Fiftieth Annual Allerton Conference

Allerton House, UIUC, Illinois, USA

October 1 - 5, 2012

Millimeter-wave Mobile Broadband with Large Scale Spatial

Processing for 5G Mobile Communication
Farooq Khan, Zhouyue Pi and Sridhar Rajagopal, Member, IEEE

becoming increasingly challenging to accommodate the



Abstract— The phenomenal growth in mobile data traffic calls

for a drastic increase in mobile network capacity beyond demand by staying in the sub 3 GHz band.
current 3G/4G networks. In this paper, we propose a
millimeter wave mobile broadband (MMB) system for the next Microwave and millimeter wave have been widely used
generation mobile communication system (5G). MMB taps into for long-range (e.g., a few kilometers) point-to-point
the vast spectrum in 3 – 300 GHz range to meet this growing communication for many years. However, the antennas and
demand. We reason why the millimeter wave spectrum is component electronics used in these systems, including
suitable for mobile broadband applications. We discuss the power amplifiers, low noise amplifiers, mixers, oscillators,
unique advantages of millimeter waves such as spectrum synthesizers, are too big in size and consume too much power
availability and large beamforming gain in small form factors. to be applicable in mobile communication. Notwithstanding
We also describe a practical MMB system design capable of the foregoing, efforts have been made in recent years to
providing Gb/s data rates at distances up to 500 meters and utilize millimeter waves for high data rate communication
supports mobility up to 350 km/h. By means of system [2]. For example, LMDS (Local Multipoint Distribution
simulations, we show that a basic MMB system is capable of Service) operates in frequencies from 28 – 30 GHz was
delivering an average cell throughput and cell-edge throughput conceived as a Broadband Wireless Access (BWA) solution.
performances that is 10 – 100 times better than the current 20- In the United States, 7 GHz of spectrum in the 60 GHz band
MHz LTE-Advanced systems. has been allocated for unlicensed use and about 13 GHz of
I. INTRODUCTION spectrum has been allocated for high-density fixed wireless
services in the 70 – 80 – 90 GHz band. The availability of the
Mobile communication has been one of the most 60 GHz band as unlicensed spectrum has spurred interest in
successful technology advancements in modern history. The Gb/s short-range communication. Several industrial standards
combination of technology breakthroughs and product that include WirelessHD, IEEE 802.15.3c and IEEE
innovation has made mobile devices an indispensable part of 802.11ad have been developed for the 60 GHz band.
life for more than 5 billion people. The increasing popularity Moreover, significant progresses have been made in low-
of smart phones and tablet computers suggests that the cost, low-power 60 GHz RFIC and antenna solutions [4-6]
demand for mobile broadband will continue to grow in the with commercial availability of products capable of
foreseeable future. delivering Gb/s data rates for distances up to 20 meters.
The current 4G LTE system uses advanced technologies In this paper, we explore the possibility of using 3 – 300
such as OFDM, MIMO, Turbo Code, Hybrid ARQ, and GHz spectrum for 5G mobile broadband communication. In
sophisticated radio resource management algorithms. These Section II, we discuss the millimeter wave spectrum and the
technologies are made possible by decades of extensive propagation characteristics in this band. The MMB network
research in wireless communication and systems. Still, architecture and air interface design are described in Section
continuous improvements in air interface are being III and the unique features in MMB design with large scale
considered by introducing new techniques such as carrier spatial processing are discussed in Section IV. We then
aggregation, higher order MIMO, Coordinated Multi-Point provide system simulation results of an MMB system in
(CoMP) processing, etc. However, further improvements in Section V. We conclude the paper in Section VI and provide
spectral efficiency appear extremely challenging [1-3]. a few suggestions for future research.
Another possibility to increase capacity in a given area is to
increase frequency reuse by deploying smaller cells, e.g. II. MILLIMETER WAVE SPECTRUM AND PROPAGATION
femto-cells. However, the number of cells that can be A. The 3 – 300 GHz spectrum
deployed in a given area can be limited due to several cost
factors such as site-acquisition, equipment installation, and Almost all commercial radio communications including
backhaul provisioning. Therefore, small cells alone are not AM/FM radio, TV, cellular, satellite communication, GPS,
expected to meet the orders of magnitude increase in mobile and Zigbee/Bluetooth have been using the Ultra High
data traffic demand in a cost effective and scalable manner. Frequency (UHF) band with frequencies around 300 MHz –
3 GHz. However, this band is rapidly becoming crowded due
Mobile communication systems today are deployed in the to continuous increase in demand for services that can benefit
sub-3 GHz spectrum. As the mobile data demand grows, it is from its favorable propagation characteristics. Around 500
MHz of the spectrum in the 300 MHz – 3 GHz band has
already been made available for LTE and there is current
Farooq Khan, Zhouyue Pi, and Sridhar Rajagopal are with Dallas
Technology R&D Center, Samsung Telecommunications America, focus on releasing even additional spectrum such as from the
Richardson, TX 75082 USA (e-mail: {fkhan, zpi, srajagop} TV white space bands etc. from broadcasters and to make it
@sta.samsung.com). available for cellular communication.

