scispace - formally typeset

Journal ArticleDOI

5G Uniform Linear Arrays With Beamforming and Spatial Multiplexing at 28, 37, 64, and 71 GHz for Outdoor Urban Communication: A Two-Level Approach

17 Aug 2017-IEEE Transactions on Vehicular Technology (Institute of Electrical and Electronics Engineers (IEEE))-Vol. 66, Iss: 11, pp 9972-9985

TL;DR: A two-level beamforming architecture for uniform linear arrays is proposed that leverages the formation of two or more spatial lobes for the angles-of-departure and angles- of-arrival even for line-of the-sight (LOS) transmission.

AbstractMultiple-input multiple-output (MIMO) spatial multiplexing and beamforming are regarded as key technology enablers for the fifth-generation (5G) millimeter wave (mmWave) mobile radio services. Spatial multiplexing requires sufficiently separated and incoherent antenna array elements, while in the case of beamforming, the antenna array elements need to be coherent and closely spaced. Extensive 28-, 60-, and 73-GHz ultra-wideband propagation measurements in cities of New York City and Austin have indicated formation of two or more spatial lobes for the angles-of-departure and angles-of-arrival even for line-of-sight (LOS) transmission, which is an advantageous feature of mmWave channels, indicating that the transmitting and receiving array antenna elements can be co-located, thus enabling a single architecture for both spatial multiplexing and beamforming. In this paper, a two-level beamforming architecture for uniform linear arrays is proposed that leverages the formation of these spatial lobes. The antenna array is composed of sub-arrays, and the impact of sub-array spacing on the spectral efficiency is investigated through simulations using a channel simulator named NYUSIM developed based on extensive measured data at mmWave frequencies. Simulation results indicate spectral efficiencies of 18.5–28.1 bits/s/Hz with a sub-array spacing of 16 wavelengths for an outdoor mmWave urban LOS channel. The spectral efficiencies obtained are for single-user (SU) MIMO transmission at the recently allocated 5G carrier frequencies in July 2016. The method and results in this paper are useful for designing antenna array architectures for 5G wireless systems.

Topics: Antenna array (64%), Smart antenna (63%), Spatial multiplexing (63%), Beamforming (62%), Antenna (radio) (59%)

Summary (2 min read)

Introduction

  • Abstract—Multiple-input multiple-output (MIMO) spatial multiplexing and beamforming are regarded as key technology enablers for the fifth-generation (5G) millimeter wave mobile radio services.
  • Hybrid beamforming can be employed by having a group of elements connected to one RF chain or an array of sub-arrays, where each sub-array has several interconnected antenna elements but its own RF chain.
  • Phase shifts are digitally controlled by realizing discrete phase shifts which causes quantization phase error leading to formation of quantization lobes, which occur at the grating lobe angles during beam steering.

A. Current Limitations in SU-MIMO

  • In the case of SU-MIMO, more than one spatial stream is exchanged between two arrays.
  • The antenna spacing and array orientation can significantly affect the system performance [19], [20].
  • Employing a specific hybrid precoding technique or comparison of the existing hybrid precoding techniques such as for optimizing the hardware resources in mmWave MIMO is not the focus of this paper.
  • Rather, their paper provides an alternative approach to the Rayleigh distance criterion that enables SM for SU-MIMO in LOS mmWave channels.
  • The existing 3GPP channel models [1][45] do not include the effects of directional local scattering at the Tx and Rx in an outdoor urban environment for mmWave propagation, yet real-world measurements in New York City show the existence of directional propagation in urban environments, leading to the formation of SLs [2], [46], [47].

B. System-Level Architecture

  • In order to meet the contradictory requirements of beamforming which requires co-polarized closely spaced antenna elements typically at 𝜆𝜆 2 with high coherence, and SM which requires no coherence between antenna elements to ensure simultaneous separate parallel data streams [19], a twolevel (2L) hybrid beamforming architecture is proposed as illustrated in Fig.
  • Under low SNR conditions when employing beamforming, the same architecture can be configured with the sub-arrays resulting in formation of a larger array, which the authors call “super-array” in this paper.
  • MMWAVE CHANNEL MODEL A. 3D Statistical Spatial Channel Model.
  • With an increase in the number of sub-array elements, the azimuth HPBW becomes less than 6.3° (±3.3°) and is taken as 7° which is the lower limit in NYUSIM v1.5.
  • The number of independent non-zero rows and columns depends on the amount of scattering, reflection and the length 𝐿𝐿𝑦𝑦 of the Tx and Rx arrays.

C. MIMO Channel Matrix Condition Number

  • The channel condition number and its statistical properties are important for characterization of the MIMO channel.
  • The correlation between the channel paths increases if angular separation of the channel paths decreases.
  • A low channel condition number usually corresponds to a high rank and vice versa; the matrix has full rank (the highest rank) when the channel condition number is equal or close to 0 dB (the lowest theoretical condition number).
  • The channel condition number is an important design parameter in MIMO systems as it has been shown to drastically affect the detection, error and performance of linear Rxs in MIMO systems.
  • Performance of linear detectors such as zero-forcing (ZF), maximum-likelihood (ML) and minimum mean square error (MMSE) detectors has been investigated indicating strong dependence of the detector performance on the channel condition number [64]-[70].

A. Sub-Array Antenna Element

  • The azimuth and elevation sub-array radiation pattern would depend on 𝑀𝑀 in each sub-array and their individual element radiation pattern (𝐸𝐸(𝜃𝜃)) .
  • Both circular and rectangular patches have similar gain, beam position and efficiency.
  • A circular pin-fed antenna patch is employed in this paper and its design parameters are given in Fig. 5 [80], where 𝐷𝐷, 𝑆𝑆𝜋𝜋 , 𝑅𝑅, 𝐻𝐻, 𝜖𝜖𝑟𝑟 and 𝑡𝑡𝑡𝑡𝑚𝑚𝑡𝑡 are the patch diameter, feed offset, feed pin radius, substrate height, relative permittivity and loss tangent due to the substrate medium respectively.
  • The dominant mode is 𝑇𝑇𝑀𝑀11 and the radiation pattern is a single lobe with maximum in the direction normal to the plane of the antenna.

