884H1: Topics in Wireless

Communications
Project Assignment
Candidate Number: 218831
Due date: 26/05/2019

1
Contents
CONTENTS .............................................................. 2
BACKGROUND ........................................................ 3
WIRELESS WAVEFORMS ......................................... 4
M-PSK ................................................................... 4
M-QAM ................................................................. 4
OFDM ................................................................... 4
4G LTE ................................................................... 6
SOFTWARE DEFINED RADIO ................................... 8
HARDWARE AND FLEXIBILITY........................................ 8
MOBILE COMMUNICATIONS SIGNALS ............................ 8
2G ................................................................................ 8
3G ................................................................................ 9
4G ................................................................................ 9
MM-WAVE ........................................................... 10
CHANNEL CODING ................................................ 12
BLOCK CODES ......................................................... 12
PERFORMANCE ....................................................... 12
SIMULATION........................................................... 13
Hamming and Reed-Solomon ................................... 13
LDPC .......................................................................... 14
OFDM ................................................................... 15
SIMULATION........................................................... 15
SPATIAL DIVERSITY ............................................... 17
RX DIVERSITY ......................................................... 17
TX DIVERSITY .......................................................... 18
MIMO ................................................................... 20
MULTI-USER MIMO ............................................... 20
BEAMFORMING .................................................... 22
BEAMFORMING IN 5G .............................................. 22
BEAMFORMING AT THE RECEIVER ............................... 22
BEAMFORMING AT THE TRANSMITTER ......................... 23
TYPES OF BEAMFORMING ......................................... 24
Digital Beamforming ................................................. 24
Analog Beamforming ................................................ 25
Hybrid Beamforming ................................................. 25
BEAM STEERING ...................................................... 25
REFERENCES ......................................................... 28

2
Background transmission (Channel Coding) and multicarrier
modulation (OFDM) to the application specific
The mobile communication systems are one of the
techniques such as Spatial Diversity and MIMO.
constant pillars of human development, first
allowing users to communicate with each other Finally, a description of Beamforming, a mature
using large and inefficient systems compared to technique already used in several other industries,
what is available today. Each generation of mobile including its theory, mathematical background, use
communications comes at the early years of the in 5G and simulation.
decade while having at least one revision at the
middle of its lifetime; each of these delivers
fascinating new features and improvements in both
speed and reliability.
As 2G introduced the implementation of digital
communications instead of being analog, many
other characteristics were to be invented, such as
text messages and multimedia messages. This
convention is still relevant, as every following
generation keeps the use of digital communications.
The goals set to each generation are the driving
motivation for engineers to come up with newer and
more advanced techniques that makes the most of
the available resources. For example, the need of
faster data throughput and to support a vast number
of users led to the implementation of ODFM as the
second modulation technique and the multiple-
access scheme of 4G (More about this in the
Wireless Waveforms section).
The use of mobile communications moved from
being a commodity to an increasing necessity since
industries now depend on these technologies, and
people are used to its inclusion in their lives. The
previous statement means that the upcoming
generation of mobile communications must support
features for the current industries and the
newcomers to use the communication
infrastructure reliably. This decade’s generation,
5G, is already being deployed in some parts of the
world and its technical proposal includes several
new technologies; some of these are already mature
while others still present an engineering challenge.
Some of these technologies are described in the
following sections, going from the already known
techniques for modulation (Wireless Waveforms),
message coding for correcting errors in

3
Wireless Waveforms M-QAM
As mobile communications evolve, technologies and Quadrature Amplitude Modulation is a Single-
techniques are updated, affecting characteristics in Carrier form of modulation like M-PSK in which both
time and frequency domain. Since the first amplitude and phase are used to code the input
generation of mobile communications, features like signal. In QAM, amplitude levels and phase angles
multiplexing/Multiple-Access, modulation, number are meant to create square-shaped constellations
of carriers (or general bandwidth availability) have (although non-square constellations are used, as
been engineering challenges to overcome to offer a well). The main performance difference between
service of quality to end-users. QAM and PSK is the spatial distance between
constellation points which is (roughly saying)
M-PSK proportional to Bit Error Performance (BER),
M-ary Phase Shift Keying is a form of single-carrier, making QAM more error robust at the cost of power
phase modulated, constant amplitude modulation efficiency.
scheme in which the input signal is coded into a
correlated sinewave with variable phase. The
number of possible phase angles is defined by the M-
ary number (M), this number also represents the
name of the modulation scheme and the number of
bits per symbol (k) at the modulated signal.
𝑘 = log 2 𝑀
𝑀 = 2𝑘
Equation 1. Relationship between the M-ary number (M) and the Figure 2. Various M-QAM constellation diagrams.
number of bits per symbol (k).
OFDM
Orthogonal Frequency Division Multiplexing is the
multiplexing (and Multiple Access) technique used
in 4G systems. OFDM is a multicarrier technique
with several advantages to single carrier
modulation being fading robustness and bandwidth
efficiency the main features of it.
As its name implies, OFDM uses the orthogonality
property to place sub-carriers close to each other in
the frequency domain without overlapping,
increasing the bandwidth efficiency abruptly. The
Sub-Carrier’s bandwidth is used as the Sub-Carrier
spacing and its inverse as the symbol time.
Practical applications of OFDM implement FFT/IFFT
to combine all the different subcarriers into a single
time-domain signal (To avoid using multiple
oscillators) and a so-called “Cycle-Prefix” as guard-
Figure 1. Various M-PSK constellation diagrams. interval between symbols to avoid shifts in the
frequency domain and Inter-Symbol Interference

4
The number of subcarriers is given by the next
higher power of two that results in more than the
required subcarriers to meet the design goals.
𝑁𝑠𝑢𝑏−𝑐𝑎𝑟𝑟𝑖𝑒𝑟𝑠 = 2𝑛 > 𝑁𝑟𝑒𝑞𝑢𝑖𝑟𝑒𝑑−𝑆𝐶 , 𝑛 = 1, 2, 3, …

Ideally, every subcarrier should be used to carry

user data, but some of them are used as pilots (a
known signal used for channel estimation), guard-
band or simply not used (such as the DC null
carrier). Designers can decide on how sub-carriers
are allocated, affecting the overall efficiency of the
system. Figure 4. Frequency spectrum for OFDM waveform (a)