978-1-4673-4539-2/12/$31.00 ©2012 IEEE 1517

 3 GHz 57 64 164 200 300 GHz PG
t t

Pr
54GHz 99 GHz 99 GHz

Pr r
r
All cellular mobile
communications
Potential 252 GHz Pt
available bandwidth
Pt
60 GHz Oxygen Water vapor (H2O)
absorption band absorption band

Figure 1. Millimeter wave spectrum Figure 2. Omnidirectional and directional transmission

As depicted in Figure 1, within the 3 GHz – 300 GHz

band, up to 252 GHz spectrum can potentially be available The maximum effective aperture area of an antenna is related
for wide area mobile broadband communication. We exclude to the directivity or gain of the antenna as follows [8]:
the 57 – 64 GHz band due to severe oxygen absorption. The
spectrum in the range of 164 – 200 GHz might also be ( )
unsuitable due to absorption by water vapors, which can be
up to tens of dBs in a band centered around about 180 GHz An antenna with a larger aperture has more gain compared
[7]. to a smaller one as it captures more energy from passing
radio waves and by antenna symmetry argument it also
Currently, FCC has allocated the entire spectrum up to radiates more energy in that direction. In antenna theory,
300 GHz for different services. Traditional uses of the the isotropic antenna is a theoretical antenna with
millimeter wave bands include radio astronomy, space-borne a directivity of 0 dBi ( ). From ( ) and (3) we can
cloud radars, passive imaging of airport runways, imaging to obtain the Friis transmission equation:
display hidden contraband, weapons and nonmetal objects,
and other military applications. Therefore, we do not expect ( )
the entire millimeter wave spectrum to be allocated for
mobile broadband applications. Even if 40% of the Let us now assume a directional transmit antenna with
millimeter wave spectrum were to be made available for transmission gain, , directed towards the receive antenna in
mobile broadband applications, it would be 200 times more Figure 2. The ratio of received power to the transmit power
than the spectrum current allocation for mobile can be written as:
communications. Our proposal here is to start considering
some of these frequency bands for mobile communications.
Furthermore, there is a possibility of large chunks of
contiguous spectrum at ~20 – 50 GHz such as 23, 28, 38, 40, The gain for this antenna with aperture area can be
46, 47, 49 GHz. These bands are currently being used for derived from (3):
fixed applications such as backhaul in the US and the EU. To
that end, we can conservatively start exploring these lower
mmWave frequencies in the 20 – 50 GHz initially due to the
device efficiency and channel characteristic considerations.
From ( ) and ( ), we can write the ratio of received power
B. Propagation loss to the transmit power for a directional transmission as:
Transmission loss of millimeter waves is accounted for
principally by free space loss [7], given by: ( )

where is the free-space loss in dB, is the carrier We note that for given aperture areas of transmit and
frequency in GHz, and is the distance between the receive antennas, shorter wavelengths (higher frequencies)
transmitter and receiver in meters. can propagate longer compared to longer wavelengths (lower
It is often assumed that higher frequencies propagate frequencies) as the transmission at higher frequencies
poorly in free space compared to lower frequencies. This becomes more directive (narrower beams). For example, a
only applies when path loss is calculated at a specific beam at 30 GHz will have about 22 dB gain (narrower beam)
frequency between two isotropic antennas or half-wavelength compared to the beam at 2.4 GHz frequency if the antenna
dipoles. Let us consider the case of omni-directional areas are kept constant. Therefore, millimeter wave
transmission between two isotropic antennas as shown in frequencies (3 – 300 GHz) are unlikely to suffer any inherent
Figure 2. The ratio of the power received by the receiving disadvantage relative to the lower frequencies (< 3 GHz) as
antenna, , to the power input to the isotropic transmitting long as directional transmissions using beamforming are
antenna, , is given by: considered. Furthermore, narrow beams and directional
transmissions reduce interference and increase spatial
o n nn multiplexing and spatial division multiple access (SDMA)
o p capabilities for cellular applications.