B. Super-Array Element Spacing and Grating Lobes

  • 𝐵𝐵𝑚𝑚𝑡𝑡𝑒𝑒 𝑗𝑗2𝜋𝜋𝜆𝜆 𝑑𝑑 𝑚𝑚𝑚𝑚𝑚𝑚𝛿𝛿𝑁𝑁𝑇𝑇𝑚𝑚𝑡𝑡=1 (7) where ∑ 𝐵𝐵𝑚𝑚𝑡𝑡𝑒𝑒 𝑗𝑗2𝜋𝜋𝜆𝜆 𝑑𝑑 𝑚𝑚𝑚𝑚𝑚𝑚𝛿𝛿𝑁𝑁𝑇𝑇𝑚𝑚𝑡𝑡=1 is the array factor 𝐴𝐴𝐴𝐴𝑆𝑆𝐴𝐴for the super-array.
  • 𝐴𝐴𝐴𝐴𝑆𝑆𝑆𝑆𝑆𝑆𝐴𝐴)𝐴𝐴𝐴𝐴𝑆𝑆𝐴𝐴 (8) The array pattern 𝐴𝐴(𝜃𝜃) of 𝐶𝐶𝑀𝑀4 super-array with two subarrays and elements with radiation pattern 𝐸𝐸(𝜃𝜃) obtained is plotted in Fig. 6 with MATLAB R2016a® where the subarray separation is 𝑑𝑑 = 2𝜆𝜆 .
  • The elevation HPBW remains constant and is taken as 45° that is the upper limit in NYUSIM v1.5.
  • As the number of MIMO channels increases for the same lateral separation of the sub-arrays, the channel condition number increases indicating unfavourable propagation conditions for MIMO SM.

A. Channel Condition Number

  • With a view to improve the channel condition number for 𝐶𝐶𝑀𝑀4 in Fig. 7, 𝑑𝑑 is increased from 2𝜆𝜆 to 4𝜆𝜆 for the 5×5 MIMO 28 GHz channel.
  • The channel condition number decreases for super-array with a higher number of elements as this increases the sub-array spacing indicating favourable MIMO propagation even for further increases in number of 𝑁𝑁𝑅𝑅 × 𝑁𝑁𝑇𝑇 elements at L2.
  • A. Poon and M. Taghivand, “Supporting and enabling circuits for antenna arrays in wireless communications,” Proc. IEEE, vol.

Did you find this useful? Give us your feedback

...read more

Content maybe subject to copyright    Report

AbstractMultiple-input multiple-output (MIMO) spatial
multiplexing and beamforming are regarded as key
technology enablers for the fifth-generation (5G) millimeter
wave (mmWave) mobile radio services. Spatial multiplexing
requires sufficiently separated and incoherent antenna
array elements, while in the case of beamforming, the
antenna array elements need to be coherent and closely
spaced. Extensive 28-, 60-, and 73-GHz ultra-wideband
propagation measurements in cities of New York City and
Austin have indicated formation of two or more spatial
lobes for the angles-of-departure and angles-of-arrival even
for line-of-sight (LOS) transmission, which is an
advantageous feature of mmWave channels, indicating that
the transmitting and receiving array antenna elements can
be co-located, thus enabling a single architecture for both
spatial multiplexing and beamforming. In this paper a two-
level beamforming architecture for uniform linear arrays is
proposed that leverages the formation of these spatial lobes.
The antenna array is composed of sub-arrays, and the
impact of sub-array spacing on the spectral efficiency is
investigated through simulations using a channel simulator
named NYUSIM developed based on extensive measured
data at mmWave frequencies. Simulation results indicate
spectral efficiencies of 18.5-28.1 bits/s/Hz with a sub-array
spacing of 16 wavelengths for an outdoor mmWave urban
LOS channel. The spectral efficiencies obtained are for
single-user (SU) MIMO transmission at the recently
allocated 5G carrier frequencies in July 2016. The method
and results in this paper are useful for designing antenna
array architectures for 5G wireless systems.
Index TermsArrays, beamforming, 5G, mmWave, spatial
multiplexing, SU-MIMO
Manuscript received March 16, 2017, revised July 10, 2017, accepted
August 13, 2017.
This work was supported in part by the Engineering and
Physical Sciences Research Council under Grant EP/K031953/1. Copyright (c)
2017 IEEE. Personal use of this material is permitted. However, permission to
use this material for any other purposes must be obtained from the IEEE by
sending a request to pubs-permissions@ieee.org.
Jaswinder Lota is with the University of East London, London UK (e-mail
j.lota@uel.ac.uk) and is an honorary faculty member with the Department of
I. INTRODUCTION
pectrum allocation for fifth-generation (5G) cellular
systems are classified by the 3
rd
Generation Partnership
Project (3GPP) international cell phone standard body as being
from 0.5 GHz up to 100 GHz [1]. The spectrum above
30 GHz
is known as the millimetre wave (mmWave) band, with 28 GHz
also being regarded as mmWave for its proximity to 30 GHz
spectrum. The vast spectrum at frequencies above 28 GHz
offers wide channel bandwidths that will support high peak data
rates of several Gigabits per second. Such throughput speeds
will be required for high-definition (HD) video, low latency
content, and high data rate transfer between data centers and
virtual interaction between people and machines [2]. Video
continues to be the major application generator for mobile data
traffic growth accounting for 51 percent of global mobile data
traffic in 2012, and it is predicted to account for 75 percent of
global mobile data traffic by 2020 [3]. The 28 GHz band is
attractive as it enables mobility on mmWave due to 850 MHz
of contiguous bandwidth in the United States, and has been a
frequency of major focus for academic research and
prototyping efforts; whereas the 38 GHz band [4] is particularly
suited for ultra-high data rates and has initial agreement from
the International Telecommunications Union (ITU) World
Administration Conference as a global spectrum allocation. The
64-71 GHz spectrum would enable development of new
innovative unlicensed applications and promote next generation
high-speed wireless links with wider connectivity and higher
throughput [5]. The Federal Communications Commission
(FCC) in July 2016 allocated the 28 GHz, 37/39 GHz and 64-
71 GHz frequencies both as licensed and unlicensed bands for
the 5G mobile radio services (MRS) [5].
Multiple-input multiple-output (MIMO) has already been
used in 4G long term evolution (LTE) cellular networks, and is
regarded as one of the technologies likely to be adopted in 5G
Electronic & Electrical Engineering, University College London, London
WC1E 7JE UK (e-mail: j.lota@ucl.ac.uk). S. Sun and T. S. Rappaport are with
NYU WIRELESS and Tandon School of Engineering, New York University,
Brooklyn, NY 11201 USA (e-mail: ss7152@nyu.edu; tsr@nyu.edu). Andreas
Demosthenous is with the Electronic & Electrical Engineering Department,
University College London, London WC1E 7JE UK (e-mail:
a.demosthenous@ucl.ac.uk).
5G Uniform Linear Arrays with Beamforming
and Spatial Multiplexing at 28 GHz, 37 GHz, 64
GHz and 71 GHz for Outdoor Urban
Communication: A Two-Level Approach
Jaswinder Lota, Senior Member, IEEE, Shu Sun, Student Member, IEEE,
Theodore S. Rappaport, Fellow, IEEE, Andreas Demosthenous, Senior Member, IEEE
S