OFDM is often represented as a grid of “Resource As shown in Figure 4, 64 Sub-Carriers are stacked in
Elements”, each one representing a Sub-Carrier in the ODFM resource grid from which 11 of them act
both time and frequency domains. as band guards, and the rest are data sub-carriers.
The horizontal axis represents the resource grid in
MATLAB’s Wireless Waveform Generator App helps
the time domain as OFDM symbols; 100 are
to create various kinds of modulated signals (as
displayed as configured in Table 1; this resource
shown in Figure 2 and Figure 3). An OFDM
grid does not include a DC null nor pilot sub-carriers.
waveform was generated using the following
Figure 5 shows the frequency spectrum of the
parameters using MATLAB. generated OFDM signal and can only show the
Parameter Value specified signal bandwidth. Fifty-three data sub-
FFT Length 64 carriers (Result of the subtraction of 11 from the
Guard Band Sub- 6 and 5 initial 64) with 1 MHz sub-carries were defined
Carriers resulting in a total bandwidth of 53 MHz (Or 64 MHz
DC Null No considering the guard-band).
Cyclic Prefix Length 16
OFDM Symbols 100 Parameter Value
Sub-Carrier Spacing 1 MHz FFT Length 128
Pilot Input No Guard Band Sub- 6 and 5
Table 1. Parameter table for OFDM waveform (a). Carriers
DC Null Yes
Cyclic Prefix Length 16
OFDM Symbols 100
Sub-Carrier Spacing 1 MHz
Pilot Input 11:12:120
Table 2. Parameter table for OFDM waveform (b).

Figure 3. Resource Grid for OFDM waveform (a)

5
 Figure 6. Type 1 frame (FDD)

Figure 5. Resource Grid for OFDM waveform (b)

Table 1 shows a different configuration for a second

OFDM signal. This new set increases the number of
sub-carriers to 128, keeps the original amount of
guard-band sub-carriers, includes arbitrarily placed
pilots every 12 subcarriers (starting from 11 and
ending in 120) and the DC null sub-carrier. Figure 7. Type 2 frame (TDD)

4G LTE Both frames may have the following symbols.

The LTE standard uses OFDM and assigns fixed
• Physical Downlink Shared Channel (PDSCH):
values to the parameters in the past tables and
User data is transmitted in this channel.
assigns new special functions to resource elements.
• Primary Synchronization Chanel (PSCH):
The LTE standard features the following parameter
Used for synchronization in cell search.
values.
• Secondary Synchronization Channel (SSCH):
Parameter Value Used to synchronize timing and to transmit
Sub-Carrier spacing 15 kHz cell group ID during a cell search.
Symbol time 66.6 µs • Physical Broadcast Channel (PBCH): Carries
Sub-Carriers per 12 cell information.
Resource block • Physical Downlink Control Channel
Symbols in a Resource 7 (PDCCH): Used for forwarding error
block correction, resource allocation and Uplink
Sub-Frame duration 0.5 ms schedule grant.
Available BW 1.4, 3, 5, 10, 15, 20 MHz
• Physical Control Format Indicator Channel
Available Resource 6, 15, 25, 50, 75, 100
(PCFICH): Indicates the number of PDCCH
Blocks
symbols per subframe.
Duplexing Mode TDD/FDD
Table 3. LTE standard numerology. • Physical Hybrid ARQ Indicator Channel
(PHICH): Carries H-ARQ information.
LTE supports two frame structures, one for • Physical Multicast Channel (PMCH): Carries
Frequency Division Multiplexing and the other for Multicast data.
Time Division Multiplexing and their arrangement
of each frame depends on its duplexing The following grid was generated in
characteristics. [Link]

6
and outputs a visual reference showing all the The resulting LTE resource grid can be seen on this
resource grid details such as block number, frame website (Figure 38). This grid uses the type 2 frame
number, sub-frame number, symbol and channel (TDD) and shows one frame (10 Sub-frames → 20
type. These grids are also found in the Appendix. slots per frame → 140 symbols per frame) and the
contents of these are divided into Downlink, Uplink
Parameter Value and Special sub-frames. Like type 1 frame, every
Duplexing Mode FDD
downlink sub-frame starts with PCFICH, PHICH and
Channel Bandwidth 1.4 MHz
PDCCH channels to schedule and grant resources.
Cyclic Prefix Length Normal
Uplink sub-frames are assigned to the UE, and the
No. of Tx Antennas 2
base station does not define them. Special sub-
MBSN Subframe None
Physical Layer Cell ID 0 frames happen between UL and DL sub-frames,
Control Format Indicator 1 allowing synchronization and a guard-time between
PHICH factor 1 Ul and DL.
PHICH duration 1 Symbol
Table 4. LTE Resource grid setting for FDD frame type.

The resulting LTE resource grid can be seen on this

website (Figure 37). This grid shows 6 RBs and half
a frame (10 Sub-frames (Half of the total type 1
frame) → 20 slots per frame → 140 symbols per
frame). Each sub-frame starts with several PCFICH,
PHICH and PDCCH to schedule and configure the
contents of the upcoming slots in the given sub-
frame. Sub-frames can be dedicated to transmitting
user data or synchronization, the latter type can only
be seen in sub-frame 0 and 5, and the remaining sub-
frames are used for data transmission. In every sub-
frame, some symbols are used as a reference of
power transmission from an eNode B, and its
arrangement depends on the number of Tx antennas
and port number.

Parameter Value
Duplexing Mode TDD
Channel Bandwidth 1.4 MHz
Cyclic Prefix Length Normal
No. of Tx Antennas 2
MBSN Subframe None
Physical Layer Cell ID 0
Control Format Indicator 1
PHICH factor 1
PHICH duration 1 Symbol
TDD Uplink/Downlink 4 DL, 4 UL and 2
Configuration Special Sub-frames
Special Sub-frame 9 DwPTS symbols
Configuration and 1 UpPTS
symbol
Table 5. LTE Resource grid for TDD frame type.

7
Software Defined Radio and simulate RF systems and the much higher
processing speed for RF applications.
Software Defined Radio (SDR) is a term to refer to
radio systems with almost all its functionality done The SDR device used for the following sections is the
in Software, instead of hardware as it is commonly RTL-SDR, a low-cost, receive-only SDR capable of
found in most of the RF applications in recent times. receiving signals up to 1750 MHz which uses a
An ideal SDR, as shown in Figure 8, would have an monopole antenna. The SDRSharp software was
FR Front End only consisting of a power amplifier used to interface with the RTL-SDR.
and high-speed Analogue to Digital converter. At the
same time, the remaining physical layer functions Mobile Communications Signals
such as modulation, synchronization or encoding Second, third and fourth generations of mobile
are done using DSP. communications are still available for people to use,
each one of these generations having a particular
SDR systems are expected to work in a wide range of space in the frequency spectrum. Most of them are
the frequency spectrum and to perform various
available in the range supported by the RTL-SDR.
operations as a dedicated hardware implementation
would have. For example, using a low-cost SDR unit
for hearing a local FM radio station and for getting
an OFDM signal using QAM modulation in the 2.4
GHz band are both widespread uses.