1518
C. Penetration and other losses per km at mmWave frequencies [11]. In those cases, fallback
While signals at lower frequencies can penetrate more mechanisms to lower-frequency cellular bands can be
easily through buildings, millimeter waves do not penetrate considered as a part of the MMB system design.
most solid materials (e.g., brick wall and concrete) very
well. Millimeter-wave penetration through some materials D. Doppler
such as drywall, glass and wood is comparable to that at 2.5 The Doppler of a millimeter-wave channel depends on
GHz as shown in Table 1. Millimeter-wave signals the carrier frequency and the mobility. The maximum
transmitted from outdoor base stations may largely be Doppler shift is given by
confined to streets and other outdoor structures although
some signals might reach inside the buildings through glass
windows and doors etc. The coverage deep inside buildings
in this case can be provided by other means such as femto- where is the mobile speed, is the carrier frequency, and
cell or Wi-Fi solutions. is the speed of light. The coherent time . The
Table 1 Attenuations for different materials [9] maximum Doppler shift and the coherence time for carrier
Material
Thickness Attenuation Attenuation frequency of 3 – 60 GHz with mobility of 3 – 350 km/h are
[cm] 2.5GHz [dB] 60GHz [dB] shown in Table 2.
Drywall 2.5 5.4 6.0
Office Whiteboard 1.9 0.5 9.6 Table 2. Time-domain variation for millimeter wave
Clear Glass 0.3/ 0.4 6.4 3.6 Maximum Carrier Frequency (GHz)
Mesh Glass 0.3 7.7 10.2 Doppler (Hz) 3 10 30 60
3 8.3 27.8 83.3 166.7

Mobile speed
Foliage losses for millimeter waves are significant and 30 83.3 277.8 833.3 1666.7

(km/h)
can be a limiting impairment for propagation in some cases.
120 333.3 1111.1 3333.3 6666.7
An empirical formula has been developed in [10] to
calculate the propagation through foliage: 350 972.2 3240.7 9722.2 19444.4

Coherence Carrier Frequency (GHz)

Time (ms) 3 10 30 60
where is the propagation loss through foliage, is the 3 120.00 36.00 12.00 6.00
Mobile speed

carrier frequency in MHz and is the depth of foliage

(km/h)

30 12.00 3.60 1.20 0.60

traversed in meters, applicable for . The formula
in Equation ( ) is applicable for frequencies in the range of 120 3.00 0.90 0.30 0.15
200 MHz – 95 GHz. In Figure 3, we plot penetration losses 350 1.03 0.31 0.10 0.05
for foliage depth of 5, 10, 20 and 40 meters. We note, for
example, that at 40 GHz frequency and 10 meters foliage
penetration, the loss can be about 10 dB higher than the loss The Doppler shift of each incoming wave at various
at 3 GHz frequency. We can therefore expect that the angles at the receiver is different, resulting in a phenomenon
millimeter-wave signals will be severely attenuated by large called the Doppler spread. In a rich scattering environment
foliage depths. with equal receiver antenna gain in all directions, the power
spectrum of the Doppler spread exhibits a “U” shape
characterized by:

–
{ √ ( )

where is the maximum Doppler shift. In the case of MMB,

the transmitter and receiver beamforming significantly reduce
the angular spread of the incoming waves, which in turn
reduces the Doppler spread. In addition, as the incoming
waves are concentrated in a certain direction, there is a non-
zero bias in the Doppler spectrum, which will be largely
compensated by the automatic frequency control (AFC) in
the receiver. Therefore, the time-domain variation of an
Figure 3. Foliage penetration losses MMB channel is likely to be slower than that shown in Table
2.
Millimeter wave transmissions can experience significant
attenuations in the presence of heavy rain. Raindrops are
comparable in size to the millimeter radio wavelengths and E. Multipath
therefore cause scattering of the mmWave radio signals. The multipath channel characteristic is a key factor in
Monsoon at a rate of 150 mm/hour, which is very rare, can determining the numerology for the MMB systems. With
jeopardize communications with losses of up to tens of dBs narrow transmitter and receiver beams, the multipath