to increase peak data rates along with beamforming for low
signal-to-noise ratio (SNR) scenarios such as cell edge users
[6]-[8]. Beamforming approaches that are suitable for mmWave
frequencies can be broadly classified as analog, hybrid analog-
digital, and low resolution digital, each having specific
implications for deployment in mmWave MIMO channels.
Analog beamforming requires analog phase shifters which are
adaptively adjusted to vary the phases of antenna elements,
thereby increasing the antenna gain to counter the path loss in
line-of-sight (LOS) and non-line-of-sight (NLOS) mmWave
propagation [9]. Phase shifters can be active elements, which
suffer phase-shifter loss, noise and nonlinearity, or they can be
passive, where passive phase shifters have the advantage of low
power consumption and reduced nonlinearity, but incur more
insertion loss and occupy more area [10]. Analog beamforming
is employed with beam training algorithms and acquiring the
channel state information (CSI). Hybrid (analog-digital)
beamforming requires either precoding or combining
techniques both in the baseband and radio-frequency (RF) level
[11]-[14] and can be designed through different approaches
using phase shifters [15], switches [16] and lenses [17]. Hybrid
beamforming can be employed by having a group of elements
connected to one RF chain or an array of sub-arrays, where each
sub-array has several interconnected antenna elements but its
own RF chain.
MmWave MIMO transmission requires multiple antenna
elements to provide beamforming gain to compensate for
higher path loss on account of mmWave incurring higher
attenuation in the first meter of propagation due to Friis law
[18]. Making a highly directive antenna with small beamwidths,
steerable over large angle ranges for the angle of arrival (AOA)
at the receiver (Rx), would ensure a high gain. This adaptive
beamforming requires multiple elements with high coherence
to enable beamforming and with beam steering that requires co-
polarized antenna elements closely spaced typically at
,
where is the carrier wavelength. MIMO transmission requires
spatial multiplexing (SM) which is ensured by separate spatial
paths of parallel transmissions, which mandates a contrary
requirement of ensuring that there is no coherence between
antenna elements transmitting parallel data streams
simultaneously, i.e. antenna elements that are either cross-
polarized, orthogonal in spatial beam patterns, and/or relatively
spaced far apart [19]. This requires an antenna element spacing
of greater than
which reduces coherence among antenna
elements, but leads to formation of grating and quantization
lobes which reduce the available angle range for beam steering.
Continuous phase shifting requirements from 0° to 360° are
expensive and typically not used in practice. Phase shifts are
digitally controlled by realizing discrete phase shifts which
causes quantization phase error leading to formation of
quantization lobes, which occur at the grating lobe angles
during beam steering. Meeting both MIMO and beam steering
requirements simultaneously in a single architecture is
challenging. In this paper these are explained followed by an
architecture that addresses these challenges. Simulations are
undertaken with a mmWave channel simulator developed by
New York University (NYU) from extensive field data,
NYUSIM v1.5, to characterize the MIMO channel conditions
at the recently allocated FCC 5G frequencies of 28, 37/39, and
64-71 GHz. Section II details current limitations in single-user
(SU)-MIMO with beam steering along with the proposed
hybrid architecture. The mmWave MIMO channel model is
described in Section III, based on a 3D statistical spatial channel
model (SSCM) along with how the channel coefficients are
obtained for the MIMO channel matrix H. The section details
degrees of freedom (DOFs) and channel condition number for
H that quantify SM required for MIMO transmission. Section
IV details the uniform linear array (ULA) element design and
the array pattern, formation of the grating and quantization
lobes. The channel condition number and spectral efficiency
values obtained from simulations for various sub-array spacings
are given in Section V, followed by conclusions in Section VI.
II.
TWO-LEVEL HYBRID BEAMFORMING WITH SPATIAL
MULTIPLEXING
A. Current Limitations in SU-MIMO
In the case of SU-MIMO, more than one spatial stream is
exchanged between two arrays. Due to the multipath sparsity, the
channel propagation matrix can be near-singular and
conventional MIMO capacity will degrade significantly. The
antenna spacing and array orientation can significantly affect the
system performance [19], [20]. SU-MIMO capacity for a full
digital mmWave array has been studied in [21] for LOS and in
[20] for a two-path channel. When the LOS-path is dominant,
multiplexing gain is largely limited to the gain achievable by
LOS-MIMO, which relies on careful placement of Tx and Rx
antennas. For a full digital array, the LOS-MIMO capacity,
which can be achieved at the Rayleigh distance, depends on the
orientation of Tx and Rx arrays, their distance R, the element
spacing and the number of antenna elements. The Rayleigh
distance criterion leads to a full-rank and orthogonal MIMO
channel matrix, but generally requires impractically large
antenna space and array size. Nevertheless, results in [20][27]
indicated that in principle it is possible to achieve the maximum
multiplexing gain in mmWave MIMO channels with LOS
transmissions by carefully designing the geometrical distribution
of the antennas at both link ends. Using aligned ULAs at both
ends showed that the channel vectors experienced by different
Tx/Rx antennas can be mutually orthogonal if the antenna
spacings and the end-to-end distance satisfy the Rayleigh
distance criterion. This approach of Rayleigh distance criterion,
however, relies on careful placement of Tx and Rx and is not
practical for outdoor urban communication, which may require
varying distances between Tx and Rx based on the different
requirements of cell design and geometry. For a setup with a
carrier frequency of 38 GHz, two parallel ULAs with 16
elements, and R = 500 meters, achieving system capacity
requires an element spacing of about 0.5 m (~ 63 wavelengths).
In [21], system throughput is examined for arrays with closer
element space. It is shown that the maximum distance to support
multiplexing communications over LOS-MIMO channels is
mainly determined by the product of the aperture sizes of the Tx
and Rx antenna arrays, instead of the numbers of antennas at
both ends. For communication distance in the order of kilometers,
the multiplexing gain is limited to 4, even for a large array size
of 5 m.
Secondly, the array configurations based on the Rayleigh
distance criterion for ensuring LOS SU-MIMO are highly

