Lau, “Two-tier precoding for FDD multi-cell massive
MIMO time-varying interference networks,” IEEE Journal on Sel.
Areas in Communications, Sep. 2014.
[111] A. Adhikary, E. Al-Safadi, M. Samimi, R. Wang, G. Caire, T. Rap-
paport, and A. Molisch, “Joint spatial division and multiplexing for
mm-wave channels,” IEEE Journal on Sel. Areas in Communications,
Sep. 2014.
[112] K. S. Gilhousen, I. Jacobs, R. Padovani, A. J. Viterbi, L. Weaver, and
C. Wheatley, “On the capacity of a cellular CDMA system,” IEEE
Trans. on Veh. Technology, vol. 40, no. 2, pp. 303–12, May 1991.
[113] P. Bender, P. Black, M. Grob, R. Padovani, N. Sindhushayana, and
A. Viterbi, “CDMA/HDR: A bandwidth efficient high speed wireless
data service for nomadic users,” IEEE Communications Magazine,
vol. 38, no. 7, pp. 70–77, Jul. 2000.
[114] J. Li, X. Wu, and R. Laroia, OFDMA Mobile Broadband Communi-
cations: A Systems Approach. Cambridge University Press, 2013.
[115] C. E. Shannon, “A mathematical theory of communication (part 1),”
Bell System Tech. Journal, vol. 27, pp. 379–423, 1948.
[116] A. Ghosh, J. Zhang, J. G. Andrews, and R. Muhamed, Fundamentals
of LTE. Prentice-Hall, 2010.
[117] S. Hong, M. Sagong, C. Lim, S. Cho, K. Cheun, and K. Yang, “Fre-
quency and quadrature-amplitude modulation for downlink cellular
OFDMA networks,” IEEE Journal on Sel. Areas in Communications,
Sep. 2014.
[118] A. Barbieri, D. Fertonani, and G. Colavolpe, “Time-frequency packing
for linear modulations: spectral efficiency and practical detection
schemes,” IEEE Trans. on Communications, vol. 57, no. 10, pp. 2951–
2959, Oct. 2009.
[119] F. Rusek and J. Anderson, “Multistream faster than Nyquist signaling,”
IEEE Trans. on Communications, vol. 57, no. 5, pp. 1329–1340, May
2009.
[120] J. Anderson, F. Rusek, and V. Owall, “Faster-than-Nyquist signaling,”
Proceedings of the IEEE, vol. 101, no. 8, pp. 1817–1830, Aug. 2013.
[121] D. Dasalukunte, F. Rusek, and V. Owall, “Multicarrier faster-than-
Nyquist transceivers: Hardware architecture and performance analy-
sis,” IEEE Trans. on Circuits and Systems I, vol. 58, no. 4, pp. 827–
838, Apr. 2011.
[122] B. Farhang-Boroujeny, “OFDM versus filter bank multicarrier,” IEEE
Signal Processing Magazine, vol. 28, no. 3, pp. 92–112, May 2011.
[123] V. Vakilian, T. Wild, F. Schaich, S. ten Brink, and J.-F. Frigon,
“Universal-filtered multi-carrier technique for wireless systems beyond
15
LTE,” in Proc., IEEE Globecom Broadband Wireless Access Workshop,
Dec. 2013.
[124] G. Wunder, P. Jung, M. Kasparick, T. Wild, F. Schaich, Y. Chen,
S. Brink, I. Gaspar, N. Michailow, A. Festag, L. Mendes, N. Cassiau,
D. Ktenas, M. Dryjanski, S. Pietrzyk, B. Eged, P. Vago, and F. Wied-
mann, “5GNOW: non-orthogonal, asynchronous waveforms for future
mobile applications,” IEEE Communications Magazine, vol. 52, no. 2,
pp. 97–105, Feb. 2014.
[125] N. Cassiau, D. Ktenas, and J.-B. Dore, “Time and frequency synchro-
nization for CoMP with FBMC,” in 10th International Symposium on
Wireless Communication Systems, Aug 2013, pp. 1–5.