1519
components of a millimeter wave link are limited. Studies mobile station . With transmitter and receiver beamforming,
show that the root mean square (RMS) of the power delay the transmitter and the receiver can choose and
profile (PDP) of mmWave channels in an urban environment judiciously such that the SINR in Equation ( ) is
is of the order of 1 – 10 ns [7]. However, it is noted that the maximized. The choice of and can adapt to the
transmitter and receiver antenna gains used in [7] are higher long-term shadowing or the short-term fading, resulting in
than those used for MMB. Therefore, it is possible that in an different signaling overhead and performance.
MMB system longer paths can be observed and the
coherence bandwidth may be smaller than those reported in Transmitter and receiver beamforming that adapt to long-
these studies. term shadowing (i.e., the long-term beamforming) can
support very high mobility (up to 350 km/h). To illustrate
III. MMB AIR INTERFACE DESIGN this, assume that planar antenna arrays are used for
beamforming. The maximum beamforming gain of a planar
A. Beamforming
array can be approximated as [12],
Beamforming is a key enabling technology of MMB
system. For MMB transceivers, a large number of antennas
(e.g., 16 – 1024) can be packed into a small area due to the
short wavelengths, thus enabling beamforming with large where is the half-power beamwidth (HPBW) in the E-
gains. When employed in both the transmitter and receiver, plane and is the HPBW in the H-plane. Assume
beamforming can significantly improve the signal strength , the relationship between the antenna gain and the
and suppress interference from neighboring MMB cells, beamwidth becomes
resulting in much higher cell throughput and cell-edge
performance than traditional cellular systems.
√
As shown in Figure 4, beamforming in MMB systems can
be achieved in the digital baseband, analog baseband, or RF
For a planar array with a 30 dB gain, °, which
front end. With digital baseband beamforming and multiple
translates into 10 meters at a distance of 100 meter from the
RF chains, it is possible to transmit multiple data streams
base station. For a mobile station with 350 km/h speed, it
simultaneously, thereby enabling SDMA and MIMO type
takes about 100 ms to travel 10 meters. As the beamforming
transmission schemes. However, the cost of implementing
gain will mostly be less than 30 dB for MMB, we expect the
RF chains for every antenna element can be high for MMB
adaptation rate of long term beamforming to be less than 10
systems. With analog baseband beamforming or RF
Hz.
beamforming, the beamforming weights are applied to the
same signal transmitted via multiple antennas. This approach The adaptation rate of transmitter and receiver
reduces the number of RF chains but may also limit the beamforming to exploit the short-term fading (i.e., the short-
multiplexing capability of the system. term beamforming) needs to be faster than the Doppler of the
channel. As a result, it is only applicable when the coherent
time of the channel is much longer than the delay of the
DAC PA PA channel state information feedback. In an MMB system, by
DAC choosing the system design parameters appropriately, the
DAC PA PA short-term beamforming can be made to support short-term
(a) Digital baseband beamforming (b) Analog baseband beamforming
fading with Doppler up to 1 kHz.

PA
B. Frame structure
Beamforming weights
DAC OFDMA and single-carrier FDMA were chosen to be the
Up-converter
PA multiple access schemes of 4G systems due to a variety of
(c) RF beamforming reasons (e.g., flexibility in support multiple bandwidths,
Figure 4. Three different beamforming approaches simpler equalizer and MIMO receiver, and ability to support
efficient multiple access, etc.). For largely the same reasons,
we propose the use of OFDM and single-carrier waveform
The SINR of the signal from BS to MS , , can be for MMB. Table 3 shows a possible MMB OFDM
represented as numerology with 1/8 CP and system bandwidths of 250
‖ ‖ MHz, 500 MHz, and 1 GHz.
∑‖ ‖ ( ) Table 3. OFDM numerology (1/8 CP)
Short CP Configurations
where is the channel from BS to MS , is the System Bandwidth (MHz) 250 500 1000
transmission power of BS , and are the angle of Sampling rate (MHz) 276.48 552.96 1105.92
departure (AoD) and angle of arrival (AoA) for the link OFDM symbol length (FFT size) 1024 2048 4096
between BS and MS , respectively. The transmit antenna CP length 128 256 512
pattern at BS and the receiving antenna pattern at MS are Subcarrier spacing (kHz) 270
denoted by and . Therefore, and OFDM symbol duration (us) 3.70
CP duration (us) 0.46
represent the transmit beamforming gain and receiver
Number of OFDM symbols per slot 30
beamforming gain for the channel from base station to Slot duration (us) 125