constrained to a limited beam steering angle due to formation of
grating lobes [23]-[27], thus the optimal spacing between the
antenna elements for the Rayleigh distance criterion may not be
optimal for beam steering. Therefore an antenna configuration
based on the Rayleigh distance criterion would be severely
constrained in terms of the beam steering angle.
In view of the challenging requirements for SU-MIMO,
extensive investigation of the state-of-the-art hybrid precoding
techniques is undertaken as reported in [28]-[43]. This indicates
that most of the existing work details performance of
architectures that are applicable to multiple-user MIMO (MU-
MIMO). In these cases, Rx antenna arrays can be separated by
considerable distances as these are for multiple users which can
ensure sufficient spatial separation that enables SM. Similar Rx
antenna array separations are not feasible for SU-MIMO on a
single mobile user equipment. Secondly, SM in mmWave
propagation depends on both the channel and antenna properties.
Antenna properties such as the radiation pattern, the sub-array
spacing and orientation are intrinsic to implementing the ULA
design which must be considered for obtaining the channel
coefficients. These parameters are crucial to analysis of SM but
also for the beamforming constraints such as formation of
quantization and grating lobes in relation to the ULA architecture.
Some analyses for SU-MIMO hybrid precoding are given in [28],
[30], [36], [42], [43] but parameters assumed are for NLOS
mmWave channels. Also, previous work did not systematically
study the effects of sub-array spacing in mmWave for SU-
MIMO, or formation of quantization and grating lobes which is
an area of current research [19], [44]. Having multiple sub-arrays
reduces hardware complexity at the expense of less overall array
flexibility for beam scanning and MIMO which needs careful
analysis. Overall, in the existing approaches, SU-MIMO is a less
attractive option in mmWave cellular systems due to limited
multiplexing gain and dependency on the distance relationship.
The main motivation for employing hybrid precoding in
mmWave MIMO is for reducing the hardware resources.
Employing a specific hybrid precoding technique or comparison
of the existing hybrid precoding techniques such as for
optimizing the hardware resources in mmWave MIMO is not the
focus of this paper. Rather, our paper provides an alternative
approach to the Rayleigh distance criterion that enables SM for
SU-MIMO in LOS mmWave channels. Encouraging results for
up to 500m Tx-Rx separation distance are reported which are
practical to implement for urban outdoor mmWave cell sizes.
The proposed architecture is not constrained to any beam
steering angle due to formation of grating lobes. As the proposed
architecture is based on sub-arrays, a suitable hybrid precoder
can be accordingly employed to offer reduction in hardware
resources.
Furthermore this novel approach leverages the unique
characteristic of mmWave propagation which is formation of
one or more spatial-lobes (SLs) even in LOS channels due to a
rich scattering environment, which is not reported earlier in the
context of enabling SM for SU-MIMO in LOS mmWave
channels. The antenna elements can be co-located averting the
limitation of space requirements as SM is ensured due to
different angles of departure (AODs) and AOAs of the SLs even
in LOS propagation. The existing 3GPP channel models [1][45]
do not include the effects of directional local scattering at the Tx
and Rx in an outdoor urban environment for mmWave
propagation, yet real-world measurements in New York City
show the existence of directional propagation in urban
environments, leading to the formation of SLs [2], [46], [47].
SLs conveniently represent the mmWave radio channel because
they implicitly account for directionality, a key differentiator of
future wireless cellular and mobile systems operating in the
mmWave spectrum compared with today’s ultra-high frequency
(UHF) and microwave systems.
B. System-Level Architecture
In order to meet the contradictory requirements of
beamforming which requires co-polarized closely spaced
antenna elements typically at
with high coherence, and SM
which requires no coherence between antenna elements to
ensure simultaneous separate parallel data streams [19], a two-
level (2L) hybrid beamforming architecture is proposed as
illustrated in Fig. 1. The architecture follows a 2L hierarchy
wherein at level 1 (L1) each sub-array at the Tx and Rx is
employed to adaptively form highly directive beams to provide
beamforming gain. Each sub-array is an analog array consisting
of antennas connected with adjustable phase shifters in the RF
chain. Each sub-array is connected to a baseband processor via
a digital-to-analog converter (DAC) in the Tx or an ADC in the
Rx. Level 2 (L2) is the resulting MIMO system that enables
SM of these beams to increase channel capacity by supporting
simultaneous parallel data transfer based on the selected spatial
directions of the sub-arrays formed from L1. A similar approach
has been proposed in [26] for E-band (70-95 GHz) and
employed for SM for a 60 GHz indoor link [27]. In [26],
simulated values indicate higher spectral efficiency as the
physical spacing of the sub-array antennas is increased, where
sub-arrays with highly directive pencil beams are employed in
telescopic dish configurations. However, these highly
directional sub-array antennas are constrained for beam steering
due to formation of grating lobes at 10° off boresight, since the
increase in sub-array separation reduces the available angular
sector for which grating lobes do not occur. One can increase
the sub-array spacing to increase the number of SM paths but
this leads to formation of the grating lobes, severely restricting
the beam steering angle. In [27], analysis was restricted to only
MIMO SM and no beamforming was employed in an indoor
environment. The 2L approach presented in this paper can
combine SM with beamforming for ULAs to enable prototype
development of handsets and point-to-point Txs/Rxs in outdoor
LOS or NLOS urban environments for SU-MIMO.
In Fig. 1 the number of MIMO transmitting elements at L2 is
, wherein each of the transmitting elements is actually a sub-
array of many antenna elements. At L1 each sub-array has
elements. For mmWave propagation it is observed that two to
five SLs occur and two to three is the number at which most if
not all the energy will be received [46]. For example in Fig. 1
the number of MIMO transmitting elements
is three,
wherein each of the transmitting elements is a sub-array with
elements which in Fig. 1 is 4. Likewise, the number of MIMO
Rx elements
is two. Accordingly, at L2 there is an
×
mmWave MIMO channel. Under low SNR conditions when
employing beamforming, the same architecture can be
configured with the sub-arrays resulting in formation of a larger
array, which we call “super-arrayin this paper. In Fig. 1, the