Figure 9. Mobile Cell Signals compared to the RTL-SDR frequency

range. [1]

Sadly, neither spectrogram nor frequency response

plots will not allow seeing the modulation scheme of
any generation. Also, time-domain properties are
Figure 8. Ideal SDR architecture. [1] hard to notice using the SDRSharp software.
Hardware and Flexibility 2G
Unlike old radio units using specialized hardware to The second generation of mobile communications is
receive and process a particular kind of incoming defined by its multiple access technique: TDMA.
signal, SDR hardware is generic enough to process Time-Division Multiple Access assigns a time slot to
data of several frequencies and modulation each user to transmit data. These messages are 200
schemes; an FPGA inside the SDR radio often does kHz wide and are located in between 890-915 for
this feature. Field Programmable Gate Arrays the Uplink and 935-960 for the Downlink. [1]
(FPGA) is a hardware programmable device, instead
of software programmable as the commonly known
microcontrollers. An FPGA is configured using
Hardware Device Language (HDL) that describes the
intended behaviour of the FPGA. FPGAs, in the
context of DSP, are faster and efficient than
microcontrollers; because to the dedicated DSP
slices inside the FPGA, the available tools to design

8
 Figure 11. Spectrogram signal from 3G at ~888 MHz from the
RTL-SDR

Figure 10. Spectrogram signal from 2G at 898 MHz from the 4G

RTL-SDR.
As mentioned in Wireless Waveforms - OFDM, the
3G LTE standard supports deployments of 1.4 to 20
The MA technique of this generation is Code- MHz, so, probably, the RTL-SDR will not be able to
Division Multiple Access. CDMA is based in the sample the entire spectrum. As LTE uses OFDMA, a
spread spectrum technique, in which many UE can multicarrier system is expected to be found while
use the same frequency spectrum by spreading their using the SDR unit, specifically sub-carriers with
messages using pseudo-random coding. The width both spacing and width of 15 kHz.
of the CDMA spectrum is 5 MHz, which is almost
double the RTL-SDR can sample. CDMA is mostly
found at 1700 MHz and above, but some signals are
still found in the 800-900 MHz range. [1]

Figure 12. Spectrogram signal from 4G at between 800.5 – 810.5

MHz from the RTL-SDR

9
mm-Wave Model UMa UMi RMa
Distance 10 – 5000 m >1 m
The millimeter-wave (mm-Wave) is the frequency
Frequency 0.5 – 100 GHz 73 GHz
spectrum between 30 and 300 GHz, in which the
BS Height 25 m 0 – 50 m
wavelength of the signals is between 1 and 10 mm.
UE Height 1.5 – 22.5 m 0 – 10 m
The mm-Wave comes with a much higher path loss Table 6. Model constraints for 3GPP path loss model.
attenuation and its inherited properties such as
oxygen and water absorption, foliage blockage and A comparison for each model was made, consisting
precipitation attenuation of the electromagnetic of the individual comparison of UMa, UMi and RMa
waves. Besides, the mm-wave multipath behaves environments using both LOS and NLOS scenarios.
differently across surfaces; this means engineers The model parameters are constrained by the
must consider rapidly changing channels. Finally, restrictions in Table 6 and were lastly selected, as
the mm-Wave is not capable of penetrating walls or shown in Table 7.
most solid objects, making signal reception an
Model UMa UMi RMa
additional challenge to what has already been
Distance 10 – 5000 m
discussed. [2, 3, 4]
Frequency 6-100 GHz
BS Height 25 m 10 m 35 m
UE Height 10 m
Table 7. Model parameters.

The 3GPP model was compared to the Free Space

path loss equation, which ideally describes the
attenuation a signal experience.
𝜆
𝑃𝐿𝑑𝐵 = −20 log10( )
4𝜋𝑓𝑑
Equation 2. Free Space Path Loss.

Figure 13. Atmospheric absorption of electromagnetic waves. [2] As seen in Error! Reference source not found., the
free space path loss is the less attenuated model,
Entities such as the 3rd Generation Partnership serving as a baseline for both 3GPP models. For
Project (3GPP), the 5G Channel Model (5GCM), every model, the NLOS scenario is the most
Mobile and Wireless Communications Enablers for attenuated, and the LOS scenario is very similar to
the Twenty-Twenty Information Society (METIS) the free space path loss, excepting the RMa scenario
and the Millimeter-Wave Based Mobile Radio Access in which there is a constant 10 dB difference.
Network for 5G (mmMAGIC) have developed Different to what is displayed in Figure 13, in any of
different models for the mm-Wave in environments these simulations the atmospheric impact of on the
such as Urban Macrocell (UMa), Street Canyon Path Loss is present. For every scenario, the Path
(UMi) and Rural Macrocell (RMa) for both Line of Loss is proportional to fc2 and d2.
Sight and No Line of Sight scenarios. The following
simulations use the 3GPP model specified in the Changing the frequency of these models shows
3GPP TR 38.901, which is a document describing the similar patterns, only changing the scale of the
path loss model, its constraints, derivation, and vertical axis.
assumptions. [2]