1520
 The OFDM / single-carrier numerology is carefully A typical link budget for such a MMB communication
chosen to cater to a number of engineering considerations. system is shown in Table 4 . A 28 GHz system is chosen as a
The 270 kHz subcarrier spacing, for instance, is small enough design example that captures the impact of higher frequency
to stay within the multipath coherent bandwidth on the one operation while simultaneously meeting device
hand while being wide enough to keep small FFT / IFFT implementation and feasibility constraints. Our goal is to
sizes (4096 points for 1 GHz system bandwidth), and to provide a minimum of 1 Gb/s for the mobile uplink and 4
accommodate inaccuracies of low-cost VCXOs. A carrier Gb/s for the downlink over 0.25 km distance within a 500
frequency of 28 GHz and a 20 ppm VCXO generates a clock MHz bandwidth. RF component efficiency can be poor at
drift of no more than 560 kHz – about twice the proposed mmWave frequencies, although there is active research in
subcarrier spacing. This enables simple design of improving mmWave RF component efficiency [4-6]. We
synchronization and system acquisition. The chosen Cyclic assume 23 dBm as the transmit power for the uplink and 40
Prefix (CP) provides sufficient margin for accommodating dBm for the downlink. We assume a 7 dB noise figure [5].
the longest path [13-16] in different deployment scenarios, Channel models for mmWave communication have been
and the potential increase of delay spread in the case of small traditionally studied for outdoor, fixed environments for
antenna arrays (e.g., smart phones with small form factors) or LMDS applications [13] and for indoor, mobile for 60 GHz
wider beams (e.g., control channel transmissions) while applications. We assume an additional loss of 20 dB to
keeping a small CP overhead (11.1%). account for shadowing, fading and NLOS communication.
1 frame = 10ms
However, deeper understanding of the mmWave channel is
needed for outdoor, mobile environments to understand path
1 Frame
loss, angular spread, delay spread, NLOS beamforming and
= 10 subframes
1 subframe = 1ms
blocking issues and is a focus of on-going research in this
area [14-16].

1 subframe
IV. LARGE SCALE SPATIAL PROCESSING
= 8 slots
1 slot = 125us
The antenna aperture size decreases with the square of the
carrier frequency. This means that a 100X more antennas
1 slot can be used at 20 GHz to capture the same energy as an
CP

CP

CP

CP

equivalent 2 GHz antenna. However, a single digital chain

4.16us 3.7us
= 30 OFDM / Single-Carrier symbols
0.46us
per antenna is likely to be prohibitive for MMB systems due
to area, power and cost con n o uppo ng 100’ o
Figure 5. MMB frame structure (1/8 CP) digital chains. Hence, a hybrid approach consisting of both
analog and digital spatial processing looks promising for
One configuration of the MMB frame structure is shown MMB systems and has been investigated in [17-20].
in Figure 5. The basic transmission time interval (TTI) of Figure 6 shows a possible architecture for hybrid large
MMB is a slot, which has 125 us duration. The durations of scale spatial processing with Nt antennas, transmitting Ns
subframe and frame are chosen to be 1 ms and 10 ms, the data streams to K users. To do so, the transmitter is equipped
same as the subframe and frame duration in LTE, to facilitate
with Ns < NtRF < Nt RF chains. The signal is first precoded
interoperability with 4G systems.
digitally by a baseband precoder, followed by a phase-only
per antenna RF precoder using phased antenna arrays.
C. Link budget
The key factors that determine the link budget of an
MMB system are the transmission power, the transmitter and
receiver beamforming gains, and the path loss.