Tx super-array consists of 12 elements (4 from each of the 3
sub-arrays) and the Rx super-array consists of 8 elements (4
from each of the 2 sub-arrays). The element spacing at L1, i.e.
within each sub-array is given by

, which remains
, while
the distance is the separation of the sub-arrays, which are
considered to be single elements for MIMO at L2.
Fig. 1 MmWave MIMO hybrid beamforming.
The proposed architecture enables generating the channel
coefficients based on the number of N
T
, N
R
and the half-power
beamwidth (HPBW) for a single beam that is transmitted through
the channel. At L2 accordingly the individual number of sub-
array elements M are reflected through the HPBW for a single
beam. It is worth noting that the HPBW is for the entire antenna
array i.e. the super-array. In case of low SNR conditions the
architecture enables beamforming via the antenna elements
within each sub-array, such that one or more weights are
generated to be applied to the individual ULA antenna element
signals. During normal/high SNR conditions MIMO SM is
carried out using sub-arrays, with each sub-array treated as a
single radiating element in a MIMO super-array system that
sends different data streams through each sub-array [19]. The
antenna element spacing can be interleaved or localised among
the sub-arrays, the later as in Fig. 1 offers smaller grating lobes,
larger LOS-MIMO capacity of a given array size and is practical
for hardware implementation as compared to the former [44]. In
case of low SNR during beamforming all the elements in the
entire super-array transmit the same symbol to compensate for
the high path loss with the beamforming gain. While in case of
normal or high SNR conditions for meeting the requirements of
high data transfer with MIMO, the three transmitting sub-arrays
in Fig.1 would transmit two different symbols for SM. Since the
antenna element spacing does not change during both modes of
transmission this circumvents formation of grating /quantization
lobes during beamsteering
Various super-array configurations are possible for the ULA
architecture in Fig. 1. Some of the configurations presented in
this paper are detailed in Table I, where is the total number of
antenna elements at L1,

is the array configuration with
denoting the number of elements in each sub-array, and
the
entire array length considering all sub-arrays in cm at carrier
frequency in GHz. If the lateral separation between the sub-
arrays is increased, with a view to reduce correlation between
sub-arrays, the visible region for beam steering of the super-
array (composed of sub-arrays) decreases, due to formation of
grating and quantization lobes.
III.
MMWAVE CHANNEL MODEL
A. 3D Statistical Spatial Channel Model
The MIMO channel model is based on a 3D SSCM for urban
LOS and NLOS channels developed from extensive 28-, 60-,
and 73-GHz ultra-wideband propagation measurements in
cities of New York City and Austin [46], [47]. The model
generates channel impulse responses (CIRs) that match
measured field data at wide range of distances and over local
areas based on the time clusterspatial lobe (TCSL) modeling
framework. The approach extends the 3GPP model through
directional root mean square (RMS) lobe angular spreads (ASs)
and is consistent with the 3GPP modelling framework.
Based on the 3D statistical channel model in [46], [47], a
MATLAB-based statistical simulator, NYUSIM v1.5, has been
developed by NYU [48], [49] that can generate 3D AOD and
AOA power spectra along with omnidirectional and directional
power delay profiles (PDPs) that match measured field results.
3GPP has unrealistically large number of strong eigenvalues
which are not found in measured mmWave channels. In order
to realistically quantify performance, NYUSIM v1.5 is
employed in this paper for simulating the MIMO channel as the
simulator is built from field data which gives more realistic
results [50].
TABLE
I
S
UPER-ARRAY CONFIGURATIONS FOR FIG. 1 WITH LENGTHS IN CM.
N
T

( = 2)

( = 4)

( = 8)

( = 16)
N
L
37
L
64
L
71
L
28
L
37
L
64
L
71
L
28
L
37
L
64
L
71
N
L
28
L
37
L
64
L
71
2
8
2.83
1.64
1.47
16
8.03
6.08
3.51
3.16
32
16.60
12.56
7.26
6.54
64
33.75
25.54
14.76
13.30
3
12
4.45
2.57
2.32
24
12.32
9.32
5.39
4.85
48
25.17
19.05
11.01
9.92
96
50.89
38.51
22.26
20.07
4
16
6.08
3.51
3.16
32
16.60
12.56
7.26
6.54
64
33.75
25.54
14.76
13.30
128
68.03
51.48
29.76
26.83
5
20
7.70
4.45
4.01
40
20.89
15.81
9.14
8.23
80
42.32
32.02
18.51
16.69
160
85.17
64.45
37.26
33.59
6
24
9.32
5.39
4.85
48
25.17
19.05
11.01
9.92
96
50.89
38.51
22.26
20.07
192
102.3
77.43
44.76
40.35
7
28
10.94
6.32
5.70
56
29.46
22.29
12.89
11.61
112
59.46
45.00
26.01
23.45
224
119.4
90.40
52.26
47.11
8
32
12.56
7.26
6.54
64
33.75
25.54
14.76
13.30
128
68.03
51.48
29.76
26.83
256
136.6
103.3
59.76
53.87