[126] G. Fettweis, M. Krondorf, and S. Bittner, “GFDM - generalized
frequency division multiplexing,” in Proc., IEEE Veh. Technology
Conf., Apr. 2009, pp. 1–4.
[127] R. Datta, D. Panaitopol, and G. Fettweis, “Analysis of cyclostationary
GFDM signal properties in flexible cognitive radio,” in Proc., Intl.
Symposium on Communications and Info. Technol., Oct. 2012, pp.
663–667.
[128] N. Benvenuto, R. Dinis, D. Falconer, and S. Tomasin, “Single carrier
modulation with nonlinear frequency domain equalization: An idea
whose time has come – again,” Proceedings of the IEEE, vol. 98,
no. 1, pp. 69–96, Jan 2010.
[129] N. Lee, R. W. Heath Jr., D. Morales-Jimenez, and A. Lozano, “Base
station cooperation with dynamic clustering in super-dense cloud-
RAN,” in IEEE GLOBECOM’13 Workshop on Cloud-Processing in
Heterogeneous Mobile Communication Networks, Dec. 2013.
[130] Z. Zhu, P. Gupta, Q. Wang, S. Kalyanaraman, Y. Lin, H. Franke,
and S. Sarangi, “Virtual base station pool: towards a wireless network
cloud for radio access networks,” in Proc., ACM International Con-
ference on Computing Frontiers, 2011, p. 34.
[131] “Network function virtualisation - an introduction, benefit, enablers,
challenges and call for action,” white paper, Oct. 2012. [Online].
Available: http://portal.etsi.org/NFV/NFV White Paper.pdf
[132] “Network function virtualisation - network operator perspectives
on industry progress,” white paper, Oct. 2013. [Online]. Available:
http://portal.etsi.org/NFV/NFV White Paper2.pdf
[133] “AT&T domain 2.0 vision,” white paper, Nov. 2013. [Online].
Available: http://att.com
[134] “Software-defined networking, the new norm for networks,” Apr.
2012. [Online]. Available: http://www.opennetworking.org
[135] S. Sezer, S. Scott-Hayward, P. Chouhan, B. Fraser, D. Lake, J. Foin-
negan, N. Viljoen, M. Miller, and N. Rao, “Are we ready for SDN?
implementation challenges for software-defined networks,” IEEE Com-
munications Magazine, vol. 51, no. 7, pp. 36–43, Jul. 2013.
[136] [Online]. Available: http://www.greentouch.org
[137] “Special issue on energy-efficient wireless communications,” IEEE
Journal on Sel. Areas in Communications, vol. 29, no. 8, Sep. 2011.
[138] “Special issue on energy efficiency in communications,” IEEE Com-
munications Magazine, vol. 48, no. 11, Nov. 2010.
[139] G. Auer, V. Giannini, C. Desset, I. Godor, P. Skillermark, M. Olsson,
M. Imran, D. Sabella, M. Gonzalez, O. Blume, and A. Fehske, “How
much energy is needed to run a wireless network?” IEEE Wireless
Communications, vol. 18, no. 5, pp. 40–49, Oct. 2011.
[140] C. Saraydar, N. B. Mandayam, and D. Goodman, “Efficient power
control via pricing in wireless data networks,” IEEE Trans. on Com-
munications, vol. 50, no. 2, pp. 291–303, Feb. 2002.
[141] F. Meshkati, H. Poor, and S. Schwartz, “Energy-efficient resource
allocation in wireless networks,” IEEE Signal Processing Magazine,
vol. 24, no. 3, pp. 58–68, May 2007.
[142] S. Buzzi and H. Poor, “Joint receiver and transmitter optimization for
energy-efficient CDMA communications,” IEEE Journal on Sel. Areas
in Communications, vol. 26, no. 3, pp. 459–472, Apr. 2008.
[143] S. Buzzi, G. Colavolpe, D. Saturnino, and A. Zappone, “Potential
games for energy-efficient power control and subcarrier allocation in
uplink multicell OFDMA systems,” IEEE Journal of Sel. Topics in
Signal Processing, vol. 6, no. 2, pp. 89–103, Apr. 2012.