10
Figure 14. Path Loss for the UMa scenario. Figure 16. Path Loss for the UMi scenario.

Figure 15. Path Loss for the RMa scenario

11
 Hamming Reed- Bose-
Channel Coding Solomon Chaudhuri-
Chanel Coding describes techniques to correct data Hocquenghem
that was received incorrectly during transmission M 𝑀≥3 𝑀>1 𝑀≥3
over an unreliable channel; these can be classified N 𝑁 = 2𝑀 − 1
into Forward Error Correction (FEC) and Automatic k 𝑘 =𝑁−𝑀 𝑘 <𝑛−1 𝑘 <𝑚 ∙𝑡+𝑛
Repeat reQuest (ARQ). [1] dmin 𝑑𝑚𝑖𝑛 = 3 𝑑𝑚𝑖𝑛 𝑑𝑚𝑖𝑛
1> >2
=𝑁−𝑘−1 2𝑡
• Forward Error Correction: In FEC, a 𝑑𝑚𝑖𝑛 = 2𝑡 − 1
codework (Concatenation of transmitted Table 6. Block Coding equations for Hamming, RS and BCH
user data and redundancy) is received and codes. [1]
decoded. Decoding can find errors and
correct them using the redundant segment Performance
of the codeword. The purpose of coding is to enhance BER
• Automatic Repeat reQuest: ARQ requires a performance at the cost of data throughput.
feedback channel from the receiver to 𝑘
transmitter to request retransmission of 𝑅=
𝑛
data in the event of an incorrect reception.
Equation 3. Coding ratio for most linear block codes. [1]
Block Codes As seen in Equation 3, the coding ratio is linear,
FEC techniques can be classified into Block and unlike the BER relation with the same parameters.
Convolutional coding. The main difference between Hence, optimization can be done to achieve the best
both categories is that convolutional codes are BER in a specific SNR value while getting an
iterative (Or even recursive), and block codes are acceptable throughput. As a summary, having a low
not. Block Coding uses linear algebra to encode at throughput ratio does not translate to high BER
the transmission and try to get the bits that have performance.
been corrupted in transmission. There are many
algorithms for encoding using block codes, but all of
them are based on the following parameters:

• Codeword Length (N): This is the length of

the encoded message. Designers can define
this parameter by choosing a positive
integer power of two (M).
• Message Length (k): This is the length of the
user message to be transmitted. The number
of bits to be encoded is given by the
codeword length and the parameter M.
• Minimum Distance (dmin): This parameter
defines the error correction capability (t) of
the coding algorithm. It is given by the
smallest number of differences between Figure 9. Theoretical BER performances for BPSK over an AWGN
every possible codeword combination. channel using linear block coding algorithms.

Depending on the coding algorithm, equations and

conditions may differ from one algorithm to
another. The equations that relate all these
parameters are the following.

12
Simulation Reed-Solomon respectively. Values for N and k were
A Simulink model was designed to simulate a single assigned by using the equations in Table 6.
user transmitting encoded data using 8-PSK for
Hamming and RS, and QPSK for LDPC over an AWGN
channel (Not fading because MATLAB does not offer
the theoretical BER plots for fading channels). Three
linear block coding algorithms were tested:
Hamming, Reed-Solomon (RS) and Low-Density
Parity-Check, specifically the later has some
simulation constraints such as narrow valid SNR
range (According to MATLAB documentation [2],
community forums and community files)

Hamming and Reed-Solomon

The first two coding schemes are the simplest ones,
both Hamming and Reed-Solomon are linear block
codes based on the theory given in Block Codes. The
model parameters for simulation are the following.

Bernoulli Binary
Sample Time 0.01 Seconds – 100 Hz
Samples per Hamming 57
Frame RS 52
AWGN Channel
Eb/No Defined by BERtool
Number of bits 3 Figure 10. (Up) BER performance for Hamming encoded 8-PSK
modulated system over AWGN. (Down) BER performance for
per symbol Reed-Solomon encoded 8-PSK modulated system over AWGN.
Symbol Period 0.022 Seconds
Encoders Both models were designed for comparing
N Hamming 63 numerology and performance, having identical
RS 15 detection and correction capabilities (dmin = 3, t = 1);
k Hamming 57 meaning only a single bit can be corrected.
RS 13
Table 7. Model parameters for Hamming and Reed-Solomon The Hamming encoded plot presents a slightly
simulations. better BER performance compared to the uncoded
scenario when EbNo is higher than 6, due to the
Values as the sample time for the Bernoulli Binary
probability of having more than one corrupted bit is
block are arbitrary since there are no fading
reduced at this power. The coding rate for this
conditions to meet. Samples per Frame in the same
system is ∼90%, so no significant BER performance
block is defined by the encoder’s input signal
boost is achieved even with the high N and K
constraint being these k and k times m for Hamming
numbers.
and Reed-Solomon, respectively. The remaining
parameters in the model are set to their default The RS encoded plot presents a worse BER
value. performance at lower EbNo, this is because the
amount of parity bits in the codeword is too small,
To define both encoders, arbitrary values for M
but at EbNo higher than nine a performance boost is
where given, these being 6 and 4 for Hamming and
achieved. The coding rate for this system is 86%;

13
hence this system is objectively worse than the last, instead of hard detection (As shown in MATLAB’s
offering less BER performance and less throughput. documentation). The remaining blocks have the
same parameter values as the Hamming and Reed-
LDPC Solomon simulations.
Low-Density Parity Code is the FEC algorithm
proposed for 5G. It is a linear block code in which This model’s parity check matrix is given by the
both parity bits and message are equally protected, DVB-S2 specification, which consists in an irregular
and the parity check matrix is created randomly as a LDPC matrix with 32400 message bits and 64800
sparse matrix (few 1s and many 0s to increase the total bits, six 1s in the first row and seven 1s in the
Hamming distance). [2] remaining rows. The code rate for this encoder is ½.
[2]
Bernoulli Binary
Sample Time 0.01 Seconds – 100 Hz
Samples per Frame 32400

AWGN Channel
SNR Defined by BERtool
LDPC Encoder/Decoder
Parity-check Dvbs2ldpc(1/2)
matrix
Demodulator
Decision Type Approximate Log-
Likelihood ratio
Noise Variance 1/db2pow(SNR)
Table 8. Model parameters for the LDPC simulation. Figure 11. BER performance for Hamming encoded 8-PSK
modulated system over AWGN.
The LDPC encoder/decoder in Simulink was only
able to work using a different overall configuration LDPC encoder could not output results for SNR
of the whole block diagram. AWGN block had to be higher than 0.8 dB, but when compared to the
set to SNR instead of Eb/No and the demodulator uncoded plot, the error performance shows to be
block was set to Approximate Likelihood ratio significantly enhanced.

Figure 19. Block Diagram Model for Hamming, Reed-Solomon and LDPC coding.