B. Parameters and Antenna Properties.
The simulator settings for the channel parameters and the
antenna properties employed for simulations in this paper are
listed in Table II. UMi, UMa and RMa denote urban
microcell, urban macrocell and rural macrocell settings,
respectively. Co/Cross is the polarization between the Tx and
Rx antenna arrays. Typical values for the barometric pressure,
humidity, temperature, rain rate and foliage attenuation have
been used with nil foliage loss. For the antenna properties,
ULAs have been considered at the Tx and Rx with a variable
antenna spacing d (i.e. the L2 sub-array spacing), the number
of transmitting antenna elements (varying from 2 to 8, which
is the number of sub-arrays at L2) is
, and number of
receiving sub-arrays is
. The Tx and Rx azimuth and
elevation HPBWs are for the super-array at L2. In Section
IV, the ULA antenna element has been designed with CST
MICROWAVE STUDIO® and array patterns developed
using MATLAB R2016a® generating HPBW input
parameters for NYUSIM v1.5. For various configurations in
Table I, the maximum azimuth HPBW is 12.6° (±6.3°) and
elevation HPBW is 88° (±44°). With an increase in the
number of sub-array elements, the azimuth HPBW becomes
less than 6.3° (±3.3°) and is taken as 7° which is the lower
limit in NYUSIM v1.5. The elevation HPBW remains
constant and is taken as 45°, the upper limit in NYUSIM v1.5.
A larger elevation HPBW would not affect the SM since for
horizontal ULAs the azimuth HPBW is more critical.
TABLE
II
NYUSIM
V1.5 SIMULATION SETTINGS USED IN THIS STUDY.
Channel Parameters
Frequency
(0.1-100 GHz)
28,
37,
64, 71
As per frequencies allotted by the FCC on
which this paper is based.
RF Bandwidth
(0-800 MHz)
800
Typical bandwidth that is expected for
mmWave 5G, and also the maximum
bandwidth that NYUSIM supports
Scenario
(UMi/UMa/RMa)
UMi
Urban micro cell, as we are considering
urban outdoor environment.
Environment
(LOS/NLOS)
LOS
LOS is considered in this paper.
Tx Rx Sep.
(10-500 m)
500
The maximum distance considering the
worst case scenario. The cell size in
mmWave 5G is unlikely to be greater than
500 m in view of high
free space path loss.
A
ccordingly, this is also the upper limit for
NYUSIM v.1.5.
Tx Power
(0-30 dBm)
30
A normal cellular base station has 30 dBm
of power transmission.
Number of Rx
Locations
100
Since the simulation output values are
random, 100 values are considered for
obtaining median values. Simulations show
that any increase beyond 100 results is
marginal dividend.
Barometric Pressure
(mbar)
1013.25
Normal atmospheric conditions.
Humidity (0-100%)
50
Normal atmospheric conditions.
Temperature °C
20
Normal atmospheric conditions.
Rain Rate (0-150
mm/hr)
8
Normal atmospheric conditions.
Polarization (Co/Cross)
Co
Co polarization antennas are considered in
this paper.
Foliage Loss
No
Nil foliage loss is considered.
Distance Within
Foliage (m)
0
Nil distance within foliage.
Foliage Attenuation
(dB/m)
0.4
Normal foliage attenuation.
Antenna Properties
Tx Array Type
(ULA/URA)
ULA
Horizontal ULAs are considered in this
paper.
Rx Array Type
(ULA/URA)
ULA
Horizontal ULAs are considered in this
paper.
No. of Tx elements
(sub-arrays)
N
T
Relevant Parameter.
No. of Rx elements
(sub-arrays)
N
R
Relevant Parameter.
Tx Antenna (sub-array)
Spacing (in λ, 0.1-100)
d
Relevant Parameter.
Rx Antenna (sub-array)
Spacing (in λ, 0.1-100)
d
Relevant Parameter.
Tx Antenna
(Super-Array)
Azimuth HPBW
(7°-360°)
12.6,
7
In Section IV, the ULA antenna element has
been designed with CST MICROWAVE
STUDI and array patterns developed
using MATLAB R2016a® generating
HPBW input parameters for NYUSIM v1.5.
Eqs.
(5)-(9) in Section IV are used to plot
array pattern in Fig 6 from
which the HPBW
for azimuth and elevation values are
extracted and detailed in TABLE II. This
happens to be 12.6 degrees (azimuth) and 45
degrees (elevation) for the configuration with
lowest number of array elements C
M4
and are
taken as such for simulatio
n. For C
M8,
the
number of elements is doubled, accordingly
the HPBW is half of 12.6 degrees which is
6.3 degrees (azimuth) and 45 degrees
(elevation
-does not change as the number of
elements are changing in azimuth only). But
the NYUSIM v1.5 has a lower limit of
HPBWs of 7 degrees for azimuth and
elevation. Thus for simulations in this paper,
an azimuth of 7 degrees is taken. SM gets
worse as the HPBW increases, thus the
NYUSIM v1.5 gives a worst case scenario
which can be accepted; as 6.3
-degree azimuth
HP
BW is expected to achieve even higher
spectral efficiencies. The narrower the
azimuth beam, the higher the SM gain.
Accordingly, C
M16
azimuth HPBW is 3.13
degrees and for CM
32
this decreases to 1.565
degrees. HPBW is taken as the lower bound
in NYUSIM 1.5 which is 7 degrees.
Accordingly, even better spectral efficiencies
are expected for these configurations as we
are assuming worse HPBW values than
expected.
Tx Antenna
(Super-Array)
Elevation HPBW
(7°-45°)
45
Rx Antenna Azimuth
HPBW (7°-360°)
12.6,
7
Rx Antenna Elevation
HPBW (7°-45°)
45
C. MIMO Channel Matrix Degrees of Freedom
Multi-carrier transmission is one of the technologies being
considered for mmWave. Considering orthogonal frequency
division multiplexing (OFDM) transmission, each resolvable
multipath component contributes to the MIMO channel
coefficients for an OFDM sub-carrier. Assuming ULAs at both
the Tx and Rx, the channel coefficient is given by [51]:
,
(
)
=
,,

,,

,,


,,


,,
(1)
wher
e
,
(
)
is the MIMO channel coefficient between the

transmit antenna and the

receive antenna for the sub-
carrier , represents the

resolvable multipath
component, is the amplitude of the channel gain, denotes
the phase of the multipath component, represents the time
delay,
and
are the antenna element spacing at the Tx and
Rx, respectively, while and denote the azimuth AOD and
AOA, respectively. All of the above parameters are extracted
from post simulations undertaken by NYUSIM v1.5 with
settings in Table II, with a Tx height of 10 m and a Rx height
of 1.5 m above the ground. Typical values obtained for a single
28 GHz simulation that results in 8 resolvable multipath
components are given in Table III.