[144] G. Miao, N. Himayat, Y. Li, and D. Bormann, “Energy efficient design
in wireless OFDMA,” in Proc., IEEE Intl. Conf. on Communications,
May 2008, pp. 3307–3312.
[145] B. L. Ng, Y. Kim, J. Lee, Y. Li, Y.-H. Nam, J. C. Zhang, and K. Sayana,
“Fulfilling the promise of massive MIMO with 2D active antenna
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
array,” in Proc., IEEE Globecom Workshop on Emerging Technologies
for LTE-Advanced and Beyond-4G, Dec. 2012, pp. 691–696.
[146] L. Venturino, A. Zappone, C. Risi, and S. Buzzi, “Energy-efficient
scheduling and power allocation in downlink OFDMA networks
with base station coordination,” to appear, IEEE Trans. on Wireless
Communications, 2014.
[147] S. He, Y. Huang, H. Wang, S. Jin, and L. Yang, “Leakage-aware
energy-efficient beamforming for heterogeneous multicell multiuser
systems,” IEEE Journal on Sel. Areas in Communications, Sep. 2014.
[148] G. Koutitas, “Low carbon network planning,” in Proc., European
Wireless Conference, Apr. 2010, pp. 411–417.
[149] E. Oh and B. Krishnamachari, “Energy savings through dynamic base
station switching in cellular wireless access networks,” in Proc., IEEE
Globecom, Dec 2010, pp. 1–5.
[150] S. Zhou, J. Gong, Z. Yang, Z. Niu, and P. Yang, “Green mobile access
network with dynamic base station energy saving,” in Proc., ACM
MobiCom, vol. 9, no. 262, 2009, pp. 10–12.
[151] Z. Niu, Y. Wu, J. Gong, and Z. Yang, “Cell zooming for cost-efficient
green cellular networks,” IEEE Communications Magazine, vol. 48,
no. 11, pp. 74–79, November 2010.
[152] A. Conte, A. Feki, L. Chiaraviglio, D. Ciullo, M. Meo, and M. Marsan,
“Cell wilting and blossoming for energy efficiency,” IEEE Wireless
Communications, vol. 18, no. 5, pp. 50–57, October 2011.
[153] M. Marsan, G. Bucalo, A. Di Caro, M. Meo, and Y. Zhang, “Towards
zero grid electricity networking: Powering BSs with renewable energy
sources,” in Proc., IEEE Intl. Conf. on Communications Workshop
on Green Broadband Access: Energy Efficient Wireless and Wired
Network Solutions, June 2013, pp. 596–601.
[154] H. S. Dhillon, Y. Li, P. Nuggehalli, Z. Pi, and J. G. Andrews, “Funda-
mentals of heterogeneous cellular networks with energy harvesting,”
to appear, IEEE Trans. on Wireless Communications.
[155] D. Ng, E. Lo, and R. Schober, “Energy-efficient resource allocation in
OFDMA systems with hybrid energy harvesting base station,” IEEE
Trans. on Wireless Communications, vol. 12, no. 7, pp. 3412–3427,
Jul. 2013.
[156] P. Skillermark and P. Frenger, “Enhancing energy efficiency in lte with
antenna muting,” in Proc., IEEE Veh. Technology Conf., May 2012,
pp. 1–5.
[157] C.-L. I, C. Rowell, S. Han, Z. Xu, G. Li, and Z. Pan, “Toward green
and soft: a 5G perspective,” IEEE Communications Magazine, vol. 52,
no. 2, pp. 66–73, Feb. 2014.
[158] L. Liu, H. Wang, X. Liu, X. Jin, W. B. He, Q. B. Wang, and
Y. Chen, “Greencloud: A new architecture for green data center,” in
Proc., Intl. Conference Industry Session on Autonomic Computing and
Communications Industry Session, New York, NY, USA, 2009, pp. 29–
38. [Online]. Available: http://doi.acm.org/10.1145/1555312.1555319
[159] C. Li, J. Zhang, and K. B. Letaief, “Throughput and energy efficiency
analysis of small cell networks with multi-antenna base stations,” to
appear, IEEE Trans. on Wireless Communications.
[160] S. Rangan, “Energy efficient methods for energy efficient methods for
millimeter wave picocellular systems,” in Proc., IEEE Communica-
tions Theory Workshop, Phuket, Thailand, Jun. 2013.