14
OFDM SISO Fading Channel
Discrete 0.00001 Seconds
As explained in Wireless Waveforms - OFDM,
Path Delays
OFDM is the frequency diversity technique used in
Max Doppler 0.001 Hz
4G, based in the use of several low-bandwidth sub- Shift
carriers orthogonally placed to increase bandwidth Table 11. Model parameters for OFDM simulation.
efficiency and avoid channel impairments and
frequency selective fading. Settings such as Sample Time/Period in Bernoulli
Binary and AWGN Channel were set arbitrarily, the
For designing a basic OFDM system, parameters fading channel was configured to be in flat fading,
such as the number of subcarriers and the job of and the remaining parameters in non-mentioned
each one of them (Data, pilot or null/Guard-Band), blocks are set to default.
the modulation scheme (Many schemes can be used
for different subcarriers). For the first simulation run, default parameter
values were used in the OFDM
modulator/demodulator (FFT size of 64, 11 guard
band subcarriers, no DC null subcarrier and no
pilots; parameters such as cyclic prefix length,
number of symbols and antennas are not relevant
for this simulation). As the input signal, an array of
FFT size minus total guard-band sub-carriers is
needed to fill each data subcarrier in the OFDM
modulator. Since BPSK modulation is being used,
only a single bit can be processed at a time, so no
compensation is required in the input signal size.
Figure 21. Block Diagram model for OFDM system. The Channel Since OFDM is overly sensitive to Doppler shifting, a
block includes combination of fading and AWGN channels.
fixed low shift was selected for all the simulations
being this 0.001 Hz to avoid ISI and get consistent
Simulation
non-ideal results.
Three simulations were carried out using the
proposed block diagram model; parameters are
shown in the following table.

Bernoulli Binary
Sample 0.01 Seconds – 100 Hz
Time
Samples per 53 1013 65524
Frame
OFDM Modulator/Demodulator
FFT Size 64 1024 65535
Number of [6 (Lower boundary); 5 (Upper
Guard Bands Boundary)]
AWGN Channel
EbNo Defined by BERTool
Bits per 53 1013 65524 Figure 22. BER performance of OFDM over a Rayleigh channel
Symbol using different FFT sizes.
Symbol 0.01 Seconds – 100 Hz
Period

15
As shown in Figure 22, BER performance increases
in proportion to the FFT size and even surpasses the
theoretical BPSK without OFDM. Although
implementing an OFDM system of that size is
unpractical, actual systems used in LTE have FFT
lengths from 128 to 2048 for 1.25 MHz and 20 MHz,
respectively. [5]

Figure 23. BER performance of OFDM over an AWGN channel

using different FFT sizes.

Figure 23 shows the same model but using an AWGN

channel instead of fading. The FFT size still enhances
BER performance but its capped by the modulation’s
BER theoretical performance.

16
Spatial Diversity algorithm equation is given by the following
equation.
Spatial diversity describes techniques related to
using multiple antennas for transmitting and 𝑀−1
receiving; more advanced concepts such as MIMO ∗
𝑠 = ∑ ℎ𝑚 𝑟𝑚
and massive MIMO are based on this concept. Spatial 𝑚=0
diversity has several architectures to enhance data Equation 5. MRC equation.
throughput, bit error performance or both; these
architectures involves adding several antennas at After processing the incoming signals, the combined
the transmitter or the receiver and the required result can be further processed using a Maximum
algorithm to process the data from the additional Likelihood Detector algorithm.
data streams.
A model of this scheme was done using Simulink and
The models and simulations shown in this report is shown in Figure 24.
have the goal to increase Bit Error Performance.

Rx Diversity
The receive diversity scheme consists in adding
several antennas in the receiver side and then select
or combine the incoming signals from every
antenna. These algorithms are the Selection
Combining, Maximum Ratio Combining (MRC) and
Equal Gain Combining (EGC).
The most straightforward algorithm is the SC, which
only selects the highest fidelity signal from all the
antennas. This algorithm may require little channel
estimation to decide which antenna to use, but it can
be switched arbitrarily. As a summary, this
algorithm is the easiest to implement but delivers Figure 24. MRC block diagram model.
the weakest performance.
A simulation was performed using this model, which
MRC is an algorithm in which the signals coming is an extension of a standard 1x1 wireless
from every antenna and combines it after communications system model. The most significant
multiplying them with their respective channel differences are the MIMO fading channel block,
conditions (hm). This requirement means that MRC which holds the same fading parameters while
needs precise channel estimation to get a complex adding the number of antennas at transmitter and
representation for both magnitude and phase receiver. Besides, the MRC algorithm is performed
response of the channel. The received signal from an by the conjugation, product, and sum of elements
antenna is given by the following equation. blocks; as seen in Equation 5. As the gain channel in
𝑟𝑚 = ℎ𝑚 𝑠𝑛 + 𝑛𝑚 the MIMO fading block is a 4-dimensional matrix
with redundancies, a squeeze block is used to get the
Equation 4. Received signal from every antenna. parameter channels as a matrix with dimensions Rx
Where m represents the antenna number, Sn times Tx.
represents the sent message at the instant n and nm The parameters not mentioned in the previous
represents the noise profile of the channel. The MRC paragraph are set as a default or a commonly known
parameter.

17
 Once two equations are given, the contents of the
message can be estimated using simple algebra.
These equations can be extended to be used for
more messages and antennas; the proportion of
used Space-Time coding combinations is called Rate.
Since Space-Time coding is a commonly used feature
in wireless communications, Simulink provides a
library for this in the Communications Toolbox. By
using these blocks, the algorithm can be easily
implemented in models for simulations.

Figure 25. BER performance for Rx diversity using the MRC

algorithm.

As shown in Figure 25, BER performance is

enhanced by adding more receive antennas. Since
the most considerable gap in performance is
between 1x1 and 1x2, it is implied that adding more
antennas do not add a liner increase in performance.
Hence, studies and simulations must be done to
intelligently design a spatial diversity application Figure 26. Block diagram model for Alamouti Space-Time coding.
that is both high-performance and cost-efficient.
The simulation, similar to the Rx Diversity model,
Tx Diversity consists of a conventional wireless communications
The Tx diversity adds multiple antennas at the system model, only adding the MIMO fading channel
transmitter side and then apply a space-time coding and the Space-Time coding block from the
algorithm; the one reviews is known as the Alamouti communications toolbox. The number of
code. This technique is based on the resulting transmitting antennas is changed to see the impact
addition of signals at the receiver due to the of this parameter in BER, while the number of
multipath properties of the channel, as it assumes receiving antennas is fixed to one. The parameters
the worst-case condition in which the addition is required for the Space-Time encoder and decoder to
always destructive. The space-time part of the work is the number of Tx and Rx antennas and the
algorithm is to send two messages (S0 and S1) over coding rate, this last being set to ¾ (3 messages in 4
two antennas, the first antenna sends S0 and the instants) for the simulations for 2 and 3 Tx antennas.
second antenna send S1. Then, the same two The channel conditions and the remaining block
messages are sent, but using different antennas and parameters are set as default or to a value that
phases; the first antenna sends -S1* and the second makes the model meet its mathematical
antenna sends S0*. requirements.