Citations
More filters

Journal ArticleDOI
TL;DR: Results show that the answers to channel performance metrics, such as spectrum efficiency, coverage, hardware/signal processing requirements, etc., are extremely sensitive to the choice of channel models.
Abstract: Fifth-generation (5G) wireless networks are expected to operate at both microwave and millimeter-wave (mmWave) frequency bands, including frequencies in the range of 24 to 86 GHz. Radio propagation models are used to help engineers design, deploy, and compare candidate wireless technologies, and have a profound impact on the decisions of almost every aspect of wireless communications. This paper provides a comprehensive overview of the channel models that will likely be used in the design of 5G radio systems. We start with a discussion on the framework of channel models, which consists of classical models of path loss versus distance, large-scale, and small-scale fading models, and multiple-input multiple-output channel models. Then, key differences between mmWave and microwave channel models are presented, and two popular mmWave channel models are discussed: the 3rd Generation Partnership Project model, which is adopted by the International Telecommunication Union, and the NYUSIM model, which was developed from several years of field measurements in New York City. Examples on how to apply the channel models are then given for several diverse applications demonstrating the wide impact of the models and their parameter values, where the performance comparisons of the channel models are done with promising hybrid beamforming approaches, including leveraging coordinated multipoint transmission. These results show that the answers to channel performance metrics, such as spectrum efficiency, coverage, hardware/signal processing requirements, etc., are extremely sensitive to the choice of channel models.

147 citations


Cites background from "5G Uniform Linear Arrays With Beamf..."

  • ...At the cell edge, the boresight array gain is the key factor for coverage, as sidelobes off of the main beam will be far below the boresight power levels [49]....

    [...]

  • ...In general, the antenna array gain on boresight is approximately proportional to the number of antenna elements, but also depends on the radiation angle, and the specific gains and patterns of the antenna elements [49]....

    [...]



Journal ArticleDOI
TL;DR: A new massive MIMO channel model is suggested that embraces both the spatial- and frequency- wideband effects, and discusses issues to design a practical massive M IMO system.
Abstract: Massive MIMO, especially in the millimeter- wave frequency bands, has been recognized as a promising technique to enhance spectrum and energy efficiency, as well as network coverage for wireless communications. Most research in massive MIMO just uses the extended conventional MIMO channel model by directly assuming that the channel dimensionality becomes large. With massive numbers of antennas, however, there exists a non-negligible propagation delay across the large array aperture, which then causes a transmitted symbol to reach different antennas with different delays, thereby rendering conventional MIMO channel models inapplicable. Such a phenomenon is known as the spatial-wideband effect in the areas of array signal processing and radar signal processing, and introduces the beam squint effect in beamforming. However, the spatial-wideband effect and the related beam squint issue are seldom studied in massive MIMO communications. To design a practical massive MIMO system, it is important to understand when the spatial-wideband effect appears and how it affects signal transmission, how the spatial-wideband effect interacts with the frequency-wideband effect (frequency selectivity), especially for multi-carrier modulations such as orthogonal frequency- division multiplexing (OFDM), and how we should re-design the transceiver. In this article we suggest a new massive MIMO channel model that embraces both the spatial- and frequency- wideband effects, and discuss these issues.

51 citations


Cites background or methods from "5G Uniform Linear Arrays With Beamf..."

  • ...Furthermore, broadband communications over a millimeter-wave (mmWave) frequency band make massive MIMO attractive since the tiny antenna size facilitates packing a tremendously large number of antennas in a small area [2, 3]....

    [...]

  • ...For ease of illustration, we consider mmWave-band communications in sparse channel scenarios, which is reasonable from [2]....

    [...]


Journal ArticleDOI
TL;DR: The theoretical and experimental results of a new approach for multitarget vital-signs monitoring using an electromagnetic-based Doppler radar that prevents the phase collision of the signatures of the targets and hence provides multiperson detection capability to the system are presented.
Abstract: This paper presents the theoretical and experimental results of a new approach for multitarget vital-signs monitoring using an electromagnetic-based Doppler radar. A phased-array radar is designed and implemented using a hybrid beamforming architecture to generate two simultaneous beams. The proposed system significantly mitigates the phase collision problem in the presence of multiple targets. Comprehensive discussions on the theory of multibeam systems alongside detailed simulations are provided. For the purpose of demonstration, a prototype dual-beam phased-array continuous-wave Doppler radar has been designed and implemented at 2.4 GHz. The system is fully characterized, and the measurement results confirm the feasibility of the proposed method. The experimental measurements show that for the first time, the breathing rates of two individuals can be monitored at the same time and using the same frequency. Several practical aspects of the system are examined, and a pilot study on the subject tracking is presented. The proposed dual-beam system prevents the phase collision of the signatures of the targets and hence provides multiperson detection capability to the system.

40 citations


Cites background from "5G Uniform Linear Arrays With Beamf..."

  • ...There is not a unique approach for designing such systems [2], [58], [59], [64]....

    [...]

  • ...A thorough explanation of these systems can be found in [52], [54], and [58]....

    [...]

  • ...both from academia [2], [53], [58] and industry [52], [59] and are now considered as an unrivaled and plausible solution for the innumerable users and unprecedented demands of higher service quality....

    [...]


Journal ArticleDOI
TL;DR: Mixed-signal CMOS integrated circuits designs are proposed for the implementation of DVM multibeam algorithms along with low-complexity digital realizations to achieve hybrid beamforming for mmW applications.
Abstract: This paper proposes a low-complexity wideband beamforming subarray for millimeter wave (mmW) 5G wireless communications. The multibeam subarray is based on using a novel delay Vandermonde matrix (DVM) algorithm to efficiently generate analog true-time-delay beams that have no beam squint. A factorization for the DVM leading to low-complexity analog realizations is provided and complexity analysis for real and complex inputs is derived. The DVM is a special case of a Vandermonde matrix but with complex nodes that lack any special properties (unlike the discrete Fourier transform matrix). Error bounds for the DVM are established and then analyzed for numerical stability. Mixed-signal CMOS integrated circuits designs are proposed for the implementation of DVM multibeam algorithms along with low-complexity digital realizations to achieve hybrid beamforming for mmW applications. Analog–digital hybrid mmW multibeam beamforming circuits and systems are designed, for example, with eight beams at 28 GHz and simulated in cadence for functional verification.