Table 4. MMB downlink link budget

Link Budget Analysis Downlink Uplink
Transmit power (dBm) 40.00 40.00 23.00 23.00
Transmit antenna gain (dBi) 25.00 25.00 12.00 12.00
Carrier frequency (GHz) 28.00 28.00 28.00 28.00
Distance (km) 0.50 0.25 0.50 0.25
Free space propagation loss (dB) 115.32 109.30 115.32 109.30
Other losses (shadowing, fading) 20.00 20.00 20.00 20.00
Receive antenna gain (dB) 12.00 12.00 25.00 25.00
Received power (dBm) -58.32 -52.30 -75.32 -69.30
Bandwidth (GHz) 0.50 0.50 0.50 0.50
Thermal noise PSD (dBm/Hz) -174.00 -174.00 -174.00 -174.00
Figure 6. MMB Large Scale Hybrid Spatial Processing system
Noise figure 7.00 7.00 7.00 7.00
Thermal noise (dBm) -80.01 -80.01 -80.01 -80.01
SNR (dB) 21.69 27.71 4.69 10.71 The channel characteristics for a mmWave MIMO
Implementation loss (dB) 3.00 3.00 3.00 3.00 channel are considerably different to that of “ gul ”
Spectral efficiency 6.23 8.21 1.31 2.79 MIMO channel. The latter is typically characterized by high
Data rate (Gbps) 3.11 4.11 0.65 1.39

1521
penetration and omni-directional reception at the mobile. is 1°, i.e., there are 60 beams in the base station transmit
The number of paths arriving at the mobile is typically much beamforming codebook for each sector. Further details on
higher than the number of antennas. In contrast, the the simulation methodology and design considerations are
mmWave MIMO channel is more reflective, and discussed in [22].
beamforming techniques are needed to capture the energy Mobile stations with random orientations are randomly
from appropriate directions. The number of antennas dropped. The resolution of mobile station receiver
supported is typically much greater than the number of paths beamforming is also 1°, i.e., there are 360 beams in the
at the mobile and the nn l l ly mo “g om ” mobile station receiver beamforming codebook.
as opposed to being omni-directional.
Current phased antenna array design implementations
MIMO precoding is primarily a baseband precoding can attain a resolution of 5° [5]. Although it could be argued
mechanism where the the number of digital chains are equal that a resolution of 1° for transmit and/or receive
to the number of antennas, and the channel matrix is beamforming may be difficult to realize in practice, such
relatively small. In contrast, the mmWave MIMO precoding precision in resolution is not required since the HPBW used
comprises substantial RF processing with phased antenna by the antenna arrays for steering in azimuth and elevation
arrays and the number of digital processing chains are fewer can cover a wide range of resolution angles without
than the number of antennas resulting in a relatively larger significant performance loss.
matrix, However, since the channel is more geometric, the 1
number of principal components in the channel matrix is 0.9
relatively few, resulting in a sparse matrix. Table 5 compares 0.8
design aspects between cellular communications below 3 0.7
GHz and mmWave cellular communication. 0.6