𝑟0 = ℎ0 𝑆0 + ℎ1 𝑆1 + 𝑛0
𝑟1 = −ℎ0 𝑆1∗ + ℎ1 𝑆0∗ + 𝑛1
Equation 6. System of equations for Alamouti Space-Time coding.

18
Figure 27. BER performance for Tx diversity using the Alamouti
Space-Time coding.

As seen in Figure 27, the BER performance is

enhanced as more antennas are added to the system.
Unlike Rx diversity, similar improvements can be
seen in 2 and 3 transmitter antennas, while a minor
one is noticed when adding the fourth antenna.

19
MIMO The model in Figure 29 represents a 3-user MIMO
system in which a total of 9 data streams are
Multiple-Input Multiple-Output is a spatial diversity
assigned unequally to each user.
technique that involves having several antennas at
both receiver and transmitter to acquire the same User Data Streams
performance benefits that spatial diversity offers. 1 4
An advanced deployment of this, called Massive 2 3
MIMO will be used in 5G systems, increasing the 3 2
number of antennas in base stations from around Table 12. Datastream assignment per user.
10-20 to hundreds [8], allowing multiple
Each user’s data is modulated and then
simultaneous users and increased throughput and
concatenated into a single vector for every antenna
BER performance, while also implementing other
to be used in the MIMO fading channel. After adding
techniques such as Beamforming and Full-duplexing
noise to all the received signals, the estimation
to avoid interference and allow better performance
algorithm is used to convert from a vector with as
when communicating over the mm-Wave spectrum.
elements as Rx antennas to a vector with as
The current record of highest bandwidth efficiency
elements as data streams. Finally, each user’s data
is 145.6 bps/Hz, achieved using a 128 Massive
streams are separated, demodulated and compared.
MIMO array. [9]
The transposition blocks after concatenation and
Multi-User MIMO before de-modulation are used to satisfy model
MU-MIMO consists on having several data streams compatibilities instead of mathematical
in a multi-antenna system, where the number of requirements.
antennas at the transmitter side constraints the
number simultaneous data streams for every user Simulations were carried out variating the number
and the number of antennas at the receiving side of receiving antennas, increasing them for each user
must be equal or higher than the number of to have an additional antenna every new simulation.
transmitting antennas. Same as Spatial Diversity, the The number of antennas was 9, 12 and 15 receiving
ratio between Tx and Rx antennas increase both antennas.
BER and throughput, instead of a tradeoff as Spatial
Diversity does; but only BER performance will be
displayed as a performance metric.
Channel estimation is needed to get an overall
combination of all the signals from the receiving
antennas, and this is done by merely solving the
matrix representation of the incoming signals.
𝑹 = 𝑯𝑺 + 𝑵
𝑯 𝑹 = 𝑺 + 𝑯−1 𝑵|𝑵=0
−1

𝑺 ≈ 𝑯−1 𝑹
Equation 7. MIMO Signal combining.

Where R is the received signal matrix, H the channel

gains for every Tx and Rx antenna combination and Figure 28. BER performance for MU-MIMO.
S is the transmitted signal. In case the number of Tx
The simulation shows two main characteristics. The
and Rx antennas is not equal, the inverse of matrix H
first one is that every user has similar BER
will not be directly obtainable, and the pseudo-
performances in every scenario. The second is that
inverse algorithm must be used instead.

20
adding the first additional antenna per user as in the
9x12 scenario enhances the BER performance, while
the 9x15 scenario shows a more modest
performance difference.

Figure 29. Multi-User MIMO simulation model.

21
Beamforming Beamforming at the Receiver
Beamforming is widely used in several fields such as The purpose of applying a receiving beamformer is
signal processing, radar (sonar) or biomedical. As to get signals from the desired direction without
for mobile communications, Beamforming is used to getting as much interference from other sources as
transmit or receive signals over an antenna array a single omnidirectional antenna; this is called the
optimally, and it has been considered for the 5th combiner stage. Assuming a linear antenna array
generation of mobile communication systems and such as the one represented in Figure 30, an
was recently included in the IEEE 802.11ax standard upcoming signal with angle θ would arrive first to X0
(Commonly called WIFI6) as a main feature. and then to X1 after a few instants, this lapse is
Beamforming allows the transmitter to focus power represented as τ which remains constant for all
into a specific direction and receiver to only expect antenna elements.
signals from a specific direction, without
considering the remaining incoming signals.
Beamforming is done by placing an antenna array
consisting of several equally spaced omnidirectional
antennas, the shape of this array result in the
capabilities and performance of the beamforming
system.

Beamforming in 5G
5G goals are to enhance data throughput, drastically
reduce latency and increase the number of
simultaneous users. These goals will be achieved by
using the so-called millimetre wave spectrum (mm-
Wave) which is defined as the frequency spectrum Figure 30. A general structure for the combiner stage at the
receiving beamformer. [11]
with a wavelength between 1 and 10 mm, and this
spectrum has a range between around 30 and 300 𝑑 ∙ 𝑠𝑖𝑛(𝜃)
GHz. As of today, most of the available commercial 𝜏=
𝑐
solutions involving RF are located in the sub 6 GHz
Equation 8. Arrival time difference from one antenna element to
spectrum; this means using the mm-Wave spectrum the following. [11]
is a significant improvement that comes with several
engineering challenges such as to design cost- 𝑀−1
∗
effective hardware capable of generating and 𝑦(𝑡) = ∑ 𝑥𝑚 (𝑡)𝑤𝑚
processing super-high frequencies and develop 𝑚=0
techniques for avoiding the natural constrains the Equation 9. The received signal from the antenna array. [11]
mm-Wave has in terms of channel characteristics.
Equation 9 describes the received signal before
For example, the attenuation due to gas absorption demodulating using the carrier frequency using
(Pathloss) for a 60 GHz signal is around 10 dB/km, combining weights (wm). For this, the signal Xm (the
while a 700 MHz signal path loss is around 0.1 received modulated signal) is expressed as a phasor
dB/km [10]. This only means that for using the mm- for more accessible mathematical analysis, and then
Wave and achieve all the goals 5G proposes, a highly Equation 9 is simplified into Equation 11.
efficient transmission and reception scheme is
needed only for being able to communicate over the 𝑥𝑚 (𝑡) = 𝑒 𝑗𝜔(𝑡−𝜏𝑚)
mm-Wave. Equation 10. Phasor representation of the input signal from the
antenna element. [11]