[161] R. Coase, “The federal communications commission,” Journal of Law
and Economics, vol. 2, no. 1, pp. 1–40, Oct. 1959.
[162] Y. Benkler, “Open wireless vs. licensed spectrum: Evidence from
market adoption,” Harvard Journal of Law and Technology, vol. 26,
no. 1, pp. 71–163, Oct. 2012.
[163] M. Matinmikko, M. Palola, H. Saarnisaari, M. Keikkila, J. Prokkola,
T. Kippola, T. Hanninen, M. Jokinen, and S. Yrjola, “Cognitive radio
trial environment: First live authorized shared access-based spectrum
sharing demonstration,” Proc., IEEE Veh. Technology Magazine, vol. 8,
no. 3, pp. 30–37, Sep. 2013.
[164] J. Khun-Jush, P. Bender, B. Deschamps, and M. Gundlach, “Licensed
shared access as complementary approach to meet spectrum demands:
Benefits for next generation cellular systems,” ETSI Workshop on
Reconfigurable Radio Systems, Cannes, France, pp. 1–7, Dec. 2013.
[165] H. Zhou, R. Berry, M. Honig, and R. Vohra, “Complexity of allocation
problems in spectrum markets with interference complementarities,”
IEEE Journal on Sel. Areas in Communications, vol. 31, no. 3, pp.
489–499, Mar. 2013.
16
[166] R. Berry, M. Honig, and R. Vohra, “Spectrum markets: Motiva-
tion, challenges, and implications,” IEEE Communications Magazine,
vol. 48, no. 11, pp. 146–155, Nov. 2010.
[167] T. Nguyen, H. Zhou, R. Berry, M. Honig, and R. Vohra, “The impact
of additional unlicensed spectrum on wireless services competition,” in
Proc., IEEE Intl. Symposium on New Frontiers in Dynamic Spectrum
Access Networks (DYSPAN), May 2011, pp. 146–155.
[168] H. Zhou, R. Berry, M. Honig, and R. Vohra, “Investment and competi-
tion in unlicensed spectrum,” in Proc., 46th Conference on Information
Sciences and Systems (CISS), Mar. 2012, pp. 1–6.
[169] ITU-R 2020 Vision Workshop, “IMT vision towards 2020 and beyond
” development”,” in IMT-2020 (5G) China Promotion Group, Feb.
2014.
[170] ——, “Views on IMT beyond 2020,” in ARIB 2020 and Beyond Ad
Hoc, Feb. 2014.
[171] ——, “5G vision and requirements of 5G forum, Korea,” in 5G Forum
Korea, Feb. 2014.
[172] METIS ICT-317669-METIS/D1.1, “Scenarios, requirements and KPIs
for 5G mobile and wireless system,” available at www.metis2020.com,
Apr. 2013.
[173] METIS ICT-317669-METIS/D4.1, “Summary on preliminary trade-off
investigations and first set of potential network-level solutions,” Sep.
2013. [Online]. Available: www.metis2020.com
[174] ITU-R WP5D, “Work plan, timeline, process and deliverables for
“IMT-2020” development,” Feb. 2014.
[175] A. Scrase, “Summit key messages,” Nov. 2013. [On-
line]. Available: http://docbox.etsi.org/Workshop/2013/201311
FUTUREMOBILESU\MMIT/ZZ CONCLUSIONS.pdf
[176] ITU-R WP5D, “Working document towards a preliminary draft: New
report ITU-R M.[IMT.above 6 GHz],” Feb. 2014.
[177] D.-E. Meddour, T. Rasheed, and Y. Gourhant, “On the role of infras-
tructure sharing for mobile network operators in emerging markets,”
Computer Networks, vol. 55, no. 7, pp. 1576–1591, May 2011.
[178] A. Ghanbari, J. Markendahl, and A. Widaa, “Cooperation patterns in
smallcell networks - risks and opportunities to distinguish the win-win
model,” in Proc., 24th European Regional ITS Conference, Oct. 2013.