CDF
0.5
Table 5 Comparing regular and mmWave cellular communication
0.4
Feature Current cellular mmWave cellular
0.3
Antenna size Large antennas Small antennas
Limited MIMO Massive MIMO 0.2
MIMO order
(BS & MS) (BS & MS) 0.1
BF Digital Processing Hybrid spatial processing 0
architecture (Digital MIMO & BF) (Digital MIMO & RF BF) -18 -15 -12 -9 -6 -3 0 3 6 9 12 15 18 21 24
Multipath
#Paths >> #Antennas #Antennas >> #Paths Geometry (dB)
channel
Joint digital and analog
Precoder Cellular MMB, NoBF MMB, RxBF
Digital optimization optimization
design MMB, TxBF MMB, TRBF
(phase only analog)
Figure 7. Geometry distribution in an MMB system
V. SIMULATION RESULTS
The performance of MMB system is studied using system The mobile station geometry distribution in the MMB
level simulations. The system configuration largely follows system is shown in Figure 7. Base station Tx power of 40
the system simulation methodology of [21] with a few dBm per sector is assumed for both the MMB systems and
changes to model the MMB-specific features. The list of the cellular reference system. A 4G LTE-A cellular
important system parameters are shown in Table 6. The path reference system is assumed for comparison with 4 Tx
loss formula, more realistic to the mmWave channel, based MIMO per sector, 3-sector cell and 20 MHz system
on [13] is assumed. bandwidth as in [23]. W ou b m o m ng (“NoBF”),
Table 6. MMB system simulation configuration
geometry of MMB performs worse than a comparable 4G
Cell layout 19-cell wrap around d gn (“ llul ”). How , w b m o m ng, at the
Number of sectors per cell 6 mob l on only (“RxBF”) o b on only (“TxBF”),
Site-to-site distance 500 meters provides a better performance than cellular design. Joint
Carrier frequency 28 GHz transmit and receive b m o m ng (“TRBF”) gn n ly
System bandwidth 500 MHz improves the geometry of the mobile stations by enhancing
Path loss formula 157.4 + 32log10d (d in km)
BS Tx power per sector 40, 43, 46, 49 dBm
the d d MS’ 5 p n l signal strength by at least 6 dB
BS antenna configuration 48-element (12x4) horn antenna as well as suppressing the interferences to other MSs.
array per sector Beamforming at the BS and the MS appears to be an
Lognormal shadowing STD 12 dB essential component of the MMB system design.
Mobile station noise figure 7 dB
MS antenna configuration 2 arrays, each with 4 elements The system throughput and cell-edge performance of the
System overhead 30% designed MMB system are shown in Figure 8. A single
Implementation loss 3 dB stream transmission and simple round-robin scheduling are
assumed for the simulations. With 40 dBm BS Tx power, a
Each base station is deployed with 6 horn antenna arrays cell throughput of 8.37 Gb/s (including all 6 sectors in a cell)
of 12x4 elements each, an antenna array covering a 60° and a cell-edge throughput of 680 Mbps can be achieved. As
sector. The resolution of base station transmit beamforming we raise the BS Tx power to 49 dBm, a cell throughput of

1522
9.07 Gb/s and a cell-edge throughput of 882 Mbps are REFERENCES
achieved. [1] M. Dohler, R. W. Heath, A. Lozano, C. B. Papadias, R. A.
V l nzu l , “I PHY L y D d?”, IEEE Communications
These performance numbers from Figure 8 demonstrate Magazine, ,vol.49, no.4, pp. 159-165 April 2011
that with mmWave communication, it is possible to achieve [2] Z. Pi, F. K n, “ n n odu on o m ll m -wave mobile
a 10–100 times improvement over the current 4G systems. b o db nd y m ,” IEEE Communications Magazine, vol. 49, no. 6,
This gain is in line with the increased system bandwidth pp. 101-107, June 2011
used the MMB system in our example (25 times more [3] T. L. M z , “Non oop llul wireless with unlimited
bandwidth compared to the 20 MHz)). With more numb o b on n nn ,” IEEE Trans. Wireless
sophisticated technologies including spatial multiplexing, Communication, vol. 9, no. 11, pp. 3590 – 3600, Nov. 2010
SDMA, MIMO, and channel sensitive scheduling, further [4] C. H. Do n, S. Em m , . M. N kn j d, R.W. B od n, “M ll m
w CMOS d gn,” IEEE Jou n l o Sol d-States Circuits, vol.40,
performance improvements of the MMB system with Tx and no.1, pp. 144- 155, Jan. 2005.
Rx beamforming can be expected. [5] M. Tabesh, J. Chen, C. Marcu, L.-K. Kong, S. Kang, E. Alon, A. M.
N kn j d, “ 65nm CMOS 4-Element Sub-34mW/Element 60GHz
P d y T n ,” IEEE International Solid-State Circuits
Conference (ISSCC), pg 12-14, March 2011.
Cell throughput Cell-edge throughput
[6] T. S. R pp po , J. Mu do k, F. Gu z, “ S o n 60 GHz
9100 950 Integrated Circuits and Systems for Wireless Communications,
Proceedings of the IEEE, August 2011, Vo.99, No 8, pp. 1390-1436.
9000 900
[7] “M ll m w p op g on: Sp um m n g m n mpl on ,”
Federal Communications Commission, Office of Engineering and

Cell-edge throughput (Mbps)

8900 850
Cell throughput (Mbps)