22
 𝑀−1 𝑀−1
𝑗𝜔𝑡 −𝑗𝜔𝜏𝑚 ∗
𝑦(𝑡) = 𝑒 ∑𝑒 𝑤𝑚 𝑃(𝜔, 𝜃) = ∑ 𝑒 −𝑗 𝑚 𝜋 sin(𝜃) 𝑤𝑚
∗

𝑚=0 𝑚=0

Equation 11. Received signal from antenna array using phasor Equation 17. Antenna array response considering condition for
notation. [11] avoiding spatial aliasing and Direction of Arrival. [11]

Furthermore, the response of the beamforming Where sin(θ) can be taken as the normalized
system and matrix expression are given by the frequency (Ω) for design and simulation purposes as
following equations. the antenna array can be modelled as a low pass FIR
filter.
𝑃(𝜔, 𝜃) = 𝒘𝐻 𝒅(𝜔, 𝜃)
𝒘 = [𝑤0 𝑤1 … 𝑤𝑚 ]𝑇 This equations and theory are for Narrowband
𝒅(𝜔, 𝜃) = [1 𝑒 −𝑗𝜔𝜏1 … 𝑒 −𝑗𝜔𝜏𝑀−1 ]𝑇 Beamforming, in which the bandwidth of the
incoming signal is not very large. Beamformers are
Equation 12. Array response using vector notation.
categorized depending on their Fractional
Like Digital Signal Processing theory, aliasing can Bandwidth (FB); a narrowband beamformer may
affect beam directivity by placing the antenna have a FB lesser than 1%. [12]
elements farther away. Hence, the elapsed time 𝑓𝐻 − 𝑓𝐿 𝐵𝑊
between two elements receiving the same message 𝐹𝐵 = = < 1%
(𝑓𝐻 + 𝑓𝐿 )/2 𝑓𝐶
is too large, and the message’s direction of arrival
cannot be estimated. For a constant distance, the Equation 18. Narrowband condition. [12]
following equation describes the elapsed time of
For example, a 5 GHz carrier (fC) could transmit a
arrival between antenna elements.
signal with 50 MHz as bandwidth.
𝑑 sin(𝜃)
𝜏𝑚 = 𝑚 𝜏 = 𝑚
𝑐 Beamforming at the Transmitter
All the previous equations and theory for the
Equation 13. Time constant definition assuming constant
distance and Direction of Arrival. [11] combiner stage is applicable for the precoder stage
as well. Only a few differences in concept abstraction
𝑗 𝑚 2𝜋 𝑑 𝑠𝑖𝑛(𝜃)
𝑑 sin(𝜃)
− appear in theory behind distancing each antenna
𝑥𝑚 (𝑡) = 𝑒 𝑗𝜔(𝑡−𝑚 𝑐
)
= 𝑒 𝑗𝜔𝑡 𝑒 𝜆
element in the linear array such as naming renaming
Equation 14. The received signal from each antenna element wm as the precoder weights. While having many
considering constant distance and Direction of Arrival. [11]
closely aligned antenna elements improves the
Moreover, applying the Nyquist Theorem to avoid performance of the receiving beamformer, at
spatial aliasing. transmission having this situation diminishes the
contribution of each antenna and the whole array
𝜆 behaves as a single omnidirectional antenna
𝑑≤
2 element.
Equation 15. Condition for avoiding spatial aliasing. [11]
Increasing the distance between antenna elements
𝑥𝑚 = 𝑒 𝑗𝜔𝑡 −𝑗 𝑚 𝜋 sin(𝜃)
𝑒 reduces the transmitted power in specific directions
while maintaining constant power at the direction
Equation 16. The received signal from each antenna element
perpendicular to the antenna array, this is called
considering the condition for avoiding spatial aliasing. [11]
directivity enhancement. A critical point is placed at
half the wavelength of the carrier frequency (λ/2);
spacing each antenna element at a distance of λ/2
will enhance directivity without generating
secondary lobes while using a distance greater than

23
λ/2 will still enhance directivity while also Figure 31 shows how changing spacing distance can
generating secondary lobes orientated parallel to affect the radiation pattern from making the array
the antenna array. behave as an omnidirectional antenna to enhancing
directivity and generating secondary lobes.
Creating secondary lobes is often an undesired
consequence of over separating antenna elements.
However, it is a design constrain when a finite and
little amount of antenna elements can be placed in
the array; this can be solved by:

• Reducing the distance between elements:

This will decrease directivity while reducing
the power of secondary lobes. A balance can
be reached between these two parameters.
• Increasing the number of elements in the
array: More secondary lobes will be
generated, but all of them with lower power
and higher directivity.
• Modifying carrier frequency: Increasing the
wavelength to the desired value can have the
same result as reducing space between Figure 32. Radiation Pattern for an antenna array with multiple
elements placed at a constant distance of λ/2 at a carrier
elements. frequency of 5 GHz.
• Applying a window tapper: Window
functions, such as the ones used in DSP Figure 32 shows how increasing the number of
systems, are useful for reducing the power antenna elements in the array can improve the
of secondary lobes. Commonly used directivity of the main beam by adding several low
algorithms such as Hamming, Hanning, power lobes.
rectangular or Chebyshev windows may be
implemented for beamformers. Types of Beamforming
As stated before, Beamforming has been widely
studied, and many forms of techniques and
implementations are available now for many
different applications. The main category of
techniques is the beamformer architecture type,
which can be analog, digital or hybrid (Analog-
Digital). Each type has its benefits and
disadvantages, as well as individual performance.

Digital Beamforming
This type of beamformer is the “simplest” design and
mainly relies on a baseband processor to generate
the phased output signal already described. The
main advantage is that a single design pattern can be
followed to create a beamformer; this means a
Figure 31. Radiation Pattern for an antenna array with two
combination of RF chain and power amplifier is
elements (M = 2) placed at different distances at a carrier required to drive each antenna. That combination is
frequency of 5 GHz. implemented as antennas are needed.

24
This feature also causes the main disadvantage of
digital Beamforming. Having a whole RF frontend to
drive a single antenna is highly inefficient and
poorly cost-effective since precoding and combining
are done in a DSP processor, and each RF frontend
have considerable power requirement [13]. Digital
beamforming also allows multiuser MIMO and
higher spatial diversity.