[179] R. Berry, M. Honig, T. Nguyen, V. Subramanian, H. Zhou, and
R. Vohra, “On the nature of revenue-sharing contracts to incentivize
spectrum-sharing,” Proc., IEEE INFOCOM, pp. 845–853, 2013.
[180] B. Leng, P. Mansourifard, and B. Krishnamachari, “Microeconomic
analysis of base-station sharing in green cellular networks,” Proc.,
IEEE INFOCOM, 2014.
[181] W.-C. Liao, M. Hong, H. Farmanbar, X. Li, Z.-Q. Luo, and H. Zhang,
“Min flow rate maximization for software defined radio access net-
works,” IEEE Journal on Sel. Areas in Communications, Sep. 2014.
[182] Y. Zhou and W. Yu, “Optimized backhaul compression for uplink cloud
radio access network,” IEEE Journal on Sel. Areas in Communications,
Sep. 2014.
[183] N. Golrezaei, K. Shanmugam, A. G. Dimakis, A. F. Molisch, and
G. Caire, “Femtocaching: Wireless video content delivery through
distributed caching helpers,” in Proc., IEEE INFOCOM, Mar. 2011,
pp. 1107–1115.
IEEE JSAC SPECIAL ISSUE ON 5G WIRELESS COMMUNICATION SYSTEMS
Jeffrey G. Andrews [S’98, M’02, SM’06, F’13]
received the B.S. in Engineering with High Dis-
tinction from Harvey Mudd College, and the M.S.
and Ph.D. in Electrical Engineering from Stanford
University. He is the Cullen Trust Endowed Professor
(#1) of ECE at the University of Texas at Austin,
Editor-in-Chief of the IEEE Transactions on Wireless
Communications, and Technical Program Co-Chair
of IEEE Globecom 2014. He developed CDMA sys-
tems at Qualcomm from 1995-97, and has consulted
for entities including Verizon, the WiMAX Forum,
Intel, Microsoft, Apple, Samsung, Clearwire, Sprint, and NASA. He is a
member of the Technical Advisory Boards of Accelera and Fastback Networks,
and co-author of the books Fundamentals of WiMAX and Fundamentals of
LTE.
Dr. Andrews received the National Science Foundation CAREER award in
2007 and has been co-author of nine best paper award recipients: ICC 2013,
Globecom 2006 and 2009, Asilomar 2008, the 2010 IEEE Communications
Society Best Tutorial Paper Award, the 2011 IEEE Heinrich Hertz Prize, the
2014 EURASIP Best Paper Award, the 2014 IEEE Stephen O. Rice Prize, and
the 2014 IEEE Leonard G. Abraham Prize. He is an elected member of the
Board of Governors of the IEEE Information Theory Society.
Stefano Buzzi [M’98, SM’07] is currently an As-
sociate Professor at the University of Cassino and
Lazio Meridionale, Italy. He received his Ph.D. de-
gree in Electronic Engineering and Computer Sci-
ence from the University of Naples ”Federico II”
in 1999, and he has had short-term visiting ap-
pointments at the Dept. of Electrical Engineering,
Princeton University, in 1999, 2000, 2001 and 2006.
His research and study interest lie in the wide area
of statistical signal processing and resource alloca-
tion for communications, with emphasis on wireless
communications; he is author/co-author of more than 50 journal papers and 90
conference papers; Dr. Buzzi was awarded by the Associazione Elettrotecnica
ed Elettronica Italiana (AEI) the ”G. Oglietti” scholarship in 1996, and was
the recipient of a NATO/CNR advanced fellowship in 1999 and of three CNR
short-term mobility grants. He is a former Associate Editor for the IEEE
Communications Letters, and the IEEE Signal Processing Letters.
Wan Choi [M’06, SM’12] received the B.Sc. and
M.Sc. degrees from the School of Electrical En-
gineering and Computer Science (EECS), Seoul
National University (SNU), Seoul, Korea, in 1996
and 1998, respectively, and the Ph.D. degree in the
Department of Electrical and Computer Engineering
at the University of Texas at Austin in 2006. He is
currently an Associate Professor of the Department
of Electrical Engineering, Korea Advance Institute of
Science and Technology (KAIST), Daejeon, Korea.