Technology, Bulletin Number 70, July, 1997

8800 800 [8] D. M. Pozar, Microwave engineering, 3rd ed., Wiley 2005.
8700 750 [9] C. R. nd on, T. S. R pp po , “In-Building Wideband Partition
Lo M u m n 2.5 nd 60 GHz”, IEEE T n on on
8600 700 Wireless Communications, Vol. 3, No. 3, May 2004.
[10] E. V ol , R. E p l nd nd K. C. ll n, “M ll m -Wave
8500 650 Propogation Characterestics and Channel Performance for Urban-
Subu b n En onm n ,” U. S. Department of Commerce, NTIA
8400 600
Report 88-239, December 1988.
8300 550 [11] ITU-R P.838-3, “Specific attenuation model for rain for use in
40 43 46 49 prediction methods”, 2005.
Base station transmit power per sector (dBm) [12] R. S. Ell o , “B mw d nd d yo l g nn ng y ,”
Last of Two Parts, The Microwave Journal, pp. 74 – 82, January 1964
Figure 8. MMB system and cell-edge performance [13] P. Som , . l., “P op g on m u m n nd mod l ng o LMDS
do nn l n S ng po ,” IEEE T n . on V ul T ., M y
2003
VI. CONCLUSION [14] T. S. Rappaport, E. Ben-Dor, J.N. Murdock, Y. Qiao, J. Tamir,
“C llul nd P -to-Peer Broadband Millimeter Wave Outdoor
In this paper, we analyzed the millimeter-wave spectrum propagation measurements and Angle of Arrival characteristics using
(3 – 300 GHz) and discussed its suitability for mobile d p b m ng,” n d p p nd p n on, IEEE R d o
broadband communication. We proposed a millimeter-wave and Wireless Symposium (RWS) 2012, Santa Clara, CA, Jan. 15,
mobile broadband (MMB) system for 5G mobile 2012.
communication that can outperform LTE-Advanced systems [15] T. S. Rappaport, E. Ben-Dor, J. N. Murdock, Y. Qiao, "38 GHz and
60 GHz Angle-Dependent Propagation for Cellular & Peer-to-Peer
by 10 – 100 times in terms of cell throughput as well as cell- Wireless Communications," IEEE International Conference on
edge performance. This performance gain is made available Communications (ICC), June 2012.
mostly through the larger bandwidth available for MMB [16] J. N. Murdock, E. Ben-Dor, Y. Qiao, J. I. Tamir, T. S. Rappaport, "A
systems at these higher frequencies. The challenges in 38 GHz Cellular Outage Study for an Urban Outdoor Campus
channel propagation characteristics can be met by developing Environment," IEEE Wireless Communications and Networking
new robust system designs and network topologies and Conference (WCNC), April 2012.
developing new technologies supporting hybrid large scale [17] O. El Ayach, R. W. Heath, Jr., S. Abu-Surra, S. Rajagopal, Z. Pi,
“Low Compl x y P od ng or Large Millimeter Wave MIMO
spatial processing with digital and RF beamforming for 100’s Sy m ,” ICC 2012, Ottawa, Canada, June 2012.
of antenna elements. However, further research is needed in [18] Z. P , “Op m l MIMO n m on w p -antenna power
channel modeling for mmWave systems for outdoor mobility on n ,” Globecom 2012, December 2012
and MIMO support, system design for enabling large scale [19] Z. P , “Op m l nm b m o m ng w p -antenna power
spatial processing, and mmWave transceiver solutions on n ,” ICC 2012, Ottawa, Canada, June 2012
focusing on high efficiency and low power to bring this [20] S. R j gop l, “Beam Broadening with Phased Antenna Arrays for
MMB system to reality and is a focus of our current research. mmWave Mobile Communication”, ICC 2012, Ottawa, Canada, June
2012
As mobile broadband traffic is projected to continue its [21] 3GPP TR 36.814, v9.0.0, “Further advancements for E-UTRA
exponential greatly stretching the capabilities of the current physical layer aspects”, M 2010
4G systems, MMB offers the opportunity to develop [22] S. Rajagopal, S. Abu-Surra, Z. Pi, and F. K n, “Antenna Array
Design for Multi-Gbps mmWave Mobile Broadband
innovative millimeter-wave mobile communication Communications”, Globecom 2011, pp. 1-6, December 2011
technologies for 5G mobile communication that can deliver [23] 3GPP TR 36.912, 9.3.0, “F b l y udy o further advancements
100 – 1000 times more capacity than the current 4G systems. for E-UTRA (LTE-Advanced) ”, June 2010

1523