25 citations


Cites background from "5G Uniform Linear Arrays With Beamf..."

  • ..., for P = 64 or 128 elements) impractical at mmW [10], [12], and [13]....

    [...]


References
More filters

Book
01 Jan 2005
Abstract: 1. Introduction 2. The wireless channel 3. Point-to-point communication: detection, diversity and channel uncertainty 4. Cellular systems: multiple access and interference management 5. Capacity of wireless channels 6. Multiuser capacity and opportunistic communication 7. MIMO I: spatial multiplexing and channel modeling 8. MIMO II: capacity and multiplexing architectures 9. MIMO III: diversity-multiplexing tradeoff and universal space-time codes 10. MIMO IV: multiuser communication A. Detection and estimation in additive Gaussian noise B. Information theory background.

7,400 citations


Journal ArticleDOI
TL;DR: The motivation for new mm-wave cellular systems, methodology, and hardware for measurements are presented and a variety of measurement results are offered that show 28 and 38 GHz frequencies can be used when employing steerable directional antennas at base stations and mobile devices.
Abstract: The global bandwidth shortage facing wireless carriers has motivated the exploration of the underutilized millimeter wave (mm-wave) frequency spectrum for future broadband cellular communication networks. There is, however, little knowledge about cellular mm-wave propagation in densely populated indoor and outdoor environments. Obtaining this information is vital for the design and operation of future fifth generation cellular networks that use the mm-wave spectrum. In this paper, we present the motivation for new mm-wave cellular systems, methodology, and hardware for measurements and offer a variety of measurement results that show 28 and 38 GHz frequencies can be used when employing steerable directional antennas at base stations and mobile devices.

5,589 citations


"5G Uniform Linear Arrays With Beamf..." refers background in this paper

  • ...and virtual interaction between people and machines [2]....

    [...]

  • ...More AOA SLs at 73 GHz arise from more prominent local scattering than at 28 GHz [46], likely due to the more thorough scanning in azimuth and elevation [47]....

    [...]

  • ...For a channel where signals depart at the Tx array of length Ly and arrive at the Rx array with a length of Ly in distinct SLs as is the case for mmWave channels, with AS of the kth SL as Θtk and Θrk for the AOD and AOA respectively, the DOF is given by [53]: min { ∑ k [Ly θtk ], ∑ k [Ly θrk ] } ....

    [...]

  • ...The diffused scattering leads the signal to arrive at the Rx in distinct SLs. Measurements undertaken for the 28 GHz NLOS channel in urban environments indicate a mean number of 1.6 SLs for both AOD and AOA....

    [...]

  • ...Formation of higher number of SLs in mmWave even for LOS scenarios implicitly account for directionality which ensure that the channel matrix H is well-conditioned when compared to existing sub mmWave mobile radio channels....

    [...]


Journal ArticleDOI
Abstract: New research directions will lead to fundamental changes in the design of future fifth generation (5G) cellular networks. This article describes five technologies that could lead to both architectural and component disruptive design changes: device-centric architectures, millimeter wave, massive MIMO, smarter devices, and native support for machine-to-machine communications. The key ideas for each technology are described, along with their potential impact on 5G and the research challenges that remain.

3,295 citations


Journal ArticleDOI
TL;DR: This paper considers transmit precoding and receiver combining in mmWave systems with large antenna arrays and develops algorithms that accurately approximate optimal unconstrained precoders and combiners such that they can be implemented in low-cost RF hardware.
Abstract: Millimeter wave (mmWave) signals experience orders-of-magnitude more pathloss than the microwave signals currently used in most wireless applications and all cellular systems. MmWave systems must therefore leverage large antenna arrays, made possible by the decrease in wavelength, to combat pathloss with beamforming gain. Beamforming with multiple data streams, known as precoding, can be used to further improve mmWave spectral efficiency. Both beamforming and precoding are done digitally at baseband in traditional multi-antenna systems. The high cost and power consumption of mixed-signal devices in mmWave systems, however, make analog processing in the RF domain more attractive. This hardware limitation restricts the feasible set of precoders and combiners that can be applied by practical mmWave transceivers. In this paper, we consider transmit precoding and receiver combining in mmWave systems with large antenna arrays. We exploit the spatial structure of mmWave channels to formulate the precoding/combining problem as a sparse reconstruction problem. Using the principle of basis pursuit, we develop algorithms that accurately approximate optimal unconstrained precoders and combiners such that they can be implemented in low-cost RF hardware. We present numerical results on the performance of the proposed algorithms and show that they allow mmWave systems to approach their unconstrained performance limits, even when transceiver hardware constraints are considered.

2,426 citations


Journal ArticleDOI
TL;DR: This article presents recent results from channel measurement campaigns and the development of advanced algorithms and a prototype, which clearly demonstrate that the mmWave band may indeed be a worthy candidate for next generation (5G) cellular systems.
Abstract: The ever growing traffic explosion in mobile communications has recently drawn increased attention to the large amount of underutilized spectrum in the millimeter-wave frequency bands as a potentially viable solution for achieving tens to hundreds of times more capacity compared to current 4G cellular networks. Historically, mmWave bands were ruled out for cellular usage mainly due to concerns regarding short-range and non-line-of-sight coverage issues. In this article, we present recent results from channel measurement campaigns and the development of advanced algorithms and a prototype, which clearly demonstrate that the mmWave band may indeed be a worthy candidate for next generation (5G) cellular systems. The results of channel measurements carried out in both the United States and Korea are summarized along with the actual free space propagation measurements in an anechoic chamber. Then a novel hybrid beamforming scheme and its link- and system-level simulation results are presented. Finally, recent results from our mmWave prototyping efforts along with indoor and outdoor test results are described to assert the feasibility of mmWave bands for cellular usage.

2,181 citations