From 1998 to 2003, he was a Senior Member of
the Technical Staff of the R&D Division of KT Freetel, Korea, where he
researched 3G CDMA systems.
Dr. Choi is the recipient of IEEE Vehicular Technology Society Jack
Neubauer Memorial Award in 2002. He also received the IEEE Vehicular
Technology Society Dan Noble Fellowship Award in 2006 and the IEEE Com-
munication Society Asia Pacific Young Researcher Award in 2007. He serves
as Associate Editor for the IEEE Transactions on Wireless Communications,
for the IEEE Transactions on Vehicular Technology, and for IEEE Wireless
Communications Letters.
17
Stephen V. Hanly [M’98] received a B.Sc. (Hons)
and M.Sc. from the University of Western Australia,
and the Ph.D. degree in mathematics in 1994 from
Cambridge University, UK. From 1993 to 1995, he
was a Post-doctoral member of technical staff at
AT&T Bell Laboratories. From 1996-2009 he was at
the University of Melbourne, and from 2010-2011
he was at the National University of Singapore. He
now holds the CSIRO-Macquarie University Chair in
Wireless Communications at Macquarie University,
Sydney, Australia. He has been an Associate Editor
for IEEE Transactions on Wireless Communications, Guest Editor for IEEE
Journal on Selected Areas in Communications, and Guest Editor for the
Eurasip Journal on Wireless Communications and Networking. In 2005 he was
the technical co-chair for the IEEE International Symposium on Information
Theory held in Adelaide, Australia. He was a co-recipient of the best paper
award at the IEEE Infocom 1998 conference, and the 2001 Joint IEEE
Communications Society and IEEE Information Theory Society best paper
award.
Angel Lozano [S’90, M’99, SM’01, F’14] received
the M.Sc. and Ph.D. degrees in Electrical Engineer-
ing from Stanford University in 1994 and 1998,
respectively. He is a Professor and the Vice-Rector
for Research at Universitat Pompeu Fabra (UPF) in
Barcelona, Spain. He was with Bell Labs (Lucent
Technologies, now Alcatel-Lucent) between 1999
and 2008, and served as Adj. Associate Professor
at Columbia University between 2005 and 2008.
Prof. Lozano is an Associate Editor for the IEEE
Transactions on Information Theory, the Chair of the
IEEE Communication Theory Technical Committee, and an elected member
to the Board of Governors of the IEEE Communications Society. His papers
have received two awards: ISSSTA 2006 and the 2009 IEEE Stephen O. Rice
prize.
Anthony C. K. Soong [S’88, M’91, SM’02, F’14]
received the B.Sc. degree in animal physiology and
physics from the University of Calgary, and the B.Sc.
degree in electrical engineering, the M.Sc. degree
in biomedical physics and Ph.D. degree in electrical
and computer engineering from the University of Al-
berta. He is currently the chief scientist for wireless
research and standards at Huawei Technologies Co.
Ltd, in the US. He serves as the vice-chair for 3GPP2
TSG-C WG3. Prior to joining Huawei, he was with
the systems group for Ericsson Inc and Qualcomm
Inc. Dr. Soong has published numerous scientific papers and has over 80
patents granted or pending. He was the corecipient of the 2013 IEEE Signal
Processing Society Best Paper Award.
Jianzhong (Charlie) Zhang [S’96, M’02, SM’09] is
currently senior director and head of Wireless Com-
munications Lab with Samsung Research America
at Dallas, where he leads technology development,
prototyping and standardization for Beyond 4G and
5G wireless systems. From Aug 2009 to Aug 2013,
he served as the Vice Chairman of the 3GPP RAN1
working group and led development of LTE and
LTE-Advanced technologies such as 3D channel
modeling, UL-MIMO and CoMP, Carrier Aggrega-
tion for TD-LTE, etc. Before joining Samsung, he
was with Motorola from 2006 to 2007 working on 3GPP HSPA standards, and
with Nokia Research Center from 2001 to 2006 working on IEEE 802.16e
(WiMAX) standard and EDGE/CDMA receiver algorithms. He received his
Ph.D. degree from University of Wisconsin, Madison.