Power Amplifier

Figure 34. Analog Beamforming block diagram. [13]

Hybrid Beamforming
Hybrid Beamforming combines the advantages of
Digital and Analog Beamforming and a few
disadvantages. It consists of performing the
precoding and combining stages at both DSP
processor and RF frontend. This combination allows
having the flexibility, multiuser and higher spatial
diversity of Digital Beamforming, the high cost and
Figure 33. Digital Beamforming block diagram. [13] power efficiency and less complicated hardware
architecture from Analog Beamforming at the cost of
Analog Beamforming slightly higher power consumption than analog
This type of Beamforming keeps the processor from Beamforming and much more complex
phasing and amplifying the output signal by adding precoding/combining algorithms.
an intermediate block between the power amplifier
and the RF chain containing a phase and amplitude Beam Steering
modifier. Adding this stage allows output signals to Beamformers can change their radiation pattern in
be precoded and combined by specified hardware, order to rotate the main lobe and secondary lobes
allowing having a single RF frontend for many for avoiding known interference or focus in the
antennas and reducing power consumption from desired direction dynamically. Beam Steering can be
the DSP processor. done in both main or secondary lobes independently
to focus and avoid upcoming waves.
A disadvantage of this type of beamformer is the
added complexity of tuning both hardware and A second weight must be added to the precoder or
software whenever modifications are needed. Since combinator weights to shift the Direction of Arrival
beamforming parameters are fixed multiuser MIMO and steer the main beam.
and higher spatial diversity are restricted.

25
 Figure 35. Hybrid Beamforming block diagram. [13]

𝑀−1 calculated using Equation 20 and φ = 30. Both are

∗ ∗
𝑃(𝜔, 𝜃) = ∑ 𝑒 −𝑗 𝑚 𝜋 sin(𝜃) 𝑤𝑚 𝑆0 shown below.
𝑚=0
𝑤𝑚 = [0.120 0.240 0.300 0.240 0.120]
𝑆0 = [1 −𝑗 −1 𝑗 1]
Equation 19. Antenna array response considering beam steering.
Finally, the radiation pattern is obtained using
𝑆0 = 𝑒 −𝑗𝑚𝜋sin (𝜙) Equation 19.
Equation 20. Steering Vector.

Where φ is the angle of steering (Direction of

Arrival/Departure).
The following example shows a Radiation Pattern of
a beamformer (Can be either Tx or Rx) consisting in
a 5-element uniform linear array (d = λ/2) pointing
to the directions ±15° with a Direction of
Arrival/Departure of 30 degrees. This beamformer
was modelled as a 4th order FIR low pass filter and
designed in MATLAB’s Filter Designer app, and its
normalized passband and stopband were set to
sin(5°) and sin(30°), respectively; bandpass weight
was set to 1 while stopband was set to 0.5. Figure 36. Radiation Pattern, for the example beamformer.

MATLAB generated the filter coefficients using the

Filter Designer. Then, the steering vector was

26
Appendix
Repository
The MATLAB and Simulink simulation files used in the previous sections are available in the following
Repository.
[Link]

LTE Grids

Figure 37. FDD LTE Grid

Figure 38. TDD LTE Grid

27
 [1 Mathworks, "Hybrid Beamforming for Massive MIMO
References 0] Phased Array Systems," 2020. [Online]. Available:
[Link]/campaigns/offers/hybrid-
beamforming-white-
[1 B. Stewart, K. Barlee, D. Atkinson and L. Crockett, [Link]?s_iid=doc_wp_AR_footer. [Accessed 12
] Software Defind Radio Using MATLAB & Simulink and April 2020].
the RTL-SDR, Strathclyde Academic Media, 2017.
[1 W. Lui and S. Weiss, "Introduction," in Wideband
[2 T. S. Rappaport, G. R. MacCartney and E. Mellios, 1] Beamforming: Concepts and Techniques, Wiley, 2010,
] "Overview of Millimeter Wave Communications for pp. 1-18.
Fifth-Generation (5G) Wireless Networks - With a
[1 B. Allen and M. Ghavami, "Narrowband Array
Focus on Propagation Models," IEEE Transactions on
2] Systems," in Adaptive Array Systems, Chichester,
Antennas and Propagation Vol. 65. No. 12, pp. 6213-
Wiley, 2005, pp. 32-54.
6230, December 2017.
[1 Peregrine Semiconductor, "Analog Beamforming—
[3 A. Nordrum and K. Clark, "5G Bytes: Millimeter Waves
3] What is it and How Does it Impact Phased-Array
] Explained," IEEE, 6 May 2017. [Online]. Available:
Radar and 5G?," YouTube, 17 Febraury 2017.
[Link]
[Online]. Available:
5g-bytes-millimeter-waves-explained. [Accessed 5
[Link]
May 2020].
&t=307s. [Accessed 11 April 2020].
[4 Federal Communications Commission, "Millimeter
[1 W. Tomasi, Advanced Electronic Communications,
] Wave Propagation: Spectrum Management
4] Error Performance ed., Harlow: Pearson, 2014, pp.
Implications," July 1997. [Online]. Available:
101-106.
[Link]
nology/Documents/bulletins/oet70/[Link].
[1 5G Learning, "A Detailed Introduction to
[Accessed 5 May 2020].
5] Beamforming," YouTube, 6 May 2018. [Online].
Available:
[5 K. Haesik, "Wireless Channel Impairment Mitigation
[Link]
] Techniques," in Wireless Communications Systems
[Accessed 7 April 2020].
Design, 1st ed., Wiley, 2015, pp. 75-122.

[6 A. Grami, "Error-Control Coding," in Introduction to

] Digital Communications, Elsevier, 2016, pp. 433-435.

[7 MathWorks, "MATLAB/Simulink Help Center,"

] MathWorks, 2020. [Online]. Available:
[Link]
[Accessed 17 03 2020].

[8 N. Amy and C. Kristen, "5G Bytes: Massive MIMO

] Explained," IEEE, 17 August 2017. [Online]. Available:
[Link]
5g-bytes-massive-mimo-explained. [Accessed 4 May
2020].

[9 N. Amy, "5G Researchers Set New World Record For

] Spectrum Efficiency," IEEE, 12 May 2016. [Online].
Available: [Link]
talk/telecom/wireless/5g-researchers-achieve-new-
spectrum-efficiency-record. [Accessed 4 May 2020].

28